Preventing the Next Housing Crisis

home for sale

mic Listen to the podcast:

Wharton real estate professor Benjamin Keys discusses a new analysis tool he an colleagues have created that could be an early-warning signal when lenders relax mortgage requirements beyond a safe level.

The kinds of loose-to-nonexistent mortgage requirements leading up to the financial crisis are not exactly making a comeback. But today some loans are brushing up close to the post-crisis regulatory lines meant to prevent a repeat performance. That is the view of Wharton real estate professor Benjamin Keys and fellow researchers who have designed a tool that could become an early warning signal whenever looser loan standards threaten to breach regulatory guardrails.

In this Knowledge@Wharton interview, Keys discusses the evolution of problematic mortgage loans, where things stand today and how to ward off future trouble. “This is a really nice indicator for when markets are looking a little bit frothy, a little bit bubbly.” Keys noted that late last year when some lenders began making loans that have echoes of the past. “They are not being called subprime or non-prime or alt-A. The new name for them is non-qualified mortgage or non-QM, ” he said. The name of his paper is “Affordability, Financial Innovation, and the Start of the Housing Boom.” His co-authors are Jane Dokko, assistant vice president of the Federal Reserve Bank of Chicago, and Lindsay E. Relihan, assistant professor of real estate economics and finance at the London School of Economics. (Listen to the full podcast at the top of this page).

An edited transcript of the conversation follows.

Knowledge@Wharton: Please briefly explain the overall idea of your paper and why mortgages that really loosen up on the credit side can lead to consequences down the road?

Benjamin Keys: When we’re thinking about the credit boom writ large, there are a few dimensions along which we might think that credit supply expanded. There’s a growing consensus that credit supply really sharply expanded and that one of the main directions that it expanded into was private label securitization. These would be mortgages that wouldn’t be held on banks’ balance sheets, and they wouldn’t be held by Fannie Mae or Freddie Mac. The GSEs (government-sponsored enterprises) instead were being held by investors privately. There are a lot of famous stories. Think of the movie, The Big Short, which depicted investors in these private-label securities. If you dig into the types of loans that are being made at that time, credit is really expanding on three dimensions. One is credit scores. They’re simply lending to people who have a weaker credit history. Second is down payments — people are just putting down smaller and smaller amounts, sometimes nothing at all. The third dimension, and the dimension that we really explore in this paper, is about the ability to pay. It’s about your income. How much is that monthly payment going to be?

“It’s an amazing feature of the housing boom. Checking your ability to pay your mortgage was simply not a requirement.”

We focus in on a set of mortgages that we describe as either “alternative mortgages” or “nontraditional mortgages” in the paper. But what we’re really getting at is any payment that’s going to be less than your traditional 30-year fixed-rate payment. This may come about through a teaser rate, where you have some artificially low rate for a couple of years. It may come about through a negatively amortizing loan, where actually the balance increases over the first few years — or an interest-only loan, where you’re paying interest but no principal for the first few years. Or, they may just stretch out the term of the loan. So now it’s a 40-year loan, and relative to a 30-year loan, your monthly payments are going to fall. We really wanted to focus in on the prevalence of these types of contracts and think about what role they played in the boom.

Knowledge@Wharton: What did you find?

Keys: There’s a chicken-and-egg problem with these types of loans. On the one hand, they may be used as an affordability tool for housing markets that have already gotten very expensive. House prices have already gone up. In that case, house prices proceed the use of these contracts, and now I can’t simply afford a 30-year fixed rate at the prices that are prevailing right now. So, a teaser rate is what’s going to get me in the door.

The other way is that the availability of credit actually fuels the rise of house prices. You can think about the causality running in either of those directions. If these products now become available, people who otherwise weren’t able to access the market — or investors who have to put very little down for a year or two — can now access the market, and they’re going to bid up prices. In the paper, we try to disentangle these two stories using a couple of different approaches.

Knowledge@Wharton: Did one dominate over the other?

