Why is important to assess risks regarding COVID-19?
Science
According to the New York Times, as of March 2022, there have been over 79 million confirmed cases of COVID-19 in the United States since the outbreak in January 2020. As the virus spread from the populace to other states, households experienced different levels of exposure to the pandemic in terms of varying local infection rates, different health conditions, different personal experiences with the infection of the virus, and different economic impacts on the local economy due to local lockdown measures. Studying the effect of the differential exposure due to varying health conditions in the US on investor risk-taking, this paper offers information on individual risk-taking behavior during unprecedented economic and health situations.
While research has studied how past experiences affect risk-taking using measures such as self-rated questionnaires and hypothetical investment experiments, I want to study how investor risk-taking is affected during the pandemic due to varying exposures by examining actual investment decisions. Investors’ risk-taking behavior, measured by both the individuals’ portfolio choice and the decision to enter the equity market, provides insight into their risk preferences, economic beliefs and expectations. Using data collected during the onset of the pandemic, the study adds to the information of predicting investor behavior during both an economic and health crisis.
I hypothesize that households that have greater exposure to the COVID-19 pandemic will take less risk when making personal financial decisions.
Specifically, households that have at least a member with health conditions that make them more susceptible if infected will modify their investment portfolio composition to be more conservative and reduce their likelihood of entering the equity market. The reduction in investment risk-taking might be due to reduced tolerance for risk, pessimistic beliefs about the economy, increased need for liquidity during the pandemic, and change in local economic conditions.
The results in this paper support the hypothesis and suggest that increased liquidity needs during the pandemic for susceptible households could be the reason for reduced risk-taking. Susceptible households with future consumption commitments, higher medical expenditures, and reduced expected labor earnings experience increased liquidity needs and will shift towards less risky assets to avoid short-term capital losses. Contrary to the findings in the literature, household short-term expectation and risk preferences do not have a significant effect on household risk taking behavior during the pandemic.
Literature Review
Attitude towards risk is a key determinant of an individual’s economic behavior under uncertainty. A number of studies have investigated the heterogeneity of risk preferences and how it changes over time. Riley & Chow (1992) examined an individual’s actual pattern of asset allocation using the Survey of Income and Program Participation. Their results show that the share of risky assets to wealth decreases with each age category until the 65-and-older category, within which it increases significantly. While education. Wealth. And total income become highly correlated. They find risky asset share decreasing. As an individual’s wealth increases. Independent of the total income after regressing the risky asset share to wealth.
Other demographic factors like marital status, race, and gender have relatively small differences in risk aversion compared to factors like age, wealth, and total income. Their results, however, can only explain the effect on household risk taking behavior rather than household risk preferences, which will require additional controls for household expectations.
Acknowledging the gap between the observed portfolio and the result of the mean-variance efficient portfolio, Bucciol & Miniaci (2011) arrive at a similar conclusion after constructing the efficient portfolio with the same variance as the observed portfolio from the 2004 wave of the Survey of Consumer Finances data with constraints on efficient portfolio weights for human capital and mortgages and including real estate into the definition of the portfolio.
Using household-level panel data from the Panel Study of Income Dynamics, Brunnermeier & Nagel (2008) find that:
Changes in liquid wealth have a significant positive effect on the probability of stock market entry and a negative effect on the probability of exit.
However, liquid wealth plays no role in explaining changes in asset allocation for households that participate in the stock market. Likewise, Sahm (2012) uses panel data on hypothetical gambles over lifetime income in the Health and Retirement Study to quantify changes in risk tolerance over time and differences across individuals and finds that changes in household income and wealth over the decade do not alter an individual’s measured willingness to take risks, which is consistent with Constant Relative Risk Aversion (CRRA) assumption.
Despite the mixed findings regarding the effect of wealth on risk tolerance, Sahm (2012) agrees with other literature that risk tolerance declines modestly with age and increases with an improvement in macroeconomic conditions. In addition, major life events of a job displacement and the diagnosis of a serious health condition that likely reduce expected lifetime income also have little impact on measured risk tolerance (Sahm, 2012).
Another strand of literature explores the effect of risk preferences and beliefs on risk-taking behaviors. Malmendier & Nagel (2011) relate measures of household risk taking to their experienced histories of stock and bond returns. Their study demonstrates that households’ risk-taking is strongly related to experienced returns: households with higher experienced stock market returns are more willing to take financial risks, participate more in the stock market, and invest more liquid assets in stocks. Experience of increased historic returns is also associated with higher expectations of future returns.
