In a recent post[1] we looked at the evolution of the FHA purchase mortgage market share broken down between areas with a high percentage of Low-Moderate Income (LMI) households and those without. While the overall FHA share has generally declined since the onset of the pandemic, its share has held up in areas with a preponderance of LMI households. There are many factors behind these trends, but a natural consideration is underwriting standards.

To examine this factor, we use the Recursion Matched Dataset[2], where we create a large sample of loans with characteristics from both HMDA and the Agency disclosure data. A very high share of mortgages can be matched using our proprietary algorithm over the years 2018-2020. The coverage ratio from the Matched Dataset is provided in a previous post[3].

We proceed by looking at three major underwriting characteristics for LMI and non-LMI areas for FHA and the GSE’s: Credit Score (CS), Loan-to-Value (LTV) and Debt-to-Income (DTI),
Most interesting is Credit Scores:

In a recent post[1], we looked at the declining FHA share of the purchase mortgage market relative to the GSE’s across a variety of price points. Another way to look at this question is by a geographic breakdown focusing on those census tracts with a Low-Moderate Income (no greater than 80% of area median income) population greater than 51% (we will call these LMI areas)[2].
​
To address this issue, we utilize the HMDA dataset, and then apply the LMI information to compute shares of originated purchase loans delivered to FHA vs GSE. This is done on both a loan count and loan balance basis.

One of the many recurring themes of these posts is that the shock of the Covid-19 Pandemic and subsequent policy response has resulted in structural changes in behavior that cause loan performance metrics to shift compared to the pre-crisis world. An interesting example of this can be found in the performance of modified loans in Ginnie Mae programs.

Modified loans in these programs are those that have been purchased out of pools by servicers that are past due that subsequently have features such as rate and term adjusted in order to bring households back to a current status. These are then often resecuritized into a new GNM pool.
​

• Securitization of modified loan volumes more than doubled after the pandemic

The Covid shock of course has resulted in a surge in delinquencies and triggered large volumes of buyouts and modifications. Below find a plot of the share of modified loans of total new issuance for FHA and VA programs since January 2020:

​The recent unprecedented surge in home prices to a record 18% jump on a year-year basis as measured by the FHFA purchase-only index brings affordability front and center to the current housing policy debates. In May 2021, indexed home prices stood 15.5% above indexed aggregate earned income, a bit less than half of the peak house price overvaluation of 29.0% reached in December 2005, just before the onset of the Global Financial Crisis.

The topic of affordability is very broad, and will be the subject of much further commentary, but in this post we look briefly at this topic through the distribution of the purchase mortgage market across securitization agencies, notably FHA and the GSE’s.

Looking at the distribution between the GSEs and FHA is informative in this issue because the FHA program is aimed at low-income borrowers. According to 2020 HMDA data, the weighted average household income for FHA borrowers of purchase mortgages was $85K while for those in conforming mortgages the figure was $228K.
​
Since the onset of the Covid-19 pandemic in early 2020, the share of FHA purchase mortgages of the total[1] delivered to agency pools as been in general decline, on both a loan count and outstanding balance basis:

​With a base consisting of relatively lower-income borrowers, it makes sense that the borrowers in this program are struggling to qualify for loans in a skyrocketing market. To check this out, we calculate the change in the distribution of loans between FHA and the GSE programs by original loan sizes:

Intuitively, larger loans comprise a greater share of the distribution of purchase loans in both programs between January 2020 and July 2021.
​
Over this period, FHA lost a bit over 5% in market share to the GSE’s in this category. The change in share by loan size bucket and the contribution of each of these to the total loss in share is given below:

​In fact, it turns out that about three quarters of the loss in FHA’s purchase market share comes from losses in loan sizes less than $250,000. Further analysis is needed to look at the fundamental and structural factors that are behind this result.

As we accumulate more data at a fine local level, the opportunities to evaluate policies derived from insights into lender, borrower and supervisor behavior grow massively. Our recent post looking at the potential impact of FHFA’s new rate mod policy[1] is our most recent example of the application of digital tools in the policy space, but as noted previously[2], the new policy framework is designed to focus on wealth creation and housing sustainability at the local level.

