Using Monte Carlo simulations to establish a new house price stress test

The focus of this paper is on the house price stress test (termed ALMO) that was designed to assess the fiscal strength of Fannie Mae and Freddie Mac and, if necessary, to trigger remedial action in order to avert a crisis. We assess whether the ALMO stress test was an adequate representation of an extremely weak housing market, given the best available information leading up to the Great Recession. A Monte Carlo simulation model is developed to estimate the severity of low probability events (i.e., severe house price declines). We illustrate the complexity and subjective nature of the process used to generate a plausible house price stress test scenarios. A major finding is that the ALMO stress test scenario severely understated (possibly by 50% or more) what an updated statistical process would have suggested. Part of this stems from idiosyncrasies related to the construction and implementation of ALMO, while other factors include a fundamental shift in the relationship between housing price appreciation and key explanatory variables - especially over the past 10-15 years, which shows a heightened role of momentum in explaining changes in housing prices. We offer several suggestions for a new stress test that include: continual updates and testing; variation across markets; and, like the recent FRB stress test, the scenario should be based on real (rather than nominal) price patterns.