Wednesday, September 12, 2018

Jotwell post on recent Zucman et al paper

A website called Jotwell (for Journal of Things WE Like Lots) encourages its contributors to write very short pieces calling out for praise particular recent articles that were published in one's field.

I tend to participate in this annually, although with the press of other obligations I skipped 2017. But they have just posted this short piece that I wrote that discusses the recent Torslov, Wier, and Zucman NBER paper, The Missing Profits of Nations, which argues that big multinational firms are shifting A LOT (40% or more) of their economic profits to tax havens.

I note in my short piece that there is an ongoing dispute among leading empirical economists regarding whether the amounts being shifted are this high, or significantly lower. My own anecdotal sense of things is that the high estimates are likely to be correct, but I accept that there is a genuine empirical dispute here for the experts in such research to hash out. But in any event Zucman et al have found a creative and interesting new way of addressing the question, as I note very briefly in my piece and they explain at greater length in their paper.

As I note in my short comment, the greater the magnitude of such income-shifting, probably the less the real responses we ought to expect to, say, the recent U.S. corporate rate cut from 35% to 21%. But even if we get only minimal real responses because they do so much income-shifting anyway, it's somewhat of a separate question whether there would be more real responses if the income-shifting were more substantially shut down.

Tuesday, September 11, 2018

Hypothetical taxation of gambling

The fall semester at NYU Law School has begun, which by now feels fine (although it was kind of tough, as it always is, to start classes before Labor Day). I am teaching the introductory Income Tax class for the first time in several years. It's always fun to see (some?) students finding out that the subject has greater interest and more depth than they had expected.

Perhaps as soon as next week, we will reach the topic of how the Internal Revenue Code treats gambling losses. In brief, what the Code does is deny deductions for net gambling losses during the year. This is probably best rationalized as a proxy for the fact that people - say, gambling in a casino or at the racetrack for an evening here and there - may gamble despite expecting to lose money, viewing it as an entertainment activity. E.g., Person 1 goes to the theater at a cost of $100, because he or she likes to attend plays. Person 2 goes to the casino, expecting to lose $100 but anticipating a sufficiently fun time at the tables while this is happening. In the case where this expectation is precisely satisfied, these two cases look pretty much the same, and the tax law treats them the same by denying the deduction in both cases. (Leaving aside the issue of gambling gains on a different evening, against which the $100 gambling loss could be deducted.)

This is of course a bit of an arbitrary rule. And it has the odd implication that if, say, I bet you $100 on the outcome of the Super Bowl (and neither of us does any other gambling during the year), then as a matter of income tax law the winner has $100 of taxable income, while the loser has a nondeductible $100 loss. Plus, if I lose $200 rather than $100, than I'm actually worse off - perhaps emotionally as well as economically - yet the rule, by denying any deduction, in effect treats me as if I had enjoyed $200 of consumption value.

It strikes me that the taxation of gambling is a more interesting topic theoretically than it is as a practical matter (where rough and ready rules such as what we have today are certainly close enough for government work,). 

To illustrate a part of what I have in mind, suppose there were 3 types of gamblers, each wholly distinguishable from the other two and known both to themselves and the authorities. Suppose further that everyone was perfectly rational, given his or her preferences, and that there were no administrative issues of measurement (as well as none of identification), and also that there were no borderline or mixed-motive cases. Then it is plausible that each should be treated under a wholly separate regime, as follows.

PLEASE NOTE, HOWEVER, THAT WHAT FOLLOWS BELOW IS A THOUGHT EXPERIMENT TO TEASE OUT THE SEEMING IMPLICATIONS OF VARIOUS IDEAS - NOT AN ACTUAL PROPOSAL FOR THE FAR MESSIER REAL WORLD!

Case 1: Taxpayer rationally expects to break even, but wants to bet because taking on the risk is fun: As an example of what I have in mind here, people generally hate and try to avoid the sensation of free fall, yet they also have been known to stand on hour-long lines to ride scary rollercoasters. Suppose, analogously, that I bet $100 on the Super Bowl, despite ordinarily being risk-averse, because it will add to my excitement and pleasure in watching the game. (Or for that matter, I may bet the money on the Patriots as a psychic hedge, because I'm otherwise rooting against them.)

Here, under the strict assumptions that I am making, there is an argument for excluding both gains and losses. There is no reason to tax-penalize the activity, as between consenting adults who are in fact on equal terms. (I'm ruling out the scenario where one is a more skillful bettor than the other, hence should actually expect to win on average.)

The standard insurance argument for income (or consumption) taxation might suggest symmetrically including gains and deducting losses. Then one might raise the Domar-Musgrave point, to the effect that the bettors could offset this by scaling up the bet. E.g., if they want to bet $100, but gains are included and losses are deducted (with loss refundability if needed), they can get there anyway if their tax rates are the same. E.g., if the counterparties both face 33% marginal rates, then instead of betting $100 without tax consequences, they bet $150 with, and get to exactly the same place.

Is it unfortunate that they can do this? Not at all, under my assumptions. The income tax as insurance responds, in rational actor scenarios, purely to undesired risk (e.g., from the "ability lottery" or from having under-diversified human capital by reason of needing to specialize). But here, by assumption, people are rationally doing what they want and like. So, while their presumed ability to offset the undesired "insurance" suggests that maybe it just doesn't matter, ignoring the bet leads directly to the desired result. There is no motivation for making them adjust (even if it does no harm under the assumption that it's easily done, and indeed that in any event they are betting given the degree of mandatory insurance).

