A pitch based on differential comparisons between non-protected monopoly industries and non-protected competitive industries.
The return to protectionism or Paul Krugman's summary for the motivation behind this pitch.
- County boundaries are hard borders. Make these more flexible with a falloff over county distances.
- No validation years nor mechanism (would be nice to validate on 2000 steel tariffs that employment change occurs).
- Make sectoral data come with its year to facilitate downloading the proper survey dataset for businesses.
- Sectoral data is incomplete. Try to improve.
- Add sectoral data for other events.
- Loss function is poor. Try to adjust
county_loss_func
(you'll wantpenalty
to reflect how indicative of the trend described in the papers the current county (row) is. Higher is worse)