Diff-in-Diff Regression for Minimum Wage Impacts

[Rachel Lee]

Hi all, I'm very new to Stata and I'm doing my own research for a class to write a paper. I'm trying to figure out the level of inequality by using two states (one has minimum wage hike and the other does not). I'm stuck here because I don't know how to code this, or how to even start. Can anyone help and point me in the right direction?

[Justin Niakamal]

There is a user-written command diff (ssc install diff), which uses minimum wage impacts from Card and Krueger 1994 as an example in the help file. Here are some helpful resources that show you how to use the command as applied to the aforementioned minimum wage study:

https://mpra.ub.uni-muenchen.de/4394...with_stata.pdf
https://www.stata.com/meeting/uk12/a...uk12_villa.pdf

You might also find this helpful:

http://www.princeton.edu/~otorres/DID101.pdf

Hope this helps.

Card, D., Krueger, A. "Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania".
The American Economic Review, Vol. 84, No. 4 (Sep., 1994), pp. 772-793.

[Konrad Franco]

Hi Rachel--

From your post I assume that you are attempting to identify the causal effect of a minimum wage increase on inequality. Difference-in-Difference (DD) usually takes the following format in the regression framework: Outcome_it = β1 +β2 Treat_i +β3 Post_t +β4 (Treat×Post)_it +ε_it. Treat is a dummy if the observation is in the treatment group and Post is a post treatment dummy.

For your purposes you likely need to take a few steps. First, create a dummy variable to indicate the time when the treatment started. For example, if we assume that the treatment (i.e., min wage increase) started in 2015, then observations before 2015 will have a value of 0 and observations during and after 2015 will have a value of 1. Second, create a dummy variable to identify the group exposed to the treatment. For example, if we were to assume that your data has only two states then the observations from the state that did not experience the min wage increase would have a value of 0 and the observations from the state that did experience the min wage increase would have a value of 1.

Here is some crude "pseudo code":

gen post = 0
replace post = 1 if year >= year of min wage increase
gen treat = 0
replace treat = 1 if state == state w/ min wage increase

regress y treat##post, robust

The coefficient for interaction between the treatment and post treatment binary variables "treat*post" is the DD estimator. Please keep in mind though that there are some strong assumptions that accompany the use of this identification strategy, including the parallel trends assumption.

There is also a handy user written (courtesy of Juan M. Villa) for DD. You can install it and look at the details/options on your own.

ssc install diff
help diff

Here is a terrific slide deck from Wooldridge that should serve as a good resource (http://www.nber.org/WNE/lect_10_diffindiffs.pdf).

It is a best practice to post sample data when you post to the list. Look in the using the "dataex" package to provide example data that is applicable to your case in the future. If you have future questions about this project or Stata please ask a more direct question so that we can provide more targeted help.

--Konrad