big standard error

[lisa carnicelli]

hello, i have a quick question. I am running a DID regression on monthly data that have been deseasonalised and detrended. for ecah variable I have 120 observations 60 before and 60 after. When I run the regression I get huge standard errors and coefficients. My model is the following : y=b0 + post + treated + post*treated
post: 1 if after legislation 0 before.
treated : 1 if treated 0 if before.
i tried adding the table but for some reason I cannot...
I get for:
post : coeff = -11.0887 st error =145.7597
treated: coeff = 41.82148 st error =311.9048
did: coeff = -87.41964 st error =475.5051
const: coeff =13.90112 st error = 94.761
could this be due to the fact that I have a small sample size ? and how can I fix it ?

[Richard Williams]

Sure, a small sample size doesn't help. But it could also be that your theory is bad or the treatment is ineffective or you have a great deal of multicollinearity. You've deseasonalized and detrended and maybe you did something wrong or maybe that wipes out effects.

I would suggest simplifying the model. Do any variables have significant bivariate correlations with Y? Are their problems with measurement, and if so is there anything you can do about, e.g. use other measures. If there is any way to get more data then by all means do, but you may not have much control over that.

As for posting output see pt. 12 in the FAQ. It is easy to do but harder to explain. It would be nice if there were a short video on this.