Tutorial 6: Stata Application of Panel Data Method The University of Queensland ECON7360
Instructor: Rigissa Megalokonomou
Problem I: Differences-in-Differences: Effect of Worker Compensation on Weeks out of Work
Background The data set for this exercise comes from the paper by Bruce D. Meyer, W. Kip Viscusi, David L. Durbin Workers¡¯ compensation and injury duration: Evidence from a Natural Experiment, published in The American Economic Review, Vol. 85, No. 3, p. 322-340.
Idea MVD studied the length of time that an injured worker receives workers¡¯ compensa- tion. On July 15, 1980, Kentucky raised the cap on weekly earnings that were covered by workers¡¯ compensation. An increase in the cap has no effect on the benefit for low-income worker, but this makes it less costly for a high-income worker to stay on worker¡¯s com- pensation. Therefore, the control group is low-income workers, and the treatment group is high-income workers; high-income workers are defined as those who were subject to the pre-policy change cap. Using random samples both before and after the policy change, MVD were able to test whether more generous workers¡¯ compensation causes people to stay out of work longer. They started with difference-in-difference analysis, using log(durat) as dependent variable.
Use the data in injury.dta for the following questions.
(i) Load the data and describe the data.
(ii) Estimate the differences-in-differences model with regression with and without ro- bust standard errors for Kentucky data.
(iii) Re-estimate (ii) using Kentucky data by adding explanatory variables, male, married, a full set of industry and injury type dummy variables. Use robust standard errors. How
1
does the estimate on the interaction term, afchnge ¡¤ highearn, change when these other factors are controlled for? Is the estimate still statistically significant?
(iv) What can you say about the small R-squared from parts (ii) and (iii)?
(v) Estimate (ii) using Michigan data using robust standard errors. Compare the esti- mates on the interaction term for Kentucky and Michigan. Is Michigan estimate statistically significant? What do you make of this?
Problem II: Differences-in-Differences: Minimum Wages and Employment
Background The data set for this exercise comes from the paper by David Card and Alan Krueger Minimum Wages and Employment: A Case-Study of the Fast-Food Industry in New Jersey and Pennsylvania, published in the American Economic Review, September 1994, vol. 84.
Idea Your basic microeconomic theory of the firm in competitive factor markets tells you that factor demand curves slope downwards. This implies that if minimum wages are binding, we would expect employment to fall. This conventional wisdom among economists was challenged using this and other data by Card and Krueger. Their research design was based on New Jersey raising its minimum wage from $4.25 to $5.05 on 1 April 1992 while the minimum wage in neighboring Pennsylvania remained unchanged. They collected data on wages and employment in 65 fast-food restaurants in Pennsylvania and 284 in New Jersey in Feb/March 1992 (i.e. before the rise in the NJ minimum wage) and in Nov/Dec 1992 (i.e. after the rise). They use a difference-in-difference design to investigate the impact of minimum wages on employment.
Use the data in minwage.dta for the following questions.
(i) Load the data and describe the data.
(ii) Type the following command. In Stata, type: tab nj after, su(fte) means . Derive the difference-in-difference estimate from this table.
(iii) Estimate the differences-in-differences model for regressions with and without ro- bust standard errors. Why the difference in the standard errors is sizeable?
(iv) Now estimate the following model in levels: In Stata, xtreg fte nj after njafter, fe robust . What is the coefficient of interest?
2
(v) State key assumption for DID estimator.
(vi) Why is all the estimated impact of the minimum wage the same in all these models?
3