One of the villains of the current financial crisis is "securitization." The alphabet soup of securities structures--CMOs, CDOs and SIVs--is roundly blamed for the current financial world's mess.
The irony is that it was not so long ago that emerging countries looked to securitization as a savior for the problems that they faced in developing capital markets. Specifically, many countries (particularly in Latin America) looked to Fannie Mae/Freddie Mac Mortgage Backed Securities as models for instruments for providing housing finance, and others (such as India) looked at special purpose vechicles as a potential method for getting around the poor financial conditions of local government attempting to finance infrastructure.
So which is it: villain or savior? Well, of course the answer is neither. Securitization is just an instrument, and when applied appropriately under appropriate circumstances, is a useful instrument. So let's begin by dispensing with the notion that securitization is it self a villain, and then talk about why it is not a savior either.
I believe investors made two fundemental mistakes about subprime mortgages. First, some investors thought US house prices would never fall nationally, in part because they never had (in nominal terms) in the post-was era. So long as house prices rose, these investors reasoned, mortgage borrowers would retain a powerful incentive not to default; consequently, default risk for all mortgages was deemed to be low. True story--around 2005 I was in the elevator of a large investment bank, and one person said to another, "you can't make a bad real estate loan." That happens to be the moment that I began to worry about the subprime market.
The problems with this line of reasoning were two: just because house prices had never fallen nationally didn't mean that they couldn't, and even if house prices did rise nationally, if they fall regionally (as they did in the US Midwest in the 1970s, in Texas in the 1980s and in New England and California in the 1990s), one will still see defaults. This was an underwriting issue, not a securitization issue.
The second problem is that Wall Street Ph.D.s thought they could outsmart bad underwriting. This reflected insufficient modestly about how confident we can be about parameters. The idea behind Collatoralized Debt Obligations was that one could combine subordinated securities (those that were in the first loss position) and use diversification to get a very precise estimate of the losses that one could expect from those securities. Suppose that the expected loss of a security was ten percent with a standard deviation of ten percent. By combining 30 securities, one reduces the standard deviation by 1/(sqrt(30)), or by more than 1/5, so investors can have confidence that the actual realized loss would fall within a narrow band. Investors could then use the knowledge to further slice and dice.
For this to work, however, one needs to know that the parameter estimates for expected losses and standard deviaion of losses are correct. For a whole host of reasons, we didn't have remotely enough information about parameter stability to make these sorts of judgments. Something we need to remember is that as our models get cleverer, we start losing degrees of freedom. But we also need to remember that the vanilla MBS security structure, and even simple senior-subordinated tranching, worked very well for a very long time. Securitization is a good way to match up households with capital markets, and remains true today.
But the recent crisis suggests that securitization is no magic bullet for emerging countries. For securitization to work, investors need to understand the loans that are being securitized, and that means the loans must be underwritten robustly and consistently. For this to happen, emerging economies will need stronger financial infrastrucure (such as well developed banking systems) and property rights infrastructure (so investors can have confidence in collatoral). I remember when I did a Bank mission in one very low-income country, I was asked about whether it should develop a Fannie Mae. This was a country whose courts couldn't enforce foreclosure rules, and that had no long term sources of finance. If any good news arises from the current crisis, it is that emerging countries might focus on getting fundementals right, instead of hoping for a magic securitization scheme to solve all their problems.
Monday, November 2, 2015
A Post on the Growth Commission Blog
It is here. I wrote:
Counterparties
I participated last week in a Berkeley-UCLA conference on the Mortgage Meltdown.
Two distinguished scholars said things that surprised me. One noted that the net position of derivatives was zero (a correct statement) and therefore derivatives were nothing to worry about. The other said that the size of the subprime losses would likely be around $400 billion, which should have been managable. I should note that I made a similar statement about subprime losses around a year ago, but I didn't really think through the implications of it.
Had the $400 billion of subprime loan losses been held with equity, the implications of the crisis would have been much smaller. This is why the collapse of the tech bubble, while meaningful to the economy, resulted in only a mild recession. The problem is that the $400 billion in losses on subprime mortgages and the derivatioves they support are being realized by highly levered institutions, and so losses precipitate a chain of events that go well beond he original losses.
