Retail risk: Investors’ portfolios during the pandemic

Individual participation in financial market investment expanded significantly over the past decade, with considerable growth during the COVID-19 pandemic. The share of U.S. households with at least some stock holdings reached a record high as of 2022, according to data from the Federal Reserve,¹ and the monthly percentage of individuals moving money from checking to brokerage accounts was three to four times higher during the pandemic than in the preceding years (Wheat and Eckerd, 2024a, 2024b).^2 3 Consequently, a broader segment of the population has engaged in opportunities to generate additional wealth for retirement, home purchases, and savings for times of hardship, albeit while accepting market risk.

A significant factor in the rise of investing participation was the widespread availability of online self-directed brokerage accounts.⁴ These accounts provided easier access to capital markets for individuals who might otherwise be unable to invest, including those with limited liquid assets⁵ or without employer-provided retirement accounts. Given this expansion, policymakers have increasingly focused on understanding the tradeoffs between encouraging more individuals to invest and the heightened financial risks, especially for inexperienced investors.⁶ Timely data on investment risk in retail investors’ portfolios are limited.⁷ Prior research on behavior of retail investors predates the recent growth in investing and does not provide a direct measure of the potentially changing connection between financial markets and household portfolios.^8 9

Our research aims to fill these gaps by examining investment portfolios in self-directed brokerage accounts. We use a new de-identified dataset of 500,000 investors spanning the years 2019 to 2023.¹⁰ The dataset includes holdings in stocks, mutual funds, ETFs, fixed income, cash and cash-equivalents. We measure risk to quantify the trade-offs retail investors made when selecting investments. Generally, higher risk implies a greater potential for losses, and to compensate, investors typically expect higher returns.

Our most notable finding is that market risk in portfolios rose substantially in 2020 and remained elevated through 2023, indicating that the average investor faced greater fluctuations in their portfolio value relative to the stock market. Going forward, these investors may also expect higher potential long-term wealth accumulation, given the historical tendency for the market to rise over time. Coupled with the research showing broader participation in investing, our findings indicate a stronger link between individuals’ wealth and financial markets than before the pandemic. Policymakers, including those at the Federal Reserve, should consider this relationship when assessing the impact of macroeconomic policy on individuals.

This report also examines how portfolio risk varied by investor sub-groups, including cohorts based on account start date, gender, and age. Our results indicate that investors establishing new accounts in 2020 and 2021 took on more risk than those in adjacent periods, likely influenced by unique market conditions and heightened attention to stock trading during the pandemic. We also find that women’s portfolios held less investment risk than men’s, and younger investors held more market risk than older investors.

These differences by cohort, gender, and age highlight the importance of providing investors with educational information about investment risks. While our analysis does not determine whether investors chose the appropriate level of risk for their financial objectives, our results highlight the need for accessible education tailored to diverse goals. Financial literacy can help investors better leverage the opportunities that self-directed accounts offer for wealth accumulation.

Our overall findings are summarized as follows:

Background: Investment risk and retail investors’ portfolios

We use granular data to assess risk in investment portfolios. Investment risk describes the trade-offs investors make when selecting investments. Securities with higher risk metrics typically have a greater likelihood or magnitude of potential losses, and investors expect higher returns to compensate. We consider two general risk categories: market (or systematic) risk, which affects the entire stock market, and idiosyncratic (or unsystematic) risk, such as single-stock volatility, which can be mitigated through diversification.

Our findings measure three portfolio risk metrics, encompassing both systematic and unsystematic risk along with a measure that captures the portion of return variation attributable to the former. In calculating these metrics, we use the S&P 500 as our market benchmark in regression models with portfolio and security returns. Appendix 2 provides further detail on the construction of these metrics.

1. Market Risk: Measures systematic risk by capturing the change in portfolio returns relative to changes in the S&P 500, known as beta, estimated by a simple regression. A beta greater than one indicates that the portfolio fluctuates more than one-for-one with percent changes in the market, on average.
2. Likeness to Market Benchmark: Measures how closely a portfolio tracks the S&P 500, measured as the R² from a regression model. A higher R² indicates more systematic risk; the portfolio’s returns closely follow the market, typical of a diversified portfolio of stocks or index funds. A lower R² suggests greater idiosyncratic risk, as the portfolio deviates more from the market benchmark, and less diversification.
3. Idiosyncratic Risk: A measure of unsystematic risk. It captures percent changes in the portfolio value not explained by the overall market, quantified as the standard deviation of portfolio return residuals from the market risk regression. Higher idiosyncratic risk indicates greater exposure to random, diversifiable shocks specific to individual firms or industries and less portfolio diversification.
    

