# Regression coefficient for a logged dependent variable

I am running into a puzzling econometric issue.

Here is my data set-up and the model:

I am using a two way fixed effects model on a panel data of states. For regression results, I take natural log of my dependent variable $$Y$$. My Treatment variable is a tax policy dummy which takes a value zero for the years prior to the policy and one after it is adopted by the states. States adopt this tax policy at different times, so my treatment timing varies in the sample.

Here is the econometric issue:

When I run my panel data regressions, I obtain a regression coefficient for my treatment variable which is 1.14 and it is statically significant. This result implies that the treatment policy increases the dependent or the $$Y$$ variable by 114%. This makes no sense. Further since my dependent variable is measured on a log scale, there isn't an issue of outliers in the $$Y$$ variable which might have caused this.

Question -- Any ideas what could be creating this puzzling issue? How can I troubleshoot this problem?

• If Dimitriy's post isn't helpful, maybe you could elaborate further on why this doesn't make sense. The increase is likely greater (i.e., $e^{1.14}$). Also, care to share your output? – Thomas Bilach Oct 16 '20 at 23:48