Do tax cuts starve or stoke the government beast?

Kevin Drum links to a Jonathan Rauch column in the Atlantic Monthly (non-subscribers can click here to read the whole thing), which summarizes William Niskanen's finding that starving the government of tax revenue doesn't starve the beast of government spending -- if anything, the trend is the exact opposite. From Rauch's story: Even during the Reagan years, Niskanen was suspicious of Starve the Beast. He thought it more likely that tax cuts, when unmatched with spending cuts, would reduce the apparent cost of government, thus stimulating rather than stunting Washington?s growth. ?You make government look cheaper than it would otherwise be,? he said recently. Suppose the federal budget is balanced at $1 trillion. Now suppose Congress reduces taxes by $200 billion without reducing spending. One result is a $200 billion deficit. Another result is that voters pay for only 80 percent of what government actually costs. Think of this as a 20 percent discount on government. As everyone knows, when you put something on sale, people buy more of it. Logically, then, tax cuts might increase the demand for government instead of reducing the supply of it. Or they might do some of each. Which is it? To the naked eye, Starve the Beast looks suspiciously counterproductive. After all, spending (as a share of the gross domestic product, the standard way to measure it) went up, not down, after Reagan cut taxes in the early 1980s; it went down, not up, after the first President Bush and President Clinton raised taxes in the early 1990s; and it went up, not down, following the Bush tax cuts early in this decade. Niskanen recently analyzed data from 1981 to 2005 and found his hunch strongly confirmed. When he performed a statistical regression that controlled for unemployment (which independently influences spending and taxes), he found, he says, ?no sign that deficits have ever acted as a constraint on spending.? To the contrary: judging by the last twenty-five years (plenty of time for a fair test), a tax cut of 1 percent of the GDP increases the rate of spending growth by about 0.15 percent of the GDP a year. A comparable tax hike reduces spending growth by the same amount. Again looking at 1981 to 2005, Niskanen then asked at what level taxes neither increase nor decrease spending. The answer: about 19 percent of the GDP. In other words, taxation above that level shrinks government, and taxation below it makes government grow.... [C]onservatives who are serious about halting or reversing the dizzying Bush-era expansion of government?if there are any such conservatives, something of an open question these days?should stop defending Bush?s tax cuts. Instead, they should be talking about raising taxes to at least 19 percent of the GDP. Voters will not shrink Big Government until they feel the pinch of its true cost. Without necessarily endorsing the "starve the beast" theory of political economy, my first reaction is to ask about lagged effects. As I've understood it, the starve the beast idea does not say that government spending will immediatekly go down as deficits rise; it argues that eventually the increase in deficits creates market and political pressure to cut government spending. My guess is that if you lagged taxes by five years you might get a different result. I see that this paper made the blog rounds a few years ago -- but it does not appear to have been published. Furthermore, the link to the original conference paper is not not working. Still, the argument is provocative enough for readers to chew on. UPDATE: Sebastian Mallaby sure seems convinced.