* read in data for kentucky from
* meyer, dubin viscusi
use kentucky

* output log file
log using kentucky.log, replace


* generate log duration
gen ldurat=ln(durat)

* sort the data by highearn and afchnge
sort highearn afchnge

* gets means of ldurat for
* 2x2 table
by highearn afchnge: sum ldurat


* get the treatment effect
gen treat=highearn*afchnge

* run difference in difference regression
reg ldurat highearn afchnge treat


* get other covariate
gen lage=ln(age)
gen lprewage=ln(prewage)
gen high_x_lpre=highearn*lprewage

xi i.industry i.injury

* this next regression replicates the results in 
* column (i) of table 6, page 336

reg ldurat highearn afchnge treat male married lage lprewage high_x_lpre _I*


*log close