We organize our analysis of retail investor risk dynamics around three findings. Finding 1 examines how exposure to market risk evolved from 2019 through 2023. Here, we focus solely on market risk due to limitations in accurately capturing changes in the other metrics over a short-time window. Findings 2 and 3 explore how all three types of portfolio risk varied across different investor cohorts and demographic characteristics, providing a comprehensive view of portfolio risk among retail investors. Appendix 2 provides further details on the construction of these metrics.

Growth in retail investment activity during the pandemic was substantial. As of mid-2020, individuals were three to four times more likely to transfer money from checking accounts into retail brokerage accounts than they were from 2010 to 2015 (Wheat & Eckerd, 2024a). While increased investment participation is well-documented, relatively little is known about how individuals managed their risk in self-directed brokerage accounts during this time period.

To document the evolution of market risk, we examine the exposure of retail investor holdings to changes in the U.S. stock market. We compute the beta of securities held in portfolios using the S&P 500 as the market benchmark.¹¹ Figure 1 shows the average portfolio beta in our sample from 2019 to 2023. To focus our analysis on the risk-taking of typical investors, we present the equal-weighted¹² average beta.¹³ These results indicate that investors’ portfolios assumed greater market risk during the pandemic bull market. Average portfolio beta began to rise sharply in March 2020, peaking in December 2021 at nearly 15 percentage points higher than the 2019 level. Although market risk levels started to decline after this peak, they remained 10 percentage points higher by mid-2023 than in 2019.

Changes in market risk are driven by two mechanisms. First, active trading decisions by investors to buy or sell securities can alter a portfolio’s beta. Portfolio beta increases when investors make net purchases of riskier securities than those previously held. Second, portfolio beta can change passively when investors hold quantities of shares constant, but changes in market prices alter the relative value of securities in the portfolio. Passive increases in beta result from relative price increases of high-beta securities already held by investors, which occur automatically during market upturns in the absence of active rebalancing. To identify the contributions of these two mechanisms to changes in overall market risk, we decompose changes in average portfolio beta following the approach of Calvert et al. (2009).

Figure 2 presents the results of our decomposition on a quarterly basis over our sample period. The analysis suggests that both active and passive factors contributed meaningfully to the rise in average portfolio beta from the first quarter of 2020 through the first quarter of 2021. Market declines in 2022, particularly in the second quarter, drove a passive decline in portfolio beta, partially offset by sustained trading activity to increase risk.¹⁴ From March 2020 to the December 2021 peak, active changes accounted for approximately two-thirds of the total average increase in market risk, while passive changes contributed the remaining one-third.¹⁵ The large contribution of active trading to increases in beta suggests investors had a strong appetite for market risk during this period, which they retained past the pandemic period.

A consideration of policymakers during a February 2021 Congressional hearing¹⁶ was that online self-directed brokerage accounts—amid the pandemic context of heightened media attention on investing, income growth, excess savings, and a bull market—might attract new investors prone to increased risk-taking. While this event was unprecedented, existing literature suggests that trading behavior in self-directed brokerage accounts was more sensitive to biases like heightened attention to the market during this period (Barber et al. 2023). Moreover, market returns early in an investor’s experience can have lasting effects on their risk perception (Malmendier and Nagel, 2011). The historic influx of new investors during the pandemic may be riskier than those who opened accounts before or after, potentially affecting their wealth in the short- and long-term.

To explore this possibility, we segment our investor sample into three cohorts based on account opening dates: 1) a pre-pandemic cohort (January 2019 to February 2020); 2) a pandemic cohort (March 2020 to June 2021); and 3) a late-pandemic cohort (July 2021 to September 2023). We require these cohorts to have at least three months of investing behavior before analyzing their portfolio characteristics to avoid the volatility that occurs when investors initially fund their accounts.

Figure 3 compares market risk, likeness to the market benchmark, and idiosyncratic risk for each of the three investor cohorts. Consistent with strong market performance following the initial pandemic volatility in March 2020, we find significantly higher investment risk metrics in the pandemic cohort’s portfolios. Specifically, portfolios of the pandemic cohort had 12.6 to 11.7 percent higher beta than the cohorts that preceded and followed, respectively.

As an illustrative example to demonstrate the significance of these differences in dollar terms, consider a hypothetical portfolio with the pandemic cohort’s beta and an initial investment of 60,000 dollars (near our sample average). From March 2020 to December 2021, following the total S&P 500 return of 84 percent, this portfolio would have gained an additional 6,400 dollars compared to a portfolio with the pre-pandemic beta. Conversely, during the 15 percent S&P 500 correction in 2022, this portfolio would have lost 1,100 dollars more than a portfolio with the pre-pandemic beta. Portfolios with higher beta gain more during market upswings but incur greater losses when markets correct.

