The Credit Spread Puzzle in the Merton Model - Myth or Reality?

07 October 2015
12:30  - 13:30

Peter Feldhütter, London Business School



We test the Merton model of credit risk using data on individual firms for the period 1987-2012. We find that the model matches the average level of investment grade spreads and furthermore captures the time series variation of the BBB-AAA spread well with a correlation between 84% and 92%. A crucial ingredient to the success of the model is that we use default rates for a long period of 92 years to calibrate the model. In simulations we show that such a long history of ex post default rates is essential to obtain estimates of ex ante default probabilities that have a reasonable level of precision; using default rates from shorter periods as done in most existing studies will often lead to the conclusion that spreads in the Merton model are too low even if the Merton model is the true model. Finally, we show that using data on individual firms - rather than a “representative firm” - is important for matching the slope of the term structure of credit spreads.

image attachment Back