The mortgage market was the starting point for several of the post-Lehman crises: the subprime crisis, the Irish crisis, the Spanish crisis, and many more. It is a market typified by massive information asymmetries, and it has been argued that a market based on highly asymmetric information contributed to the buildup of bad mortgage debt during the first half of the last decade.
In tracking down the links between asymmetry of information and bad mortgage debt, most research has focused on asymmetric information concerning repayment characteristics of borrowers (Elul 2009, Keys et al. 2010). Similarly, much of the recent policy discussion that has aimed to address the issue of asymmetric information in mortgage lending has focused on proposals that alleviate information asymmetries about borrower characteristics, for instance the debate on the removal of low-documentation mortgages.
A different perspective, collateral value of property
But there is another key source of information asymmetry: the collateral value of properties. Due to the illiquid and heterogeneous nature of housing, knowing a property’s true worth is difficult. In recent research (Stroebel 2012) I empirically analyse the sources and magnitude of asymmetric information between lenders, and find these lenders to differ significantly in their information about true underlying housing collateral values. From a policy perspective, the identification of collateral values as a key source of asymmetric information in mortgage lending helps to develop proposals to improve the functioning of the market. For example, it suggests that better credit scoring technology and the more extensive sharing of borrower information will not address all forms of asymmetric information and that policies to address asymmetric information about collateral quality are also important.
I focus on lenders that compete to originate mortgages used to purchase newly constructed properties. This market is characterised by property developers regularly providing home buyers with mortgage financing offers through vertically integrated mortgage lenders. These integrated lenders might have better information than non-integrated lenders about aspects of construction quality that are difficult for non-integrated lenders to observe. Such information might include, for example, the skills of the individual subcontractors constructing neighboring homes within a development. I find that such asymmetric information about collateral quality is a significant source of adverse selection in this market. In addition to testing for the presence of asymmetric information and uncovering its sources, I also quantify the impact of this asymmetric information on the cost of mortgages, which I find to be significant.
A model of integrated lenders and non-integrated lenders
A simple theoretical model is used to guide the analysis of the competition between the better informed integrated lender and other non-integrated lenders. In the model, an integrated lender obtains an informative signal about the quality of the housing collateral, while competing lenders only know the average collateral quality. The better informed integrated lender conditions its financing offer on its superior information (offering lower interest rates for mortgages secured by higher-quality collateral) and thereby subjects non-integrated lenders to adverse selection, similar to a ‘winner’s curse’, where the winner tends to overpay. As true house quality is revealed over time, those homes initially financed by an integrated lender should thus experience larger price increases relative to ex-ante similar homes financed by non-integrated lenders. This effect is bigger when the integrated lender's signal about collateral quality is more precise. To compensate for the adverse selection, non-integrated lenders charge higher interest rates to break even than if they were competing only against equally informed lenders. Interest rates rise by more for borrowers whose repayment is more sensitive to changes in collateral values, for example because they make a smaller downpayment.
I show empirically that such asymmetric information between competing lenders is in fact an important feature in the financing of newly developed homes. I construct a dataset of all housing transactions and associated mortgages in Arizona between 2000 and 2011 to track the return of properties following their initial sale. About 85% of new homes are in developments with an active integrated lender, and, when present, the average market share of these integrated lenders is about 73%. I find that in developments with an integrated lender, those houses financed by the integrated lender outperform ex-ante similar houses in the same development financed by non-integrated lenders by an average of 40 basis points annually. When I consider the distribution of returns, I find that the 40 basis point mean return difference is usually driven by a lower probability of the integrated lender financing houses that experience very significant capital losses (i.e. a thinner left-tail in the distribution of returns conditional on observable characteristics). This is suggests that the information of the integrated lender is related to the relative likelihood of low-probability, high-cost events, such as the cracking of foundations. I also find that mortgages financed by an integrated lender are over 40% less likely to enter into foreclosure than ex-ante observationally similar mortgages financed by other lenders.
An important result is that the annual outperformance of the integrated lender's collateral portfolio is larger (about 100 basis points) amongst houses built on ‘expansive soil’, a high-clay content soil that makes housing returns more sensitive to unobservable aspects of construction quality such as the care with which the foundation was poured. This result provides additional evidence that the construction quality of the housing collateral is a significant source of asymmetric information.
I also compare the return and foreclosure probability for the ownership duration of the second owner of the house. The relative outperformance of those houses initially financed by the integrated lender remains the same. This result confirms that the outperformance is to a large extent explained by asymmetric information about the housing collateral, not the borrower, since the identity of a possible second owner of the house was not known to any lender at the time the mortgage was granted to the initial owner.
To further test the theory, I analyse textual descriptions of houses in ‘for sale’ property listings. I scan these property listings for evidence of significant depreciation and identify listings that signal damage to the property. I find that when they are listed for resale, those houses initially financed by integrated lenders are less likely to contain evidence of damage to the property than ex-ante similar homes financed by non-integrated lenders, suggesting that the integrated lender's outperformance can be best explained by differential depreciation rates of houses, not differential initial pricing.
I also analyse the cost to borrowers in terms of higher interest rates that result from this asymmetric information. I find that non-integrated lenders charge an average interest rate premium of about ten basis points annually for otherwise similar mortgages when competing against an integrated lender. This higher interest rate compensates non-integrated lenders for the adverse selection in the presence of an integrated lender. As predicted by the model, the interest rate increase is larger for mortgages with a low downpayment, rising to almost 50 basis points annually for mortgages with a downpayment of less than 3%. For those mortgages the repayment probability is more sensitive to changes in collateral values. Non-integrated lenders thus need to charge higher interest rates to break even when facing adverse selection on collateral quality.
In addition to help inform the debate about policies to limit the impact of information asymmetries in mortgage lending, this project also provides insights into the lending behaviour of financial institutions in the pre-crisis period 2000-2007. It has sometimes been argued that due to the lack of ‘skin in the game’ – where high-ranking individuals invest in the companies they run – generated by securitisation or agency problems within firms, many loan officers no longer had incentives to distinguish between borrowers and collateral of differential quality, which could help to explain the lower quality of mortgages originated. In contrast, the evidence uncovered in this project is highly consistent with lenders actually attempting to price cross-sectional differences in collateral quality in a highly sophisticated manner.
This project provides new evidence on the role of integrated lenders during the recent construction boom, and suggests that rather than making low-quality mortgages in order to sell more houses, integrated lenders were actually able to select an equilibrium portfolio of mortgages that was of higher quality than that of competing non-integrated lenders.
Elul, R (2011), “Securitization and mortgage default", Federal Reserve Bank of Philadelphia, Working Paper, 9-21.
Keys, B J, T Mukherjee, A Seru, and V Vig (2010), “Did securitization lead to lax screening? Evidence from subprime loans", Quarterly Journal of Economics, 125 (1), 307-362.
Stroebel, J (2012