Add control dummies to a difference in difference study

[David Cremers]

Hi all,

I am doing a study on the effect of Private equity on the employment dynamics. In which I do a difference in difference study to compare the Productivity (= revenue / employees), compensation (employee cost / employees), and log employees between PE deals (treatment) and NON PE deals (control group). My model is explained below:

• where represents the firms in the sample and represents the time periods relative to the deal announcement.
• is a dummy variable for the country (5 countries in West Europe)
• is a dummy variable for the industry (4 different categories)
• is the number of employees in firm i at time t,
• is the average compensation per employee in firm i at time
• is the productivity in firm i at time t.
• is the treatment indicator variable, with for private equity-owned firms and for non-private equity-owned firms.
• is the log of the dealsize of entity i.
• is a time indicator variable, with for the year of the deal and for after the deal.
• is the error term.

I ran my regression in Stata without controls in the following way:

Treatment = 0 for NON PE, 1 for PE
Post = 0 for pre deal, 1 for post deal
B3 = the interaction dummy between Post and Treatment

and tested for multicollinearity:

But when I run my model using 'reg'. I have not been succesfull adding control variables to my 'didreg' command. I get the feedback from Stata that the control part 'is not allowed'.

The is multicollinearity while it was not there previous. Where is my mistake and how do I Fix it?

[David Cremers]

No response needed, of course deal size is related to both Post and B3 so that caused Multicollinearity