Let's start with the loans themselves. Suppose a bank owns subprime loans, and assumes a loss rate of 5 percent, and holds sufficient capital to bakstop those loans. Now suppose the loss rate doubles. The bank may still be solvent (i.e., have positive value), but its capital position has been eroded, perhaps to the point that it can no longer make loans. This is not hypothetical--I have recently talked to some community bankers who have told me this is exactly their position.
The inability to make loans means the value of the banking business falls--and so share prices fall. This means it is difficult for banks to recapitalize, because issuing stock is expensive. Credit markets freeze up, and the implications of losses arising from subprime get transmitted to the general capital stock--banks lose the ability to finance P&E. The reduces the future value of the broader economy beyond the initial subprime loss.
But now let's consider a bank that was smart enough to buy credit default swaps so that its balance sheet would remain healthy in the event of poor subprime performance. The bank gives up cash flow to pay premiums to insure against future problems. But the bank is assuming that the counterparty with which it has contracted can make good on its obligations. If the counterparty is highly levered, a subprime meltdown will produce bankruptcy. It is here that the ex post net derivative value goes negative: the payouts from the insurance company are smaller than the losses incured by the bank, because the value of the insurance company is truncated at zero. This creates its own set of multiplier effects.
Two distinguished scholars said things that surprised me. One noted that the net position of derivatives was zero (a correct statement) and therefore derivatives were nothing to worry about. The other said that the size of the subprime losses would likely be around $400 billion, which should have been managable. I should note that I made a similar statement about subprime losses around a year ago, but I didn't really think through the implications of it.
Had the $400 billion of subprime loan losses been held with equity, the implications of the crisis would have been much smaller. This is why the collapse of the tech bubble, while meaningful to the economy, resulted in only a mild recession. The problem is that the $400 billion in losses on subprime mortgages and the derivatioves they support are being realized by highly levered institutions, and so losses precipitate a chain of events that go well beond he original losses.
Let's start with the loans themselves. Suppose a bank owns subprime loans, and assumes a loss rate of 5 percent, and holds sufficient capital to bakstop those loans. Now suppose the loss rate doubles. The bank may still be solvent (i.e., have positive value), but its capital position has been eroded, perhaps to the point that it can no longer make loans. This is not hypothetical--I have recently talked to some community bankers who have told me this is exactly their position.
The inability to make loans means the value of the banking business falls--and so share prices fall. This means it is difficult for banks to recapitalize, because issuing stock is expensive. Credit markets freeze up, and the implications of losses arising from subprime get transmitted to the general capital stock--banks lose the ability to finance P&E. The reduces the future value of the broader economy beyond the initial subprime loss.
But now let's consider a bank that was smart enough to buy credit default swaps so that its balance sheet would remain healthy in the event of poor subprime performance. The bank gives up cash flow to pay premiums to insure against future problems. But the bank is assuming that the counterparty with which it has contracted can make good on its obligations. If the counterparty is highly levered, a subprime meltdown will produce bankruptcy. It is here that the ex post net derivative value goes negative: the payouts from the insurance company are smaller than the losses incured by the bank, because the value of the insurance company is truncated at zero. This creates its own set of multiplier effects.
The Presiden't Commission on Tax Reform and Housing
Andy Reschovsky and I just wrote a paper on this. Here is part of the conclusion:
The United States government currently spends about $175 billion per year to subsidize homeownership. The lion’s share of these subsidies operate through the tax system, with the largest single tax subsidy being the mortgage interest deduction. Despite these subsidies, ho-meownership rates for certain groups of Americans, notably African Americans, Hispanics, and households with modest incomes, are substantially below the average homeownership rate. For several decades now, economists have argued that the use of a mortgage interest credit instead of the current deduction would both encourage homeownership and more equitably distribute ho-meownership tax subsidies across the income distribution. In fact, in their final report, issued in November 2005, the President’s Advisory Panel on Federal Tax Reform recommended eliminat-ing the mortgage interest deduction and replacing it with a non-refundable 15 percent mortgage tax credit.
