Reexamining the Empirical Relevance of Habit Formation Preferences

We reexamine the empirical relevance of habit formation preferences with micro-data on households' portfolio choices. We first derive the analytical solution to the risky asset share in a theoretical model with both habits and time-varying labor income. Our analytical results indicate that (1) for each household, there are two channels through which the risky asset share responds to wealth fluctuations, habits and household income; (2) across households, there are heterogenous responses through the habit channel: those who experience large negative income shocks reduce their share of risky assets; and (3) two potential mis-identification problems arise when both two channels and the heterogeneity are ignored. Contrary to the existing literature, our empirical results find positive evidence of habit formation preferences after correcting the two mis-identification problems.