JFQA | Cryptocurrency Trading: Investment or Gambling?

Authors | Sudheer Chava, Fred Hu, Nikhil Paradkar

Source | JFQA

Compiled by | Yan Zilin

1. Introduction

Since the advent of Bitcoin in 2009, the cryptocurrency market has experienced explosive growth. During this period, thousands of crypto tokens—digital assets created on a blockchain (a decentralized distributed digital ledger)—have emerged. These tokens can represent various assets and utilities: well-known cases like Bitcoin and Ethereum primarily serve as mediums of exchange or stores of value, while other crypto tokens can be used to access specific products or services on blockchain platforms or represent ownership of physical and digital items. Alongside the market expansion, consumer interest has surged, with over 20% of American adults having invested in, traded, or used cryptocurrencies (CNBC (2022)), and the estimated number of global crypto investors reaching 580 million (Crypto.com (2024)).

Despite the significant growth of the retail crypto investor population, direct evidence regarding the characteristics of these investors is scarce due to the anonymous nature of blockchain. Meanwhile, the surge in crypto investors has raised concerns among policymakers, especially considering the extreme volatility of the crypto market. For example, the total market capitalization of cryptocurrencies reached nearly $2.8 trillion in November 2021, then fell to $1.2 trillion in June 2022, and rebounded to $2.6 trillion by May 2024 (Forbes (2024)). This dramatic volatility raises concerns that retail investors may not fully understand the associated risks. Specifically, the returns on crypto tokens exhibit a high positive skew, indicating a small probability of achieving extremely high returns (Liu and Tsyvinski (2021), Liu, Tsyvinski, and Wu (2022)). This return pattern resembles lottery products, making it highly attractive to investors with strong gambling preferences (Kumar (2009)). Therefore, this paper explores whether gambling preferences can predict retail interest in the crypto market. Understanding whether retail investors view crypto tokens as lottery-like products can help policymakers determine appropriate disclosure standards and regulatory frameworks (such as the legislative proposals put forth by Lummis and Gillibrand (2023)).

In the absence of direct data, this paper draws on the research of Da, Engelberg, and Gao (2011), using Google search interest as a proxy for retail attention, focusing on two prominent types of crypto tokens: Initial Coin Offerings (ICOs) and Non-Fungible Token (NFT) series. Unlike tokens that serve as general currencies, ICOs focus on project investment, while NFTs emphasize digital ownership and collectibles. Consistent with the view that gambling preferences predict interest in crypto, this paper finds that regions with higher per capita lottery sales exhibit significantly greater attention to crypto tokens. This finding is robust to other gambling-related demographic characteristics identified by Kumar (2009) and Kumar, Page, and Spalt (2011). To address concerns that "attention does not equal investment," this paper documents that attention to crypto wallets also surged before and after token issuance, and higher attention correlates with greater fundraising amounts and participant numbers. Additionally, this paper rules out other explanatory pathways such as advertising, risk preferences, or distrust in institutions.

This paper further investigates the token-level factors influencing gambling-driven attention. First, ICOs and NFT projects launched during the crypto market bubble attract more attention from regions with high gambling tendencies. Second, in the ICO market, tokens with lower opening prices (i.e., more "lottery-like" characteristics) and those lacking "Know Your Customer" (KYC) protocols (Li, Shin, and Wang (2021)) also generate greater interest in these regions. Moreover, this paper uses the gradual legalization of sports betting across U.S. states as a natural experiment and finds that when legal sports betting is permitted, regions with high gambling tendencies show a relative decline in attention to token issuances. This indicates that crypto tokens are largely viewed as substitutes for traditional gambling products by retail investors.

Finally, this paper examines the relationship between crypto attention and consumer credit outcomes. Using data from Equifax, this paper finds that in regions with high gambling tendencies, the default rates on consumer credit soar following periods of high crypto attention, particularly among financially constrained subprime borrowers. Lagged analyses show that the rise in attention precedes the increase in default rates.

This paper contributes to multiple literatures: first, it provides a new perspective on the characteristics and motivations of retail investors in the ICO market (Li and Mann (2025), Lee and Parlour (2021), Cong, Li, and Wang (2021, 2022), etc.); second, it enriches the NFT literature by revealing the relationship between retail attention and primary market performance (Kong and Lin (2021), Borri, Liu, and Tsyvinski (2022), Oh, Rosen, and Zhang (2023)); third, it extends the literature on how gambling preferences influence financial product prices and trading volumes (Barberis and Huang (2008), Bali, Cakici, and Whitelaw (2011), Kumar (2009), Green and Hwang (2012)); fourth, it connects research on retail investor behavior (Barber and Odean (2000, 2008), Welch (2022), Fedyk (2022), Barber et al. (2022)); finally, it adds to the emerging literature on crypto investor characteristics (Dhawan and Putniņš (2023), Hackethal et al. (2022), Kogan et al. (2024), Aiello et al. (2023), Divakaruni and Zimmerman (2024), Sun (2023)), demonstrating that gambling preferences are a significant predictor of retail interest in the crypto market.

