Abstract

Many studies have been conducted on the National Football League’s (NFL) strategies for expanding its brand internationally. However, the NFL has recently implemented and amplified many strategies that were not as prevalent beforehand. This Literature Review will examine these new strategies. It will also examine the reasons that previous strategies were ineffective and unsuccessful. The paper specifically examines the strategy of “Glocalization”, the act of using preexisting networks and influential people in foreign nations, to appeal to the widest possible audience.  The paper uses this context to investigate how the NFL uses the Global Markets Program, social media, fan clubs, and international games to attract new fans. The paper explores how these strategies can not only bring in new fans, but also cause deeper levels of fandom for already existing fans. These more dedicated fans can also bring the league more money through ticket and merchandise sales. The study concludes that glocalization is a successful strategy that combats the alienating feelings of past NFL international endeavors. The study also concludes that one area of growth for the NFL is the development of the playing of the sport internationally. The Literature Review addresses how the NFL needs to develop grass roots leagues, international equipment, and international facilities because fans who play the sport are also more likely to watch it.

            Keywords: National Football League, international expansion, glocalization

Introduction

The National Football League (NFL) was founded in 1920 with 14 teams (Jozsa, 2014). Of those 14 teams, only 2 exist today and neither from their original city (Ibid). The NFL added four teams in 1922(Ibid). The first NFL season looked very different to the current league: the average attendance was just over 4000, the playoffs did not exist, and the champion was decided through a vote (Crepeau, 2020).

The NFL needed to expand its viewership besides fans watching live, in person games. In 1956, CBS became the first network to televise games (Felser, 2008). By broadcasting games, the NFL could broadly appeal to fans across the country, not just fans attending games. Therefore, by 1960, television viewership became increasingly important. NFL commissioner Peter Rozelle came up with the idea of a unified television deal with revenue sharing (Ibid). In April 1961, he signed a 2 year 9.1 million dollar contract with CBS granting all broadcasting rights to all NFL regular season games. The American Football League (AFL), at the time a competitive league to the NFL, sold their broadcasting rights to NBC for 42 million dollars in 1964 (Crepeau).

            The natural next step forward for the NFL was to expand its market. In 1966 the NFL and AFL merged (Felser). The league reorganized to have 26 teams. This merger was a smart strategy for the NFL. The AFL had preexisting teams and fans that were brought into the NFL fanbase. They had stars to bring in new fans such as Joe Namath, the quarterback of the New York Jets. By creating this merger, the NFL was able to nearly double the number of locations of their teams. Their appeal across the country was even larger.

            The NFL implemented new strategies to mythologize the league. In the 1970s the NFL created Monday Night Football which brought in many viewers on a new day of the week (Crepeau). They marketed rivalries to deepen fandom by creating animosity for other teams. In 1989, the NFL created John Madden Football, a video game in which people could control their favorite team(Ibid). Sports bars started to open that catered towards specific fanbases.

            The Jacksonville Jaguars and the Carolina Panthers were the NFLs most recent expansion in 1996, bringing the league to its current 32 teams (Jozsa). The NFL continued to look at new places to add fans.

Literature Review

Global Markets Program

Firstly, to attract long lasting fandom in international markets, the NFL adopts the strategy of assigning teams to specific nations through the Global Markets Program (NFL, 2026). Within these nations, NFL teams can market, hold fan events, and have local sponsorships. This assignment forces teams to individualize specific strategies for nations and not have an ineffective broad strategy. One major obstacle in NFL globalization is that the NFL can feel like an alien spectacle. To combat this feeling, markets are oftentimes assigned to teams based upon preexisting connections between the cities and their target nation.

For example, the Saints are assigned to France because New Orleans was formerly under French control. Teams such as the Miami Dolphins are assigned to Spanish speaking countries due to their city’s large hispanic population. The Rooney family who owns the Steelers has Irish origins (Ibid). These assignments create an environment where distant fans are already connected to a team.

Next, these teams participate in glocalization and collaborate with popular, prerecognizable brands and people within their assigned nation. Many teams partner with soccer clubs and players to reach into a known fanbase. During the 2026 World Cup, teams were advertising towards tourists to increase international fandom.

Additionally, teams partner with local fashion brands to pack merchandise in stores. The Jets, for instance, collaborated with trapstar London, a British streetwear brand.

The specificity of this marketing would be impossible if each team was forced to market to all nations. The Global Markets Program allows teams to create a cohesive narrative for fans to follow. The NFL forms television deals with international markets so distant fans can follow their assigned teams season long. Teams have year long, not sporadic, advertising in their assigned nations to create long lasting relationships. Consequently, distant fans become permanent fans and do not just exist during international games. Overall, the Global Markets Program lets teams be ingrained in international society and create long lasting fandom.

International Games

            The NFL also builds international fandom by hosting games abroad. The NFL international series started in 2007 when the Giants defeated the Dolphins 13-10 in Wembley stadium in London with over 80,000 fans watching. International games garner large foreign support. International fans desire an experiential factor in their NFL experience. Going to games provides a different, better experience than just watching on television. At games, fans are surrounded by similarly minded fans who also want to support football, creating a fan on fan connection. Additionally, fans want to experience tailgates and halftime shows that are unique to the in person experience. A higher level of fandom is possible when fans can support teams in person and not just distantly because they gain a personal connection to the team.

In 2012, Shahid Khan became owner of the Jacksonville Jaguars and every year since then (besides 2020 due to the Covid-19 pandemic) decided to hold one home game in London. This decision was revolutionary because he opted to give up revenue from a home game to create international fandom. Consequently, the Jacksonville Jaguars have cultivated a large distant fan base in London despite being a small market team with a bottom five record since Khan’s takeover (Drew, 2024).

The Jaguars are increasingly relying on the London fanbase with two home games scheduled to be played in London for the 2026-2027 season. These games have grown increasingly popular overseas. The 2024-2025 season saw a 35% increase in ticket sales over the year before (Ibid). These games grant fans the opportunity to see the teams they support in person. Fans receive a positive memory from attending a game and are therefore more likely to support the team in the future.

These NFL games also participate in glocalization, attempting to adapt to already existing platforms for the games. The NFL utilizes pre-existing infrastructure for other sporting events. Games in the United Kingdom are played in Tottenham Hotspur and Wembley stadium, both premier league stadiums. In Melbourne, games are played on the Melbourne Cricket Ground (NFL, 2026). These venues provide both a convenient location for the NFL and a familiar spot for locals.

NFL merchandising is also an important part of NFL international games. Many international fans do not own NFL merchandise before games. NFL stores at international games are oftentimes larger than the stores at domestic games because new fans are anxious to buy new merchandise. They want to support a team and have memorabilia to remember their in game experience (Drew).

Social Media Marketing

The NFL also adopts glocalization when it comes to marketing through social media. Firstly, the NFL creates accounts for foreign countries. The NFL has created accounts for Brazil (1.1 million followers; NFL Brasil, 2026), Mexico (683,000 followers; NFL Mexico, 2026), Spain (147,000 followers, NFL España, 2026), and France (147,000 followers, NFL France, 2026) . These accounts post content in their nation’s home language, making it simpler for fans to follow along.

During the offseason these accounts post updates on players’ lives with captions in their native language. These posts are essential because they create a narrative outside of the game. This storytelling allows fans to build a deeper connection to the players. This strategy is especially important for international fans because they are able to see similarities between themselves and players who live in another country that may be unexpected.

During the regular season, highlights are posted on these accounts. This provides fans with an opportunity to watch plays and stay up to date on games they could have missed due to time zone differences. This strategy limits any differences due to time restraints.

Additionally, social media is an important space for fan on fan engagement. Many United States fans are able to experience these interactions during in person games and at bars. These opportunities are less common for worldwide fans. If an international fan wanted to watch a game at home they could see posts in their social media feed creating an energy and sense of community over events happening in the game. (Brattain) Furthermore, social media creates a space for online discourse. International fans are able to debate their opinions with each other online. International NFL accounts commonly create posts that encourage fans to debate in the comment section. For example, NFL Brasil posted power rankings with fans arguing in the comments (NFL Brasil, 2026). Fans are able to debate and argue online creating a greater fan-fan connection (Brattain). A study of Brazilian NFL fans found that Twitter, now known as X, provided fans with a virtual space for online fandom. This online connection deepened fandom(Moura).

            These accounts also participate in glocalization by posting specific localized content. They collaborate with local celebrities to endorse the game which touches on a preexisting market. These accounts post many futbol players to create a connection between the sports. For instance, NFL UK posted a video of Jaguars Kicker Cam Little exchanging jerseys with Fulham FC player Antonee Robinson (NFL UK, 2026). Accounts will also post the opposite by showing NFL players at international events assimilating to British culture. British comedian Troy Hawke was posted welcoming Jaguar players to London by teaching them British expressions (NFL UK, 2026). These posts make the NFL feel less alien because the assimilation is double sided.

The NFL also sponsors many local influencers to support their events. They bring many influencers to games in their international series. Additionally, Influencers are brought out to announce picks during the NFL Draft. This strategy utilizes the large preexisting reach of influencers in these international markets.

One way that the NFL recognizes these fans is through the NFL international fan of the year program (NFL). Each team chooses an international fan who is involved in creating a community in their home country. Many chosen fans are social media influencers who support the NFL.

The NFL will also host events with influencers. For example, the NFL hosted the Sao Paulo creator cookout before 2025’s football game in Brazil. They brought together popular Brazilian creators to cook while answering NFL trivia questions as promotional material for the International Series. (NFL, 2025)

International Supporter Groups

The NFL also utilizes supporter groups to build lasting fandom for teams. Supporter Clubs are an organized group of fans that watch and support a sports team together. Distant supporter groups organize watch parties at bars to watch games together.

            Supporter clubs are necessary to build interpersonal relationships between fans. As previously mentioned, it is more difficult for international, distant fans to be able to find a community to be a part of who supports a team. While social media can provide this community, fan clubs create a deeper, in person social interaction. Fan clubs create a stronger sense of team affiliation by facilitating a community around these games (Katz, 2020). One distant Steelers fan in the Midwest stated,

“You know I have always said, I can sit at home and watch the games in my pajamas and yell at the TV. But why not do it with 100+ other people who are feeling the same way you do. We all want that sense of belonging with fans like us” (Stensland, 2020).

            While not international, this fan club provided a fan with likeminded people to be surrounded by during games.

            Moreover, fans experience a stronger sense of affiliation by forming smaller communities within the larger community of the team (Katz). Supporter groups sometimes maintain smaller groups within the larger group to provide an even more interpersonal connection. Theoretically, if a person is developed farther into the fan group, they will have a deeper connection (Ibid).

            The most important fans in the fan groups are the most connected fans, not necessarily the most vocal. The most connected fans are those who are involved with the most people and subgroups within the fan club (Ibid). Many of these people are founders and leaders of international fan clubs. They are able to lead large groups because they are interconnected to everyone. It is important for the NFL to market towards these people because they pass down information to everyone else.

            The NFL also uses the International Fan of Year award to recognize leaders in these fan groups. Many are founders and leaders of fan clubs. Stephen O’Brien, for example, is the founder of the UK and Irish Packers Fan Club, where British Packers fans can connect and watch games together (NFL, 2025). The Packers honored him and he announced a draft pick at the NFL draft. The NFL understands that if they can market towards and support these leaders, they can also market toward the fans they influence.

NFL International Player Pathway

The NFL International Player Pathway is a pipeline for international players to join the NFL. International players train in Florida and perform a combine before the NFL draft. Since its inception in 2017 five players have been drafted from this program (NFL, 2026). Additionally, the NFL allows a seventeenth practice squad spot for international players from this program (Ibid). It allows premier athletes a chance at becoming an NFL player.

International fans have a built-in support of players from their home countries. They connect with players who grew up in a similar place and manner to them. Fans are likely to support a player and therefore also support the team they compete for. Many teams from the Global Markets Program are assigned to their country because they drafted a player from that country. For example, the Eagles are assigned to Australia because they drafted Pro Bowl Offensive Lineman Jordan Mailata who was born there.

Discussion

The NFL has had many previous attempts to reach international fans. Firstly, the NFL held the American Bowl, a preseason game, internationally from 1986 to 2005 . These preseason games utilized a similar strategy to the current international series. They attempted to gain international fandom by giving foreign nations international games. These games were very successful because they brought in large international crowds to NFL games. Specifically, the American Bowl in Mexico City had over 110,000 fans attend the game. This success led the NFL to realize that international games of a premier product were an important strategy to implement later on.

            The NFL also tried to start leagues internationally. They started the World League of American Football, (WLAF) in 1989. It originally included teams from both North America and Europe. At first, the league was successful in Europe at first, but the American teams were a failure. The NFL wanted it to be similar to a g league for the NFL. NFL fans in the United States did not want to support something that was not a premier product.

            The league was put onto hiatus for two years. It was rebranded to NFL Europe. There were six teams within the league. The league was very unprofitable for the NFL. It lost 30,000,000 dollars per year. Attendance for the league was very low towards the end. London, one of the most populated cities in the league, had attendance of under 10,000 people per game. It was rebranded to NFL Europa. Germany was the only country where the league was successful. At the end of the league, five of the six remaining teams were located in Germany.

            The league failed for many reasons. Firstly, the league failed to secure broadcasting rights in Europe. While the games were taking place in Europe, it was difficult to watch the game. This difficulty made a deeper fandom less likely. Additionally, the league felt alien and out of place. It appeared in these countries without taking initiative to assimilate into their culture. It was difficult to build fandom because people had no good reason to connect to a team.

            The most important strategy that the NFL is adopting is glocalisation. Past attempts at expanding the NFL into other countries felt forced and unnatural. The NFL adopted the strategy of globalization to combat this issue. They do not want their sport to feel alien to the countries they expand into.

The NFL partners with already popular local brands because they need to build into an already built in fanbase. Brands already have customers who support them. By partnering with these brands, the NFL feels natural and normal in these nations. They also partner with local influencers. They are able to tap into a preexisting network to spread their influence. They use important, connected figures to spread good word about the league. They similarly use highly connected individuals within fan clubs to build fandom. They work with the individuals who have the greatest influence within the fan club to connect to new fans of the NFL.

            The NFL needs to improve their overseas player development. While the International player pathway exists, it is not very successful. Only two players were selected from the international player pathway during the 2026 NFL draft: Seydou Traore and Uar Bernard. The lack of international talent in the NFL is especially apparent when compared to other American sports leagues. Only 3% of NFL players are international. Comparatively, 30% of the National Basketball Association (NBA) and 26% of the Major League Baseball (MLB) are foreign born. 0 NFL Most Valuable Players (MVPs) have been international, while the last 8 NBA MVPs are all foreign born .

If the NFL builds superstar talent abroad, international fans will have a more personal connection with the league. One important factor that the NFL misses out on is player development in foreign nations. It is necessary for the NFL to work on building foreign youth talent.

Firstly, the NFL needs to teach football in other nations. They somewhat already do this by sponsoring and supporting flag football leagues. They commonly feature and advertise this content with their games. However, teaching flag football is not enough to build talent. If the NFL wants to build real talent in other nations they need to support the teaching of actual football.

Also, there is an important correlation between playing and supporting a sport. Athletes who grow up playing a sport are much more likely to watch and follow the sport. They have a deeper emotional connection towards football meaning a likelier deeper level of fandom.

Another way the NFL can support the growth of playing football on an international level is by making football more accessible. Other sports such as soccer or basketball are much easier to play at any given notice. A person can bring a soccer ball to a field at a park or a basketball to an open basketball court and they would be able to improve their skills easily. It is much more difficult for a person to develop football skills, especially in a foreign nation. Firstly, there are much fewer facilities to practice at in these areas. Other sports built facilities to facilitate the growth of their sport. Fifa, for example, built soccer courts in less affluent areas, so lower income individuals would have an increased ability to play the game. They were given access to facilities that they wouldn’t normally have. Similarly, there are public basketball courts in many foreign nations. The NFL needs to build football facilities to grow the playing of the sport internationally.

Similarly, the NFL needs to provide better access to football equipment. One problem with football is that more equipment is necessary to fully play tackle football. Players need pads, helmets, and mouthguards, not just footballs to learn the game. This equipment can be expensive and hard to come across. The NFL needs to provide easier, cheaper access to this equipment for players to learn the full game and develop into better players.

Overall, the NFL needs to support the development of youth players and leagues to build a deeper emotional fandom and to develop players to compete in the league.

Conclusion

The NFL has developed and grown many successful international strategies within the past decade. Most importantly the NFL has understood glocalization and adapting to countries to make american football not feel like an alien product. The NFL has used this strategy within the Global Markets Program. They realized that they can appeal more towards international fans if they have a team to support. By assigning teams to countries the NFL allowed each team to specialize towards their assigned market and deepen the level of fandom because each team has a given fandom to support. The NFL has also broadened their reach by increasing the amount and locations of international games. These games deepen fandom because a greater connection is created when fans can watch their favorite teams live. The National Football League also utilizes social media by creating accounts for different nations and partnering with local influencers. These local partnerships help them reach the broadest possible audience. They also fan clubs, understanding that the most influential fans are the ones who are the most connected. The NFL needs to improve their development of the playing of football internationally to create a deeper fandom. Future studies in this field can examine how the NFL is attempting to grow the playing of the sport internationally. It can focus on the effectiveness of those strategies and what new more productive strategies can be put in use.

References

Brattain, L. (n.d.). The NFL and the digital era: How social media has redefined the National Football League. Whittier College. https://poetcommons.whittier.edu/cgi/viewcontent.cgi?article=1062&context=scholars

Crepeau, R. C. (2020). NFL football: A history of America’s new national pastime. University of Illinois Press. https://books.google.com/books?hl=en&lr=&id=9jj6DwAAQBAJ&oi=fnd&pg=PA1911&dq=NFL+growth&ots=bLIhqdRJaf&sig=CWH4kJZZBgN58c-mpEC7W1_xvX8#v=onepage&q=NFL%20growth&f=false

Drew, M. (2024, September 4). The NFL’s global blitz: What the 2024/25 ticket sales data reveals about the league’s international growth. SportsPro. https://www.sportspro.com/opinions/stadiums-and-venues/nfl-2024-25-ticket-sales-data-viagogo-international-games/

Felser, L. (2008). Birth of the new NFL: How the 1966 NFL/AFL merger transformed pro football. Simon and Schuster. https://books.google.com/books?hl=en&lr=&id=HiL53_048JgC&oi=fnd&pg=PP1&dq=NFL+merger&ots=boujqLcw6k&sig=lwjPIFfZXr6CYWoAtTRQB2m5H6Q#v=onepage&q=NFL%20merger&f=false

Global markets program now active in 22 international markets. (2026, March 30). NFL.com. https://www.nfl.com/international/global-markets-program

International fan of the year: Green Bay Packers. (2025). NFL.com. https://www.nfl.com/honors/fan-of-the-year/international/2025/packers

Jets unveil exclusive collaboration with Trapstar London x Mitchell & Ness. (2025, October 9). New York Jets. https://www.newyorkjets.com/news/jets-unveil-exclusive-collaboration-with-trapstar-london-mitchell-ness-10-09-2025

Jozsa, F. P., Jr. (2014). National Football League strategies: Business expansions, relocations, and mergers. Springer. https://books.google.com/books?hl=en&lr=&id=RFUlBAAAQBAJ&oi=fnd&pg=PA1&dq=NFL+merger&ots=ckmqVDcx_h&sig=V81kz1W0ov3MWfUnyaYrG2mSKac#v=onepage&q=NFL%20merger&f=false

Jozsa, F. P., & Guthrie, J. J., Jr. (1999). Relocating teams and expanding leagues in professional sports: How the major leagues respond to market conditions. ABC-Clio. https://www.google.com/books/edition/Relocating_Teams_and_Expanding_Leagues_i/6k7EEAAAQBAJ?hl=en&gbpv=0

Katz, M., Baker, T. A., III, & Du, H. (2020). Team identity, supporter club identity, and fan relationships: A brand community network analysis of a soccer supporters club. Journal of Sports Management.

Martin, S. M. (n.d.). Internationalizing sports brands: The impact of marketing in the NFL’s entry into the Spanish market. Comillas.

Moura, B. M., & Maranhão, A. L. (2019). Consumption attachments of Brazilian fans of the National Football League. International Journal of Sports Marketing & Sponsorship. https://www.emerald.com/inmr/article-pdf/17/3/251/1284252/inmr-02-2019-0015.pdf

[nfl]. (2025, September 13). Had to make things interesting. Creator Cookout: São Paulo Edition full episode is live on YouTube [Video]. Instagram. https://www.instagram.com/reel/DOjXbkiAGQj/

NFL international fan of the year. (2025). NFL.com. https://www.nfl.com/honors/fan-of-the-year/international/

NFL International Player Pathway. (2026). NFL.com. https://www.nfl.com/international/player-pathway

NFL international players. (2026). NFL.com. https://operations.nfl.com/programs-initiatives/international-growth/International-player-pathway

[nflbrasil]. (2026, October 15). Power ranking da NFL Brasil após a semana 5! O que você achou da lista? Confira o ranking completo no perfil [Photograph]. Instagram. https://www.instagram.com/p/Dehf4f1E1cy/?hl=en

[nfluk]. (2026, October 17a). An annual tradition! Troy Hawke is back to give the Jaguars a proper welcome ahead of the London Games [Video]. Instagram. https://www.instagram.com/p/Demp9jfMDRN/?hl=en

[nfluk]. (2026, October 17b). A pleasure, Cameron Little [Photograph]. Instagram. https://www.instagram.com/p/DemLvbjIQIA/?hl=en

About the author

Shane Handman

Shane Handman is currently a senior at Harvard Westlake in Studio City, California. He has an interest in economics, business, math and the world of sports. Besides academics, he is the captain of his cross country team at his school.

The post The National Football League’s Strategies for International Expansion appeared first on Exploratio Journal.

Abstract

Separating market-wide movements from stock-specific returns is central to factor modeling, while it remains unclear whether more flexible factor models produce residuals that are more useful as trading signals in the Korean stock market. This study compares three models with increasing levels of flexibility which are principal component analysis (PCA), instrumented component analysis (IPCA), and conditional autoencoder (CAE). Using daily data of 156 KOSPI stocks, each model is trained on 70% of the valid sample and applied to the remaining 30% to produce out-of-sample residuals. The residuals are then tested through statistical arbitrage and idiosyncratic momentum strategies, ultimately evaluating which factor models’ residuals produce better or worse trading strategy performances. CAE produced the most tradable residuals within both statistical arbitrage and idiosyncratic momentum strategy. IPCA’s residuals performed better than PCA’s in the statistical arbitrage strategy, but the PCA’s residuals outperformed IPCA’s in idiosyncratic momentum strategy. Another result was that across all three factor models, idiosyncratic momentum strategy substantially outperformed the statistical arbitrage strategy. The finding from these results was that the most flexible factor model that is nonlinear and has characteristic dependent loadings produce the best residuals for trading signals in South Korea. However, it doesn’t imply that the more flexible factor models are, the more tradable the residuals are, since the result of the PCA and IPCA are different. Also, since the OU process didn’t work, it could be suggested as the reason why statistical arbitrage strategy underperformed compared to momentum strategies, suggesting how KOSPI residuals contain stronger information about persistent trends of stocks than short term mean reversion, as can be concluded from the comparison of idiosyncratic momentum strategy and statistical arbitrage strategy. However, there are several limitations to be considered in this study such as the size of the stock data, quality of the data, disregardments of small implementation details such as transaction costs, etc. and survivorship bias.

I. Introduction

Stock returns are affected by many variables, but they could be said as being mainly driven by two components: market-wide factor movements which affect all the stocks in the market, and stock-specific components which only affects a specific stock. The decomposition of the systematic risk, which is the market-wide force, and the idiosyncratic risk, the stock-specific effect, is what a factor model does. If a factor model removes the systematic risk accurately, the remaining idiosyncratic residual could be used to optimize asset pricing and reveal meaningful information such as temporary mispricing or stock specific trading signals.

Because factor decomposition is an important part in quantitative investing and portfolio optimization, factor models have evolved continuously throughout financial modeling history. Principal component analysis (PCA), which has been used as early from Chamberlain & Rothschild (1983) and Connor & Korajczyk (1986) and has continuously been popular among many researchers, is a linear factor model in which returns are represented as weighted sums of latent factors. Evolved from this is the Instrumented principal component analysis (IPCA) introduced by Kelly et al. (2019) [2], which is still linear but has its betas vary as linear functions of stock characteristics. Inspired by the IPCA model, Conditional Autoencoder was proposed by Gu et al (. using a nonlinear neural-network structure with changing betas by characteristics. [3,4] 

These three factor models have been tested and shown to be successful in U.S. stock data [5], but their effectiveness in the recent Korean stock market is less clear. Although there are prior studies that have applied factor models to Korean markets, very few are applied on recent data, including 2025 and beyond. Furthermore, there are only a few studies comparing the effectiveness of different factor models in the Korean stock market.

Therefore the primary objective of this study is to determine which factor model, leveled by its degree of flexibility, is most effective in the recent Korean stock market, measured by the trading performance from its residuals. The factor models of varying flexibility are PCA, the linear with fixed betas, IPCA, linear with linear-varying betas, and CAE, the nonlinear with nonlinear-varying betas. The evaluation of the factor model this study uses is based by  comparing them through the residuals they produce, which one produces the most effectful residuals that are most effectful for two financial methods, statistical arbitrage and idiosyncratic momentum strategy. The factor models are evaluated by comparing the residuals they generate, with the comparison focusing on which model produces the most tradable residuals that yield the strongest performance in two residual-based strategy.

II. Methodology

Data and Factor Model

In this study, I use daily data of KOSPI stocks from July 4, 2016, to July 1, 2026, from Yahoo. The sample initially consisted of the top 200 KOSPI stocks by market capitalization. Preferred shares were excluded, and approximately ten years of daily closing prices were collected. From the initial 200 stocks, only stocks with valid closing prices on at least 98% of the trading days within the sample period were retained, changing the sample size to 156 stocks. This filtering step was necessary because the factor models, residual estimates, OU parameters, and out-of-sample parameters all require aligned return observations. After collecting the closing prices, they were converted into daily returns. Excess returns were calculated by subtracting a 2.5% annual risk-free rate, converted to a daily rate by dividing by 252.

The first factor model used in this study is principal component analysis (PCA). It is a linear factor model with fixed factor loadings, beta. I first computed the covariance matrix of daily excess returns, performed eigendecomposition, and sorted the eigenvectors by eigenvalues to get the top k=5 principal components. Each stock’s fitted common return is reconstrcted by projecting excess returns onto the retained principal components.

The second model is instrumented principal component analysis (IPCA). IPCA is a linear factor model with its betas changing over time as linear functions of stock characteristics. While Kelly at al. (2019) analyzes 36 firm characteristics, due to small data set and limitations, this study uses 4 firm characteristics that are shown in the table below and an intercept. IPCA extends on the PCA and estimates the gamma matrix and factors by alternating least squares.

This table lists the firm characteristics and their definitions in this study  

The thrid model is the conditional autoencoder (CAE). CAE is a nonlinear model with a neural network mapping stock characteristics into betas, therefore a nonlinear specification of beta [1]. The model encodes stock returns, extracts the underlying factors, and reconstructs each stock’s return using the learned factor loadings and latent factors. The CAE uses the same characteristics inputs as IPCA.

For both IPCA and CAE, the hyperparameters are tuned using optuna, an optimization software that optimizes hyperparameters through using efficient sampling and pruning algorithms to search parameter spaces dynamically. [7] Hyperparameters of the IPCA are k and the ridge penalty gamma, hyperparameters of the CAE are hidden layer size, k, learning rate, and number of training epochs.

After each model is fitted, the residuals are obtained through the excess returns minus the model’s fitted common return component. The residuals are used as the main component in comparing factor models throughout the study. 

OU Parameters and Strategy Design

For each model and stock, I take the residuals and model it as the Ornstein-Uhlenbeck process (OU process), a model explaining a value that idiosyncratically drifts but means revert to an overall industry level. [6]

dXi(t) = κi (mi− Xi(t)) dt + σidWi(t), κi > 0.                 (1)

I fit an AR(1) model to each stock’s residual series, which is to regress each day’s residual on the previous day’s residual. The AR(1) coefficients are then converted into three OU parameters: κ as the speed of the mean-reversion [6], σ as the residual volatility, and m as the long-run residual mean. The estimated values of  ,  , and   for the three factor models are presented in the table below. The AR(1) coefficient b measures the day-to-day persistence of the idiosyncratic component. The R² measures how much of each stock’s residual variation is explained by its own value.

This table shows the average  (annual), average b (AR coefficient), and average R² of the residuals from PCA, IPCA, and CAE from the OU process.

These estimated parameters reveal significant limitations in modeling the daily KOSPI residuals as a continuous OU process. By dividing the average annual    by 252 trading days, a daily reversion speed of approximately 3.6 across all three models can be observed. The mean reversion time, calculated as the reciprocal of the daily   , is therefore about 0.27 trading days. Such rapid reversion makes it practically impossible to exploit using daily closing prices and strongly suggests that the models are fitting to high frequency noise. Furthermore, the AR(1) coefficients are negative for all three models, indicating that the residuals are oscillating rather than continuously mean reverting, which fundamentally violates the assumptions of an OU process.   

Despite these violations, I test statistical arbitrage, a strategy highly reliant on OU dynamics, as our first trading strategy. As the primary objective of this study is to compare factor models through the performance of residual based strategies, including statistical arbitrage serves as an essential baseline. Retaining this strategy also allows the demonstration of its unexpected underperformance under broken model assumptions and contrast it against other trading strategies.

 Statistical arbitrage is a strategy to exploit from short-term deviations of prices from a long-term equilibrium. [8] In this study the deviation is not of the raw stock returns’ deviation but one of the residuals. From the estimated OU parameters, a trading signal can be generated for statistical arbitrage strategy. This signal measures how far a stock’s residual is from its estimated equilibrium. If the signal is strongly positive, the strategy opens a short position expecting residual to decline; if the signal is strongly negative, the strategy opens a long position, expecting residual to rise. Positions are closed when the signal is short. [8] The following below is the threshold I use for this study.

Open short position (s_i,t> 0.5)  close short position (s < 0.25)

Open long position (s_i,t<-0.5)  close long position (s>-0.25) [8]

In this study, stocks with AR(1) R² < 0.02 are excluded and signals are shifted by one day before applying returns to avoid look ahead bias. Following this strategy it creates three statistical arbitrage portfolios: Statistical Arbitrage-PCA(StatArb-PCA), StatArb-IPCA, and StatArb-CAE.

The second trading strategy is the idiosyncratic momentum, also called as residual momentum. Conventional momentum strategies such as raw momentum strategy rank stocks by past performance and buy the stocks with the strongest past returns. I use the idiosyncratic momentum strategy, which ranks stocks on residuals instead of total returns. [9] It allows signals to be based on stock-specific performances after removing common factor movements, which improves the result. For each factor model, the strategy calculates a 12-month minus 1-month residual momentum score. It skips the most recent month to avoid short-term reversal effect. The strategy ranks stocks by this score each day, buys the top five stocks, and then weights them equally. This creates three idiosyncratic momentum portfolios: Idiosyncratic Momentum-PCA (IdioMom-PCA), IdioMom-IPCA, IdioMom-CAE. Also, a raw momentum benchmark is retained and tested too as a trading strategy for comparison.

Ultimately, the factor models and its residuals are evaluated by comparing the trading performance of strategies based on testing them through the out-of-sample testing method. The valid sample, from August 7, 2019 to July 1, 2026, is divided into an approximately 70% training period- August 7, 2019 to May 28, 2024; 1180 days- and a 30% test period-May 29,2024 to July 1, 2026; 507 days. Although the price data spans from July 4, 2016, the valid sample begins on August 7, 2019 because the 756day rolling market beta requires approximately three years of warmup before producing valid observations, leaving 1678 usable trading days. First, PCA, IPCA, and CAE are fitted to the in-sample data, producing in-sample residuals. OU parameters and idiosyncratic momentum lookback history are estimated from these in-sample residuals. The fitted models are applied to the test period to produce out-of-sample residuals. Finally, the statistical arbitrage and idiosyncratic momentum strategies are evaluated on the out-of-sample residuals, with their trading performance compared across models via Sharpe ratio and more.

