Lawler: Update on Mortgage/Treasury Spreads

CR Note: A year ago, housing economist Tom Lawler explained why the spread between mortgage rates and the 10-year Treasury yield was wider than “normal”. See: Lawler: Mortgage/Treasury Spreads, Part I and Lawler: Mortgage/Treasury Spreads Part II: “Decomposing” the Widening This Year.

From housing economist Tom Lawler:

Last May I wrote a two-part piece on why the spread between the 30-year fixed-rate mortgage rate and the 10-year Treasury rate had widened significantly from the end of 2021 to early May of last year, and explained why changes in the yield curve, changes in market-implied interest rate volatility, and the end of the Federal Reserve’s misplaced MBS purchase program all logically and rationally resulted in a widening of the mortgage/10-year Treasury spread.  I also said that it was unlikely that mortgage/10-year Treasury spreads were unlikely to revert to what some argue is the “normal” level (there is no such thing) any time soon.  Since then, mortgage/10-year Treasury spreads have on average been wider than there were last May, for the most part for logical and rational reasons based on what has happened to the yield curve and implied interest rate volatility.

Note: In this chart I am using the Optimal Blue 30-Year Fixed-Rate Conforming Index, based on actual daily rate-lock activity. This rate is available daily back to 2017 on the Federal Reserve Bank of St. Louis’ FRED database. One word of caution: the index is based on rate-lock activity throughout the day while the daily Treasury yields are rates as of last afternoon. Significant intraday rate volatility can cause some “spurious” fluctuations in the spreads.

As I noted back then, there is no reason why mortgage rates should be anchored solely to the 10-year Treasury yield.  The timing of expected cash flows (regular payments and pre-payments) from a pool of mortgages don’t even remotely resemble the timing of known cash flows from a 10-year Treasury security, and mortgage valuation is based on discounted cash flows across the whole yield curve.  As such, one would expect that the shape of the yield curve would impact the spread between mortgage rates and the 10-year Treasury, with a steeper curve producing a narrower mortgage/10-year spread and an inverted curve producing a wider mortgage/10-year spread.

In addition, while the timing on cash flows on 10-year Treasuries are known with certainty, that is obviously not the case for mortgages, which allow the borrower to prepay his/her loan in whole or in part at any time without penalty.  Expected mortgage cash flows are very much dependent on the path of interest rates:  rising interest rates generally result in somewhat slower prepayments (a negative to the investor because interest rates are up), while lower interest rates result in substantially faster prepayments, a negative to investors because interest rates are down.

Since mortgage cash flows are very much interest-rate path dependent, the “expected” value of mortgage cash flows across all possible interest-rate paths is heavily dependent on the probability distribution of future interest rates.  If market expectations of future interest rate volatility were low, one would expect relatively low mortgage/Treasury spreads, while if market expectation of future interest rate volatility were high, one would expect relatively wide mortgage/Treasury spreads.

The shape of the yield curve also impacts that probability distribution of future interest rates.  A very steep yield curve implies a market expectation of increasing future interest rates, and future rate simulations are typically “centered” around this rising “base case” scenario (with possible adjustments for estimated term premia).  An inverted yield curve, in contrast, implies a market expectation of declining future interest rates, with future rate simulations centered around this falling “base case” rate scenario.  As such, a simulation of future rates with an inverted curve “produces” more rate scenarios where borrowers exercise their prepayment option, implying that mortgage/10-year spreads will be narrower with a steep curve and wider with an inverted curve.

To sum up, one would expect mortgage/10-year spreads to be (1) wider when interest rate volatility is high; and (2) wider when the Treasury yield curve is inverted.  And when interest rate volatility is high AND the yield curve is very inverted, one would expect the mortgage/10-year spread to be VERY wide.

With this in mind, here are a few charts to consider.  First, here is the spread between the 10-year Treasury and the 1-year Treasury the beginning of 2017 through June 2, 2023.

Next, here is a chart of the ICEBofAML MOVE Index, which is a measure of market-implied rate volatility based one-month options on Treasury futures across multiple maturities.  Historical daily data are available on Yahoo Finance (ticker MOVE) back to November 12, 2002.

As these two charts show, the Treasury yield curve is extremely inverted, while market-implied interest rate volatility is historically very high.  This combination would logically and rationally be expected to be associated with historically very wide mortgage/10-year Treasury spreads.  (See 1st graph above!)