2. Data and Descriptive Statistics

This section introduces the data sources used in the study and the descriptive statistics of the variables in the regression analysis. A. Data Sources

1. Retail Attention This paper adopts the method of Da et al. (2011), using online attention obtained from Google Trends as a proxy for investment behavior. Its advantage lies in capturing investors' search intentions in private settings. The study employs the Search Volume Index (SVI), which ranges from 0 to 100. Data is collected at a more granular designated market area (DMA) level, covering 209 DMAs in the U.S. For each token project, the area with the highest attention (SVI=100) represents the highest relative popularity of that project in that region.
2. Initial Coin Offerings (ICOs) ICOs are a method for blockchain startups to raise funds. Unlike IPOs, these tokens do not represent equity but typically represent some utility within the project's ecosystem. (1) Sample Selection: Integrated from ICOBench.io, projects that did not meet the fundraising soft cap and those that U.S. investors could not participate in were excluded. (2) Data Volume: A total of 937 completed ICOs from January 2016 to December 2018 were ultimately selected. (3) Contributor Identification: Wallet addresses were obtained from white papers, and the number of unique wallet addresses was tracked using Etherscan.io to infer the actual number of contributors.
3. Non-Fungible Tokens (NFTs) NFTs represent ownership of unique items (such as artworks) on the blockchain. (1) Sample Selection: Data is sourced from the largest trading platform, OpenSea. Since Google Trends often shows 0 for low trading volume entries, this paper focuses on the top 100 NFT series by trading volume from 2017 to 2022. (2) Filtering Criteria: Projects with a total supply exceeding 10,000 or a minting average price of 0 were excluded. The final sample includes 46 NFT series.
4. Regional Demographic Characteristics This paper uses per capita lottery sales as a proxy for regional gambling tendencies. Data was manually collected from state gaming commissions and aggregated to the DMA level. To avoid "look-ahead bias," all demographic data is set at a 2015 baseline to capture static cross-sectional differences.
5. Consumer Credit Characteristics Default data (90 days overdue considered as default) was obtained from Equifax. This paper calculates default rates at the DMA-year-month level and compares subprime (subprime, < 620) and non-subprime (≥ 620) populations based on credit scores. B. Descriptive Statistics
6. Regional Characteristics: In 197 DMAs with lottery data, the average annual lottery spending per adult is $199, but there is significant regional variation (ranging from less than $1 to over $800).
7. ICO Characteristics: ICOs raised an average of $26.3 million (about 40% of the hard cap). 36% of projects required KYC (identity verification), and 57% of projects made their code public on GitHub.
8. NFT Characteristics: The median issuance volume of NFT series in the sample is approximately 9,200. The vast majority (about 90%) are active on Twitter and Discord, and 85% of projects promote "rare items."

3. Regional Gambling Tendencies and Retail Crypto Attention

This study examines how differences in regional gambling tendencies affect the attention received by crypto tokens through estimating the following general regression model: where SVI represents the attention to ICO or NFT series i in designated market area (DMA) d during the issuance period. The core coefficient measures the impact of gambling tendencies at the DMA level on crypto attention. This paper uses per capita lottery sales as a proxy for gambling tendencies and controls for regional demographic characteristics and project fixed effects. Key Conclusions:

1. ICO Attention Analysis: The study finds a significant positive correlation between per capita lottery sales and ICO attention. Specifically, for every one standard deviation increase in gambling tendencies, the attention received by ICOs increases by approximately 12.8%. This conclusion remains valid after adding regional demographic variables or project fixed effects for robustness checks. This indicates that regions with higher gambling tendencies show greater retail interest in ICOs.
2. NFT Series Analysis: The research on NFTs shows an even more significant association. For every one standard deviation increase in gambling tendencies, the attention received by NFT series increases by approximately 20%. Although the attention to NFTs is more concentrated geographically than that of ICOs, the predictive power of gambling preferences for interest in NFTs remains very strong.

