Matched pair design and control variables

#1
Hi everybody,

Currently I'm doing a Master thesis and I have a question related to that. I'm running a regression and I created a matched pair design sample (based on firm size, industry and year). However, my regression already has these variables added as control variables. I have read papers (in top journals) who also run a matched pair design regression based on a certain variable (for instance return on assets), but also have added this variable as a control variable. To me this seems a bit double over, but probably I'm missing something.

Can somebody tell me the difference between matching on a certain variable and adding a certain variable as a control variable. And more specifically: What is being analyzed extra when creating the matched pair sample when you already ran a regression with this control variable?

I hope my question is clear and somebody can help me with an answer.

Many thanks in advance.
 

Karabiner

TS Contributor
#2
A variable is a possible "control variable" if it is associated with
one or more independent variables., and also with the dependent
variable.

If the "control variables" are used as matching criteria, then there is no
longer an association with "group", but still with the dependent variable.
Therefore, model error will be reduced and group effects have a greater
chance to be detected.

With kind regards

Karabiner
 
#3
Thank you for your response Karabiner, much appreciated. What exactly do you mean by "group" in the sentence
If the "control variables" are used as matching criteria, then there is no longer an association with "group"?

Kind regards,

Jens