Dear Statalist users,

I am using a triple difference-in-difference strategy to try and establish the effect of higher tuition fees on salary earned by graduates. This is using the 2012 tuition fee policy change and there are two control groups: Foreign students and Scottish universities. Therefore, only Home students at English universities qualify for the treatment effect.

I first ran a standard difference-in-difference comparing Home and Foreign students at English universities and the interaction term had a coefficient of -0.0251. But when I ran the triple difference-in-difference, I got a coefficient of 0.13489 on the triple interaction term. So not only has the sign switched, the magnitude has significantly increased. It seems strange and wrong to me that the treatment effect would increase salary by 13.5%.

I have attached my log file. I am comparing the DID regression on page 13 vs the triple DID regression on page 14.

Any help on this would be greatly appreciated!