In this paper, we estimate a model of housing tenure choice and housing expenditures using data from the one-percent PUMS from the 2000 Census. The model allows us to determine the impact of alternative tax policies on the user cost of owning relative to renting. Because of our large sample size, we are able to estimate quite precise effects for individual racial/ethnic groups. Our results are very robust, with the tax variable proving to be highly significant in re-gressions using data for both 1990 and 2000, for recent movers, and for various racial/ethnic groups. The results of our housing model are used in a tax simulation model that we have con-structed based on 2004 federal and state tax law. Our simulation model allows us to calculate federal income tax liabilities of all taxpayers under existing tax law and under a variety of alter-native tax policies aimed at increasing the rate of homeownership.
Specially, we simulate the impact of the housing credit proposal made by President Bush’s Advisory Panel on Federal Tax Reform. Because we are convinced that the elimination of the mortgage interest deduction is politically impossible, we also simulate a plan that involves a refundable mortgage interest tax credit that provides every household with the option of utilizing the credit or the existing mortgage interest deduction, whichever one provides the largest tax savings.The striking thing about the housing proposals of the President’s Tax Reform Panel is how little they do beyond redistributing the homeowner tax subsidy. The net result of replacing the mortgage interest deduction with a credit would actually be a decrease in the overall ho-meownership rate (by half a percentage point). This decrease would occur because the additional homeownership among low- and moderate-income households would be insufficient to offset the decrease in homeownership among middle- and high-income households, some of whom would choose to rent in response to a reduction in the size of the mortgage tax subsidy they would re-ceive, Although a credit is more beneficial than a deduction for low- and moderate-income families, its impact on the homeownership rate is limited because the proposed credit is non-refundable, and therefore provides no homeownership incentive to households who pay no in-come tax.
On the other hand, because many homeowners who are not itemizers become eligible for the credit, the Tax Reform panel’s proposal does not do very much to increase federal income tax revenue. Much of the benefit of the proposed plan flows to households who are already owners in the form of reduced housing costs. We estimate that if there were no changes in how households finance their housing, the Treasury would gain $4.9 billion in revenue, about 1.6 per-cent of total 2004 income tax revenue. However, if there is a substantial shift away from debt toward equity for the financing of homes, the net revenue impact of the proposal will be much smaller.
Statistics and the WSJ Op-ed page
There is an embarrassingly bad piece in today's WSJ op-ed page:
http://online.wsj.com/article/SB119397079767680173.html?mod=opinion_main_commentaries
The two writers follow the number of executions and the number of murders across time, conclude that there is a strong correlation between executions one year and number of murders the next, and then say the fact that they haven't controlled for anything is probably OK, because of Occam's razor.
Aside from the fact that they don't deal with the special issues presented by time series data (correlations produced by two series of data with strong trends in them will be spuriously high), their Occam's razor approach pretty much ignores how serious people do social science research. Richard Freeman, the Harvard Labor economist, says something like: "it had better be there in the correlation, it has better be there in the OLS regression, it had better be there in the IV regression, and it had better be there in the high-tech Maximum Likelihood estimate.
I am open minded about the possibility that the death penalty is a deterrent. But this morning's piece does nothing to pursuade me; worse, by polluting social science, it undermines social science.
http://online.wsj.com/article/SB119397079767680173.html?mod=opinion_main_commentaries
The two writers follow the number of executions and the number of murders across time, conclude that there is a strong correlation between executions one year and number of murders the next, and then say the fact that they haven't controlled for anything is probably OK, because of Occam's razor.
Aside from the fact that they don't deal with the special issues presented by time series data (correlations produced by two series of data with strong trends in them will be spuriously high), their Occam's razor approach pretty much ignores how serious people do social science research. Richard Freeman, the Harvard Labor economist, says something like: "it had better be there in the correlation, it has better be there in the OLS regression, it had better be there in the IV regression, and it had better be there in the high-tech Maximum Likelihood estimate.
I am open minded about the possibility that the death penalty is a deterrent. But this morning's piece does nothing to pursuade me; worse, by polluting social science, it undermines social science.
Subscribe to:
Posts (Atom)