To look at this issue, we need to have data at hand that tells us which local areas have a preponderance of low-income borrowers on which we can overlay the HMDA data set, which reports census tract level indicators related to the policy issue at hand.

It turns out that the income data can be found in the American Community Survey (ACS)[3]. A key facet of this survey is information regarding the share of every census tract where low and moderate income (LMI) people comprise less than or equal to 51% of the total population. This data is available through the HUD Exchange[4].

With that, we now have a robust tool for analysis. An immediate challenge in this regard is to come up with specific queries out of the myriad of possibilities that demonstrate their power. Below finds a chart that provides a big-picture view of lender behavior in LMI neighborhoods broken down between banks and nonbanks[5]. Specifically, we look at the trend in purchases from third-party originators of both FHA and conventional loans:

​Assigning letters to economic recoveries (“V”, “L”, “U” etc.) has become a standard part of the economist’s toolkit for expressing a view on the nature of a particular forecast. The Covid-19 crisis has added a new letter to the lexicon, “K”. In a “K-shaped” recovery, some segment of the population experiences relatively strong growth, while others are left behind. Since housing tenure is an essential determinant of the distribution of household wealth, it is not surprising that we can clearly see this shape in the relative trends in house prices versus rents:

Sometimes, future trends can be seen in the weeds. In this case it’s the 12 FHA and 1 VA mortgages (out of tens of millions) that were securitized this month in Ginnie Mae pool G2 CA8080, the very first RG pool, issued by PNC Bank, delivered to the GNMII20C program. This pool type was first announced by Ginnie Mae last December 4[1], and consists entirely of loans that were bought out of pools and cured with partial claims. These are eligible for resecuritization after 6 months without a missed payment. A previous announcement was made by Ginnie Mae last June that prohibited loans in forbearance from being bought out of pools and resecuritized into any existing pool type[2]. This rule was enacted after large banks purchased a massive number of loans in forbearance and resecuritized them immediately, leading to concerns on the part of investors[3].

Is there anything interesting about these loans?

The loans were all originated in 2011-2013, so they are pretty seasoned. Note rates range from 3.75% - 4.25%. Underwriting characteristics vary considerably, with credit scores ranging from 533 to 829, for example. While original LTV’s are generally high (8/13 greater than 90) home price appreciation over the last 8-10 years likely implies that borrowers have considerable equity.
​
More of this to come as forbearance programs begin to run out later this year.

In a recent post, we discussed the application of the FHA Neighborhood Watch dataset to understanding the market landscape for this program[1]. Peering a bit deeper, more insights can be obtained. We just updated this dataset through December so it is an opportune time to take a look at FHA loan performance.
​
First, the share of FHA loans in pools [2]continued to decline at the end of the year:

​The loans in pools fell by about 60,000 in December while the total fell by 40,000 implying that perhaps 20,000 loans were purchased out of pools, and presumably modified as foreclosures are currently forbidden. Interestingly, the number of loans in pools new issuance with mods rose for the first time since July:

It shows even though most of the loans are expected to be cured by partial claims, modification remains a tool to work out delinquent loans. We will have separate pieces focusing on partial claims in future posts.

Now what about delinquencies? What is the delinquency rate[3] of loans in the FHA program?

​As servicers may buy serious delinquent loans out of pools, and banks tend to hold conventional loans not FHA loans on their balance sheet, the overall FHA delinquency rate reported by FHA Neighborhood Watch data is generally higher than that for loans in pools. When COVID-19 first struck last spring, the 30-day delinquency rate spiked, narrowing the gap with the total figure, but many of these cured as labor markets recovered. More recently, lenders have picked up the pace of purchasing delinquent loans out of pools, as they have the financial incentive to modify the loans to allow the borrowers to become current and then resecuritize them. A key question for 2021 is when forbearance programs expire, how many borrowers will be able to work with lenders to keep their homes, and how many will lose them? Stay tuned.