Though their findings show that the effect of personally-experienced return histories is due to the impact on beliefs, risk preferences remain to be a channel that affects risk taking. In terms of risk-taking behavior during the economic crisis in 2008, findings by Necker & Ziegelmeyer (2016) and Guiso et al. (2018) show that the observed decrease in household risk-taking attributed to an emotional reaction to the crisis, similar evidence during the COVID-19 pandemic found by Angrisani et al. (2020) using two rounds of elicited risk preferences from professional traders and students before and during the outbreak of the pandemic.
Their results show that the reduced risk-taking is due to a change in beliefs by the market participants rather than a change in risk appetite. The stability of risk preference can also be observed from the results of a survey of more than 8,000 households conducted by Hanspal et al. (2020). The survey studied how exposure to the COVID-19 stock market crash affects expectations and planned behavior and found no systematic tendency to actively rebalance portfolios in response to the passive reduction in equity portfolio shares.
Aside from the change in expectations and risk preferences, risk-taking during the COVID-19 pandemic is also affected by varying regional exposure. Hanspal et al. (2020) conducted a repeated survey across a panel of graduate students at Wuhan University of Science and Technology before and after the outbreak of the pandemic. Using the variation in exposure across different Chinese cities and provinces to measure the impact of the pandemic on risk-taking, they find that subjects’ allocations of wealth to hypothetical risky investments decrease monotonically based on the strength of their exposure to the pandemic.
Graeber et al. (2020) also performed a similar study using the socio-economic consequences of COVID-19 survey data in Germany and found that the increased self-rated risk aversion can be observed particularly among better educated and more affluent respondents. Himanshu et al. (2021) also find a shift in preference towards risk-free assets like insurance, gold, bank deposits, and public provident funds during the pandemic.
In addition, the health conditions of the respondent and the respondent’s social connections also affect risk-taking. Guenther et al. (2021) find healthier participants display significantly higher risk tolerance in self reported risky behavior of a non-financial nature, such as not obeying the guidelines regarding social distancing. Knowing someone with COVID also impacts households’ confidence in the economy and adjusts their risk behavior to take fewer risks (Yue et al., 2020).
This paper is inspired by the framework and variables from the studies discussed above. Referencing the study by Brunnermeier & Nagel (2008), I use the change in financial wealth risky asset share and liquid wealth risky asset share as the dependent variable for measurements of households’ asset allocation and performed probit regressions on equity market entrants. Graeber et al. (2020) inspired this paper’s usage of longitudinal and regional variations in infection rates as the proxy of regional differences in exposure to COVID-19. And the control of pre-COVID health conditions and change in short-term economic expectations.
Based on the above discussions. This paper hypothesizes that households that are susceptible to COVID-19 if infected. Will be less willing to take risks when making personal financial decisions. Specifically, households with members who have health conditions like heart, lung conditions, diabetes. Or are immune-compromised that make them more susceptible to the virus will modify their risk-taking behavior to be more conservative. The reduction in investment risk-taking might be due to reduced tolerance for risk. Pessimistic beliefs about the economy. Increased need for liquidity during the pandemic. And changes in local economic conditions.
Empirical Methodology
Data
This paper uses data from the 2018 and 2020 waves of the Health and Retirement Study. Which is a longitudinal panel study featuring a sample of households with at least one member who is age 55. Or older in the United States conducted by the Institute for Social Research (ISR) Survey Research Center (SRC). At the University of Michigan. The data includes the economic. Health. Marital. And family status of older Americans. On both the individual and the household level. The 2018 imputed household-level income and wealth information. And sampling weights, extracted from the 1992-2018 RAND dataset. And merged with the two waves of HRS core data. To allow inferences to the U.S. population of households.
The regional COVID-19 infection rate data is obtained from the “Trends in Number of COVID-19 Cases and Deaths in the U.S. by region” data reported to the Centers for Disease Control and Prevention. With monthly averages corresponding to the month of interview of the HRS survey. The regional unemployment data is retrieved from the Local Area Unemployment Statistics. From the Bureau of Labor Statistics and weighted using the 2020 Census Apportionment Results from the Census Bureau. To create monthly unemployment rates by census region. And merged with the HRS data.
The greatest advantage of the HRS data is that, compared to other panel surveys on income and wealth. Moreover, the 2020 wave of HRS core data includes COVID-related information on household infections of COVID-19. Pre-pandemic health conditions of household members are also included to identify households with members that are susceptible to COVID-19 if infected. This information provides the basis for measuring how susceptibility to COVID-19 affects the risk-taking behaviors of households. The varying exposure to COVID-19 is also measured by the regional infection rates represented by the 7-day moving average number of new cases within each region. Regional unemployment rates are included to control for the impact on the local economy. Both households’ risk preferences and expectations, the two determinants of households’ risk-taking behavior, are included in the HRS data.
