Temperant Portfolio Choice & Background Risk

The abstract introduces the core research question: Does earnings uncertainty and borrowing constraints deter households from participating in the stock market? The study employs an empirical approach, aligning with theoretical predictions on portfolio choice when earnings are uninsurable. A key innovation is the use of a self-assessed proxy for the correlation between earnings and financial risks, sourced from the DELTA-TNS 2002 cross-sectional survey. This allows for a nuanced investigation into the impact of income risk on investment behavior, moving beyond simple observations of aggregate data. The study specifically investigates whether this relationship varies depending on the perceived correlation between earnings and financial risks, creating a hypothesis that can be tested using this data set.

The study's findings reveal a significant relationship between income risk and stock market participation, but only under specific conditions. While income risk does not deter stock market participation for households reporting a negative correlation between their earnings and financial risks, it does have a significant negative effect for those reporting a non-negative correlation. This aligns with established economic theory, suggesting that the perception of the relationship between income and market risk is critical. The use of the DELTA-TNS 2002 survey, which has the novel advantage of incorporating a self-assessed measure for the correlation between these two risk factors allows for a better understanding of the underlying mechanisms at play, improving upon previous research that has lacked this critical dimension. The fact that the findings align with established theoretical predictions lends weight to the methodological approach and the importance of the observed correlation.

The research relies on the DELTA-TNS 2002 survey, a cross-sectional survey of approximately 4000 French households, focusing on individuals aged 35-55. Of these, 2460 provided usable data for the analysis. The survey includes information on various household characteristics, including earnings, income, wealth, socio-economic status, and demographic details. Importantly, the survey incorporates questions designed to capture individual risk aversion and, crucially, a self-assessed measure of the perceived correlation between earnings and financial market risks. While innovative, it's acknowledged that the self-assessed nature of this correlation measure might introduce limitations, as subjective perception might not perfectly reflect actual co-movement between earnings and financial asset returns. The limitation of this particular methodology should be addressed in future research endeavors, possibly by incorporating longitudinal data, which is noted as lacking in similar studies.

The section begins by outlining the classical theory of portfolio choice, which operates under the assumption of complete markets where all individual risks are tradable. Key determinants of household stock demand within this framework are identified as the expected excess return of stocks over riskless assets, the objective risk of stocks (measured by variance), and households' subjective risk perception (measured by risk aversion). However, the paper notes that the complete markets assumption is unrealistic due to severe informational restrictions that prevent most households from insuring their labor income—their primary source of lifetime income. This necessitates a reconsideration of the standard model to incorporate the realities of incomplete markets, where labor income risk is non-tradable. This critique of the classical model is based on the observed limitations of information asymmetry and the reality of non-tradable income risks that lead to a departure from the predictions of this model.

The paper then discusses the implications of relaxing the complete markets assumption. The non-insurability of earnings risk fundamentally alters the portfolio choice problem, creating an important connection between portfolio and labor market decisions. The paper cites several extensions to the standard model, integrating factors such as non-tradable and undiversifiable income (Koo, Davis & Willen, Viceira), borrowing and liquidity constraints (Koo, Haliassos & Michaelides), indivisibilities (housing), age (Gollier & Zeckhauser, Viceira, Campbell & Viceira), and information/transaction costs (King & Leape, Vissing-Jorgensen, Haliassos & Michaelides). The availability of more recent household-level datasets has enabled more empirical research to test these extended models. The study itself utilizes household-level survey data to explore these theoretical advancements, particularly focusing on the French household context. The increased availability of granular household-level data, compared to aggregate data, has led to a better understanding of these theoretical concepts and the nuances of household financial decision making.

A critical aspect of the theoretical framework revolves around the correlation between earnings and financial risks. The paper highlights surprising empirical findings showing that these risks are often independent in the aggregate but can exhibit different correlations depending on how income is categorized (Davis & Willen, Heaton & Lucas). Some research, using longitudinal datasets, suggests a negligible impact of this correlation on household risky asset demand (Vissing-Jorgensen, Massa & Simonov). Yet, existing theory and other empirical work underscore the crucial role of this correlation, especially in addressing puzzles such as the home bias. The paper uses this theoretical backdrop to build upon the earlier limitations of the models, testing the implications of imperfect markets using household level survey data which considers the correlation between income risk and market risk. This addition allows for a deeper understanding of the complex relationships between these critical risk factors.

The concept of 'temperance' is introduced, where households facing uncontrollable background risk (like uninsurable labor income) reduce their exposure to other avoidable risks (endogenous risks) to manage their overall risk exposure (Pratt & Zeckhauser, Kimball, Gollier & Pratt). This implies a substitution effect between risks. The theoretical framework shows that in the presence of a correlated background risk (income), households' optimal choice of exposure to other controllable risks will be impacted. This theoretical model emphasizes the importance of considering the correlation between background risk, which is beyond the household's control, and endogenous risk, which can be altered by adjusting portfolio decisions. The paper then incorporates these theoretical implications into the study design to better understand the impact of the correlation between labor market income and stock market performance on household investment decisions. This leads to specific implications about the study's empirical strategy.

