Tax policy colloquium, week 7: Zach Liscow, part 2

My prior blogpost offered some background regarding Zach Liscow’s “Democratic Law and Economics.”  Liscow has been in the forefront among those questioning the merits of following the “double distortion” line of argument to conclude that “legal rules” or other regulatory policy should respond purely to efficiency concerns, leaving distribution to be handled by the “tax system.”

While earlier work by Liscow and others (such as Sanchirico) has challenged the accuracy and completeness of the assumptions that underlie the admonition that distributional issues be ignored outside the “tax” realm, here he accepts the analysis, at least arguendo, but says: What if following it leads to too little redistribution because voters, while not otherwise averse to it, really dislike cash transfers? (This is the demogrant side of the Mirrlees tax model.) While the paper’s formal model defines this as a universal aversion among voters to cash transfers (held even by poor people who would receive the transfers), its textual discussion invokes beliefs about entitlement to pre-tax market income. So we might think of it informally as concerning the higher tax rates that are needed to fund demogrants, rather than about the demogrants themselves.

The paper further posits that this is not generalized anti-redistributive sentiment, but merely reflects “policy mental accounts.” This draws on the behavioral economics insight that how an individual chooses to spend a given dollar may reflect which pot of money or transaction he assigns it to – leading to departures from consistent rational choice, although perhaps understandable as a heuristic or rough rule of thumb to guide choice.

This in turn implies that voters (presumed to influence policy outcomes) who oppose high tax rates to fund large demogrants might be perfectly happy with redistribution accomplished by different means. Perhaps one might think of this as involving the endowment effect on the tax side (i.e., coding precluded market returns differently than those that were first earned then taxed), plus greater tolerance of in-kind than cash benefits on the benefit side.

The paper therefore posits that inefficient redistribution through legal rules might be an overall policy improvement if there is space for it, but not for the first-best of doing it the Kaplow-Shavell way.

Here is a very simple example that I think can be used to help illustrate the paper’s analysis. It’s taken from one of the central cases discussed in the paper, but here I spell it out a bit more.

Suppose the Department of Transportation (DOT) is deciding whether to spend $$ saving a rich person an hour of travel time (via airport upgrades), or a poor person the same hour (via mass transit upgrades). Suppose further that, based on willingness to pay, the rich person values the hour saved at $63, and the poor person at only $25. (The paper derives this from actual data noted in the paper.

OPTION 1, spending the money on mass transit, benefits the poor person by $25 and the rich person by zero.

OPTION 2, spending the money on airports, benefits the poor person by zero and the rich person by $63.

Cost-benefit analysis, as done at the DOT and elsewhere, commonly uses willingness to pay to discern value. So the “efficient” choice is Option 2, spending the money to help rich people because they place greater value on their time. Instead choosing Option 1, e.g., based on valuing people’s time equally and then using benefit to the poor as a tiebreaker, is inconsistent with the view that only the tax system should consider distributional issues.

Let’s now further strengthen the case for Option 2. Using it in lieu of Option 1, but with the addition of a cash transfer from the rich taxpayer to the poor taxpayer, can create a Pareto improvement relative to choosing Option 1.

Again, under Option 1 the parties gain 25 (poor) and 0 (rich).

Under Option 2, they gain 0 (poor) and 63 (rich).

Suppose we adopt Option 2 but the rich person pays the poor person anywhere between $26 and $62 in cash.

Under Option 3a (Option 2 plus a $26 side payment,) they gain 26 (poor) and 37 (rich).

Under Option 3b (Option 2 plus a $62 side payment), they gain $62 (poor) and $1 rich).

Both of these options are Pareto-superior vs. Option 1. So, while this is not exactly the double distortion argument in action, it supports the same conclusion: Do the most efficient thing possible outside the tax system (using willingness to pay as people’s own measure of utility effects on them), and then, with the economic pie having been made as large as possible, use tax-funded cash grants to create a Pareto improvement relative to the case where we used inefficient legal rules to address distributional concerns.

