The credit crisis has forced MBS investors to adopt new tactics. Notably, they have to take a closer look at the collateral backing mortgage bonds. But the analysis goes beyond that. Investors now need to look at home prices in a geographic region and consider the health of local economies where the homes are actually located. Ahead of the crisis, little attention was paid to whether a property was owner-occupied or if there were multiple liens on a single property. Now, issues like these make a world of a difference. Then there is the question of what actually makes a borrower walk away from a loan? Some observers believe geographic location has much to do with a homeowner's decision to default on their mortgage. Having clues to the roots of a default can help market participants better size up the value of a pool of whole loans or securities out for bid. ASR recently met with some of the market's leading analysts and observers to learn about the leading indicators of a default in a mortgage-backed bond. That discussion, which was moderated by ASR's contributing editor Aleksandrs Rozens, was sponsored by 1010data - a provider of tools and analytics used by many mortgage bond investors today. Interestingly, the panelists agreed that even if the data to look deep in the mortgage bonds were available ahead of this credit crisis, few investors would have held off from snapping up the riskiest mortgage securities. Why? The appetite for risk was just too great.
Rozens:What kinds of tools are people in the industry, originators/dealers, using to get an edge when they look at credit information about homeowners?
MAYER: One of the things that can be done is to try to get a sense not only of the obligation somebody has for their mortgage, but also to try to look at information about the borrower's credit. They can also look at the underlying credit associated with mortgages. When we've done that kind of analysis, for example, we've found that roughly a quarter of the borrowers who have mortgages tend to substantially reduce their borrowings immediately after walking away from a mortgage, which is consistent with the idea of strategically defaulting. In other words, you're defaulting not because you can't afford to make the payments but because you're choosing to reduce other kinds of debt and it's not because of a job loss.
BERGANTINO: The data we currently have access to for projecting mortgage performance includes a combination of loan-level at origination credit characteristics such as mortgage amortization type, borrower credit score, loan-to-value ratio (LTV), reported occupancy status, etc. as well as various dynamic economic information such as movements in interest rates, home prices and unemployment rates. When we combine all of these data we find that a borrower's equity position, as measured by their home price index adjusted combined LTV, is highly correlated with the probability of becoming delinquent and eventually defaulting on their mortgage. In particular, the data shows that when a borrower's combined LTV exceeds 100%, meaning that the borrower has no equity or negative equity in the property, default rates increase substantially. While this result could be an indication of strategic default, extremely high LTVs are typically associated with weak local economies, and so we are also likely capturing a reduced ability to pay in addition to any tendency of borrowers to walk away from underwater mortgages.
GREEN: Not so long ago, there were many investors whose RMBS credit analysis consisted merely of checking the ratings assigned to the bond, and maybe looking at the percentage of loans remaining in the deal that were delinquent as of the most recent date. As investors became sophisticated, and information technology in the industry improved, smart money players began doing more detailed granular loan level analysis. Then, as mortgage credit began to deteriorate rapidly in 2007, an increasing number of market participants also began tracking the amount of equity or negative equity in every home backing every loan in a every deal in the entire non-agency universe at every point in time. However, investors soon realized they still needed to take it a step further. To more accurately estimate the equity of a borrower, one needs to know the balance of all liens on the property. In the past, information about the existence of a second (lien) or a HELOC was not that easy to obtain. From 2005-2007, many borrowers picked up second liens or took out a HELOC several weeks or months after they got the first lien, thereby reducing the equity dramatically in the home long after the first lien was securitized in a deal. So now, to get a more complete picture, many investors on our platform have begun to leverage updated consumer credit information from credit bureaus that are linked to loans. These periodic updates provide investors with timely and more detailed consumer information every month, including the balances of any new or subordinate lien or HELOC on the property. These investors are also looking at numerous other credit variables provided by the credit bureaus such as changes in credit scores of current borrowers. The idea is to gain an edge to better predict which current borrowers will go delinquent in the future. Because there are so many new variables to look at, it makes having a powerful, fast analytics platform increasingly important.
Rozens:The mortgage investor has come along way from just relying on ratings?
GREEN: Right. We've gone from just looking at ratings from a rating agency, to individual loan level analysis, to tracking the amount of equity on every loan at every point in time, to comprehensive consumer credit updates on borrowers. It's happened in a very short period of time due to everything that's gone on in the world in the past three or four years.