III. Results

In total, 7 strategies are compared through the Out-of-sample testing, which are StatArb-PCA, StatArb-IPCA, StatArb-CAE, Idiomom-PCA, Idiomom-IPCA, Idiomom-CAE, and RawMom. Their performance is evaluated by annual returns, annual volatility, sharp ratio, max drawdown, and hit rate. KOSPI is also added as a benchmark. The results are summarized in Table 1, and the sharp ratio is visualized as a bar graph in Figure 1, as it serves as the primary measure of strategy performance in this study. 

This is a table of annual returns, annual volatility, sharpe ratio, maximum draw downs, and hit rate of the 7 strategies: statistical arbitrage PCA, statistical arbitrage IPCA, statistical arbitrage CAE, idiosyncratic momentum PCA, idiosyncratic momentum IPCA, idiosyncratic momentum CAE, and KOSPI’s Out-of-Sample performance. The test window is from May 29, 2024 to July 1, 2026.

From the summary table and figure 1, I can first compare the factor models within each strategy by their performance, as well as compare the performance across strategy type. Within the statistical arbitrage strategy, and from looking at the sharpe ratio of them, CAE residuals perform much better than IPCA and PCA, and the IPCA residuals performs better than the PCA residuals. This indicates that the higher flexibility a factor model has, the improved tradability residuals have when applied to statistical arbitrage strategy. 

Within the idiosyncratic momentum strategy, CAE residuals still outperform the other two factor models, but a difference from the statistical arbitrage is that PCA performs better than IPCA. Although it varies a little, this still shows how the highest flexibility factor model out of the three, CAE’s residuals perform the best within the strategy with the highest sharpe ratio. 

However, the biggest and clearest overall result is the difference between the two strategy families. The momentum strategies produced much higher annual returns, higher sharpe ratios, and overall better results. From all trading strategies, raw momentum was the most effective strategy in our sample too. 

IV. Discussion

The results indicate that, among the three factor models, the CAE produced the most effective residuals for both the statistical arbitrage and idiosyncratic momentum strategies. This supports the idea that more flexible factor models can generate more tradable residuals. PCA is transparent and stable, but its factor loadings are fixed over time, making it relatively restrictive. IPCA improves on this by allowing factor loadings to vary linearly with stock characteristics. While both PCA and IPCA remain linear models, CAE further allows nonlinear relationships between stock characteristics and factor loadings. The stronger performance of CAE residuals across both trading strategies suggests that this additional flexibility enables the model to better isolate stock-specific return components.

However, the results also show that the choice of trading strategy is just as important as the flexibility of the factor model. Although CAE produced the best-performing residuals among the factor models, the raw momentum strategy outperformed all residual-based strategies, suggesting that simple momentum signals were more effective than residual-based alternatives in the KOSPI market during the sample period. Furthermore, after raw momentum, all three idiosyncratic momentum strategies outperformed the statistical arbitrage strategies.

This drastic difference in performance between momentum strateigies and statistical arbitrage strategies could be explained by the estimated OU parameters. The parameter estimates from the methodology section has revealed that the daily residuals do not follow an OU process. Therefore, the statistical arbitrage strategy, relying heavily on mean reversion and OU derived signals, underperformed relative to momentum strategies. Furthermore, the breakdown of the OU process and the superior performance of momentum strategies suggest that  daily KOSPI residuals in this study’s sample are more effective for capturing persistent stock-specific trends than for exploiting short-term mean reversion. Therefore, while increasing factor model flexibility could improve the quality of the residuals, the profitability of residual-based trading also depends on whether the trading strategy is consistent with the underlying behavior of those residuals.

However, there are several limitations to be considered importantly in this study. First, the stock universe of the data is small and the sample period is relatively short compared with larger asset-pricing studies. This makes the results sensitive to the particular market period and to the stocks in the analysis, which could not make the same result when performed to a larger data set. Second, the data is from Yahoo Finance, which may contain missing values and data quality problems. Third, the back tests in this study do not fully include transaction costs, bid-ask spreads, liquidity constraints, and more. Especially in trading performance studies where small implementation details can strongly affect returns and the results, this must go through consideration. Lastly, Survivorship bias is another important limitation. The sample in this study is built from current large KOSPI stocks that have enough data during the study window, meaning that firms that have disappeared and merged are not fully represented. This can make performance appear better than it would be in reality since only the surviving firms remain in the dataset. In order to eliminate survivorship bias, a stronger design of data to include firms that were in the KOSPI universe at each point in time even if they later disappeared could be used.

Acknowledgment

I would like to express my gratitude to Prof. Panos Parpas at the Department of Computing at Imperial College London for his mentorship and guidance on financial modeling and mathematics studies.

References

1. Kelly, B. T., & Xiu, D. (2021). Factor models, machine learning, and asset pricing. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3943284
2. Kelly, B. T., Pruitt, S., & Su, Y. (2017). Instrumented Principal Component Analysis. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.2983919
3. Gu, S., Kelly, B. T., & Xiu, D. (2019). Autoencoder Asset Pricing Models. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3335536
4. Kim, E., Cho, T., Koo, B., & Kang, H.-G. (2023). Conditional autoencoder asset pricing models for the Korean stock market. PLOS ONE, 18(7), e0281783. https://doi.org/10.1371/journal.pone.0281783
5. Ryou, H.-Y., Kim, E., & Kang, H.-G. (2025). Application of Instrumented Principal Component Analysis in Korean Stock Market. Industrial Engineering & Management Systems, 24(4), 571–585. https://doi.org/10.7232/iems.2025.24.4.571
6. Avellaneda, M., & Lee, J.-H. (2008). Statistical Arbitrage in the U.S. Equities Market. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.1153505
7. Akiba, T., Sano, S., Yanase, T., Ohta, T., & Koyama, M. (2019). Optuna: A Next-generation Hyperparameter Optimization Framework. https://arxiv.org/pdf/1907.10902
8. Papanicolaou, G., & Yeo, J. (2016). Risk Control of Mean-Reversion Time in Statistical Arbitrage. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.2868525
9. Blitz, D., Huij, J., & Martens, M. (2009). Residual Momentum. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.2319861

About the author

Joheun Lee

Joheun Lee is currently a sophomore in The Webb Schools in Claremont, California. She has a deep interest in math along with finance, AI, and data science, and enjoys connecting theories with real-life problems. Beyond academics, she is passionate in playing volleyball in her school.

The post Comparing Linear and Nonlinear Factor Model Residuals through Statistical Arbitrage and Idiosyncratic Momentum in Korean Stock Market appeared first on Exploratio Journal.

1. Introduction

Nowadays, artificial intelligence has become increasingly important in many areas, both academically and in real life. Many tasks that were once completed by humans can now be done by AI. In finance, AI is very important because it can analyze large amounts of data in a short period of time and help investors make decisions. See Blackrock (2023).

The stock market is one of the most important parts of the economy because it connects directly to the companies that need money with investors who want to raise their money. This paper will mainly focus on the American stock market. When people buy stocks, they buy small ownership shares in companies. If a company performs well, the stock price will rise, and investors can earn money through capital gains or dividends. On the other hand, the stock market also contains risks, prices can change quickly due to company news, interest rates, investor emotions, and global events. See Investors Business Daily(2024). Because of this, many investors try to find ways to ‘beat the market,’ meaning they want to earn higher returns than the majority of people.

At present, more investors and companies are now using artificial intelligence to analyze data and make investment decisions. AI can deal with large amounts of information much faster than humans, including financial statements, stock prices, and news. This could help investors find patterns, predict price movements, and make better trading decisions. See Blackrock (2023). However, the main concerns are whether AI can provide reliable information to investors and whether AI can provide efficient output in a highly competitive stock market.

AI can improve investment research and decision-making because it allows investors to analyze large amounts of market information more quickly and efficiently. AI can help identify patterns, predict possible price movements, and support better trading decisions. However, since financial markets are competitive, any kind of strategy can quickly become less effective once many investors realize it and start using it. Therefore, this paper will examine how AI can help investors understand the market and improve their decisions, while others realized using AI as a tool to invest. Also considering consistently beating the market, especially for investors without strong data, technology, and risk management. See Blackrock (2023).

This paper will examine whether artificial intelligence can help investors beat the stock market. First, it will explain the basic nature of stocks, how the stock market operates, and the factors that influence stock prices. See Investor’s Business Daily (2024). Second, it will discuss normal and abnormal market returns and explain how benchmarks such as the S&P 500 can be used to determine whether an investor has outperformed the overall market. See Maverick (2015). Third, it will examine the meaning of ‘beating the market’ through Efficient Market Theory and explain why consistently earning abnormal returns is difficult. See Investopedia (n.d.-d). The paper will then discuss several possible ways investors may gain an advantage, including fundamental analysis, investor behavior, technology, and artificial intelligence. It will examine how AI is used in financial markets through algorithmic and high-frequency trading and use DeepSeek as an illustration of AI’s ability to perform financial analysis. See BlackRock (2023), See U.S. Securities and Exchange Commission (2020), See Lu et al. (2025). Finally, the paper will consider the limitations of AI, including its dependence on data, changing market conditions, competition, and unequal access to technology, as well as the policy implications created by the increasing use of AI in financial markets. These sections will provide the basis for evaluating whether AI can create an advantage that allows investors to beat the market and whether that advantage can continue over the long term.

The main question of this paper is whether artificial intelligence can help investors beat the stock market. To answer this question, several related questions will also be considered. What does it actually mean to ‘beat the market,’ and how can investors determine whether their returns are truly above normal market returns? If stock prices already reflect available information under Efficient Market Theory, where could AI find an advantage? The paper will also examine whether AI’s ability to process information faster, identify patterns, and automate trading can create abnormal returns. Finally, it will consider whether these advantages can remain effective over time after accounting for risk, competition, investment costs, and the increasing use of similar AI technology by other investors.

2. Nature of Stock Market

Before understanding whether artificial intelligence can help investors ‘beat the market,’ it is important to understand the fundamental nature of the stock market and what stocks actually are. A stock represents a small ownership share in a company. When investors buy stocks, they become partial owners of that company and are entitled to a portion of its value. Companies offer stocks mainly to raise money for growth, research, and expansion. Investors hope that the company will increase in value over time, which raises the stock price. Investors can also earn money through dividends, which are payments made by companies to shareholders from their profits. See Investor’s Business Daily (2024).

The stock market is the system where stocks are bought and sold. It acts as a market that connects investors who want to buy stocks and those who want to sell them. Prices in the stock market are determined by supply and demand. If more people want to buy a stock than sell it, the price usually goes up. If more people want to sell, the price usually goes down. This change in price is what makes the stock market both profitable and risky. See Investor’s Business Daily (2024).

In the United States, there are two major stock exchanges: the New York Stock Exchange (NYSE) and the NASDAQ. The NYSE is one of the oldest and largest stock exchanges in the world and is known for listing many large, established companies. The NASDAQ is a fully electronic exchange and is more heavily focused on technology companies such as Apple, Microsoft, and Amazon. The NYSE usually has more stable companies, while the NASDAQ has more fast-growing technology companies that often change prices more quickly. Both exchanges function in similar ways, but they have different structures and the types of companies. These exchanges allow stocks to be traded efficiently and in real time, which is essential for modern financial markets. See Investopedia (n.d-b).

Another important concept in understanding the stock market is the idea of stock indices. The stock index is a measurement that tracks the performance of a group of stocks. Instead of looking at one company, an index shows the overall direction of a market or sector. One of the most commonly used indices is the S&P 500, which tracks 500 of the largest companies in the United States. Because it represents a wide range of industries, it is often used as a benchmark to measure the performance of investors and funds. Another major index is the Dow Jones Industrial Average (Dow Jones), which tracks 30 large and well-known U.S. companies and is often used to represent the overall stock market. For example, when people say they are trying to ‘beat the market,’ they usually mean they are trying to outperform the S&P 500. See Maverick (2015), See Investopedia (n.d.-a).

Stock prices are influenced by many different factors. Company performance is one of the most important factors, including revenue, profit, and future growth potential. Economic conditions such as interest rates, inflation, and unemployment also affect stock prices. Fear and excitement can cause prices to rise or fall even when there is no major change in a company’s fundamentals. Global events such as wars, pandemics, or political changes can also impact the entire market. See Investor’s Business Daily (2024).

Because of these factors, the stock market is often difficult to predict. Even though data and analysis can help investors make better decisions, there is always risk. This uncertainty is one of the main reasons why many investors and financial professionals attempt to use advanced tools, including artificial intelligence, to improve decision accuracy and efficiency.

Overall, the stock market is a complex system where companies try to raise the money and investors attempt to earn money. It is influenced by economic data, company performance, global events, and human behavior. Understanding these basic structures before analyzing whether AI can provide an advantage is essential.

3. Normal vs. ‘Abnormal’ Market Returns

It is also important to understand what normal market returns are and how investment performance is measured before determining whether investors can ‘beat the market.’ In this section, the discussion focuses on the American stock market, especially U.S. market indices such as the S&P 500 and the Dow Jones Industrial Average. Since individual investments can have very different performances, investors need a standard to compare their results. By comparing investment returns with historical market performance, investors can determine whether their returns are above the overall market. See The Investopedia Team (2025).

3.1 Normal Market Returns

Normal market returns represent the typical returns investors can expect from investing in the overall stock market over a long period of time. Historically, the stock market has produced average annual returns of approximately 9–10% over periods of 10 years. See Maverick (2015). But this does not mean investors can earn exactly 9–10% every year. Because this average is calculated from many years of market growth and decline. See O’Shea (2026).

Annual market returns can change significantly depending on economic conditions, company performance and global events. During economic growth, companies may gain higher profits, causing stock prices to increase. However, during economic downturns or financial crises, stock prices may decline.

Because yearly returns can be unpredictable, investors usually focus on long-term performance instead of short-term results. A strong investment strategy is not measured by earning a high revenue in one year but consistently produce high revenue over many years. For example, an investor who earns around 10% annually over several decades may be considered successful because they have achieved returns similar to the overall market. See O’Shea (2026).

Understanding normal market returns is important because it provides a scale for evaluating investment performance. If an investor earns returns close to the market average over a long period, they are performing similarly to the overall market. Many investors choose index funds that track major market indices because achieving market-level returns is often more realistic than consistently earning higher returns. See Downey (2025).

3.2 Abnormal Market Returns

Abnormal returns refer to returns that are significantly higher or lower than the expected return. Abnormal returns usually refer to earning returns above a market standard. For example, if the overall market produces an average annual return of 10%, an investor who consistently earns 15% has abnormal returns. See The Investopedia Team (2026).

However, consistently achieving abnormal returns is difficult because financial markets are highly competitive. When new information becomes available, many investors quickly analyze and respond to it. As more investors discover and use similar strategies, the advantage of that strategy often becomes smaller. This makes it challenging for any investor to maintain above-average returns over a long period of time. See Downey (2025).

A temporary high return does not necessarily prove that the investor can consistently beat the market. Short-term success may occur by market conditions, taking higher risks, or totally by accident. To demonstrate the ability to consistently beat the market, the investor must achieve higher returns consistently over many years while considering all the factors such as risk, costs, and changes in market. See The Investopedia Team (2026), See Downey (2025).

This difference between normal and abnormal returns is essentially while analyzing whether artificial intelligence can help investors beat the market. AI may allow investors to deal with enormous amounts of information faster, identify the patterns, and make decisions more efficiently. However, the crucial point is whether AI can create a long-term advantage that produces consistent abnormal returns in nowadays society. See BlackRock(2023), See Downey (2025).

3.3 Figure 1.Dow Jones diagram Over the Past 50 Years

See Macrotrends (n.d.).

This graph shows the long-term performance of the Dow Jones Industrial Average over the past 50 years. Even though the market experienced several periods of decline, such as the dot-com bubble, the 2008 financial crisis, and the COVID-19 market crash, the overall trend has increased over time. This shows that stock market returns can change significantly from year to year, investors who hold stocks in the long term have historically benefited from staying invested. The graph also explains why investors focus on average returns over many years instead of only looking at one year of performance. Some years may have significantly higher or lower returns, but the stock market has produced an average annual return of around 9–10% over a long period of time. See Thune, K (2023).

3.4 Standard Deviation and Investment Risk

Standard deviation can also help explain the risk shown in the graph. Standard deviation measures the distance between yearly returns and average return. If the standard deviation is high, it means the returns are more spread out, so the market has more ups and downs. If the standard deviation is low, it means the yearly returns are closer to the average and more stable. In the Dow Jones graph over past 50 years, the market does not increase smoothly every year. There are many sharp increases and decreases during different periods. This shows that even though the long-term average return may be around 9–10%, the actual return in a single year can be much higher or much lower. Therefore, standard deviation is important because it shows that investors must consider risk, not only average return, when evaluating stock market performance. See Investopedia (n.d.-c).

4. The Meaning of ‘Beating the Market’

Although investors may achieve returns above the market average, consistently producing abnormal returns is extremely difficult. This leads to an important question in finance: if many investors are trying to find advantages, why is it still difficult to consistently beat the market?

4.1 Efficient Market Theory (EMT)

Efficient Market Theory, also known as the Efficient Market Hypothesis (EMH), explains that stock prices reflect available information, including company performance, economic conditions, and public news. When new information appears, investors react quickly, causing prices to adjust. Because many investors compete using the same information, it is extremely difficult to consistently earn returns above the overall market. The implications of Efficient Market Theory are important for investors. If markets are efficient, investors cannot consistently beat the market by only using public information. Therefore, many investors choose passive investing strategies, such as index funds that follow major market indices.(See Investopedia, n.d.-d).

4.2 Three Forms of EMT

EMT has three forms. Weak-form efficiency states that stock prices already reflect past prices and trading patterns. Semi-strong efficiency states that prices reflect all publicly available information, including financial reports and company announcements. Strong-form efficiency argues that prices reflect both public and private information, however this form is controversial because investors cannot legally use private information to gain benefits. This paper will focus on Strong-form efficiency. (See Investopedia, n.d.-d, See U.S. Securities and Exchange Commission, 2023).

4.3 Assumptions for EMT to Hold

Efficient Market Theory helps explain why ‘beating the market’ is such a challenging goal. When information set public and already reflected in stock prices, investors cannot gain an advantage by using the same information available to everyone else. If a strategy creates an advantage, other investors may quickly discover and use the same strategy, reducing its effectiveness. For example, if an investor identifies a stock that appears undervalued, other investors may notice the same opportunity and purchase the stock. Their actions will cause the stock price to rise, reducing the≠ advantage. As more investors compete for the same opportunities, it becomes more difficult to maintain abnormal returns over a long period of time. Therefore, short-term success does not always prove that an investor has a reliable strategy, as high returns may result from favorable market conditions, higher risks, or chance (See Investopedia, n.d.-d).

5. Five Ways Investors May ‘Beat’ the Stock Market: A critical Assessment

Efficient Market Theory explains that consistently beating the market is difficult because available information is quickly reflected in stock prices. However, investors may achieve returns above the market through different strategies. Here are five possible ways include using non-public information, fundamental analysis, investor behavior, advanced technology, and taking greater risks.

5.1. Using Non-Public Information

One possible way to gain an advantage is by having important private information before it becomes public, which is also known as insider trading. Since the information has not yet been reflected in the stock price, an investor may be able to act before the rest of the market. This challenges the strong form of EMT, which states that stock prices reflect both public and private information.

 It also violates the EMT assumption that all relevant information is public because some investors have information that others do not. This advantage is not sustainable because it disappears once the information becomes public, and trading based on non-public information is illegal. See U.S. Securities and Exchange Commission (2023).

5.2. Finding Undervalued Stocks Through Fundamental Analysis

Investors can use fundamental analysis to find stocks they believe are undervalued. They can analyze a company’s financial statements, revenue, profits, debt, management, and future growth. If an investor finds a strong company whose stock price is lower than its actual value, the investor may earn abnormal returns when the price increases. 

This can violate the EMT assumption that the ability to extract information from a given signal is uniformly distributed because some investors may be better at analyzing the same financial information than others. This advantage may be sustainable for skilled investors, but it can decrease when other investors recognize the same opportunity.

5.3. Taking Advantage of Investor Behavior

Investor behavior can also create opportunities to beat the market. Investors may be influenced by fear, excitement, overconfidence, and herd behavior. For example, negative news may cause investors to be afraid and no willingness to hold the stock, causing its price to fall significantly. An investor who recognizes that the market has overreacted may buy the stock and benefit if the price later recovers. 

This violates the EMT assumption that people form their expectations in a rational manner because emotions and biases can affect their decisions. This advantage may continue because investor biases still exist, but predicting this correctly is difficult.

5.4. Taking Greater Risks

Investors may also achieve higher returns by taking greater risks. For example, an investor may invest heavily in a small number of stocks or choose faster-growing companies. If these investments perform well, the investor may earn significantly higher returns than the overall market. Higher returns should be considered with the amount of risk. For example, if the S&P 500 earns 10% and an investor earns 15% by taking significantly greater risks, the additional return will be related to the additional risk. Therefore, risk and return should both be considered when determining whether an investor has successfully beaten the market. 

This method does not directly violate an EMT assumption because efficient markets can still provide higher expected returns for investors who accept greater risk. This strategy may produce higher returns over time, but those returns do not necessarily represent an advantage over the market because they come with greater risk.

5.5. Using Hyper Technology, Data, and Artificial Intelligence

In this section, l  will examine more deeply how technology and artificial intelligence may give investors an advantage in the stock market. Technology and AI may give investors another advantage. AI can process enormous amounts of information, including financial reports, stock prices, economic data, and news, much faster than humans. It can also identify patterns that may be difficult for individual investors to discover. If AI identifies useful information or patterns before other humans, it may help investors make better decisions quicker and earn abnormal returns. See BlackRock (2023). 

This violates the EMT assumption that the ability to extract information from a given signal is uniformly distributed because investors with better AI technology may analyze information faster and more effectively than others. This advantage may become less sustainable as AI becomes more widely used and more investors discover the same patterns and opportunities.

These five methods show separate ways investors are able to gain returns above the market, including using different information, fundamental analysis, investor behavior, advanced technology such as AI, or taking greater risks. Each method involves different opportunities and risks, and its effectiveness can vary depending on market conditions and how it is used.

6. Nature and Impact of High-Frequency and Algorithmic Trading-A deeper look

One technology in investing is algorithmic trading. Algorithmic trading uses computer programs to automatically make trading decisions based on certain rules, market information, or mathematical models. Instead of having an investor watch the market, the computer can continuously monitor stock prices, trading volume, price changes, and other information. Once certain conditions are reached, the system can automatically execute a trade. This allows investment decisions and trades to be completed much faster than humans. See U.S. Securities and Exchange Commission (2020).

6.1 Nature of Algorithmic Trading and High Frequency.

6.1.1 Nature of Algorithmic Trading

These algorithms can be designed for different purposes. For example, an algorithm may compare the current price of a stock with its historical prices, monitor changes in trading volume, or compare prices between related securities. When the system identifies a condition that matches its strategy, it can immediately decide whether to buy or sell. This is especially useful because financial markets constantly produce new information, making it difficult for investors to monitor every change at the same time. Algorithms can also follow the market continuously without being affected by tiredness or emotions such as fear and excitement. See U.S. Securities and Exchange Commission (2020).

Before an algorithm is used in the real market, investors can also test it using historical market data. This process, commonly known as back testing, allows investors to examine how a trading strategy would have performed under previous market conditions. A strategy that performs successfully in historical testing may then be further tested and adjusted before real money is invested. Past performance does not guarantee that the same strategy will continue to work because future market conditions can be different. See U.S. Securities and Exchange Commission (2020).

6.1.2 Nature of High Frequency

High-frequency trading, also known as HFT, is a more advanced type of algorithmic trading that focuses heavily on speed. High-frequency traders use powerful computers and fast connections to analyze market information and complete a large number of trades within a very short period. The profit from an individual trade may be very small, but making a large number of trades make these small profits become larger. Speed is especially important because a trading opportunity may only exist for a very short period before other investors recognize it and stock prices adjust. See U.S. Securities and Exchange Commission (2020).

High-frequency trading operates on extremely short time periods. HFT systems continuously receive information about prices, buy and sell orders, and trading volume from financial markets. The computer analyzes this information and automatically submit, change, or cancel orders when it identifies a trading opportunity. Because these decisions are automated, the entire process can occur much faster than a human investor could analyze the information and manually place a trade. See U.S. Securities and Exchange Commission (2020).

One important factor in HFT is latency, which refers to the delay between receiving market information and responding to it. Even an exceedingly small delay can matter because many HFT firms may be competing for the same opportunity. For this reason, firms invest heavily in powerful computers, high-speed networks, and technology that allows them to receive market data and execute orders as quickly as possible. Some firms also place their computer systems close to the exchanges’ systems, a practice known as co-location, to reduce communication delays. See U.S. Securities and Exchange Commission (2020).

For example, a stock may temporarily have slightly different prices across markets. A high-frequency trading system may recognize this small difference and quickly buy at the lower price and sell at the higher price. The profit from one trade may be extremely small, but the strategy can become meaningful when it is repeated many times. As other trading systems recognize the same price difference, their buying and selling causes the prices to move closer together and the opportunity disappears. This explains why HFT companies invest heavily in computing power and fast connections. Even an exceedingly small difference in the time needed to receive information and execute a trade can affect whether the opportunity still exists. See U.S. Securities and Exchange Commission (2020).

6.2 Artificial Intelligence in Algorithmic Trading

Artificial intelligence can improve algorithmic trading by allowing computer systems to analyze larger and more complicated information. Traditional algorithms usually follow rules created by humans, such as buying or selling when a stock reaches a certain price or certain condition. AI models can analyze relationships between different types of data and identify patterns that may be difficult for humans to recognize. This can help trading systems react more efficiently to changing market conditions.See BlackRock (2023).

Figure 2. Process of AI-Supported Algorithmic Trading

See Coin Bureau (n.d.).

Figure 2 shows how an AI-supported trading system can move from receiving market information to executing a trade. Market information such as stock prices, trading volume, financial news, and economic data can first be collected and analyzed by an AI model. The system may then identify a pattern or trading opportunity and generate a buy or sell signal. If the conditions of the trading strategy are satisfied, the algorithm can automatically execute the trade. This process can occur continuously and much faster than a human investor could complete the same steps.

Machine learning can make this process more advanced because the system can learn relationships from historical data instead of depending only on fixed rules created by humans. For example, an AI model may consider price movements, trading volume, volatility, and other market information together to estimate the probability that a stock price will increase or decrease. The model can then use this prediction as one factor when deciding whether to trade. This does not mean that AI knows what the stock price will be in the future, but it can use patterns in data to estimate possible outcomes. See BlackRock (2023).

AI can also analyze information that is more difficult to represent with simple trading rules. For example, financial news and company announcements contain large amounts of written information. AI systems can analyze this information and identify whether it may have a positive or negative effect on a company. Combining this information with stock prices and trading data allows an algorithm to consider several types of information at the same time. This creates a major difference between a simple rule-based algorithm and a more advanced AI-supported trading system. See BlackRock (2023).

6.3 Impact of Algorithmic and High-Frequency Trading on the Market

Algorithmic and high-frequency trading can also affect the overall market. Research reviewed by the U.S. Securities and Exchange Commission found that some forms of high-frequency trading can improve market liquidity and price efficiency during normal market conditions. For example, automated trading can increase competition between buyers and sellers and help stock prices respond quickly to latest information. See U.S. Securities and Exchange Commission (2020).

Liquidity is important because it represents how easily investors can buy or sell an investment without causing a large change in its price. When automated traders continuously participate in buying and selling, there may be more opportunities for other investors to complete their trades. Algorithmic trading can also contribute to price discovery because new information can be analyzed and reflected in stock prices very quickly. See U.S. Securities and Exchange Commission (2020).

On the other hand, these systems contain risks. A computer can execute many trades extremely quickly, which means that an error in an algorithm may also affect the market very quickly. During unusual periods of market stress, some algorithmic trading activities may increase costs for investors. An algorithm that works successfully during normal conditions may also react differently when the market experiences an unusual event that was not considered when the system was developed. This shows that speed and automation can improve the trading process, but they also require careful control and risk management. See U.S. Securities and Exchange Commission (2020).

Faster price discovery means that when essential information becomes available, trading activity can help move a stock toward a price that better represents that information. Therefore, algorithmic trading does not only affect investors using the technology; it can also influence how quickly the overall market reacts to new information.See U.S. Securities and Exchange Commission (2020).

The development of algorithmic trading also changes the competition between investors. Investors with faster computers, better algorithms, more data, and stronger AI models may be able to analyze and respond to information before other investors. As more financial institutions develop similar technology, these differences can become smaller and trading opportunities may disappear more quickly. See BlackRock (2023).

This competition can also make the market itself more efficient. When an algorithm discovers a price difference and trades on it, its actions help move the price toward a level that reflects the available information. Therefore, algorithmic trading can create an advantage for investors who recognize opportunities quickly, while at the same time causing those opportunities to disappear more quickly. The success of algorithmic trading is not only based on finding useful patterns, but also on how quickly and accurately investors can act on those patterns before the rest of the market responds. See U.S. Securities and Exchange Commission (2020), See Investopedia (n.d.-d).

6.4 DeepSeek as an Illustration of AI in Financial Analysis

DeepSeek is a Chinese artificial intelligence company founded by Liang Wenfeng. Before establishing DeepSeek, Liang had a strong background in mathematics, artificial intelligence, and quantitative investing. He studied electronic information engineering at Zhejiang University and later earned a master’s degree in information and communication engineering. His early interest in using mathematics and computer technology to analyze financial markets eventually led him into quantitative investing. See Reuters (2025).

Liang later co-founded High-Flyer, a Chinese quantitative hedge fund. Unlike traditional investors who may depend heavily on personal experience and individual judgment, quantitative investment firms use mathematical models, large amounts of financial data, and computer algorithms to identify investment opportunities and make trading decisions. High-Flyer increasingly used artificial intelligence and computing technology as part of this process. This experience gave Liang and his team a strong connection between AI, data analysis, and financial markets. See Reuters (2025).

In 2023, Liang expanded his focus beyond using AI mainly for quantitative investing and established DeepSeek to focus on developing more advanced artificial intelligence models. DeepSeek began researching large language models that could perform tasks such as reasoning, coding, mathematics, and information analysis. The company developed several AI models and released DeepSeek-V3 in December 2024, followed by the reasoning-focused DeepSeek-R1 in January 2025. DeepSeek-R1 received significant international attention because of its reasoning performance and the company’s open approach to releasing its models.See Reuters (2025), See DeepSeek (2024).

The company has a connection to quantitative investing through High-Flyer. Before developing general-purpose AI models, Liang and his team already had experience using algorithms, large datasets, and computing power to analyze financial markets. DeepSeek therefore provides an interesting example of how technology originally developed and applied in a quantitative investment environment can contribute to the development of more advanced AI systems capable of analyzing many different types of information. Liang remains closely connected to High-Flyer today, while also leading DeepSeek, showing that his work in artificial intelligence continues to have a strong connection with his background in quantitative investing. See Reuters (2025).

DeepSeek can be used as an example of how AI can support financial analysis. Investors often need to read financial statements, compare companies, analyze economic information, and perform calculations before making investment decisions. These tasks can require investors to examine enormous amounts of information from different sources. AI can make this process more efficient by organizing and analyzing information much faster than a human investor.

For example, investors can use AI to analyze a company’s revenue, profits, expenses, and debt. AI can also summarize financial documents, compare different companies, and explain complicated financial information. An investor could provide financial information from several companies and ask an AI model to compare their revenue growth, profit margins, debt levels, and other financial measurements. Instead of manually searching through every part of a financial report, the investor could use AI to identify essential information and then examine it more carefully. This allows investors to process larger amounts of information in less time.

DeepSeek can also perform reasoning and numerical tasks that are important in financial analysis. Financial decisions often require more than simply finding information. Investors may need to calculate financial ratios, recognize relationships between different numbers, and determine what changes in financial performance could mean for a company. An AI model that combines information extraction with numerical reasoning can therefore help investors move from collecting information to analyzing it.

