Non-bank institutions continue, and increase, their dominance in mortgages, most vividly in the Ginnie Mae space.
Private label securities (PLS) account for an almost negligible amount of the U.S residential mortgage market, with an “extraordinary dominance of agency-backed MBS.”
And there are some significant differences between the GSEs in terms of their projections for originations, housing starts and home sales — Freddie Mac’s more bullish than Fannie Mae — and in their effective guarantee fees.
Those are among the points that particularly stand out to Mortgage Media’s Dave Stevens in the Housing Finance Policy Center’s reference guide for mortgage and housing market data, “At A Glance” (colloquially referred to as the “Chartbook”). Stevens recently took part in a conversation with the Urban Institute’s HFPC researchers on many of the main points of this month’s chartbook; following are some of his main takeaways:
MBS over PLS: Among the big stories, Stevens said, is “the extraordinary dominance of agency-backed MBS (mortgage-backed securities)” — to the near exclusion of PLS (private label securities) — in the overall U.S. residential mortgage market. The combined agency MBS market continues on a steady upward climb, now at $6.65 trillion. That’s far ahead of other categories: unsecuritized first loans at $3.25 trillion, second loans at $0.54 trillion — and PLS lagging at $0.46 trillion.
“You’ll see that Ginnie Mae, the growth there is extraordinary,” he said, pointing to performance in not only FHA but VA and USDA, “where they become better execution over Freddie/Fannie.” Ginnie is at $2 trillion in issuance now, topping Freddie Mac’s $1.8 trillion. (Fannie Mae is at $2.9 trillion — it’s slowly rising, but hasn’t seen anywhere near the dramatic rise that Ginnie has since 2008.)
As opposed to the private label. “What’s really different about the market despite what we’re seeing in changes in changes on DTI and FICO, primarily the institutional participation and respective programs is today — there’s literally no PLS.” Stevens points to the first lien origination volume chart — the “skinny little blue sliver” on the 2017 and 2018 columns is where PLS is executing, he noted. (As opposed to the years before the housing crisis, in which PLS accounted for a much bigger share — a big wide blue sliver, if you will.) “This is really a market that’s driven and dominated totally by agency, and some portfolio and private label has not come back.” Stevens said. “That’s a good thing relative to FICO and DTI, in some levels. It’s not good to the fact that there’s not a good functioning PLS market.”
30-year dominance: Fixed-rate, 30-year mortgages are what’s dominating the mortgage product space. Stevens noted that the American Enterprise Institute for Public Policy put out a paper on this topic recently, suggesting that the 30-years, combined with DTI and FiCOs, is adding stress to the market — and he cautions attention to “risk layering that isn’t in place, versus the previous period, when there were so many other products in the marketplace, particularly ARM.”
Refinancing: The HARP program helps underwater homeowners refinance their homes and lower their monthly payment (or pay off faster and build more equity). And, Stevens notes, HARP and quantitative easing drove rate and term refinancing — “and really helped get a lot of folks out on negative equity loans and high rate loans into a more stable loan product, and it actually improved their credit position.”
The cash-out refinancing picture is considerable, driven by equity created in the housing market. “No one is going to refi out of a 3.5 rate into a 4.5 — but they will go for a cash out,” Stevens notes (especially with higher interest rates, HFPC’s Laurie Goodman later noted). It’s considerable, but barely comparable to the way things were in the bubble period leading up to the Great Recession: Cash-out refinancing volume peaked above $80 billion around 2006. These days, it’s less than $20 billion.
Nonbank originators: “It is really stark what’s happening in the non-bank space,” Stevens said. The nonbank origination share of loan issuances has been steadily rising since 2013, especially the nonbank share in Ginnie Mae, considerably higher than non-banks’ involvement through the GSEs. The non-bank share in Ginnie reached a new record in December 2018 with 83 percent. (Meanwhile, the non-bank share in the GSEs averages in the area of 55 to 65 percent.) So what’s driving the median percentage of non-bank involvement in issuances is entirely the activity in the Ginnie space.
The fact is, banks are more skittish, having dealt with the fallout from the housing crisis and the Great Recession, whereas the nonbanks weren’t as active in the sphere until post-recession and didn’t necessarily feel the pain. As Stevens puts it: “The reality, gang, is the reason why banks are not as pronounced in this program, and why they’re putting greater overlays on the Ginnie programs — which I think we all know — is that they are still living with the hangover effects of False Claims Act, certifications that haven’t been fixed, taxonomy which hasn’t been implemented, and other servicing fixes which are really needed at FHA. The nonbanks are far more comfortable — I think they don’t have the legacy effect that the banks had. Because they weren’t as active in the space prior to (recession), so therefore they haven’t felt the pain post-recession, because they weren’t there as aggressively. And now, we’re seeing them really leading the space.”