XU: We use loan level data from two sources: Loan Performance and BlackBox. We use these datasets in conjunction with Equifax data on 1010data's platform. This is not only to get the historical data of the performance, such as delinquencies, but we're using, combined with the House Price Index, the Equifax and the borrower's new credit score to get the current LTV and combined loan to value (CLTV) level. That gives us more updated data. We looked into the issue of strategic defaults, which are very interesting. We looked into borrowers in states where the judgment after the mortgage foreclosures is not allowed, such as California and Arizona. We compared these with the localities, such as Florida and Nevada, which allow the banks to pursue the borrowers after they default on their mortgage. We found that under the same LTV circumstances, the California and Arizona borrowers are 40% more likely to default compared to those in Florida and Nevada. We think strategic defaults are happening on a big scale.
GOODARZI: It's very exciting to see that people are using our data. It's exciting to see the split that existed in the analysis of data, which was between static and dynamic information. The dynamic piece was the delinquency and the balance information, and everything else was pretty much fairly static as far as the loan was concerned. What we're provided with is all or many of the other attributes. Owner occupancy was mentioned, current CLTV was mentioned. When these pieces come to bear, they give you a more accurate representation of what the borrowers are up to now, as opposed to what they were up to, even on the assumption that borrower was telling us the truth. If you were interested in looking at fraud as a possible event, then you could certainly look to see what was declared earlier. An example is following through to see if owner occupancy is in fact true over a long period of time or throughout the life of the loan. We know that 25% of the population that declares themselves an owner occupant is no longer an owner occupant within 24 months after the loan's origination. We have been releasing that data, and we know that tracking that information is going to give an added lift to the models that you're building and the analysis that you're doing across the board. So the other piece of this is how predictive these variables are. It's one thing to say we have additional information and it's another thing to say it's relevant to analysis. In our analysis and testing, the added lift to the model ranges depending on which segment of the market and of the loan population you look at. It could range anywhere from 10%, 15%, 20% to about 30%. The added lift observed in being able to predict defaults and delinquencies is the reason we're all very excited about the added variables.
MAYER:I would highlight the concerns about strategic default as one of the things a lot of us are trying to pay attention to, and to try to understand whether there are clusters of these going on and whether there are changes in people's attitudes about paying debt. One of the stunning things is if you look, for example, at loans with a CLTV of 180 or more - so people are way leveraged and very far out of the money, the vast majority of those loans will still be current. If they were current then, they will be current now. So looking at indicators that suggest a contagion effect is something that's pretty important, not only in the local data but overall. A part of seeing some market stabilization has been the people who were prone to walk away have left. Hopefully the population that's remaining is more willing to take those responsibilities seriously.
GOODARZI: What we're seeing on strategic defaults is that certainly they're more likely to be investors; not investors as declared by the borrower but as declared by the data. They tend to be taking on larger amounts of debt, and also the types of property that they're purchasing are larger and more expensive than the median in that neighborhood. That in and of itself is very interesting to hone in on to identify who will be a candidate for this kind of behavior, as opposed to identifying them postmortem. It's a little bit more exciting to actually identify them a few months, maybe a year, before they actually pull the trigger on that decision.
MAYER:Throughout the LTV curve, it's surprising, if one had a very strategic view of borrower behavior, the number of people who still make their payments given the large number of underwater borrowers in the country. That's what most lenders thought when they originated the loans. They thought people would continue to make payments because historically people who are underwater are more likely to default, but still the majority of them still make their payments.
ROZENS: Let's talk about the contagion effect that you had mentioned. This survey that Fannie Mae had of 3,000 homeowners and renters who personally knew someone who had defaulted on a loan were twice as likely to consider it. Whether they go through it or have that strategic default that's a different story, but it was interesting just knowing someone who's done it made a big difference.
BERGANTINO: As I alluded to earlier, updated LTV is often an indicator of the economic stress faced by a borrower as well as their incentive to continue servicing their mortgage. If home prices have fallen 60% in a particular region, then there has probably been a fairly large increase in the unemployment rate in that area as well. So correlations between LTV and default are likely picking up a combination of strategic behavior and an inability to pay on the part of the borrower.
