Using data from the US Health and Retirement Study, we study the causal effect of increased health insurance coverage through Medicare and the associated reduction in health-related background risk on financial risk-taking. Given the onset of Medicare at age 65, we identify our effect of interest using a regression discontinuity approach. We find that getting Medicare coverage induces stockholding for those with at least some college education, but not for their less-educated counterparts. Hence, our results indicate that a reduction in background risk induces financial risk-taking in individuals for whom informational and pecuniary stock market participation costs are relatively low. (Gollier and Pratt, 1996). Risks related to income, entrepreneurship and health have often been suggested as instances of a background risk that is negatively associated with risky asset ownership. 3 A lower background risk, however, may not suffice to induce investment in risky assets. In fact, in a number of standard life-cycle portfolio models incorporating background risk the optimal level of risky assets is zero after the introduction of participation costs (Haliassos and Bertaut, 1995;Vissing-Jørgensen, 2002).
4Stock market participation costs can be both pecuniary (e.g., brokerage fees) and nonpecuniary (e.g., time spent to find the most suitable assets to invest in, to consult with financial advisors, to monitor market developments), and typically vary by education. A higher level of human capital is typically associated with higher financial resources and more efficient information processing, making both aforementioned costs easier to bear. Hence, it is natural to expect the impact of a reduction in background risk on stockholding to differ across education groups due to the education-induced variation in stock market participation costs.We find that Medicare eligibility induces individuals with at least some college education to invest in stocks. Our preferred estimates suggest an increase in total stockholding, ranging from 12 to about 25 percentage points for this education group, depending on the method used. This is in line with the increase in stockholding prevalence observed in the data for this group. On the other hand, we find no effect of Medicare on 3 See e.g., Guiso, Jappelli and Terlizzese (1996), Heaton and Lucas (2000), Viceira (2001), Rosen and Wu (2004), Edwards (2008), and Yogo (2009. 4 The intuition for this result is as follows (see Haliassos, 2002 for a more detailed exposition): given that expected returns from stocks exceed those of riskless assets, a household will be discouraged from stock investment only because stockholding increases too much the riskiness of consumption. When the household invests no money in stocks, however, stocks returns are not correlated with consumption, and thus at the margin of zero stock investment the household should prefer to invest in stocks rather than in a riskless asset in order to take advantage of the equity premium.3 stockholding for those without a...