In a recent post, we discussed macro factors driving delinquency rates across the mortgage landscape.[1] In this brief update, we make two comments, one fundamental and the second technical: First, we look at the possibility that the relatively greater financial distress observed among FHA borrowers compared to conforming borrowers at present is due, at least in part, to the relatively higher debt service levels held in the former cohort. Second, we perform an exercise matching FHA loan in HMDA to those in the Ginnie Mae disclosures to measure natural disasters’ impact on loan performance.
Debt and Delinquency [DQ] In our prior post, we noted the large differential in DQs across consumer sectors: autos, credit cards, and mortgages since interest rates began to rise in early 2022. We now have an additional quarter of data from the Federal Reserve Bank of NY which is summarized here: Recursion data was utilized in a story entitled “Fannie, Freddie Loan Requests Surging” published by Commercial Mortgage Alert on September 13, 2024. The story reports that the GSEs have recently received a “flood” of loan applications, which could indicate that CMBS production could pick up in the last half of 2024 after a lackluster start to the year. The story states “In the eight months through August, Fannie purchased $27.8 billion of multifamily mortgages, down 22% from the prior-year period, according to data from Recursion Co.”
Recursion is pleased to be a trusted source for information utilized by key public and private-sector decision-makers in the mortgage industry ecosystem. With many homeowners locked into their properties by low mortgage rates, and listings for existing homes are generally low, the onus for property availability in desirable areas falls on the homebuilders with new homes. The newly released HMDA data shows the population is flowing towards areas of higher climate risk[1], with the understanding that our research covers new homes with a mortgage only. HMDA data does not come with client risk measurements. However, FHFA provides a list of census tracts as “Designated Disaster Areas (DDAs)”, which are located “in a county designated by the federal government as adversely affected by a declared major disaster under the Federal Emergency Management Agency’s (FEMA) administration, where housing assistance payments were authorized by FEMA”[2].We incorporate these into our HMDA Analyzer, which contains lender and borrower information. By so doing, we can perform complex analyses of mortgage origination in many dimensions.
We first take a look at the heat map for 2023 DDAs provided by FHFA[3], noticing that DDAs are concentrated in coastal states, especially Florida and Texas: On August 9, 2024, Commercial Mortgage Alert (CMA) published an article discussing the impact of a policy change instituted by Fannie Mae in late June that requires a 50-50 split on all yield spread premiums with originators, removing a threshold of $100,000 previously in place.
According to CMA, “In the five weeks through Aug. 2, the agency saw a 47.7% year-over-year decrease, to $157.2 million, in the issuance of multifamily loans with sub-$5 million balances. That compares to a 13.7% drop across all Fannie loans, according to data from Recursion Co.” We are proud to be an essential source of information to market participants on these important policy issues. Regular readers of Recursion blogs will recall that the final version of the HMDA data[1] is generally released about mid-year and contains data of interest to those looking into important issues related to the role of housing in social trends. In this post, we will focus on both borrower and geographic breakdowns by loan class as HMDA 2023 is out. To perform the socioeconomic analysis, it is often very useful to link HMDA data with outside databases from other publicly available data sources such as FHFA Duty-to-Server, or FEMA National Risk Index, that provides various characteristics at the census tract level.
Low-income households Ginnie Mae discloses low-to-moderate borrower exposure at the pool level. Using its definition, Recursion constructed the flag in HMDA. We start by looking at the share that low-to-moderate income households obtained in purchase mortgage originations from 2014-2023[2]. For the first time in a while, second liens have come to the forefront of mortgage industry conversation. To a large degree, this is natural because of the unprecedented rise in home prices that we have experienced since the Covid shock. These seconds can be so-called “piggyback” loans that are used to keep the first lien under the conforming loan limit at origination, or they may be “closed seconds” that are used by consumers to extract equity from gains in home price valuations. This product may be superior to the other traditional form of equity extraction, cash-out refinancings, as this vehicle requires that the entirety of the original mortgage be refinanced, not just the extraction amount, often from a much lower level. Finally, while not strictly a loan, Home Equity Lines of Credit (HELOCs) are also popular for this purpose.