Research also shows DeepSeek’s ability to perform financial tasks. BizFinBench tested 25 AI models using 6,781 real-world financial questions involving calculation, reasoning, information extraction, and financial knowledge. DeepSeek-R1 performed especially well in numerical calculation and information extraction. See Lu et al. (2025).

These abilities could be useful when investors need to analyze many companies at the same time. A human investor has limited time to read financial reports from hundreds of companies, while an AI system can process information much more quickly. AI could therefore be used as an initial screening tool to identify companies or financial information that may deserve further analysis. This does not mean that DeepSeek can automatically identify which stocks will increase in price, but it can reduce the amount of time required to complete some parts of investment research.

At the same time, DeepSeek has limitations. The research found that no single AI model performed best in every financial task. Powerful performance in calculation or information extraction also does not necessarily mean that the model can accurately predict future stock prices. A company may have strong financial results while its stock price falls because investors already expected even better results. Stock prices can also be affected by interest rates, economic conditions, investor behavior, and unexpected events that are difficult for an AI model to predict. Therefore, AI can make financial analysis faster and more efficient, but investors still need to evaluate its results carefully instead of completely depending on AI when making investment decisions. See Lu et al. (2025).

7. Limitations

Artificial intelligence can help investors analyze financial information and make investment decisions, but it also has several important limitations. Financial markets are affected by many different factors, and AI cannot perfectly understand or predict every change. Its effectiveness depends on the quality of its data, market conditions, and how investors use its results.

One limitation is AI’s dependence on data. AI models use enormous amounts of historical information, including stock prices, financial statements, economic data, and news, to identify patterns and make predictions. If the data is inaccurate, incomplete, or biased, the results may also be inaccurate. Historical patterns also do not always represent what will happen in the future.

Unexpected events, Wars, pandemics, political changes, natural disasters, and sudden company problems can quickly affect stock prices. These situations may be very different from the historical information used by an AI model. As a result, a model that performs successfully during normal market conditions may perform poorly when the market suddenly changes.

AI models can also make mistakes when analyzing financial information. They may misunderstand information, make calculation errors, or provide conclusions that appear reasonable but are incorrect. The DeepSeek example discussed previously demonstrates this limitation. DeepSeek-R1 performed strongly in several financial tasks but did not perform best in every area. Investors cannot assume that one AI model will always provide accurate financial analysis. See Lu et al. (2025).

Competition can also reduce the advantage provided by AI. If an AI model discovers a useful market pattern, other investors have the same opportunity using similar technology. As more investors find out, AI will reduce the original advantage. This connects with Efficient Market Theory because competition causes new information to become reflected in market prices more quickly.

Another limitation is unequal access to technology. Large financial institutions may have better data, stronger computers, professional programmers, and more advanced AI models than individual investors. Investors may also become too dependent on AI and follow its recommendations without carefully checking the results. AI can support investment decisions, but human judgment and risk management remain important.

Overall, problems with data, unexpected events, AI errors, competition, and unequal access to technology can all affect the performance of AI in financial markets. These are the limitations when AI is used for investment analysis.

8. Policy Implications

As artificial intelligence becomes widely used in financial markets, regulators need to consider its potential risks. AI can improve financial analysis and trading, but incorrect use may affect investors and market stability. Policies should allow technological development but protect investors.

One important issue is transparency. Some AI models are difficult to understand, so financial institutions should remain responsible for decisions made using their systems. Human supervision is also important because AI can make mistakes or respond incorrectly to unusual market conditions.

Regulators also need to prevent misleading claims about AI. Some companies may exaggerate how they use AI or what their technology can achieve. The U.S. Securities and Exchange Commission has described misleading claims about AI as ‘AI washing’ and has taken enforcement actions against investment advisers for making false or misleading statements about their use of AI. See U.S. Securities and Exchange Commission (2024).

Market stability and data security are also important. Algorithmic trading systems can execute many trades very quickly, meaning errors can spread rapidly. AI systems may also use large amounts of sensitive financial information, which needs to be properly protected.

Financial institutions should also be required to test and monitor AI-based trading systems before and during their use. Because automated systems can execute many trades within a very short period, an error can spread quickly through the market. Firms should have risk controls that can limit or stop trading when unusual behavior is detected. See U.S. Securities and Exchange Commission (2020).

Overall, financial policies should balance the development of AI with investor protection, transparency, market stability, and data security.

9. Conclusion

AI can help investors beat the market in some situations because it can process large amounts of information quickly, identify patterns, and react faster than humans. Investors with better data, stronger models, and faster technology may gain an advantage before other investors recognize the same opportunity. See BlackRock (2023), See U.S. Securities and Exchange Commission (2020).

The main issue is whether this advantage can continue over a long period. If more investors use similar AI systems, they may discover the same opportunities, causing stock prices to adjust and reducing the original advantage. This is connected with Efficient Market Theory because competition causes information to become reflected in prices more quickly. See Investopedia (n.d.-d).

AI also has limitations. It depends on historical data, can make mistakes, and may perform poorly when market conditions suddenly change. The DeepSeek example shows that AI can perform well in some financial tasks, but no model performs best in every area. See Lu et al. (2025).

Therefore, AI can beat the market during certain periods, but there is not enough evidence to show that it can consistently beat the market over the long term after considering risk, competition, costs, and changing market conditions. AI is better understood as a powerful tool that can improve investment decisions rather than a guaranteed way to beat the market.

References:

BlackRock. (2023, June 15). How AI is transforming investing. BlackRock. https://www.blackrock.com/us/individual/insights/ai-investing

Coin Bureau. (n.d.). What is an AI trading bot? Coin Bureau.
https://coinbureau.com/education/what-is-an-ai-trading-bot

DeepSeek. (2024). DeepSeek-V3. DeepSeek.
https://www.deepseek.com/en/news/deepseek-v3/

Downey, L. (2025, November 15). Efficient market hypothesis (EMH): Definition and critique. Investopedia.
https://www.investopedia.com/terms/e/efficientmarkethypothesis.asp

Investopedia. (n.d.-a). Dow Jones Industrial Average vs. S&P 500: What’s the difference? Investopedia.
https://www.investopedia.com/ask/answers/difference-between-dow-jones-industrial-average-and-sp-500/

Investopedia. (n.d.-b). Stock market: Definition, how it works, and key concepts. Investopedia.
https://www.investopedia.com/terms/s/stockmarket.asp

Investopedia. (n.d.-c). Understanding standard deviation in investment risk management. Investopedia.
https://www.investopedia.com/ask/answers/021915/how-standard-deviation-used-determine-risk.asp

Investopedia. (n.d-d). What are the differences between weak, strong, and semi-strong versions of the efficient market hypothesis? Investopedia.
https://www.investopedia.com/ask/answers/032615/what-are-differences-between-weak-strong-and-semistrong-versions-efficient-market-hypothesis.asp

Investor’s Business Daily. (2024, May 22). What is the stock market? Moneyworks from Investor’s Business Daily.
https://get.investors.com/moneyworks/how-to-start-investing/what-is-the-stock-market/

Lu, G., Guo, X., Zhang, R., Zhu, W., & Liu, J. (2025). BizFinBench: A business-driven real-world financial benchmark for evaluating LLMs. arXiv.
https://arxiv.org/abs/2505.19457

Macrotrends. (n.d.). Dow Jones – 100 year historical chart. Macrotrends.
https://www.macrotrends.net/1319/dow-jones-100-year-historical-chart

Maverick, J. B. (2015). S&P 500 average return and historical performance. Investopedia.
https://www.investopedia.com/ask/answers/042415/what-average-annual-return-sp-500.asp

O’Shea, A. (2026, March 4). What is the average stock market return? NerdWallet.
https://www.nerdwallet.com/investing/learn/average-stock-market-return

Reuters. (2025, January 29). High-Flyer: The AI quant fund behind China’s DeepSeek. Reuters.
https://www.reuters.com/technology/artificial-intelligence/high-flyer-ai-quant-fund-behind-chinas-deepseek-2025-01-29/

The Investopedia Team. (2025, December 31). What does it mean to ‘beat the market’? Investopedia.
https://www.investopedia.com/ask/answers/03/100903.asp

The Investopedia Team. (2026, May 22). Understanding abnormal returns: Causes and impact on investments. Investopedia.
https://www.investopedia.com/terms/a/abnormalreturn.asp

Thune, K. (2023, January 2). What is the average return of the stock market? Seeking Alpha.
https://seekingalpha.com/article/4502739-average-stock-market-return

U.S. Securities and Exchange Commission. (2020, August 5). Staff report on algorithmic trading in U.S. capital markets. U.S. Securities and Exchange Commission.
https://www.sec.gov/files/Algo_Trading_Report_2020.pdf

U.S. Securities and Exchange Commission. (2023). Insider trading. Investor.gov.
https://www.investor.gov/introduction-investing/investing-basics/glossary/insider-trading

U.S. Securities and Exchange Commission. (2024, March 18). SEC charges two investment advisers with making false and misleading statements about their use of artificial intelligence. U.S. Securities and Exchange Commission.
https://www.sec.gov/newsroom/press-releases/2024-36

About the author

Yiqi Xu

Yiqi is a Grade 12 student in Ontario, Canada, interested in mathematics, finance, artificial intelligence, and computer science. He is especially interested in understanding financial markets and the growing role of technology in investing. He enjoys exploring new ideas in math, coding, and finance.

The post Can you use AI to ‘beat’ the market? appeared first on Exploratio Journal.

1.0 Introduction

Social Security’s Old-Age and Survivors Insurance trust fund is projected to reach insolvency by 2032 [1], forcing policy-makers to now address certain queries regarding the distribution of the program’s benefits and the process in which it collects capital. The most glaring question lies in the differences between the demographics that constitute the United States: how does the contribution-to-benefit ratio differ between demographics given each subgroup’s life expectancy, and how does this ratio impact the Social Security system(specifically the Old-Age and Survivors Insurance, also known as the OASI)? This paper examines the question, specifically within the Indian American Subgroup (U.S. residents that identify as Asian Indian). It will analyze the life-expectancy of Indian Americans to further understand how long they contribute and benefit from the Social Security trust fund along with providing a comparison to other subgroups (African Americans, Whites, Latinos, etc). 

The Indian-American demographic is characterized as an anomaly from both an economic and physiological standpoint, justifying a specialized study. The median household income of Indian-Americans is $145,000, an income that is substantially higher than the median for overall Asian-Americans($100,000) [2]. The Asian-Born, first-generation subgroup also displays exceptionally high longevity, in which they have a life-expectancy significantly longer than their US-Born counterparts [3]. 

Beyond the anomalous nature of Indian-Americans, existing literature tends to be monolithic. That is, most studies treat Asian-Americans as a single group, disregarding the group’s immense heterogeneity; studying Indian-Americans exclusively allows for a granular understanding of how each demographic affects the Social-Security system. 

An analysis of how this demographic affects the financial sustainability of the program compared to other demographics serves of great importance given the current fiscal health and poor outlook of the Social Security trust fund. Due to a variety of socio-logical and economic factors, Social Security is expected to reach insolvency by 2032-2033, with the OASI trust fund triggering automatic benefit cuts of roughly 22% [1]. Understanding which demographics are net contributors versus net beneficiaries further allows for beneficial policy-making, effectively mitigating these projected cuts in benefits. 

Given that different demographics have different life-expectancies, the contribution-to-benefit ratio varies significantly between subgroups; Black-Americans, for example, generally have lower life-expectancies than other subgroups to which they receive, on average, 25 percent less benefits than White-Americans [4]. Understanding the visible disparity between demographics in the amount of benefits they collect further drives changes in policy that accommodate for each demographic’s general life-expectancy. 

Ongoing debates persist regarding the reforms needed to ensure the program’s fiscal health. Some economists propose to cut the benefits retirees receive or increase the minimum age of collection, while others argue for changes on the contribution side such as an increase in the tax rate or the taxable annual earning maximum. This study aims to provide further insight into how these reforms would affect each demographic, and how each demographic is currently affecting the program. 

Analyzing the effects of immigration is also a necessity. Throughout the history of the program, immigration has shown to be a major beneficiary; comprising 19% of the US labor force despite being 14.3% of the total population, immigrants strengthen Social Security’s payroll-tax base immensely. An examination of this specific population (Asian Indian U.S. residents, a population that is 66% foreign-born [5]) provides initial insight into how the demographic affects the Social Security trust fund, promoting the crucial changes that need to be made in the policy surrounding the program. 

2.0 The Nature of the Social Security System

Social Security operates as a “pay-as-you-go” system; that is, it does not function as a personal savings account [6]. When one pays a Federal Insurance Contributions Act (FICA) tax, they are not accumulating money for their own future use; rather, the payroll taxes paid today directly fund the benefits for those that are currently eligible for the program (consisting of retirees, workers with disabilities, surviving spouses, and low-income individuals) [6]. 

The Great Depression, primarily caused by the stock market crash and systemic banking failures of 1929, put millions of Americans under immense economic strain. The disasters of the period effectively wiped out life savings for the elderly and heightened job scarcity [7]. Then, as a part of Franklin D. Roosevelt’s “New Deal” (a series of economic, social, and political reforms)[8], the 1935 Social Security Act was enacted. The legislation was renowned for providing an economic safety net for Americans; it distributed benefits for retirees funded by payroll taxes and provided a temporary income to those who lost their jobs [9]. As the country’s fiscal stress eased, the program began to broaden its scope beyond workers, and further accommodated for dependents and the disabled [10]. 

Contributions are collected through a mandatory payroll tax split evenly between employees and employers. As of 2026, the Old-Age, Survivors, and Disability Insurance tax (OASDI) is at a rate of 12.4%, with employers and employees each paying 6.2% of the total rate [11]; those that are self-employed are mandated to pay the full 12.4% tax rate [12]. The annual taxable income for the OASDI is $184,500, implying that employees contribute an annual maximum of $11,439 [13]. Given this wage cap, it is ineffectual to split each demographic subset based on their raw annual income for an analysis of Social Security contributions; every worker that earned an annual income above $184,500 contributes the same amount to the trust fund. Rather, each demographic subset should be stratified based on whether they pass this annual taxable income for the OASDI tax. This cap also persists for the benefits that retirees receive. A retiree’s past 35 highest annual income years is averaged into monthly earnings known as the Average Indexed Monthly Earnings(AIME); this AIME is calculated with the $184,500 cap, meaning the maximum AIME one could have is $14,358 [14]. The retiree’s AIME is then distributed through a bucket system, in which 90% of the first $1286 of the AIME, 32% of the amount between $1286 and $7749, and 15% of anything above $7749 all get summed as the retiree’s monthly benefits check [15]. This system is displayed in Table 1. 

Table 1: 2026 Social Security Benefit Formula. 

Source: Social Security Administration, Office of the Chief Actuary, Author Made

Given the program’s pay-as-you-go nature, the system’s solvency is dependent on having enough current workers to accommodate the current beneficiaries. The program’s projected insolvency is largely due to the downward trajectory of the worker-to-beneficiary ratio. In 1960, this ratio was 5:1, meaning that for every 5 workers there was 1 beneficiary drawing out from the program. Today, the ratio has lessened to 3:1, marking a considerable decline, sparking a collective anxiety of the program’s ability to provide 100% of its promised benefits[16]. The decreasing trend in the worker-to-beneficiary ratio from 1955 to 2025, along with the predicted trend from 2025 to 2075 is shown in Figure 1. 

Figure 1: Ratio of Workers to OASI Beneficiaries Over Time.

Source: Peter G. Peterson Foundation, Social Security Administration 

Along with the worker-to-beneficiary ratio, the average life expectancy of workers dictates the system’s financial health as it determines how long a beneficiary draws money out of the program. Decades of literature has indicated a robust relationship between one’s demographic and their life expectancy [17], further indicating that the distribution of different demographics would have a profound impact on the Social Security system [18]. Benefits may be claimed as early as age 62 and as late as age 70 (one can still start collecting benefits past 70 but their monthly payment will not increase anymore). Depending on when one begins collecting, they may see an increase or decrease in their calculated monthly benefits. To receive 100% of their calculated benefits, one must start collecting at their full retirement age which is 66 or 67 depending on their year of birth. These adjustments are actuarially designed so that a beneficiary with average life expectancy receives approximately the same total value regardless of when they claim [19]. Along with the age requirement, one has to accumulate 40 credits to be eligible for benefits in which they can earn a maximum of 4 credits yearly [19]. Assuming the annual maximum number of credits is earned each year, it would take a worker 10 years to fulfill their working requirement to receive benefits. This prerequisite also alters the way in which immigrants affect the Social Security program. Many immigrant workers enter the country and leave the country on an H-1B visa in less than 10 years, meaning they contributed to the trust fund but did not collect benefits. 

3.0  Demographic Facts about Indian-Americans

First-generation Indians occupy a substantial portion of America’s foreign-born population, comprising 6 percent of the total foreign-born population of 50.2 million [20]. The growth of India’s foreign-born population has accelerated faster than the total foreign-born population’s growth; the Indian immigrant population grew 78% from 2010 to 2024 which is triple the 26% growth rate of the overall U.S. foreign born population [20]. This growth has persisted for 3 decades: since 2000, the Indian population in the U.S. has grown roughly 174%, from 1.8 million to 4.9 million [5]. Such rapid growth is consequential for the Indian-American subset’s general effect on the Social Security system as it entails a disproportionately large working population, meaning the demographic has an anomalous worker-to-beneficiary ratio. California hosts the largest Indian population(960,000), second is Texas(570,000), and the third is New Jersey(440,000)[21]. Given that Indians are the second-largest immigrant group in America [20], general statistics regarding their population are imperative to understanding the status of the Social Security system as a whole. The five states with the highest Indian population densities are shown in Figure 2.   

Figure 2: Indian Americans as a Percentage of the State Population (Top Five Most Indian-Populated States)

Source: New India Abroad, Pew Research Center, Author Made

66% of the Indian population in the U.S. consists of first-generation immigrants that primarily moved to the U.S. as a working young adult [22]. Indians also tend to naturalize relatively quickly: the median years that immigrants hold a green card before naturalizing is 7.5 years, while it is 6 years for the Indian demographic specifically [20]. Precisely, 51% of Indian immigrants are naturalized American citizens [21]. It is important to clarify that one does not need to be a citizen to collect Social Security benefits, but citizenship still serves as an adequate predictor of how long one will reside in the U.S.; given that immigrants generally wait 5 to 8 years to naturalize, it is likely they will live in the U.S. for a substantial amount of time. That said, 40% of Indian-Americans are now considering moving out of the United States, primarily due to three factors: frustration with the political climate under Donald Trump (58% of the 40%), rising cost of living(54% of the 40%), and personal safety concerns(41% of the 40%) [23]. Still, literature has indicated that Indians who live in the U.S. for a substantial amount of time tend to rarely leave the country. 

The age distribution of Indian Americans tends to be skewed towards the working age adults, indicating more years of payroll tax contribution relative to their population size. Indeed, the median age of Indian immigrants is about 41 which is younger than most immigrant groups [24], while the median age of the overall Indian American population is 34.2. The median age for U.S. born Indian Americans is 13.4, and six-in-ten U.S.-born Indian Americans are under 18 with just 1% being 65+ [5]; this statistic further underscores the young, soon-to-be worker majority of this Indian-American population, in which the demographic has a large worker-to-beneficiary ratio that will persist for some time. 

Indian Americans have a bachelor’s degree attainment rate of 78%, the highest of any origin group in the country [25]. The highly-educated nature of the demographic is further accentuated with the statistics regarding the distribution of occupations and income within the demographic. 75% of employed Indian immigrants worked in management, business, science, and the arts compared to the 36% of all U.S. immigrants overall [26]. 21% of all active U.S. physicians are Asian, and with India being the single largest source country for international medical graduates, it’s reasonable to assume there would be a relatively high concentration of Indian doctors in America [27]. Indian American households have the highest median income of any ethnic or racial group in the U.S., registering a median income of $151,200 in 2023, an income well above the $105,600 median income for the overall Asian American population [5].

3.1 Life Expectancy Data for Indian Americans vs. Different Demographic Groups

Life Expectancy serves as the central variable in connecting demographic composition to Social Security’s financial outlook as it determines how much time a beneficiary draws out of the program. As of 2023, national life expectancy stood around 78.4 years overall, having risen 0.9 years from 77.5 in 2022 as the country was in the midst of recovering from the Covid-19 pandemic [28]. The recovery from the pandemic was not uniform across demographics, in which Whites had an expectancy of 78.4 years during 2023 and Hispanics had an expectancy of 81.3 years; African-Americans had the lowest expectancy of 74.0 years despite experiencing a substantial increase from previous years driven by decreased mortality from Covid-19, heart disease, homicide, diabetes, and cancer [29].  Asians Americans consistently rank the highest in life-expectancy amongst other demographics within the country; in 2023, Asians had a staggering expectancy of 85.2 years [29]. These life expectancies for the four demographics in 2023 are displayed in Figure 3. 

Figure 3: Life Expectancy at Birth by Demographic Group, 2023

Source: CDC NCHS, KFF, Author Made 

As mentioned earlier, however, analyzing the statistics of Asian Americans as a whole offers limited utility. The “Asian American” category treats Chinese Americans, Vietnamese Americans, and Indian Americans all as statistically inseparable even though these subsets of the Asian American population have vastly different statistics in their socioeconomics and health outcomes. Using National Vital Statistics System mortality data and American Community Survey population estimates from 2012–2016, Baluran and Patterson found that life expectancy at birth varied for as much as a decade between these different Asian American demographics, with the Vietnamese American population having a life expectancy of 77.5-79.7 years while the Chinese American population have a life-expectancy of 86.8-91.3 years [30]. This range in life-expectancy even exceeds the difference between White and Black populations in America. Baluran and Patterson ranked the life expectancies of the six largest Asian American subgroups from greatest to least: Chinese, Indians, Koreans, Japanese, Filipinos, and the Vietnamese [30].  Their paper did not include the specific life expectancy range for Indians. 

This places Indian Americans well above the average expectancy for Asian American demographics, but also substantially higher than White, Black, and Hispanic populations. It’s worth mentioning, however, that Indian Americans fall behind Chinese Americans, a rather unexpected finding given the two demographics’ socioeconomic characteristics. As previously stated, Indian American households have the highest median income of any ethnic or racial group in the U.S.[5], and the highest rate of bachelor’s degree attainment of all six Asian subgroups [31], raising questions as to why they do not rank the highest in life-expectancy. Literature has confirmed, however, that the correlation between socioeconomic status and life-expectancy, while strong, is not perfect [31]. Ranking highly within the socioeconomic hierarchy does not entirely translate to an increased life-expectancy [31]. 

Along with this, Asian Indian Americans do not necessarily have an upwards trend in life expectancy. In their study, Zheng used the years of life lost metric (YLL) to measure the gap between how long someone actually lives and how long they could’ve lived (this age is capped at 84) [31]. The YLL increased by 1.52 years for Asian Indian men with a bachelor’s degree and by 1.1 years for Asian Indian women with a bachelor’s degree. Among men without a bachelor’s degree, Vietnamese, Korean, and Asian Indian Americans experienced the largest YLL increases, ranging from 2.37 to 3.74 years, from 2000–2004 to 2020–2022 [31]. For women without a bachelor’s degree, YLL increased by 0.22 years among Asian women overall, and the Asian increase was mainly driven by Vietnamese, Korean, and Asian Indian women. The Covid-19 pandemic was not the primary culprit of this pattern, as the unfavorable YLL pattern was already evident before the pandemic, between 2000–2004 and 2015–2019.[31] Zheng traces 75% of the shortfall amongst Asian ethnicities to three specific causes of death: circulatory diseases (heart disease/stroke), cancer, and diabetes [31]. The changes in years of life lost for each race from the study is shown in Figure 4. 

Figure 4: Change in Years of Life Lost by Race/Ethnicity, Sex, and Education, 2000–2022, Decomposed by Pre-Pandemic and Pandemic Periods. 

Source: Zheng et al. (2026)

Heart disease, in fact, has proven to be somewhat of a commonality amongst Indians. Indian Americans tend to have a substantially higher frequency of attracting heart disease relative to other Asian subgroups, particularly ischemic heart disease. Literature has indicated that these levels of heart disease are associated with hypertension, diabetes, smoking, and central obesity [32]. Utilizing NCHS mortality data from 2003-2012, one study found that Indian American men lost an average of 17 years of life to ischemic heart disease, and 724 years of potential life per 100,000 population in 2012 alone, a notably higher figure than other Asian subgroups [33]. This increased risk of cardiovascular disease amongst the Indian demographic could likely be the primary reason as to why their life expectancy underachieves given their socio-economic well-being. 

The “healthy immigrant” phenomenon likely plays a factor into why Indian Americans have a substantially higher life expectancy compared to Indians that reside in India [34]. Immigration is an arduous process, requiring a certain threshold of physical health and material wealth. To immigrate, one needs to pass several medical screenings as they may face rejection if they were tested to be ill, frail, or have a chronic untreated condition [35]. Furthermore, one would generally need to have some level of affluence before immigrating, and are usually required to have a guaranteed employment route once they reach America. The H-1B Visa (71% of all H-1B visas go to Indian nationals [20]), for instance, requires that Indian immigrants have a bachelor’s degree and a job offer from a U.S. employer [36]. Given that education and income has a direct, positive correlation with health outcomes[37], it’s fair to assume that many of the immigrants that fulfill the H-1B Visa requirements are healthy and would generally have an extended life-expectancy. This healthy immigrant phenomenon effectively filters the Indian population, choosing the healthiest and most affluent to come to America[34]. 

4.0 Impact on the Social Security System of the IA data

To make a direct, comparable measure of each demographic group’s contribution to the OASI trust fund, the OASDI tax formula(6.2%) was applied to each demographic’s year-round median individual full-time earnings. Individual earnings were used instead of household income as the tax rate is levied on individuals, not households. The median individual earnings for Indian Americans, the broader Asian American population (including Indian Americans), African Americans , Hispanic Americans, and White Americans were calculated from the U.S. Census Bureau’s 2024 American Community Survey 1-year Public Use Microdata Sample (PUMS) [38]. Census-assigned person weights were applied to each respondent, as each respondent’s earnings represented an estimated, varying number of U.S. residents’ earnings. The median individual earnings for all demographic groups are from 2024 and expressed in 2024 dollars. It is worth noting that the figures drawn from this dataset throughout the study are independently calculated estimates derived from sample data rather than official published statistics. The median full-time individual earnings and annual social security contributions by demographic group is shown in Table 2. 

Table 2: Median Full-Time Individual Earnings and Annual Social Security Contributions by Demographic Group, 2024. 

Source: U.S. Census Bureau’s 2024 American Community Survey 1-year Public Use Microdata Sample (PUMS), Author Made 

At the median, Indian Americans contribute more than double the amount contributed by Hispanic/Latino and African American workers. The income disparities between demographics are especially visible when looking at each group’s annual payroll tax dollars. Indian Americans also contribute substantially more than the broader Asian American aggregate ($6924.04 vs. $5035.64), further underscoring how extracting statistics for the entire Asian population substantially understates the economic position of the Indian subgroup specifically. Every group’s median individual earnings sit well below the $184,500 cap, meaning their total earnings proportionally translate to how much they contribute to the program. 

For each demographic, calculating the proportion of full-time, year-round workers whose annual earnings exceed the $184,500 taxable maximum is necessary to understand what share of each subgroup contributes to the program proportional to their annual earnings, versus what share is shielded above the cap. Using the same 2024 ACS PUMS dataset as the previous section, this analysis calculates that percentage for each demographic group. The share of full-time workers earning above the social security taxable maximum by demographic group is shown in Table 3.

Table 3: Share of Full-Time Workers Earning Above the Social Security Taxable Maximum by Demographic Group, 2024. 

Source: U.S. Census Bureau’s 2024 American Community Survey 1-year Public Use Microdata Sample (PUMS), Author Made

While it was established that no demographic’s typical worker has their earnings exceed the $184,500 maximum (as indicated by the analysis on the median full-time individual earnings for each demographic), the proportion of Indian Americans that do earn past this maximum is substantially higher than the other subgroups; at 13.84%, Indian Americans are capped at more than three times the rate of White workers (4.03%), roughly nine times the rate of Hispanic workers (1.53%), and over ten times the rate of Black workers (1.34%). Notably, Indian Americans also exceed the cap at nearly double the rate of the broader Asian American aggregate (7.57%); this further underscores how pan-ethnic Asian statistics substantially understate the affluent economic status of the Indian subgroup. 

This finding has a direct and counterintuitive implication for the wage-cap framework established earlier in this paper: for this specific 13.84% of the Indian American population, a portion of their true earnings is not captured by the Social Security tax, meaning their contribution to the program understates their true income advantage relative to other groups. Analyzing a median-based comparison therefore does not account for this nuance as the Indian demographic’s aggregate contributions to the trust fund are not directly proportional to their full economic standing. Indeed, a substantial share of the demographic’s total earning power goes untaxed as it sits above the cap. Nevertheless, Indian Americans remain among the top contributing demographics examined in this paper. Because a larger share of the population reaches the taxable maximum than any other group, more Indian American workers pay the full $11,439 ceiling contribution than in any other demographic. 

4.1 Duration of Contribution

It is essential to establish a general idea of what proportion of each demographic is old enough to be collecting Social Security benefits versus still within the working age.  To do so, the percentage of each demographic that is 67 years or older was calculated; 67 was chosen as the baseline age for calculating the metrics as it is the full retirement age when one can receive 100% of their benefits[39]; it is important to clarify, however, that different demographics may begin collecting benefits at different ages due to a variety of the mentioned factors such as income. The median age was also calculated for each demographic to understand how close each group is, on average, to the retirement age. Both metrics were calculated using the same 2024 ACS PUMS dataset used for the previous contribution metrics to ensure relative consistency. 

Table 4: Median Age and Share of Population at Full Retirement Age (67+) by Demographic Group, 2024. 

Source: U.S. Census Bureau’s 2024 American Community Survey 1-year Public Use Microdata Sample (PUMS), Author Made

As shown in Table 4, Indian Americans currently skew heavily toward the working age rather than retirement age relative to most other groups examined. At a median age of 36.1, Indian Americans are more than seven years younger than White Americans (43.2), and only 7.4% of the Indian American population has reached the full retirement age, the second-lowest share of any group in Table 4. This distribution suggests that the Indian American demographic currently has a very small share of its population collecting benefits, with the overwhelming majority still contributing. These findings are anticipated as large-scale immigration from India is relatively recent [20]. Along with this, the typical age at which one immigrates to the U.S. is around 31 [40] (this age is not Indian-immigrant specific). Since newly arrived immigrants enter the country as a young adult and large-scale Indian immigration has only recently begun, it is expected that most Indian Americans have not lived to reach the full retirement age of 67. 

A separate but relevant question from each population’s current age composition shown in Table 4 is how long a newly arrived Indian immigrant specifically needs to remain in the U.S. before becoming eligible for benefits in the first place. This requires connecting the 40-credit/10-year work requirement to the age at which Indian immigrants typically arrive. Approximately 60% of Indian immigrants have resided in the U.S. for 10 or more years [5]; while it is important to clarify that living in America for 10 or more years does not inherently mean these immigrants were employed for 10 or more years, this suggests that a large portion of the demographic has had sufficient time in the country to meet the eligibility threshold. 

Van Hook and Zhang’s analysis of immigrant emigration patterns further provides context as to how many immigrants work in the U.S. long enough to pass the eligibility threshold. They reveal that emigration is most common among immigrants in their first 0-4 years of U.S. residence. Once immigrants surpass the 10-14 year mark, they are 30-33% less likely to emigrate, which is the same window in which 40-credit eligibility is typically achieved [41]. Since roughly 60% of Indian immigrants have already resided in the U.S. for 10 or more years, a substantial share of Indian immigrants who arrive as working-age adults likely accumulate the 40 credits to qualify for their Social Security benefits. Those that don’t achieve the 40 credit threshold for eligibility are classified as the “Contribute and Leave” population, the subgroup of working Indian immigrants that annually contribute for as long as they reside in the U.S. but cannot claim payments from the program. The exact total dollar value of contributions made by this group is currently unavailable, and the statistics needed to manually calculate this figure (such as the number of people within this group) are also not accessible in published sources. 