Keys: Yes, we really find a “Tale of Two Booms.” In the early booms — we’re thinking here of the late 1990s and early 2000s — prices started to rise in places like Boston and San Francisco really early on. A lot of that is the early momentum around the tech boom, the dot-com boom. You start to see prices rise. But they rise in line with incomes in those markets — at least at first. And we’re really focused in on the start of the boom. What sets off the rise in house prices? You see this sort of early boom in the Bostons and San Franciscos being less driven by these types of products. In contrast, the frothiest, bubbliest markets — think of Las Vegas or Phoenix, Florida and parts of California like the Inland Empire — that’s where it looks like the use of these affordability products preceded the sharp rises in house prices. That’s interesting because it suggests that it wasn’t really an affordability issue in these places. It was really a credit supply issue, that now a different set of borrowers can access these markets, and they’re relaxing this particular dimension of access. And that’s going to lead to more bidders willing to bid up the price of houses.

Knowledge@Wharton: What’s the net result? Increasing home ownership is thought to be net good. But if it’s just a sugar high — it’s just prices going up — and they’re just chasing prices, which will someday reverse and cause problems, then it’s not a good thing.

Keys: At least in the context of the recent housing boom and bust, it’s pretty clear that there were a lot of over-inflated prices. So, the sugar high that you’re describing is exactly the way to think about it. There was a big crash in house prices, and in some markets a disproportionate crash. It’s an amazing feature of the housing boom. Checking your ability to pay your mortgage was simply not a requirement. We look at things like the low-documentation or even no-documentation loans, where you simply could state your income and say, “Sure, I make $150,000.”  And no one really went to check it.

This gets back to some of the mechanisms about the housing crisis writ large. Why were we willing to allow this to happen? You really have to start to focus on this private-label securitization market, and a market where investors are really far removed from the people making the loans on the ground. You think of the sort of now canonical stories of German pension funds, investing in a portfolio of subprime mortgages coming out of Florida.  They’re not checking incomes. They’re not checking people’s bank accounts. You start to put this distance between those groups. That’s some research that I’ve done in the past, thinking about these low-documentation loans, in particular.

This study really focuses in on this ability-to-repay dimension, and you’ve seen some interesting reforms in the market in exactly this space. There are still ongoing discussions about the ability-to-pay requirements related to mortgage underwriting these days.

Knowledge@Wharton: Are new rules are making loosey-goosey lending a lot more difficult?

Keys: It’s quite a bit more difficult. There are two policies that really came out of the reforms from the Dodd-Frank Act — the Qualified Mortgage and the Qualified Residential Mortgage. Those relate to the ability for lenders to securitize loans and not retain a risk on their books, and those loans cannot have teaser rates. That’s an important piece to this. They also can’t have 40-year mortgage terms.

We saw this at the tail end of 2018. As interest rates started to go up, we started to see some lenders — even very traditional lenders — starting to make these loans that have some shadows of the types of loans that were so popular during the boom. And now these are not being called subprime or non-prime or “alt-A.” The new name for them is non-qualified mortgage or non-QM. I think people are going to start to hear more about these types of loans.

Knowledge@Wharton: How close to the line do they get?

Keys: There’s quite a bit more documentation in terms of income and assets. That’s probably the key difference right now. But you’re starting to see slightly different standards around documentation. This has always been a challenge for folks who are self-employed, who don’t have traditional incomes and don’t have a W-2 that they can report quite as easily.

“We look at things like the low-documentation or even no-documentation loans, where you simply could state your income and say, ‘Sure, I make $150,000.’  And no one really went to check it.”

But then you’re starting to see some things that look like teaser rates. You’re starting to see some contracts that look like maybe interest-only contracts, where you’re only paying interest on the loan for two, three, five years, and then you start to repay the principal at that point.

So yes, you’re starting to see these nibbles around the edges of those requirements, and the rules are quite clear on this. That means that these lenders are either going to need to hold these loans on their books as whole loans, where they’re going to be exposed to the risk directly, or if they try to securitize these loans on a secondary market, they’re going to have to bear some of that residual risk.

