Transaction costs, Income Risk and Household Portfolio Allocation: Evidence from French Panel Data

This paper has considered the dynamics of financial portfolio choice by French households, using data from the ECHP over the period 1994-2001. The panel data allow us to test a number of theoretical predictions which are difficult to analyse with cross-section data: life cycle effects, importance of transaction costs, temperance behavior. The age effects that we estimate are corrected for cohort effects: in particular, the profile for equities is positive in cross-section, but hump-shaped in panel. The Mundlak approach to correct for correlation between the individual error terms and the explanatory variables, allows us to estimate separately average (permanent) and transitory (current) effects for those variables which vary over time. For the case of income, this specification allows us to consider the influence of income risk on portfolio choice. We find a positive (even small) effect of transitory income on the probability of holding risky assets: households with the riskiest income also have riskier portfolios. So, these result contredict risk substitution theory. Last, the inclusion of lagged dependent variables underlines the importance of transaction costs in household portfolio choice, notably with respect to the restricted degree of diversification typically found in survey data. These costs seem particularly important for equities.