It’s important to also account for the population of Indian Americans that reside in a different country after retirement. For Social Security retirement benefits specifically, if these Indian Americans are U.S. citizens they can receive payments while living anywhere in the world, with only a small list of restricted countries as exceptions. For non-citizens, however, they cannot receive payments for any month in which they have been outside the U.S. for more than six consecutive months, unless they meet specific exceptions (as mentioned previously) [42]. 

Medicare, by contrast, does not transfer the same way. Medicare generally does not cover any medical care received outside the United States, regardless of citizenship status or years of U.S. residence. If a retiree who was eligible to claim Medicare benefits returns to India, they could not use it for care received at a hospital in India. This means a returning Indian American retiree’s earned Social Security cash benefit remains payable in India, but their earned Medicare health coverage effectively does not. This implies that the total value of benefits actually accessible to a returning retiree is less than the full value of what they contributed across both programs, further complicating the contribution-to-benefit ratio for the Indian American demographic. 

4.2 Life Expectancy Translated Into Years of Collection

Life expectancy helps determine how many years a beneficiary is expected to draw from the OASI trust fund after becoming eligible. 

Life expectancy for each demographic in 2024 was calculated combining two data sources: one providing mortality data, and the other providing population counts. Mortality data were obtained from the Centers for Disease Control and Prevention’s CDC WONDER Multiple Cause of Death, 2018–2024, Single Race database, which provided 2024 all-cause death counts by age and race. Corresponding 2024 population estimates by age and race were obtained from the same 2024 ACS PUMS dataset used for the prior calculations. 

Age was grouped into year intervals: <1, 1–4, 5–9, … 95+. The death rate for each of these age groups was calculated by combining the mortality data with the age-specific population estimates. This rate was then converted into a death probability for a hypothetical set of 100,000 people per demographic. For each age group, the number of individuals expected to survive was carried forward to the next age interval. Total years lived was then calculated for each age group by multiplying the number of hypothetical individuals in that age group by the interval length, while subtracting the years not lived by individuals who died before the interval ended. Thus, individuals who survived the entire interval contributed the full number of years in that interval, whereas individuals who died contributed only the estimated portion of the interval that they lived. The total years lived metric for each age group was summed and divided by the original hypothetical population of 100,000 to obtain the life expectancy at birth for the given demographic.

Years of expected benefit collection were calculated by subtracting 67 (the age that was previously used for the age distribution metrics) from life expectancy, allowing for a consistency across the different calculated metrics. The calculated life expectancy at birth and years of expected benefit collection by demographic group is shown in Table 5. 

Table 5: Calculated Life Expectancy at Birth and Years of Expected Benefit Collection by Demographic Group, 2024.

Source: CDC WONDER’s Multiple Cause of Death, 2018–2024, Single Race database and U.S. Census Bureau’s 2024 American Community Survey 1-year Public Use Microdata Sample (PUMS). Author Made

Some variation from each demographic’s true life expectancy is expected as these estimates reflect mortality conditions in 2024 alone rather than a multi-year average, and they are calculated estimates derived from sample data. The calculations also utilize two separate data systems, further increasing the figures’ potential error. These inconsistencies may partially explain the narrow gap between Indian American and Asian American life expectancy, a gap that literature has indicated to be wider. 

Despite these inherent limitations in the calculations, the figures still underscore the general distribution of life expectancies between demographics. Indian Americans, with the highest calculated life expectancy at 86.2 years, are expected to draw from the OASI trust fund for 19.2 years after reaching full retirement age, nearly three times as long as African Americans (7.0 years), who have the lowest life expectancy of any group examined at 74.0 years. The gap between these two groups amounts to 12.2 additional years of benefit collection, a difference that compounds substantially when multiplied across a beneficiary’s monthly payments.

These figures highlight one of the primary concerns regarding the program’s distribution of benefits. A benefit formula that is uniform for each demographic produces unequal lifetime returns between demographics once differential mortality is accounted for. Two workers of different ethnicities with identical earnings histories and contributions would still likely receive a different amount of benefits primarily because they live differing amounts of time, which varies systematically across demographic groups (as indicated in this study). 

Notably, the ordering of groups by years of benefit collection does not align with the ordering of groups by contribution established earlier in this section. Hispanic Americans, who had the lowest median individual earnings ($48,732) and therefore the lowest annual contribution of any group examined, are nonetheless expected to collect benefits for 15.3 years; this is 3.6 years longer than White Americans, who contribute substantially more annually. African Americans contribute more annually than Hispanic Americans yet are expected to collect for less than half as long. This misalignment between contribution capacity and collection duration is precisely what makes the contribution-to-benefit ratio a more informative metric than either variable examined in isolation. 

Indian Americans, conversely, contribute the most annually of any demographic examined, while also being positioned to collect benefits for the longest period. Whether this results in a net-positive or net-negative position relative to the trust fund depends on the demographic’s contribution-to-benefit ratio. 

4.3 Contribution-to-Benefit Ratio for Indian Americans

Utilizing the 2024 ACS PUMS dataset, the annual earnings were first capped at the taxable maximum of $184,500; the cap needs to be implemented when calculating any contribution or benefit amounts as all residents earning above the cap are effectively contributing and benefiting the same amount to the program. After this cap was applied, the average annual earnings was calculated for all Indian U.S. residents which was $70,658.65. These earnings produce a contribution-to-benefit ratio of either 0.25 or 0.51, depending on whether the employer’s matching contribution is counted. On the employee side alone, a worker earning this amount pays $4,380.84 annually in payroll taxes, amounting to $153,329.27 over an assumed 35-year career. Counting the employer’s matching 6.2%, total contributions to the trust fund on that worker’s behalf reach $8,761.67 annually and $306,658.54 across the same period. On the benefit side, this earnings level yields an AIME of $5,888.22, which the bend-point formula converts into a monthly benefit of $2,630.11 at full retirement age. Multiplied across the 19.2 years of expected collection established earlier, this produces a total lifetime benefit of $605,977.50. The Contribution-to-Benefit Ratio for Indian Americans, along with the metrics used to calculate the ratio are shown in Table 6. 

Table 6: Contribution-to-Benefit Ratio for Indian Americans, 2024

Source: Author’s calculations using U.S. Census Bureau 2024 ACS 1-Year PUMS and Social Security Administration benefit parameters

The employee-share ratio of 0.25 reflects the individual worker’s perspective: an Indian American retiree is projected to receive roughly four dollars in lifetime benefits for every dollar withheld from their own paycheck. The combined ratio of 0.51 reflects the trust fund’s perspective, accounting for the full 12.4% the program actually receives, and indicates that Indian Americans are projected to collect approximately twice what is paid in on their behalf. The combined ratio is the more appropriate measure for analyzing the demographic’s effects on the program’s solvency since it includes the total share of what is being contributed towards the program, combining both the employee and employer share. For analysis of the effects on the individual, the employee-share ratio is the measure of importance as it isolates the employee’s burden and benefits from the total share. 

Under both measures, the final ratio falls well below 1.0, indicating that Indian Americans are projected to receive more in lifetime benefits than is contributed on their behalf. Given the high income of Indian Americans, they tend to be the highest contributors to the program. Nevertheless, they also tend to collect benefits for the longest, primarily due to their simultaneously high life expectancy. Because of this, the demographic’s contribution-to-benefit ratio is still unfavorable to the trust fund even while they provide substantial contributions. It is also important to note that the AIME benefit formula is inherently structured to favor lower earners; because the replacement rate declines from 90% to 32% to 15% as AIME rises, each additional dollar of earnings produces proportionally less additional benefit, while the same 12.4% tax applies to every dollar below the cap.

5.0 Policy Implications

After analyzing the income and age distribution of each subgroup, along with a calculated contribution-to-benefit ratio for Indian Americans specifically, two policy questions are worth analysis. Is it defensible for the program to be distributing lifetime benefits unevenly on the basis of life expectancy, an uncontrollable variable that varies substantially between each person and subgroup? Secondly, which populations strengthen the trust fund’s position, and which strain it?

No single reform clearly addresses both questions. Indeed, attempting to solve one persisting limitation of the program may worsen another. Indian Americans contribute more per worker than any group examined. Their high life expectancy, however, means they are projected to collect substantially more than what is paid on their behalf, marking a strain on the trust fund. African Americans occupy the reverse position: their short expected collection period eases pressure on the fund, but it also means they proportionally recover less of what they contributed than any other group, through no choice of their own. When examining reform options on both sides of the program, it is important to note that gains on one aspect of the program comes at the expense of the other. 

5.1 Reform Options Bearing on the Contribution Side

A frequently proposed reform lies in the $184,500 taxable maximum. Literature indicates that increasing or completely removing this cap would improve the solvency of the program, but may lead to adverse effects beyond the program. In their study, Bagchi used a life-cycle general-equilibrium framework to evaluate the effect of eliminating this cap, finding that eliminating the cap would limit potential future benefit reductions to under 4%, compared to nearly 12% if the cap remains in place. This improvement would come at the cost of high earners potentially saving and working less, which would reduce labor, capital, output, and overall welfare[43]. 

Removing the cap would fall unevenly across demographic groups. Referring back to Table 3, the Indian American demographic has the highest percentage of high earners compared to the other groups, with 13.84% of Indian American full-time workers earning above the taxable maximum; removing the cap would therefore disproportionately draw from Indian American earnings as the 13.84% of Indian American workers that exceed the threshold would then have to contribute directly proportional to their earnings. Conversely, this cap removal would have little effect on the African American demographic where only 1.34% of their population is currently above the threshold. Indeed, removing the taxable maximum would affect each demographic to substantially different degrees. 

Completely removing the cap may weaken incentives for human-capital investment, where workers feel less inclined to be more productive and skilled [44]. The more modest and less radical approach would be to partially increase the taxable maximum, meaning those earning above the current threshold would contribute more, though still not on all of their earnings. This increase would still disproportionately affect each demographic but not to the same extent as full removal would. 

Adjusting the tax rate is also a potential solution to remedying the program. In their study, Kitao modeled the effects on the program if the contribution rate were to increase to 16.4% split evenly between the employer and employee (compared with the original 12.4%)[45]. Raising the rate to 16.4% would directly increase payroll-tax revenue, improving program solvency. With this rate increase, however, consumption and capital are predicted to fall 5.5% and 5.3% respectively. Along with this, the average hours worked is expected to rise by 2.1% and the average work life would increase from 44.0 to 44.5 years [45]. This approach would likely appeal to middle-older aged workers who are nearing their retirement; this age group will only be paying the increased tax rate for a little period of time while still receiving all of their promised benefits, including the benefits during the time in which they contributed through the original 12.4% tax rate. Younger workers would likely oppose this reform because they will have to contribute relatively more over a longer period of time while receiving the same benefits that they would have originally received through the 12.4% tax rate. It is important to clarify that each demographic will still proportionally contribute different amounts of money due to the different percentages of people in each demographic that pass the taxable maximum threshold. 

5.2 Reform Options Bearing on the Benefit Side

Raising the full retirement age is also a commonly proposed benefit-side reform, delaying the time-frame in which one could begin collecting their benefits. Much like the contribution-side reforms, however, these reforms would fall unevenly across demographic groups. 

Increasing the age at which one can collect benefits will uniformly decrease the total years retirees collect their benefits for, meaning less money is drawn out of the program. While this would alleviate some of the current fiscal stress of the program, it would also create a burden for retirees. The Social Security Administration estimates that raising the full retirement age to 68 would close 23% of the program’s actuarial imbalance, while raising it to 70 would close 31%. Still, these changes would reduce average benefits by roughly 6.8% and 18.6%, respectively [46]. These benefit reductions have differing degrees of severity for each demographic group. For Indian Americans, raising the full retirement age to 68-70 would still translate to between 16.2 and 18.2 years of benefit collection, a substantial collection period. African Americans would face a much worse change if the retirement age increased to 68-70; their expected years of benefit collection would drop to between 4 and 6, marking a 14% to 43% reduction from the already low original years of benefit collection they had.  

Another reform that could be made on the benefit side is to modify the bend-point benefit formula itself. Modifications could be applied in two ways: lowering the existing replacement rates, or subdividing the current brackets into additional tiers. Under the latter approach, for instance, the 15% rate currently applied to all earnings above $7,749 might be replaced by an 8% rate on earnings between $7,749 and $8,500 and a 7% rate on earnings above $8,500. Monthly benefits would be reduced through both modifications, though it is important to also account for how it burdens lower and higher earners. The previously stated example (the subdivided 15% into 7% and 8% tiers) would be classified as a progressive reform, as it would reduce benefits primarily only for higher earners, leaving those below the $7,749 threshold unaffected. Conversely, lowering the initial 90% replacement rate would substantially hurt lower earners, as much of their AIME falls within the initial $1,286. 

Adjusting this formula provides a potentially viable and sustainable approach to managing how much money is drawn out of the program, but it is important to recognize how each modification to the formula would influence the existing wage gap and affect groups across income levels and demographics. A reform like this would also be correcting for earnings differences, not life-expectancy differences, though it may somewhat offset those life-expectancy differences (this would not fully offset those disparities, as the paradoxical cases discussed earlier demonstrate that life expectancy does not reliably track earnings. Hispanic Americans, for instance, have the lowest median earnings of any group examined yet the second-longest expected collection period). 

5.3 What These Findings Do Not Support

This study does not support a demographically-differentiated benefit formula. Adjusting one’s contribution rate, bend-point formula, assigned retirement age, or taxable maximum based on their ethnic demographic is legally impermissible. Beyond this approach’s unethical nature, it would also not effectively remedy the program’s growing insolvency. Studies analyzing each demographic’s contributions or benefits from the program rest on group averages and general life expectancies, studying patterns from aggregate outcomes; these generalized analyses say nothing about how long any particular individual will live. Any reform responding to these disparities must therefore operate through variables the program already recognizes (earnings, claiming age, contribution history) rather than through demographic classification itself.

6.0 Concluding remarks and Further Research

This paper examined the interaction between life expectancy, demographic characteristics, and contribution-to-benefit patterns in order to assess the current condition of the Social Security program, its projected trajectory, and the reforms available to address its long-term solvency. This analysis focused particularly on the Indian-American demographic given their anomalous characteristics and their frequent aggregation into a single ‘Asian American’ category that obscures variation between Asian subgroups. 

The findings indicate that Indian Americans occupy a distinctive position on both the contribution and benefit side of the program. They contribute more per worker than any group examined and have the highest share (13.84%) of workers above the taxable maximum out of the examined groups. They also have the highest calculated life expectancy at 86.2 years, yielding 19.2 expected years of benefit collection. Their combined contribution-to-benefit ratio of 0.51 indicates they collect roughly twice what is paid in on their behalf. Contribution capacity and collection duration do not align across groups, and neither variable alone describes a group’s relationship to the program. Because benefits are paid for as long as a beneficiary survives, life expectancy is a key determinant for how much a worker receives their benefits. 

These conclusions rest on stated constraints: independently calculated life expectancy estimates, an assumed flat 35-year earnings history, and ratios describing group averages that cannot be attributed to individuals. These limitations point towards future work that could be done. Probabilistic modeling through a Monte Carlo approach would replace this paper’s fixed assumptions with distributions of earnings, career length, and mortality, producing ranges rather than point estimates. In the future, studies specifically analyzing second-generation Indian Americans would also provide insight into how this group will affect long-term solvency. Beyond the Indian American demographic, this disaggregation analysis could be applied to other Asian subgroups that are aggregated into the single ‘Asian American’ category. 

Disaggregating the Indian American population from a broad statistical category revealed a group that is simultaneously the program’s largest per-worker contributor and among its largest net beneficiaries.The comparison across groups further showed that contribution capacity and collection duration move independently of one another, meaning that a program applying identical rules to every participant returns those contributions unequally, along a dimension the benefit formula does not measure. With the trust fund projected to reach insolvency within the decade, the window for reform is narrowing faster than the groundwork required to guide it, making further research on the program’s sustainability an increasingly urgent prerequisite rather than a refinement to be pursued afterward. 

References

1. Picchi, Aimee. “Social Security Insolvency Now Projected for 2032, Putting Benefits at Risk of a 22% Cut.” Cbsnews.Com, 9 June 2026, https://www.cbsnews.com/news/social-security-trust-fund-insolvency-2032-trustees-report/. 
2. Pew Research Center. “Indian Americans: A Survey Data Snapshot.” Pew Research Center, 6 Aug. 2024, https://www.pewresearch.org/2024/08/06/indian-americans-a-survey-data-snapshot/. 
3. Mehta, Neil K., et al. “Life Expectancy Among U.S.-Born and Foreign-Born Older Adults in the United States: Estimates From Linked Social Security and Medicare Data.” Demography, vol. 53, no. 4, July 2016, pp. 1109–34, https://doi.org/10.1007/s13524-016-0488-4.
4. Johnson, Richard W, and Karen E Smith. “Limitations on Social Security Benefits for Black Retirees: Exploring Structural Racism’s Impact.” Urban Institute, 20 Apr. 2023, https://www.urban.org/research/publication/limitations-social-security-benefits-black-retirees.
5. Budiman, Abby. “Indians in the U.S. Fact Sheet.” Pew Research Center, 29 Apr. 2021, https://www.pewresearch.org/race-and-ethnicity/fact-sheet/asian-americans-indians-in-the-u-s/. 
6. Peterson Foundation. “Budget Basics: How Does Social Security Work?” Peterson Foundation, 27 Nov. 2024, https://www.pgpf.org/article/how-does-social-security-work/. 
7. FDR Library. “Great Depression Facts – FDR Presidential Library & Museum.” Fdrlibrary.Org, Franklin D. Roosevelt Presidential Library and Museum, 2016, https://www.fdrlibrary.org/great-depression-facts. 
8. Pan, Guanchen. “Analysis of Roosevelt’s New Deal: The Decisive Role the Government Should Play When Facing the Crisis.” Lecture Notes in Education Psychology and Public Media, vol. 4, no. 1, EWA Publishing, May 2023, pp. 405–09, https://doi.org/10.54254/2753-7048/4/2022101. 
9. “Social Security.” LII / Legal Information Institute, https://www.law.cornell.edu/wex/social_security. 
10. “Public Benefits for People with Disabilities | Financial Resilience Center.” National Disability Institute, https://www.nationaldisabilityinstitute.org/financial-resilience-center/benefits/. 
11. Social Security Administration. “FICA & SECA Tax Rates.” Www.Ssa.Gov, https://www.ssa.gov/oact/progdata/taxRates.html. 
12. “​FICA Tax: What Does It Mean and How Is It Calculated?” Paychex, https://www.paychex.com/articles/payroll-taxes/how-to-calculate-fica-rate. 
13. Social Security Administration. “Contribution and Benefit Base.” Ssa.Gov, 2025, https://www.ssa.gov/OACT/COLA/cbb.html. 
14. Social Security Administration. “Social Security Benefit Amounts.” Www.Ssa.Gov, https://www.ssa.gov/oact/cola/Benefits.html. 
15. “Primary Insurance Amount.” Www.Ssa.Gov, https://www.ssa.gov/oact/COLA/piaformula.html. 
16. Peterson Foundation. “The Ratio of Workers to Social Security Beneficiaries Is at a Low and Projected to Decline Further.” Peterson Foundation, 29 Oct. 2024, https://www.pgpf.org/article/the-ratio-of-workers-to-social-security-beneficiaries-is-at-a-low-and-projected-to-decline-further/. 
17. Dwyer-Lindgren, Laura, et al. “Life Expectancy by County, Race, and Ethnicity in the USA, 2000–19: A Systematic Analysis of Health Disparities.” The Lancet, vol. 400, no. 10345, July 2022, pp. 25–38, https://doi.org/10.1016/S0140-6736(22)00876-5..
18. Hendley, Alexa A, and Natasha F Bilimoria. “Minorities and Social Security: An Analysis of Racial and Ethnic Differences in the Current Program.” Social Security Bulletin , vol. 62, no. 2, 1999, https://www.ssa.gov/policy/docs/ssb/v62n2/v62n2p59.pdf. 
19. Social Security Administration. “Benefits Planner: Retirement | Retirement Age and Benefit Reduction | SSA.” Www.Ssa.Gov, 2023, https://www.ssa.gov/benefits/retirement/planner/agereduction.html. 
20. Batalova, Jeanne, and Ari Hoffman. “Indian Immigrants in the United States.” Migrationpolicy.Org, Migration Policy Institute, 7 Dec. 2022, https://www.migrationpolicy.org/article/indian-immigrants-united-states. 
21. Kumar, Arun. “Indian Americans Second-Largest Asian Group in the US: Pew.” The American Bazaar, 6 May 2025, https://americanbazaaronline.com/2025/05/06/indian-americans-second-largest-asian-group-in-the-us-pew-462477/. 
22. “Indian Americans Now 2Nd Largest Asian Group in US Pew.” New India Abroad, 2024, https://www.newindiaabroad.com/english/features/indian-americans-now-2nd-largest-asian-group-in-us-pew. 
23. Singh, Ritu. “American Dream Fading? 40% Of Indian Americans Consider Leaving The US, Survey Reveals Top Reasons.” Www.Ndtv.Com, NDTV, 23 Apr. 2026, https://www.ndtv.com/feature/american-dream-fading-40-of-indian-americans-consider-leaving-the-us-survey-reveals-top-reasons-11396699. 
24. “VisaVerge.” VisaVerge, 3 May 2025, https://www.visaverge.com/india/indian-american-population-in-us-reaches-new-record-census-shows/. 
25. theworldstatisticswebsite@gmail.com. “Indian Population in US 2025 | Statistics & Facts – The World Data.” –, 14 Oct. 2025, https://theworlddata.com/indian-population-in-us/. 
26. “Asian Indian Immigrants | EBSCO.” EBSCO Information Services, Inc. | Www.Ebsco.Com, 2020, https://www.ebsco.com/research-starters/social-sciences-and-humanities/asian-indian-immigrants. 
27. MyAsianVoice. “1 In 5 Medical Doctors in the U.S. Are Asian – MyAsianVoice.” MyAsianVoice, 6 Dec. 2024, https://www.myasianvoice.com/1-in-5-medical-doctors-in-the-us-are-asian. 
28. Arias, Elizabeth, et al. National Vital Statistics Reports, vol. 74, 2025, https://www.cdc.gov/nchs/data/nvsr/nvsr74/nvsr74-06.pdf. 
29. KFF. “Life Expectancy at Birth Increased Across All Race and Ethnicity Groups In Recent Years, Erasing the Declines Experienced During the COVID-19 Pandemic.” KFF, 6 Mar. 2026, https://www.kff.org/racial-equity-and-health-policy/racial-disparities-in-life-expectancy/. 
30. Baluran, Darwin A., and Evelyn J. Patterson. “Examining Ethnic Variation in Life Expectancy Among Asians in the United States, 2012–2016.” Demography, vol. 58, no. 5, Sept. 2021, pp. 1631–54, https://doi.org/10.1215/00703370-9429449.
31. Zheng, Hui, et al. “Heterogeneity in Asian Americans’ Mortality Trends, 2000–2022.” The Journals of Gerontology Series B, vol. 81, no. 2, Oxford University Press, Dec. 2025, https://doi.org/10.1093/geronb/gbaf248.
32. Kaur, Prabhdeep, et al. “Risk Factors for Cardiovascular Disease in Rural South India: Cohort Study.” BMJ Open, vol. 9, no. 10, Oct. 2019, p. e029759, https://doi.org/10.1136/bmjopen-2019-029759.
33. Iyer, Divya G, et al. “Years of Potential Life Lost Because of Cardiovascular Disease in Asian-American Subgroups, 2003-2012.” Journal of the American Heart Association, vol. 8, no. 7, 2 Apr. 2019, p. e010744, https://doi.org/10.1161/JAHA.118.010744. PubMed Central. 
34. Riosmena, Fernando, et al. “Explaining the Immigrant Health Advantage: Self-Selection and Protection in Health-Related Factors Among Five Major National-Origin Immigrant Groups in the United States.” Demography, vol. 54, no. 1, Duke University Press, Jan. 2017, pp. 175–200, https://doi.org/10.1007/s13524-016-0542-2. 
35. may. “Why You May Need to Provide Evidence You Are in Good Health.” Immigration New Zealand, 2022, https://www.immigration.govt.nz/process-to-apply/applying-for-a-visa/providing-evidence-and-documents-to-support-your-visa-application/health-requirements/why-you-may-need-to-provide-evidence-you-are-in-good-health/. 
36. U.S. Citizenship and Immigration Services. “H-1B Specialty Occupations, DOD Cooperative Research and Development Project Workers, and Fashion Models | USCIS.” Www.Uscis.Gov, 5 Feb. 2021, https://www.uscis.gov/working-in-the-united-states/h-1b-specialty-occupations.
37. “Understanding the Relationship Between Education and Health – NAM.” NAM, 21 Nov. 2024, https://nam.edu/perspectives/understanding-the-relationship-between-education-and-health/. 
38. United States Census Bureau. “Explore Census Data.” Census.Gov, https://data.census.gov/table/ACSDT1Y2024.B20017D?g=010XX00US. 
39. “A Guide on Taking Social Security.” Schwab Brokerage, https://www.schwab.com/learn/story/guide-on-taking-social-security. 
40. “Immigrants Are Coming to America at Older Ages.” CIS.Org, https://cis.org/Report/Immigrants-Are-Coming-America-Older-Ages.
41. Van Hook, Jennifer, and Weiwei Zhang. “Who Stays? Who Goes? Selective Emigration Among the Foreign-Born.” Population Research and Policy Review, vol. 30, no. 1, Apr. 2010, pp. 1–24, https://doi.org/10.1007/s11113-010-9183-0.
42. “Social Security Payments Outside the United States.” Social Security Administration, https://www.ssa.gov/international/payments.html?tl=0. 
43. Bagchi, Shantanu. “Can Removing the Tax Cap Save Social Security?” The B.E. Journal of Macroeconomics, vol. 17, no. 2, June 2017, https://doi.org/10.1515/bejm-2016-0091.
44. Blandin, Adam. “Human Capital and the Social Security TAX CAP.” International Economic Review, vol. 62, no. 4, Wiley, June 2021, pp. 1599–626, https://doi.org/10.1111/iere.12525.
45. Kitao, Sagiri. “Sustainable Social Security: Four Options.” Review of Economic Dynamics, vol. 17, no. 4, Elsevier BV, Oct. 2014, pp. 756–79, https://doi.org/10.1016/j.red.2013.11.004. 
46. “Distributional Effects of Accelerating and Extending the Increase in the Full Retirement Age.” Social Security Administration Research, Statistics, and Policy Analysis, https://www.ssa.gov/policy/docs/policybriefs/pb2011-01.html. 

About the author

Urjit Galera

Urjit is a senior at the Bronx High School of Science. His deep interest in economics and social welfare policy paired with his Indian-American background is what drove him to pursue this research. Through this project, Urjit aims to show how disaggregating broad demographic categories reveals patterns that aggregate data conceals, bringing greater precision to debates over Social Security’s future.

He hopes to continue his research endeavors in university, exploring the intersection of public health, economics, and data science.

The post An Analysis of the Social Security Experience of Indian Americans appeared first on Exploratio Journal.

Abstract

In 2025, an agreement between the House and the NCAA led to a major change in college sports by finally allowing universities to pay their athletes. This article examines the problem that this change has caused: can universities afford the rising cost of paying their college athletes? According to legal analyses and financial reports, the study points to four pressures caused by the new system. The first of these is that a spending limit linked to revenue increases the financial gap between the wealthy Power Four conferences and the smaller Group of Five, even though the College Football Playoff is expanding in order to improve access. Second, at individual universities, the majority of the funds available are spent on football and men’s basketball, placing a heavy strain on resources. Third, because enforcement is weak, schools are able to get around the spending cap by making arrangements with third parties. Fourth, the universities meet these costs by relying on limited sources such as student fees and tuition increases. The analysis concludes that for sustainable viability it will be necessary to create genuinely new revenue streams, for example through sport-related real estate development, rather than reallocating existing revenues to a permanent and growing expense.

Keywords: name, image, and likeness (NIL); House v. NCAA; the manner in which college athletics are financed; revenue sharing; Title IX; and competitive equilibrium.

Borrowed Time: Can Colleges Reimburse Their Athletes in the Light of House v. NCAA?

For more than a hundred years, college sports have been based on the concept of amateurism. The athletes took part in order to promote school pride and to get a scholarship, not because they were being paid. Even if a quarterback filled a stadium with 100,000 people in attendance, sold thousands of jerseys, and brought in millions of dollars for his university, he would still be ineligible if he earned a single dollar. The NCAA considered this rule to be what made college sports special since it clearly distinguished college sports from professional sports (Henderson, 2023). However, behind this principle a multi-billion-dollar industry developed at the expense of the athletes who were prohibited from benefiting from it.

That contradiction could not continue forever. The first indications of weakness came in 2014, when Ed O’Bannon took the NCAA to court over the use of his image without being paid and won (Henderson, 2023). As pressure increased, the Supreme Court gave a uniform decision in the case of NCAA v. Alston in 2021, proving that the NCAA’s rules could not be maintained under legal scrutiny (Henderson, 2023). A short time later the NCAA gave in and allowed athletes to get money from their name, image, and likeness (NIL). The final element to collapse was the 2024–25 House agreement, which enabled colleges to pay athletes directly, the bigger programs being allowed to spend up to $20.5 million annually (Ropes & Gray, 2025). In under four years college sports had gone from paying athletes nothing to establishing a $2.75 billion NIL market, with $1.95 billion going directly to the athletes (Opendorse, 2025).

That figure represents only the beginning of the earnings that athletes could potentially obtain. While college athletes get only about 13 per cent of the revenue which they help to generate, NFL players receive 48 per cent under their collective bargaining agreement and considerably less than the nearly equal 50/50 divisions found in most professional entertainment (De Lima, 2025). As college sports move towards a fully professional system, the pressure to pay athletes more will only grow, and the funds required to meet that demand are already becoming more difficult to obtain.

The tension will determine the character of the coming period in college sports. At present, universities are burdened by a massive new cost which did not exist a few years ago and all signs suggest that this cost will keep on increasing, even if the income meant to cover it is at most highly uncertain. This leads us to the main question of the paper: can colleges actually afford to meet the growing cost of paying their athletes? In order to find out whether or not schools can keep up with the demands, this analysis looks at four major pressures—specifically, the growing financial gap between the richest conferences and the others, the way in which money is allocated among the different sports at a university, the legal risks involved, the failure to enforce the spending limit, and the limitations of the funding sources on which universities currently rely.

Literature Review

The Widening Gap Between the Power Four and the Group of Five

The financial gap between the richest college football programs and all the others has grown as a result of the House ruling. This gap is not merely about the facilities that the schools have; it is now clearly seen in the amount that each school can pay its athletes. The settlement permits the participating schools to spend up to 22% of their athletics revenue on athlete compensation, and this percentage rises each year—projecting a sum of $33 million per school by 2034–35 (Ropes & Gray, 2025). In this era when there is free market movement of players, the programs have to use up their entire allowance if they are to remain competitive in both recruiting and in keeping players, since if one school decides not to spend, its competitors will very likely do the same instead (Webster & Carr, 2026). In theory all Division I schools are required to abide by the same limit. In practice, however, only the wealthiest programs can afford to meet that limit.

The gap in spending power is huge. Power Four schools can pay their athletes much more because the House limit is based on a percentage of each school’s revenue (Webster & Carr, 2026). The Group of Five conferences are the American, Conference USA, the MAC, the Mountain West, and the Sun Belt. Conferences such as the MAC and Conference USA each earned less than half a billion dollars (Webster & Carr, 2026). These differences stem directly from television contracts: the Big Ten’s media agreement is worth about $8 billion and the SEC’s is approximately $7.1 billion, while smaller conferences receive only a small share of that amount (Webster & Carr, 2026). The amount that a school can spend on athletes is simply a reflection of the television money its conference already has. A school in the Power Four can easily afford the full $20.5 million, while a school in the Group of Five cannot come close.