GOODARZI:Part of that economic stress is going to come out in the kinds of relationships that the borrowers have, above and beyond their mortgage. What they do with their student loan, their auto loan, their credit cards, the behavior that they exhibit are all going to roll into their mortgage. Granted, in the course of the last two years or so, the time distance between the delinquency in a credit card and in a mortgage has shrunk overall. But, we're still talking about several months, in the order of four to six months prior to a mortgage delinquency that you can observe that there has been delinquencies in other areas. Credit card behavior, rolling debt in general, is going to pop in before you see any delinquency behavior occurring on the mortgage.
GREEN:I've spent a lot of time looking through the vast amount of updated credit information available. There are many ways it can be used to augment a model. There are many variables one needs to explore, so making good use of this data requires a fast and flexible analytics platform capable of handling very large volumes of data. There are many curves you have to run. It's a very interactive process that requires many calculations on dozens of variables and hundreds of millions of rows of data. So to build a model that captures all of this is not a simple task. But not every subscriber to the credit updates through our platform uses it to build complicated models. There are traders who use the information just to do better, more in depth surveillance. They may want to know which specific securitized deals have borrowers with healthier looking credit profiles within the "still current" bucket of loans. They then use the credit data to rank order loans based on a few key credit variables. Clearly the new credit data does not diminish the importance of home movements and mortgage rates, which are still extremely important. The more negative equity in the home, the more likely you are to strategically default on a macro level. However, many buyside market participants want to determine all relevant information on the borrower for every single loan that they're exposed to, not just borrower equity because there are underwater borrowers who still pay their loans. So it's almost the opposite of what was going on years ago, where people were exposed to a bond and they made their decision based on let's say the name of the originator. Speaking of originator names, I was listening to Congress question a Goldman Sachs executive about what was going on in the marketing process of a certain controversial CDO deal back in 2007, and I think Senator [Carl] Levin was grilling one of the executives because the executive knew that this deal was full of bonds backed by New Century loan collateral and therefore the deal was bad because everybody knew New Century had a bad reputation. Well, I think that is a bit of a simplistic view. There probably were some New Century loans that were very bad, but the name New Century didn't automatically make the loans bad. The loans in those bonds were bad because the borrowers likely each had multiple liens on their properties, the borrowers' credit score likely dropped three months after getting the loan, the borrower never moved into the house after receiving the loan. It wasn't simply a bad loan just because the originator's name was New Century; that's not what made the loan bad. What made the loan bad was poor credit characteristics of the borrower, information which can now easily be quantified.
Rozens: In terms of lead clues to what security will see a higher default rate, what do you look for most?
GOODARZI: LTV is the right place to start; there's no question about that. Credit behavior is going to be an indicator of what the borrower is going to do in the short term. A lot of times what people try to do is try to use these short-term behavior indicators for the long-term health of the bond that they're analyzing; that would be the wrong way to look at it. Modeling traditionally has been this business of predicting what the bond is going to do 360 months out. It is more prudent to look at the short-term value of all the variables that you have at hand. I would include in that house prices and interest rates. Those two are only valuable for a short window. If you try to go beyond a short window, you can start to lose the value of even those indicators that people have used traditionally.
XU: I would agree with Afshin, but I would add that if you're trading a bond, you have to have that.
GOODARZI: You have to have assumptions, that's true. Well, three months is certainly too short. I think four years is the minimum. But certainly if you make some assumptions out for the rest of the life of the deal.
BERGANTINO: Since for the purposes of bond valuation we have to project cashflows for the full maturity of a mortgage, we try to embed in our long-run models what appear to be the more stable relationships between the borrower behavior that we are trying to model and the explanatory factors that we have available to us (and that need to be projected forward in order to run the models). What one could do to supplement this approach is to build an augmented, short-term model that reflects any current deviations from long-run relationships as well as any additional information that is highly predictive of borrower behavior but is difficult to project far into the future, such as the wide array of credit bureau data that is the focus of this roundtable discussion. This short-term model could then be used to provide better projections of borrower performance over the next 12 to 24 months while letting the enduring relationships embedded in the long-run model dominate projections farther out.
GREEN: I just want to make the comment that it's very refreshing to hear someone thinking more than six to twelve months out, perhaps something not enough of us in the Wall Street community were doing before the financial crisis started.
ROZENS: You're thinking a lot can change in twelve months.
