Google search intensity, mortgage default and house prices in regional residential markets

Abstract

The internet provides a new way for households to access relevant information, while their online search behaviour may also contain information for their concerns and intentions, or even be used to predict real economic activity. This thesis explores the use of Google search data to predict mortgage default and regional house price dynamics in an empirical macroeconomic framework. The thesis is composed of three independent empirical studies.

The first study examines the dynamic interdependence between mortgage default and house price across different housing market segments, i.e., top-tier vs. bottom-tier houses and recourse states vs. non-recourse states, based on a Panel VAR model. In particular, this study uses the Mortgage Default Risk Index (MDRI) proposed by Chauvet et al. (2016). It captures the intensity of Google search for keywords and phrases such as “mortgage foreclosure” or “foreclosure help” and measures the potential default risk of households. It is shown that shocks to house price returns have a significantly stronger effect on actual foreclosures in non-recourse states than in recourse states. The results suggest that borrowers are financially sophisticated and strategic as they are less likely to default in recourse states. Additionally, the MDRI has a stronger negative impact on top-tier home price returns, while the foreclosure rate of homes more pronouncedly decreases bottom-tier home price returns. These findings hold for the entire sample and recourse states. However, in non-recourse states, the impacts of the MDRI and the HF on bottom- and top-tier house price returns are about the same.

The second study examines the impact of house prices on the foreclosure rates in the local housing market and explores whether the MDRI helps predict future house prices and foreclosures. In particular, this study uses an error correction framework to capture both the long-run equilibrium fundamental component of house prices as well as the short-run dynamics of house prices, including the component of bubbles. It is found that the MDRI shows a negative impact on both components of house prices but, more importantly, a negative impact on foreclosure rates. Furthermore, it is shown that foreclosure rates are negatively affected by the fundamental component of house prices but are not sensitive to their bubble component. This study sheds new light on the predictive power of household sentiment derived from Google searches on prices and foreclosure rates in local housing markets.

The third study recognizes that, by searching online, households are transmitting information to and simultaneously receiving information from the Google Search engine. While they might divulge information about their financial concerns or vulnerability, they are also gathering information and learning through their search behaviour. This chapter aims to examine the comprehensive impact of the disclosure and information-learning effects of online searches on mortgage default. To that end, based on the assumption of different pre-existing knowledge of households, this study defines two kinds of Google search activities of households, i.e., naïve and sophisticated searches, and practically performed by aggregating the search activities for different query terms. It is found that sophisticated search activity has a positive impact on mortgage delinquency but a negative impact on foreclosure starts, while naïve search activity only positively affects foreclosure starts. The results suggest that the Google search activity of households is a combination of information disclosure and information-learning processes. Furthermore, borrowers are more likely to learn from sophisticated online searches, and they can use the information to avoid foreclosure starts.