Economic significant on logit regression.

[Vincent Tawiah]

Hello members, Please how do i calculate and interpret the economic significance from logit and ordered logit regression results.
The dependent variable is binary (0,1) and independent variables are scores between (0, 1, 2, 3).
Thanks for your help

[Carlo Lazzaro]

Vincent:
welcome to this forum.
Too broad/vague a question: without further details, see -logit- and -ologit- entries in Stata .pdf manual.

[Vincent Tawiah]

Hello Carlo,
Thanks for the comment.
Unfortunately, i could not access this see -logit- and -ologit- entries in Stata .pdf manual. Please, can you attached here or resend the link?

In some articles, the economic significance was calculated as the coefficient of the variable times the standard deviation. So i want to know if I can apply the same thing for logit regression.
thanks

[Carlo Lazzaro]

Here you are (actually the links below point you to the Stata 13 .pdf manual, but I think it would suffice for your purposes):
https://www.stata.com/manuals13/rlogit.pdf;
https://www.stata.com/manuals13/rologit.pdf

[Vincent Tawiah]

Hello Carlo, Thanks for the references, I have read through carefully plus other materials but I couldn't find anything about calculating economic significance. Am revising an article for a top journal and the editor requested calculating the economic significance. I am not sure if I can apply the standard formula of the coefficient of the variable times the standard deviation.

thanks

[Carlo Lazzaro]

Vincent:
sorry, but, despite being an economist, I've never heard about economic significance.

[Richard Williams]

Perhaps the listcoef command, part of Long and Freese's spost13 suite of commands, can do what you want. Type

findit spost13

I briefly describe the command at

https://www3.nd.edu/~rwilliam/xsoc73994/L04.pdf

[Phil Bromiley]

In my field, we do talk about substantive or economic significance meaning not a test but rather whether the change in the predicted dv for a reasonable change in the rhs variable of interest is large enough to be of interest. As it well known, with large samples you can get statistical significance even if the effect doesn't change the predicted value much at all.

While there are effect size indicators often used in some fields (check the subject index for that available in Stata), I usually use margins to look at the change in the predicted value for a given change in the variable of interest. This might be a one standard deviation change or it might be something else that makes sense given the measure.

[Vincent Tawiah]

Thanks so much for the comments