High correlation between variables

[Shakhzod Erkinov]

Hi all,

I am doing a research regarding the effect of family ownership and firm performance in the context of Indian companies.

The correlation between roa and roe is very high (82%). In addition the correlation between tobinsq and roe as well as roa is also high. Isn't it a problem since their way of calculation are similar to each other?

Thank you in advance.

[Phil Bromiley]

You have three measures of firm performance (ROE, ROA, market to book). There is no problem that these are highly correlated - they should be correlated.

You may get a problem if you try to enter the three on the right hand side of a model. Essentially, if you ran a regression with ROE, ROE, and Market to Book on the right hand side, the estimate would only be able to use the variance in each that is not in the other two to estimate the parameters. However, entering all three doesn't make much sense - you'd be estimating the influence of ROE holding constant ROA and MTB, etc. but it makes no sense to think about holding ROA and MTB constant if ROE is varying.

[Ihab Man]

Dear Phil Bromiley
I hope you keep healthy.
please, as you have mentioned above the ROE and ROA in the same regression ( on the right hand side) its not good because they are highly correlated .Well if I have two var and the correlate negatvie ( -.71) this also I should not put them in the same regression model ? Thank you so much