Making sense of two Regression Models

[Anupam Ghosh]

I am looking at the effect of a treatment on county-wise wages and unemployment.

I have a Wage_Standardized_t and %Change_Wage_Standardized_t. Now, I run the following two regression models:

Wage_Standardized_t = a1+b1T+E1 ---- (1)

%Change_Wage_Standardized_t = a2+b2T+E2 ---- (2)

Now b2 < 0 (not statistically significant) and b1> 0 (statistically significant). Model-1 has an overall R2 = 0.92 and Model-2 has an overall R2 = 0.19.

Then, is it the case that, the treatment increases wages but not having the treatment would have increased wages even more? I am confused! Also, based on the R2 values, should I select one or the other?

[George Ford]

Are you trying to estimate a difference-in-difference model?

If T is just a trait of some area, then what you have is T increasing wages but not increasing the growth in wages. But if it's not 2x2DD, then there's no presumption that you've actually measured the true effect of the treatment T.

[Anupam Ghosh]

I am estimating a county and year fixed effects model for both.

[Carlo Lazzaro]

Anupam:
the most fruitful way to post questions on this forum is following the FAQ.
Therefore, why not posting what you typed and what Stata gave you back so that interesyed listers cam switch from guess-working to (hopefully) more helpful replies? Thanks.