The House admitted this fact, noting that the Group of Five programs produce considerably less revenue and will therefore spend much less than the amounts allowed by rule (Webster & Carr, 2026). The “cap” does not actually provide a level playing field; it is a ceiling that only the wealthy can attain. For the majority of the Group of Five schools, the more pressing question is not how to reach the cap but whether or not it makes financial sense to join at all. Some have decided that it does not; the entire Ivy League chose not to opt in, and smaller programs such as the University of North Carolina at Asheville decided to remain outside, stating that opting in “would not be advantageous for [its] long-term success” (Ropes & Gray, 2025). In some cases, the cost of trying to keep up may even lead to varsity sports being cut or the school dropping down to a lower division entirely (Ropes & Gray, 2025).

The Irony of Playoff Expansion

The fact that this increasing gap is so obvious is that it runs up against a change which was meant to have the opposite effect. Beginning with the 2024 season, the College Football Playoff (CFP) raised the number of teams from four to twelve in order to make the sport more accessible and to give a greater number of programs a real opportunity to win a national title. In the new system, the highest-ranked team from the Group of Five is guaranteed a place, thus ensuring that a smaller program will reach the sport’s most important event every year (NCAA, 2026). Effectively, the expansion was a commitment to access – a structural move aimed at putting an end to the situation in which a small number of dominant programs had controlled the postseason for a decade.

It is absolutely ironic. As the CFP grew in size, NIL and revenue sharing had greatly widened the gap in talent among the universities concerned. This led to a first-round series of mismatches, since all four of the first-round games were one-sided, each being decided by an average margin of about 19 points and with only one lead change in the four games taken together (Boston Globe, 2024). Boise State, the champion of the Mountain West and the first Group of Five team ever to make it to the playoff, suffered a loss and was defeated 31–14 by Penn State (Yahoo Sports, 2025). Although the playoff had in fact kept its promise of providing access, that access had not brought about any real competitiveness. NIL and revenue sharing go against the very aim of expansion, since they cause the best talent to concentrate at the schools that can afford to pay the most. A team from the Group of Five can now qualify, but it will then have to face teams whose rosters cost many times as much to put together. As demonstrated by Boise State’s heavy defeat at the hands of Penn State, access without the means to compete results in only nominal inclusion.

It must be acknowledged that the different studies do not all reach the same conclusion regarding whether NIL actually results in a concentration of talent at the top. Li and his colleagues (2025), in Management Science, found that talent became more evenly distributed as a result of NIL, with some programs that were ranked lower attracting higher-quality recruits. However, this finding relates to a different matter: while it concerns the recruits who join the programs, it says nothing about a department’s ability to keep its roster from year to year. A well-funded program might lose one recruit and yet still end up spending a great deal more than a smaller school over the entire period of its roster, both through the transfer portal and in terms of retention. Hence, the dispersion observed in one recruiting class does not offset the structural benefit that the House cap gives to the Power Four.

The Divide Within: Revenue Sports, Non-Revenue Sports, and Title IX

The problem is not merely evident in the state of affairs at schools but also shows itself in the increasing division within each department between sports that generate money and those that do not. Even though the settlement did not specify exactly how the schools should allocate their funds, the formula used to calculate the back payments provided a clear precedent: of the about $2.8 billion paid out to former athletes, 95 per cent was awarded to athletes in Power Five football and in both men’s and women’s basketball (75 per cent to football, 15 per cent to men’s basketball and 5 per cent to women’s basketball), while only 5 per cent was given to all the other sports. This has now triggered new discussions regarding budgets and Title IX since the majority of the money is being focused on a small number of programs (Ropes & Gray, 2025). Schools have begun to adopt this approach. For instance, LSU stated that football would receive 75 per cent of its revenue-sharing money and men’s basketball 15 per cent (Cronkite News, 2025). Despite the fact that a university may have twenty or more varsity sports, only two or three of them receive substantial compensation.

This presents a new structural issue. Previously, funds from football and men’s basketball were used to support all the other campus sports. Nowadays, the settlement returns a fixed amount of that money to the athletes who have earned it. Since expenses are increasing, the amount of money available to support the other sports is decreasing. For this reason, schools are having to alter their budgets, increasing student fees, and are considering cutting or reducing their varsity sports (Loeb & Loeb, 2025). Moreover, the settlement has substituted scholarship limits with roster limits, a change which caused a delay in court approval since it would have meant that some athletes would lose their places (Ropes & Gray, 2025). It is now possible for walk-on and backup players to lose their positions so that the department can afford the athletes.

The fact that the distribution is uneven leads to concerns and also involves legal risks since the majority of the money goes to men. Nearly 90 per cent of the settlement payments are made to male football and basketball players (Wingert Grebing Brubaker & Walshok, 2026), which means that male athletes are paid about 90 per cent more than female athletes (Venable, 2025). Eight of the female athletes immediately appealed to the Ninth Circuit just five days after Judge Claudia Wilken had approved the settlement, challenging the back-pay rule under Title IX (Venable, 2025). In order to avoid this problem, the settlement concentrated on the antitrust issue and excluded the Title IX claims (Syracuse Law Review, 2025). The departments are now setting up long-term payment plans based on a distribution that the courts have not yet confirmed. Should a later decision require a more equal split between men and women, the schools would have to either pay the football players less or increase their spending on women’s sports in order to comply. If they decide to increase their spending, their costs will go up again.

Rules Without Teeth: Enforcement and Cap Circumvention

A salary cap only works if it is enforced, and any third-party deals worth $600 or more have to be reported through this portal within five business days and then checked to make sure they have a valid business purpose (Bradley, 2026). Bryan Seeley, who is the chief executive of CSC, said that all the deals had to be legitimate and should not be used as a means of paying for favors (ESPN, 2025).

The problem is that income from third-party NIL agreements remains completely outside the cap, a flaw which critics have called a major disadvantage. While there are restrictions on how much a school can pay an athlete directly, there are no limits on the amount the athlete earns from outside sources, which means that the main way the rule is worked around is by means of agreements with third parties. Schools have used multimedia rights partners, sponsors, and apparel brands to avoid the cap by channeling revenue from their athletic departments into artificial NIL arrangements (Yahoo Sports, 2026). It comes as no surprise since SEC commissioner Greg Sankey stated that changes were necessary because the present system allows the cap to be circumvented through third-party NIL arrangements (Yahoo Sports, 2026). The amount involved is considerable—the CSC has approved almost $300 million in third-party deals since the system was introduced (Yahoo Sports, 2026), and one general manager from the Power Four said that some programs spend more than $40 million in order to field competitive teams (NIL-NCAA, 2026).

Enforcement is in question even when it is attempted. Even though the CSC obtained its first final arbitration in 2026, stopping a number of football transactions in Nebraska which had been deemed improper ‘warehousing’, the lawyers for the House plaintiffs argued that the CSC had exceeded its authority (Buchanan Ingersoll & Rooney, 2026). The CSC is a private body that has been contracted by Deloitte and is not subject to the Freedom of Information Act (Webster & Carr, 2026). It has now been asked to keep a close eye on billions of dollars. The system relies on an entity with unclear legal powers and proceedings that are closed. If the cap does not in fact restrict spending, the amount that schools are budgeted against will not be the real sum they will have to pay.

Where Does the Money Actually Come From?

The one genuine problem that schools have to deal with is how they are to pay for it; in other words, the majority of them cannot afford to meet these expenses by drawing their money from their current athletics, as the table showing the amount earned by the Power Four conferences in comparison to the others demonstrates (see Table 1). According to the NCAA data, only a small number of departments are able to meet their own costs, and the rest have to depend on student fees, institutional support, or go into debt (Minding the Campus, 2026). Because the costs involved reach eight figures even in departments which are already running at a loss, universities have to look for other sources of funding.

Table 1

Total Athletic Revenue by Conference, 2023–24 (Knight-Newhouse College Athletics Database)

The most evident measure has been the imposition of higher fees: Clemson charges $150 per semester, South Carolina charges $300 per year, Minnesota has a fee of $200 per year and West Virginia charges $125 (SportsEpreneur, 2026). Florida has given approval for auxiliary funding amounting to $22.5 million for its sports program (Front Office Sports, 2025), and Virginia Tech has raised its tuition in order to cover the costs of the athletes (Ropes & Gray, 2025). In some instances, the organisations have had to rely on fundraising and on reorganising: Syracuse has launched a $50 million campaign in order to attract and keep athletes (University Business, 2025), and Kentucky has restructured its athletic department into a non-profit LLC (Ropes & Gray, 2025). Even the conferences which have the greatest resources are now considering private equity as a means of stabilising their cash flow (Webster & Carr, 2026).

There is here a marked irony which once again leads us back to the original issue. The universities that are expected to be the ones getting the greatest amount of financial support—for example, Ohio State, Texas, and Alabama—charge their students very little or nothing at all since television revenue covers the costs. Universities that do not have television revenue, on the other hand, have to pass the cost on to their students (SportsEpreneur, 2026). The worst impact is felt by the schools that are least able to absorb it. A student going to a Group of Five school may now have to pay a compulsory fee in order to finance a payment system that will never make her school competitive with the programs it was meant to help it catch up with. Tuition, fees, and donations are all limited resources. If the cap reaches $33 million by the mid-2030s without a corresponding increase in revenue, the current funding system will become mathematically unworkable for a large number of Division I schools.

A Countervailing Benefit

It is inaccurate to describe the years after the House era as having been merely a difficult time, since one important advantage has developed: Name, Image, and Likeness (NIL) payments are now causing top athletes to remain in college for longer. At the 2026 NBA Draft, all of the twenty players selected in the first round had played college basketball, a first occurrence since 1994 (NBA, 2026). A year before that, in the second round of the 2025 draft, a record number of senior players were picked, a situation which can be directly attributed to NIL, because the players received payment to stay on rather than face the risk of ending up with an uncertain draft position (Sports Illustrated, 2025). For many decades, this incentive has encouraged able athletes to leave college early; NIL has reversed that trend so that more of their academic work is completed and those who do not make it to the professional level have a more practical alternative. About 60% of former NBA players are in financial difficulty five years after they retire (Henderson, 2023), and athletes who stay on for longer experience a rapid influx of wealth while still at university, where they can learn how to manage it—Henderson (2023) points out that states such as Georgia and Louisiana already have requirements of this type. Yet this benefit should not be overestimated, since it is just as unequally distributed as all the other aspects of the issue—only those who can afford to stay are the ones receiving large amounts of money, and this money is concentrated in the well-off programs mentioned above.

The cost, which became compulsory as a result of the agreement in the House, is always increasing, yet it has never provided a reasonable means of financing it nor any method of bringing the costs under control. The two problems exacerbate one another: the costs keep on rising, and the situation still remains unbalanced. The easiest way to address this issue is to fix the meaningless spending cap. As long as third-party NIL deals aren’t counted in the salary cap, the $20.5 million cap means nothing, and schools will end up spending with no way to keep track. The cap will only begin to matter if third-party agreements are included and the College Sports Commission gets access to information on all third-party agreements. A salary cap that can’t be enforced is actually worse than having no cap, because it creates a false sense of control and rewards the colleges which find ways around the rules. The second step is designed to close the growing skill gap between powerhouse and underdog schools in their sports teams. A salary cap will not single-handedly help the Group of Five, since these schools cannot even reach the current limit. A more realistic approach would be a tiered system, where the salary cap matches each conference’s revenue, which would be paired with a revenue-sharing fund sponsored by the most profitable media deals. The Power Four conferences, which make billions from TV, can afford to share some of their income to help the schools that make their schedules possible. You might think this is charity, and these Power Four schools would never agree to it. But it is not just charity. In fact, it is in their own interest, because if games become too lopsided, fans will lose interest and TV money will disintegrate. The strongest and most sustainable solution, in my opinion, is for schools to create new sources of income instead of just trying to nurture their current sources.The ways in which money is currently earned rely on student fees, higher tuition charges, and fundraising. These methods are very limited, and in fact student fees might even interfere with the school’s academic objectives. As Webster and Carr (2026) note, the most viable solution is to make use of real estate connected with sports, for example by setting up mixed-use developments that have shops, housing and entertainment situated near stadiums. This type of development generates income throughout the year and, crucially, is not included in the revenue-sharing pool. Professional sports teams have long used this method since it helps to meet the increasing costs of players. Nowadays wealthier athletic departments can adopt this strategy, but Group of Five schools will have to rely on the kind of shared investment that I referred to earlier. However, the basic principle remains unchanged in all situations: if a new cost has to be covered, a new source of income must be found.

Conclusion

In just four years college sports has gone beyond the boundary that it had been nearly at for a hundred years, and this shift is now permanent. The House settlement did more than simply allow schools to pay athletes; it placed limitations on a financial system which was not designed to cope with such expenses, lacked the ability to regulate spending, and provided no safeguards for schools and sports that are falling behind due to a constantly rising and mandatory cost. As a result, the gap between the Power Four and the other schools has grown wider, pressure has been placed on non-revenue sports, the Title IX problem has still not been resolved – a circumstance which could lead to further cost increases – and a funding model has been put in place that will not be sustainable for most of Division I if it persists into the mid-2030s. The problem cannot be solved by redistributing the money that is already available, since that money is all that is available. Whether or not college sports continues to be a national tradition or becomes something that only a small number of wealthy schools can afford will depend on how quickly schools are able to find new ways of meeting the increasing costs of athletes.

References

Boston Globe. (2024, December 22). Blowouts dominated the first round of the College Football Playoff. Will the quarterfinals be any better? https://www.bostonglobe.com/2024/12/22/sports/college-football-playoff-first-round/

Bradley. (2026, January 21). Enforcing after House: The College Sports Commission and the future of NIL regulation. https://www.bradley.com/insights/publications/2026/01/enforcing-after-house-the-college-sports-commission-and-the-future-of-nil-regulation

Buchanan Ingersoll & Rooney. (2026, May 28). The College Sports Commission prevails in NIL arbitration. https://www.bipc.com/college-sports-commission-prevails-in-nil-arbitration

Cronkite News. (2025, July 10). What the NCAA settlement means for ASU, athlete pay and Title IX. https://cronkitenews.azpbs.org/2025/07/10/asu-house-settlement-athletes-title-ix-2025/

De Lima, M. (2025, August 28). College football’s NIL revolution surges toward $2 billion. Sports Illustrated. https://www.si.com/fannation/college/cfb-hq/nil/college-football-nil-revolution-surges-toward-2-billion

ESPN. (2025, July 31). College Sports Commission loosens prohibition on NIL payments. https://www.espn.com/college-sports/story/_/id/45867746/college-sports-commission-loosens-prohibition-nil-collectives-payments

Front Office Sports. (2025, June 23). Colleges are raising student fees to pay for athlete revenue-sharing. https://frontofficesports.com/colleges-are-raising-student-fees-to-pay-for-athlete-revenue-sharing/

Henderson, T. (2023). It’s about time: An analysis of name, image, and likeness legislation in the United States. Jeffrey S. Moorad Sports Law Journal, 30(1), 61–90. https://digitalcommons.law.villanova.edu/mslj/vol30/iss1/3

Knight-Newhouse College Athletics Database. (2025). College athletics financial database (FBS). https://knightnewhousedata.org/fbs

Li, I., Lu, F., & Qiao, M. (2025). Does personalized pricing increase competition? Evidence from NIL in college football. Management Science. https://pubsonline.informs.org/doi/10.1287/mnsc.2024.06423

Loeb & Loeb. (2025, July 1). College sports enters a new era: House settlement greenlit, student-athletes to be paid. https://www.loeb.com/en/insights/publications/2025/07/college-sports-enters-a-new-era-house-settlement-greenlit-studentathletes-to-be-paid

Minding the Campus. (2026, January 26). College athletics is breaking universities. https://mindingthecampus.org/2026/01/26/college-athletics-is-breaking-universities/

NBA. (2026). 5 takeaways: College players rule 1st round of 2026 NBA Draft. https://www.nba.com/news/five-takeaways-2026-draft-1st-round

NCAA. (2026, February 3). How the 12-team College Football Playoff will work: Teams, schedule, bids. https://www.ncaa.com/news/football/article/2026-02-03/how-12-team-college-football-playoff-will-work-teams-schedule-bids

NIL-NCAA. (2026). NCAA revenue sharing and NIL estimates. https://nil-ncaa.com/

Opendorse. (2025). NIL at four: Monetizing the new reality. https://biz.opendorse.com/wp-content/uploads/2025/07/NIL-at-Four-Monetizing-the-New-Reality_July2025.pdf

Ropes & Gray. (2025, June 20). House v. NCAA settlement approved: Era of direct payments to college athletes begins. https://www.ropesgray.com/en/insights/alerts/2025/06/house-v-ncaa-settlement-approved-era-of-direct-payments-to-college-athletes-begins

Sports Illustrated. (2025). Second round of 2025 NBA Draft sees record number of seniors selected. https://www.si.com/nba/draft/newsfeed/second-round-of-2025-nba-draft-sees-record-number-of-seniors-selected-01jys8zec3pe

SportsEpreneur. (2026, June 11). Student athletic fees: Who really pays for college sports? https://sportsepreneur.com/student-athletic-fees-college-sports/

Syracuse Law Review. (2025, June 19). Title IX and the House settlement: Playing for keeps. https://lawreview.syr.edu/title-ix-and-the-house-settlement-playing-for-keeps/

University Business. (2025, July 21). How student-athlete revenue-sharing is changing college finances. https://universitybusiness.com/how-new-athlete-revenue-sharing-is-causing-staggering-financial-change/

Venable. (2025, July 14). Title IX goes head to head with antitrust: NCAA NIL settlement challenged by female student-athletes in House v. NCAA. https://www.venable.com/insights/publications/2025/07/title-ix-goes-head-to-head-with-antitrust-ncaa

Webster, N. R., & Carr, J. (2026). From NIL to real estate: Strategic revenue adaptation in the post-House era of college athletics. Frontiers in Sports and Active Living, 7, Article 1728507. https://doi.org/10.3389/fspor.2025.1728507

Wingert Grebing Brubaker & Walshok. (2026, February 28). The House v. NCAA effect: How the $2.8 billion settlement is rewriting the rules for California college sports. https://wingertlaw.com/2026/02/28/house-v-ncaa-settlement-california-guide/

Yahoo Sports. (2025). What is the Group of 5 in college football? Explaining the conferences with CFP automatic bid. https://sports.yahoo.com/article/group-5-college-football-explaining-080002236.html

Yahoo Sports. (2026, June 17). With Protect College Sports Act under scrutiny, NIL deals with “associated entities” creating confusion. https://sports.yahoo.com/college-football/article/with-protect-college-sports-act-under-scrutiny-nil-deals-with-associated-entities-creating-confusion-204731149.html

About the author

Will Nemo

Will is a senior at Riverdale Country School in New York City. His academic interests lie in finance, economics, and the business of college athletics. Outside academics, Will plays quarterback for his school’s varsity football team and follows the college sports industry closely.

The post The Cost of Keeping Up: Athlete Pay and Financial Sustainability in College Sports  appeared first on Exploratio Journal.

1. Introduction

Environmental, social and governance (ESG) ratings are a framework to score companies based on three broad criteria. Firstly, the environmental dimension considers a firm’s impact on the natural ecosystem. Secondly, the social dimension evaluates the firm’s relationship with its stakeholders: customers, suppliers and communities. Lastly, governance focuses or the board structure and diversity.

Rating the ethics of a company isn’t a new practice. Before ESG, there was corporate social responsibility (CSR) in the 1990s and even prior to that there was socially responsible investing (SRI) in the 1970s ((Krantz, 2024). But neither CSR nor SRI ever gained the recognition ESG has in the past decade. This can be accounted for due to a key difference in their foundations. Measures like CSR and SRI believed and conveyed that businesses should behave ethically because it is morally the correct thing to do. In contrast, ESG ratings aim to score a company’s sustainability and vulnerability to external factors – variables that link directly to monetary returns. As a result, the rigorous set of variables that ESG evaluates makes it an excellent metric to assess a company’s performance through factors that don’t appear on the traditional financial balance sheets. Hence, more and more investors are starting to incorporate these ratings, along with the financial data, to determine a company’s financial return on investment (ROI).

However, the author’s evaluation of rating methodology of Crisil ESG Ratings & Analytics Ltd, a central ESG provider in India, reveals a crucial flaw in current ESG calculation that significantly hinders ESG’s ability to indicate financial performance. Currently a uniform ESG pillar weightage framework of 40% governance 35% environment and 25% social is used across all sectors. While this may be apt for sustainability indication it hinders financial performance indication as the uniform weightage scheme doesn’t account for sector-specific materiality differences. For example, the environment pillar may be far more important for an energy company then for an information technology, for whom governance may be more important. To address this limitation, this section empirically investigates whether sector-specific ESG pillar weighting systems improves the strength of the relationship between ESG scores, and return on capital Employed (ROCE) and aims to find the most optimum ESG weighting systems for 12 sectors. Thus, the research paper aims to answer the question:

What is the optimal sector-specific weighting system of environmental, social, and governance pillars for explaining variation in ROI, and do these optimized models outperform equal-weight ESG frameworks in the Indian market?

The research paper is structured as follows. It begins with a literature review on the importance of considering a company’s qualitative aspects while investing and studies regarding correlation between ESG and ROI. Followed by this is the development of a sector specific ESG weightage scheme through correlations between ESG and return on capital employed (ROCE). The paper concludes by outlining limitations and identifying avenues for future research.

2. Literature Review

2.1 The Efficiency of Traditional Financial Metrics

Investors have relied on and further developed financial analysis for nearly a century. As a result, most indicators of future growth have already been discovered and are heavily used by investors. The Efficient Market Hypothesis, as developed by Fama (1970) states that in a semi-strong efficient market publicly available financial information is rapidly priced in. Consistent with this, Green, Hand, and Zhang (2013) show that most well-documented return predictors lose statistical significance once widely known, as arbitrage activity erodes whatever informational edge they carried.  

These findings collectively suggest that due to most forms of analysis of traditional financial metrics being well known company stocks tend to be priced accordingly. Due to this, it has become increasingly difficult to generate excess returns through these forms of financial analysis and investors must look for other sources of information.

2.2 Unexplained Volatility and the Case for Qualitative Metrics

Traditional financial variables explain less of stock price behavior than is commonly assumed. Roll (1988) documented that financial and industry factors only accounted for a small part of individual stock variation; the remainder being determined by firm specific qualitative factors. This is because financial analysis can’t predict regulatory exposure, reputational events, stakeholder dynamics, and governance failures.

More recent empirical work further proves this trend. Edmans (2011) finds that firms with strong employee satisfaction earn returns of roughly two to three percent annually above market benchmarks and these results persist over time. Similarly, Hartzmark and Sussman (2019) indicate that investor invest more money in response to sustainability signals even in underlying financial variables remain constant. This indicates qualitative firm characteristics influence share prices through channels that conventional financial metrics miss. 

This pattern along with the previous section show that quantitative financial data is priced quickly and efficiently, while qualitative signals are under processed. That gap is where investors can earn excess returns. ESG frameworks are an excellent cumulation of these qualitative indicators. The question is whether it reports this information in a way that is useful for predicting financial performance or does its goal of reporting sustainability cause it to stray from being an indicator of future financial performance.

2.3 Correlation Between ESG and Financial Performance

There is extensive literature on ESG and financial performance. The broad findings can be summarized by Friede, Busch, and Bassen (2015) review of over 2,000 empirical studies, finding that 90% reported a non-negative relationship between ESG scores and financial performance, with the majority showing a positive one. While this isn’t perfect it shows potential.

The study by Khan, Serafeim, and Yoon (2016) distinguish between ESG factor that are financially material to a given industry and those that are not. Firms outperform their competitors when they score well on the material ESG factors while the immaterial ones show no correlation at all.

These two studies, taken together show that currently ESG scores have not reached their financial performance indication potential, primarily due to composite scores not accounting for differing ESG factor materiality across different firms. This is a result of the uniform ESG weightage scheme across all sectors. Hence, this literature review shows that developing a new sector- specific ESG weightage system, so that the final ESG scores give more importance to the more material factors based on the sector, can make ESG an indicator of future financial performance. Furthermore, this would allow investors to identify companies not valued for their qualitative factors allowing them to earn excess returns.  

3. Proposed Sector-Specific Optimized ESG Weighting scheme and Correlation with ROCE

3.1 Methodology

3.11 Variables

Table 3.11  Variables Note. Author made

3.12 Data Sourcing and Sector Classification 

The study uses ESG scores published by Crisil for the financial years 2023 and 2024. The sample comprises 748 listed Indian companies spanning 12 sectors. The 12 sectors were decided to ensure simplicity as well as homogeneity in regard to main pillar importance in each sector.

The following sector classification has been used:

Table 3.12  Sector classification Note. Author made

Firm-level ROCE data was collected from two independent, reliable financial databases: “Screener” and “Equitymaster”. This ensured consistency and cross-verification of financial performance data. The ESG data for composite scores and main pillar scores was obtained from Crisil’s archives. Companies with incomplete ESG pillar scores or missing ROCE data were excluded.

3.13 Framework of Quantitative Analysis

The strength of the relationship between ESG scores and ROCE was measured using Spearman’s rank correlation coefficient ( . This is beneficial as outliers are extremely common in sectoral ROCE data, making it important to reduce their impact of empirical results as done by this metric. All correlations were computed on a sector-by-sector basis to account for structural heterogeneity across industries and correlations were reported to an accuracy of two decimal places to enforce accuracy.

A python program was developed to generate sector specific main pillar weightage systems that were more optimized ROCE indicators based on Crisil’s individual main pillar scores, environmental, social and governance. For each firm i in sector s, the composite ESG score was defined as:

subject to:

Two sets of weightage systems were generated:

1. Unconstrained Model:

The objective of this model was to generate weightage systems that maximized ESG’s correlation with ROCE. No lower bound for a main pillars weightage was set, which means the model could deliver a weightage scheme in which the optimum weightage for a main pillar for a sector is 0%.

1. Constrained Model:

The objective of this model remains the same as the unconstrained model but a lower bound for a main pillars optimum weight was set at 15%. By setting this minimum at 15% the conceptual integrity of ESG evaluation is preserved.

For each sector, three correlations were calculated to draw comparisons:

1. Crisil’s composite ESG score and ROCE
2. Unconstrained optimized ESG score and ROCE
3. Constrained optimized ESG score and ROCE

3.2 Baseline: Correlation Under Crisil’s Fixed Weightage Scheme

Crisil’s ESG framework calculates composite scores through a uniform weighting structure across sectors. Governance with the highest weightage (40%), followed by environmental (35%), and lastly social (25%). This allows comparability but assumes that environmental, social, and governance risks affect all industries equally. To test this assumption, sector-wise correlations between ESG scores (under Crisil’s weighting scheme) and ROCE were computed.

Table 3.2 Sector-Wise Correlation Between ESG and ROCE (Crisil Weightage Scheme) Note. Author’s calculations based on raw data from souces:18, 19, 20 and 21

The results demonstrate substantial heterogeneity not only in the magnitude but also in the economic interpretation of the ESG–ROCE relationship across sectors. 

Industries such as communication services, informational technology demonstrated strong positive correlations due to the structural importance of governance quality and regulatory compliance (governance and environment have high weightages in Crisil’s weightage scheme). In contrast, financials consumer discretionary – auto (-0.02) show near-zero correlations. This can be accounted to the fact that in financials, ROCE may be influenced more heavily by macroeconomic cycles driven through context rather than ESG scoring differentials. As for auto manufacturing, majority of the companies aren’t automobile brands themselves but brand that produce automobile parts. Due to the massive information asymmetries in India and more lenient production rules these companies tend to a low correlation with ESG in present. Additionally, cyclical demand fluctuations, technological transitions (e.g., EV shifts), commodity price volatility, and high fixed-cost structures materially shape ROCE. For this very reason, it was separated from the consumer discretionary sector to not obscure the correlation of the rest of the companies in that sector.

Relatively moderate positive correlations in energy, industrials, healthcare, consumer staples, consumer discretionary – durables and services, materials, real estate and utilities) suggest partial alignment but also potential miss weighting of pillars. For instance, environmental factors may be economically dominant in energy, but Crisil’s model still gives the environmental pillar the lowest weightage.

Collectively, these patterns indicate Crisil’s weightage scheme is able to predict financial performance more strongly for companies that whose ESG factor materiality aligns with Crisil’s weightage scheme. The findings therefore provide substantive economic reasoning for investigating sector-optimized weighting models rather than relying on a single standardized ESG structure.

3.3 Optimized ESG Weightage Scheme 

To test whether sectoral customization improves explanatory power, ESG weights were optimized to maximize correlation with ROCE. A python program was used to come up with unconstrained and constrained weightage schemes to maximize ESG’s correlation with ROCE.

Table 3.31  Sector-Wise Unconstrained Optimization System’s Correlation with ROCE Note. Author calculated based on raw data from Equitymaster, Screener and Crisil Ratings and Analytics

As the results indicate, the unconstrained weighting system results in theory significant jumps in ESG’s correlation with ROCE. However, in practice these could vary significantly based on the data set used and since a lot of weightages for environmental, social or governance pillars are near zero in some sectors this weighting scheme pulls away from the very concept of ESG. Thus, it is of more relevance to determine and the optimized weighting system that is constrained at a minimum of 15% weightage for each main pillar as this keeps the core, multidimensional concept of ESG intact while allowing for a generalization as the most appropriate sector-specific weightage scheme in the Indian Market.

Table 3.32  Sector-Wise Constrained Optimization System (Minimum 15%) Note. Author calculated based on raw data from Equitymaster, Screener and Crisil Ratings and Analytics

Under the constrained optimization system, communication services record the strongest correlation under the constrained framework, reflecting a very tight alignment between ESG performance and capital efficiency. Information technology and energy also demonstrate relatively strong positive associations due to the government-dominant allocations (G ≈ 0.68) that influence oversight, capital discipline, and regulatory management. Moderate correlations are observed in consumer staples, real estate, industrials, and utilities as while the optimization of weightages accounts for sector materiality ESG factors there are sector specific aspect in each that are not accounted by ESG factors.

However, financials and consumer discretionary – auto continue to exhibit weak correlations despite governance-heavy weightings (0.70 in both sectors). This suggests that even when pillar emphasis is adjusted to reflect sector logic, ESG scores may not translate strongly into ROCE variation due to reasons explained in detail in 3.2 which ESG factors simply can’t account for. ESG performance may influence long-term strategic positioning or risk mitigation in these sectors.

These findings indicate that while sector-specific optimization improves statistical alignment in several industries with most demonstration a strong or moderate correlation it cannot uniformly overcome structural characteristics that weaken the ESG–ROCE transmission mechanism. The constrained framework proposed therefore demonstrates the strengths and limits of ESG factor’s ability to indicate financial performance: its predictive power is considerable when firms are affected by the qualitative factors ESG reports but gets weaker when financial outcomes are driven primarily by macroeconomic or capital-cycle forces.

3.4 Comparative Performance: Crisil vs Proposed Sector-Specific Weighting Scheme

Table 3.4 Comparative Performance: Crisi’s Weightage scheme vs Author’s Proposed Weightage Scheme  Note. Author calculated based on raw data from Equitymaster, Screener and Crisil Ratings and Analytics

As indicated by the data, applying sector specific weightages drastically improves the correlation between ESG and ROCE across most sectors. The most significant improvement is seen in real estate (consumer discretionary – auto and financials have high percentage improvement but correlation is still near-zero) as the proposed weightage scheme gives allows for governance to drive the composite score as in India it is the most material to financial performance. Energy Consumer Staples and Utilities have also shown strong improvements due to improved sector-specific materiality considerations by the proposed weightage scheme. Moderate improvements can be seen in Consumer Discretionary – Durables & Services, Information Technology, Healthcare, Materials. Only two sectors Communication Services and Industrials showed small improvements as for both Crisil’s weightage scheme fairly rated these companies through their weightage system leaving little room for improvement.  

Taken together the table demonstrates the need for sector-specific weightage schemes for ESG calculation. The proposed weightage scheme gives importance to ESG pillars based on their sector specific materiality to financial performance allowing for ESG to become a substantial indicator of financial performance.

3.5 How to Calculate ESG Scores for Multi-Sectorial Companies.

The weightage schemes this research paper has presented makes it clear as to the method used to calculate the most reliable ESG ratings for most companies. However, there are certain companies who don’t entirely fit into one sector. An example of this is reliance limited.

