cluster

[Ihab Man]

Dear Andrew Musau
Please, I have seen your posts related to clustering, and I hope you can help me with my questions.
Please, I have a panel data set for 500 companies from 11 countries with regular period (2000-2010) with 9 explanatory variables (not dummies). On the other hands my dependent variable is dummy if the company issue securities 1 and 0 otherwise and I will see which from 9 explanatory variables motivate the company to issue it. For example company 1 issue only in year 2007 I generate that this company equal 1 on year 2007 and the rest years equal zero and so on and those company not issued (all years equal zero.) is it correct the producers yes ?
I used xtlogit, fe but not work and many observation dropped, then I tried to use xtlogit, re but I have seen previous studies said that they are clustering the standard error at company level. Well, please, can you explain me why they did that? And it will work for my case? If yes? How can I do that in the command but without vce? And in this case can I use cluster at year level?
Another question please, can you tell me what is the correct command for my case to winsorize all variables in the model at the 1st and 99th percentiles? And please how can separate the sum statistic into two group (company issued, company not) Thank you so much in advance.
Best regards

[Andrew Musau]

On the other hands my dependent variable is dummy if the company issue securities 1 and 0 otherwise and I will see which from 9 explanatory variables motivate the company to issue it. For example company 1 issue only in year 2007 I generate that this company equal 1 on year 2007 and the rest years equal zero and so on and those company not issued (all years equal zero.) is it correct the producers yes ?

In terms of how you define your dependent variable, that is the correct way to code it.

I used xtlogit, fe but not work and many observation dropped, then I tried to use xtlogit, re

As fixed effects models look at the within variation, it will drop observations of firms that do not vary (outcome switches from 0 to 1 or 1 to 0) over the sample period. This in itself should not be a problem as long as you do have enough firms where there is variation. So do not discount the fixed effects results and if you have to use random effects, use the Hausman test to make sure that you can justify this choice.

I have seen previous studies said that they are clustering the standard error at company level. Well, please, can you explain me why they did that? And it will work for my case? If yes? How can I do that in the command but without vce?

With clustering, you assume that observations within clusters are not independent but are across clusters (in your case firms). This is reasonable as the decision to issue securities this year for a company has strong dependence on what happened in the previous year. In your case, you have enough observations and clusters, so there should be no issue. The only way to do this in xtlogit is to use the option -vce(robust)- in RE or -vce(bootstrap)- in FE. The clusters will be based on your panel variable which I assume is firm.

And in this case can I use cluster at year level?

No, not with xtlogit and I do not see any compelling reason why you should do this.

Another question please, can you tell me what is the correct command for my case to winsorize all variables in the model at the 1st and 99th percentiles? And please how can separate the sum statistic into two group (company issued, company not) Thank you so much in advance.
Best regards

Start a new thread with an appropriate title and ask this question.