The two remaining risk metrics shown in Figure 3 indicate that the pandemic cohort’s portfolios tracked the overall market less closely and held more idiosyncratic risk than the other cohorts, suggesting potential concentrated holdings of more volatile single stocks and less diversified portfolios. Specifically, R-squared—measuring likeness to the market benchmark—was 44 percent for the pandemic cohort, compared to 45 percent for the earlier cohort and 57 percent for the later cohort. Idiosyncratic risk—measured as the monthly standard deviation of portfolio returns—was higher at 6.1 percent for the pandemic cohort, relative to 5.0 and 4.0 percent for the earlier and later cohorts, respectively.

Demographic differences in investor cohorts offer important context for interpreting these results. The pandemic cohort of investors were 2.5 times more likely to be men and 1.1 times more likely to be Generation Z and Millennials.¹⁷ As shown in Finding 3, investors with these demographic characteristics have riskier portfolios generally, but demographic differences do not fully explain the gaps in risk by cohort, a result described statistically in Appendix 3. In addition, interpretation of these results should be considered in the context of the limitations of our data, which are discussed in Appendix 4: Research Limitations.

These estimates suggest that policymakers’ considerations about the heightened risk-taking behavior of new investors during the pandemic may have an empirical basis. This cohort may have been influenced by unique market conditions and heightened attention to stock trading during the pandemic period, which may have led to investment strategies with higher risk.

Previous research on investment holdings in brokerage accounts shows that portfolio risk can vary by investor gender and age.¹⁸ These demographic differences can have implications for the long-term impacts of risk on wealth accumulation and the importance of tailored financial education. Here, we investigate our risk metrics—market risk, likeness to the market benchmark, and idiosyncratic risk—for portfolios by each demographic subpopulation.

Figure 4 illustrates the differences in portfolio risk by investor gender. Consistent with prior research, women’s portfolios were lower beta, more aligned with the market benchmark, and lower in idiosyncratic risk, suggesting more investment in index funds and less volatile investments. In contrast, men’s portfolios were held higher market risk, less aligned with the market benchmark, and higher in idiosyncratic risk, suggesting a more concentration in single-stock equities and less diversification.¹⁹

These results align with prior research that suggests women may be more risk-averse investors than men (Jianakoplos and Bernasek, 1998; Barber and Odean, 2001). Economic factors contributing to this difference include women’s greater vulnerability to poverty, lower lifetime earnings, less wealth, and longer lifespans, which necessitate higher savings (Bernasek and Shwiff, 2001). As women are likely to hold an increasing share of wealth due to generational wealth transfers in the coming years,²⁰ tailoring investing education for goals that might differ by gender is increasingly important.

We also find differences by age group generally consistent with traditional investing patterns over the lifecycle: individuals tend to reduce market risk as they approach retirement (Ameriks and Zeldes, 2004). Figure 5 shows portfolio risk metrics by age as captured by investor generation.²¹ Consistent with earlier evidence on the typical lifecycle profile of investment risk,²² market risk was higher in the portfolios of younger individuals. Our results for likeness to the market benchmark and idiosyncratic risk were less directional with age. Younger investors’ portfolios exhibited more likeness to the market benchmark and higher idiosyncratic risk compared to older investors. This may be because older investors’ portfolios included other types of risk, such as interest rate risk from holdings of fixed income securities, making them less correlated with the S&P 500 but still relatively low in volatility.²³

Appendix 3 shows that age and gender differences in risk persist when controlling for cohort effects and portfolio characteristics.

Conclusion

We examine retail investor risk during a period of significant economic changes from 2019 to 2023, using a dataset from Chase Wealth Management covering the portfolios of 500,000 investors in self-directed brokerage accounts. Our analysis reveals that market risk in retail investors’ portfolios increased by an average of 15 percent from 2019 to 2021, driven by active trading and the appreciation of riskier securities. The rise in market risk only partially receded to 10 percent by the end of 2023, indicating a lasting impact of pandemic-era volatility on household portfolios. We find that investors who opened accounts during the pandemic between 2020 to 2021 assumed more investment risk compared to those in adjacent periods. The results suggest investors may have been influenced by the unique market conditions and heightened attention to stock trading during the pandemic. Additionally, our analysis shows notable differences in portfolio risk by age and gender. Portfolios of men and younger investors were more sensitive to market returns, exhibiting higher levels of risk compared to those of women and older investors.

