Diff in Diff -> what is the minimum amount of treated observations?


New Member
Can I use Diff in Diff when I only have a single treated observation?
The case is that I observe the changes of a tax reform on real estate sales on the real estate prices (measured by an index). If I used the Index as my dependent variable I do not fulfill the 3rd assumption of the OLS, as there is autocorrel.
I suggested using the logged returns, which have a nice normal distribution and are not correlated. But if I use the log returns I can only mark a single observation on the date of the tax increase (quarterly data) as treated.
I get a nice significant effect, a decent R^2 and good error term plots; nevertheless I dont know whether an event study can be made with jsut 1 treated observation.

I hope you guys can help me, I am new here =):)


Less is more. Stay pure. Stay poor.
Not following. So you have data for one time period, geographic area???

Provide a little more info. please!