This paper exploits vast granular data – with over one million county-month observations – to estimate a dynamic panel data model of weather’s local employment effects. The fitted county model is then aggregated and used to generate in-sample and rolling out-of-sample (“nowcast”) estimates of the weather effect on national monthly employment. These nowcasts, which use only employment and weather data available prior to a given employment report, are significantly predictive not only of the surprise component of employment reports but also of stock and bond market returns on the days of employment reports.
J. Wilson, Daniel. 2017. “Clearing the Fog: The Predictive Power of Weather for Employment Reports and their Asset Price Responses,” Federal Reserve Bank of San Francisco Working Paper 2017-13. Available at https://doi.org/10.24148/wp2017-13