The process of securitizing mortgages has received a lot of negative attention during the financial crisis. An oft-made claim is that by collateralizing risky subprime mortgages, securitization drove an unprecedented expansion of mortgage credit to borrowers with bad credit histories. This increase in available credit fed the housing bubble, which ultimately burst after the expansion of subprime credit had run its course. The implication is that without Wall Street's insatiable appetite for mortgage-backed securities (MBS), less bad credit would have been extended, housing prices would not have soared so high, and the subsequent housing bust would not have been so bad.
On the surface, a link between securitization of subprime mortgages and the housing bubble seems both intuitive and plausible. After all, the vast majority of subprime mortgages were sold by originators to private institutions, not the government-sponsored housing-finance agencies like Freddie Mac and Fannie Mae. These private institutions operated in the secondary mortgage market, where loans are securitized and sold to investors around the world. Figure 1A in Mayer and Pence (2008) provides a nice illustration of this pattern.
Higher housing prices, not MBS, may have encouraged subprime lending
But the arrow of causality may not run from the expansion of securitized subprime credit to higher housing prices. Rather, expectations of higher housing prices may have encouraged more lending to subprime borrowers, whose loans were subsequently securitized. (Any loan is a good loan if prices are rising, because the collateral that backs the loan is getting more valuable over time.) Both interpretations are equally validated by aggregate data. To answer the chicken-or-egg question of what comes first in the logical chain—higher housing prices or increased subprime securitization—you need to perform a more disaggregated analysis.
In a recent paper, Taylor Nadauld of Ohio State and Shane Sherlund of the Federal Reserve Board of Governors attempt to solve this identification problem. Their analysis is based on a change in regulations that could have affected the level of securitization but was plausibly unrelated to either housing prices or the demand for subprime credit. Consequently, by examining the effects of this regulatory change, the authors can ask whether changes in securitization had true, causal effects on the amount of credit extended to subprime borrowers.
Did capital requirements reduction increase the demand for MBS?
The regulatory change at the heart of the paper involves the required capital levels at five broker-dealers: Bear Stearns, Morgan Stanley, Goldman Sachs, Lehman Brothers, and Merrill Lynch. By some accounts, the rule reduced capital requirements at these five institutions by up to 40 percent. Specifically, in April 2004, the Securities and Exchange Commission amended a series of rules that had the effect of reducing capital requirements for the five broker-dealers (hereafter referred to as consolidated supervised entities, or CSEs). The change was made in response to the European Union's Conglomerates Directive that required U.S. broker-dealer affiliates to show proof that their consolidated holding companies were subject to supervision by a U.S. regulator. The rule change established an alternative method of calculating capital requirements for the CSEs, which were not already subject to consolidated capital regulation from a regulatory authority. Basically, the CSEs could use their own internal risk-based models to calculate a capital adequacy measure consistent with those put forth by the Basel Committee on Banking Supervision. (For further detail, see the Nadauld and Sherlund paper.)
The authors hypothesize that the reduction in capital requirements increased the institutions' demand for purchasing subprime mortgages from the primary market for the purpose of securitizing them either to sell to other investors or to hold themselves. That is, lower capital requirements either increased their demand to invest in subprime MBS themselves or increased their capacity as intermediaries to securitize the mortgages and sell to other investors.
Authors use two-stage strategy to support findings
The authors use a two-stage econometric strategy to identify the impact of this exogenous increase in the CSEs' demand to purchase subprime mortgages to securitize, which would have raised the supply of credit to subprime borrowers. In the first stage of their analysis, they verify that the regulatory event did indeed raise secondary-market purchases at the affected institutions. In particular, they ask whether securitization activity among the five institutions increased by more than the securitization activity of institutions that were not affected by the regulatory event. The answer is yes. According to data on privately securitized mortgages from FirstAmerican LoanPerformance, the CSE banks securitized about 32 percent more loans on average than did their non-CSE counterparts after 2003.
In the second stage, the authors ask whether the increase in CSE securitization is linked to an increase in subprime credit supply. For this step, they obtain ZIP-code–level data on mortgage originations and securitization activity from Home Mortgage Disclosure Act data and then merge these data with the LoanPerformance data. Essentially, in this stage, the authors ask whether ZIP codes that experienced higher CSE securitization activity (relative to non-CSE securitization activity) also experienced higher levels of subprime mortgage originations. The answer again is yes. The authors interpret their findings as evidence that increased demand for the CSEs to securitize mortgages resulted in increased access to subprime mortgage credit at the household level.
Analysis may need refinement
In our opinion, this paper is one of the few that has come up with a reasonable way to identify the effect of the secondary mortgage market on the ability of households to obtain mortgages. But the paper needs to address two issues in order to offer a more convincing analysis.
The first issue concerns the authors' measure of subprime credit supply. They use the ratio of subprime mortgages originated to total housing units in a given ZIP code. But this is a measure of mortgage credit issued in equilibrium. Many factors could create cross-sectional variation in this variable (across ZIP codes) that have nothing to do with differences in access to credit. For example, differences in homeownership rates, the fraction of homeowners with a mortgage, and wealth and income differences could all affect the quantity of mortgage lending in a ZIP code without explaining differences in access to credit. Of course, the authors try their best to control for such factors in their estimates, but ultimately it is impossible to control for all of them.
The second substantive issue concerns the link between the regulatory event and demand and supply for MBS in the secondary market. The authors argue that the regulatory event could have affected the secondary mortgage market through two channels. First, they argue, relaxing capital requirements may have increased the CSE banks' demand for highly rated subprime MBS. We know for certain that the five CSE institutions were heavily involved in the supply of MBS to other investors, but we also think that these institutions were investors as well. It shouldn't be too hard for the authors to find evidence of this connection, but what would be even more convincing, and perhaps more difficult, would be to review whether the CSE banks substantially increased their holdings of subprime MBS after the regulatory event.
The second potential channel involved relaxing the constraints associated with the supply of subprime MBS. In this case, capital is needed to warehouse mortgages during the process of creating securities. In addition, most deals required over-collateralization, which usually meant that the issuer would take the first-loss position. If these constraints were binding for these institutions before the regulatory event (that is, the secondary market had pent-up demand for subprime MBS), then the relief on capital requirements after the event may have resulted in increased supply. This hypothesis seems a little far-fetched to us, not to mention virtually impossible to test in the data, so the authors may be better off focusing on the demand-side effects.
By Kris Gerardi, research economist and assistant policy adviser at the Atlanta Fed (with Boston Fed economists Christopher Foote and Paul Willen)
Note: The authors were given an opportunity to respond to this blog posting. As of this publishing, the author has not commented.