Here is a table showing various interest rates as well as the MOVE index for 12/30/2021 compared to 5/31/2023, as well as a table showing various spreads for those two dates.

The first table shows that from 12/30/2021 to 5/31/2023 (1) the 30-year FRM increased by almost the same amount as the “current coupon” MBS yield; (2) that the increase in Treasury yields varied massively depending on the maturity of the Treasury security; and (3) that the MOVE interest rate volatility index increased sharply.

The second table shows that from 12/30/2021 to 5/31/2023 (1) while the mortgage rate to 10 year Treasury spread widened by 125 bp, the MBS option-adjusted spread widened by only 56 bp, and the MBS “Z spread” – or the spread over the entire Treasury curve assuming zero interest-rate volatility – widened by only 109 bp; and (2) the spread between the 10 year Treasury rate and the one year Treasury rate went from +114 bp to -154 bp, a swing of -268 bp!  Obviously, the widening in spread between MBS and 10-year Treasury rates over this period vastly overstated the degree to which MBS had become “cheap.”

Note that the option-adjusted spread on the current coupon 30-year MBS was negative at the end of 2021, and in fact was negative during almost all of 2021.  Stated another way, after adjusting for the option embedded in MBS, MBS yields were lower than Treasury yields during almost all of 2021!  This highly unusual spread behavior almost certainly reflected the massive (and probably ill-advised) purchases of MBS by the Federal Reserve over this period, which severely distorted/inflated MBS prices.  While the historical “normal” MBS OAS – meaning periods without Fed intervention – depends on the time period one looks at, a reasonable “normal” value would be about 30 bp.

If a “normal” MBS OAS were, say, 30 bp, then on May 31, 2023, MBS yields (and mortgage rates) relative to the entire Treasury curve after adjusting for market-implied interest rate volatility were only about 20 bp above normal, even though the mortgage/10-year spread was 308 bp!

Now let’s look at some “quick and dirty” regression results.  Below is a simple regression where the dependent variable in the spread between the 30-year FRM (Optimal Blue) and the 10-year Treasury, and the independent variables are the spread between the 10-year and 1-year Treasury yield and the MOVE index.  First, here are regression results using daily data from 1/3/2017 to 5/31/2023.

Note the negative and significant sign on the coefficient of the 10/1 spread, and the positive and significant sign on the MOVE index.

The overall “fit” of the regression is not that great, however, and in looking at the actual/fitted residuals (not shown) it is clear that actual spreads far exceeded fitted spreads during the worst of the pandemic while during the period when mortgage OAS were negative, when Fed intervention was clearly impacting the MBS markets, actual spreads were well below fitted spreads.

While one possibly could attempt to deal with the pandemic and Fed issues with dummy variables or other measures, an alternative simply would be to exclude certain periods So below is a regression with the same dependent and independent variables but excluding the period from 2/3/2020 to 3/31/2022.

Again, the coefficient on the 10/1 spread is negative and statistically significant, while the coefficient on the Move Index is positive and statistically significant.  Moreover, the overall “fit” for a spread equation is very good.

Note that these results are not really a “model;” linear regressions assume linear relationships that may not be accurate; the 10/1 spread is a crude way to incorporate “the curve;” and the Move index is a measure of one-month Treasury volatility, while a “good” model would incorporate the whole term structure of volatility.  However, all of these variables are publicly available, and the equation actually produces a pretty tight fit for the mortgage rate.

To folks who like to look at simple “R squared” measures comparing mortgage rates to the 10-year Treasury, note that the above equations are for the mortgage/10-year spread, and not the mortgage rate itself.  Here is an Excel generated chart showing fitted vs. actual LEVELS of the mortgage rate derived from the spread equation.

To sum up, the spread between the 30-year fixed-rate mortgage and the 10-year Treasury is at historically very wide levels because (1) the yield curve is incredibly inverted by historical standards; and (2) interest rate volatility has been historically high.  Models that take into account the timing of mortgage cash flows and the option embedded in mortgages suggest that the mortgage/10 year spread is at most only slightly wider than it “should” be.  A material narrowing of the mortgage/10-year spread back to what some people refer to as a “historical norm” will only happen if the yield curve returns to “normal” and/or interest rate volatility returns to “normal.”

CR Note: This was from housing economist Tom Lawler. See also: Could 6% to 7% 30-Year Mortgage Rates be the "New Normal"?