Impact on currencies

We calculate the average exchange rate variation in the days following a substantial increase in climate attention. The following figure illustrates the average cumulative exchange rate variation following these specific days. 

To ensure that our observations are not influenced by idiosyncratic patterns in our dataset, these averages are computed relative to the average cumulative variation observed in the remaining 85% of business days in our sample. Our findings reveal that adverse climate news shocks are linked to the contemporaneous appreciation of countries that are comparatively more exposed (i.e., possessing higher β values).

In the previous figure, we illustrate this effect among pairs of countries consisting solely of advanced economies (left panel), developing economies (right panels), or a combination of advanced and developing countries (middle panel). The appreciation pattern is observed across all country pairs. Furthermore, it tends to persist longer when at least one emerging country is involved in the analysis.

In our appendix, we show that our results are unchanged when we focus on either a GDP-weighted global index or a Twitter volume-weighted global index. We also replicate the same exercise by focusing on nonconsecutive top-15% days. Specifically, we look at cumulative returns over a 10-day window and consider only days with no overlapping time windows. Our results are qualitatively unchanged.

Type of news. We take advantage of our text analysis results and estimate the following regression:

where St denotes the share of tweets focused on transition risks at day t, normalized between 0 and 1. This share is computed by using both positive and negative tweets classified as related to transition risk. In the below figure, we depict the composite coefficient bk + ck · St for different values of St. This specification enables us to identify whether our results are primarily driven by transition or physical risk. We find that the results are stronger when St = 1, i.e., in top days when attention is entirely about transition risk.

We repeated this exercise using the share of tweets related to transition risk with a negative sentiment and obtained virtually identical results. Even though in the model we do not distinguish between physical and transition climate news, in the data we find that currency appreciation is mostly driven by negative news about transition risk.