What’s Next for the Housing Market?

Even though today’s jobs report has been much better than expected, overall the economy still shows signs of slipping into a recession and, further, the Fed remains on the path to increase interest rates very significantly. Consumption, the most important component of GDP, is starting to slow down with retail sales, adjusted for inflation, declining for a few months now. Even job growth, when based on the household survey of employment, is not as robust.

With all that given, here is what I expect for the housing market looking ahead:

The housing market is slowing with declining new/existing home sales, housing starts etc . Capital Economics is predicting a 5% decline in house prices by mid-2023 while Moody’s is predicting a 5% decline if there is a recession (0% if no recession).

One argument consistently put forward is that, even if there is a house price decline, we are very unlikely to see borrower defaults comparable to that from the 2007 housing crash. This is because today’s underwriting standards are much tighter resulting in much better borrower credit quality and also there are none of the lending excesses today that were seen in 2007. That is no doubt true but it doesn’t answer the question of how bad could things actually get? Since the 2007 recession was the most severe economic recession with the largest national house price decline since the Great Depression (I am excluding the pandemic-induced recession of 2020), saying that the current situation will not get as bad as the 2007 recession does not provide much information.

1990-91 recession

To address the question of how bad could things actually get, I use the 1990-91 recession as a reference point and note that the national decline in house prices resulting from that recession was 3% from peak to trough using the Case-Shiller Index (CSI), which was a smaller total decline than what Capital Economics and Moody’s recession scenario are predicting just over the next year. It was, however, a very different story in some regions, particularly California and New England. The CSI shows a cumulative decline of nearly 27% in Los Angeles over a 6-year period from Mar 1990 to Mar 1996 and a cumulative decline of 12% for Boston over a 3-year period from Mar 1989 to Feb 1992. The FHFA index, even though it shows no national decline, shows a decline of 13% for Massachusetts over 1989 Q4 to 1997 Q3 (see Table 7).

Nature of house price declines – variation across regions, extended period of decline

What this means is that even if there is little to no house price decline at the national level, there could be considerable variation around this with some regions suffering large declines. Further, such house price declines tend to play out over a period of 5-7 years before bottoming out.

A key observation from decades of house price data is that house prices usually decline after an excessive run-up and the bigger is the run-up, the bigger is the subsequent decline. For example, the National CSI rose 37% over the 3 years prior to hitting its peak in July 2006 but then dropped 27% before bottoming out in Feb 2012, while for the 3 years prior to 1990, it rose only 16% and then, as noted above, dropped 3%. This pattern can also be seen at regional levels. The CSI for Los Angeles rose 65% over the 3 years prior to the 1990 recession and then fell 27% bottoming out after nearly 6 years.  The same index rose 82% over the 3 years prior to its peak in April 2006 and then dropped 41% before bottoming out in Feb 2012.

Examples of currently overvalued regions

In this regard, both Redfin and Moodys have recently compiled lists of metros that they consider the most overvalued and are, therefore, most likely to see house price declines. Let’s take two cities that are among the top in both lists – Phoenix and Tampa. Over the 3 years prior to May 2022 (the last month for which the CSI is available), Phoenix and Tampa have seen house price increases of 78% and 70% respectively (and may increase some more before reaching their peaks). By comparison, over the 3 years prior to their 2006 peak, Phoenix and Tampa house prices rose by 88% and 73% and these were then followed by declines of 56% and 48% respectively with the market, in both cases, taking 5+ years to reach its bottom. Phoenix and Tampa have, therefore, experienced house price run-ups over the past 3 years that are getting close to their corresponding run-ups before 2007.

Current house price run-up as compared to prior ones

In addition, the house price run-up over the last few years has been more widespread geographically and this is unlike what was seen in 1990, when only a few geographies had big run-ups. It may even be broader than what we saw in 2007 since the National CSI has already increased 46% over the 3 years up to May 2022 while over the 3 years prior to July 2006 it increased 37% (the corresponding numbers using the FHFA index are 35% and 27% respectively) likely reflecting the fact that, in the 2007 housing crash, the largest increases and subsequent declines were seen in CA, FL, AZ and NV with smaller movements elsewhere.

Implication for house price forecasts

The conclusion, therefore, is that there is considerable risk of large house price declines in a number of markets across the country with the declines likely to be highest in those areas that have seen the highest run-ups. These declines may not get to the magnitude of what was experienced after 2007 but their impact can still be quite painful and can continue to depress local economic conditions for several years.

