This paper estimates the effect of marginal tax rates on demands for various financial assets, using the 1983-1989 Survey of Consumer Finances (SCF) panel. In the cross-section, marginal tax rates appear to have a strong and statistically significant influence on portfolio allocations, even after controlling for income and wealth. For example, people with high tax rates tend to hold larger shares of their assets in equities, which are taxed relatively lightly due to the treatment of capital gains. Cross-sectional estimates could be biased, however, if the marginal tax rate is correlated with unobserved influences on portfolio choice, such as financial sophistication, preferences for risk or liquidity, or degree to which portfolio choices are constrained by Social Security. These unobserved variables, like the marginal tax rate, are likely to be related to income and wealth in a positive and non-linear fashion, so that the marginal tax rate may serve as a proxy for them. This paper utilizes the SCF panel to estimate fixed-effects and correlated random-effects (Chamberlain, 1984) models, which effectively control for this unobserved heterogeneity. Identification is provided by the Tax Reform Act of 1986, which changed marginal tax rates by very different amounts for different people. The empirical results confirm that, without controlling for unobserved heterogeneity, marginal tax rates have a strong influence on portfolio choices. After controlling for unobserved heterogeneity, however, the magnitude of the effect is much smaller and often statistically insignificant. These results are found to be robust to efforts to distinguish between unobserved heterogeneity and state dependence (i.e., slow adjustment of portfolios).
