Model Stability and the Subprime Mortgage Crisis

摘要: We study the potential model instability problem with respect to mortgage default risk and examine what extent it helps explain shock during recent crisis. find that econometric models based on historical data can be unstable over time. Due temporal shifts in parameters, prediction of 2006 vintage subprime loans hazard Logit estimated 2003 loan generate 40 percent fewer defaults than actual number, assuming perfect forecast house price change. also combined impact parameter bad HPI growth enlarges under-prediction rate but marginal is larger forecast. Our findings have important implications regarding limitations risk, improvements, economic capital, regulatory reform.