Securitization, Recourse Uncertainty, and Crash Risk

Asset securitizations have become a prominent type of financial transaction in recent decades. According to the Securities Industry and Financial Markets Association, $3.3 trillion worth of mortgage- and other asset-backed securities were issued in 2006. This number fell to $1.6 trillion in 2008 and rebounded to $2.4 trillion in 2016, compared with $1.5 trillion issuances of corporate debt in 2016. While research documents various benefits from securitization, the complexity and financial reporting opacity of these transactions have raised concern among academics, practitioners, and standard-setters. In particular, given the complex structure and disclosures, market participants have found it difficult to assess the true level of risk transferred by banks to investors in asset-backed securities, and thus faced significant opacity-induced recourse uncertainty. In a new paper, we examine the extent to which this uncertainty is associated with bank-specific stock price crash risk, defined as negatively skewed future returns. This examination is important, as stock price crashes have devastating effects on shareholder welfare, and awareness of such an association could be useful in managing risk through applications such as value-at-risk (VaR).

In a typical securitization, the issuer (i.e., a securitizing bank) transfers a pool of financial assets, such as mortgages and credit card receivables, to a special-purpose entity, a bankruptcy-remote entity that finances the acquisition of these assets by issuing debt in the form of beneficial interests. These interests typically consist of a hierarchy of risk-return tranches where the least (most) risky tranche earns the least (highest) return. Since banks have superior information (relative to investors) about the credit risk of the transferred assets, they provide some form of recourse to protect investors against potential losses from these assets. Securitization structures can differ substantially in the extent to which banks retain the risks associated with the securitized assets. Banks often provide contractual (“explicit”) recourse by retaining first-loss contractual interests. When the securitized assets under-perform, banks may also provide non-contractual (“implicit”) recourse to make up some portion of the losses not covered by the retained interest in securitized assets. Additionally, in nearly all securitizations, banks provide contractual representations and warranties that the securitized assets have stipulated characteristics. Violations of these clauses require banks to buy back the assets if requested by the purchasers.

Due to the complex structure and inadequate financial reporting, it is difficult for market participants to assess the extent of risk retained by the bank, regardless of the recourse form. Bank managers have incentives stemming from a variety of reasons (e.g., career concerns) to withhold bad news but not good news. The difficulty in evaluating the level of recourse by market participants is likely to enable the managers to conceal negative news about recourse for an extended period, which in turn increases the likelihood of future crashes. A good example is the 2007 bankruptcy of New Century, the second-largest subprime mortgage lender during 2006. The bankruptcy examiner lists two key accounting and disclosure issues that contributed to the collapse of New Century. First, the company significantly understated its repurchase reserve, a liability related to loans that New Century had sold but was potentially obligated to repurchase when the borrower defaulted shortly after the sale or because New Century failed to comply with the representations and warranties. Second, New Century overstated the value of retained interests in securitizations by using artificially low discount rate, prepayment, and credit loss forecasts in the valuation models. We expect the recourse uncertainty faced by market participants to be positively associated with banks’ future stock price crash risk.

A positive relation between recourse uncertainty and crash risk is not obvious, however. Extensive information about securitizations is available from public sources other than banks’ financial reports, including credit ratings of asset-backed securities, securitization prospectuses, and data vendors (e.g., Asset-Backed Alert). Research shows that these sources provide useful information such as issuer and deal characteristics that helps market participants assess risks. The richness of this information may forestall the strategic bad-news-hoarding behavior by bank managers. Another feature of securitizing banks is that they are under scrutiny by bank regulators, who pay close attention to disclosures of potential adverse events. The stringent bank regulation may limit the ability to hide negative recourse news. As such, whether recourse uncertainty is positively associated with future crash risk is ultimately an empirical question.

