The effect of industries on national culture

[Rik Velthoven]

Dear Statalist,

I got a question about a regression model I want to set up. I will first briefly explain my problem. I am currently doing research about the effect of national culture on the abnormal return of firms during the announcement period. I have a panel dataset which covers about 2000 different firms engaging in a merger during the time period 1996-2016. I the first regressions I have run the CAR3 is my depedent variable and national culture is my idependent variable. The CAR3 is the combined return of the target and acquirer firm surrounding the announcent period. Where the CAR3 is the combined return of 1 day before the announcement till 1 day after the announcement. National culture is the relative distance in culture between two countries measured by the Hofstede Index. My first regressions looks as follows;

CAR3=constant+B(National Culture)+B(control variables)

I found a negative effect of naional culture on the combined returns of the target and acquirer firm. So far so good.

However want to test in what way different industries influence the effect of National culture on the combined announcement returns (CAR3). I have 10 different industries in my sample. The problem I am facing is that I don't know which regression model to run. I was thinking about including a dummy variabel for industry and making an interaction effect with national culture. the regression would then looks as follows:

CAR3=constant+B*(National_Culture)+B*(dummy_indust ry*National_Culture)+ B*(control variables)

I was wondering if this is a propper approach to do it. I thank you for your time on beforehand

Kind Regards,
Rik

[Clyde Schechter]

The approach of using an interaction model is appropriate. A correctly specified regression equation would include industry indicators (dummies) themselves in addition to their interaction with national culture. Now, it may be that you are working with firm-level panel data and a fixed-effects model and the industry indicators would end up being omitted due to colinearity anyway, but there is no harm in including them in the model anyway and letting Stata drop them for you. In fact, it is a safer way to proceed, because if Stata does not drop them, then you will be alerted to the fact that there is some error in the coding of your data that has broken the anticipated colinearity.

If you are not doing a fixed effects regression, then the model would be mis-specified if you leave out the industry indicators themselves.

Finally, do use factor variable notation when you code this so that you will be able to use the -margins- command afterwards. Use of -margins- greatly simplifies the interpretation of interaction models. If you are not familiar with factor variable notation, read -help fvvarlist-. If you are not familiar with the -margins- command, I recommend the excellent Richard Williams' https://www3.nd.edu/~rwilliam/stats/Margins01.pdf, which is a model of clarity and includes several worked examples similar to yours.

[Rik Velthoven]

Thanks a lot for your response I will have a look at the suggested file you attachted