Controlling for a variable in a Probit Regression

[Theocharis Iosifidis]

Dear Forum members,

Happy new year to everyone!

I am in the final stage of finishing my MSc in Finance Thesis and I encounter a problem for which I would like to raise the following question: So, basically, I am using an Ordered Probit model in order to find out which macroeconomic variables affect the choice of payment in M&A transactions. I have three different cases: The value of 2 for Cash-only deals, the value of 1 for a combination of Cash and Stock deals, and the value of 0 for Stock-only deals. This variable which is named Method of Payment is my dependent variable. My independent variables are among others the rate of inflation, unemployment, and the interest rate that banks charge to non-financial institutions for loans above EUR 5million. My professor asked me to make the loan interest rate a "control variable" which I did not initially understand what is because I already included it as an independent variable. But then I came across this paper https://www.stathelp.se/en/regressio...oEmlrKgOZSHnno which explains what it means to control for a variable. However, I cannot understand how to implement this in Stata, since it refers to an OLS regression and not to an ordered probit regression as in my case. Can anybody assist me on that, how is it possible to control for a variable in Stata for an ordered probit regression?

Thanks in advance,

Theocharis

[Andrew Musau]

There is no reason for you not to directly seek clarification from your supervisor. He or she is paid to offer you guidance. That said, the supervisor probably means that you should not discuss or interpret the coefficient of the control variable. All you have to do is to state that the regression controls for the loan interest rate. As far as estimation goes, there is no difference in how independent and control variables are treated. The distinction is only relevant in terms of what your focus is, which is the effect of the independent variables on the outcome.

[Theocharis Iosifidis]

There is no reason for you not to directly seek clarification from your supervisor. He or she is paid to offer you guidance. That said, the supervisor probably means that you should not discuss or interpret the coefficient of the control variable. All you have to do is to state that the regression controls for the loan interest rate. As far as estimation goes, there is no difference in how independent and control variables are treated. The distinction is only relevant in terms of what your focus is, which is the effect of the independent variables on the outcome.

Dear Andrew,

Thank you for your help!

But what is the point of doing that if I may ask?

One of my components for the final output is the loan interest rate. I want to show that when the loan interest rate increases then this is negatively related to a cash payment and positively related to a stock payment (the reason is that cash deals are most of the times done via debt-raising from banks and thus when rates go up, debt becomes more expensive and thus the interest payments, indicating a lesser willingness to pay with cash). If I do not include this variable in my model by controlling it then it is the same as completely removing it right? In any case, may I kindly ask for some help on how to code this "controlling" in stata? I know that you can control in OLS regressions via the xtreg command followed by a comma for the variable that you want to control but this doesn't work for an ordered probit model. When I insert the command oprobit followed by the independent variables and then by a comma for the loan interest rate I get an error message with respect to the command.

Thanks in advance,

Theocharis

[Andrew Musau]

One of my components for the final output is the loan interest rate. I want to show that when the loan interest rate increases then this is negatively related to a cash payment and positively related to a stock payment (the reason is that cash deals are most of the times done via debt-raising from banks and thus when rates go up, debt becomes more expensive and thus the interest payments, indicating a lesser willingness to pay with cash).

In this case, you are treating the variable not as a control variable but as an independent variable.

If I do not include this variable in my model by controlling it then it is the same as completely removing it right? In any case, may I kindly ask for some help on how to code this "controlling" in stata? I know that you can control in OLS regressions via the xtreg command followed by a comma for the variable that you want to control but this doesn't work for an ordered probit model. When I insert the command oprobit followed by the independent variables and then by a comma for the loan interest rate I get an error message with respect to the command.

As I said in #2, in terms of estimation, control variables are entered into the regression as independent variables, same as other independent variables. The confusion, I suppose, arises from the terminology of an independent variable in a model versus an independent variable in a regression. In a model, independent variables are the focus of your study. Control variables are part of the model, but you are really not interested in their association with the outcome variable. In a regression, both the model control variables and independent variables are independent variables (regressors). So the way I interpret your supervisor's remarks is that you should not come up with any hypotheses relating to the loan interest rate and you should not discuss or interpret its coefficient, beyond stating that the regression controls for it. Bottom line, talk to your supervisor about why it's not a good idea to have this variable as an independent variable in the model. I can only guess that he or she want you to focus on something else.