Below find the share of second liens, by both loan count and by the original loan amount from HMDA[1]: One of our ongoing themes is that pervasive structural change implies that traditional relationships between measures of market performance and economic fundamentals cannot be relied on to be stable over time. As a result, a broader view across markets is called for. Below see 90+Day delinquency rates as reported by the Federal Reserve Bank of New York’s Household Debt and Credit Report:[1] There are quite a few points to take away from this chart. Note the peaks around the time of the Global Financial Crisis (2009-2010) and the onset of the Covid Crisis (2020-2021). These correspond well with comparable spikes in the unemployment rate (10% in October 2009 and 14.8% in April 2020). But the composition is quite different. During the GFC mortgage, DQs spiked very much higher for mortgages than those for car loans, while the pattern was reversed during Covid. This can reflect differences in policy between the two periods (mortgage forbearance in 2021 but not 2011), but also, this likely reflects consumer’s views on the value of paying off different types of debt. In 2010, house prices were plummeting, so incentives to pay towards an underwater mortgage were low, but they needed their cars to get to their jobs, so they continued to pay off their auto loans. During the more recent Covid crisis period, homeowners in distress had access to forbearance programs. This brings us to the more recent period. We see that delinquency rates across products started to rise in late 2022 across debt classes. Of note is the sharp increase in 90-day DQs for credit cards from 7.7% in Q4 2022 to 10.7%, above the Covid crisis peak of 10.0% in 2021Q1 and the highest reached since 10.9% in Q2 2012. Car loan DQs rose modestly from 3.7% to 4.4% over this period, while mortgage DQs inched slightly higher from 0.4% to 0.6%. The superior performance in mortgages is no doubt due to the post-Covid surge in house prices, which has created a powerful incentive for homeowners to stay current on their mortgages. But the interesting observation is that these increases occurred even though the unemployment rate has remained less than or equal to 4.0% for the past 18 months. Clearly there is a class of distressed borrowers out there in this relatively strong labor market. Since the NY Fed data contains no detail at all in terms of these borrowers, we take a deep dive into mortgage delinquencies based on agency disclosures. We start by looking at total delinquencies of the total books of business for the GSEs and FHA: It's interesting to note that the trend in all cases is flat-to-down, in contrast to the slight rise in the NY Fed data[2]. There are many possible explanations for this, including that our data do not contain VA or non-QM loans, and that the NY Fed data is based on a 5% sample while ours covers the entire universe of outstanding mortgages in the GSE and FHA programs[3]. For our purposes, the most interesting program is FHA since these borrowers tend to have lower incomes, lower credit scores, and less experience managing debt compared to other mortgage holders, implying that they have relatively higher vulnerability to adverse economic conditions[4]. The thing is, as mentioned earlier, conditions are not that adverse. To find trends in distress in the FHA program we look at modified loans as these started out as loans that were determined to be unsustainable to begin with. In fact, performance here has deteriorated on trend over the past couple of years: The 90-day DQ rate for these loans almost doubled from 4.1% in February 2022 to 8.0% in March 2023 and has been in a volatile but flat trend since that time. We look for fundamental trends to scale this trend, in general, from labor market indicators, which we feel is the key driver of mortgage performance. There are many such candidates, but here we look at the Kansas City Federal Reserve Labor Market Condition Indicators (LMCI)[5]. These indicators consist of 24 distinct labor market variables. The indicator we use is the LMCI Level of Activity index. It is scaled so that a number above zero indicates a stronger level of activity than the long-term average. In the chart above, we invert this indicator so as to see the correlation with rising distress in FHA modified loans. It seems that modified loans are sensitive to deceleration in labor market activity. We can dig deeper into this phenomenon by breaking the data down into different categories: Here, we break the FHA universe into four categories: ET loans, mods ex ET, RG and other purposes. ET is extended term, which are generally 40-year mortgages. RG is Reperforming, which is not a mod but a loan leaving forbearance with a partial claim. All four categories exhibit quite different behaviors with ET loans displaying sharply worse performance than the other categories. It’s important to note that the ET share is quite small but growing: The above chart points out that while looking at DQ charts is informative, they can be somewhat misleading in situations where programs have been around for different lengths of time. One way to get around this is to look at particular cohorts. Below find DQ’s for the 2022-2024 cohort: It’s interesting to note here the stability in the Mod ex-ET and other categories. Growing distress is visible for the newer RG and ET programs, although the DQ rates for both peaked earlier this year. Finally, another way of looking at distress over time in a consistent pattern across programs is via Early Payment Defaults (EPDs). These are the share of loans that experience two or more months of missed payments in the first six months after origination: EPDs jumped with the onset of Covid but quickly fell back after forbearance and other policy programs were instituted. More recently, we can see distress in the various mod programs while loans in the Other category exhibit little such behavior. This difference is interesting because the incentive to pay for more recently-originated mortgages is very much less than those prior to 2021 given the recent jump in home prices. To conclude, examining the behavior of nonpayment of debt is a very complex task, which varies by loan class, and subclass. It depends on the financial health of borrowers, and the state of the economy, particularly with respect to labor markets. It is important to avoid making broad statements about trends in DQs without taking these factors into consideration. Finally, we do not see broad signs of distress in consumer debt markets (particularly credit cards) spilling over to mortgages, except possibly in the most distressed cohorts. We conclude that mortgages are not yet a source of general financial concern, but the situation needs to be closely monitored, particularly within the most distressed category of borrowers. [1] https://www.newyorkfed.org/microeconomics/hhdc
[2] This contradicts a widespread view that FHA loan performance has deteriorated in recent quarters. In fact, the 30-day DQ rate for FHA has been in an upward trend over the past two and a half years. (To as high as 6.5% recently, according to FHA Loan Performance Trends Report). Most of these however, cure before hitting the 90-day threshold. [3] See New York Fed Household Debt And Credit Report (https://www.newyorkfed.org/microeconomics/hhdc), p44, paragraph 1. [4] For 2023 issued GSE and FHA loans (disclosed by Agencies), the WA credit score is 754 vs. 676; while the share of first-time home buyer of purchase loans is 46.5% vs. 75.0%. In addition, according to HMDA 2023, the WA borrower’s income is $301k for conventional loans and $112K for FHA loans. [5] https://www.kansascityfed.org/data-and-trends/labor-market-conditions-indicators/ In recent months, we have witnessed increasing attention being focused on the Government mortgage programs, particularly VA. Two issues in particular have generated considerable discussion. First, last year VA allowed its forbearance programs to expire without a backstop for distressed borrowers wishing to avoid foreclosure. Instead, VA implemented a voluntary foreclosure moratorium for their servicers, which has been extended to the end of 2024. And very recently, VA implemented a new program, the Veterans Administration Servicing Purchase Program (VASP), to help households in need, with full availability required by December 31, 2024[1].
Second, there has been a noticeable pickup in VA prepayment speeds compared to other major programs, including FHA[2]. This is particularly evident in higher coupon securities: |
Archives
September 2024
Tags
All
|