Space-Time Modeling of Natural Disaster Impacts

Abstract

County-level estimates of employment, unemployment, and the unemployment rate are not produced directly from a sample survey; rather, they are developed through models that use information on the labor force from a number of statistical programs such as the CPS (Current Population Survey), CES (Current Employment Statistics), and State Unemployment Insurance (UI) programs. These sources of information input as well as the models themselves are affected in various ways by disaster events such as Hurricane Ike (Brown et al. 2006). We use a pace-time dynamic panel model to quantify the impact of Hurricane Ike on Texas county-level employment. Natural disasters such as a hurricane produce impacts that fall on the immediate region plus neighboring regions at the time of the disaster event, and these effects continue to impact the own- and other-regions through time as the initial impact emanates through space over time. Our space-time model can be used to quantify how the initial impact of a disaster event in one region (or regions) influences all regions over time by exploiting historical patterns of space, time and space-time dependence between the number of establishments and employment.