Implications

This report introduces a new source of data on retail investors’ portfolio holdings in self-directed brokerage accounts, aiding policymakers in understanding the expanding connections between financial markets and individuals’ financial health. Implications of this research apply to the following domains:

• Macroeconomic policy and financial health: Our results show retail investors increased the market risk of their portfolios compared to their holdings prior to the pandemic. While our analysis does not directly measure the impact of this risk on individuals’ broader financial health, it shows that investors increased the sensitivity of their wealth stock market volatility. Coupled with recent research showing increased participating in investing (Wheat and Eckerd, 2024a, 2024b), this heightened market exposure indicates a stronger connection between the U.S. stock market and individuals’ investment wealth since the pre-pandemic period. Policymakers, including the Federal Reserve, should consider these findings when assessing macro-financial linkages, particularly the impact of monetary policy on individuals, such as wealth effects on consumption and consumer sentiment.
• Equitable access to capital markets and wealth: Access to capital markets is a key tool for longer-term wealth accumulation, and self-directed brokerage accounts help lower barriers for investing. Increased participation also means more individuals are exposed to investment risk, with the trade-off of potential losses for higher expected returns. Since risk exposure varies in investor sub-populations, there may also be differential outcomes by investor cohort, gender, and age. Policymakers can use these results to consider the impact of policies that would support increased access to self-directed brokerage accounts.
• Financial literacy and tailored education: With the broadening of investment participation and increased exposure to market risk, attaining financial literacy becomes crucial for investors to understand their choices and manage risks effectively. Investment education is especially important for self-directed brokerage account holders who manage their own risk. While our analysis does not evaluate if the risk is appropriate for each investor’s goals, the results suggest a need for access to tailored education, in particular by gender, age, and investment experience. While women are likely to hold more wealth in the future, they also face higher gaps for retirement savings and may need different information than men to manage these needs.²⁴ Financial education should also focus on diversification and risk management for newer and younger investors to help ensure they understand the risks involved in investing.
• Future research: The new data on individual portfolio holdings provide a valuable resource for future research on the relationship between financial markets and household finances. Further studies could explore the consumption wealth effects of household investment, the role of retirement funds for economic security, and behavioral trends in trading.

The increased exposure to market risk in retail investors’ portfolios during the pandemic, along with differences in portfolio risk by cohort, gender, and age, indicates the need for policymaker awareness, financial literacy, and ongoing research.

Appendix 1: Data and Summary Statistics

Our research uses a new de-identified dataset from Chase Wealth Management containing the portfolios of 500,000 retail investors in self-directed brokerage accounts from 2019 through 2023.²⁵ The dataset includes month-end snapshots of holdings in stocks, mutual funds, ETFs, fixed income, and cash.²⁶ We observe the quantities of positions held and closing market prices, and we approximate flows in and out of securities by changes in these quantities and prices. The data encompasses both standard investment accounts and tax-advantaged Individual Retirement Accounts (IRAs). For investors with both types of accounts, holdings are aggregated at the individual level. The dataset is also enriched with investor age and gender, supporting analysis of heterogeneity across demographics. For a discussion of data limitations, see Appendix 4: Research Limitations.

We apply the following filters to curate the data used in this report. After aggregating positions across all self-directed accounts at a customer-month level, we require investors to have an average total portfolio wealth of at least 1,000 dollars and a minimum of 12 months of investment history. We also censor extreme outliers in the averages we report.

Table 1 presents summary statistics of investment portfolios and demographic characteristics. On average, investors in our sample have 60,000 dollars in investment balances, hold 7 positions, and trade once or twice a month in their self-directed brokerage accounts. The sample is predominantly male, with Millennials as the largest generation. The pandemic cohort—those who opened brokerage accounts in 2020 and 2021—constitutes half of the sample, while the other two cohorts split the remaining half.

Table 1: Summary statistics of investment portfolios and demographic characteristics

Appendix 2: Modeling Risk with our Data

Investment portfolios hold securities with varying risk levels. Using standard statistical methods, we run OLS regressions to calculate risk factors for both portfolios and individual securities. Specifically, we regress monthly returns of each security or portfolio on the S&P 500. Each regression assumes time-invariant risk characteristics either at the portfolio or security level. The regression for securities informs Finding 1, while the regression for portfolios informs Findings 2 and 3. Both analyses require a minimum of 12 months of historical data.