XU: I think Afshin did bring up a very important point. There is a limitation on the model. Not only does the predictive power of data seems to fade over time. The data do not necessarily cover a host of other factors, such as how the servicers behave given the same borrower information. Servicer A may do things totally different from Servicer B. For example, one servicer would only liquidate things that are really easy to liquidate to realize as little losses as possible. These servicers would leave other properties that had huge negative equity in the deal. Other servicers would liquidate those who have more negative equity and they realize losses upfront. You cannot necessarily model that kind of thing just using the borrower's data. But as modelers, we need to see that there are some things that cannot necessarily be modeled and we need to be very vigilant about this.
Rozens:What kind of consumer debt do you look at most closely?
XU: At the borrower level, we mostly only look at the origination mortgage data. The other information we look at is for some research purposes, and not necessarily for trading. For example, we have found that before borrowers default on their mortgage, they would probably first default on their auto loans or credit cards. Interestingly, if you look at the data over time you're going to see that this relationship actually changed a bit. In recent years, people are more likely to default on their mortgage before they default on a credit card or auto loan. I think there is a rationale behind this, because if you default on your mortgage, the banks will wait for months and, with the government policy now, maybe wait for years before you lose your home.
GREEN: You need a car to find a new job.
XU: If you default on your car, within weeks the car is gone. The same thing goes for the credit cards. The data actually enables us to see a trend. But how does this affect trading in credit card deals? It's difficult because of the revolving feature of these deals.
MAYER: One of the things that is interesting to look at is the relationship between first and second liens. One of the things that people have puzzled a lot about is why there are people defaulting. There are more people in our cut of the data that are delinquent on the first than the second, rather than vice versa, and why is that? Because if you miss an underwater second payment they're probably not going to take your house, but if you miss your first lien, you may eventually lose your home. It gets to the issue of the servicers because the servicers of the second have been much more aggressive. Looking at how these things are serviced have pretty strong predictive powers in terms of what the resolutions are going to be and there are huge differences across servicers.
XU: That's exactly the same thing that we've found. One reason I would add is that in today's market, the second lien tends to have zero equity on the loan. For example, in California, before a servicer forecloses on a home, he needs to make a decision whether to go after the house or go after the borrower. If you go after the borrower then you forfeit your rights to go after the house. If you have a second lien, it's a no-brainer, you go after the borrower because it doesn't make sense to go after the house. It may force the borrower into a bankruptcy and then you go after his wages and his bank account. But if you are a servicer of a first lien mortgage, then it's a hard decision whether to go after the borrower or the house. Most likely you'll go after the house, because this is the biggest asset the borrower has. Certainly as the servicing rights are sold, the servicer is trying to, in effect and in some cases, move the bad loan off their books or move it on to a special servicer. But even when you account for the movement to special servicers, there's still a higher rate of default in those loans that those servicers are selling off. They are seeing something that the rest of us are not in the information that they're using to make that decision, which is encapsulated in that sale. So utilizing that data is going to be very, very interesting.
Rozens:So should we watch for sales of servicing rights?
GOODARZI: It's intuitively meaningful too that a servicer dealing more directly with a consumer is seeing something either in their conversations with the borrower; they're collecting some information in their call centers; they're seeing some behavior that's not reflected in the delinquency, prompting them to get rid of the loan.
XU: Basically what you're saying is before they sell the loans, they're cherry picking a little bit?
GOODARZI: Potentially there might be some cherry picking. In fact in the transaction themselves, they're signaling there's something less desirable about that, especially when you're talking about some of the more prominent players.
GREEN: There is the argument that the first month or two the borrower is used to a certain envelope color coming in the mail and now it's a new company, with a new envelope color they are not used to seeing. They may just be throwing the new envelope away assuming its junk mail, i.e., servicer disruption.
GOODARZI: That will turn into some sort of delinquency period.
GREEN: For a period or two, and does it then level off? I'm not saying you're not on to something, but temporary servicer disruption is something to consider.
XU: We do notice there are differences between loans held onto by the original servicer and the loans that have been sold to a third party. There are several causes for the difference in the performance. There are some cases that the servicer may have bought out the delinquent loans and masked it as prepayments, if the rights stay with the originator's own book. So in that case, you'll see that prepayments actually are faster with loans serviced by the originator. When we model this, we first look at the servicer's history to see what they're doing. Another reason is the lending channel may tend to sell more of the wholesale loans versus the retail loans. The wholesale loans perform a lot worse.
Rozens: Does anybody here follow how much the paper that's originated is from independent brokers? I understood there was about 60% to 70% of the total mortgage market was from independent brokers, that's probably changed now.
