Using detailed disaggregated Swedish household administrative data on portfolio holdings and labor income, this paper investigates retail investors’ behavior of seeking skewness in their portfolio choice. I develop a model of rational portfolio choice in which investors optimally hold portfolios with a (positively) skewed return distribution to hedge against (negatively) skewed labor income risk. I find empirical support for the model’s predictions. I find that investors trade off their portfolio’s Sharpe ratio against higher skewness, which explains the suboptimal Sharpe ratio found in previous studies. I also find that skewness seeking is more pronounced for investors with (i) higher overall risk in their labor income, (ii) higher downside risk in their labor income, and (iii) less wealth. Further, I find that investors hold more assets that provide insurance against the time-varying downside risk in their labor income.
Cyclical Background Risk and Portfolio Choices: Evidence from Sweden
with Sylvain Catherine and Paolo Sodini
Using Swedish administrative panel data on individual’s wages and portfolio holdings, we show that countercyclical labor income downside risk reduces households’ willingness to invest in financial market. We start by computing the cross-sectional variance and skewness of wage growth by occupation and year from 2001 to 2013. Then, we show that occupations for which these measures of labor income risk correlate with stock market fluctuations have lower participation rates and invest a smaller share of their financial wealth in risky asset. In line with theoretical predictions, these effects are stronger for individuals with modest financial wealth. Finally, we also show that households invest less in assets whose returns negatively correlates with downside risk in their profession.
Multifractal Volatility with Shot-Noise Component
with Laurent Calvet
We develop in this paper a discrete-time multifractal volatility model. Based on the Markov Switching Multifractal (MSM) model of Calvet and Fisher (2004), we construct a more realistic multifractal model, labeled Shot-Noise Multifractal. The model is designed to capture the jump and decay pattern in volatility process along with other stylized facts. By construction, both GARCH class and MSM model fail to capture the fact that volatility level jumps up after exogenous shock and has gradually reversed. Compared to GARCH and MSM model, the Shot-Noise Multifractal model is able to generate more jumps in the process and have fatter tails of return. Statistically, SNM performs better than GARCH and MSM both in and out of the sample.