For a multi-sector company like Reliance Industries Limited, the ESG score should be a weighted sum of the scores calculated using each sector’s optimal weightage.

The General Formula:

Where:

• ​W_sector is the percentage of revenue from that sector
• ESG_sector is the ESG score calculated using the optimal weights for that specific sector.

4. Strengths and Limitations of the Quantitative Analysis

This study has several methodological strengths. Data used is from reputable sources: Screener, Equitymaster and Crisil Ratings and Analytics. Furthermore, the study only uses data available to everyone publicly which makes this highly relevant for retail investors and smaller firms, not just larger corporations. The clear segregation into 12 sectors makes using the proposed weightage scheme straightforward and the system also accounts for multi-sectoral companies. Additionally, the use of ROCE as the dependent variable strengthens the analysis by focusing on operational capital efficiency instead of market-based performance. This aligns ESG evaluation with fundamental financial productivity. Statistically, the Spearman’s correlation was used to reduce weightage of outliers as when data is limited outliers could significantly harm reliability of the conclusion. Lastly, the implementation of both unconstrained and constrained optimization frameworks allows for differentiation between purely statistical maximization and theoretically grounded calibration. 

However, certain limitations must be acknowledged. The study is conducted over a limited period, due to lack of publicly available ESG data in India. This potentially restricts insights into the stability of optimized weightages overtime and across economic cycles. Secondly, ESG scores are published annually as an aggregate which doesn’t account for intra-year variation in sustainability performance that could have affected ROCE. Sizes of sector-level datasets may have also affected overall results. Most importantly, the dataset is exclusively composed of Indian companies. As a results, the proposed weightage scheme is only directly applicable to India and potentially other economies at similar stages of development. This is because the sector-specific weightage system gives importance to the pillars based on their materiality due to the Indian context such as local information asymmetries and the tendency of Indian government to impose regulation.  For example, it is likely that a highly developed country has low information asymmetry regarding business practices (between businesses and consumers), whereas in an underdeveloped or develop this asymmetry may be higher. Therefore, the optimized weightage schemes identified in this study are most appropriate for India or countries with comparable institutional and developmental characteristics. Due to this same concept, these ESG weightage systems are also only relevant in India while it remains in its current broad development stage.

5. Concluding Remarks

This research has examined the evolving role of ESG as an indicator of corporate financial performance within the Indian context. While frameworks like those provided by Crisil have significantly improved the accessibility of sustainability data, this study identifies a critical gap in the application of uniform, non-sector-specific weighting systems. Such models fail to capture the unique operational parameters and types of risks inherent to different industries, hence diluting the correlation between ESG performance and financial returns.

Through empirical analysis and quantitative modeling, this paper produces a sector-specific weighting scheme, tailored to the environmental, social, and governance priorities of specific industries, that serves as a far more robust indicator of ROCE. By improving the framework of existing ESG practices as financial performance indicators this paper aims for ESG ratings to be more widely used alongside financial metrics while predicting the future share performance of a firm. 

The proposed framework also carries broader implications for markets and society. By providing a stronger and more consistent correlation between ESG performance and financial returns, this framework increases the likelihood that sustainability considerations will be integrated into investment decision-making. This could lead to qualitative factors regarding companies’ sustainability and ethics also being priced in (as financial metrics are now) in the long run. Once this has happened more companies are likely to make most decisions keeping consequences regarding sustainability and ethics in mind with the profit-maximizing goal of improving their stock price. This will result in huge positive externalities as more firms demonstrate responsible environmental stewardship, ethical governance structures, and socially sustainable practices. In the long run, such alignment between financial incentives and sustainability outcomes has the potential to reinforce ethical corporate conduct and hence have considerable benefits to society from an economic and social perspective. 

6. Future Scope for Research

The author intends to pursue further research and publication in the areas outlined below, building upon the analytical framework developed in this study.

1) Cross-Country Optimal Weightage Comparative Studies: Applying this optimization framework across multiple countries at varying stages of development would determine whether sectoral ESG relevance is development dependent. It would provide insights into how differing governments’ regulation tendencies and information asymmetries across the development spectrum affect the ideal ESG weightages for financial performance indication. Moreover, conclusions from such a study could be significant as it could lead to a significant alteration in the methodology of major multinational ESG providers like MSCI transforming the ESG landscape completely and leading to much higher correlations between ESG and ROI globally. 

2) Multi-Year Longitudinal Analysis Across Economic Cycles: Extending this study over a larger time frame once there is Indian ESG data available over more years would make this system far more reliable and would also allow for insights into how the optimum ESG weightages change across sectors through economic cycles. 

3) High-Frequency or Quarterly ESG Data: If quarterly ESG metrics become available in India, studies on how ESG shocks influence ROCE within shorter intervals. This would be useful to determine the speed of ESG-financial transmission mechanisms which would be useful for investors as it will indicate the average time they have to act on information reported by ESG.

4) ESG in Stress Periods: Since ESG is primarily an indicator of sustainability, such a study would indicate whether firms with higher ESG ratings are able to outperform competitors during stress periods.  

References

“Crisil ESG Ratings & Analytics Limited.” CRISIL, 2021, www.crisilesg.com/en/home.html.

“Crisil ESG Ratings & Analytics Ltd Gets SEBI Nod to Offer ESG Ratings.” CRISIL, 2024, www.crisilratings.com/en/home/newsroom/press-releases/2024/04/crisil-esg-ratings-and-analytics-limited-gets-sebi-nod-to-offer-esg-ratings.html.

Edmans, Alex. “Does the Stock Market Fully Value Intangibles? Employee Satisfaction and Equity Prices.” Journal of Financial Economics, vol. 101, no. 3, 2011, pp. 621-640.

Equitymaster – Get Unbiased Equity Research on Indian Stocks and Share Market Updates for Value Investing in India. Equitymaster, 2025, www.equitymaster.com.

“ESG Ratings 2023.” CRISIL, 2023, www.crisilesg.com/en/home/esg-ratings/historical-ratings/esg-ratings-2023.html.

“ESG Ratings Criteria and Methodology.” CRISIL, 2025, www.crisilratings.com/content/dam/crisilesg/criteria-and-methodology-for-crisil-esg-ratings/archives/criteria-and-methodology-for-esg-ratings-feb-2025.pdf.

“ESG Ratings List 2024.” CRISIL, 2024, www.crisilesg.com/en/home/esg-ratings/historical-ratings/esg-ratings-2024.html.

Fama, Eugene F. “Efficient Capital Markets: A Review of Theory and Empirical Work.” The Journal of Finance, vol. 25, no. 2, 1970, pp. 383-417.

Friede, Gunnar, et al. “ESG and Financial Performance: Aggregated Evidence from More than 2000 Empirical Studies.” Journal of Sustainable Finance & Investment, vol. 5, no. 4, 2015, pp. 210-233.

Green, Jeremiah, et al. “The Supraview of Return Predictive Signals.” Review of Accounting Studies, vol. 18, no. 3, 2013, pp. 692-730.

Hartzmark, Samuel M., and Abigail B. Sussman. “Do Investors Value Sustainability? A Natural Experiment Examining Ranking and Fund Flows.” The Journal of Finance, vol. 74, no. 6, 2019, pp. 2789-2837.

Khan, Mozaffar, et al. “Corporate Sustainability: First Evidence on Materiality.” The Accounting Review, vol. 91, no. 6, 2016, pp. 1697-1724.

Krantz, Tim. “The History of Environmental Social and Governance (ESG).” IBM, 8 Feb. 2024, www.ibm.com/think/topics/environmental-social-and-governance-history.

Roll, Richard. “R2.” The Journal of Finance, vol. 43, no. 3, 1988, pp. 541-566.

“Stock Screener and Fundamental Analysis Tool for Indian Stocks – Screener.” Screener, www.screener.in.

About the author

Nimay Shah

Nimay is an IB Diploma student with a keen interest in finance, economics, and quantitative research. His work focuses on applying mathematical and data-driven methods to improve investment analysis, particularly in sustainable finance and ESG.

Through this research, Nimay aims to bridge the gap between academic theory and practical investing by developing more effective approaches to evaluating corporate performance. He hopes to pursue business, finance, or economics at university and continue exploring the intersection of data, markets, and sustainability.

The post ESG–ROI Linkages in India: Developing Sector-Specific Weightage Frameworks for Optimized Financial Performance Indication appeared first on Exploratio Journal.

Abstract

India’s Unified Payments Interface (UPI) has reshaped the country’s financial landscape, establishing India as the global leader in real-time digital payment adoption. However, this rapid expansion has been accompanied by a corresponding rise in digital payment fraud, raising concerns for consumer protection and the effectiveness of structural systems. Applying Routine Activity Theory, the paper analyzes literature to explain this outcome as the product of a convergence between motivated offenders, suitable targets, and absence of capable guardians. Recent studies show that victimization remains consistent despite awareness of digital payment frauds. Educated and technologically aware users continue to fall victim to complex psychological manipulation. Once targeted, victimization largely depends on the exploitation of cognitive biases. Institutional guardianship remains India’s most persistent weak point. Victims recover a fraction of the stolen funds, redressal mechanisms are largely unsatisfactory, and legal loopholes leave a majority of victims outside of the protection system. The paper concludes with recommendations to put a stop to legal loopholes, modernize legal frameworks, and add an additional layer of protection at the point of transaction rather than relying solely on user vigilance.

Introduction

India’s creation and adoption of the Unified Payment Interface (UPI) is an example of one of the most rapid financial digitisation initiatives in recent history, expanding from 20 million transactions in  FY 2016-17 to over 241.62 billion in FY 2025-26 (Ministry of Finance, 2026). This represents a 12,000-fold increase.  In 2012, over 87% of all transactions in India were conducted in cash, and fewer than 10% of the population had ever used any non-cash payment instrument (Iyer, 2017). Today, India accounts for 48.5% of worldwide real-time digital payment transactions, making it the global leader in digital payments (Singh and Agarwal, 2025). Furthermore, of the total retail digital payment transactions, UPI accounted for over 81% of them, emerging as the largest real-time retail payment system globally (Ministry of Finance, 2025). Between April 2016 (at launch) and March 2026, the number of banks live on UPI increased from 21 to 703. These statistics prove that UPI has evolved from merely being a means of payment to a financial infrastructure critical to India’s economy.

However, the same expansion that produced this success also generated a fraud crisis of proportionate scale. According to the Association of Certified Fraud Examiners, fraud is any activity that relies on deception in order to achieve a gain. Fraud becomes a crime when it is a knowing misrepresentation of the truth or concealment of a material fact to induce another to act to his or her detriment. Within the UPI ecosystem, fraud commonly involves social engineering and other deceptive practices that manipulate users into authorising transactions or disclosing sensitive credentials. The number of reported UPI fraud cases rose from 195,000 in 2020-21 to 7,25,000 in 2022-23, while the amount lost increased from ₹111 crore (approximately $15 million as per the exchange rates for FY 2020-21) to ₹573 crore (approximately $72.4 million as per the exchange rates for FY 2022-23) in the same period (Agrawal, 2026). Across all digital payment modes, fraud cases increased from 1,19,699 in 2020-21 to 14,57,000 in 2023-24, recording a compound annual growth rate (CAGR) of 130.03%, which outpaces the 40-105% annual growth in UPI transactions (Kumar and Dharshan, 2026). Iyer (2017) documented that card fraud in India increased in direct proportion to card adoption throughout 2011–2016, with losses doubling from ₹4 billion ($89.58 million as per the 2011 exchange rates) to ₹7.68 billion ($113.56 million as per the 2016 exchange rates) as card usage expanded. Moreover,  according to Wadkar et al. (2025), UPI fraud accounts for 47.25% of all cybercrime cases reported in India since 2020.Understanding why fraud has scaled at this rate requires examining not only the architecture of UPI itself, but the behavioral and structural conditions that enable exploitation. Statistics alone cannot explain why UPI fraud has scaled at this rate or why users remain vulnerable despite high awareness. This paper applies Routine Activity Theory (Cohen and Felson, 1979) to examine the structural and behavioral conditions that enable this crisis.

Theoretical Framework:

According to Routine Activity Theory (Cohen and Felson, 1979), a criminal event occurs when three elements converge in both time and space: a motivated offender who perceives an opportunity for gain, a suitable target, and the absence of a capable guardian who can prevent the crime. Additionally, Cohen and Felson argue that changes in routine activities, such as shifts in work patterns, leisure activities, and household arrangements, impact the convergence of motivated offenders, suitable targets, and the lack of capable guardians. This, in turn, affects crime rates. Scholars have since extended RAT beyond street crime to fraud contexts. Holtfreter et al. (2008) conducted a study that extended the Routine Activity Theory (RAT) to the context of consumer fraud victimization,  using a sample of 922 participants. Their study found that a one-unit increase in the buying routine index (a measure of how frequently respondents engaged in consumer activities such as online shopping) was associated with a 61% increase in the odds of being targeted for fraud.

Holtfreter et al. (2008) also tested Self-Control Theory alongside RAT, finding that while low self-control significantly increased the probability that a targeted individual would become a victim (302% per unit increase), it had no effect on whether the individual was targeted in the first place. This targeting/victimization distinction is directly relevant to the UPI fraud paradox examined later in this paper: why do 52.5% of users become victims despite 96.2% awareness? Once targeted, a user’s capacity for deliberate, cautious decision-making (rather than prior knowledge alone) determines victimization outcomes.

DeLiema (2018) reinforces this by applying RAT to study financial fraud against older adults, examining 53 cases of financial exploitation. The study investigated interpersonal financial exploitation amongst adults aged 65 years or older. The research found that social isolation, which means a lack of capable guardians like trusted friends, family, or oversight from institutions, was the main factor allowing these crimes to happen. The study also highlighted that perpetrators actively worked to create this isolation. They limited the victims’ social interactions to increase dependence before taking advantage of them.

Scale and Nature of the Problem:

Before examining specific fraud techniques, it is important to establish why the architecture of the Unified Payments Interface (UPI) itself creates a structurally suitable target for fraud. Understanding these design characteristics provides the foundation for explaining how offenders exploit the system. The simple and easy-to-use nature of UPI makes it an appealing option for online banking: the user’s phone number is linked to their bank accounts, debit cards, and credit cards through a platform, which allows for instant transactions at no cost. Singh and Agarwal (2025) state that while this attribute provides convenience, it also means that a single phone number becomes the single point of compromise that gives the attacker access to all of the user’s financial information simultaneously.

Sharma, Matharu and Sinha (2026) surveyed 183 UPI users in the Shimla district. As the study was confined to a single district and employed a relatively small sample, its findings have limited generalisability to the wider Indian population. The study found that although 96.2% of respondents reported being aware of UPI fraud, 52.5% had nevertheless become victims. The authors concluded that awareness alone may not translate into safe behaviour, highlighting behavioural factors such as overconfidence and susceptibility to social engineering. By contrast, Singh and Agarwal (2025) surveyed 280 respondents in Varanasi, excluded 30 inconsistent responses, and analysed 250 valid cases. Using the Friedman ranking test, the study identified awareness (mean rank = 6.72) and education (mean rank = 6.42) as the most effective countermeasures against UPI fraud. The two studies are not necessarily incompatible. Rather, when taken together, they suggest that awareness and education are seen as the best defences, but they do not always prevent fraud in practice. The difference in district, sample size, and research design may also contribute to the contrast.

In a comparative study of India, Sweden, and the United States of America, it was found that losses due to card fraud in India grew at an average annual rate of 12.19% between 2011 and 2016 (Iyer, 2017). Sweden, however, despite having the highest percentage of card transactions among the countries studied, demonstrated the lowest average annual fraud growth rate of 4.83%, which is nearly a third of the growth rate in India (Iyer, 2017). According to the study, India’s higher growth rate could be a result of card payment systems being introduced relatively recently, when compared to Sweden and the USA.

Furthermore, Sweden’s digital payment ecosystem functions within a complex regulatory framework managed by Finansinspektionen, the Swedish Financial Supervisory Authority. This agency sets security rules, checks compliance, and inspects payment platforms. Digital payment providers must follow the Payment Services Act, anti-money laundering laws, counter-terrorist financing regulations, the General Data Protection Regulation, and the Payment Card Industry Data Security Standard. They are also required to implement strong authentication, encryption, tokenization, and transaction monitoring systems as necessary security measures (Abraham et al., 2023).

According to Mukhopadhyay and Mukhopadhyay (2024), despite becoming victims, a large proportion of people were unwilling to leave UPI as a means of payment. Only 7.4% of fraud victims switched entirely to cash, 8.8% reduced their use of digital payments, and 51.5% reported no reduction in UPI usage. The Pearson correlation coefficient between fraud victimisation and reduced UPI usage is -0.408, a moderate negative relationship that confirms victims are largely unwilling to exit the platform. Holtfreter et al. (2008) demonstrated that a 1-unit increase in routine consumer activity, such as purchasing something online, translates into a 61% increase in the odds of being targeted for fraud. The data indicates that since victims themselves are substantially reluctant to use different means of payment, people will almost always be exposed to and within the environment in which motivated offenders operate.

It is difficult to determine the true scale of the problem due to issues regarding under-reporting. Mukhopadhyay and Mukhopadhyay (2024) found that 20.6% of respondents who had experienced UPI fraud did not report the incident to the concerned authorities. This finding is based on a survey of 70 valid respondents conducted over a five-day period (20–25 April 2023). Sirajutheen and Abirami (2026) found that, during the “Golden Hour” (the critical period immediately following a cyber fraud during which prompt reporting maximizes the likelihood of freezing fraudulent transactions and recovering funds) of a fraud incident, 40% of fraud victims would incorrectly call their bank’s IVR (Interactive Voice Response) helpline instead of the National Cyber Crime Helpline (1930). Meanwhile, post-fraud grievance redressal mechanisms prove to be largely ineffective, with 33% of people finding them unsatisfactory and only 6% describing the bank’s response as very satisfactory (Sirajutheen and Abirami, 2026). While under-reporting is an issue in and of itself, the low recovery rate of assets further highlights the shortcomings of guardianship. According to Sharma and Singh (2024), ₹2294 crore ($274.14 million as per the 2024 exchange rates) was lost in cyber fraud, but only ₹57 crore ($6.81 million as per the 2024 exchange rates) was successfully recovered, showing a recovery rate of 2.5%. The low recovery rate indicates that the existing institutional mechanisms for recovering funds after UPI fraud are objectively ineffective. Such limited recovery could reduce the effectiveness of formal reporting systems and suggests that current guardianship mechanisms provide insufficient protection to victims.

How Fraudsters Operate and Why People Become Victims

One of the most prominent modalities of UPI fraud is social engineering (the psychological manipulation of the target’s behavior). Sharma and Singh (2024) present social engineering attacks in which fraudsters impersonate trusted authorities and create a sense of urgency regarding issues such as KYC compliance, blocked accounts, or failed transactions. Similarly, Wadkar et al. (2025) studied two UPI fraud cases involving educated and technologically aware victims who were deceived by messages warning of imminent electricity disconnection and directing them to malicious links. These studies demonstrate that digital payment fraud takes advantage of psychological pressure and trust rather than being a result of mere technological ignorance.

Other prevalent techniques include phishing and vishing (voice phishing), fake QR code scams, fraudulent payment request schemes in which fraudsters send payment requests disguised as payment receipts, and SIM swap attacks that redirect incoming OTPs to the fraudster’s device (Sharma and Singh, 2024; Singh and Agarwal, 2025). Sirajutheen and Abirami (2026) report that while awareness of phishing (67%) and vishing (70%) is relatively high among UPI users, awareness of more complex schemes (such as mule account operations, in which victims unknowingly launder money by renting their bank accounts to fraudsters) is relatively low, with only 6% of respondents being familiar with the term.

Singh and Agarwal (2025) point out that the platform connects a debit card, credit card, and bank account to one phone number. This means that one successful manipulation gives the offender access to several financial tools at once. Sharma and Singh (2024) also mention that digitization has not only widened the market for fraud but has also made criminals more skilled. Offenders keep adapting their techniques to keep up with the changes in the platforms they target.

Survey data from Sharma, Matharu, and Sinha (2026), drawn from 183 UPI users in Shimla district, indicates that 68.3% of fraud victims are graduates or postgraduates, challenging the assumption that higher educational attainment inherently protects individuals from digital financial fraud. As established above, 96.2% of the same respondents reported being aware of UPI fraud, yet 52.5% had nevertheless experienced victimization (Sharma et al., 2026). Singh and Agarwal (2025), meanwhile, identified fraud awareness as the highest-ranked preventive measure among surveyed stakeholders, assigning it a mean rank of 6.72. Taken together, these findings expose a structural gap: awareness and education are the most widely recommended individual-level defences against digital payment fraud, yet their presence has failed to prevent victimization at scale. Sirajutheen and Abirami (2026) conducted a correlation analysis that revealed a strong positive relationship between the frequency of digital usage and victimization (r = .825). This finding aligns with the Routine Activity Theory, which suggests that increased routine exposure creates greater opportunities for motivated offenders. Holtfreter et al. (2008), in a study of 922 adults, established a crucial distinction between the determinants of fraud targeting and the determinants of fraud victimization. Their logistic regression models demonstrated that routine consumer activity predicted whether individuals were selected as targets, producing a 61% increase in targeting odds per unit increase in activity. Individual traits, however, did not affect targeting at all.

Holtfreter et al. (2008) established that while RAT explains targeting, individual factors determine whether a targeted person becomes a victim. In the UPI context, fraudsters systematically exploit cognitive biases.

Several cognitive and social biases contribute to vulnerability to fraud. Overconfidence bias leads users to underestimate fraud risk and disregard security protocols and updates. FOMO (Fear of Missing Out) is exploited through fabricated urgency, such as claims of limited offers or imminent account suspension, to push users toward hasty decisions. Reciprocity bias, along with sympathy exploitation, is described as common in romance and distress scams. Confirmation bias, anchoring, authority bias, and groupthink are identified as social influences that interact with these factors to further increase susceptibility to manipulation and fraud (Sharma, Matharu, & Sinha, 2026).[1]  This reveals that exploiting the architecture of human decision-making has a vital role in achieving UPI fraud.

The comparison with DeLiema’s study is instructive here. In elder fraud, cognitive vulnerability arose from neurological decline, which is a form of diminished capacity that reduced the victim’s ability to evaluate threats accurately (DeLiema, 2018). In UPI fraud, cognitive vulnerability is manufactured when fraudsters design scenarios that activate the biases listed above, effectively inducing a temporary incapacity in otherwise competent individuals. The complex nature of the attack is what converts the targeted individual into a victim, regardless of education and awareness.

Victims and non-victims attribute fraud causation differently. Victims emphasize financial illiteracy and technological complexity, while non-victims emphasize cybersecurity vulnerabilities and slow investigation. This divergence suggests that victimization is associated with a shift in how individuals understand the threat (Singh and Agarwal, 2025).

Systematic Failure

The institutional response to UPI fraud victimization has largely been a refusal to accept responsibility. Wadkar et al. (2025) document that when two fraud victims reached out to a pseudonymous Bank ABC, the bank maintained that the transactions were authorized because the victims had shared their UPI PIN or one-time password (OTP). This was the case regardless of whether the victims did so under deceptive circumstances, leading the bank to conclude that the transactions were non-refundable. The RBI’s own framework reinforces this position, specifying that if a transaction occurs as a result of the customer sharing their credentials, the loss rests with the customer until the bank is formally notified (Wadkar et al., 2025).

Sharma and Singh (2024) state that social engineering attacks trick customers into disclosing information by impersonating legitimate authorities and manufacturing crises. However, the victim does not share credentials negligently. Rather, they are deceived into doing so. Yet the legal framework treats engineered deception and careless disclosure as equivalent, and holds the victim equally liable in both cases.

One of the most prominent legal gaps in India’s UPI consumer protection framework is the APP (Authorized Push Payment) loophole. According to Agrawal (2026), as per RBI[1] ’s 2017 Consumer Protection Circular, banks are obligated to cover losses from unauthorized transactions if reported within three days. APP fraud, however, operates by deceiving the victim into authorizing the transfer themselves, so the transaction is technically classified as authorized and falls entirely outside the protection framework.

The failure to provide an appropriate institutional response is evident in the recovery statistics. In 2022, the recovery rate was determined to be 2.5% (Sharma and Singh, 2024). This statistic highlights a significant failure of systemic guardianship. Furthermore, it was revealed that only 6% of respondents rated their bank’s grievance redressal process as “very satisfactory,” while 33% found it “unsatisfactory.” These perceptions align with the recovery outcomes, indicating that the current system is not effectively providing reliable protection  Sirajutheen and Abirami, 2026, p.301).

The 20.6% non-reporting rate among UPI fraud victims is not clear evidence of a lack of action. The authors suggest two possible explanations for non-reporting: that the grievance redressal mechanism may be difficult for victims to access, or that victims may lack sufficient awareness of how and where to report UPI fraud (Mukhopadhyay and Mukhopadhyay, 2024).

Furthermore, in elder fraud cases, capable guardians (financial institutions, social services, family members) consistently failed to identify victimization until losses had already occurred, at which point intervention was too late (DeLiema, 2018). When we look at UPI, the transactions are much more rapid, and the window for intervention is narrower. Therefore, the convenience of quick transactions also brings forth the issue of timely prevention by guardians.

Recommendations

The most significant reform that Indian policymakers can make is to close the APP liability loophole. ​​The United Kingdom introduced a mandatory reimbursement of up to £85,000 (approximately $113,000 as per the 2026 exchange rates) for APP fraud, requiring payment service providers to compensate eligible victims and strengthening incentives for fraud prevention (Agrarwal, 2026). A comparable reform in the Indian context would require banks and payment platforms to compensate victims of social engineering fraud, removing the current presumption that authorized transfer equals voluntary consent.

Legal frameworks governing digital payments must evolve continuously to match the pace of technological change (Ballaji, 2024). Amending the Information Technology Act (2000) and the Payment Settlement Systems Act (2007) is recommended to address evolving forms of digital payment fraud, support a fraud management framework, and implement AI for fraud detection (Banerjee et. al, 2025).

It has been established that a majority of users aware of digital payment fraud have become victims (Sharma, Matharu, and Sinha, 2026). Therefore, policymakers should design interventions that reduce the probability of a person falling victim to fraud at the moment of the transaction. An example of such could be a form of QR authentication before the approval of a transaction.

Furthermore, since 40% of users were unaware of the National Cybercrime Helpline (1930) during a fraud incident (Sirajutheen and Abirami, 2026), a simple display of the helpline in any UPI platform (such as Google Pay) would allow for timely action and early intervention.

Additionally, while many studies support using AI in fraud detection, there is little guidance on how to implement these systems effectively. Future research should focus on building and testing practical AI-based fraud detection frameworks that can be integrated into digital payment systems.

Several methodological limitations in the current literature also need to be addressed. Future research should use nationally representative samples rather than regional ones, as this would allow the findings to be generalized across India. Furthermore, future work should test which specific cognitive biases result in relatively greater victimization, so that prevention efforts can be prioritized. Finally, comparative research across a wider set of countries would help identify which regulatory and infrastructural conditions most effectively curb fraud growth.

Conclusion

This paper explains why India’s digital payment revolution driven by UPI has produced a growing class of fraud victims. The observations suggest that UPI fraud stems from structural factors that create opportunities for fraud, rather than solely from individual negligence. UPI’s architecture, in which a single phone number links a user’s bank accounts, debit cards, and credit cards, creates a single point of vulnerability, while its convenience ensures routine digital activity that places users in the realm of motivated offenders.

Routine activity governs who is targeted, but awareness and education alone do not determine who becomes a victim. Victimization depends on whether fraudsters succeed in manufacturing the cognitive conditions, such as urgency, authority, and overconfidence, that override otherwise informed judgment.

One of the most critical issues is the absence of effective guardianship. Recovery rates for stolen funds remain minimal, and redressal mechanisms are widely regarded as unsatisfactory. Additionally, the Authorized Push Payment loophole allows banks to treat engineered deception and careless disclosure as equivalent, leaving a large category of victims without reimbursement.

Therefore, closing the gap between UPI’s convenience and its vulnerability will require structural changes, not merely individual reforms. Tackling the APP loophole, updating the legal frameworks governing digital payments, addressing methodological limitations, and introducing safeguards at the point of transaction are of utmost importance. As UPI continues to expand as the backbone of India’s digital economy, the strength of its guardianship mechanisms is likely to determine whether this expansion will occur without a proportionate rise in victimization.

References

1. Abraham, Eva, et al. “Exploring the Regulatory Demands and Evolution of Payment Security Regulations for Digital Payment Platforms in Sweden.” Law and Economy, vol. 2, no. 10, 2023, pp. 18–26, https://doi.org/10.56397/LE.2023.10.03.
2. Agrawal, Sristy. “Cyber Fraud in Digital Payments: A Global Study on UPI Scams and Online Financial Crimes.” Indian Journal of Law and Legal Research, vol. 8, no. 2, 2026, pp. 6991–7014.
3. Association of Certified Fraud Examiners. “Fraud 101: What Is Fraud?” ACFE, https://www.acfe.com/fraud-resources/fraud-101-what-is-fraud. Accessed 4 July 2026.
4. Ballaji, Nima. “Consumer Protection in the Era of Digital Payments: Legal Challenges and Solutions.” Beijing Law Review, vol. 15, 2024, pp. 1268–1290, https://doi.org/10.4236/blr.2024.153076.
5. Banerjee, Adhip, et al. “Digital Payments in India: Exploring Emerging Issues in the Legal and Regulatory Framework.” Krytyka Prawa, vol. 17, no. 1, 2025, pp. 78–96, https://doi.org/10.7206/kp.2080-1084.753.
6. Cohen, Lawrence E., and Marcus Felson. “Social Change and Crime Rate Trends: A Routine Activity Approach.” American Sociological Review, vol. 44, no. 4, 1979, pp. 588–608. JSTOR, https://doi.org/10.2307/2094589.
7. DeLiema, Marguerite. “Elder Fraud and Financial Exploitation: Application of Routine Activity Theory.” The Gerontologist, vol. 58, no. 4, 2018, pp. 706–718, https://doi.org/10.1093/geront/gnw258.
8. Holtfreter, Kristy, et al. “Low Self-Control, Routine Activities, and Fraud Victimization.” Criminology, vol. 46, no. 1, 2008, pp. 189–220.
9. Iyer, Swarnalakshmi M. A Case Study on Monetary Fraud in a Cashless Economy. 2017. Purdue University, Master’s thesis. Purdue e-Pubs, https://docs.lib.purdue.edu/open_access_theses/1289.
10. Ministry of Finance, Government of India. Coordinated Efforts of Government, RBI and NPCI Accelerate Growth in Digital Payments. Press Information Bureau, 2025, Press Release ID 2240723, https://www.pib.gov.in/PressReleasePage.aspx?PRID=2240723&reg=3&lang=1.
11. Ministry of Finance, Government of India. UPI Completes 10 Glorious Years, Emerges as World’s Largest Real-Time Payments Platform, Anchoring India’s Digital Economy. Press Information Bureau, 2026, Press Release ID 2257087, https://www.pib.gov.in/PressReleasePage.aspx?PRID=2257087&reg=3&lang=2.
12. Mukhopadhyay, Nibedita, and Mahadeb Mukhopadhyay. “UPI Frauds: A Study on UPI Usage, Awareness and Impact in India.” International Journal of Research in Commerce and Management Studies, vol. 6, no. 6, 2024, pp. 179–197, https://doi.org/10.38193/IJRCMS.2024.6616.
13. Sharma, Ashish, and Yogender Singh. “Cyber Frauds in India’s Digital Payment Ecosystem: Risk, Impacts, and Regulatory Responses.” Educational Administration: Theory and Practice, vol. 30, no. 5, 2024, pp. 15326–15332, https://doi.org/10.53555/kuey.v30i5.8951.
14. Sharma, Reetika, et al. “Socio-demographic and Behavioral Determinants of UPI Fraud Vulnerability: A Descriptive Study from Shimla District, Himachal Pradesh.” Journal of Forensic Science and Research, vol. 10, no. 1, 2026, pp. 9–14, https://doi.org/10.29328/journal.jfsr.1001110.
15. Singh, Amit Kumar, and Krishna Kumar Agarwal. “An Overview of Digital Payment Frauds: Causes, Consequences, and Countermeasures.” Journal of Informatics Education and Research, vol. 5, no. 1, 2025, pp. 2297–2311.
16. Sirajutheen, T., and V. Abirami. “A Study on Analysis of Financial Fraud in the Indian Banking Sector.” International Advanced Research Journal in Science, Engineering and Technology, vol. 13, no. 3, 2026, pp. 297–302, https://doi.org/10.17148/IARJSET.2026.13347.
17. Vinoth Kumar, S., and Dharshan S. “An Analytical Study of Financial Frauds in E-Payment Systems in India.” International Journal of Creative Research Thoughts, vol. 14, no. 3, 2026, pp. d326–d331.
18. Wadkar, Prashant, et al. “Cybersecurity Challenges in Digital Payments: A UPI Fraud Case Study from India.” International Journal of Science, Architecture, Technology, and Environment, vol. 2, no. 10, 2025, pp. 227–232, https://doi.org/10.63680/ijsate1025030.024.

