Shifting rate dynamics add to MBS negative convexity

Brian Landy, CFA | February 26, 2021

This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors. This material does not constitute research.

MBS prepayment speeds look set to fall sharply after the recent rise in interest rates, but models may not be ready for this new environment. Lenders have little room to keep origination volumes up by tightening the spread between primary and secondary mortgage rates. Primary mortgage rates at this point should closely track changes in MBS yields and become more volatile, a shift from last year’s pattern of widening as MBS yields fall and tightening as MBS yields rise. This should increase negative convexity at any given MBS price point and lower effective OAS.

At current interest rates, roughly half of mortgages have moved out-of-the-money to refinance (Exhibit 1). A mortgage is considered in-the-money if its note rate is at least 75 bp higher than the prevailing mortgage rate. The most recent mortgage rate, using Bankrate.com’s daily survey, is 3.15%. Higher rates should be fully reflected in April prepayment speeds, which are reported in May. Speeds that month should fall to roughly 24 CPR, down from roughly 30 CPR in January.

Exhibit 1: Half of mortgages are less than 75 bp in-the-money

Speeds marked in the graph are seasonally adjusted. The actual speed will be faster in the summer months and slower in the winter months.
Source: Fannie Mae, Freddie Mac, eMBS, Amherst Pierpont Securities

The primary-secondary spread is unlikely to compress much as rates move higher, since it appears have tightened close to its pre-pandemic levels (Exhibit 2). The spread is calculated by comparing the average note rate of 30-year mortgages originated each month to a lagged average of the MBS current coupon. The average spread from 2014 through 2019 was roughly 127 bp. However, this baseline needs to be increased by 12.5 bp, assuming a 4x IO multiple, to account for the 50 bp refinancing fee the GSEs began to charge in December. The spread in January originations was 150 bp, only 11 bp higher than the adjusted pre-pandemic average. And the spread in February originations has fallen slightly below the average, although so far this covers only about 30,000 loans.

Exhibit 2: Primary-secondary spreads are near normal levels

The pre-COVID average is measured from 2014 through 2019. The average is adjusted 12.5 higher to account for the 50 bp refinance fee the GSEs began charging in December. February origination data is preliminary and only includes roughly
Source: Fannie Mae, Freddie Mac, eMBS, Amherst Pierpont Securities

The primary-secondary spread is also unlikely to widen if rates move lower. Typically, the spread widens when rates fall because of capacity constraints. But the industry will suddenly have significant excess capacity as refinancing volumes fall, and mortgage rates should closely track any drop in MBS yields.

MBS at a given price point–$103 or $105, for instance—should show more negative convexity in today’s market compared last year’s, all else equal, since the primary-secondary spread is unlikely to widen or tighten much in future rate moves. Normally the spread movements act to reduce the volatility of mortgage rates, which adds convexity to MBS and boosts OASs. For intuition, consider the extreme case in which the spread is perfectly negatively correlated with MBS yields, which means the mortgage rate would never change. Mortgages would behave more like Treasury debt, with very stable cash flows and prices affected only by shifts in the discounting rate. MBS would show positive convexity, and option cost would drop to zero.

Some models may not properly capture current primary-secondary spread movements. Many rely on surveys like Bankrate.com’s or Freddie Mac’s to determine the current mortgage rate, but these surveys have tended to overstate the spread during the pandemic. These models may tighten the spread in response to higher interest rates and therefore forecast a mortgage rate that is too low. These models will underestimate how much speeds will slow in April.

Models may also improperly assume that the primary-secondary spread will widen if rates were to retrace their steps lower, not recognizing that mortgages face higher volatility due to excess capacity. Those models will produce OASs that are too high since they will not capture the more negative convexity in MBS in an excess capacity origination environment.