Wednesday, July 15, 2015

Housing Tax Policy, A Series: Part 43 - First Time Buyers and the Housing Crisis

Bill McBride points to this interesting paper from FHFA regarding first time home buyers and the housing crisis.  From the abstract:
First-time homebuyers are younger and have lower credit scores, home equity, and income than repeat home- buyers, and therefore are comparatively less likely to withstand financial stress or take advantage of financial innovations available in the market. The distributional make-up of first-time homebuyers is different than that of repeat homebuyers in terms of many borrower, loan, and property characteristics that can be determined at the time of loan origination. Once these distributional differences are accounted for in an econometric model, there is virtually no difference between the average first-time and repeat home-buyers in their probabilities of mortgage default.
One takeaway is that the high prevalence of first time buyers late in the housing boom increased the danger of subsequent defaults even if banks were not loosening credit standards, because of these distributional differences.  There are some interesting details to consider here.

First, if there was an increase in the number of first time buyers during the boom, and first time buyers have lower credit scores, then the relative strength of credit scores in general during that period means that, after adjusting for the effect of the larger number of first time buyers, banks had to be tightening credit standards even more sharply among repeat buyers for the average credit score to remain high.  This is further evidence that banks were accounting for the factors that might have been raising FICO scores of repeat homebuyers.

The paper begins:
While there are widely-perceived benefits of homeownership and overwhelming support for homeownership in the United States, the Great Recession also brought to the fore the perils of unsustainable homeownership. During the crisis, a record number of homeowners became unable to pay their mortgages and many lost their homes. According to a Core Logic report in October 2014, there were 7 million foreclosure completions in the United States since the second quarter of 2004 when the homeownership rate peaked (equivalent to 15 percent of all mortgages) and over 5 million of those foreclosure completions occurred since the financial crisis began in September 2008.
The narrative that homeownership was unsustainable is a prior here, not an interpretation of the data.  The author even points out that there were less than 500,000 foreclosures per year for the first four years after homeownership peaked, and at almost all of the foreclosures happened after September 2008 - after home values in the major cities had fallen 25% from their peaks.  A record number of homeowners didn't become unable to pay their mortgages during the crisis.  A record number of homeowners became unable to pay their mortgages after the crisis.  The fact that more than 5 million homeowners were foreclosed on after September 2008 is says little or nothing about the sustainability of homeownership in 2004.  Delinquencies and foreclosures spiked in late 2008 and 2009, but homeownership has been declining fairly linearly from 2005 to today.  There is enough of a parallel here to make these issues look like they are related, if we are primed to see them as related.

If we weren't primed to make this connection, would this even be an interpretation that would seem plausible?  Since 2005, mortgage originations have remained high, then collapsed.  Home prices continued to rise, then leveled off, then collapsed, then leveled off, then recovered.  Delinquencies and foreclosures were low, then a little high, then exploded, then slowly declined.  And through all of these extreme fluctuations, homeownership rates have fallen steadily by 1/2% per year.

First Time Homebuyers, Home Values, and Interest Rates, and the Timing of Events

Here is a chart from the paper.  We can see here that repeat homebuyers peaked in 2003.  Really, it is more descriptive to say that repeat buyer originations were at a high plateau from 1998 to 2007, with a peak from 2001-2003 and decline from 2003 to 2007.  First time homebuyers peaked later, in 2007.

There are several interesting things to think about.

1) Home purchases by repeat buyers peaked in 2003.  This was relatively early in the price boom and before the sharpest increases in price.  The portion of homebuyers that were existing homeowners has fallen consistently since 2003.  Consider this trend in light of the notion that the height of the boom was fueled by homeowners who were propping up their spending with real estate capital gains.  Originations by repeat buyers were about as high in 1998 and 1999 as they were in 2004-2007, even though home prices in the 1990s had been declining relative to incomes.  New home purchases were increasingly made by new buyers, who would not have had balance sheets that were artificially boosted by real estate gains.  In fact, as prices rose, these buyers would have needed to save more to increase their down payments just to keep Loan To Value levels (LTV) stable, as property values rose.

