The Impact of Weather Events on U.S. Unemployment Statistics

Hurricane Katrina is well known as one of the most impactful weather events in U.S. history.  It is also well-known that Katrina had a large impact on unemployment statistics collected by the U.S. Bureau of Labor Statistics (BLS) shortly after the event.  The impact of Katrina on the weekly initial claims for unemployment was larger than any other weather or economic event in history.  For the 5 weeks following Katrina landfall, an excess of 379 000 initial claims for unemployment were made in the United States above what was expected from routine economic activity.  It is less well-known that a wide variety of weather events can be detected from unemployment statistics.  Every year there are typically one or two weather events which cause otherwise unaccounted for deviations in the initial claims data.  Most weather events cause an increase in unemployment claims due to economic disruption, but some cause a decrease in claims, and some move unemployment claims (which would have occurred anyway due to seasonal factors) to earlier or later parts of the calendar.   There are strong regional variations in how weather events impact unemployment, with a strong signal from early or later winter weather in northern states.  Climatologically warmer states have fewer weather events impacting unemployment data, though rare tropical disasters have the greater impact when they occur.  The BLS develops seasonal corrections to raw unemployment data in order to remove the strong seasonal signal (which by itself is of interest to climatologists).  BLS does this correction for national data, but not for state data.  In order to better illustrate the regional impact of weather on unemployment, this paper develops a method for state-based seasonal correction that differs from the complex process used by the BLS to correct national data.  It is the purpose of this paper to draw attention to the impact of major weather events on unemployment statistics, and to describe the developed technique for analyzing BLS data on a state-by-state basis.