RDP 2013-05: Liquidity Shocks and the US Housing Credit Crisis of 2007–2008 1. Introduction

There is extensive anecdotal evidence to suggest that a significant tightening in credit conditions, or a ‘credit crunch’, occurred in the US housing market following the collapse of the loan securitisation market in late 2007. For example, the Federal Reserve's Senior Loan Officer Opinion Survey indicates that the number of US banks that tightened lending standards rose sharply for both prime and subprime mortgages in the December quarter 2007 (Figure 1). The term ‘credit crunch’ has become so commonplace that the Economist magazine has created a Credit Crunch board game and the term is now officially part of the English language, having been recently included in the Concise Oxford English Dictionary.

Despite its popularity as a concept, there has been surprisingly little formal testing of whether a credit crunch, in fact, occurred in the US housing market in 2007–2008. This is probably because the necessary conditions to test the hypothesis are quite strict. The most generally accepted definition of a credit crunch is attributable to Bernanke and Lown (1991) who define it as a ‘significant leftward shift in the supply curve for bank loans, holding constant both the safe real interest rate and the quality of potential borrowers’ (p 207). This definition requires two key conditions to be satisfied to establish a credit crunch: first, there should be a fall in credit that is caused by a decline in credit supply rather than demand; and, second, the fall in credit supply must be exogenous in the sense that it is not caused by an increase in the credit risk of potential borrowers. In other words, the fall in credit supply will generally be caused by factors affecting the size and composition of financial institutions' balance sheets, such as a tightening in financing conditions.^[1]

The other reason why there has been little formal testing for a credit crunch is that it is difficult to conduct adequate econometric tests. There are two common econometric problems in identifying a credit crunch. First, a crunch typically coincides with a general decline in economic activity, which also causes the demand for credit to fall (simultaneity bias). Second, even if the decline in credit can be traced to a fall in supply, this may be an endogenous response by lenders to a decline in the quality of potential borrowers associated with the economic downturn (selection bias). The difficulty in separately identifying the effects of changes in credit supply and demand is highlighted by the Federal Reserve's Senior Loan Officer Survey, which shows that the demand for mortgage credit also fell sharply around 2007–2008 (Figure 2). It is therefore possible that this decline in demand drove the overall fall in lending rather than a decrease in credit supply.

This paper tests whether the fall in mortgage credit over 2007–2008 was caused by a reduction in credit supply that, in turn, can be traced to a fall in the level of financing available to US mortgage lenders, which caused them to become liquidity constrained. I will refer to this as the ‘liquidity constraints hypothesis’. I use application-level information on new mortgage loans to assess how US mortgage lenders' lending behaviour changed as a result of the tightening in financing conditions.

In the first part, I estimate a model that identifies the liquidity shock based on each mortgage lender's reliance on securitised lending in the period prior to 2007. In this framework, the closure of the securitisation market acts as the ‘treatment’, the ‘treatment group’ are the mortgage lenders that were reliant on securitisation before 2007 and the ‘control group’ are the mortgage lenders that were not dependent on securitisation. I will refer to the treated lenders (those reliant on securitised lending) as the originate-to-distribute (or OTD) lenders, and the control group of lenders, which were not dependent on securitised lending, as the non-OTD lenders.

The novel aspect of this study is that the causal effect of the liquidity shock on mortgage lending is identified through variation in the lending activity of OTD and non-OTD lenders that grant credit to the same borrower (where a particular region is broadly defined as a ‘borrower’). Specifically, I assume that mortgage lenders that originate loans in the same Census tract (a tract is similar to a postcode) face the same demand conditions and the same risk profile of loan applicants. Under this assumption, a reduction in credit by OTD lenders relative to non-OTD lenders within a tract implies that a negative lender liquidity shock, and hence a decline in credit supply, caused the overall fall in mortgage credit.

I find strong evidence to indicate that the OTD lenders disproportionately reduced mortgage credit supply following the liquidity shock. The negative liquidity shock caused by the shutdown of the securitisation market explains about 14 per cent of the average decline in mortgage credit during the crisis. Moreover, I find that the link between lender funding liquidity and mortgage lending holds even after controlling for unobservable lender characteristics, such as changes in banks' assessment of borrowers' risks.

I also examine which borrowers were most affected by the reduction in credit supply due to the tightening in lender financing conditions. Theory suggests that lenders re-balance their portfolios towards less risky loans when economic conditions deteriorate (Bernanke, Gertler and Gilchrist 1996). During recessions, the share of credit flowing to borrowers with more severe asymmetric information and agency problems, such as small firms, decreases. This ‘flight to quality’ has been identified in a range of empirical studies (e.g. Lang and Nakamura 1995; Popov and Udell 2010). However, more recent research indicates that there may also be a ‘flight to home’ effect when economic conditions deteriorate (e.g. De Haas and Van Horen 2012; Giannetti and Laeven 2012). Specifically, lenders re-balance their asset portfolios towards local borrowers when the economy weakens, as lenders are typically better informed about local borrowers than non-local borrowers. The flight to home effect co-exists with, but is distinct from, the flight to quality effect. To the best of my knowledge, this paper is the first to examine whether the flight to quality and flight to home effects are relevant to residential mortgage lending.

I find that all mortgage lenders reduced credit to risky borrowers, though the effect was not disproportionately larger for the OTD lenders. This points to a general flight to quality by US mortgage lenders during the crisis. I find limited evidence for a flight to home caused by the liquidity shock; while the OTD lenders increased the share of credit to local borrowers relative to the non-OTD lenders, the differential effect is not significant.

Footnote

It is effectively a leftward shift of the credit supply curve where the quantity of credit is measured on the x-axis and the loan interest spread is measured on the y-axis. [1]