Relative risk aversion once more: An analysis of Japanese households' financial asset holding pattern

It is widely accepted that as the total assets increse, households tend to diversify their portfolios. In other words, absolute risk aversion is decreasing. On the other hand, the proportion of risky assets may increase or decrease depending on whether relative risk aversion (RRA) is decreasing or increasing, and its direction is still left open as an empirical question. This study examines the constancy of RRA from Japanese individual households' financial asset holding data collected in 1984. Constant RRA implies that the proportion of risky assets in one's portfolio is constant regardless of the amount of total assets. A casual observation of household portfolio holding pattern suggests that this implication is clearly violated by the data, because there are substantial proportion of households which do not hold any risky assets. Zero-holding, however, may be interpreted as a result of fixed transaction cost incurred by individual investors when they hold risky assets. Then, we pose a question, ‘Do investors hold constant proportion of risky assets, when they decide to hold them?’ In order to explain a substantial number of zero risky asset holders in the sample, we propose a portfolio selection model with a transaction cost, and estimate the model using a variant of Heckman's two-step method. In estimation we control for individual investors' socioeconomic characteristics, as well as income and total assets. The construction of the model imposes nonlinear restrictions on the two estimators, from which we can test the specification of the model. The estimation results suggest that there is a statistically significant decreasing tendency linked to total assets but that its rate of change tapers off as total assets increase. Our results are consistent with the previous studies which tended to support constant RRA for the higher asset holders, and complement previous studies in explaining lower asset holders' investment behavior.