Here at the ASI we like taxes to be as predictable, as flat, as broad-based and as non-distortionary as possible—not to mention as low as possible. Until we've convinced everyone that we don't really need the government for most of the things it does now, we're going to need to raise revenue somehow. We want to do this in a way that reduces social welfare (and the economic activity that produces the goods to consume that produces the individual welfare that we sum to get social welfare) as little as possible.
Now we may sometimes need to use 'Pigovian' taxes—ones that discourage certain activities because they have negative outcomes on others—but most choices do not have substantial externalities. And in a society where property rights are clear and extensive, most substantial externalities will be priced in. For example, when roads are owned, their owners charge what we might call 'congestion charges': lots of problems arise only when some crucial good is un-owned and thus un-priced.
But generally we're just picking up revenue somewhere to pursue some government activity we view as worth the costs. Any non-Pigovian tax is going to reduce economic activity and welfare, but some more than others. For example, taxing investment into capital disincentivises most the activities which bring us greater productivity and wealth in the future. By contrast, if we could magically know, and tax, each individual's innate ability we wouldn't distort any decisions at all—because no decisions could change their tax liability.
The upshot of all of this is that broad-based consumption taxes are the best method of raising tax we can actually do. A 20% (or higher) tax on any good at any time leaves us as free to decide between options as no taxes, even though we have less in total to go round. By contrast any tax on capital or savings biases us in favour of current consumption over future consumption (and an income tax is partly a tax on savings).
The IFS tells us that scrapping all UK VAT exemptions would have raised £26-28bn in 2010-11 (since which we have grown substantially in real and nominal terms). In their view we could compensate everyone fully and still have £3bn left over.
A new job market paper from Bibek Adhikari at Tulane University in New Orleans takes this result further. Because VATs are usually implemented country wide, Adhikari builds 'synthetic controls'—essentially imaginary countries made up of weighted bits of other countries that didn't implement VATs—to properly test the effect of large-scale consumption taxation.
He finds that switching to consumption taxation leads to more capital invested per worker and higher total factor productivity (a measure of how good we are at using inputs), thereby raising output per head. In his words:
Five years after the reform, TFP of the treated group is 9.9 percent higher compared to the synthetic group and at its highest, the TFP of the treated group is 11.6 percent higher than the synthetic group.
So the ASI was right then!