To further elaborate on my question, assume that I have a time series dataset of Tax X and Tax Y, where in Tax X is paid by 100% of the sample while Tax Y is paid by 75%. Both taxes differ with regards to how they are implemented and collected, but are ultimately based on business revenues. Now if I am trying to figure out the relationship between Tax Y and Tax X to estimate Tax Y for the 25% of the sample for which there is no Tax Y data, is it kosher to regress Tax Y on Tax X? It is clear to see that Tax X is not causing Tax Y and that both taxes are being determined by individual business revenues, which is not known.


There can be a linear relationship between variables with no causal relation directly involved. It's fair to check if the linear model is a good fit in order to make estimations.

I personally don't know any single model which needs something like "direct causality assumption".

| cite | improve this answer | |
  • $\begingroup$ Interesting. I don't have too much experience with time series data, but I was under the impression that since you have a greater degree of ability to identify and establish things like Granger causality, regression specifications with spurious relationships like the one I laid out were not considered useful. $\endgroup$ – Naj S May 28 '15 at 20:42
  • $\begingroup$ +1 In the absence of a causal idnetification strategy, linear regression is just a handy descriptive or predictive tool. Apply as necessary, where necessary. $\endgroup$ – conjugateprior May 28 '15 at 22:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.