The jtrain.dta dataset contains information on job training, and other variables, at a variety of manufacturing facilities. Here is a list with brief definitions of the variables in the data. Label your answers a., b., c., d., e.

1. year1987, 1988, or 1989

2. fcodefirm code number

3. employ# employees at plant

4. salesannual sales, $

5. avgsalaverage employee salary

6. scrapscrap rate (per 100 items)

7. reworkrework rate (per 100 items)

8. tothrstotal hours training

9. union=1 if unionized

10. grant=1 if received grant

11. totraintotal employees trained

12. hrsemptothrs/totrain

13. lscraplog(scrap)

14. lemploylog(employ)

15. lsaleslog(sales)

16. lreworklog(rework)

17. lhrsemplog(1 + hrsemp)

18. lscrap_1lagged lscrap; missing 1987

19. grant_1lagged grant; assumed 0 in 1987

20. lavgsallog(avgsal)

Use the command "xtset fcode year" to inform STATA that this is a panel dataset.

Run a pooled regression of lhrsemp on lavgsal lscrap grant

a. Report below the coefficient and t-statistic for lscrap.

b. All else equal, a 1% decrease of avgsal leads to a ____ change in hrsemp.

Run a fixed effects regression with the same variables from above, where the fixed effects entities are the firms. For reasons that will be apparent below, I recommend using the dummy variable technique you utilized in Homework 6.

c. All else equal, when a firm gets a grant this will increase the lhrsemp by ____. The confidence interval for this increase is between ____ and _____.

d. The intercept for firm number 418107 is ___.

e. With these control variables, the relationship between lscrap (a measure of the waste produced during production) and lhrsemp (the hours per employee spent in training) is negative and significant. Speculate on why this might be the case.