Imputation of the data comprises two parts. First, in order to analyze the data at the household level, I took the average value of numeric variables like age, years of education, and self-rated risk-taking across household members as the household variable for age and education. Household employment status is represented using a binary variable that indicates any unemployment within the household between 2018 and 2020.
Both marital status and change in marital status are represented using binary variables. Household-level categorical variables like COVID-19 exposure, and health (except for whether the household is susceptible to COVID-19, which is 1 if any one of the members is susceptible) are imputed by first taking the average of the responses of individual household members and then rounded to the nearest category. Second, since the 2020 HRS RAND imputation of wealth and income is not yet available, the wealth and labor income variables in 2020 are imputed based on the missing percentage of 2018 values.
Wealth is imputed using the sum of total assets and total liabilities of the households. Households with missing asset/liabilities categories that sum up to 30% of the total asset/liabilities value in 2018 are discarded. Households with missing values that sum up to less than 30% of the total value in 2018 are then imputed using the 2018 values, under the assumption that no significant changes took place during the 2-year period. Similar imputation method is applied to impute the labor income.
Households with less than 30% missing value of labor income in 2020 are imputed using labor income values in 2018, assuming no significant changes. The sample becomes restricted to households with positive liquid and financial wealth in both years. After discarding households that are extremely levered or with liquid or financial wealth less than $10,000.
Model Specification and Estimation Strategy
This paper uses a linear regression and a probit regression to compare the effects of the COVID-19 pandemic on susceptible and non-susceptible households’ risk-taking behaviors. Following Brunnermeier & Nagel’s (2008) model, the models included demographic controls like household average age, average education, dummy variables for any unemployment or retirement from 2018 to 2020, change in marital status, and change in household size. Wealth and income control variables include the change in log liquid and financial wealth (corresponding to the risky asset share of liquid/financial wealth as the dependent variable), and dummy variables for non-zero income in 2018 and 2020.
Background risk controls include the change in percent of labor income to liquid wealth, change in percent of home equity to liquid wealth, and change in percent of business equity to liquid wealth (for the liquid wealth risky asset share regression), and change in percent of labor income to financial wealth (for the financial wealth risky asset share regression). Local infection rate, local unemployment rate, and the change in self-rated health conditions become also included in the regression to account for health and regional economic conditions. Change in household risk preference becomes measured using the elicited risk preference in the HRS survey. Household risk-taking behaviors are measured by both the change in risky asset share in liquid and financial wealth. And equity market entrance between 2018 and 2020.
Household susceptibility becomes a binary variable. One hat takes the value of 1 if any member of the household has heart or lung conditions, diabetes. Or immune-compromised. The vector of controls becomes described as above. And the interaction between household susceptibility. And whether the household knows someone with COVID-19 is also included. The regressions become computed with and without the expectation and change in risk preference variables. The expected sign for susceptibility should be negative in both regressions.
A household that is susceptible to COVID-19 may prefer to hold assets in more liquid forms rather than stocks to prepare for emergency needs and will therefore have a smaller risky share. The second set of regressions aims to explain the effect of household susceptibility on equity market entrance between 2018 and 2020. The probit regression uses equity market entry (had a positive equity position in 2020 but 0 in 2018) as the dependent variable and includes the same controls and fixed effects as the liquid wealth risky share regression in the first set of regression models.
Results
Financial wealth and liquid wealth risky share
The first set of regressions investigates the effect of exposure to COVID-19 on financial wealth and liquid wealth risky shares. Since the exposure variables do not show significant effects in the preliminary regressions (see appendix for results), the regressions only include the exposure variable in the interaction with household susceptibility to test for varying effects.
The effect of household susceptibility to COVID-19, as measured by whether any of the members in the household have health conditions that have higher risks if infected, is negatively correlated with financial wealth risky share but the magnitude of the effect is close to zero for liquid wealth risky share regressions. In terms of the financial wealth risky share regressions, results show that susceptible households’ financial wealth risky share is estimated to be 2-3 percentage points lower
compared to non-susceptible households. Similar effects are not found with liquid wealth risky share measures. The effect of the interaction between susceptibility and COVID exposure is not significant and close to 0. This means that the impact of exposure to COVID-19 does not vary by susceptibility. In terms of expectation variables, the change in short-term household expectations does not have a significant effect on household risky shares. The effect of household risk preferences is close to 0.
Household risk preferences are also not affected by household susceptibility therefore the change in risk preferences does not vary by household susceptibility (see appendix for regression results). Regional variables like unemployment rate and infection rates also don’t have a significant effect.
