Abstract: (preliminary; no paper yet)

Housing is most homeowners’ most important non-human asset, and housing costs are one of the largest determinants of the cost of living.  House prices and housing costs are both subject to substantial risk, the costs of which depend crucially on their correlations with other exposures in households’ portfolios.  Using Census data from 1940–2010, we estimate the relationships between house prices, rents, and wages.  We find substantial location-specific risk in these prices, the patterns of which are consistent with spatial equilibrium models that feature large labor demand shocks. In particular, a location’s rents and house prices covary strongly positively with its wages. This implies that for a typical working household, owning is riskier than renting.  Relative to having no exposure to the housing market, renting decreases total risk exposure because rent risk hedges labor earnings risk.  Owning, by contrast, increases total risk exposure because risk in the eventual sales price of the house is both large and strongly positively correlated with wages.  Analyses of a life-cycle model suggest that the risk-exposure advantage of renting is large on average and highly dependent on a few key characteristics of households and locations. These results have important implications for government policies that affect housing tenure choices and the elasticity of housing supply