0
$\begingroup$

I am trying to make a regression on the following model:

Y = β0 + β1 X1 + β2 X2 + β3 X3 + β4 X4 error

I am studying the effect of X1 on Y and added X2, X3 and X4.

The data I have is cross sectional data (70 observations) and the variables are:

Y is the Market-to-book value X1 is Social response performance (a score given out of 100) X2 is total assets X3 is debt-to-equity ration X4 is the revenue growth

Each stock has 2 observations, one for 2020 and the other for 2019 (I don't think this is relevant because it is not time-series)

The r-squared i get is very low although this model has proven a very strong r-squared in research. Am I doing something wrong?

Is it the values are very far from each other?

For example the total assets values are above tens and hundreds of millions. For some observation it is (100500300 -51000200 - 900000000 .. etc). On the other hand the social responsible scores are all under 100 and the revenue growth values are always under 0.5.

I appreciate the feedback.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

The scale of the variables isn't going to affect the $R^2$

All you can really do, if you don't have other explanatory variables is to investigate non-linear associations (either via transformation of variables or introducing non-linear terms) and interactions.

On the other hand, $R^2$ isn't always important. See here for example:

Is $R^2$ useful or dangerous?

Improve this answer
$\endgroup$
4
  • $\begingroup$ Thank you. The ANOVA (tests the null hypothesis that multiple R in the population equals 0) has a Sig. of .95 and the variables have a significance level of 0.7, 0.75 and 0.8. I am suspecting I am doing something wrong. Could you please help me? $\endgroup$
    – Santi1980
    Commented Jun 6, 2021 at 8:28
  • $\begingroup$ The data are like that: Variable one is price to book ration (independent) and it is market cap divided by book equity value. Variable 2 (dependent) is the Socially responsible score. A rating that is given out of 100. Variable three (dependent) is the total assets (values that are very large and reaches billions). Vriable 4 (dependent) is the debt to equity ratio in percent which is the book value of debt divided by book value of equity. Variable 5 (dependent) is the revenue growth (this year's revenue minus last year's revenues dvided by last year's revenues). $\endgroup$
    – Santi1980
    Commented Jun 6, 2021 at 8:33
  • $\begingroup$ The data itself is organized such that each company has 2 observation (one for each year). I am considering 35 companies, so the total number is 70. I do not think this is relevant for a cross sectional regression though, because it does not take time into consideration and it deal with each observation regardless of the company id and the years. So, I just choose the first variable as dependent and the rest are independent. Although I am only interested in the effect of variable 2. Am I correct or should I warn SPSS that some variables are just control variables (3,4 and 5). $\endgroup$
    – Santi1980
    Commented Jun 6, 2021 at 8:37
  • $\begingroup$ Everything you've said indicates that this is simply not a good model for these data. You should check the usual model assumptions - linearity and independence. It is likely that these data are no independent since you have repeated measures withing company so you should defnintely deal with that - for example by fitting random intercepts for company (this would then be a mixed effects model). $\endgroup$
    – Robert Long
    Commented Jun 6, 2021 at 9:40

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.