The results for wealth variables in the regressions match the results from Brunnermeier & Nagel (2008): although the estimates for change in log liquid wealth are positive and significant at the 1% level, the magnitude of the effects are close to 0. The coefficient results for change in log financial wealth in columns 1, 3, and 5 are negative and significant at the 0.1% percent level which is consistent with Brunnermeier & Nagel’s (2008) findings in both sign and magnitude.
Equity market entrance between 2018 and 2020
Results show that household susceptibility to COVID-19 has a negative effect on market entry as expected. A susceptible household expected on average to be 1.9 less likely than a non-susceptible household to enter the equity market between 2018 and 2020, after including household expectations and risk preferences. The effect, however, is not statistically significant. The effect of COVID-19 exposure also does not vary by household susceptibility since the interaction terms are not significant. Regional variables, household short-term expectations, and household risk preferences also don’t have any significant effects.
My results also match the findings of Brunnermeier & Nagel (2008) in terms of the effect of liquid wealth on equity market entry: the change in log liquid wealth is positively correlated with stock market entry. This effect is significant at the 0.1% level. Household average years of education also positively correlated with stock market entry: a one-year increase in household average years of education estimated to increase the household’s probability of entering the stock market by 1.8%. This effect is also significant at the 0.1% level.
Discussion and Conclusion
In conclusion, the evidence shows that households that are susceptible to COVID 19 take less financial risk by having smaller financial wealth risky shares than households that are not susceptible to COVID-19. Susceptible households are also less likely to enter the equity market than non-susceptible households. Households’ short term expectations and risk preferences do not have a significant effect on either financial risk-taking, measured by risky share, or equity market entry.
Overall, older households in the U.S. that have members with health conditions that increase their risk of mortality if infected reduce risk-taking by investing a smaller proportion of their financial wealth in risky assets during the pandemic. Household susceptibility also reduces the possibility of equity market entry. This result is consistent with the previous literature that increased exposure to the pandemic reduces observed household financial risk-taking (Hanspal et al., 2020; Yue et al., 2020).
One possible explanation is that households with at least one member infected by COVID-19 have a higher chance of experiencing symptoms of long-COVID and will prefer to hold less risky assets to avoid short-term capital losses due to an increased need for possible health and medical expenditures. Also, older individuals are more likely to develop more severe symptoms of COVID-19 if infected.
The health impact is greater for older individuals than for younger adults. Older working individuals whose health conditions impacted by the infection may choose to move up their retirement dates. And shift toward a more conservative portfolio due to a decrease in expected labor earnings. Susceptible households therefore have higher liquidity needs during the pandemic and are willing to take less financial risk. However, contrary to the literature’s findings, I did not find the changes in risk-taking behaviors to be a result of a change in beliefs as in Hanspal et al.’s paper. Since subjects in Hanspal et al.’s study are mostly graduate students. Whereas the households in the HRS data become mostly households with people aged 55 and older. It is possible that the effect of exposure to COVID-19 varies by age group.
There seem to be varying effects of wealth on changes in asset allocation when measured using liquid wealth and financial wealth risky shares. This effect of household susceptibility is only significant when measured in change in financial wealth risky share but not in the change in liquid wealth risky share. The R-squared values are also larger for financial wealth risky share regressions than for liquid wealth risky share regressions. While there is currently no reasonable explanation for this difference, future research can explore the difference between these two wealth measures using higher quality data.
Given the timeliness of this topic, there becomes limited data available regarding the COVID-19 pandemic. While the HRS data incorporates useful information regarding the health and expectations during the pandemic, the available raw data includes incomplete responses for the 2020 wave that need to be imputed since the imputed HRS RAND is not available for 2020. Due to the limitation of the imputation process. The sample only includes 2444 households. And the labor income becomes measured using household labor income. Without accounting for pension and transfer income. Since half of the households in the sample are retirees that receive pension or IRA income, not including social benefits, pension, and IRA income as additional income measures in the model may potentially bias the results. Imputation of wealth and labor income values based on 2018 HRS RAND entries also adds to the possibility of measurement error.
Future research can further explore this topic using datasets with more detailed wealth and income information to get a larger sample and explore the effect of different wealth measures on household risk-taking during the pandemic. It would also be useful to explore the duration and/or long-term effect of the changes in household financial risk-taking during the pandemic.
Why is important to assess risks regarding COVID-19? Written by Tianshu Zhang
References for Why is important to assess risks regarding COVID-19?