2) Entry level home sales have been low since the crisis.  This is a product of the broken mortgage market.  As I pointed out in this recent post, banks are generally only lending generously to the top quarter or so of households (by credit quality).  Entry level home buyers can't qualify for mortgages.  But, note the disconnect between entry home sales levels and the relative number of new buyers versus repeat buyers.  What we are seeing is that households that kept their homes are sitting on capital losses that have damaged their financial flexibility.  The data in this study counts households that have not owned a home in 3 years as new homebuyers, so I suspect that many of these new buyers are former homeowners who sold or defaulted on their previous homes.  In any case, there is not a lack of new homebuyers.  There are a relatively normal number of new homebuyers.  Where we are short is of households around the median financial level who can get mortgage credit in today's environment and households who have weathered the real estate bust with enough home equity to facilitate a transaction.

3) The increase in first time buyers in 2004-2007 coincided with rising short term interest rates.  The yield curve inverted by mid-2006.  So, the top years for first-time homebuyers post-dated the period where ARMs would have been tempting but dangerous.  By the time first-time homebuyer originators topped out in 2007, the Fed Funds rate was already on its way back down.  By that time, ARM mortgage would have been advantageous for homebuyers.  The surge of first time homebuyers, given the timing of events, was not due to a prevalence of teaser loans, and clearly was not due to excessively loose monetary policy.

4) Here are two graphs regarding incomes of homeowners since 1992.  Homeownership rates began increasing in 1995 and peaked in 2004.  During this period, the average dollar banks lent out went to the household at about the 85% income level.  The income profile of the typical mortgage holder wasn't changing.  At the peak of the boom, from 2004 to 2007, after homeownership rates began to decline but prices were still high, the average relative income of mortgage holders increased.  During that time, the ownership rate of the top 40% of households (by income) increased and the decrease in homeownership came from the bottom 60% of households.  We might be tempted to look at the figure from the paper and assume that the jump in first-time buyers was part of a last gasp effort by the mortgage industry to rope in marginal households.  There are many anecdotes along these lines.  The data says that lower income households were net sellers during that period and that the buyers at the end of the boom were increasingly high income.

5) Remember that homeownership rates peaked in 2004.  So, the period with the sharpest rise in first-time buyers, between then and 2007, happened when there was a fall in the ownership rate.  There must have been an unusual churn in the pool of homeowners during that time.  As pointed out above, that churn, on net, was of higher income households buying and lower income households selling.

6) Rent inflation was especially high during this period.  And, the problem cities (generally in the Case-Shiller 10 Index) which had especially high rent inflation, saw their steepest home price gains, relative to the rest of the country, during the 2004-2005 period.  Homeownership peaked in 2004, and housing starts collapsed at the beginning of 2006.  All of this suggests that during the final phase of the boom, the collapsing housing supply was creating pressures on rent, and that the buyers during this period tended to be high income households in large cities who had had the means to own, but had previously not purchased for other reasons.  They were purchasing a hedge against rent inflation, which, after a decade of being high, was now accelerating.  It must have been exasperating.

Since prices topped out at the end of 2005, but rents were rising sharply, this period from 2006-2007, when first-time homebuyers surged, actually had sharply falling Price-Rent ratios.

7) It is interesting that while mortgage originations and total mortgage levels were still growing into 2007, homeownership rates had been falling since 2004 and housing starts had collapsed precipitously starting at the beginning of 2006.  Also, note in this graph how multi-unit structures built for sale (the sort of homes one might see in the core of a large metro area) surged in 2004.  Also, note that household real estate leverage began to move up sharply at about the same time as the collapse in housing starts.

This has been a mystery for me.  I have been blaming the crisis on tight monetary policy that started all the way back in early 2006, when the yield curve first inverted and monetary growth was collapsing.  But, if mortgage issuance continued to grow at a healthy pace for nearly two more years, why did housing starts collapse back at the beginning of 2006.  This new information on first time homebuyers makes the picture more clear.

The fall in homeownership and the collapse in housing starts were the first stark signs of the liquidity crisis.  By then, households weren't using home equity LOCs as ATMs; households were getting out of the real estate game - especially households with lower incomes.  Between the first-time buyer originations shown above through Fannie and Freddie, and the decline in homeownership, nearly 3 1/2 million households must have exited homeownership during the 2005-2007 period, plus however many first time buyer originations happened outside the enterprises during that period.