BERGANTINO: It was much higher during the boom period, when there were a large number of mortgage conduits that weren't banks per se, but rather were aggregators of mortgages and issuers of mortgage backed securities.
ROZENS: Mr. Xu, you mentioned that originators or servicers in some cases would buy-out the loan. The idea would be then to resell the property as quickly as possible.
XU: Well, there are several reasons. They may need to buy-out anyway because they have reps and warranties on these loans and they rather buy it out at an earlier stage rather than wait to be deeply delinquent. If the servicing right changes to a third party, then this guy may fall into not paying this month and they keep trying to call him rather than immediately reporting to the seller and request a buy-back. In that case it's more likely for these loans to show a delinquency, especially when you're using the OTS method. With the OTS method,
if you're due on April 1st and you're still not paid up on April 30th, it's not a delinquency at all. You become delinquent on May 1st. So for a loan to become thirty-day delinquent, you should wait until June 1st. There are different pay cut delays that may even delay it further. If you are a servicer who originated the loan, and you know that you need that loss, you probably won't wait for that long to buy it back and that's just gone. Nobody would know that this loan was ever delinquent. And if you're a third-party delinquent, you probably would wait several months before you file a request for a buyback.
Rozens:What about loss severities and recovery lags?
BERGANTINO: I think that the credit bureau data is most valuable in predicting future prepayments and delinquencies for borrowers who have never been delinquent on their mortgage. The same factors that provide early warning signs of an imminent default are also likely to indicate an inability to refinance. For example, a seemingly high quality borrower who has seen a decline in their credit score since origination, or has subordinate liens not visible in the standard mortgage databases, or is delinquent on one or more of their non-mortgage accounts or no longer lives at the property associated with their mortgage is much less likely to qualify for a new mortgage than would an otherwise similar borrower whose credit file shows none of these warning signs. Moreover, there appears to be a surprising amount of variation in credit bureau data within the universe of borrowers with clean pay histories. For example, we found that approximately 20% of the borrowers in a random sample of borrowers wtih clean mortgage pay histories, had seen their vantage scores fall by 50 points or more since origination and about the same percentage had subordinate liens that were not visible in the mortgage databases that we typically have access to. As for severities, to the extent that credit bureau data can improve projections of the time from initial delinquency to liquidation it would be useful for estimating severities on securitized loans. This is because servicers of securitized loans typically advance the scheduled principal and interest due each month to investors and then recover these advances along with advanced property taxes and hazard insurance premiums at the time of liquidation.
GOODARZI:On the prepayment side, we have some indicators that can help. For example, if a borrower is out there actively looking for another mortgage, what happens mechanically is somebody's pulling credit on them and when they pull credit that's a hard inquiry that gets tag. We see that information and we can make that information available.
GREEN: If the borrower's credit health is seen as stable and improving, a credit inquiry may indicate an imminent voluntary prepayment, while if they're looking for a loan and their credit health is deteriorating than conversely, an inquiry may predict an imminent involuntary prepayment even if the borrower is still current.
GOODARZI: Which brings up the point that there are those out there who are aggressively cleaning up their credit so that they can qualify for a mortgage. Once they do, they tend to migrate back down to where they were before they started the clean up; that behavior certainly happens in some segment of the overall universe. But mortgage inquiries in general are going to give a heads up in terms of prepayment behavior when the option is in the money, as it were. Certainly that is information that is available.
MAYER:Back in the days when we thought house prices were going up forever and credit risk wasn't something we worried very much about, everything was about prepayment rates. In the interest rate environment we are in, you really worry about prepayment speeds in one sense when rates are declining, and about how quickly people prepay out of the pool. On the other hand, at this point, just getting cash is a pretty good thing. People worry a lot about credit risk, but prepayments are an incredibly important thing, in terms of the value of the pool, even just getting a few more payments for some of the positions that are deeply subordinated.
GREEN:I think it's a case by case basis. For a modeler, a certain variable may only provide a little bit of a lift on a big macro sense; it may not be a huge value added to the model's ability to predict voluntary prepayments. But for a proprietary trader looking at a particular $20 million home loan pool or a couple of tranches that you're bidding on that day, a combination of new variables to look at may give you a little bit of insight into voluntary prepayments, severity and recovery time.
Rozens: If we were doing this industry wide analysis in '05 and '06 and early '07, what do you think people would have done differently?
