7,649
edits
Changes
Jump to navigation
Jump to search
Estimating Unobserved Complementarities between Entrepreneurs and Venture Capitalists (view source)
Revision as of 19:34, 25 May 2021
, 19:34, 25 May 2021→Dataset Rebuild
**Crucial difference: Only investments within ComboKeys_Code20 are included in the history counts. Before anything in RLJoinerFF or even RoundlineBase were included, which created the impression of overcounts.
**The vast majority of missing distances were caused by missing firm addresses for just 391 VCs (some of these were state UN, and later removed). However, we have zip codes for almost all of them. I built firmbogeoplus (see [[VCDB20]]) to add in zcta centroids from the U.S. Census gazetteer where available. However, 47 zips weren't in the ZCTA lookup, and they account for 113 firms that participated in 3,997 pairs (real and synth). So I also ran GeocodeOneKey.py with a (no header) Zip\tZip input (MissingFirmZips.txt), manually added three records (02801,85292,91399) and loaded up the result as zipgeoaddon (from MissingFirmZips-Geocoded.txt) in Load.sql.
**Also required that matches have a code(/20/100)
**Changed the Experience table to count experience in previous years rather than previous rounddates, as then experience doesn't change within a market except through the match.
*Changed the join to the ranking. New build uses PlaceYearRanking, and joins on placename, state, year (placename is available in PortCoSuper).
**There was an error with Centerville Ohio appearing in the ranking twice each year from 1998 to 2001.