Angrisani, M., Cipriani, M., Guarino, A., Kendall, R., & Ortiz de Zarate, J. (2020). Risk Preferences at the Time of COVID-19: An Experiment with Professional Traders and Students. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3609586
Brunnermeier, M. K., & Nagel, S. (2008). Do Wealth Fluctuations Generate Time Varying Risk Aversion? Micro-evidence on Individuals. American Economic Review, 98(3), 713–736. https://doi.org/10.1257/aer.98.3.713
Bucciol, A., & Miniaci, R. (2011). Household Portfolios and Implicit Risk Preference. Review of Economics and Statistics, 93(4), 1235–1250. https://doi.org/10.1162/REST_a_00138
Centers for Disease Control and Prevention. (2022). Trends in Number of COVID-19 Cases and Deaths in the US Reported to CDC, by State/Territory. Retrieved from https://covid.cdc.gov/covid-data-tracker/#trends_dailycases.
Graeber, D., Schmidt, U., Schroeder, C., & Seebauer, J. (2020). The Effect of a Major Pandemic on Risk Preferences—Evidence from Exposure to COVID-19. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3724461 Why is important to assess risks regarding COVID-19?
Guenther, B., Galizzi, M. M., & Sanders, J. G. (2021). Heterogeneity in Risk-Taking During the COVID-19 Pandemic: Evidence From the UK Lockdown. Frontiers in Psychology, 12, 643653. https://doi.org/10.3389/fpsyg.2021.643653 Why is important to assess risks regarding COVID-19?
Guiso, L., Sapienza, P., & Zingales, L. (2018). Time varying risk aversion. Journal of Financial Economics, 128(3), 403–421. Why is important to assess risks regarding COVID-19?
https://doi.org/10.1016/j.jfineco.2018.02.007
Hanspal, T., Liao, Y., & Liu, Y. (2020). Risk-taking during a Global Crisis: Evidence from Wuhan. 41.
Hanspal, T., Weber, A., & Wohlfart, J. (2020). Exposure to the COVID-19 Stock Market Crash and its Effect on Household Expectations. 88. Why is important to assess risks regarding COVID-19?
Health and Retirement Study, (2018 Core) public use dataset. Produced and distributed by the University of Michigan with funding from the National Institute on Aging (grant number NIA U01AG009740). Ann Arbor, MI, (2021). Retrieved from https://hrsdata.isr.umich.edu/data-products/2018-hrs core. Why is important to assess risks regarding COVID-19?
Health and Retirement Study, (2020 Core) public use dataset. Produced and distributed by the University of Michigan with funding from the National Institute on Aging (grant number NIA U01AG009740). Ann Arbor, MI, (2022). Retrieved from https://hrsdata.isr.umich.edu/data-products/2020-hrs-core. Why is important to assess risks regarding COVID-19?
Himanshu, Ritika, Mushir, N., & Suryavanshi, R. (2021). Impact of COVID‐19 on portfolio allocation decisions of individual investors. Journal of Public Affairs, 21(4). https://doi.org/10.1002/pa.2649
Malmendier, U., & Nagel, S. (2011). Depression Babies: Do Macroeconomic Experiences Affect Risk-taking?*. The Quarterly Journal of Economics, 126(1), 373–416. https://doi.org/10.1093/qje/qjq004
Necker, S., & Ziegelmeyer, M. (2016). Household risk-taking after the financial crisis. The Quarterly Review of Economics and Finance, 59, 141–160. Why is important to assess risks regarding COVID-19?
https://doi.org/10.1016/j.qref.2015.03.006
Riley, W. B., & Chow, K. V. (1992). Asset Allocation and Individual Risk Aversion. Financial Analysts Journal, 48(6), 32–37. https://doi.org/10.2469/faj.v48.n6.32 Sahm, C. R. (2012). How Much Does Risk Tolerance Change? Quarterly Journal of Finance, 02(04), 1250020. https://doi.org/10.1142/S2010139212500206 The Covid-19 Pandemic. The New York Times. Retrieved March 6, 2022, from https://www.nytimes.com/news-event/coronavirus
U.S. Bureau of Labor Statistics (2022). Local Area Unemployment Statistics. Retrieved from https://www.bls.gov/lau/#data.
U.S. Census Bureau (2021). 2020 Census Apportionment Results. Retrieved from
https://www.census.gov/data/tables/2020/dec/2020-apportionment-data.html. Yue, P., Gizem Korkmaz, A., & Zhou, H. (2020). Household Financial Decision Making Amidst the COVID-19 Pandemic. Emerging Markets Finance and Trade, 56(10), 2363–2377. https://doi.org/10.1080/1540496X.2020.1784717
Why is important to assess risks regarding COVID-19? Written by Tianshu Zhang
Why is important to assess risks regarding COVID-19?
