Hello,
I would like to ask for advice on how to plot the parallel trend line (critical assumption) to carry out difference in difference analysis. I have firm-level data set in 50+ countries from 1990 to 2000. Some countries adopted a law at different points in time. For example, France implemented it in 1995, and Belgium in 1998. More than half of the countries (control group) do not have such law in place. I am going to include year fixed effect and firm fixed effects. Because of the year fixed effect and firm fixed effect, the variable for treatment group and the variable for post-treatment years get dropped. Thus, the model will look like this:
y_i,t = beta0 + beta1*treatmentinplace + X_i,t + i fixed effect + t fixed effect
where i = firm, t = year, X = a set of controls, and treatmentinplace = 1 if a country has the law in effect in a given year.
Before I carry out the analysis however, I need to make sure that parallel trend assumption holds. Usually this is done visually using a chart. But I am not sure how I can plot these trends effectively since the treatment takes place at different years for different countries. First of all, I am thinking to take mean of y variable for each country in a given year and plot them. I thought about having a x axis of the chart to represent the number of years since the treatment. For example, for France, year 1995 would be 0, 1994 would be -1, 1996 would be 1... For Belgium, year 1998 would be 0, 1999 would be 1, and so on. However, the problem with this approach is that it would be difficult to include control group (where no such law exists) in the chart with this x-axis.
If you have any comments, I would really appreciate them. Thank you!
I would like to ask for advice on how to plot the parallel trend line (critical assumption) to carry out difference in difference analysis. I have firm-level data set in 50+ countries from 1990 to 2000. Some countries adopted a law at different points in time. For example, France implemented it in 1995, and Belgium in 1998. More than half of the countries (control group) do not have such law in place. I am going to include year fixed effect and firm fixed effects. Because of the year fixed effect and firm fixed effect, the variable for treatment group and the variable for post-treatment years get dropped. Thus, the model will look like this:
y_i,t = beta0 + beta1*treatmentinplace + X_i,t + i fixed effect + t fixed effect
where i = firm, t = year, X = a set of controls, and treatmentinplace = 1 if a country has the law in effect in a given year.
Before I carry out the analysis however, I need to make sure that parallel trend assumption holds. Usually this is done visually using a chart. But I am not sure how I can plot these trends effectively since the treatment takes place at different years for different countries. First of all, I am thinking to take mean of y variable for each country in a given year and plot them. I thought about having a x axis of the chart to represent the number of years since the treatment. For example, for France, year 1995 would be 0, 1994 would be -1, 1996 would be 1... For Belgium, year 1998 would be 0, 1999 would be 1, and so on. However, the problem with this approach is that it would be difficult to include control group (where no such law exists) in the chart with this x-axis.
If you have any comments, I would really appreciate them. Thank you!
