I have a panel data of patents from the grant year of the patent to the year that patent is traded to other entities (1,059,024 patent-year observations.). 34,375 patents exposed to a policy in year t (the exact year of each treatment patent may be different), There are 34,375 treatment patents, based on the observed characteristics of the treatment patents in t-1 year (i.e., one year before the patent is exposed to the policy), we apply 1-to-1 nearest-neighbor matching without replacement and with 0.5 caliper to estimate propensity scores on which treated and untreated patents were matched based on. Treatment observations whose propensity scores are higher than the maximum or less than the minimum propensity scores of the controls are dropped. The treatment patents that fail to be matched to a control were discarded. We end up with 11,091 unique pairs of matched patents (i.e., 11,091 treatment patents and the corresponding 11,091 control patents).
Then, I try to conduct difference in difference analysis on these matched pairs. There are 99,374 patent-year observations generated by these 11,091 pairs of matched patents. I posted my code, while it seems that I should add pair id of these matched patent pairs. But I am not sure whether I should add i.pair_id, this dummy variable to the model? Is the specification of the model correct?Thanks!
xi:logit Y post*treatment controls, or vce (cluster record)
Then, I try to conduct difference in difference analysis on these matched pairs. There are 99,374 patent-year observations generated by these 11,091 pairs of matched patents. I posted my code, while it seems that I should add pair id of these matched patent pairs. But I am not sure whether I should add i.pair_id, this dummy variable to the model? Is the specification of the model correct?Thanks!
xi:logit Y post*treatment controls, or vce (cluster record)