There was still a lot of housing sales activity, but this was facilitating a huge churn out of real estate.  See in the graph of equity and mortgage levels how home prices were still the same in February 2007 as they had been in March 2006.  But, households lost nearly 10% of their home equity during that period!  You might think that this is because they were tapping their home equity LOC's.  But, that is incorrect.  Home equity lines of credit stopped growing in the summer of 2005.  They were flatlined by then.  (They did briefly rise again at the beginning of 2008, and especially after the disastrous FOMC meeting in September 2008, where LOC levels shot up as households grasped at any cash they could access.)

So, the drop in real estate equity that came after the Fed pegged the Fed Funds Rate at 5.25% wasn't from irresponsible households sucking equity out of overpriced homes.  It was due to this transfer of ownership between households.  The more leveraged new first time buyers were focused on the large, supply-constrained metro areas and had higher incomes than the sellers.  The new buyers were reacting to the supply problem, which as of the beginning of 2006, had turned from chronic to acute.  The liquidity crunch was partially being played out in a collapse of home equity.

(As a side note, I wonder if the liquidation of home equity that came from this transfer of ownership was one factor that buoyed the stock market until late 2007.)

Characteristics of First Time Buyers

There are a lot of graphs in the paper showing characteristics of first-time and repeat buyers.  Some can fit the conventional narrative.  There was a positive skew in Payment-to-Income and Debt-to-Income ratios that raised the average PTI and DTI during the boom for both first-time and repeat buyers.  At the risk of seeming stubborn, I will point out that:
(1) In the aggregate during this period, households were adjusting to rent inflation by maintaining nominal housing expenditures, thus lowering real housing expenditures.  But, rent inflation, especially in the supply-constrained large metro areas, has been creating cost pressures for both renters and buyers.  In other words, the positive skew here was coming from the higher cost of housing, centered in supply-constrained metro areas, not necessarily from a higher cost of owning relative to renting.
(2) Counterintuitively, payments should increase in a low real interest rate environment.  Also, an environment where rent inflation is persistently higher than non-rent inflation is mathematically similar to having even lower real interest rates, and mortgage payments should be relatively higher at the inception of the mortgage, since the advantage of ownership comes from lower payments in future years.

There are several characteristics which don't fit the standard narrative, and which point to the constricted mortgage markets that have been in effect since the crisis.  Incomes for both first-time and repeat buyers were stable through the boom, confirming the data I have previously referenced from the Survey of Consumer Finances.  In the later boom years, incomes of repeat buyers appears to have risen at least as strongly as first-timers, and in the constrained post-bust market, incomes of all buyers have had to be very high.  Data from several sources (survey data, conventional loans, and subprime loans) now fails to back up the idea that homeownership and home prices were inflated through systematic predatory lending to lower income households.

FICO scores increased throughout the boom for both first-time and repeat buyers.  In 2006 and 2007, when there was a surge of first-time buyers, their average FICO scores did fall back to the average FICO levels of the 1990s.  Since this same cohort had rising incomes, this may reflect the influx of households with higher incomes, but some other characteristic that created obstacles to mortgage credit, who were willing to take non-prime terms in order to escape rent inflation exposure.  Since the crisis, FICO scores for both types of buyers have risen to above previous norms, mostly because only about 10%-15% of mortgages are going to FICO scores under 700 points, which was close to the average FICO score for new mortgages in the mid-1990s.

Average Loan-to-Value (LTV) was stable until 2004.  From 2004-2006 LTVs decreased for both first-time and repeat buyers.  LTVs then increased to a level above the previous norm for first-time buyers for one year, 2007, before falling to levels below the previous norm since then for both types of buyers.  By 2013, LTV had climbed back to the previous norm.

As with other reports I have looked at, delinquencies appear to be overwhelmingly a product of capital losses beginning in 2007.  First-time buyers have a naturally higher default rate, because of the difference in their characteristics, but considering that the most significant difference is probably a higher LTV for first-timers, it is surprising that there weren't more defaults among that group as prices dropped.  It looks like many homeowners have decided to ride out the crisis with negative equity.


  1. "If we weren't primed to make this connection, would this even be an interpretation that would seem plausible?"