This is a highly stylized and simplified example. But it’s useful to illustrate the line of argument in the Liscow paper. In effect, he accepts the entire thing at least for argument’s sake, but adds a political economy constraint: Suppose that in practice Option 3a or b would happen, in a mass society as opposed to one with just one rich and poor person negotiating, only via higher labor income taxes to fund larger demogrants. And suppose that aversion to high taxes or cash grants means that 3a and b simply won’t happen. So our only choices are Option 1 or Option 2.

Suppose further that, due to other aspects of voter belief systems, they’d be fine with selecting Option 1 – for example, based on the belief that people’s time should count equally and that tiebreakers favoring the poor are okay. But if the regulators believe that only efficiency should drive non-tax decisions, we’ll get Option 2.

In effect, the paper argues that Option 1 might actually be better than Option 2, if we assume both (a) that there is too little redistribution overall due to mental accounting rule disparaging high tax rates and cash grants, and (b) that there will be no marginal redistributive effects to the choice of Option 2 over Option 1. (In effect, nothing will happen towards implementing Option 3 variants.) So the DOT should employ distributional analysis, rather than purely efficiency-driven cost-benefit analysis, in the course of deciding whether it’s better to implement Option 1 or Option 2.

Choosing Option 1 might be here viewed as a standard “leaky bucket” problem in redistribution. The rich lose $63 while the poor gain $25, causing the analysis to depend at least in part on the marginal utility of these values at the applicable income levels. And again, the fact that one might have been able to use a less leaky bucket, if the public didn’t object to the standard optimal tax model, is ruled out of bounds as politically unavailable.

 The if-then logic of the paper is unassailable. It’s a basic second-best thing, aka, the best shouldn’t be the enemy of the better-than-nothing. If there are two paths to addressing distributional concerns, and the better one is unavailable in circumstances where the worse one might be available, then of course one shouldn’t rule out the latter, but should duly consider it.

The harder and more interesting question concerns whether and to what extent it might have significant policy relevance. Here are some quick thoughts about that:

1) Assuming voter control, or positing a fixable asymmetry? The paper posits that voter influence over political outcomes makes it relevant that people have inconsistent views, such that they might dislike redistribution done via taxes and cash benefits, but be fine with it when done by means that a welfarist with an advanced economic understanding of policy instruments might deem clearly inferior. The posited set of viewpoints strikes me as clearly plausible. The assumption that voters actually influence political outcomes sufficiently strikes me as less so. There are well-known studies by the likes of Martin Gilens, Larry Bartels, Benjamin Page, etc., suggesting that the policy views of the 99% have startlingly little influence on actual policy choices in Washington.

However, there is a different reason why the paper’s line of argument might be politically efficacious. Tax policymaking in Washington occurs in a highly politically charged realm in which the players are only marginally subject to influence by what people in the academic and think tank realms are saying. (An example of such influence, however, might be recent academic work by the likes of Diamond, Saez, and Zucman pushing out the Overton window so that 70 percent top bracket income tax rates, along with the use of wealth taxes or similar instruments, are now considered more plausible than they were previously.)

But regulatory policy et al is potentially subject to area-specific influence by specialists and experts, who might even have some discretion despite any political overlords from the Executive Branch or Congress who have the power to rein them in. If they have been thinking that the regulatory process should look solely at efficiency, because that is the climate of intellectual thought under which they have been trained (whether or not they are actually familiar with Atkinson-Stiglitz or Kaplow-Shavell), then it’s not impossible that suasion to the effect that distributional considerations should count here too might affect their judgments.

In other words, one could claim in support of the efficacy of the Liscow paper’s project that it’s addressing an asymmetry, in which the tax realm doesn’t much follow optimal tax theory recommendations re. what it should do, but the regulatory realm does follow the prescription that one should leave all distributional issues to the tax system. Moving towards distribution-conscious cost-benefit analysis might conceivably make a difference here, subject to the “political general equilibrium” question of how this will actually play out in the end overall.

2) General equilibrium political playout: ‘political Coase theorem” versus the baseball game metaphor – As the Liscow paper concedes, the distribution-conscious approach that it urges for regulatory policymaking might not matter after all if what David Weisbach, among others, has dubbed the “political Coase theorem” might apply at the end of the day.