About the Author

Aadhavan Singh

Aadhavan is currently a Grade 12 student at La Martiniere College, India. He is a national-level shooter and passionate trekker.

The post Exploring How India’s Digital Payment Revolution Created a New Class of Fraud Victims: An Analysis of UPI Scams appeared first on Exploratio Journal.

Introduction

“87% of neighborhoods in San Francisco undergoing gentrification were once redlined as hazardous” (“From Redlining to Gentrification: The Policy of the Past that Affects Health Outcomes Today”). Redlining was a discriminatory practice in which banks and government agencies denied services such as insurance and mortgages to residents of neighborhoods with large African American and other minority populations. Over time, this practice caused significant disinvestment and economic deterioration of neighborhoods it affected, leaving many people in a fixed place of poverty that prevented upward socioeconomic mobility.

Established in 1933 as a part of Franklin Delano Roosevelt’s “New Deal” program to lift the country out of economic depression, the Home Owners’ Loan Corporation (HOLC) provided mortgage relief to homeowners at risk of losing their homes through foreclosure, in order to stabilize the housing market. As a part of this process, the HOLC created numerous residential security maps, grading different neighborhoods A-D (A being a highly secure and desirable neighborhood & D being a hazardous neighborhood) in about 200 different U.S. cities. Though aiming to create stability in the housing market, these letter grades were primarily based upon the socioeconomic, racial, and ethnic makeup of the neighborhoods’ residents, which facilitated the development of discrimination in the U.S. housing market. With the Johnson administration’s passage of the Fair Housing Act in 1968, any means of housing discrimination on the basis of race, sex, familial status, nationality, and disability were prohibited, thereby marking the closure of the HOLC’s rampant redlining practices of the mid-20th century. However, despite the practice’s prohibition, redlining had already caused a large amount of socioeconomic damage on the communities it had impacted, through a ripple effect of discrimination prevalent in many of society’s pillar institutions (ie schooling, employment, healthy food access)

In this paper I aim to answer the following question: What are the long-term economic impacts of 1930s-era HOLC redlining in the San Francisco Bay Area? By focusing on the quantitative relationships between redlining tract coverage and census data of racial demographics, household income, educational access, and employment, this paper aims to quantify the extent of redlining’s association with economic disparity throughout the late 20th and early 21st centuries through distributional impacts, categorical comparisons, and regression analysis.

From conducting distributional, categorical, and regression analyses, there are several patterns that can be concluded about the long-term economic impacts of New Deal-era HOLC redlining in the Bay Area. Firstly, the distributions holistically showed minimal change in shape over time indicating that the economic indicators measured stayed consistent over the course of several decades. This consistency goes to show that the economic impacts caused by redlining are solid and aren’t weak enough to change over time. Secondly, the regression analysis showed a negative relationship between income and redlining tract coverage with an increasing slope magnitude over time and a statistically significant relationship between the two variables indicated by the very low p-values.  

Section 1: Distributional Impacts

To examine the persistent long-term economic impacts of redlining across different HOLC-graded neighborhoods (graded on desirability), I first generated histograms of median household income (1980-2000), unemployment (1980-2000), and white-occupied housing (1980-2020) across HOLC grades A-D, which reveal several long term distributional economic trends.

Histograms were chosen because they allow one to see the persistence or change of these economic patterns over time by noticing a persistence of change in the spread and shape of the census data across HOLC grades. The incorporation of histogram series by time period with the four different grades per serie allows one to see variation within each grade and how the socioeconomic indicator of one grade changed over time relative to another. Each histogram plots the relative frequency of a given census variable across HOLC tracts A-D as defined by the 1930s HOLC redlining maps. By using relative frequency histograms rather than raw count histograms, the distributions are normalized which allows for fair and accurate comparison across grades that have a different number of census tracts. To ensure accurate comparability over time, the relative frequency histogram that plots median household income (1980-200) across HOLC grades A-D were converted from nominal US dollars to real 2020 US dollars using the Bureau of Labor Statistics’ annual average CPI values.

Subsection 1A: Median Household Income

The first socioeconomic indicator is real median household income. This variable measures the typical earning level of all households in a neighborhood, which is a strong indicator of overall economic opportunity in that neighborhood. Because nominal incomes change with inflation, the median household incomes were converted from nominal to real 2020 U.S. dollars using annual average CPI values from the U.S. Bureau of Labor Statistics. 

The first observation that can be drawn from Figure 1.1 is that the shape of the distributions of median household income for each HOLC grade remains approximately the same consistently across all three dates of measure. This indicates that aggregate household income for each grade persisted over the period of measure.

The second observation that can be drawn from Figure 1.1 has to do with the differences in intergrade clustering and spread of the data. Grade A’s distribution has a slight right-skew with a moderately-large income range that persists across the three dates of measure. Moving from the Grade A distribution to the Grade D distribution, the data gradually shifts left with each grade closer to D. Grade D’s distribution shows one, a high level of clustering at the left side of the left side of the histogram and two, an income range that is significantly smaller than that of Grade A’s distribution. These two observations continue throughout the three dates of measure.

From these observations, it can be concluded that Grade A Bay Area neighborhoods have a larger range of median household income values with fewer observations of lower median household income values. This conclusion is in accordance with the inference that those in a grade with greater security and desirability tend to have higher job opportunity, educational access, and healthcare access. The subtle right-skew in Grade A’s distribution indicates that a smaller proportion of tracts have exceptionally high incomes which increases the mean and pulls the distribution’s tail rightward. The gradual leftward shift of the distributions from Grade A to Grade D indicates that as the magnitude of risk assessed by the HOLC for the tracts is inversely proportional to the median household income for the tracts. Moving to Grade D, it can be concluded that tracts judged to have lower security and desirability by the HOLC have a smaller range of median household income values and a high proportion of observations of lower median household income. This conclusion is in agreement with the inference that individuals in this grade tend to have lesser job opportunity, lesser educational access, and lesser healthcare access, traits which undermine ability to achieve high household income.

In addition, the moderately-large range of income denoted in Grade A’s distribution hints at a more discrete but key principle of 1930s HOLC redlining: racial prejudice. As observed in Grade D’s distribution, there was a visible clustering of tracts toward the left of the histogram around the lower income values. Intuitively, this logic should apply to Grade A as well but in an opposite fashion: clustering of tracts toward the right of the histogram around the higher income values. However, in Grade A’s distribution there is a significantly-sized range and spread of income, much larger than that of Grade D’s distribution, indicating that Grade A neighborhoods contained households with a variety of median household income values. From this distributional observation, it can be understood that HOLC redlining grading was deeply influenced by racial prejudice, more so than objective economic data for a lot of the time.

Subsection 1B: Unemployment

The second socioeconomic indicator is percent of civilians unemployed (16+). This variable measures the share of the working-class population per grade that are actively seeking employment but cannot find work. Redlined neighborhoods often see disinvestment and lower amounts of socioeconomic mobility so distributional unemployment is a strong indicator of the severity of redlining experienced by grade.

Similar to the distribution of median household income across HOLC grades, the first observation drawn from Figure 1.2 is that consistently across all three dates of measure, the shape of the distribution for each respective HOLC grade remains approximately the same, with only a few minor shifts, which indicates that unemployment shares for each grade persisted over time. This persistence in unemployment suggests that the socioeconomic division caused by 1930s HOLC redlining has stayed relatively intact in the latter half of the 20th century, with census level unemployment shares continuing to mirror structural inequalities imposed by HOLC redlining maps.

The second observation drawn from Figure 1.2 is concerned with the differences in intergrade clustering and spread of the data as well. Generally, these differences are the same as in Figure 1.1 but a mirror flip. Grade A’s distribution shows high left-clustering, a small right-skew, and a very high proportion of observations on that side of the histogram. Moving from the Grade A distribution to the Grade D distribution, the data gradually gains overall spread/range and magnitude of right-skew while also losing peak height. Grade D’s distribution shows a range and median greater than that of Grade A. This pattern of inter-grade shifting from A to D remains approximately the same over the period of measure.

From these observations, it can generally be concluded that the magnitude of risk assessed by the HOLC for the tracts is directly proportional to the percent of civilians (16+) unemployed for the tracts. The association between these two variables is consistent with the inference that redlined neighborhoods tend to have lower economic opportunity, in this case job access, resulting in a higher unemployment rate in these tracts. While the median percent of civilians (16+) unemployed increases moving from Grade A to Grade D, the magnitude of right skew also increases which means that there is a higher degree of variability in tracts deemed less secure and desirable by the HOLC. Economically, this increase in variability in percent of civilians (16+) unemployed from Grade A to Grade D means that the magnitude of risk assessed by the HOLC for the tracts is associated with a higher degree of economic volatility in the tracts. These tracts that have experienced a high amount of redlining have also seen uneven patterns of disinvestment, reinvestment, demographic/population change, gentrification, industrial restructuring, factors which are greatly responsible for a high variation in unemployment in these tracts. In addition to an increase in magnitude of right skew from Grade A to Grade D, a decrease in peak height is also observed. This means the magnitude of risk assessed by the HOLC for the tracts is inversely proportional to the proportion of observations made. Similar to an increase in spread, a decrease in peak height also denotes a proportional relationship between degree of redlining and degree of economic volatility observed. From a statistical standpoint, the decrease in peak height, or flattening, of the distribution means that a fewer proportion of tracts share a common unemployment rate for civilians (16+). This flattening represents a growing internal heterogeneity suggesting that redlining produced a fragmented and nonuniform economic landscape in the areas it greatly impacted.

Subsection 1.C: White Share

The third socioeconomic indicator is percent of white-occupied houses. This variable measures a specific racial composition of occupied housing units. Historically, redlined neighborhoods in the mid 20th century saw a large amount of urban flight in which primarily caucasian residents departed urban areas for suburban neighborhoods. As a result of this exodus, over time, redlined neighborhoods saw a change in white residents so by measuring this variable, we will see how the magnitude and direction of this change across grades and the fluctuations in change over time. 

Similar to the previous two distributions, the first observation that can be drawn from Figure 1.3 is that across the five dates of measure, the shape of the distribution for HOLC grades A, B, and C remain approximately the same, with only a few minor shifts. This indicates that for these grades, the distribution of percent of white occupied homes persisted over the course of 4 decades. Likewise to the other socioeconomic indicators measured in Figures 1.1 and 1.2 respectively, the persistence of shares of white occupied homes suggests that the census-level outcomes of the racial bias  ingrained into HOLC redlining maps has seen very minimal change in the highly to moderately secure and desirable tracts (A-C), indicating that housing segregation on the basis of race has persisted over time in the Bay Area.

The second set of observations that can be drawn from Figure 1.3 concerns the shape of each grade’s distribution over the period of measure: what the shape means and for Grade D, what a shift in the distribution’s shape means. Firstly, moving left to right from Grade A to Grade D, there is a gradual flattening of the distribution, most significantly moving from Grade A to Grade B, and there is a shift from a right-skew to a left-skew. Moving vertically down the figure from 1980 to 2020, the distributions of Grades B, C, and D see a subtle and gradual increase in aggregate peak height over time. In Grade D, there is a sharp increase in proportion of observations over time at the 7-20% range of the 2020 distribution.

From these observations, it can be concluded that magnitude of risk assessed by the HOLC for the tracts is inversely proportional to the percent of white-occupied houses for the tracts over time. As the HOLC grade declines in security and desirability, the distribution becomes increasingly concentrated toward lower percentage values of white-occupied houses. This inverse association, shown by the shift of a right-skew to a left-skew from Grade A to Grade D in each date of measure, confirms our understanding of HOLC redlining having a basis of racial prejudice, as lower-tier grades and areas with higher levels of redlining tend to have a greater minority demographic and a lower majority, in this case caucasian, demographic. The flattening of the distributions in the earlier dates of measure (1980 and 1990 primarily) indicate an increase in demographic variability as tracts become more “hazardous.” While secure and desirable areas remain racially homogenous with a high proportion of caucasians, areas with lower security and desirability are more racially heterogenous reflecting both racial segregation’s influence in HOLC redlining as well as subsequent population shifts driven by disinvestment and suburbanization. The increase in overall peak height for Grades B, C, and D but retainment of a roughly flat shape relative to Grade A’s distribution indicate that there is a greater amount of census data concerning white-occupied housing moving forward in time, but the demographic trends revealed by this data remain persistent as mentioned in detail above. The sharp increase in peak height in 2020 at the 7-20% range of the Grade D histogram indicates that there is a significantly higher proportion of observations concentrated around the lower percent of white-occupied houses in recent times, a trend that can be explained by “white-flight” and suburbanization. This migratory occurrence is defined as the migration of primarily caucasian residents from urban to suburban areas, which results in the economic degradation of urban areas through decreased funding and overall maintenance.

Section 2: Categorical / Group-Level Comparisons

To further explore the economic impacts of 1930s HOLC redlining in the Bay Area, I used categorical plots to visualize redlining’s impact through median household income and educational attainment across HOLC grades A-D.

Definitionally, a categorical plot allows for comparison of a single numerical variable across four levels of categorical variables, in this case, each level corresponding to a HOLC grade A-D. The important distinction that needs to be made in order to understand what the differences are in the displaying of categorical variables (HOLC grades) is the difference in interpretation over time for the histograms vs. the categorical plots. The histogram focuses on distributional change within a singular HOLC grade over time, which provides insight into intra-grade socioeconomic change over the period of measure. On the other hand, the categorical plot focuses on the change in inter-grade variation over time for a given socioeconomic indicator.  Due to this difference, I chose to incorporate categorical plots to supplement the distributions in order to visualize comparative disparities between HOLC grades.

Furthermore, in connecting the two models, a single bar within a categorical plot can be interpreted like a distribution for the HOLC grade that the bar represents, through the spread of data points within the bar.

Subsection 2A: Median Household Income

The first socioeconomic indicator to be analyzed categorically is median household income (real 2020 USD) across HOLC grades A-D. In the figure that assesses this variable, there are three categorical plots, one plot for each period of measure, 1980, 1990, and 2000 respectively.

The first observation that can be drawn from Figure 2.1 is the clear downward trend from Grade A to D that is consistent across all three dates of measure, with Grade A having the highest peak and Grade D having the lowest peak. The second observation that can be drawn from Figure 2.1 concerns intra-grade variability indicated by the distribution of data points within each HOLC Grade’s bar. Consistently across all three dates of measure, Grade A tracts show the greatest intra-grade variability, indicated by a greater spread of data points within the bar. As the HOLC grade declines however, there is a decrease in intra-grade variability, indicated by the increasing clustering of data points within each bar. The third observation that can be drawn from Figure 2.1 is an increase in the peak height for each HOLC Grade’s bar over time, while the vertical separation between each bar remains approximately equivalent over time. 

From the first observation, it can be roughly concluded that magnitude of risk and insecurity determined by the HOLC and the median household income are inversely proportional for the tracts. This observation is consistent with the basis of redlining, in which redlined neighborhoods saw high levels of disinvestment and economic deterioration, preventing upward socioeconomic mobility for residents of these tracts, therefore explaining why residents of redlined tracts tended to have lower median incomes than those of non-redlined tracts. From the second observation concerning intra-grade variability, the higher levels of variability in tracts deemed less risky and more secure by the HOLC indicate that in these tracts, there is a greater degree of economic opportunity and mobility, providing an explanation for a wider spread in Grade A and a tighter clustering in lower grades, where many residents of redlined tracts are stuck in a lower socioeconomic position and unable to advance upwards. From the third observation concerning the increase in relative heights of each HOLC Grade’s bar over time, we can conclude that across all tracts, redlined or not, general development over the period of measure through technological innovation and globalization promoted the median household incomes for each HOLC Grade. Despite this overall increase, the vertical separation for each catplot between each HOLC Grade remains approximately equivalent suggesting that relative disparities between grades persisted despite overall income growth over the period of measure.

Subsection 2B: Educational Attainment

The second socioeconomic indicator to be analyzed categorically is educational attainment across HOLC grades A-D. This variable will be measured by the percent of adults 25 and older with 4 or more years of college education. In the figure below, there are three categorical plots, one plot for each period of measure, 1980, 1990, and 2000 respectively.

The first observation that can be drawn from Figure 2.2 is the clear downward trend from Grade A to Grade D across all three dates of measure, with Grade A having the highest peak and Grade A having the lowest peak. The second observation that can be drawn is the gradual upward shift in each grade over time, shown by the increasing peak height for each HOLC Grade’s bar. In connecting these two observations, we see that the difference in peak heights between the HOLC Grades decreases over the period of measure, with the bars for Grades B-D increasing by a greater amount than the bar for Grade A, revealing a flattening of the downward trend from Grade A to D.

From the first observation, it can be concluded that magnitude of risk and insecurity determined by the HOLC is inversely proportional to the percent of adults 25 years of age and older with 4 or more years of college education for the tracts. This conclusion aligns with the economics of redlining in which neighborhoods that experienced the negative implications of the practice saw less economic opportunity as well as overall disinvestment in their communities. Tracking back to Figure 2.1, we see that neighborhoods in lower tier HOLC grades have generally a lower aggregate median household income. Due to the aggregate financial status of these neighborhoods, we can infer that one of the reasons that households in redlined neighborhoods saw lower degrees of educational attainment was they comprehensively had less disposable income to spend on privileges like textbooks, tutoring services, or in this case, a college education.

From the second observation, it can be concluded that the moderate to lower tier grades saw upward educational mobility over the period of measure. This pattern can be primarily attributed to neighborhood redevelopment and gentrification throughout the Bay Area, particularly in tracts close in proximity to university centers, like UC Berkeley, or booming industries that have attracted more college-educated residents in search of work on a domestic and international level, like Silicon Valley and the South Bay. Coinciding with these inferences, the Bay Area saw several urban redevelopment projects in parts of Oakland, Emeryville, and San Francisco’s Mission District in which communities experienced infrastructural improvement in housing and schools. In a research report titled Engaging Schools in Urban Revitalization: The Y-PLAN (Youth – Plan, Learn, Act, Now!), authors Deborah L. McKoy and Jeffrey M. Vincent define the Y-PLAN initiative in the context of the urban Bay Area: 

“West Oakland, California, is an industrial area suffering the abandonment and blight common to other neighborhoods after the loss of manufacturing employers, a process that began in the 1950s…Stepping into this environment in 2000 was the Y-PLAN (Youth—Plan, Learn, Act, Now!), a model for youth civic engagement in city planning that uses urban space slated for redevelopment as a catalyst for community revitalization and education reform. Sponsored by the Center for Cities & Schools at the University of California (UC), Berkeley, Y-PLAN facilitates positive community outcomes by partnering graduate student “mentors,” local high school students, government agencies, private interests, and other community parties to work on a real-world planning issue. The Y-PLAN is both a pedagogical tool and a planning studio that addresses specific issues in local communities” (McKoy & Vincent, 2007, p. 1).

Due to the implementation of programs such as the Y-PLAN that educationally mobilized the Bay Area’s youth in impoverished areas, many of these neighborhood areas saw educational improvement reflected by the increasing peak heights for tracts in HOLC Grades B-D in the categorical plot. As a result of the flattening of the downward trend, it can be concluded that the educational disparities within the Bay Area were ameliorated over the period of measure, in part due to youth initiatives and other programs seeking educational improvement.

Section 3: Regression Analysis of Raw Median Household Income

To determine the impact of redlining tract coverage on median household income, I also conducted ordinary least squares (OLS) linear regressions for both the raw and logarithm of median household income as a function of redlining tract coverage from a 1930s HOLC map. While the histograms and catplots provide distributional insight and categorical comparison, respectively, into disparities across HOLC grades, the regression allows for high quantitative precision in determining the extent to which redlining coverage can accurately predict income for a Bay Area neighborhood. More specifically, the categorical plots allowed us to see a general slope trend based on the peaks of each bar, but the regression expands upon this observation by making it more specific numerically.

Because nominal incomes change with inflation, the median household incomes were converted from nominal to real 2020 U.S. dollars using annual average CPI values from the U.S. Bureau of Labor Statistics. The purpose of conducting a log-transformed regression was to linearize the relationship and allow for the slope coefficient to be interpreted as a percent change in income for a one-unit increase in tract coverage.

The two key regression outputs that will be analyzed in this section are slope coefficient and p-values, which together, can assess the magnitude and certainty, respectively, of the relationship between redlining tract coverage and median household income over time. For the raw income model, the slope coefficient is interpreted as the predicted change in real median household income by the OLS regression line per one-unit (or 100%) increase in redlining tract coverage. For the log-transformed income model, the slope coefficient is interpreted as the predicted percent change in real median household income by the OLS regression line per 1.0 (or 100%) increase in redlining tract coverage. Given that the distribution of income is skewed, taking the logarithm of income compresses high-income and low-income outliers, which creates a more symmetric and homoscedastic residual distribution. The patterns observed in the logarithm-transformed regression are largely identical to those observed in the untransformed regression, so I decided to include logarithm-transformed regression analysis in the first appendix of the paper. The p-values for each model (raw income and log-transformed income) represent the probability of observing each respective slope by random chance. A low p-value (less than the alpha level of 5%) indicates that the observed relationship is unlikely due to random chance.

Several neighborhood wealth-related trends can be concluded from OLS regression analysis of raw and log-transformed real median household income (2020 dollars) as a function of redlining tract coverage from 1980-2000. While quantitative, these patterns share similarities with the distributional trends derived from the histograms showing real median household income (2020 dollars) across HOLC grades A-D described in detail earlier in the paper.

Here, our parameter beta represents the true slope of the population regression line relating the explanatory and response variables of redlining tract coverage and raw income respectively. The null hypothesis is that beta is equal to zero and the alternative hypothesis is that beta is less than zero.

Figure 3.1 Raw Income as a function of Redlining Tract Coverage

Figure 3.1 confirms a negative association between income and redlining tract coverage based on the slope coefficients. Firstly, the slope coefficient for the 1980 OLS regression is -15,205.29. This means that for each additional one-unit (or 100%) increase in redlining tract coverage, the OLS regression line predicts about a $15,205.29 decrease in real median household income (2020 dollars) in 1980. Secondly, the slope coefficient for the 1990 OLS regression is -29,742.12. This means that for each additional one-unit (or 100%) increase in redlining tract coverage, the OLS regression line predicts about a $29,742.12 decrease in real median household income (2020 dollars) in 1990. Thirdly, the slope coefficient for the 2000 OLS regression is -46,819.90. This means that for each additional one-unit (or 100%) increase in redlining tract coverage, the OLS regression line predicts about a $46,819.90 decrease in real median household income (2020 dollars) in 2000.

While I noted that the association between income and redlining tract coverage remains negative, from the slope coefficients in Figure 3.1, it can also be observed that the magnitude of the decrease in income increases with each subsequent date of measure. From a graphical standpoint, this can be understood as an initially negative regression line that gets steeper and steeper in the negative direction over time. So what does this mean in the context of redlining? From this observation, it can be concluded that the negative impact of redlining coverage on household income worsens overtime, as the same increase in coverage is met with a greater magnitude of decrease in income over the period of measure. This trend can be best explained by the extensive impact of disinvestment in higher redlined areas. Following the HOLC’s classification of neighborhoods as “hazardous” via their redlining maps, banks and financial institutions chose not to lend money to businesses and individuals, insure mortgages, or fund development in these areas due to the high-risk attributed to these areas by the HOLC. As a result of this lack of financial and infrastructural support from banking institutions, these areas deteriorated over time through an aggregate decrease in property values as well as less socioeconomic mobility and in this case, access to high-paying jobs. Due to the extensive decline of the economic makeup of these areas caused by disinvestment, it makes sense why the slope coefficient of the income vs. redlining tract coverage regression decrease over time.

Figure 3.1 also displays very low p-values that all fall significantly below the general alpha level of 5%. Therefore, we can ascertain a relationship between redlining tract coverage and income. Firstly, the p-value for the 1980 OLS regression with raw income as the dependent variable is  . Assuming that there’s no association between redlining tract coverage and real median household income in 1980, the probability of obtaining a sample of this size and observing a negative relationship between redlining tract coverage and real median household income (2020 dollars) as or more extreme than a magnitude of approximately $15,205.29 per percentage point of redlining tract coverage by random chance is less than approximately   (rounded 0.0002%). Therefore, the OLS regression model provides significant statistical evidence that tracts with higher redlining coverage are associated with lower real median household income in 1980 for this sample. Secondly, the p-value for the 1990 OLS regression with raw income as the dependent variable is  . Assuming that there’s no association between redlining tract coverage and real median household income in 1990, the probability of obtaining a sample of this size and observing a negative relationship between redlining tract coverage and real median household income (2020 dollars) as or more extreme than a magnitude of approximately $29,742.12 per percentage point of redlining tract coverage by random chance is less than   (rounded 0.0008%). Therefore, the OLS regression model provides significant statistical evidence that tracts with higher redlining coverage are associated with lower real median household income 1990. Thirdly, the p-value for the 2000 OLS regression with raw income as the dependent variable is  . Assuming that there’s no association between redlining tract coverage and real median household income in 2000, the probability of obtaining a sample of this size and observing a negative relationship between redlining tract coverage and real median household income (2020 dollars) as or more extreme than a magnitude of approximately $46,819.90 per percentage point of redlining tract coverage by random chance is less than   (rounded 0.0002%). Therefore, the OLS regression model provides significant statistical evidence that tracts with higher redlining coverage are associated with lower real median household income 2000.

The consistently low p-values across all three dates of measure (less than the alpha level of 5%) confirms the high statistical strength between redlining tract coverage as an independent variable and median household income as a dependent variable. Furthermore, this aspect of the regression corroborates the trends observed from the slope coefficients of the decreasing negative association between these two variables, indicating that it is extremely unlikely this relationship occurred by random chance.

Conclusion

From distributional impacts, to categorical plots, to regression analysis, we can conclude several long term economic impacts that vary by extent to which a given census tract was redlined and how it changed over time. With the histograms, we saw that a lot of the economic patterns surrounding income and unemployment maintained shape for a given grade distribution over time indicating that a lot of the economic impacts stayed stagnant over the course of a few decades. With the regression analysis, we saw that the magnitude of negative slope for raw income as a function of redlining tract coverage greatly decreased in the 20 year period from around $30,000 to $42,000 in 1980 and 2000 respectively. With the log transformed regression, the slope coefficients were approximately equivalent for each year indicating that income dropped by the same percentage each year which shows relative stability in the influence of redlining, not due to external factors.

Looking forward, I would love to dive deeper into how gentrification interacts with these socioeconomic patterns of redlining. By definition, gentrification is the process in which a poorer area is infrastructurally improved as a result of a wealthier demographic of people moving in. This economic improvement is seen over time through improved housing, healthcare, and new business. By exploring gentrification in the context of redlining, I would ask whether modern reinvestment in previously redlined areas resulted in economic growth, ameliorating the negative effects of 1930s-era HOLC redlining.

While in this paper, I primarily studied solely socioeconomic indicators measuring the effects of redlining (unemployment, income, racial demographics, etc.), I would be curious to explore more health-related effects of this same era of redlining. By conducting research on food deserts, urban areas with limited access to good-quality fresh food, through measuring, for example, the number of fast food restaurants in a given census tract as well as illnesses through measuring, for example, the number of diabetes occurrences or hospital visits in a given census tract, I would bring in a new dimension of human health and biology in association with the redlining I explored in this paper. 

Bibliography

De los Santos, H., Jiang, K., Bernardi, J., & Okechukwu, C. (2021, May 26). From redlining to gentrification: The policy of the past that affects health outcomes today. Harvard Medical Journal. Retrieved October 28, 2025, from https://info.primarycare.hms.harvard.edu/perspectives/articles/redlining-gentrification-health-outcomes

Jonathan Schroeder, David Van Riper, Steven Manson, Katherine Knowles, Tracy Kugler, Finn Roberts, and Steven Ruggles. IPUMS National Historical Geographic Information System: Version 20.0 [dataset]. Minneapolis, MN: IPUMS. 2025. http://doi.org/10.18128/D050.V20.0

McKoy, D. L., & Vincent, J. M. (2007, June). Engaging schools in urban revitalization: The y-PLAN. Association of Collegiate Schools of Planning. https://doi.org/10.1177/0739456×06298817

Nelson, R. K., Winling, L, et al. (2023). Mapping Inequality: Redlining in New Deal America. Digital Scholarship Lab. https://dsl.richmond.edu/panorama/redlininghttps://dsl.richmond.edu/panorama/redlining.

Appendix A: Regression Analysis of the Logarithm of Median Household Income

While the raw-income regression shows the dollar change in income as a function of redlining tract coverage, by taking the logarithm of income and performing the same regression function, we see the percent change in income as a function of redlining tract coverage for the same three dates of measure. Though not visible from the regression outputs, the log-transformed regression mitigates the effect of outliers and influential points that undermined the strength of the relationship between raw-income and redlining tract coverage, providing us slope coefficients and p-values that are optimal for ensuring an accurate relationship between the two variables.

Here, our parameter beta represents the true slope of the population regression line relating the explanatory and response variables of redlining tract coverage and logarithm of income respectively. The null hypothesis is that beta is equal to zero and the alternative hypothesis is that beta is less than zero.

Figure A1 Logarithm of Income as a function of Redlining Tract Coverage

Similar to the raw income regression, the results of the log-transformed regression shown in Figure A1 confirm a negative association between redlining tract coverage and income as well. The slope coefficient for the 1980 OLS regression is -0.30. This means that for each additional one-unit (or 100%) increase in redlining tract coverage, the OLS regression line predicts about a 30% decrease in real median household income (2020 dollars) in 1980. The slope coefficient for the 1990 OLS regression is -0.28. This means that for each additional one-unit (or 100%) increase in redlining tract coverage, the OLS regression line predicts about a 28% decrease in real median household income (2020 dollars) in 1980. The slope coefficient for the 2000 OLS regression is -0.32. This means that for each additional one-unit (or 100%) increase in redlining tract coverage, the OLS regression line predicts about a 32% decrease in real median household income (2020 dollars) in 1980.

Similar to the raw income regression, the results of the log-transformed income regression evidenced in Figure A1 show very low p-values. Therefore, we can ascertain a relationship between redlining tract coverage and income. The p-value for the 1980 OLS regression is  . Assuming that there is no relationship between redlining tract coverage and the logarithm of real median household income in 1980, the probability of obtaining a sample of this size and observing a linear relationship between redlining tract coverage and the logarithm of real median household income (2020 dollars) with a slope coefficient of -0.30 or less by random chance alone is less than  %. Therefore, the OLS regression model provides significant statistical evidence that tracts with higher redlining coverage are associated with lower proportional levels of real median household income in 1980. The p-value for the 1990 OLS regression is  . Assuming that there is no relationship between redlining tract coverage and the logarithm of real median household income in 1990, the probability of obtaining a sample of this size and observing a linear relationship between redlining tract coverage and the logarithm of real median household income (2020 dollars) with a slope coefficient of -0.28 or less by random chance alone is less than  . Therefore, the OLS regression model provides significant statistical evidence that tracts with higher redlining coverage are associated with lower proportional levels of real median household income in 1990. Lastly, the p-value for the 2000 OLS regression is  . Assuming that there is no relationship between redlining tract coverage and the logarithm of real median household income in 2000, the probability of obtaining a sample of this size and observing a linear relationship between redlining tract coverage and the logarithm of real median household income (2020 dollars) with a slope coefficient of -0.32 or less by random chance alone is less than  . Therefore, the OLS regression model provides significant statistical evidence that tracts with higher redlining coverage are associated with lower proportional levels of real median household income in 2000.