A. Robustness Checks: Alternative Gambling Preference Indicators This paper references existing research (Kumar (2009)) and uses various socioeconomic characteristics as alternative indicators of gambling preferences. The findings show that in regions with a higher proportion of Catholics, severe income inequality, higher unemployment rates, and a larger proportion of minorities, attention to crypto tokens is significantly higher. Conversely, regions with higher education levels, higher marriage rates, or higher income levels show lower attention to crypto tokens. This further confirms the high consistency between interest in crypto assets and traditional gambling psychological characteristics. B. External Validation: Does Attention Equal Investment? To verify whether "attention" effectively reflects "investment behavior," this paper conducted two tests:

1. Crypto Wallet Attention: The study finds that during token issuance periods, regions with high gambling tendencies see a simultaneous surge in search volume for crypto wallets like MetaMask and Coinbase Wallet. Since participation in ICOs/NFTs requires such wallets, this provides strong evidence for the conversion of attention into actual investment intentions.
2. Primary Market Performance: By introducing "anchor tokens" to compare the absolute search heat of different projects, the study finds that high-attention ICO projects raise more funds, have a higher proportion of reaching fundraising caps, and see a significant increase in the number of contributing participants on the first day; high-attention NFT series can raise more funds, have more minting wallets, and significantly shorten the time required for minting (an increase of one standard deviation in search heat can reduce minting time by approximately 71 days). C. Excluding Other Explanatory Pathways This paper examines other potential channels that may interfere with the conclusions, finding that:
3. Anti-establishment and Institutional Distrust: Using the vote share of the Libertarian Party and the complaint rate of the Consumer Financial Protection Bureau (CFPB) to measure regional distrust, it is found that these factors do not explain the association between gambling tendencies and crypto attention.
4. General Risk Preferences: Introducing survey data to measure regional risk preferences shows that they do not replace the explanatory power of gambling tendencies for interest in crypto.
5. Regional Advertising Expenditure: For the NFT sample, regional advertising spending by crypto exchanges was controlled, and it was found that even after considering the impact of advertising marketing, regional gambling tendencies remain the core variable predicting crypto attention. Conclusion Summary: Empirical results consistently indicate that regional gambling preferences are the core driving force behind retail attention to crypto tokens, and this attention directly translates into primary market fundraising performance, rather than being solely driven by institutional distrust, general risk preferences, or marketing strategies.

4. Factors Driving Gambling-Type Token Attention

In this section, this paper explores various factors that moderate retail investors' gambling-type attention to crypto tokens, including the characteristics of the tokens themselves and changes in the external gambling environment.

A. Token Characteristics Analysis

This paper examines specific token attributes that may trigger retail gambling psychology.

1. Low Price Characteristics (Lottery-Like Attributes): According to existing literature (Kumar (2009)), low price is a core characteristic of lottery-like stocks. Empirical findings show that ICO projects with lower opening prices on the first day receive significantly more attention from regions with high gambling tendencies compared to higher-priced projects. The interaction term coefficient indicates that low-priced tokens see an additional increase in attention of about 5% in these regions.

2. Verification Protocols (KYC) and Risk Preferences: Price manipulation behaviors such as "pump and dump" (P&D) schemes are common in the crypto market, and such projects often have weak KYC (Know Your Customer) checks. The study finds that ICOs lacking KYC protocols attract extremely high attention from retail investors in high gambling tendency regions, indicating that these investors are more inclined to participate in high-risk, poorly regulated projects.

3. Market Bubble/Prosperity Period Effects: This paper defines the second half of 2017 to early 2018 as the "prosperity period" of the ICO market and the price surge phase of the NFT market from 2021 to 2022 as the "explosion period." Regression results show that token projects launched during these two periods receive significantly more attention from high gambling tendency regions compared to non-bubble periods. For NFTs, attention from high gambling tendency regions during the bubble period is approximately 23% higher than during non-bubble periods.

B. Impact of Sports Betting Legalization

To further confirm that crypto attention is driven by gambling preferences, this paper uses the phased legalization of sports betting across U.S. states as a natural experiment. If crypto tokens are viewed as substitutes for gambling, then the attention to tokens should decline when legal gambling channels emerge. This paper estimates the following regression model:

where PostSG is a dummy variable that takes the value of 1 when sports betting has been legalized in the state where DMA d is located and the ICO occurs after the legalization date.

Main Conclusions:

1. Significant Substitution Effect: Empirical results show that after the legalization of sports betting, attention to ICOs in the relevant regions significantly declines.

