I can't access the NBER paper so I can't see the methodology - how did they choose which papers to include? I'm always wary of studies that tout the size of the

I can't access the NBER paper so I can't see the methodology - how did they choose which papers to include? I'm always wary of studies that tout the size of their data set. Did they dedupe wire stories? Control for which newspapers were included? I feel like more often than not, it turns out these effects are mostly due to mix shift / selection bias / other methodological artifacts.