«Elena Loutskina University of Virginia, Darden School of Business & Philip E. Strahan Boston College, Wharton Financial Institutions Center & NBER ...»
FINANCIAL INTEGRATION, HOUSING AND ECONOMIC VOLATILITY
University of Virginia, Darden School of Business
Philip E. Strahan
Boston College, Wharton Financial Institutions Center & NBER
The Financial Crisis and the Great Recession illustrate the sensitivity of the economy to a
housing bust. This paper shows that financial integration, fostered by deregulation allowing
banks to form nationwide branch networks, amplified housing-price volatility and increased the economy’s sensitivity to local housing-price shocks. We exploit variation in credit-supply subsidies across local markets from the Government-Sponsored Enterprises to measure housing price changes unrelated to fundamentals. Using this instrument, we find that a 1% rise in housing prices causes a 0.25% increase in economic growth. This effect is larger in localities more financially integrated with other markets through bank ownership ties. Financial integration thus raised the effect of collateral shocks on the economy, thereby increasing economic volatility.
Financial integration may dampen or amplify economic shocks. Morgan, Rime and Strahan (2004) – MRS hereafter – show theoretically that integration’s effect on volatility depends on the sources and magnitudes of shocks hitting the local economy. With integration, local economies become more insulated from shocks to the supply of local finance (e.g. local bank capital). During the 1980s and early 1990s, these shocks were a major source of businesscycle instability (Bernanke and Lown, 1991). For example, the number of bank and S&L failures during the 1980s averages more than 150 per year (Kroszner and Strahan, 2008), and the
Integration makes local economies less sensitive to these financial disturbances because capital can flow in from external sources and thus allow investment to continue, even if local lenders are distressed. MRS show empirically that state-level banking integration fostered by deregulation during the 1970s and 80s lowered volatility of local economies in these years.
MRS’s theoretical model, however, also shows that integration, by allowing financial capital to flow away from depressed areas and into booming ones, can amplify local cycles. For example, if collateral values rise sharply in a locality, borrower debt capacity and demand for credit increases; integration helps bring financial resources from abroad to satisfy higher credit demand. The influx of credit from external sources raises growth above what would have been possible in a stand-alone, or dis-integrated, financial system. These flows correspondingly reduce collateral values from areas with relatively weak credit demand because these market face capital outflows. Thus, capital flows generated by credit demand shocks will reduce comovements in collateral values across financially integrated markets.
Beyond its effects on capital flows, integration is also associated with lower investment by lenders in private information about local business conditions, borrower credit quality and housing-price fundamentals (Loutskina and Strahan, 2011; Romero-Cortes, 2011). As a result of securitization, for example, residential mortgage credit supply responds more now to changes in the market value of collateral than in the past because lenders condition their credit decisions more on public signals (e.g. borrower FICO scores and loan-to-value ratios) and less on private information (Rajan, Seru and Vig, 2010). Both of these forces – more ‘flighty’ capital and more reliance on public information – may increase collateral volatility and raise the sensitivity of local cycles to variation in collateral values. Consistent with these ideas, we find that financial
sector have a quantitatively substantial causal impact on local economies, and that the transmission of these housing-price shocks increases with financial integration.
The analysis proceeds in three steps. First, we document a positive relationship between financial integration and the magnitude of local house-price shocks. To do so, we measure financial integration at the level of the Central Business Statistical Area (CBSA), the US Census Bureau’s definition of a city. The measure (In-CBSA ratio) is based on the ownership of bank branches across CBSAs, equal to the fraction of local deposits owned by a banking company also owning branches in other CBSA markets. So, a CBSA in which all of its branches are owned by banks with branches in other CBSAs would have In-CBSA ratio = 100%.
We find that the volatility of shocks to CBSA-level housing price growth increases with financial integration. The effect increases in magnitude when we use variation across states in restrictions on interstate branching as an instrument for financial integration (Rice and Strahan, 2010). Thus, there is a robust difference in local house-price volatility between more- and lessintegrated local markets. This result reverses that of MRS, who use data from the 1970s and 1980s, when shocks to the financial sector were an important source of business-cycle variation.1 Our results, however, are consistent with the theoretical argument that, in the absence of shocks to financial institutions, integration amplifies the impact of collateral shocks. To test this mechanism, we compare shocks for all unique pairs of local markets. If integration increases capital flightiness in response to collateral values shocks, then integration between pairs of markets ought to reduce the correlation between shocks across markets. Using housing price
1 Like MRS, we have also tested whether the amount of deposits in external markets, as a second integration measure, affects volatility. This second integration measure is also positively related to volatility in some specifications, although its magnitude is smaller and less significant than our primary integration measure.