BERGANTINO: Having the additional data may have provided some additional warning signals but there were other visible signs that the mortgages being originated were becoming riskier. For example, the mix of mortgage originations shifted over time from mostly amortizing fixed rate mortgages to include larger and larger shares of adjustable rate mortgages, interest only mortgages and eventually negatively amortizing mortgages. At the same time, the percentage of mortgages with limited or no income verification and initial LTVs well in excess of the traditional 80% standard kept creeping up. Interestingly, what allowed market participants to remain sanguine about these mortgages was the belief that the risk associated with their lower initial equity cushions was being offset by (1) steady and rapid growth in home prices and (2) the high initial credit scores of borrowers who were taking out these mortgages. So, in a sense, credit bureau data, in the form of higher credit scores, was part of the justification for relaxing other underwriting criteria. Moreover, it appears that the market participants who saw and capitalized on the turning point took a more macro view of what was happening, not becoming mired in the details of loan tapes and instead recognizing that the broad trends in mortgage originations were symptomatic of housing and mortgage markets that were getting overheated. In essence they got it right not by having more or better data than everyone else but rather by running and believing the home price scenarios that most other participants viewed as highly unlikely or even impossible. Thus, while it is possible that the additional data may have made a difference, I don't think it is entirely clear that it would have had the preventative effect that one would have hoped for.
MAYER: One of the things that was apparent to me, we've known for a very long time that down payments are a critical issue in predicting defaults. But, the typical subprime loans in '05, '06 and '07 were no down payment mortgages. The median purchase loan in the subprime pools had no down payment and that was apparent from the start.
BERGANTINO: But the FICO scores for those borrowers were typically 650- 675, well above the historical subprime average of 600, so market participants thought, incorrectly in hindsight, that the risk associated with lower down payments was being offset by the fact that those loans were being made to borrowers with cleaner credit histories.
MAYER:As things got bad, some of the people who saw it first were originator servicers, and they took steps to not make that so clear. They were making even worse and worse loans. You had early payment defaults. The huge percentage of defaults were happening within months of origination. Had it been clearer what was happening in tracking the existing pools, it might have been easier at that point but I'm not sure in '06, '07, we might have seen it.
GOODARZI: Well, certainly borrower actions would have been visible. So for example, if someone is to take on a $500,000 loan now and they were taking on a $200,000 loan three years ago, the credit score is going to look better, because they've taking it on. They've actually paid it off, and so they're going to look better from a FICO standpoint. The trick is not just to look at that, but at the fact that this guy went from $500,000 to $200,000 and that he doesn't have just one mortgage but he has three mortgages to his name. So all of these other factors are making this guy substantially more risky than just looking at the collateral and the fact that you're going down, further and further into riskier products is certainly a clue that something is wrong. Looking at LTVs being 100 at the get go is certainly another clue. Knowing that the person had multiple mortgages is certainly a clue. Knowing that their credit score has recently jumped up, but that jump up is coincidental with the fact that they went from $500,000 to $200,000 - all of those things are going to give you clues that things are going south fast. But would anybody have listened while they were making a tremendous amount of money is a different question.
BERGANTINO:I think it was as much the fact that anyone who was a bear in 2003 or 2004 looked foolish in 2005 and 2006. Home prices and mortgage originations not only failed to slow down after mortgage rates started increasing from their historic lows in 2003, they picked up steam. As a result, mortgages that some thought were the worst vintage in history the year before turned out to be performing quite well relative to their subordination levels. I think the rating agencies were caught off guard in the same way.
XU: I think your question has two layers. One is had we had these tools, could we have seen more clearly this train wreck coming. Hindsight is 20/20. We had the bank crisis in 2007. But in 2006, very few people knew that. A lot of those in Congress and in the White House all moved to promote homeownership. They were saying those people who cannot get credit, what can we do to help them get credit. Rather than say, no you cannot do that. Now they're saying that but in 2005, they totally sang a different tune. There was so much liquidity out there, chasing after yield. Part of the reason why we're trying to paint those lenders and brokers as bad is that they gave money to people who had no clue, etc., etc. But at that time, everybody in the link was doing something to push this forward, and I doubt if we had this tool that would have stopped people in Congress to push this, probably not.