Knowledge@Wharton: Is there a percentage of ownership they have to retain — so-called “skin in the game?”

Keys: Yes, so it’s going to vary quite a bit, but generally they’re going to have to hold at least 5% of the skin in the game. There are a bunch of different way in which they’ve talked about calculating that so it’s less of an obvious calculation.

Knowledge@Wharton: Five percent doesn’t sound like a lot to me.

Keys: It doesn’t sound like a lot, and different securitization markets have different thresholds for what the right amount of risk retention is. How do we discipline behavior in these markets? There’s certainly a lot of evidence that seems to point to the need for some form of risk retention, that that is going to align the incentives better than not. But right now, there are a lot of ways to avoid these kinds of rules.

Anything that’s sold to Fannie Mae or Freddie Mac, for instance, doesn’t require risk retention. So, Fannie and Freddie are bearing a lot of that risk. Now, they haven’t moved their standards into this direction either, but we saw this at the tail end of the boom, as well — that Fannie and Freddie started to relax their standards to try to keep up with the private market. So, we’re not nearly where we were in terms of 2004 to 2006, in terms of lax underwriting. But I think we’re starting to see those baby steps as interest rates rise.

Knowledge@Wharton: It’s a easy to imagine small mortgage shops around the country that would just loosen standards. But then, when the mortgages went to the rating agencies, things really changed. I have a picture of a cartoon in my head where they’re throwing kitchen sinks and everything else into one end of this big machine, and out the other end comes AAA-rated bonds, which is what the German pension companies presumably thought they were buying. They were saying, in effect, “OK, the rating agencies say the securities are good. I’ll trust them.”

Keys: Right. This comes back to a much bigger discussion about who do we hold accountable for some of the mistakes that were made? Certainly, the rating agencies are near the top of that list, if not at the top. If you look at the difference between the number of AAA-rated corporations versus the number of AAA-rated mortgage bonds, it’s just not even close. There are so many of these mortgage bonds out there with a AAA rating. And if you think about the way in which these loans are bundled and securitized, in principle — and this is something that I teach my students here at Wharton — these can really serve a benefit of diversification.

You can get different types of investors willing to bear different types of risks, but the model really matters. The underlying model matters a ton. And this is something the rating agencies got wrong. They just simply didn’t realize how correlated these different markets were.

If you look back at the mistakes that the credit rating agencies made, we haven’t really replaced that. One thing I think is keeping this type of lending to a relative minimum — now it’s growing quickly, but it’s still less than the 5% of the market — is we haven’t really seen a rebirth of this private-label securitization market. I think it’s a real lack of trust among the types of investors who would potentially invest in those kinds of AAA-rated bonds to say, “We learned our lesson last time around.” At least for the time being, some of these firms have long enough memories. It’s just a question of how long those memories last.

“This study really focuses in on this ability-to-repay dimension. There are still ongoing discussions about the ability-to-pay requirements related to mortgage underwriting these days.”

Knowledge@Wharton: What are some of the practical implications of this work?

Keys: We’re hoping to use this work to give policymakers a potential set of tools to keep an eye on markets. This is a really nice indicator for when markets are looking a little bit frothy, a little bit bubbly — and we start to see shifts to these types of contracts. In writing this paper, we built a county-level data set, where we aggregated over 60 million mortgages and collapsed them down.

It’s a big data problem, and my co-authors played a huge role in that to collapse the data down and to summarize this at the county-by-month level. Here’s what lending conditions look like at the county-by-month level. We’re not updating it monthly. But I think it’s something that policymakers should be using as a potential tool to track county by county where we’re seeing house prices tracking incomes more closely, and where we’re seeing them deviate. In the places where they’re deviating, are they deviating on those three dimensions I mentioned at the beginning — down payments, credit scores or ability to pay?

Knowledge@Wharton: And if problems arise?

