Do anticipated tax changes have an impact? Theory tells us they should: as households anticipate future tax liabilities, which reduce their permanent income. Thus one should see changes in consumption. Yet, there are plenty of microeconomic studies that find no significant effect of anticipated tax changes on consumption. What about output?
Karel Mertens and Morten Ravn avoid the usual VAR approach for the identification of anticipated and unanticipated tax changes. For the US, they use the Romer and Romer narratives to identify exogenous federal tax changes, and then measure the time between the signature of the tax bill and its implementation date, and if it is more than 90 days deem the tax change to be anticipated. While unanticipated tax cuts increase output and consumption, with some delay, anticipated and not yet implemented ones decrease output and have no impact of consumption. And the longer the implementation lag is, the deeper the drop in output is. And when was this particular important? When Reagan announced tax cuts and sent the US economy into recession.
Tax here is a broad word, as it encompasses all types in taxation. Of course, a tax cut on labor income is going to have a different impact than a tax cut on dividend income. Of particular interest here is that the equation they estimate is derived from a DSGE model, and they tested that it correctly measures the coefficients even in short samples. That exercise assumes that tax changes are measured by changes in tax liabilities, wherever they may be (social security, income tax or gasoline excise tax, for example). They can have very different effects depending on their nature. So it is not clear how this translates back into a model. It would be interesting to repeat the exercise by differentiating by type of tax.