Forecasting Applications to Disability Insurance Programs: Evidence for the Quebec Pension Plan Disability Program

Abstract

I propose a new method to forecast applications to disability insurance programs. My approach is based on a state space model, where the coefficient estimates on the explanatory variables vary over time, that I estimate using the Kalman Filter. Using data from the Quebec Pension Plan Disability program, I show that my model and approach is the preferred forecasting model relative to two commonly used alternative approaches. I also use my model to create out-of-sample forecasts, where I study the effects of a few hypothetical changes in the policy and economic environment on forecasts of applicants for disability benefits.