    I don't see a tension here. As long as housing prices are higher than purchase prices owners have the option to sell or refinance in some way and prevent delinquencies. You could even concoct a story based on Graph #1 where the young buyers start trying to sell en masse in 2006/2007 and that is what drives a chunk of the price decline (I am not supporting that as a good theory without more research, just an idea that occurred while thinking about my reply). The rest of the story also doesn't have the major holes that you propose (at least from the evidence you present). Generically speaking defaulting on a major loan like a house should happen only under great pressure. You don't stop making payments on your house on day 1 when you lose your job (you will have an average of 45 days before it is 30 days overdue anyway and making a single months payment from UE benefits or savings will push that to 75 days). Defaulting on large loans is often expensive and to be avoided if you think that a new job is around the corner. It is only after either your savings run dry or you start to believe that the downturn will be deep and long that you should consider cutting out your house payment from your expenses.

    1. Thanks for the input, baconbacon. Your reply is definitely the sort of thing I'll have to address as I pull this together. The details of the timing are important, and you are right that defaults wouldn't be immediate.

      But, look again at that first graph. Homeownership topped out in 2Q 2004 when the Case-Shiller 10 was at 175. 2 years later, it was at 225. And a year after that it was still at 225. The "home ATM" was still spitting out money for a long time after homeownership began to fall. The peak of the price boom, a phase that lasted several years, was associated, on net, with low income owner-occupiers selling to higher income investors.

      There's a bit of circularity in my challenge to overturn the narrative, because to me there is an obvious signal of monetary tightening when CPI rental inflation skyrockets while homebuilding takes a sharp downturn, right as the yield curve inverts at the end of the period of Fed Funds increases. But, if someone is convinced that money was too loose, they will see this as a virtue, and generally I don't think I will be able to convince them that killing the housing sector was unnecessary. I see people reference articles from 2001-2002 already calling the housing market a "bubble". If a plurality of people think that is definitive, then I doubt they will listen to me.

      But, in the end, the path to overturning the narrative regarding your points on the timing of the crisis goes back to all those peripheral details, like high rent inflation throughout the period and all of the evidence that marginal home buyers didn't, in the aggregate, have lower incomes, lower downpayments, lower credit scores, etc.

      There is a logical fallacy of making a series of arguments which each have an 80% chance of being true, but that, taken as a series, have a negligible significance (.8x.8x.8x.8x.8....), but I think a lot of these points I have found undermine the narrative in parallel - rent inflation, monetary policy, borrower characteristics, etc. So, in the end, I have to argue that the conventional story is wrong about point A, and separately, it is wrong about point B, etc. I think this is a stronger position, logically, but harder, rhetorically, because, these points must be addressed one at a time, but the observer will tend to be within a regime. We don't think of these things probabilistically. So, for me, there was a regime shift, where I allowed myself to look at the data without a bias for the conventional narrative, which mainly was the result of accepting Scott Sumner's contention that Fed policy was the trigger for the crisis. I think people have to have a sort of conversion moment on something like this, which is not an easy transition.

      Sorry, I've gone on too long.

    2. Thinking about this, maybe one of the tasks I should take on is imagining the data that we would expect to see if the conventional narrative is right, and then specifically look at differences from that in the empirical data....

    3. I would be happy to help you hash out the strongest possible case to argue against if that becomes your goal.

    4. Baconbacon, that would be awesome. I am starting to organize all of these ideas so that as I finish up over the next couple of months, I can put the whole thing into book form. I'd love to begin the book with an honest exposition of the conventional narrative so that I can then run through all of the empirical evidence and compare it. The problem I have is that there is a lot of narrative infilling as the evidence comes in. So, I think a good rhetorical technique would be to state the narrative in affirmative form that a skeptical reader would buy into, so that they will have affirmed the way they would expect the data would look if the narrative was true.

      I don't want it to be a gottcha. I want it to be honest. I just want the reader to have mentally committed to their counterfactual. And, I think it would be best to have that counterfactual done by a third party. Having seen my version of events, you might already have an idea how the counterfactual might be tweaked to my rhetorical cost or benefit, so maybe we are both compromised a bit in that regard. But, I think your input would be helpful.

      Does that make sense?

      Honestly, if you are offering, I would love to see what you can put together.

    5. This comment has been removed by a blog administrator.