Using OAS to predict excess returns

Brian Landy, CFA | January 19, 2024

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.

Investors routinely use option-adjusted spreads as an indicator of relative value in MBS. But it is unclear that consistently investing in bonds with wide OASs raises excess returns. A model that controls for other drivers of total returns helps draw out the relationship and suggests that 10 bp of OAS could improve excess returns by 2.2 bp each month. But the model leaves a fair amount of excess return unexplained. An investor could be reasonably skeptical of the relationship between OAS and excess returns, despite the statistical significance. Investors should use hedged returns—building proceeds- and duration-neutral portfolios as a supplement to, or alternative to, OAS analysis when making investment decisions.

At first glance the relationship between excess return and OAS is not apparent (Exhibit 1). Each data point compares the 1-month excess return on key coupons in Bloomberg’s MBS index for conventional 30-year MBS to the beginning-of-month OAS. This is done for coupons from 2.0% through 5.5%. Bloomberg’s excess returns are calculated by matching six key-rate durations at the start of the month using US Treasury bonds and a cash component. The hedge is held constant throughout the month. The dataset covers returns from January 2016 through December 2023. The regression line explains less than 1% of the excess returns and the slope suggests that wider OAS leads to lower returns, which is not intuitive.

Exhibit 1. The relationship between OAS and excess returns seems weak.

Source: Bloomberg, Santander US Capital Markets

Adding terms that explain the monthly changes in excess return improves the fit considerably and helps unearth a clearer relationship between OAS and excess returns. Excess returns can change when interest rates move—the hedge is not rebalanced throughout the month so there is exposure to the negative convexity of MBS, and the model durations used to set the hedge may not match the market’s empirical durations. Changes in volatility also influence returns. And mortgage excess returns vary considerably when mortgage spreads widen or tighten.

The model that includes these factors unearths a stronger relationship between OAS and excess returns (Exhibit 2). This model suggests that—after controlling for monthly changes in interest rates, volatility, and spreads—a 10 bp higher OAS should lead to a 0.99 bp increase in excess returns. But the term is only significant at the 5% level. The data was fit with a generalized additive model. A smoothing spline was used to fit the non-linear relationship between interest rates and excess returns. This means the coefficient depends on how rates move, but for this analysis it is not important to explore that further. The other terms were included as simple linear terms; smoothing splines were explored for each of these terms but did not appear to improve the fit (and may have overfit the data). The coefficients for vol change and spread change make sense—negative coefficients mean that higher volatility and wider spreads lower returns.

Exhibit 2: Controlling for rates, vol, spreads, OAS is a weak predictor of excess

Note: ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1. Model estimated using the mgcv package in R. The change in 10-year Treasury was fit using a thin-plate regression spline; the other terms are linear.
Excess return and OAS are both measured in bp, rates are in %, and volatility is Black vol (%).
Source: Bloomberg, Santander US Capital Markets

It is also possible that the OAS at the start of the month helps predict what spreads will do in the upcoming month—wide OASs may tend to tighten, and tight OASs tend to widen. Removing the term that controls for the change in OAS improves the performance of investing based on OAS (Exhibit 3). The model’s adjusted R² unsurprisingly falls, but the importance of the start-of-month OAS increases. The term is now significant to 0.1% and suggests that 10 bp wider OAS should result in 2.2 bp higher excess return each month.

Exhibit 3: Controlling for rates and vol, OAS is a stronger predictor of excess

Note: ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1. Model estimated using the mgcv package in R. The change in 10-year Treasury was fit using a thin-plate regression spline; the other terms are linear.
Excess return and OAS are both measured in bp, rates are in %, and volatility is Black vol (%).
Source: Bloomberg, Santander US Capital Markets

The models leave a significant amount of variance unexplained, and some may consider the relationship between OAS and excess returns to be relatively weak even in the last model. Prudent investors should not invest solely based on OAS—buying wide OAS bonds and selling tight OAS bonds. Option-adjusted spreads likely have more utility as a signal of potential richness and cheapness.

A better approach for assessing relative value of an MBS may be to construct portfolios of Treasuries (or swaps) that have the same market value and duration as an MBS and compare the projected excess returns. Returns can be examined under different interest rate scenarios and models can be stressed. The method can be extended to hedge convexity with short-dated swaptions, or even volatility risk with long-dated swaptions. This method helps explain whether an MBS is adequately compensating an investor for its negative convexity.