tax changes have very large effects on output. Our baseline specification suggests that an exogenous tax increase of one percent of GDP lowers real GDP by roughly three percent.
Hmmm. If the tax rate is increased from 20% to 21%, GDP falls by 3% so tax revenues still rise. But Romer and Romer were not describing a supply-side effect. If Mankiw had provided even just the entire paragraph, we would see:
The strong negative relationship between tax changes and investment also helps to explain the size of our estimated overall effect on output. Recall that we find that a tax increase of one percent of GDP lowers real GDP by about 3 percent, implying a substantial multiplier. An important part of that effect appears to be due to the procyclical behavior of investment.
But read the entire paper for yourself!