Empirical tests on International Trade

We are interested in estimating the following system of equations:

where NX/GDP is the net export over GDP of country i at time t. ΔCAIi,t denotes the percentage change of the climate attention index for country i and quarter t, ΔCAIt is the cross sectional average of CAI at date t, i.e., the global CAI index. The term control is a vector of control variables including the change in the industrial production index to control for country-specific productivity shock, and the share of Twitter volume on days when the CAI is at the bottom 5%, to control for other events that affect Twitter activities.

Given our interest on the role of global climate news shocks on the dynamics of international trade, we focus on the estimated parameter Γ. According to our model, this coefficient should be negative.

Our estimates support the prediction of our model, as they suggest that an adverse global climate shock produces an outflow of resources away from countries with a lower exposure. This result is confirmed across different ways to aggregate our country-level CAI in order to form a global index. In addition, our result is driven mainly by country pairs in which there is at least one developed economy. This is broadly consistent with the spirit of our model since we assume that risk sharing is implemented with frictionless trade.

GMM ESTIMATION OF Γ.

Notes: The figure shows estimates of Γ using expanding time windows. Each window starts in 2015:Q1 and ends on the date reported on the x-axis. The parameter Γ is estimated using GMM. Panels in different columns refer to results obtained from different ways to aggregate country-level CAI indices in order to form a global component. Panels in different rows refer to results obtained from different groups of country-pairs. Advanced economies are defined according to the IMF. Standard errors are HAC-adjusted. The shaded areas present the 90% confidence interval.