Alternative margin models for mortgage-backed securities

David Li, Roy M. Cheruvelil and Viktoria Baklanova

Abstract

Most US mortgages are traded in the form of mortgage-backed securities (MBSs) guaranteed by the US government-sponsored enterprises Fannie Mae and Freddie Mac and the government agency Ginnie Mae. A significant portion of agency MBSs trading occurs in the to-be-announced forward market. Yet, the existing margin models for TBA/MBSs mostly rely on mortgage model suites such as interest rate, prepayment and potentially other macroeconomic models, which makes the modeling process intrinsically complicated from both a model risk perspective and an operational risk perspective. In addition to these complexities, dynamics in the housing market, changes to mortgage regulatory regimes and governmental interventions always make mortgage modeling a challenge (as evidenced historically). In this paper, we conduct a study of margin models for to-be-announced MBSs using common margin frameworks for market risk such as the generalized autoregressive conditional heteroscedasticity (GARCH) t-copula and filtered historical simulation approaches. These are commonly used for other asset classes such as credit default swaps and equity-based markets but have not been widely used in the MBSs market. Such econometric models, which rely solely on market volatility and price return behavior, could potentially be used as a supplemental model framework for to-be-announced MBS margin and stress testing purposes.