Governments around the world have a common and expensive feature of their tax codes: tax subsidies to homeownership. In many countries, the deductibility of mortgage interest provides a significant subsidy to ownership. In the United States, this tax subsidy amounted to $101 billion in 2014, which is more than 20% as large as the entire federal deficit in that year (Gruber 2015). These subsidies are generally motivated by the perceived externalities of homeownership. As Glaeser & Shapiro (2003) write: “To its supporters, the home mortgage interest deduction is the cornerstone of American society. Homeownership gives people a stake in society and induces them to care about their neighborhoods and towns. By subsidizing property ownership, the deduction induces people to invest and then to have a stake in our democracy. Ownership makes people vote for long-run investments instead of short-run transfers.” Glaeser & Shapiro (2003) review a large body of empirical evidence that supports the existence of these positive externalities: homeownership is positively correlated with political activism and social connection, and homeowners take better care of their properties, leading to higher values for surrounding houses. While such correlations are suggestive, there is little causal evidence for the existence of externalities due to owning rather than renting.1 Any externality-related argument for the mortgage interest deduction requires that the policy is able to increase homeownership. Yet, we have relatively little evidence on how these tax subsidies impact housing decisions. There is a sizeable literature, reviewed below, on the effects of tax subsidies on financial decisions such as indebtedness. But there is little evidence on how these expensive subsidies impact the decision to purchase a house and the characteristics of that house. The lack of evidence on real housing responses is likely due to the fact that a convincing empirical estimate must meet three requirements. The first is exogenous variation in the mortgage interest deduction that is sufficiently large to be able to detect any effects. The second is micro data that matches tax records to real information on housing decisions. And the third is a sufficiently long time period to capture the long-run effect on homeownership, which tends to be a slow-moving outcome.
In this paper, we focus on a setting and natural experiment that meet these three criteria. The experiment is a major tax reform in Denmark in the late 1980s. This reform significantly reduced the subsidy to negative capital income (from mortgages and other borrowing) for taxpayers in the top bracket, while reducing it much less for taxpayers in a middle bracket and raising it slightly for taxpayers in a bottom bracket.
Read Study