I have a panel data model as following, $$Y_{it}=intercept +B_1X_{it}+B_2Q_{it}+error_{it}$$

where $i$ is for firm and $t$= time.

$Y$= trade credit demand, $X$= inventory cycle, as inventory moves faster firms demand more trade credit. On the other hand because trade credit means extra inventory, the cycle would be affected by it, too.

I know that the model suffers from endogeneity because $Y$ affects $X$ and in return $X$ affects $Y$. Would it be a valid procedure if I tried to work around this issue by assigning a dummy variable and not use $X$ in the analysis at all by calculating first yearly average for the cycle and assign 1 if a firm is above the average for that year and 0 if below it. Maybe the code I am using may clarify any vagueness.

in the R code below.

mydata=mydata%>%mutate(Dummy=case_when(X>=mean_x~1, TRUE~0))

and the model becomes as following:

$$Y_{it}=intercept+B_1 Dummy+B_2 Q_{it}+error_{it}$$

I hope I am clear enough. thanks in advance!


1 Answer 1



I don't see how the dummy variable you are proposing would give you the same (or even similar) analysis as in the original model.

I would recommend dealing with the endogeneity in another way. IV estimation, or one of the more recent methods, difference-in-difference, matching etc. One of the will be suitable.

  • 1
    $\begingroup$ Thank you for your answer. Let me explain the issue a bit more. I am more interested in the sign of the coefficient than its magnitute. When I run the regression as I described it, the coefficient is both significant and positive, suggesting that those that have higher rate of inventory cycle actually demand more TC, than those below the average, which is actually the asnwer I wanted to obtain. But before I turn in my paper I want to make sure that it would be, at least in principle, valid use of a dummy variable. $\endgroup$
    – krkc_bhdr
    Commented Feb 15, 2019 at 12:12
  • $\begingroup$ So then a positive dummy variable will then tell you that firms above the average will impact Y in a particular way. Is that really what you want to say? $\endgroup$
    – user22485
    Commented Feb 15, 2019 at 13:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.