The value of weather forecasts: Evidence from labor responses to accurate versus inaccurate temperature forecasts in China

Abstract

Short-term weather forecasts, a common and popular public good in the modern world, affect labor decisions regarding time allocations. This study uses a novel dataset of city-level day-ahead weather forecasts in China, collected through video transcriptions of the country’s popular TV program spanning over 2000 days since 2010. I estimate the number of hours laborers worked in a day as flexible functions of the daily maximum temperature forecast under different historical levels of forecast accuracy (represented by half-year rolling daily maximum temperature forecast root-mean-squared-error, $RMSE$). The results suggest large-magnitude (up to 4.5 and 1.2 h per day) labor decreases under uncomfortable temperature forecasts (extreme heat above $30{}^{\circ }C$ and medium cold $15{}^{\circ }C\text{–}25{}^{\circ }C$), but only when forecasts are accurate ($RMSE\approx 1{}^{\circ }C$). The economic value of accurate weather forecasts is assessed by modeling this labor adaptation to forecasts. Specifically, 930 Yuan (148 USD, in 2015 currency) is gained per worker per year, with each $1{}^{\circ }C$ decrease in the city forecast $RMSE$. For the entire country, an average 3.9% increase in city-level forecast accuracy for 2011–2015 generates a considerable social benefit of 25.3 billion Yuan (4.03 billion USD) annually from the labor sector alone, nearly covering the annual cost of the national weather forecasting system.