The Geography of Housing Market Liquidity During the Great Recession

Abstract

Using detailed micro data at the ZIP code level, this article explores the regional variation in housing market performance to account for the severity of the Great Recession. The granularity of the data, relative to a more traditional analysis at the county level, is useful for evaluating the performance of the housing market because credit and local macroeconomic variables are tied to housing valuations. The deterioration of the ability to transact (buy and sell) housing units, often referred to as housing liquidity, is an important link that connects housing outcomes with real and credit variables. The data indicate that the timing, severity, and duration of the recession varied across regions and was closely connected with the behavior of the housing market. The deterioration in housing liquidity was uniform across all house price tiers (i.e., bottom, middle, and upper end). Furthermore, there was correlation across areas between the magnitude of the declines in housing liquidity and the severity of the deterioration in house prices and macroeconomic conditions.

Introduction

The historic deterioration of housing markets during the Great Recession was characterized by plummeting home values and skyrocketing foreclosure rates. Although almost no place in the Unites States escaped unscathed, significant heterogeneity emerged across both time and space with regard to the housing market collapse. An important aspect missing in many theoretical and empirical studies of this episode is the role of liquidity in local housing markets as manifested in homes taking longer to sell. 

Examining the dynamics of ZIP-code-level data reveals that the severity, timing, and length of the Great Recession varied across regions and is linked to housing liquidity. Through­out this article, our main measure of housing liquidity is time on the market (TOM), the number of days between listing a property and selling a property. A mismatch between the number of sellers and buyers changes the transactability of houses beyond house prices. Hous­ing liquidity responds to changing macroeconomic conditions, resulting in selling delays. During the Great Recession, the deterioration in housing liquidity along with falling prices generated an imbalance between assets and liabilities, creating a debt overhang. To rebalance their portfolios, households needed to adjust their spending, liquidate leveraged assets (i.e., housing), and in some instances enter foreclosure, which in turn induced lenders to contract credit.

The National Bureau of Economic Research classifies the Great Recession as occurring from December 2007 to June 2009. However, the national housing market experienced a severe downturn nationally from at least 2006 to 2011. House prices deteriorated and mortgage delinquency increased as early as 2005 in the regions first affected by the crisis. The earlier start date for the housing crisis is not unexpected, because the subprime mortgage crisis was an important catalyst of the financial recession that followed. Besides starting earlier than the accepted recession dates, the housing market did not fully recover to pre-crisis levels until several years after the official end of the recession: Most areas did not see house prices, liquidity, or income recover to pre-recession levels until after 2011. 

Within this window of 2005 to 2011, the various regions of the United States experienced different timing of the start, trough, and end of the housing crisis. For example, the state of California witnessed fairly early declines in housing market indicators but recovered more rapidly than many other regions. Meanwhile, states in the Sunbelt, particularly Florida, Arizona, and Nevada (hereafter called the Sunbelt states), also witnessed early large declines in house prices and increases in illiquidity and mortgage delinquency. These Sunbelt states did not recover until well after the official end date of the recession. Documenting the regional variation in housing market responses (and in particular housing liquidity) during the Great Recession is one of the main goals of this article. 

The use of granular data is crucial to the study of housing. Housing markets by nature are disaggregated, as there can be significant variation within counties or metropolitan statistical areas (MSAs). The use of granular data is particularly important when looking at regional variation in the United States. Many western states have significantly larger counties by area than eastern states. Aggregating data to the MSA level, or even to the county level, causes any empirical analysis to ignore significant variation within counties. To alleviate these concerns, the analysis in this article uses ZIP-code-level data. 

Several different empirical specifications establish a strong link between explanatory housing market variables (house prices and liquidity) and local outcomes for the credit market (mortgage delinquency) and macroeconomy (income). Moreover, categorizing the geography of the housing crisis into three large regions (California, the Sunbelt states, and the other states) generates heterogeneous effects that substantially differ from those at the national level. For example, in California, house prices declined by more than the national average and mortgage delinquency increased far more than the national average. However, liquidity and income declined by less than the national averages. Although the trough was later in the Sunbelt states, income declined by large amounts and homes' TOM and mortgage delinquencies increased. 

The regression models exploit regional time windows to arrive at (non-causal) estimates of changes in house prices and liquidity on these outcomes. In all cases, the effects were highly statistically and economically significant. For example, at the national level, a one-month increase in TOM is associated with approximately a 1 percent decline in real income. To put these estimates into perspective, a 10 percent decline in house prices is associated with approximately a 1.8 percent decline in real income. For the Sunbelt states, the liquidity effects are slightly larger. Housing liquidity also has strong implications for the mortgage delinquency rate. In California, a one-month increase in TOM is associated with a 2.0 percent increase in the mortgage delinquency rate. Given this observed increase in TOM, the estimate for California predicts a 13.5 increase in mortgage delinquency. For the Sunbelt states, the estimate is 17 percent.

We also estimate alternate specifications of the models by dividing the United States into regions by housing supply elasticity rather than by geography. These estimates provide a counterfactual to the regional analysis and support the claim that liquidity was a highly significant factor in areas with a low housing supply elasticity. We also find that in areas with relatively high housing supply elasticity, liquidity was significant but had a smaller effect on outcomes. We detail in Section VI a brief theory of the interaction of housing liquidity and elasticity.