House price decline and impact on mortgage default

The commonly held view of mortgage default is that default occurs when some fraction of borrowers who suffer a financial shock such as loss of a job while also, because of house price declines, finding themselves with negative equity, decide that they are left with no better choice. It is also possible that, this time, high inflation by itself, without requiring a job loss, pinches some borrowers’ budgets sufficiently enough to tip them into default.

Typically, the most impacted are borrowers who bought their house at the peak. The data for this is very consistent. The 1989 vintage from New England and the 1990 vintage from California had the highest default rates. Similarly, at the national level, the 2007 vintage had the highest default rates followed by 2006 and then 2005. What this implies is that people who bought homes over the last year and possibly even over the last 2 years are most at risk of defaulting on their mortgages if large house price declines occur.

Likelihood of recession given current low unemployment rate

At this point, the reader would be right in thinking that a recession would need to happen to produce the kind of financial shocks mentioned above and may be wondering how such a recession could happen when the economy is adding so many jobs and when, per today’s jobs report, the unemployment rate has again reached the pre-pandemic low of 3.5%. In fact it is very typical for the unemployment rate to bottom out just before the onset of a recession as can be seen in the unemployment rate time series. For example, the unemployment rate bottomed out at 4.4% in May 2007 just 7 months before the start of the 2007 recession. For the 2001 recession, it bottomed out at 3.9% in Dec 2000 just 3 months before the start of the recession and, for the 1990 recession, it bottomed out in June 1990 at 5.2%, just 1 month before the recession began. Past recessions have followed different paths by which the economy got into recession. It is possible that this time, the financial stress of high inflation and the Fed’s efforts to stamp out inflation by raising interest rates may well send the economy into recession.

Current underwriting standards

As noted earlier, many experts argue that banks have tightened their underwriting standards after 2007 and, consequently, the credit quality of borrowers is better today than in 2007. In addition, we do not see today the excesses of mortgage lending such as low/no doc loans, negative amortization mortgages etc. That is no doubt true and, for this reason, any decline in house prices will likely result in a lower level of defaults than it did in 2007. At the same time, underwriting standards today are probably comparable to those in 1990. Underwriting of mortgages in 1990 did not yet use FICO scores but was based on LTV and criteria such as debt service coverage (e.g. ratio of mortgage or all debt payments to income). Low doc loans did not exist, except for a small volume in California, and subprime lending grew only in the late 90’s. Compared to 1990, other considerations are that borrowers today are under greater financial stress due to high levels of inflation though, at the same time, they also have a bigger financial cushion deriving from the support and reduced spending during the pandemic.

Implication for mortgage default forecasts

So, when comparing today with 1990, it seems that we could expect larger and geographically broader house price declines today and, given comparable underwriting and credit quality of borrowers, this suggests that the level of defaults would be higher than what was experienced in 1990 with the defaults, following the house price declines, occurring across more geographies unlike 1990-driven defaults (which were more concentrated in CA and New England).

So, how bad did mortgage defaults get in 1990 ? Default rates in California and New England were quite high though there isn’t much publicly available data on that. Figure 4 in this paper shows an almost 10-fold increase in foreclosures in Massachusetts after 1990 as compared to before 1989. Figure 5 shows a 3X increase for California but, as the data is only till 1994 and since CA foreclosures kept climbing steeply into at least 1996, the eventual results were almost certainly considerably worse than shown.

Financial system strength

One major distinction with respect to 2007 is that banks are well capitalized today. Most banks are now subject to Dodd-Frank stress testing (DFAST) with the largest banks subject to additional supervision by the Fed and OCC under the CCAR program. The stress scenarios used in DFAST and CCAR are at least of the magnitude of the 2007 recession implying that banks have the capital to survive a repeat of it. Since the current situation is likely to result in less defaults and losses than due to the 2007 recession, banks should be able to handle it comfortably. At the same time, the exposure of the non-bank sector to housing market risk is unclear given that non-banks have been playing a growing role in it.

Conclusion

To sum up – there is likely to be a large decline in house prices across a wide range of geographic areas over the next several years which, in turn, can lead to a number of borrowers defaulting in those areas. The economic impact of this is likely to be greater and more widespread than that of the 1990 recession but, at the same time, is very unlikely to be as severe as that of the 2007 recession. While there could be considerable economic hardship and suffering for many Americans from a potential house price decline, the banking system will hold though questions remain about the non-bank sector.

While a recession hasn’t begun yet, it is still timely for policymakers, financial institutions and other mortgage market players to start preparing for a possible housing downturn and its attendant implications.