Just in case some people doubt it, I can actually back up that claim. In a recent working paper, Guay and Normandin (2019) show that for the US since WWII, you can unambiguously pin down the effect of tax cuts in the US, but they also show that the effect of spending hikes depend on the assumption you make. They use what we call a structural VAR approach. You begin with a simple statistical model which says the current value of a group of macroeconomic variables depend on their recent past and random innovations. Then, you assume this statistical description of the economy is related to the actual causes that gave rise to the observations we make. Assuming the causal model (or structural model, hence the name structural VAR) has the same form as your statistical model, you have a chance to recover it if you can find a clever way to relate the structural innovations (the random events to which people actually react) to the statistical innovations (which are proxied by the statistical errors you get after estimating your model). Usually, we assume each statistical error is a weighted sum of structural errors -- i.e., the innovations in the data combine many innovations simultaneously. From here, the methods in the SVAR literature consists in imposing just the right amounts of restrictions to disentangle the statistical innovations and, thus, recover the structural innovations -- i.e., the so-called macroeconomic shocks, such as policy changes.
Now, what Guay and Normandin propose is that you can forgo SOME of this effort if at least one of the underlying innovation is not Gaussian: specifically, if they exhibit skewness or kurtosis, you can use that to your advantage. The reason is that skewness and kurtosis imply errors will "draw" a shape that is impossible to draw with a normally distributed variable. A simple rank test on the right matrices gets you the answer. If you follow the argument and apply it to their example, you find that no assumption structure that is consistent with the data is going to get you tax cuts that are not expansionary. In the language above, statistics alone suffice to disentangle that effect. However, it is not the case for spending hikes. We have a hard time pinning down the spending multiplier and it can go from slightly negative, all the way to above one, if I recall, depending on what we assume.
So, the nuance here is that we're more certain of what happens when taxes cause a deficit hike than when spending causes a deficit hike, at least as far as the post-war US is concerned.