Keys: Then the policy question gets a little trickier, right? Do we take the punch bowl away from the party or not? If these lending rules that are currently in place are sufficiently tight, then maybe the policymakers don’t need to do anything to react to that. They also don’t have a lot of good levers to negotiate policy at a county-by-county level, so they may have to work with state agencies or others.

One of the other big lessons of the crisis was that these lenders are going to find a regulatory home that’s most relaxed. If they’re not being regulated by folks in Washington, and they’re regulated by state regulators, this could be something that could be very useful for state regulators to keep an eye on, as well. I hope this is something that policymakers can think about building themselves.

Knowledge@Wharton: What surprised you?

Keys: Quite a bit of this was unexpected. We didn’t expect to see this divide between the two sets of booms quite as cleanly as we saw. To just see these two differences where, again, the early markets — the Bostons and San Franciscos — it doesn’t really look like these kinds of affordable mortgage products are really a key contributor.

There are definitely challenges linking incomes to house prices, given all of the frictions that come with home-building and all of the barriers to building in a lot of these cities. But then to see what happened in the later booms, and to really see these shifts away from public markets and towards these securitized loans that have these really different terms — this is really the first paper that has taken the ability-to-pay dimension so seriously. That came as a surprise to us, that it was just such a stark result, that you saw these blips in the data that just jumped out at you, of the popularity of these different contracts — to the point where in 2005 about 60% of all mortgage loans in the U.S. for purchase have at least one of these features.

Knowledge@Wharton: What’s next?

Keys: We’ll continue to try to build this data going forward, to keep an eye on these new policies. One of the interesting things about the QM policies: They haven’t been binding for the last 10 years. Lenders have been sufficiently restrained. They were burned pretty badly by the financial crisis. And certainly Fannie Mae and Freddie Mac tightened their standards, as well.

“One of the other big lessons of the crisis was that these lenders are going to find a regulatory home that’s most relaxed. If they’re not being regulated by folks in Washington … they’re regulated by state regulators”

So, for basically 10 years after the crisis, you had really little interest in taking this kind of risk. You know, why bother? Interest rates are low. People are happy to lock in a 30-year fixed rate at 4%, and that’s historically an amazing interest rate. There just hasn’t been a lot of interest in thinking about other ways to pull people into the market.

As interest rates rise over the next year or two — which is what folks are forecasting — it will be really interesting to see how these kinds of products re-enter the market and who is using them.

Knowledge@Wharton: What else should we know about this paper?

Keys: One thing that people spend a lot of time talking about is the role of the government in the housing boom and how much is it to blame? We talked a little bit about the credit rating agencies and whether they needed additional scrutiny and regulation. I certainly think they did. I think that’s the kind of stress tests that the Fed is doing now.

But one of the interesting things we find is that the Federal Housing Administration  — the FHA — was really a nonplayer in the housing crisis. Some of this is because they had extremely low loan limits. They simply couldn’t make the big, oversized loans that you needed to buy these expensive houses in these expensive markets. But one of the really cleanest things that comes out of our research is that this shift towards the private-label securities market is a shift away from the FHA.

The FHA held the line in terms of their standards during the boom, and I don’t think that’s something that gets enough attention. So, when we’re thinking about the different ways in which different federal agencies can respond to this type of risk, and we see Fannie Mae and Freddie Mac taking on more risk in 2006, and especially 2007, while the FHA didn’t — it’s useful to go back and think about the oversight of those two different groups, how they’re managed and how they’re constrained.

Citing Knowledge@Wharton


For Personal use:

Please use the following citations to quote for personal use:


"Preventing the Next Housing Crisis." Knowledge@Wharton. The Wharton School, University of Pennsylvania, 26 April, 2019. Web. 15 June, 2019 <>


Preventing the Next Housing Crisis. Knowledge@Wharton (2019, April 26). Retrieved from


"Preventing the Next Housing Crisis" Knowledge@Wharton, April 26, 2019,
accessed June 15, 2019.

For Educational/Business use:

Please contact us for repurposing articles, podcasts, or videos using our content licensing contact form.