Our sample comprises all U.S. public banks with securitized assets in at least one year during our sample period (2002-2015). We use two measures of bank-specific stock price crash risk: (1) the negative conditional skewness of bank-specific weekly returns and (2) the likelihood of the occurrence of extreme negative bank-specific weekly returns (i.e., crash weeks). Since market participants’ uncertainty about the extent of recourse is not directly observable, we construct a recourse uncertainty factor based on four proxies: securitized assets, non-performing securitized loans, charge-offs of securitized loans, and retained interests. Initial analyses reveal that more news, as captured by the number of earnings announcements, 8-K filings, and press releases, is disseminated in a crash week than a non-crash week on average. We also manually examine news during crash weeks related to high recourse uncertainty and find that more than three-quarters of crashes occur at the occurrence of adverse recourse events (e.g., write-downs of retained interests).

The primary multivariate tests involve regressions of crash risk measures on the recourse uncertainty factor, control variables (i.e., changes in share turnover, lagged negative skewness in returns, bank leverage, return-on-assets, market-to-book, market cap, return volatility, and tier 1 capital ratios), and year fixed effects. We find a positive and significant association between recourse uncertainty and crash risk. A one-standard-deviation increase in the recourse uncertainty factor is associated with an increase in the negative skewness in returns of 7.5 percent of its standard deviation, or an increase in the probability of extreme negative returns by 18.8 percent relative to the mean. The evidence suggests that banks with more recourse uncertainty are more prone to crashes than banks with less recourse uncertainty.

Despite a host of controls that are included in the regression, the results may capture the effect of securitizations on bank intrinsic risk, rather than on recourse uncertainty due to the opacity of these transactions. By bank intrinsic risk, we mean the risk inherent in banks’ fundamental businesses, independent of their financial reporting practices. We perform three sets of tests to alleviate this concern. First, we find the relation between recourse uncertainty and crash risk to be stronger for banks with a poor information environment, as captured by low analyst following, low institutional ownership, and small bank size. Second, we exploit a shock to the requirements for securitization disclosure. In 2009, the Financial Accounting Standards Board (FASB) issued two standards, SFAS 166 and SFAS 167, effective as of 2010, to improve the transparency of securitization transactions. We find that the positive association between recourse uncertainty and future crash risk is attenuated in the post-SFAS 166/167 sample. Using a subsample of banks that do not significantly reduce the mean and variance of their securitization activities following the new standards does not alter our inferences, suggesting that the attenuation of the association is attributable to the improvement in the transparency as opposed to the reduction in statistical power. Finally, if merely a tail event that happens to occur in one of the 52 weeks of a year is responsible for the relation between recourse uncertainty and crash risk, there should also be a positive association between recourse uncertainty and a positive jump. We find no association between these two variables. Our results are also robust to explicitly controlling for the skewness of quarterly cash flows over the past five years. Thus, bank intrinsic risk as reflected in the third moment (i.e., skewness) of cash flows is unlikely to explain our findings. Moreover, our results hold after controlling for a proxy for earnings management based on discretionary loan loss provisions and realized gains on securities (i.e., bank earnings management unrelated to securitization).

We contribute to the securitization literature by investigating the implications of recourse uncertainty for future stock price crash risk, a previously unexamined consequence important to market participants. This study also identifies an important type of complex transactions, securitizations, as a source of opacity that contributes to future stock price crashes, thereby adding to the crash risk literature. The results suggest that the presence of alternative information sources and stringent bank regulations in the U.S. are insufficient to prevent recourse uncertainty in securitizations from increasing bank crash risk. Our findings that the impact of recourse uncertainty on crash risk is muted in a rich information environment and after the adoption of SFAS 166/167 underscore the benefits of the rapid growth of information intermediaries generally and the FASB’s recent efforts to improve securitization disclosures specifically.

This post comes to us from professors Yiwei Dou and Joshua Ronen at New York University and Tuba Toksoz at Koc University It is based on their recent paper, “Securitization, Recourse Uncertainty, and Crash Risk,” available here.