Sensitivity of the United States Economy to Weather Variability

Jeffrey Lazo

Abstract

It has been widely claimed in the meteorology literature that one third of the US economy is weather sensitive. There appears to be little, if any, justification or empirical basis for this claim although the issue is of critical importance to the weather enterprise. This paper reports on work defining and examining the sensitivity and vulnerability of state-level economic sector productivity [gross state product (GSP)] to weather impacts.

Twenty-four years of state level super-sector economic data and historical weather observations are used to form a panel combining weather information with economic data. A translog production function is used to estimate sectoral sensitivity and vulnerability to weather impacts such as temperature (heating degree and cooling degree days) and precipitation (total and variation). Econometric modeling is discussed including issues of heteroskedasticity and auto-correlation. Holding physical inputs constant (capital, labor, energy), the resulting eleven sector models are used to fit state-sector GSP using 70 years of weather data to generate distributions of GSP variability resulting from weather variability.

The eleven super-sectors are ranked based on their degree of sensitivity to weather, states more sensitive to weather impacts are identified, and the aggregate dollar amount of variation in U.S. economic activity attributable to weather variability is calculated. We discuss potential limitations to this approach as well as plans for further model development. The results of this analysis are also discussed in the context of assessing the value of current and improved environmental (i.e., weather) forecast information. We also discuss extensions to the model which may allow us to estimate the value of historical improvements in weather forecasting and changes in weather sensitivity over time.