The empirical analysis utilizes data from the 'Mode de vie et épargne' household survey conducted by DELTA and Taylor Nelson-Sofres in 2002 (DELTA-TNS 2002). This survey included 4000 French households with heads aged 35-55, representing the French population within that age bracket. The final sample used in the analysis consisted of 2460 usable responses. The survey gathered information on various household characteristics (earnings, income, wealth, socio-economic status, demographics) and risk attitudes. Crucially, it included questions designed to capture the subjective perception of the correlation between individual earnings and financial market returns. Direct stockholding was defined as the sum of holdings across privatized public company stocks, private company listed stocks, and foreign firm stocks. Indirect stockholdings included holdings through mutual funds and managed investment accounts. Bonds were excluded from the risky asset category due to the low risk of default in the French context. The researchers used this data to test the impact of income risk and borrowing constraints on household stock market participation, using proxies for both income risk and liquidity constraints.

Due to limited data on earnings uncertainty, a proxy for subjective income variance was constructed using data from the INSEE 'Patrimoine 98' survey. By matching data from 1998 (where individuals provided probability weights for future income increases) with the 2002 DELTA-TNS data, the researchers predicted earnings uncertainty for households in the 2002 sample. This method aimed to partially address potential endogeneity issues between risk aversion, occupational choice, and stock market participation. Despite acknowledgements of the limitations of using self-assessed measures of income risk (e.g., suitability for a limited class of preferences, underestimation of very low-income events, and potential discrepancies with panel data estimates), this was the only way to conduct this particular research. The study also measured liquidity constraints using survey questions to identify 'discouraging borrowers' and 'turned-down applicants.' A dummy variable indicated liquidity constraints if a household fell into either category. This approach to measure these variables was chosen because this is a consistent measurement across other similar studies. In total, 481 households (20% of the sample) reported being liquidity constrained.

The analysis explored the impact of income risk and liquidity constraints on stock market participation, controlling for the correlation between earnings and financial returns. A key finding was that income risk only negatively impacted stock ownership for households reporting a non-negative correlation between earnings and market returns. Households perceiving a negative correlation exhibited an ambiguous relationship—higher income risk leading to less participation, but a negatively correlated income providing a hedge against market fluctuations. The research found that stock ownership was significantly less likely among those reporting a non-negative correlation and those who were liquidity constrained. The direct stockholding rate was 21%, rising to 33% when indirect holdings were included. These participation rates, while low, were comparable to prior INSEE surveys and those of other European countries. This finding is important in that it highlights the critical importance of understanding how the correlation between income and market risk affects household investment decisions.

Regression results (heteroscedastic probit estimation) showed that financial wealth and income positively influenced stock market participation. Homeownership also had a positive effect. The results support the notion of transaction costs and information barriers in accessing the stock market. Information access (internet, parental stock ownership, living in Paris) positively correlated with participation. Age had a hump-shaped effect, peaking around 46. Liquidity constraints decreased participation, reinforcing the risk substitution effect. The self-assessed correlation proxy behaved as predicted, with households reporting a negative correlation showing greater participation. Possible explanations for this were redistribution shocks dominating economic cycles (Botazzi et al., Danthine & Donaldson) and a skill premium effect for educated households with counter-cyclical earnings (Rubinstein & Tsiddon, Palacios-Huerta). Importantly, when controlling for this correlation, the study did find evidence that income risk has a significant effect on stock ownership decisions. While the impact of income risk was not significant in the baseline model, these findings highlight the importance of adjusting for the correlation between these two risk factors when conducting this type of research.

The conclusion reiterates the study's key finding: the empirical results confirm the theoretical prediction that non-negatively correlated background risks (specifically, income risk) reduce the willingness of households to bear financial risk, leading to decreased stock market participation. This aligns with the theoretical framework presented earlier and addresses the existing ambiguity in the literature regarding the impact of income risk on stock ownership. The study's findings highlight the importance of considering the correlation between earnings and market returns, particularly as it affects household investment decisions. This contrasts with some previous studies that failed to find a significant effect of income risk, likely due to their not accounting for the influence of the correlation between income and market risk, which this study does incorporate.

The conclusion also emphasizes the significant role of borrowing and liquidity constraints in reducing households' propensity to invest in risky assets. This confirms previous research and theoretical findings related to the impact of liquidity constraints on financial risk-taking behavior. The research confirms that households facing liquidity constraints are less likely to invest in risky assets due to lower risk tolerance. The analysis used a variety of metrics to measure risk and financial constraints and highlights the contribution of these factors to our overall understanding of the household decision-making process. The study's findings also quantitatively confirm the relative importance of transaction costs, information barriers, risk aversion parameters, and income risk in explaining households' stock market participation decisions, reiterating existing findings in the literature. The conclusion highlights the importance of considering both income risk and liquidity constraints in understanding the drivers of stock market participation.

The study quantifies the effects of the various factors, confirming their relative importance as observed in prior literature. Transaction costs and informational barriers, liquidity constraints, risk aversion parameters, and income risk all significantly influence households' decisions to participate in the stock market. The relative magnitudes of these effects are consistent with previous research, emphasizing the importance of each in explaining the observed behavior. The study's use of the DELTA-TNS 2002 survey on French households, coupled with its specific methodology to handle the critical correlation between income and market risks, provides new insights. While the study acknowledges the imperfect nature of the correlation proxy, it highlights its value in correctly predicting household behavior. The conclusion suggests that this framework could benefit future research, potentially using longitudinal data to validate the self-assessed correlation findings and enhance overall model accuracy.