While I am still digesting this paper and generally feel uncomforable with macro-style time-series econometrics, I like the idea of using global weather shocks rather than local weather shocks for two key reasons:
1. Because it implicitly accounts for spillovers/market responses between regions and countries;
2. Identification is more exogenous (time-series weather variation).
In other words, I believe this approach does much more to deal with some forms of adaptation and price effects than earlier panel studies that pool individual countries and purge economic spillovers with fixed effects. There is also a lot of cross-sectional identification in earlier studies (despite many claims to the contrary by the authors) due to non-linear temperature effects and the fact that temperature distributions differ so much across countries. On the face of it, I think this design is *a lot* cleaner. But I still need to get my head around some of the filtering they do up front -- I'm not a real time series / macro guy either.
I think these are thoughtful comments.
While I am still digesting this paper and generally feel uncomforable with macro-style time-series econometrics, I like the idea of using global weather shocks rather than local weather shocks for two key reasons:
1. Because it implicitly accounts for spillovers/market responses between regions and countries;
2. Identification is more exogenous (time-series weather variation).
In other words, I believe this approach does much more to deal with some forms of adaptation and price effects than earlier panel studies that pool individual countries and purge economic spillovers with fixed effects. There is also a lot of cross-sectional identification in earlier studies (despite many claims to the contrary by the authors) due to non-linear temperature effects and the fact that temperature distributions differ so much across countries. On the face of it, I think this design is *a lot* cleaner. But I still need to get my head around some of the filtering they do up front -- I'm not a real time series / macro guy either.