Extreme weather movements have been known to have an impact on asset prices. Weather data is obviously important for commodities investing, where it feeds directly into models forecasting the demand of natural gas and the supply of agricultural crops. What about other areas of application? Everyone lately has talked about the impact of bad weather on first quarter GDP numbers in the United States and the shocking revision we had on June 25 (-3%). In fact, it was the worst of the last 17 years (including Q4’08). Does bad weather systematically lead to lower Q1 GDP revisions?
In our framework of analysis, we have collected historical weather data from 800 weather stations around the United States. We have taken the daily temperature readings from each weather station and aggregated it up to the state level, and each state’s temperature is then aggregated to the national level using weights based on their respective economic contribution to GDP.
US Temperature Data
GDP Weighted by State
As shown in the above chart, temperature is highly seasonal. We can adjust this by normalizing over the previous 5 and 10 years and overlay of the first quarter of every year. In the chart below, we can see that this year, temperatures were close to 2 standard deviations below norm during Q1 compared to the last 10 years and around 1.5 standard deviations below norm over the last five years. How does this data compare to US GDP revisions historically?
Normalized US Temperature and GDP Revisions
Not every unusually low temperature generates a downside revision of GDP on the first revision. In the first quarter this year, both the weather and the GDP revision were poor. However, the correlation between normalized temperature data in the first quarter of every year and the first revision of Q1 GDP every year is 25%. Historically this is not significant, and so we consider our recent revision surprise to be a one-off.