MAYER:Every other problem with a mortgage - if you lose your job or anything else - if your house was worth more than when you had it before then you sell the house and you get out it. It's almost a tautology that you have to have. The negative equity causes default, because the reverse must be true, which is if you don't have negative equity, you're in high likelihood going to sell. But I think it's still would have buttressed the view. I was certainly surprised at the extent of fraud in the system. You say: Is that a widespread phenomenon? You can always find an anecdote for anything. So, as an economist if somebody called and said I heard somebody did X, that doesn't mean that that's the prevailing standard. What was surprising is that there were things that we looked at in retrospect, discovering the kind of people that were taking on mortgages, etc., that had these had been more widely available it would be been helpful. I spent time at the Federal Reserve while some of this was happening and the Federal Reserve didn't really have access to data. Everybody would have been likely to see things a little quicker. People were gambling on house prices going up and as long as we believed house prices were going to continue to rise, it covered over everything.
GOODARZI: What's interesting is that people have continued their thinking in a slightly different way. Now they're saying as long as the LTV supports it, I think it's a good loan. It has to be underwater or otherwise. There's continuing in that thinking that HPI is the rule of the day in every sort of economic environment. So yes, it is a very, very, important factor and it always was. There is additional information and additional clarity on the balance sheet of the borrower that you can deploy now so that you can make better judgments. Looking at the balance sheet of the borrower today and trying to, in effect, re-underwrite the loan, if you're about to invest, would your requirements be any less stringent than the original underwriter or should they be?
MAYER: To put my public policy hat on for just a second, I think what the [Federal Housing Administration] is doing right now, which is all predicated on house prices rising, because they're essentially making loans to people who are having lots of problems with 3% down. As taxpayers, we're underwriting some of those. Certainly what is happening instead is that Fannie and Freddie, not only the mortgages that they're making in the future, but the past mortgages, in underwriting those, we need a lot more clarity from the government in terms of these kinds of mortgages that are sitting around, so we can apply these kinds of due diligence in understanding what is happening with those portfolios, not only for investors, but also from a public policy perspective. I'd very much like to see us continue along this road of greater transparency, because it's got to help.
GREEN: Just going back to the big question. I think Steve makes an excellent point. Two points stick out. One, there was a lot of good loan level data available through the 1010 platform that provided warning signs in 2006 and it still happened. Second, I don't think we could have necessarily avoided the crises, but more substantive updated credit information might have helped minimize the damage by helping risk managers and investors detect the dangers at least a few months earlier. I think more granular and timely loan level analysis in general would have helped reduce the severity of the crisis. It seems risk managers didn't have access to enough granular credit data to demonstrate to senior decision makers that early pay defaults were only the beginning. The people who are left in these pools, they too pose a significant credit risk. This credit risk can be easily identified by looking at other credit performance metrics like credit card balances, auto loan payment status, etc. I think that if you study the credit data for loans in Alt-A deals, you will find that there were never enough people residing in California and Florida that make $100,000 to $300,000, which is what they would need to earn to support loans of that magnitude. I think if you gave risk managers more compelling data they could have taken to their board of directors, perhaps banks would have stopped originating ABS CDOs. By eliminating six months of bad CDOs, we get rid of some of the 2007 ABS CDO vintage. That would have helped reduce a substantial amount of pain to banks balance sheets.
MAYER: The analogy I would have would be to the sort of corporate credit rating. If you had to wait until a firm misses a bond payment, that's a long time to discover they're in trouble. And if you look at households, if you have to pay until they missed their payment, you've missed a bunch of deterioration that has happened in the interim. So I think the LTV is nice tool to help see the problems that are occurring. We're using this kind of micro data in lots of places to see where consumers are going and thinking about the economy as a whole.
GREEN: We actually have customers who don't even trade securities, don't necessarily even trade whole loans but use loan level data just to feel the temperature and over all health of a geographic region.
MAYER:We haven't looked yet to see whether its predictive if you look at the extent, within a zip code of credit utilization of neighbors.
GOODARZI: We've done a little bit of work on that. We've looked at that problem. Not so much about just looking at credit utilization, but looking at basically averages for particular zip codes, for neighborhoods, zip plus four in some cases, where we're trying to understand exactly that. Are they driven by some sort of regional effect? This logically would be very much tied to other macro measures like unemployment. Michigan, for example, is a very good example of that. In Michigan, I would expect that if 10 or 20 homes in a neighborhood are in foreclosure, certainly that's going to expand the probability that my house will be in foreclosure in the not too distant future.
MAYER:The income of your neighbor helps predict your future income as it turns out.