In a recent post we compared the average credit score at origination history from NSMO with that from the agency loan-level datasets on an apples-to-apples basis.
In today’s blog, we improved the matching by using the Analysis Weights that are provided in the NSMO dataset to compute the averages. NSMO Analysis Weights are the product of the sampling weight and a non-response adjustment, which can make the NSMO dataset more representative.
We have commented previously on the rising share of nonbank deliveries to the GSE’s in the wake of the Covid-19 crisis, but the data just released for the month of July shows this trend to be picking up at an astonishing pace. This time, let’s break the market up into two pieces: Purchase and Refi:
In a prior post, we demonstrated how the loan-level data for the CRT reference pools could produce a good estimate for delinquency rates by state in the conforming mortgage market. One way to confirm this would be to compare these to the unemployment rates by state. In fact, a scatter plot of 30+ day dq’s from the Freddie Mac STACR program vs unemployment rates reveal a striking correlation:
The MBS market is the second-most liquid market in the US after the Treasury market. The collateral for these securities is completely domestic but a significant share of the ownership is held by offshore entities. As such, investors need to be cognizant of the attitudes of overseas investors towards this market, particularly in the current highly charged geopolitical climate.
The following chart shows total holdings of Agency and GSE-backed securities over time broken down by a broad category of investors. The most recent figure is as of Q1 2020; Q2 data will be released later in September.
The National Survey Mortgage Originations (NSMO) is a quarterly mail survey jointly funded and managed by FHFA and CFPB. It provides unique and rich information for a nationally representative sample of newly originated closed-end first-lien residential mortgages in the United States, particularly about borrowers’ experiences getting a mortgage, their perceptions of the mortgage market and their future expectations. Beginning with mortgages originated in 2013, a simple random sample of about 6,000 mortgages per quarter is drawn for NSMO from loans newly added to the database.
A recent Fed paper referencing the NSMO dataset can be found here: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3422904
The paper analyzes which borrower types appear to overpay due to a lack of shopping and knowledge about mortgages based on the NSMO dataset.
The NSMO Database comes in with a long lag: the last update was on February 20, 2020. It contains 29,962 sample loans originated from 2013 to 2017. Below are several highlighted observations:
The NSMO dataset contains 2 types of data points: first, 309 borrower characteristic variables such as borrower age, sex and shopping mortgage behaviors, and second, 118 underwriting characteristic variables, such as loan size category, LTV, DTI, loan type, loan purpose, etc. Especially interesting is that credit score and delinquency status history can be traced from origination date to Sep 2019 for each loan.
The purpose of the line charts below is to show an assessment of data quality on the NSMO dataset. We compare the average credit score at origination history from NSMO with that from the agency loan-level datasets on an apples-to-apples basis. The results show credit score histories from the two data sets match quite well.
From the below table, iIt does appear that the sampling in the NSMO data selects loans with modestly higher credit scores than that in the agency loan-level data. It is an approximately 6-10 basis points across each loan size category.This may reflect a tendency of lenders to keep higher-quality loans on their books than they deliver to the agencies. This is not the case for Government programs as a greater share of these loans are delivered.
The following two tables demonstrate loan type and GSEs distribution in the NSMO dataset.
With the release of 2019 HMDA data, we now have two years of loan-level information that contains both demographic and credit characteristics. Demographic information in HMDA includes income, race, and geography down to the census tract level, while credit characteristics include DTI. Our agency loan level databases contain a richer set of information regarding lending characteristics, but limited data on geography and demographics. For institutions looking to benchmark their performance in affordable and minority lending for regulatory purposes, 2019 HMDA, with data on thousands of lenders, is an invaluable tool. If you are interested in finding out more, please reach out.
There are of course policy uses for this data as well. A significant difference between HMDA and the agency pool loan-level data is that HMDA contains data for loans held on book, the so-called “Unsold” category. This allows a comparison of loans that banks originate and keep and those they deliver. We can break this down in any number of ways, but let’s look at it for conforming loans broken down by DTI.
In the table above, we can readily observe that banks tend to keep higher-quality loans (as measured by DTI<=43) compared to those they deliver to the Enterprises. Of course, this is not a complete picture of this issue; there are many other ways to slice the data (credit score, LTV, loan size, geography). Moreover, as there is a correlation between low LTV and desirable loan characteristics for regulatory purposes (minority status, low income), we cannot simply conclude that it’s a matter of keeping the best for themselves.
A second interesting question is: did behavior in this regard change between 2018 and 2019? Below you can find a chart of the change in the distribution between unsold and delivered loans between these two years.
It appears that banks kept more of the loans associated with very low levels of indebtedness (DTI<35) in 2019 compared to 2018, while they distributed a small share of higher-risk loans across the spectrum of DTIs above that level.
Explanations for such behavior are the subject of future research.
In a recent post we looked at the agency composition of the recent surge in MBS production. We observed that Ginnie Mae’s market share in the three months to July 2020 is significantly below that experienced in the same period a year ago. The bulk of the decline is due to a drop in the refi share, while the purchase market share experienced less than a 1% decline. There are several drivers of share for purchase market share, including program design, the coupon spread between the government and conforming sectors, and differences in the credit boxes between the two. This note looks at competitiveness through the lens of the latter factor, credit boxes.
As noted in the prior post, Ginnie Mae is the securitizer for four different programs, with the two biggest being FHA and VA. VA has a fairly unique set of program requirements, so the main competition in the purchase mortgage space is between FHA and the GSEs. In the three months to July 2020 FHA lost about 2.5% of its share in the purchase mortgage space compared to the GSEs. Let’s dive into credit factors, starting with credit score:
Freddie Mac started to release forbearance data on its STACR CRT program in the July reporting cycle. Previously, the GSEs disclosed only pool-level delinquency and forbearance information . As the loans in the STACR program have a UPB representing over 50% of total deliveries to FHL balances with no obvious state level bias, they would serve as a representative sample in calculating GSE state level delinquency . Similarly, the new STACR data is helpful in assessing the impact of the COVID-19 crisis on forbearance at loan level.
From the July STACR data release, we can see a clear correlation between total delinquency rates and forbearance rates at state level. States with higher delinquency rates, such as New York, New Jersey, Hawaii, and Nevada, also have higher forbearance rates.
*The Chart can be duplicated using the above two queries
This is not a surprise as we noted in an earlier post that Freddie Mac servicers are not required to report loans in forbearance if loans are current . In fact, for STACR data, only 0.95% of current loans are in a forbearance plan, but the forbearance rate for loans in 30 day delinquency, 60 day delinquency, and 90+day delinquency are 30.64%, 92.98%, and 92.12 % respectively.