Taylor's work has always struck me as strangely naive - essentially just looking at output or consumption and looking at policy and noting whether there was a change or not. I've always figured I had to be missing something. But apparently that's really about all there is to it - and when Christina Romer raise endogeneity concerns he just says "I controlled for oil prices".
The point, Dr. Taylor, is that stimulus is a function of certain variables (not oil prices so much) that also influence output, consumption, or whatever it is you happen to be looking at. The real concern is those omitted variables, and your safest bet is looking for exogenous variation in your fiscal policy variable, not trying to track down a bunch of controls.
Romer cites the best literature to get this point across: the returns to education literature. She also praises the work of labor economists who do this stuff. I learned econometrics from a labor economist, and it's paid off big time. I think all macroeconomists should.