Many bloggers are contributing to the debate on the fiscal stimulus the Bush administration is trying to push through. Let me add my grain of salt and discuss the effectiveness of this measure under various scenarios. I bypass here the question whether a stimulus is needed at all.
Given the current state of federal government finances, I take that this stimulus would be temporary, which is fairly obvious, and that it would need to be financed by further debt: it is not accompanied by a reduction in government expenses. This debt will have to be paid back at some point, through higher taxes. Thus we are in the typical case of a shifting of taxes through time, the basic premise for the Ricardian Equivalence.
The Ricardian Equivalence states that, under certain conditions, the intertemporal substitution of taxes has no impact on consumption and interest rates, and thus on other real macroeconomic aggregates except savings and government debt. Indeed, households realize after a tax cut that they will have to pay to pay higher taxes in the future, and thus save the tax cut for the future tax increase. They do so by buying the very bonds the government is issuing to finance the tax cut.
Now, this very strong theorem holds only under some stringent conditions, basically that welfare theorems hold (no distortionary taxation, complete financial markets, no monetary non-neutralities, etc.). These conditions are unlikely to hold in reality, so the empirical question is whether a tax cut does increase consumption or not (and the not includes a decrease). The literature is far from settled on this, but the two Bush administration have given several natural experiments to study this.
Case 1: Bush the elder implemented a stimulus package in 1993 that reduced the tax withheld from salaries, to be compensated by a lower tax refund the following year. A perfect experiment for the Ricardian Equivalence. According to Shapiro and Slemrod (1995), 43% percent of households spent the supplementary temporary income.
Case 2: Bush the younger issued to every US tax payer a rebate of US$300 or US$600 in 2001. Microstudies reveal that about 22% of households decided to spend it (Shapiro and Slemrod, 2001), or that between 20% and 40% of the rebate was spent (Johnson, Parker and Souleles, 2006). Looking at credit card spending, Agarwal, Liu and Souleles (2007) find again that about 40% was spent over the next nine months. For the 2003 tax cut, it impact on consumption is estimated at 50% by Coronado, Lupton and Sheiner (2005).
Thus, the Ricardian Equivalence does not seem to hold, but tax refund are clearly not the miracle measure to encourage consumption. Those households that actually spent the tax rebate where typically credit constraint: they wanted to consume more but could not for some reason, and the cash infusion relieved them from their constraint. For example, a young person, who would want to consume more but is unable to borrow against future income. Or a temporarily unemployed who, again, cannot find credit.
This means that for the fiscal stimulus to be most effective it should be directed towards those that are the most likely to spend it right away: the poor, the homeless, the students. Or the compulsive buyers. But not tax payers. It should not be a tax rebate, as those that pay taxes are those that have higher incomes, and are less likely to be financially constrained and thus consume.