Non-bank median FICO scores versus bank median FICOs have a 36 point spread (bank at 750, non-bank at 714) — the difference due to most of the production being concentrated in the Ginnie Mae books. “It’s something that I know policymakers are watching very closely.”
Agency non-bank credit tracks about the same for the GSEs — same for Ginnie Mae, where it’s limited by policy. The DTI is higher for non-banks than banks in the GSEs and Ginnie — non-banks are more accommodating in the LTV and FICO dimensions, Stevens noted. You don’t see the same spread so much in LTV: “Banks and non-banks are both doing maximum loan to value transactions). But the banks are putting more overlays when it comes to FICO so you see the difference is showing up more elsewhere than you do on the loan to value space.”
FICO/LTV correlation: Lower average FICO scores tend to be correlated with high average LTVs (loan to value ratio), a case often seen in metropolitan statistical areas with high housing prices. Take the Riverside, San Bernardino area in California, for instance: Wages are lower, earning stress is higher, savings aren’t as great — hence FICO ratings are down and LTVs trend upward. It’s a different story in, say, San Francisco, south SF — a huge employment market, with higher wages. So the mean origination FICO for borrowers is around 772 in San Francisco, 715 in Riverside.
A delta in projections, G-fees, at GSEs: Both GSEs as well as the Mortgage Bankers Association all forecast origination volume in 2018 at $1.63-$1.65 trillion, lower than 2017 and much lower than 2016, largely due to the decline in the refi share. The 2019 volumes are expected to be close to 2018. However, there’s a considerable delta in the originations forecast between the two GSEs: A $75 billion difference. And you see the same difference show up in housing start forecasts: Fannie projects less than 1,300 thousand, Freddie predicts 1,400. And in home sales: Fannie forecasts 5,992, Freddie 6,300.
“That’s where Freddie has a much more bullish view about housing starts than Fannie … with MBA closer to the Fannie estimate than the Freddie estimate,” Stevens said. “And so that’s a number we’re going to have to look closer at, to see if Freddie’s being just overly bullish about housing starts, or who’s got this wrong — because that’s a big difference in number.”
There’s also a significant difference in effective guarantee fees between Fannie and Freddie — about 10 basis points, a huge difference (Fannie at 61.2, Freddie at 51.0.) Stevens isn’t sure of the exact reasons — he thought of MAP (market-adjusted pricing), but the gap seems too big to pin on that — and anyway, “they’re not supposed to be doing MAP.”
“So it just raises the question in my mind, whether one of the GSEs is pursuing more low down payment, perhaps lower FICO, in the LLPA (loan-level price adjustment) grids,” he noted. “…The only way that could happen would be negotiated LLPAs for select institutions. I’m not saying that’s the case, I’m just wondering in the back of my mind, why do you have a 10 bps spread?” He’s never even seen such a gulf before in his professional life, he said. “I think all of us should be thinking about where that’s coming from.”
Home price index: “The volatility risk of either excessive tightening or excess in credit availability can clearly be seen in market like L.A.,” Stevens said — with the home price index (HPI) up 180 perrcent from 2000 to the peak, down 38.2 percent from the peak to the trough, and then up 79 percent from the trough to now. Prices are currently some 10.6 percent higher than the pre-crisis peak levels.
“Ultimately at the end of the day it was HPI that buried the market,” Stevens said. “It may have been caused by other things, but that’s the ultimate severity that put the markets underwater, took down institutions, and created an impossible way to bring back a return on a defaulted homebuyer.”
First-timers: 82 percent of FHA purchased transactions are first-time homebuyers, compared to less than half — 48 percent — in the GSE space. Why? Compare the LTV and DTI figures in GSE vs. FHA space: Both are significantly higher for FHA: A 95.5 LTV in FHA vs. 87.5 for the GSEs; a DTI of 43.7 for FHA vs. 36.6 in the GSEs.
“With an FHA you can get a 97 percent loan to value on a purchase transaction, so you skew the LTVs higher — and the FHA has a more liberal DTI policy, which some are scrutinizing today in the policy sector. But that’s clearly needed when you have a shortage of supply, rising home prices and a difficulty in reaching those prices.”