Comments!

[nickeubank]

Hi All,

First, thanks to the pointer to the repo -- very helpful!

Comments looking around:

• I think your biggest issue is just statistical power. 31 stocks just doesn't give you a lot of observations to work with, especially given your in a domain with hundreds / thousands of available stocks. I think I would suggest the following:
  • Dig more for complete datasets of public company revenue sources. For example, check the Wharton data service here.
  • You could also augment your primary analysis with a second analysis that uses "industry" instead of revenue share in china. It's not perfect, but it'd be interesting if, say, service industry companies see now effect, but manufacturing companies do (since manufacturing often has supply chains in China). And you can definitely find big datasets with hundreds of stocks, their prices, and "industry".
• Yes, I can see now you're using entity fixed effects, and using the variable as continuous. Great! Make sure you're much more explicit in your writeup. Remember: someone who reads your report should basically know enough to re-create your analysis, so when your regression table doesn't print out fixed effects, you have to explicitly say what's in the model.

[nickeubank]

(But to be more clear, I think your specification is great. I think you need to explain it better in the paper, but the code is running a sound model)