Last Thursday at the colloquium, Kirk Stark presented his paper, Bribing the States to Tax Food, along with an earlier (and recently published) companion paper, The Federal Role in State Tax Reform.
The earlier piece discusses state revenue volatility, arguably a contributing factor to the horrendous fiscal problems that a number of states and localities currently face. In fat times they tend to ramp up spending, but in lean times they may find it hard (or undesirable) to shoot back down again. An appealing normative approach might be "smoothing" (e.g., have the same teacher to child ratio in a fat year as a lean one), along with updating as one's long-term revenue expectations change but without following the business cycle as such. (And with countercyclical macro policy left to the federal level, leaving aside the political problems we've recently seen with trying to execute it there.) This is what one would expect to observe with optimal political decisionmaking plus the requisite state-level access to capital markets.
A main argument of the earlier piece is that the feds should want the states to use revenue sources that are less volatile rather than more so, since state-level volatility (especially if not handled well) may impose federal costs, including the problem of bailout that could lead to moral hazard. (One could also add the undesirability from a federal macro policy standpoint of encouraging state policy to be more procyclical.)
Against this background, Stark argues in the earlier piece that the federal government undesirably pushes states towards having more, rather than less, volatile revenue sources. For example, individuals can deduct their state and local income taxes paid, but for the most part not their state and local sales taxes. The latter tend to be less volatile, because consumption is flatter from year to year than income (since individuals to a degree smooth). Likewise, state corporate income taxes, though highly volatile, are deductible as a business expense.
In response, one could argue that it's a mistake just to assume that state-level volatility is bad. Rather, perhaps one needs to compare the social costs of federal versus state revenue volatility (at least insofar as the question is where to locate volatility, not just how much there should be overall). He notes that one could imagine thinking of federal volatility as worse. To be sure, the feds have much greater economic power, since they can tax economic activity in all 50 states and can print money. So they might be better able to handle volatility in an optimal scenario. But if they fail it's much worse than if a few states do so (a diversification point). Plus, their much larger permissible credit card balance means they can run up bigger problems. E.g., states can't implicitly default via inflation the same way they can. And while the states may have political incentives to play games with employee pensions, the feds can do this with the entire federal retirement system (Social Security & Medicare) - programs no state would be tempted to run given migration between states.
The newer, work-in-progress paper that Stark presented argues that the federal government should push states towards having comprehensive sales taxes rather than exempting food that is purchased for home consumption (rather than being prepared or restaurant food). The classic argument for exempting home-prepared food from a state sales tax is regressivity, since poorer households spend a larger percentage of their budget on it than do rich households. Problems include extreme complexity from classification issues (e.g., does it matter if you get your bagel toasted? Are chocolate chips different than Hershey's Kisses?), enormous waste of the subsidy (since rich households spend more on exempted food in absolute terms), and inefficiency from tax-favoring food consumption relative to other consumption.
Stark argues that a further cost of the widespread food exemption is increased revenue volatility. State sales taxes would be less volatile if food was generally included, since people's food consumption responds less to the business cycle than their overall consumption (reflecting that one can postpone big consumer purchases but can't decide not to eat for a while). On this basis, he suggests a large federal tax credit of some kind that would only go to taxpayers in states that taxed food at the same rate as everything else in their sales taxes.
Despite the volatility point, I wonder how much of the problem with states not taxing food is really a federalism problem, instead of simply a good versus bad tax design problem. Plus, states really should be able to handle the regressivity problem through other adjustments to their fiscal systems. It's a political economy failure, not a fiscal federalism one, that leads them not to do this. So it wasn't clear to me why it makes sense for the feds to pay the states this "bribe" for improving their tax policies. Plus, if the point of the bribe is simply to get state tax laws to improve, why direct the subsidy to poor households that would be hit by an otherwise unadjusted state sales tax change? Should the incentive be directed instead at state legislatures and their general revenues? I felt one might need to specify more fully the subsidy's purpose in order to determine its optimal design.