Case 2: Taxpayer rationally expects to win, and bets in order to make money: Suppose we have a card-counter in the game of 21, or else a really good poker player, who bets so as to make money, just as other people go to the office in order to earn a salary. Now we face the standard case of risky business investment, in which gains should be taxed and losses deducted (including with loss refundability). Real world loss limitation rules, such as the fact that one cannot use net operating losses to get direct federal payouts at the applicable marginal rate, arguably reflect measurement concerns - we may fear that people are creating tax shelter losses rather than reporting real economic ones. But I have ruled all that out of bounds for purposes of my hypothetical.

This regime of including/taxing gains while deducting/refunding (at the tax rate) losses provides arguably desirable insurance through the tax system in at least two senses. First, it in effect redistributes from better gamblers to worse ones, consistent with the ability lottery scenario where people differ in "wage rate" and can't insure against this privately due to adverse selection. Second, as with any other risky business investment, it provides desired insurance that might otherwise be unavailable. To illustrate this point, I used to know a card counter who went to casinos when he could spare the time, solely to make money. He hated the short-term variation, which in fact was high enough that he could go, say, from plus $35,000 to minus $20,000 (with gambler's ruin potentially looming) in the course of a few hours. He really just wanted to earn his expected return - while also needing to manage his stress over avoiding detection (which required making some deliberately bad bets, so as to throw off the watchers employed by the gambling establishment).

It's worth noting a further assumption that may be needed here to make this approach attractive. Normally we are glad that people are willing to engage in risky activity that has a net expected payoff. But in the case of gambling, the gains are other people's losses. If one thinks of this as rent-seeking or negative externalities, the approach I'm suggesting arguably is undermined. But under my assumptions, as opposed to those that it would be reasonable to apply in the real world, this is not an issue. After all, no one is systematically losing under my gambling hypotheticals unless, as I discuss next, they are deliberately (and rationally) undertaking it as a consumption activity.

Case 3: Taxpayer rationally expects to lose, but happily gambles anyway for the entertainment value: Suppose again that we have two people. The first spends $100 to go to the theater, anticipating an enjoyable show. The second spends a few hours in the casino, expecting to lose $100, likewise regarding this as a fun way to spend the evening. And suppose that the fun comes out of the process of the gambling itself - unlike in Case 1 above, let us assume that it's not from wanting to bear risk as such.

The theatergoer faces, of course, the risk that the play will prove to be a dud (and hence revealed ex post not to have been worth $100, much less a couple of hours  that one will never get back). But at least the financial cost is known in advance. One certainly could imagine the gambler thinking about the evening in much the same way, and thus regretting that in fact it's possible to lose a lot more than the expected $100 (or to have to end the night of gambling sooner than expected).

While we may be starting here to leave far behind the actual psychology behind gambling (which surely includes the hope of winning against the odds), one could, if one liked, conceptually divide the gambler's results into a "consumption component" and an "investment component."  From this standpoint, one might say that the above gambler has what ought to be a nondeductible consumption outlay, in the amount of the $100 expected cost, along with investment variation above or below that which "ought" to be deducted or included, as the case may be.

E.g., suppose I actually break even when I ought to have lost $100. Under the hypothetical approach, I would have $100 of taxable income. (After all, I'm $100 better-off than my otherwise identical peer who actually did lose exactly $100.)  Or suppose I have a rotten night and lose $200, even though I actually should have expected to lose only $100. Now I have a deductible $100 loss.

Note that, with perfect knowledge (by the gambler and the government) of the expected loss, we don't get into Domar-Musgrave adjustments here. If I change how I am actually betting, then I change the expected loss.

If one revised the treatment of Case 1 so that gains were included and losses deducted (rather than both being ignored), it would be receiving the same treatment as Case 3 (given that the expected loss in Case 1 is zero). So those two start to collapse together, once one picks at the examples a bit.

Likewise, once we see that, in Case 2, it's really the positive expected return that one might want to tax (as "ability"), one might start feeling inclined to ask whether, in Case 3, one should want to treat lousy gamblers, who rapidly accumulate large expected losses, less favorably than the more skillful (though still loss-expecting) gamblers, who are able on average to slow the bleeding, and thus to gamble less unprofitably or for longer. This might start to push us in the direction of wanting to treat really bad gamblers more favorably than good ones, e.g., by not simply benchmarking them off the larger expected losses that reflect their lower ability in this respect. This would in effect be insurance against being the sort of gambler who is bad enough at it to have a larger than typical expected loss. (And of course I am assuming that the difference here is in ability, not effort - we're in the same realm as a wage tax that discourages work if our making this adjustment discourages people from learning how to become better gamblers.)

But here, at last, is the ACTUAL takeaway that I derive from all this: Theory, at least of very simple kinds, is more tractable than reality. It's easier to say where greatly simplified hypotheticals would lead us under particular normative views, than to reach confident judgments about the real world, in which multiple, conflicting such hypotheticals may each be more than 0% true, and yet each push us in very different directions.