Dear Statalist users,


I am Gabriele Dini, I study Economics at the University of Bologna and I'm doing the master thesis.

I want to run a Diff- in -Diff estimation.

This is the treatment:
I want to measure the effect of the so-called "80 euro tax rebate" made by Renzi's Government in 2014 on Italian households aggregate consumption.


First, I do a Propensity Score Matching to find the treated and control groups.

Then, I ran the following regression:


reg consumption eligible year14 intvar14 [pw=weight]

where:

-consumption: is the variation in consumption level between 2012 and 2014 for household "i".
I want to remark that this variable refers to panel households, for which I have data both for 2012 and 2014.

-eligible: is equal to 1 if the generic household "i" received the bonus in 2014, 0 otherwise.

-year14: is equal to 1 if year is 2014, 0 otherwise

-intvar14: is the interaction variable, that I've generated in that way: year14*eligible

-weight: are the panel weights that I found in the dataset.

Result:
I obtained an estimate for the coefficient of ye that were not statistically significant (p-value around 0.75)


So I try to add lags:

reg consumption eligible year14 intvar14 year12 intvar12 year10 intvar10 [pw=weight]

where:


-year12= 1 if year is 2012, 0 otherwise.

-year10= 1 if year is 2010, 0 otherwise.

The respective interaction variables are built as before.

However, I obtained the following result:

"year_10 omitted because of collinearity".

If I run the same regression without year14 I obtained the same result:

reg consumption eligible intvar14 year12 intvar12 year10 intvar10 [pw=weight]

"year_10 omitted because of collinearity".


I hope I show my problem clearly.

Could someone help me?

Best regards