For the most part, these two outputs of the regression reveal largely similar or identical patterns as the raw-income regression. Firstly, the slope coefficients reveal a negative relationship between redlining tract coverage and median household income that steepens over time. This steepening means that the percent decrease in median household income per 100% increase in redlining tract coverage increases with each subsequent date of measure.

Again, the p-values for this regression are far below the alpha level of 5% meaning that a relationship between redlining tract coverage and the logarithm of income is statistically significant across all three dates of measure, and not due to random chance.

About the author

Rishi Haldar

Rishi is a senior at Miramonte High School with interests in economics, mathematics, statistics, and history. He plans to attend Cornell University in the fall, where he will be studying economics in the College of Arts and Sciences. Apart from academics, Rishi is a guitarist in a band that plays local gigs (restaurants, fundraisers, etc.) and plays soccer for a club team and his high school team.

The post Redlining: Quantifying the Economic History of the San Francisco Bay Area appeared first on Exploratio Journal.

Abstract

This paper explores the complex relationship between tariffs on Gross Domestic Product (GDP) and the U.S. trade balances with its major trading partners. It investigates if imports and greater trade balances changed between the U.S and its top trading partners after tariffs were placed. The conclusion is no significant change in trade balances since the implementation of the Trump administration’s tariffs.

The evidence shows that in the months of April to July, the tariffs have not significantly changed U.S. net trade. The effects studied in this paper are a result of new trade policies by the Trump Administration which put retaliatory tariffs on most of the world. As a result, many firms and businesses frontloaded the tariffs which caused a 40% increase in imports in Q1 2025. Also in Q1, real GDP decreased by 0.5% (U.S. Bureau of Economic Analysis, 2025) according to the most recent estimate. However, imports do not reduce GDP and are only included in the calculation to support accounting principles. As a result of this misconception, many news articles written by journalists who are not economists have had misleading claims with regards to the GDP decrease. The current results of the findings could potentially be attributed to the uncertainty in the administration’s tariff policy or simply that not enough time has passed for significant changes to be observable in the data.

Introduction & Literature Review

The role of imports in shaping a nation’s economy has become increasingly significant following President Trump’s trade wars and tariffs on countries across the world. Many of these tariffs were put during the first quarter of 2025 while the others were placed on April 2, 2025, also known as Liberation Day. However, news of President Trump’s intention of using tariffs has been clear before his Inauguration and use of tariffs on foreign countries was common in his first term as well. Since his election, many businesses and firms have increased inventories and the amount of imported goods in anticipation of high tariff rates to go into effect soon.

As noted above, the heart of GDP measurement is the widely cited expenditure formula: GDP = C + I + G + (X-M) where C denotes consumption, I investment, G government expenditures, X exports, and M imports. The superficial glance at this equation shows imports as a direct drag on GDP. However, economists consistently clarify that this superficial glance is quite misleading as the negative sign simply represents an accounting principle to prevent double counting.

Bill Conerly (2025), a writer at Forbes, clarifies that “U.S. imports are neither added nor subtracted conceptually” (para. 3) for GDP. He explains that with perfect data available to statisticians, imports wouldn’t be included in a GDP calculation (Conerly, 2025).

Looking at the subtraction, Greg Mankiw notes, “this subtraction is made because imports of goods and services are included in other components of GDP,” (Mankiw, 2001, p. 499) Mankiw also notes how a purchase of an imported good raises consumption, investment or government expenditures.

The St. Louis Fed adds that “imports (foreign production) should have no impact on GDP,”(Wolla, 2018, para. 9). They explain the variable M as an accounting variable rather than an expenditure variable. It is also important to note that the imported goods will have an effect on the GDP of the country that produces them. Since it isn’t the United States, they don’t affect U.S. GDP. However, it can take into account if the goods are intermediate or partially produced in the U.S. Since the expenditure variables of C, I, and G only take final goods into account, GDP will be affected based on the amount of the goods that was domestically produced.

Keshav Srikant, a writer with Econofact, supports the net-zero effect on imports on GDP. However, he also notes how imports can potentially indirectly reduce GDP if they replace domestic consumption or if domestic government expenditures are reduced as a result of higher purchases of foreign goods (Srikant 2025). Further study of macroeconomic trends are required to make an argument for this situation as these latent variables could drive import growth and GDP declines when those two variables are not correlated.

Ultimately, imports do not directly reduce GDP and their inclusion in the components of GDP is a measure to prevent double counting.

There are many researchers who have explored the growth of imports and its, relationship with the overall economy. In fact, many specific case studies have found that an increase in imports often leads to an increase in real GDP .

A study by Peter Saunders, focused on a time series analysis of the role of imports in the economic rise of India from 1970 to 2005, analyzes the long term relationship between imports and India’s real GDP. Saunders establishes that both variables, imports and real GDP, are cointegrated using Johansen’s test of cointegration (Saunders 2010). This test proves if two variables have a long term equilibrium relationship, meaning that despite short term deviations or outliers, the variables have a long term observed relationship. A Vector Error Correction Model (VECM) examines the relationship between cointegrated variables. In the VECM used by Saunders, the results indicated that imports have positively impacted India’s economic growth in the short-term. Saunders highlights how this result defies traditional expectations that imports could be a drag on the economy (Saunders 2010).

In another study by M.Y Khan et al, about the relationship between imports and economic growth in Pakistan, a similar conclusion was reached. This study used data from 1975 to 2014 with the methodology of a Granger Causality Test. This test focuses on proving directional relationships between time series variables. The results showed that there was a bi-directional relationship between imports and economic growth in Pakistan, meaning that both time series variables mutually supported one another (Khan et al, 2019).

Research focusing on the relationship Rwandan economic growth with imports and exports showed a positive long run relationship (Al Hemzawi & Umutoni 2021) . The study concluded that a one percent increase in imports led to a 0.32% rise in Rwandan GDP. To get that correlation, the authors used a multivariate Ordinary Least Squares regression which is a way to minimize variance between variables. They also used quarterly time series data in the regression.

Immediately after the Q1 GDP contraction was announced, many news publications released misleading or false articles regarding the cause behind this result. They accredited the cause to be the tariff jumping effect and the dramatic import surge that occurred because firms and businesses rushed to purchase foreign goods before tariff prices were assigned to goods. The underlying demand was quite consistent to previous levels while business investment surged as an offset to the imports. Despite the fact that imports do not directly reduce GDP, news outlets continued to push that narrative.

For example, an AP News article stated “First-quarter growth was weighed down by a surge of imports, ” (Wiseman 2025, para. 2) while The Hill said “GDP shrank in the first quarter mostly because of lower consumer spending and a pull-forward in imports ahead of President Trump’s tariffs, ” (Burns 2025,para. 4). Many other outlets made misleading claims regarding the import surge. Although journalistic misconceptions are not uncommon, even the Federal Open Market Committee has made mistakes with regards to the effect of imports on GDP (Lemieux 2018).

On the other hand, many top economists have had different opinions. Many economists have attributed the contraction to the economic activity as a result of the imports, not by the imports directly. For example, Paul Gruenwald, a global chief economist for S&P Global Ratings, mentioned that Q1 GDP data was “distorted by the front-running of tariffs,” (2025). Gregory Daco, a chief economist at EY , added “the contraction was largely a function of economic activity being pulled forward as importers, business, and consumers rushed to get ahead of tariff implementation,” (2025). Economist Preston Caldwell ofMorningstar added that imported goods could be stored in inventories but “it just didn’t show up in the data because of measurement error,” (2025).

Some top economists also challenged the fear that this GDP result was the first domino in a potential recession. Caldwell added that this result “doesn’t mark the beginning of a recession,” (2025). Others mentioned potential for economic uncertainty further down the line as more policy was unveiled. “Demand in the first quarter looks to be driven by businesses battening down the hatches before the storm,” Chief economist Luke Tiley of the Wilmington Trust said (2025).

One potential explanation for the GDP decrease is a phenomenon called the substitution effect, a phenomenon that suggests that the tariff induced frontloading substituted for domestic purchases. If this is the case, GDP would decrease since less money would be spent toward domestic production. This has been prevalent in the past as well.

In the 1990s, the Northern American Free Trade Agreement (NAFTA) contemplated potential tariff reductions. 96% such reductions were announced far in advance, giving consumers and firms the chance to act on this information ( Khan & Khederlarian 2021). A study found that in anticipation of an upcoming tariff reduction of 1%, imports dropped by a sizable 6% in the months before the tariff implementation when compared to regular months. The study used an Herfindahl-Hirschman Index, a method to measure market concentration, and applied it to the spread of imports. Their final result articulated that firms shift their purchases to periods when lower costs can be attainable and that these anticipatory dynamics are true (Khan & Khederlarian 2021).

A potential alternate explanation is that the small decrease in government expenditures was the key factor in the GDP decrease.

There are some potential gaps in data which limit the study of GDP accounting. For example, these accounting principles say nothing about potential causality with latent variables or economic impacts. There is also difficulty in GDP data collections since it can be difficult to only count final goods. Finally, GDP data could be fixed-weighted calculations that can add error as the economy changes and price structures evolve. However, when calculated in a chain-weighted approach to account for economic evolution, there are still struggles with new goods being added.

Methodology

This analysis uses monthly trade data and monthly effective tariff rates for the United States with its largest trading partners. It also uses the same data for the European Union to use as a control. The source for monthly trade data values were the U.S. Census Bureau and Eurostat. The effective tariff rate values were gathered from trusted sources and reflect prior US tariffs and changes as newer tariffs went into effect. This analysis employs a linear regression test with net trade balances and effective tariff rates to analyze the potential correlation between the two.

Results

The data shows a minimal negative relationship between tariff rate and change in trade balances. This means that since the tariffs went into effect, there hasn’t been a significant increase or decrease in U.S. trade balances with main trading partners. The correlation coefficient was 0.0806 which confirms that in the three months since the tariffs went into effect, there weren’t any significant changes in trade balances that were caused by the changes in effective tariff rates. The coefficient of determination is 0.0065, or approximately 0, which meant that any changes that did occur in trade balances were not from the changes in effective tariff rates. Finally, the t-score value equals 0.0977 and indicates that the observed result aligns with the null hypothesis and the difference between the sample data and the population data is not statistically significant.

Discussion

Ultimately, the analysis proves that there is no correlation between the vast increases in effective tariff rates and changes in net trade balances. By contrast, as shown in Table 2, the U.S. trade deficits actually became larger for some countries such as Vietnam or the United Kingdom despite increases in effective tariff rates. Other countries such as China or Italy faced similar decreases in trade deficits despite having large differences in net trade and change in effective tariff rates. Furthermore, some countries were levied with larger tariffs than others, making predicting the change in trade balances harder to anticipate.

A potential explanation for the results of the analysis is the extreme volatility in tariff policy during the study period. Following the implementation of the “Liberation Day” reciprocal tariffs in early April, several countries experienced rapid and significant changes in their tariff rates. For example, China briefly faced tariff levels exceeding 140%, while Brazil was subjected to a 50% tariff following political disputes with the U.S. administration. In addition to these enacted measures, frequent public threats of new tariffs introduced further uncertainty into global trade markets. Simultaneously, reports of partial or full trade agreements with major partners like the EU, China, Japan, and South Korea, led to subsequent reductions in effective tariff rates. This pattern of escalation followed by negotiated de-escalation likely diluted the measurable macroeconomic impact of tariffs, complicating attempts to identify stable relationships between tariff levels and trade or GDP outcomes.

Another potential reason can be shown through the pressures of the markets. Financial Times commentator Robert Armstrong coined the current administration’s trade policies as “TACO Trade”. The acronym refers to some of the administration’s sudden reversals of tariffs. Armstrong coined the term when describing the pattern of placing large tariffs on countries which led to economic panic, shock, and stock market hits. He then explained how later reversals of these tariff policies have led to market comebacks. Additionally, the market uncertainty can be explained as how stocks look like they are trending upward and then stop due to a social media post or claim by the government. It’s possible that many firms and businesses believed that the tariff rate changes wouldn’t be in place long term and thus, no changes were found in the U.S. trade balances.

Implications for Policy and Future Research

Investigating the nuanced economic effects on GDP is key for future policy regarding tariff measures and potential trade deals. As occurred in Q1, there are potential short term distortions in GDP measurement so it’s important to keep these in mind. An area for future research is on the study of tariffs-driven import behavior and with the substitution effect’s prominence in the short and long term. This would provide key insights into how firms react to the government policy and how both parties can better facilitate economic policy.Finally, it’s important to continue to analyze changes in trade balances to see if significant changes will be present with the passing of time and more recent data.

Works Cited

Al Hemzawi, B., & Umutoni, N. (2021). Impact of Exports and Imports on the Economic Growth. MSc. Thesis, Jönköping University. Buckling up for a long ride: chief economists add detail to a downbeat outlook. (2025, May 28). World Economic Forum. https://www.weforum.org/stories/2025/05/wef-chief-economists-uncertainty-global-outlook

Burns, T. (2025, June 26). US economy shrank faster than expected, new data shows. The Hill. https://thehill.com/business/5371005-us-gdp-revised-lower-consumer-spending

Conerly, B. (2025, March 11). Understanding GDP: Why Imports Don’t Actually Reduce Economic Growth. Forbes.https://www.forbes.com/sites/billconerly/2025/03/11/understanding-gdp-why-imports-dont-actually-reduce-economic-growth/

Daco, G. (2025, May). LinkedIn. https://www.linkedin.com/posts/gregorydaco_inflation-fed-fomc-activity-7323335677599793152-c–l/

Freund, C., Pierola, M. D., & Rocha, N. (2021). How Does Trade Respond to Anticipated Tariff Changes? Evidence from NAFTA (Policy Research Working Paper No. 9561). World Bank. Gross Domestic Product, 1st Quarter 2025 (Third Estimate) | U.S. (2025, June 25). Bureau of Economic Analysis. https://www.bea.gov/news/2025/gross-domestic-product-1st-quarter-2025-third-estimate-gdp-industry-and-corporate-profits

Khan, M. Y ., Akhtar, S., & Riaz, S. (2019). Dynamic Relationship Between Imports and Economic Growth in Pakistan. Journal of Economics and Sustainable Development, 10(10), 70–77.

Lemieux, P. (2018, September 6). The St. Louis Fed on Imports and GDP. Econlib. https://www.econlib.org/imports-as-a-drag-on-the-economy/

Mankiw, N. G. (2001). Principles of Economics. Harcourt College Publishers.

Saunders, P.J. (2010). A Time Series Analysis of the Role of Imports in India’s Phenomenal Economic Growth. Indian Journal of Economics and Business, 91, 101-109.

Schonberger, J. (2025, April 30). Shrinking GDP and elevated inflation put Fed in tough spot. Yahoo Finance. https://finance.yahoo.com/news/shrinking-gdp-and-elevated-inflation-put-fed-in-tough-spot-142211609.html

Sekara, D., Dzuibinski, S., & Caldwell, P. (2025, July 16). Morningstar’s Q3 2025 US Market Outlook: Has the Storm Passed, or Are We in the Eye of a Hurricane? Morningstar. https://www.morningstar.com/markets/morningstars-q3-2025-us-market-outlook-has-storm-passed-or-are-we-eye-hurricane

Srikant, K. (2025, May 14). Fact Check: Does an increase in imports directly reduce GDP? Econofact.https://econofact.org/factbrief/fact-check-does-an-increase-in-imports-directly-reduce-gdp

Wiseman, P., & Rugaber, C. (2025, April 29). U.S. economy shrinks 0.3% in first quarter as Trump tradewars disrupt businesses. AP News.https://www.ap.org/news-highlights/spotlights/2025/u-s-economy-shrinks-0-3-in-first-quarter-as-trump-trade-wars-disrupt-businesses/

Wolla, S. A. (2018, September 4). How Do Imports Affect GDP? | St. Louis Fed . Federal Reserve Bank of St. Louis. https://www.stlouisfed.org/publications/page-one-economics/2018/09/04/how-do-imports-affect-gdp

About the author

Ishaan Bafna

Ishaan Bafna is a 12th grade student at Kingswood Oxford School with strong academic and research interests in economics and mathematics. Ishaan actively pursues opportunities that integrates analytical thinking with critical reasoning and problem-solving. Known for his intellectual curiosity and work ethic, Ishaan wishes to pursue a career at the intersection of economics, mathematics and technology.

Outside of the classroom, Ishaan is a leader of his schools Math Team and Mock Trial Team, a lead peer tutor, and a varsity golf athlete. Ishaan has interned with The Hartford Insurance as a Lean Portfolio Management Intern. He is also a National Merit Commended Scholar and a recipient of various awards at his school.

The post The macroeconomic effects of tariffs on GDP and trade balances, through the lens of Q1 2025 GDP change appeared first on Exploratio Journal.

This paper examines the stability of five leading stablecoins USDT, USDC, BUSD, TUSD, and DAI using a nonlinear machine learning model combined with an event based analysis of major depegging episodes. Fiat backed stablecoins show muted and short lived deviations from their pegs during external shocks, reflecting liquid reserves, arbitrage and institutional support, and often trade at small premiums. By contrast, the crypto collateralized DAI comoves strongly with systemic risk, embedding mark to market leverage, on chain frictions and liquidation dynamics that mirror contagion effects in the banking literature. Our approach validates and extends recent work on stablecoin fragility and shows how design choices translate into distinct patterns of resilience or vulnerability under stress, with implications for regulation and digital asset market structure.

1. Introduction

When Silicon Valley Bank collapsed in March 2023 it sent a shockwave through the stablecoin market. The news that Circle held part of USDC’s reserves at the failed bank drove its price down to about $0.87. DAI, which is backed by crypto collateral, also slipped below its peg. This episode, together with earlier events such as the 2018 USDT reserve rumor discount and the 2020 Black Thursday crisis in DAI, highlights a fundamental divide in how different types of stablecoins behave under stress.

Fiat backed stablecoins such as USDC, USDT, BUSD and TUSD mainly face redemption bottlenecks during moments of panic. Because their backing sits in cash or liquid assets, arbitrage and institutional support usually close the gap quickly, and these coins often trade at a small premium rather than a discount during calm periods. By contrast, crypto collateralized coins such as DAI embed mark to market leverage, liquidation risk and on chain frictions directly into their design. When the underlying collateral becomes volatile or gas fees spike, liquidations cascade, arbitrage slows, and prices can swing both below and above the peg. This reflects the panic-driven withdrawals and cascading effects described in Diamond and Dybvig’s model of bank runs.

In Section 1, the analysis introduces the contrasting behavior of fiat-backed and crypto-collateralized stablecoins under stress, using the USDC–SVB banking shock, the 2018 USDT reserve-rumor episode, and Black Thursday (2020) to illustrate why collateral design and on-chain frictions matter. Section 2 reviews the existing literature on stablecoin stability, systemic risk transmission, and nonlinear modeling, drawing on the work of Lyons and Viswanath-Natraj, Grobys et al., and Klages-Mundt et al. Section 3 outlines the data and methodology, combining a Gaussian Ridge Neural Network estimation of daily stablecoin prices against four macro-financial risk indexes with a structured event analysis of major depegging episodes between 2018 and 2024. Section 4 reports the main results, showing that fiat-backed stablecoins exhibit low SSE and weak correlations with systemic risk, whereas DAI shows high correlation and mixed-sign coefficients. Model predictions are compared with real-world events to demonstrate that macro shocks affect fiat-backed coins briefly, while on-chain shocks cause deeper, asymmetric deviations in DAI. Finally, Section 5 discusses the implications for stablecoin design, financial stability, and the regulation of crypto-dollar instruments.

2. Literature Review

  Stablecoins resemble fixed exchange rate regimes because they promise convertibility at par, yet their credibility depends on collateral, redemption, and confidence. Lyons and Viswanath-Natraj (2023) show that fiat backed designs such as USDT and USDC remain close to par through arbitrage and redemption and often trade at small premiums. Grobys et al. (2021) document that crypto collateralized tokens such as DAI display nonlinear dependence on systemic risk indexes. These findings echo Diamond and Dybvig (1983), where stability is sustainable in good states but fragile when coordination failures and run dynamics emerge.

  A second group of studies focuses on how stablecoin designs embed different risk channels. Klages-Mundt et al. (2020) classify stablecoins into fiat backed, crypto collateralized, and algorithmic types and show that risk profiles vary sharply across designs. Crypto collateralized coins encode mark to market leverage, on chain liquidation risk, and settlement frictions. Algorithmic designs attempt to engineer stability reflexively but can amplify feedback loops. The Terra Luna collapse in 2022 confirmed these theoretical warnings, while the USDC–SVB banking shock showed that even fiat backed coins can temporarily lose their peg. Liquidity concentration on venues such as Curve 3pool and Binance has also shown that market microstructure can transmit stress (Briola and coauthors, 2023).

  A third group of studies applies advances in financial econometrics and machine learning. Mallqui and Fernandes (2019) and Shen et al. (2020) show that radial basis function neural networks outperform linear benchmarks in predicting asset prices and volatility. Corbet et al. (2021) survey machine learning applications in crypto markets and find that neural and recurrent architectures can identify volatility patterns that GARCH style methods may miss.

  This paper extends the literature by applying a radial basis function neural network to five leading stablecoins (USDC, USDT, BUSD, TUSD, and DAI) and linking daily prices to four macro financial risk factors. By evaluating both the sum of squared errors and the correlation between predicted and observed series, the analysis captures predictive accuracy and structural co movement with systemic risk. Combining model-based results with event-based evidence from major depegging episodes shows that fiat backed coins mostly experience short lived redemption pressures, while crypto collateralized coins encode collateral volatility and on chain frictions.

 3. Data and Methodology

This study employs a dual-method approach to examine stablecoin stability. On the quantitative side, the analysis constructs and trains a Gaussian Ridge Neural Network (GRNN) to model nonlinear sensitivity of stablecoin prices to macro-financial risk indexes. On the qualitative side, structured event observation complements the quantitative modeling.

It is important to note that the sample ranges differ across stablecoins, reflecting their varied launch dates. As a result, correlation estimates are not strictly apples-to-apples. For example, USDT and DAI have longer and more volatile histories than newer entrants such as BUSD and TUSD. This difference in data coverage should be taken into account when interpreting the comparative strength of correlations across stablecoins.

A. Quantitative framework

The quantitative analysis uses a Gaussian Ridge Neural Network (GRNN) to link stablecoin price with systemic financial risk indexes. Gaussian ridge functions capture the nonlinear behavior typical of stablecoins, which remain close to their peg under normal conditions but deviate sharply during systemic or crypto-specific stress. This design enables the model to detect nonlinear fragility that linear regressions fail to capture. The model uses daily values of four macro-financial indexes—NFCIRISK, KCFSI, STLFSI4, and the 10-Year Expected Inflation Risk Premium—from the Federal Reserve’s FRED database and use them as the feature vector (the model’s inputs).

The model estimates relationships between daily stablecoin prices to four macro-financial risk factors. Let

Each stablecoin (USDT, USDC, DAI, BUSD, and TUSD) is estimated independently using identical macro financial inputs and two hidden nodes. The model optimizes weights, centers, spreads, and biases by minimizing the sum of squared errors (SSE) between observed and predicted prices.

Step 1: Input Layer to Hidden Nodes (Linear Stage). For each hidden node j ∈ {1, 2}, the model computes a separate weighted sum of the four macro- financial risk factors plus a bias term:

Step 2: Hidden Node Activation (Nonlinear Stage). ach hidden node transforms its input through a Gaussian ridge function:

Step 3: Output Layer (Linear Stage). The hidden node activations are combined to generate the predicted price (or deviation from par) of the stablecoin:

where β₀is the output bias and v₁, v₂ are weights from hidden nodes to the output node.

Step 4: Error. The model’s error term for each observation is the difference between the predicted and actual price:

where y_t is the observed stablecoin price.

Step 5: Model Fit. The model measures overall fit using the Sum of Squared Errors (SSE):

and by the Pearson correlation coefficient between actual and predicted prices:

As shown in Figure 1, the network links four macro-financial risk factors to two hidden nodes and then to an output node representing the predicted stablecoin price or deviation from par.

B. Event Observation Framework

The analysis compiles a structured dataset of major stablecoin depegging episodes between 2018 and 2024 to complement the model-based analysis. For each event, the event window and the date of maximum deviation, the affected stablecoin or coins, the lowest observed price recorded on CoinMarketCap during the episode, and a classification of the primary trigger are documented. The analysis codes triggers as either macro-financial such as external banking shocks, market stress, or regulatory actions or on-chain/DeFi, including protocol-specific failures, liquidity imbalances, or infrastructure stress. This classification enables us to distinguish between stress transmitted through traditional financial channels and stress that originates within digital-asset markets.

The analysis draws events from multiple sources including industry reports, regulatory filings, market data providers such as DeFiLlama and Kaiko, and commentary from central banks. The framework organizes each episode into six thematic drivers of fragility: market stress and panic events, regulatory and legal drivers, blockchain and infrastructure dependence, issuer behavior and transparency, liquidity concentration and market microstructure, and adoption or utility shocks.

  This event-based framework captures dimensions of fragility, confidence, governance, and infrastructure bottlenecks that lie outside the scope of purely statistical modeling. Together with the GRNN estimation, it provides a more holistic view of stablecoin stability, linking sensitivity to systemic risk factors with the historical record of crises and structural vulnerabilities.

4.  Analysis Results

A. Quantitative Results: Stablecoin Sensitivity to Financial Risk Indexes (GRNN)

After minimizing SSE, the estimated input-to-hidden weights show a sharp contrast between fiat backed and crypto collateralized stablecoins. Fiat backed coinsload near one across systemic risk indexes, consistent with a muted and proportional response to macro conditions. By contrast, DAI exhibits large mixed-sign coefficients and a high bias term.

  Figure 2 plots SSE against correlation for the five stablecoins, illustrating how DAI diverges from the fiat backed group. Figure 3 shows the average absolute input-to-hidden weights by risk factor and stablecoin, highlighting the near-unit values of fiat coins and the much larger magnitudes of DAI.

The large mixed sign coefficients for DAI provide evidence that leverage and on chain frictions transmit macro shocks directly into the stablecoin’s peg. Fiat backed stablecoins work much like a currency board or a hard peg regime: they hold reserves in cash or short term government securities and can meet redemptions quickly, which limits how far prices move when stress hits. Crypto collateralized coins such as DAI are closer to a soft peg backed by volatile assets.

Because DAI’s collateral is marked to market on chain, any rise in systemic risk immediately cuts collateral values and pushes collateral ratios toward liquidation. This sets off margin calls, liquidations, and delays in arbitrage that make price swings larger and longer. On chain bottlenecks such as gas fee spikes or thin liquidity slow down adjustment further and create the kind of liquidity spirals seen in past financial crises. The large mixed sign coefficients estimated for DAI are not random noise but evidence that leverage and on chain frictions transmit macro shocks directly into the stablecoin’s peg.

B. Event Observations: Linking Model Predictions to Real World Stress Episodes

A structured dataset of major depegging episodes between 2018 and 2024 was compiled, recording for each episode the event window, affected stablecoins, lowest observed price on CoinMarketCap, and the primary trigger categorized as either macro-financial or on-chain/DeFi. Table 1 summarizes these events. Macro events such as the SVB banking shock and BUSD’s regulatory action caused temporary but pronounced deviations in fiat backed coins. On-chain events such as Black Thursday (2020) and Curve 3pool imbalances (2023) produced sharper and more asymmetric deviations in DAI and USDT. Comparing GRNN predictions with observed prices shows that the model captures macro-financial sensitivity but underestimates DeFi-specific shocks, consistent with its input structure based on four systemic risk indexes.

 TABLE 1—MAJOR STABLECOIN DEPEGGING EPISODES, 2018–2024

Figure 4, Figure 5 and Figure 6 compare DAI price behavior with Ethereum market conditions in March 2020.

 The first figure plots daily DAI deviations from one dollar together with the Ethereum average gas price. Deviations rise when gas fees are elevated, which suggests that network congestion makes it harder to execute arbitrage or liquidations that would normally stabilize the peg. The second figure contrasts DAI daily highs and lows with the Ethereum low price over the same dates. Around mid-March, when Ethereum volatility jumps, the DAI high low spread widens at the same time, which points to stress in collateral mechanics and liquidity. Importantly, DAI moved both above and below one dollar. It fell below par when confidence weakened after collateral auctions failed to clear, and it rose above par when liquidators and arbitrageurs needed DAI to repay vault debt, which created temporary scarcity. Taken together, the patterns indicate that DAI instability during stress reflects not only broader market shocks but also on chain frictions such as high gas costs and collateral volatility.

Figure 5. Ethereum Average Gas Price in March 2020

5. Conclusion

This paper analyzes the stability of leading stablecoins using a nonlinear machine learning model combined with a review of major depegging episodes. The results reveal a clear divide between fiat backed and crypto collateralized designs. Fiat backed stablecoins show muted and short lived deviations from their pegs during external shocks, reflecting liquid reserves, arbitrage and institutional support. In contrast, the crypto collateralized DAI moves more strongly with systemic risk, embedding mark to market leverage, on chain frictions and liquidation dynamics similar to run effects in traditional finance. Linking model based evidence to historical stress events validates and extends recent work on stablecoin fragility, showing that fiat backed coins behave like tightly managed exchange rate regimes, while crypto collateralized coins resemble leveraged intermediaries whose stability depends on collateral valuation, market infrastructure and the speed of on chain adjustments. Future research could integrate real time indicators of liquidity and collateral quality, examine the market microstructure of trading venues and bridges, and test policy or design interventions such as redemption limits or insurance funds to better assess stablecoin resilience under stress.

References

Briola, Riccardo, and coauthors. 2023. “Anatomy of a Stablecoin Run: Evidence from the Terra-Luna Collapse.” Finance Research Letters, 51: 1544–6123.

Corbet, Shaen, Brian Lucey, Larisa Yarovaya, et al. 2021. “Machine Learning in Cryptocurrency Markets: A Survey.” Finance Research Letters.

Diamond, Douglas W., and Philip H. Dybvig. 1983. “Bank Runs, Deposit Insurance, and Liquidity.” Journal of Political Economy, 91(3): 401–419.

Grobys, Klaus, et al. 2021. “Stablecoins and Systemic Risk: Nonlinear Dependence and Stress Episodes.” Finance Research Letters.

Klages-Mundt, Ariah, Dominik Harz, Lewis Gudgeon, Jun-You Liu, and Andreea Minca. 2020. “Stablecoins 2.0: Economic Foundations and Risk-based Models.” AFT ’20, 59–79. ACM.

Lyons, Richard K., and Ganesh Viswanath-Natraj. 2023. “What Keeps Stablecoins Stable?” Journal of International Money and Finance, 131: 102838.

Mallqui, Daniela C., and Ricardo A. Fernandes. 2019. “Predicting the Direction, Maximum, Minimum, and Close Values of Daily Bitcoin Price Using Machine Learning Techniques.” IEEE Access, 7: 148551–148563.

Shen, Dawei, et al. 2020. “Nonlinear and Deep Learning Approaches to Crypto Asset Volatility Forecasting.” Applied Soft Computing.

About the author

Abhiram Kode

Abhiram is a rising 11th-grade student at Rock Hill High School in Frisco, Texas, with strong academic and research interests at the intersection of finance, technology, and engineering. He is deeply passionate about investment banking, fintech, cryptocurrency markets, and applied robotics, and actively pursues opportunities that blend analytical thinking with real-world problem solving.

Beyond research, Abhiram tutors mathematics at Kumon, plays varsity tennis, and participates in competitive chess and speed cubing. He has earned multiple national and international awards in mathematics and science Olympiads. Known for his curiosity, discipline, and self-driven learning, Abhiram aspires to pursue a future career that combines finance, technology, and innovation.

The post Stablecoin Stability Under Stress appeared first on Exploratio Journal.