Legal Commentary

Modern federalism, as it is practiced in U.S. courts, depends heavily on untested empirical propositions.  For instance, as everyone knows, the Supreme Court has swung back and forth a few times on whether states need some kind of judicial protection to preserve their policy autonomy, or whether instead the political process is good enough.  Similarly, the Court seems to lean more and more towards Justice Thomas' view that courts should not enforce the "dormant" commerce clause, and should leave protection of the national open market to Congress, apparently on the assumption that Congress would do a better job. 

These assumptions are testable, but until now no one has tested them.  I have.   And I find that political self-interest seems to outweigh the influence of states at least in situations when the choice between the two is presented most starkly: when Congress is deciding whether to grant a break from state taxation to a special-interest group. 

I started the study intending just to test the Justice Thomas argument (also presented in the academic liteature by folks like Marty Redish, David Super, and Ed Zelinsky): that Congress can be trusted to oversee state taxing power, and so the courts should just get out of the way.  This seemed implausible to me for two reasons: 1.  granting a *state* tax cut is costless (absent some state lobbying influence) for federal officials, and so we should expect that the feds will not fully internalize the costs of the cuts; and 2. to the extent Congress believes that there is "vertical tax competition" -- that is, if it's harder for the feds to tax if state taxes are already high -- then reducing state taxes actually benefits Congress even if there are no other rents to be earned. 

I then realized that in essence what I was looking at was an unfunded mandate.  In both "commandeering" and federal control of state taxation, the federal official can obtain political rents from a constituency without incurring any budget costs for herself.  (Many forms of preemption also have a similar political-economy dynamic, as rick h. has described).  So if I could test the tax theory, I could also test theories about whether the "political safeguards of federalism" work against commandeering and other actions that look like it. 

So what I did was find, read, and then (wtih some help from my RA's) code every federal statute Congress has ever enacted affecting state taxing power.  This took a while.  Like, two years.  Then I gathered data for some controls.  When I was done, I regressed the changes in state taxing authority on a variety of factors, including a variable indicating whether a concentrated political group stood to benefit, and another one indicating whether the state tax in question fell on an indian tribe or its property. 

Both of these, I found, were strongly statistically significant.  Conditional on enactment, if a state tax affected a special interest, congress was a lot more likely to restrict that tax.  And, if the state tax affected an indian tribe, congress was more likely to *expand* the state's authority.  There are a few stories one can tell about that, but I argue this supports my "vertical tax competition" argument: indian tribes are exempt from federal tax (arguably by the terms of the Constitution itself), and so they represent one of the few sources of income over which states and the feds do not compete.  And whaddya know?  Congress is a lot more generous when states try to tax tribes.

Interpreting these findings is complicated, and there's an extensive discussion of that in the paper.  Even if you totally buy my findings, it still doesn't necessarily follow that the "political safeguards" theory is wrong as a general matter -- I look at only a tiny sliver of all federalism-related legislation.  And even as it pertains to commandeering, all I show is that the safeguards don't work -- there are certainly arguments that that isn't necessarily a bad thing (arguments with which I largely agree).  But I think it's helpful to move on to those arguments, rather than simply slinging dueling empirical assumptions past one another. 

Comments, questions, and suggestions are very welcome.