Commercial Real Estate Loan Performance at Failed U.S. Banks

Abstract

Exposure to commercial real estate (CRE) loans at regional and small banks and thrifts has soared over the last two decades. As banks’ balance sheets become more concentrated in these types of loans, banks have become more sensitive to swings in CRE fundamentals. The concentration in CRE loans peaked in 2007, just as commercial real estate prices started a historic free fall, declining more than 30 percent in just two years. Over this same time CRE concentration has been a significant factor in recent bank failures. Default and loss models of CRE mortgages have previously been estimated using loan data from large, income-generating properties financed by insurance companies and the commercial real estate mortgage (CMBS) market. Early research used data from insurance companies (Synderman (1991), Esaki et al (1999), Vandell et al (1993), Ciochetti et al (2003), while more recently researchers have used data from the CMBS market (Ambrose and Sanders (2003), Archer et al (2002), Deng et al (2004), Seslen and Wheaton (2010), An et al (2009). Black et al (2010) found that loans in CMBS pools that had been originated by portfolio lenders, such as insurance companies or commercial banks, were of a higher quality and outperformed loans originated by conduit lenders or investment banks. The CMBS and insurance company loans used in these studies differ in structure and underlying collateral from the loans backed by bank CRE loans. Roughly a third of bank CRE loans are backed by owner-occupied CRE and another 20 to 30 percent by land and construction loans. The owner-occupied properties, which lack an external and explicit rental stream, are usually not candidates for securitisation. The loans in bank portfolios backed by land acquisition, development, and construction (ADC) projects are even less similar to those in CMBS and in insurance company portfolios. Land and construction loans are short term and the collateral is the raw land or the partially completed construction project. Finally, the loans on banks’ books backed by existing income-generating commercial properties are likely to be different from those found in CMBS pools or in insurance company portfolios.Regional and small banks also make much smaller loans than those usually seen in CMBS pools or in insurance company portfolios. Clearly, each of these types of loans has performed differently during this recent financial crisis, yet we are still dependent on default and loss models estimated using data from only one type of loan.Ours is the first paper to estimate CRE default and loss models using a loan-level dataset drawn from bank portfolios. We develop a unique dataset consisting of loan-level information on CRE portfolios for a sample of banks entering FDIC receivership over the past several years. We use this dataset to estimate a series of default and loss models. We estimate these models on the loans backed by existing CRE properties and compare the results with those from other papers that estimate CRE default using data from the CMBS and insurance companies. We then extend our analysis to the performance of land and construction loans, providing the first loan-level analysis of the performance of such loans.Full publication: Property markets and financial stability