Out-of-sample logit: applying estimated coefficients on a new series (w/o observing 0s/1s)

[John Rivers]

Hello STATA community,

This is both a methodological question and a programming question, for which I apologize a bit in advance. Perhaps the Statalist community can help me with both.

I have a database (Base A) in which I observe defaults (so, a vector of 0/1s), with corresponding (explanatory) financial data. I have another similar database (Base B) in which I observe the same regressors, except I do not know whether the firm has defaulted or not; i.e., I only have the financial data.

I was curious how I might go about applying my estimates derived from my Base A logit to say something about the probability of default of my firms in B.

From a programming perspective, is there a convenient way to do this in STATA? And, I suppose from a methodological perspective, how would I go about knowing if this is econometrically sound? I assume the two series have to be similar in certain ways (similar variance, etc).

Thank you so much for your help, and my apologies if this has been answered elsewhere. I did my best to poke around first.

John

[Maarten Buis]

Applying a model to another dataset is easy.

use data_a, clear
logit default x1 x2 
use data_b, clear
predict pr

Whether the results make sense (econometricaly, or otherwise) is just a matter of common sense: do you believe that the same mechanisms are at work in dataset a and b. If the answer is yes, then you are fine regardless of variances or means or...

[John Rivers]

Thank you, Maarten!