Dear all,
I would like to have your opinion on this matter. I have a dataset for the labor market. I estimate the impact of change in minimum wage on job retention (probability to keep the job). I want to see the impact across the distribution of incomes and I created 8 groups: group1 consists of those with wages below the new minimum wage and others are with wages above old minimum wage and below new minimum wage. I have some periods with no change in the minimum wage (4 years in row) and others with the increase in the minimum wage.
I supposed that group2 is the treatment group and all others are control groups. I omitted the group7 as I think it is with quite high wages and less affected.
I created 2 dataset which contains different samples as there are different workers. I see who keeps the job only for the fourth quarter because of institutional reasons, run the OLS and see the coefficients for each group for the "treatment" years (years with change in minimum wage) and I run the OLS also for the other dataset for the "control" years (with no change in the minimum wage). I have cross-section data and not panel and for 3 years around 3000 observations so it is not convenient to use matching as I have few obs.
Then I computed the z-test to see the difference between coefficients from treatment and control periods.
I interpret it as a kind of Difference-in-Differences because I compare the difference between groups and then between years.
I would like your opinion.
Does(Did) anyone apply it? Or have anyone read something like run DID from 2 dataset and then compare the coefficients using z-test? Is it a suitable way to proceed or it is better to append all years and do the standard DID?
Thank you very much
With kind regards,
Simona
I would like to have your opinion on this matter. I have a dataset for the labor market. I estimate the impact of change in minimum wage on job retention (probability to keep the job). I want to see the impact across the distribution of incomes and I created 8 groups: group1 consists of those with wages below the new minimum wage and others are with wages above old minimum wage and below new minimum wage. I have some periods with no change in the minimum wage (4 years in row) and others with the increase in the minimum wage.
I supposed that group2 is the treatment group and all others are control groups. I omitted the group7 as I think it is with quite high wages and less affected.
I created 2 dataset which contains different samples as there are different workers. I see who keeps the job only for the fourth quarter because of institutional reasons, run the OLS and see the coefficients for each group for the "treatment" years (years with change in minimum wage) and I run the OLS also for the other dataset for the "control" years (with no change in the minimum wage). I have cross-section data and not panel and for 3 years around 3000 observations so it is not convenient to use matching as I have few obs.
Then I computed the z-test to see the difference between coefficients from treatment and control periods.
I interpret it as a kind of Difference-in-Differences because I compare the difference between groups and then between years.
I would like your opinion.
Does(Did) anyone apply it? Or have anyone read something like run DID from 2 dataset and then compare the coefficients using z-test? Is it a suitable way to proceed or it is better to append all years and do the standard DID?
Thank you very much
With kind regards,
Simona
