Decomposing Agency MBS Yields into Risk Premia
Summary
The document considers how to decompose the forward-looking yield of an index of agency fixed-rate residential mortgage-backed securities. The proposed framework adds a duration-matched Treasury yield, the option-adjusted spread, and an adjustment intended to capture the difference between implied and realized interest-rate volatility. The questioner distinguishes this prospective yield analysis from historical return attribution, which they report can be tracked fairly closely using Treasury and at-the-money swap returns matched for duration.
The response identifies a broader set of potential drivers: prepayment, realized interest-rate volatility, basis, implied volatility, financing or leverage, liquidity, and credit risk. It notes that liquidity and credit are often treated as small for agency pass-throughs, while financing risk may still matter during funding stress. The response also points to published mortgage-market research, but gives no estimation procedure or evidence for quantifying each component. The suggested decomposition is therefore a starting framework, with risk-factor attribution and its practical estimation left unresolved.
Key ideas
- A proposed agency MBS yield decomposition combines a duration-matched Treasury yield, OAS, and a volatility adjustment.
- Mortgage prepayment risk is a distinct source of uncertainty in MBS returns.
- Potential MBS risk factors also include basis, financing, liquidity, and credit exposure.
- Agency status may reduce perceived credit and liquidity premia, while financing risk can still be material.
- The response notes that practical methods for estimating excess returns by risk factor are limited.
Tags
Full text
# Agency Fixed Rate RMBS Yield Decomposition
# Agency Fixed Rate RMBS Yield Decomposition
I'm trying to find the best way to decompose the yield on an index of fixed residential MBS securities and want to open up the question to the community. The goal is to look at this from a quantitative perspective. I have interest rate models, but no models strictly specific to mortgages. I also have all the index specific metrics I need, modified duration, option adjusted duration, average term to maturity, OAS (not my model but trusted), etc.
For return attribution, I've been able to fairly closely track total return with a duration matched Treasury total return and the total return on selling at the money swaps.
For a forward looking yield decomposition, I'm thinking of using an OAD matched Treasury yield, OAS, and a long term implied volatility to realized volatility adjustment. $$ Y_{MBS} = Y_{durationMatchedUST} + OAS + Vol Adjustment$$ where $$ VolAdjustment = \sum_{}^{1 year}Premiums_{vol= impliedVol} - \sum_{}^{1 year}Premiums_{vol= impliedVol*}$$
$impliedVol*$ is the implied volatility adjusted down by the long term difference in implied and realized volatility. The premise for $VolAdjustment$ is to capture additional expected yield from any volatility premium baked into the price of the MBS.
How else can I approach a decomposition? For the sake of this decomposition, we are assuming no ratings migration or default loss, so OAS should be premium for not knowing when payment will happen. We further adjust the premium for realized volatility. Any other adjustments needed?
## Answer by Sharad (score 1)
https://quant.stackexchange.com/a/58146
The traditional risk factor decomposition of a general MBS includes the following risk factors: Prepayment Risk, Interest-rate Risk (Realized Volatility), Basis Risk, Volatility Risk (Implied volatility), Financing/Leverage Risk, Liquidity Risk and Credit Risk. If the focus is on Agency MBS Pass-throughs then one usually assumes that Liquidity and Credit Risk are minimal. Financing/Leverage Risk is also typically neglected although this is much more questionable given the dramatic underperformance of MBS in situations where financing is hard to come by.
So much for the theory, in practice there has been too little published work on how one goes about estimating the excess return associated with these risk factors. One shining exception is: "Risk and Return in the Mortgage Market" by Amitabh Arora et al. In particular, take a look at the analysis in Section IV.Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.