2. Stronger Response in High Gambling Tendency Regions: After introducing the interaction term of "gambling legalization" and "regional per capita lottery sales," the coefficient shows a significant negative correlation. This indicates that in regions with already high gambling tendencies, the opening of sports betting has the most pronounced "crowding out effect" on crypto token attention.

3. Conclusion Summary: This finding strongly demonstrates that retail investors view crypto tokens as substitutes for traditional gambling products. When residents have legal sports betting channels to satisfy their gambling appetites, their attention to the crypto market diminishes.

5. Retail Crypto Attention and Consumer Credit Outcomes

Existing research (Barber and Odean (2000); Barber et al. (2022)) indicates that retail investors often perform poorly in traditional stock markets. If their performance in the crypto market is similarly lackluster, they may face financial difficulties. Therefore, this section studies the relationship between retail crypto attention and subsequent consumer credit outcomes, examining how this relationship varies with consumers' credit constraints. This paper measures credit constraints using credit scores and divides them into subprime borrowers (subprime, score < 620) and non-subprime borrowers (≥ 620). Given the comprehensiveness of the ICO sample compared to the NFT sample, this section focuses on the relationship between retail attention to ICOs and consumer default rates. This paper estimates the following regression model at the designated market area (DMA) --- credit segment --- year --- month level:
where the dependent variable is the change in the default rate between the current month t and the next 6 months (t+6). HighSVI is a dummy variable indicating the top third of attention for the month. Key Conclusions:

1. Association between Crypto Frenzy and Default Rates: The study finds that the interaction term between per capita lottery sales (gambling tendencies) and the ICO attention metric is significantly positive. This indicates that in regions with high gambling tendencies and high ICO attention, subsequent consumer credit default rates will significantly rise.
2. Vulnerability of Subprime Borrowers: Further analysis shows that the surge in default rates is entirely driven by subprime borrowers. In regions where high gambling tendencies and high attention coexist, the default rate of subprime borrowers rises by approximately 2.3% within 6 months. In contrast, the default situation of non-subprime borrowers (those in better financial condition) shows no significant change.
3. Lead-Lag Relationship and Pre-Trend Tests: To rule out the possibility that default behavior itself leads to an increase in attention, this paper conducts pre-trend analysis on changes in default rates. The conclusions show that during the period before the attention explosion (t-6 to t), there are no significant differences in default rates across regions (no pre-trend); in the period after the attention explosion (t+1 to t+6), the default rates of subprime borrowers in high gambling tendency regions begin to rise significantly. This temporal lead-lag relationship indicates that it is the crypto market's attention frenzy that signals subsequent credit deterioration, not the other way around. Conclusion Summary: This chapter's research demonstrates that gambling-driven investment tendencies in crypto assets can have negative economic consequences for socially vulnerable financial groups. For subprime borrowers, who already face financial constraints, participating in such high-risk, lottery-like crypto investments often comes with real financial default risks.

6. Conclusion

This paper delves into the fundamental driving forces behind retail investor participation in the cryptocurrency market, finding that gambling preferences are the core factor explaining this phenomenon. By analyzing Google Trends search data, this paper confirms that in regions with higher per capita lottery sales and a stronger speculative atmosphere, retail attention to Initial Coin Offerings (ICOs) and Non-Fungible Token (NFT) projects significantly leads other regions. This attention is not mere hype; it is highly synchronized with the downloads and usage of crypto wallets and directly positively impacts the fundraising amounts and participant numbers in the primary market.

Further moderation effect analyses show that this gambling-driven investment motivation is particularly pronounced during market "bubble periods" and when tokens exhibit "lottery-like characteristics" (such as extremely low unit prices, lack of verification protocols/KYC, and susceptibility to price manipulation). The study also finds through the natural experiment of sports betting legalization across U.S. states that when legal gambling channels emerge, the previously active attention to crypto tokens significantly declines, strongly demonstrating that retail investors view crypto tokens as substitutes for traditional gambling products.

Most critically, this gambling preference-based speculative behavior poses substantial threats to individual and societal financial health. Utilizing microdata from Equifax, the study finds that crypto attention surges in high gambling tendency regions often foreshadow rising consumer default rates in the following months, and this credit deterioration is entirely concentrated among the financially weakest subprime borrowers. This finding challenges the simplistic notion that "crypto assets are inclusive financial tools," revealing their potential predatory nature as speculative instruments on the wealth of the social underclass. In summary, this paper provides important academic evidence for global regulatory agencies: cryptocurrencies are largely viewed by retail investors as a new form of gambling tool, and regulation of such assets should not be limited to financial risks but should also consider public health and consumer protection perspectives, establishing stricter disclosure standards and entry thresholds.