One factor security regression, where s is each security, and t is each month: 

R_s,t = α_s,t + β_sRm_t + ε_s,t

One factor portfolio regression, where i is each investor:

ⁿ∑_s=1 PW_i,s,tR_i,s,t = α_i,t + β_iRm_t + ε_i,t

For these regressions, we focus on three estimates:

• Market risk, the beta coefficient β_i: Captures the change in the return of the security or portfolio relative to the S&P 500. A beta over one indicates higher volatility than the market, a beta under one indicates lower volatility. A beta of one indicates equal volatility to the market.
• Likeness to market benchmark R²: Measures how much of the market portfolio explains the portfolio's return, serving as an indicator of equity market diversification. Values close to one suggest the portfolio may primarily consist of an index fund tracking the market. Lower R² values may imply less diversification or a higher reliance on cash equivalents or fixed income rather than equities.
• Idiosyncratic risk, the standard deviation of the residuals ε_i,t: This represents the variation in the portfolio not explained by its average relationship to market-wide fluctuations, interpreted as a measure of idiosyncratic or firm-specific risk. Higher levels of idiosyncratic risk indicate greater exposure to random, undiversifiable shocks relative to the market.

Table 2 presents statistics from individual portfolio-level regressions. The average portfolio had a beta below one (note, Figure 1 shows higher betas, as it only includes the portion invested in single-stock equities, ETFs, and mutual funds). We also find an R² indicating that half of the portfolio risk is explained by the S&P 500, and an idiosyncratic risk measure of 6 percent calculated from the average monthly return standard deviation of residuals.

Table 2: Descriptive statistics of investment risk metrics from portfolio regressions

Appendix 3: Further insights into heterogeneity within portfolios

To ensure the robustness of metrics on differences in portfolio risk, we use OLS regressions to explain the variation in portfolio market risk (beta); likeness to the market benchmark (R²); and idiosyncratic risk (the standard deviation of the residuals), alongside a set of controls. The specification is as follows:

Risk Metric = α + D₀ * Cohort + B₀Age + D₁Men + B₁Log Average Portfolio Wealth + B₂Log Total Trades + B₃ Log Average Positions + B₄  Average Portfolio Weight in Equity

Our primary coefficients of interest are the binary variables for cohort, age, and gender. For controls, we include portfolio-level statistics: average portfolio wealth, total trades, average positions, and the percentage of the portfolio allocated towards risky securities. We use two specifications: the second includes portfolio weight in equities. This variable helps approximate differences driven by security selection within equities, but we also present results without it, as it is endogenous and closely related to risk metrics.

Table 3 reports the portfolio-level regression results. Differences between subpopulations by age, gender, and cohort remain robust with the inclusion of the observed controls. Consistent with Finding 2, the pandemic cohort’s portfolios had higher market risk and less diversification—as interpreted by higher betas, lower R², and higher idiosyncratic risk—compared to neighboring cohorts. Consistent with the plots in Finding 3, men’s portfolios had higher market risk and less diversification, as interpreted by higher beta, lower R², and higher idiosyncratic risk. The age results in this analysis confirms that the portfolios of younger investors had higher market risk and higher R². However, the result for idiosyncratic differs from Finding 3. The coefficient on age suggests older investors had higher idiosyncratic risk. This may be due to allocations in fixed income or low-beta stocks with more idiosyncratic risk, consistent with Korniotis and Kumar (2011).

Table 3: Heterogeneity in Portfolio Risk metrics²⁷

Appendix 4: Research Limitations

Our analytical approach has several limitations that warrant consideration. While we measure increased risk-taking in investment accounts, our analysis does not show the impact of this heightened risk directly on individuals’ broader financial health. Future research will investigate the impact of changes in investor investment wealth on consumption. Another limitation is that our analysis does not evaluate whether investors are fulfilling their financial goals with a chosen level of risk, partly due to a lack of information on investors’ objectives for their finances.

We are also constrained by our data coverage. The available time frame from 2019 to 2023 makes it difficult to contextualize changes in retail investor risk-taking over a longer historical period. This time frame also overlapped with changes to the Chase self-directed investing platform that might have increased the number of available securities or other features. Additionally, unlike other online-only platforms, the population using self-directed accounts in our sample may consist of more traditional banking customers with higher incomes than the general population. Our data may provide a conservative look at cohort effects, as there was reportedly more elevated risk-taking on other platforms not included in this analysis.²⁸ Furthermore, we do not observe investments held at other companies, meaning our analysis does not comprehensively cover all investors’ financial assets. We also do not include in our analysis any assets that individuals may hold in Chase advisory accounts; this affects only a small fraction of our sample.