Housing and Financial Asset Allocations of Heterogeneous Homeowners

Market constraints (e.g. borrowing constraint, no-short-selling constraint) are important for household portfolio choices especially for housing. A structural model naturally generates alternative portfolio regimes with different binding constraints. But empirically we cannot observe which households are constrained in safe, risky or housing finance and housing. We use a semiparametric approach on data from Wealth and Asset Survey (WAS) to determine this. We find distinct patterns of housing and financial assets allocation among homeowners by fitting a multivariate Gaussian mixture model via a censored data expectation-maximisation (EM) algorithm. Estimation results reveal that on average about 80% of the households are no-short-selling constrained in risky asset investment and with low net worth. Among other things, we find that households who are younger, less educated with lower income are more likely to be no-short-selling constrained in risky asset investment and with lower net worth. Our predicted regime classification is aligned to those of the structural model.