As background, the actual Coase Theorem that’s being invoked here holds that, if transaction costs are zero, it will make no allocative difference – although it might make a distributive difference – whether, say, (a) I have a right to pollute unless you pay me to stop, or (b) you have a right to stop me from polluting unless I pay you to let me do it. Either way, with zero transaction costs it “doesn’t matter” – in terms of allocative outcomes – which way one allocates the initial right. The idea is that the higher-valuing user will end up possessing the right. E.g., if I value polluting at $10 and you disvalue it at $12, then either (a) you’ll pay me between $10 and $12 to forbear if you have the initial entitlement, or (b) I’ll ascertain that I can’t buy the right to pollute from you at the max I’d be willing to pay ($10). So either way, the pollution doesn’t happen. (Of course, the Coase Theorem’s main message actually is that transaction costs are why it might matter who gets the right – not that it generally doesn’t matter.)

Here are two versions of what proponents have called the “political Coase theorem,” adapted to my earlier example with the choice between mass transit and airport expenditure to reduce either a poor or a rich person’s travel time.

Version 1: if the poor person has the power to get mass transit spending that she values at $25 agreed to, in lieu of airport spending that the rich person values at $63, the latter offers the former between $26 and $62 to agree to the latter. So the latter, rather than the former, ends up happening.

As adapted to more real world regulatory choices, Weisbach has noted the possibility that groups potentially subject to inefficient redistribution have an incentive to offer a Pareto deal in which the redistribution is instead done efficiently. This creates surplus that all can share, so one might ask: Why doesn’t this just happen? (In that case, the power to threaten inefficient regulation might matter, but one wouldn’t expect actually to observe it.)

The answer to the question “Why won’t that just happen?” seems pretty clear. As per the Coase Theorem in its standard application, what about transaction costs? Inertia, information costs, disaggregated political power so that different principals cover different policy areas and can’t readily trade with each other, etc., are important enough (I’d argue) that we shouldn’t simply presume that this trade is the ordinary course of things. Sure, it’s a relevant consideration, but if anything the presumption might often lie in the other direction (Why would it be able to happen?)

Version 2: If Congress has specific distributional goals that it pursues coherently and consistently, then in a sense it really won’t care what the regulators do. Or more precisely, even if it doesn’t directly rein them in, it will simply adjust its distributional bottom line so that distribution comes out in aggregate the same as if the regulators had pursued efficiency alone.

I think hardened law and economics types may be prone to finding this line of reasoning more persuasive than it actually is, because they are used to thinking about consistent rational choice by an individual with coherent preferences. But in politics you get all the issues of collective choice, along with pervasive agency costs that include political actors’ frequently greater interest in such things as personal credit-claiming, blame avoidance, and symbolic gesturing, than in substantive outcomes. Thus, even insofar as individuals fulfill the rational actor model of optimization under coherent and consistently followed preferences, collective choice institutions in a modern mass society should not be expected to do this.

Once again, while obviously one has to think about the possibility that Congress will undo (or directly rein in) distributionally minded agencies that are not adhering to pure efficiency (as well as those that ARE adhering to pure efficiency), there is really no reason here for a general presumption that it just won’t matter. That, rather, is the question to be asked.

Here is a model I prefer to the “political Coase theorem” for thinking about why, say, the left or the right might pursue particular distributional (or other) fights as zealously as they sometimes do. Each time you win a battle, you’re that one battle ahead, and it won’t necessarily be offset elsewhere even if outcomes aren’t entirely independent and uncorrelated.

Suppose a baseball team figures to win about half of its games. Bottom of the ninth with two outs, they’re down by one run but have the tying and winning runs in scoring position. So if the batter gets a hit they win, if he makes an out they lose.

Either way, they’re still a .500 team over the long run. But they’re one game ahead if he gets a game-winning hit, relative to the case where he makes the third out. And there’s no particular reason to think that this will be automatically offset. A win today doesn’t, at least inherently, make a loss tomorrow more likely than it would otherwise have been.