with each other have less similar changes in housing prices, controlling for trends (time dummies), for pair-wise fixed effects and for the similarity of industry composition. Again, we find that the effects increase in magnitude when we instrument for integration using a pair-wise combination of each area’s regulatory stance toward interstate branching.2 In the second part of the analysis, we build an instrument for house-price appreciation that exploits the importance of the Government-Sponsored Enterprises (GSEs) – Fannie Mae and Freddie Mac – in housing finance. Fannie and Freddie subsidize mortgage credit, but only for mortgages that fall below the jumbo-loan threshold (Loutskina and Strahan, 2009). Borrowers with housing demand near the jumbo-loan threshold stand to benefit from an increase in the threshold, leading to an increase in housing demand and housing prices (Adelino, Schoar and Severino, 2011). While the jumbo-loan cutoff changes uniformly across CBSAs, its effects vary across markets. For example, in Los Angeles - where about 5.3% of mortgages were made to borrowers within 5% of the jumbo-loan cutoff - the change in cut-off would have a bigger impact than in Wichita, Kansas - where this fraction was about 0.5%. Since there is both crosssectional and time-series variation in the amount of such demand (e.g. LA v. Wichita), we generate a set of instruments based on the product of the sensitivity to changes in the jumbo-loan cutoff in market i during year t-1 times the change in the cutoff itself between years t-1 and t.
The instruments depend only on the distribution of mortgage credit during the preceding year and the change in the jumbo-loan cutoff during the current year, which is the same across all local markets and depends mechanically on lags of increases in nationwide prices. Furthermore, we exploit the elasticity of the housing supply across different geographies to better capture the
2 Kalemni, Papaionnou and Peydro (2010) find similar effects following financial integration across 20 developed economies.
that these instruments pick up variation in changes in housing demand exogenous to overall economic fundamental in the local area.
We find that these instruments are powerful. Local housing prices appreciate faster in markets where credit on jumbo borrowers was more constrained in the prior year, based on the distribution of borrowers around the jumbo cutoff. This effect is stronger in markets with relatively inelastic housing supply because prices are more sensitive to changes in demand where the physical supply of housing is limited by geographic barriers.
Armed with exogenous variation in housing prices, the third part of the analysis shows that housing prices have a strong causal impact on local economic growth in employment and output. In our base model, a 1% increase in housing prices causes an increase in local GDP growth of about 0.25% and an increase in non-construction, non-finance employment growth of about 0.15%. The latter effect implies that higher prices spill over to sectors not directly affected by housing. We then show that the effects of house-price shocks are stronger in local markets with high levels of financial integration than in markets with low integration. In local areas onestandard deviation above the mean level of financial integration, a 1% housing price shock leads to a 0.30% increase in GDP growth. Taken together – higher housing price volatility and increased sensitivity to house-price shocks – the results imply that financial integration has increased economic volatility, both by amplifying variation in collateral values (house prices) and by strengthening links from collateral to the overall economy.
Our paper contributes to three strands of the literature. First, the effect of financial integration on economic volatility has been explored both across US states and also in the
Demyanyk, Ostergaard and Sorenson, 2007; Kalemni, Papaionnou and Peydro (2010)). We find that integration can amplify shocks and de-sychronize asset markets in an environment of strong credit demand and a profitable financial sector. In other settings, where financial shocks are important, integration can increase synchronization because credit supply shocks propagate across connected markets (e.g. Peek and Rosengren, 2000). Second, conventional explanations for the US housing boom blame loose lending practices as a key driver of price appreciation (e.g., Mian and Sufi (2009), Keys et al (2010), Demyanyk and Van Hemert (2010), Loutskina and Strahan (2011)). Yet these studies do little to explain why booms were concentrated in places like as Florida, Arizona and California. Financial integration can rationalize regional booms by allowing capital to flow into areas with strong credit demand.
Third, many have argued that the so-called ‘Great Recession’ has its root in the crash of housing prices beginning in the middle of 2006. Our results are consistent with this explanation but also suggest that the economic boom was itself fueled by house-price appreciation. The findings extend the work of Mian and Sufi (2009 and 2011), who show that household debt and consumption were strongly correlated with house-price appreciation during the boom.
Conversely, declines in consumer spending and financial distress across local markets during the bust are also associated with declines in housing equity. Unlike Mian and Sufi (2011), however, we go a step further and estimate the total effect of housing price shocks on the economy, and we condition this estimate on aspects of the financial system. Shocks to housing have had a large effect on the overall economy, especially in markets that are well integrated nationally.
In the next section we briefly review the forces leading to increased integration over time.
In Section III, we describe our integration measures in detail, and document their link to local
growth. Here, we first establish a first-stage model that relates changes in credit-supply subsidies from the GSEs to house-price appreciation. We then use this model to generate an instrument for housing price changes to estimate its causal impact on the economy as a whole.
Section V concludes.