Tax policy colloquium, week 10: Brian Galle's “Regulation from the Inside Out: Nudges and Price Instrument Theory for Internalties and Externalities"

Yesterday at the colloquium, Brian Galle presented the above-titled paper, available (in early draft form) here.  The following is an expanded version of more cryptic notes that I had prepared for myself to help organize the discussion at the session.

Three topics. First, what is a nudge? Second, the paper’s analysis of nudges as a middle ground between subsidies & taxes. Third, internalities & the double dividend issue.

1. What is a nudge?

Paper defines them as “behaviorally informed regulation” that relies on the fact that “innocuous little speed bumps, like the nuisance of getting back up to fetch another cup of cola, or of filling out a form to start saving for retirement, can have surprising impact on individual behavior.”

But very heterogeneous:

(1) Requiring employees to opt out of, rather than opt in to, employer pension plans. (Research suggests that opt-out creates a huge increase in take-up, even though it’s easy to change the default either way.)

(2) Suppose we want more credit card use. Gas station charges $4 per gallon with credit card, $3.75 with cash. This can be described either as a “cash discount” or a “credit card fee.” People apparently hate the latter. Suppose the government bans “credit card fees” but permits “cash discounts.”

(3) Horrible picture on cigarette packages to discourage smokers (e.g., of near-death lung cancer victims. Alternatively, to encourage smoking for fiscal reasons, suppose the government required pleasant marketing photos (e.g. Bieber photos to aid in youth marketing.)

(4) Bloomberg’s proposed ban on Big Gulps, meaning that you need two trips to get 32 ounces. (This is a regulatory ban. The reason it might be considered more of a “nudge” than Social Security forced saving is that it’s easier to reverse).

SO, what are “nudges” as used in the paper or the literature?

--Not cash to/from government.

--Not a “shove,” like sending people to prison. (A 20-year jail term for armed robbery isn’t a “nudge,” because we aren’t puzzled that people might be strongly deterred by it.)

--Not a “ban” (sufficiently reversible)

--Presumably, a nudge can be either a carrot or a stick. Those are up or down from a baseline. Consider the two cigarette photos described above. Is the first one a stick, and the second one a carrot?

--Whether cash or noncash, can’t always define a given instrument as a carrot or a stick, since those descriptions are relative to a baseline. Thus, are the existing tax benefits for employee pensions a subsidy for retirement saving? Or is the rest of the income tax a penalty for non-retirement saving?

--The paper says that “a nudge is a tax.” I agree that a non-cash cost is in some ways like a cash cost (both might make the target worse-off, & prompt various types of responsive behavior). But is using one credit card framing, rather than the other, a tax? Is pension opt-out a “tax”, other than for people who want to opt out? (But in that case isn’t opt-in a tax for those who want to do that?)

Nudges aren’t defined by their attributes, but by the rationale for using them. In particular, non-cash, if a ban then there’s an easy get-around, surprisingly big response

This is going to make it very difficult to have a “unified field” analysis.

What I think we’ll have is analysis of cash vs. non-cash ways of creating cost-benefit (including Gary Becker & prison), & analysis of non-cash (perhaps cash as well?) that yields surprisingly large responses.

2. Nudges versus cash instruments

The paper’s main argument: Nudges are a middle ground between taxes (get cash) & subsidies (pay cash). Paper mainly but not entirely agrees with what I’ll call the “standard view.” Especially if raising revenue is distortionary because lump sum taxes are unavailable, then imposing non-cash costs on people often is worse than getting tax revenues, but better than paying out cash subsidies.

Consider Gary Becker on jail time vs. fines. Even if a shove rather than a nudge, say we have a choice between $1M fine & jail time disvalued by the criminal at $1M, say the behavioral consequences are the same, & ignore all other effects of jail, such as its costing money, incapacitating criminals, & either reforming or hardening them.

Under these circumstances, the fine is free money – the Treasury gets $$, everything else is the same. Better than a lump sum tax – you get revenue INSTEAD of deadweight loss (DWL).

Even if the Big Gulp ban is a nudge rather than a shove, could apply the same analysis under some assumptions. E.g., if I place a $1 disvalue having to get in line twice, why not just charge me the money? Once again, revenue instead of DWL.

Likewise, consider the horrifying cigarette photo that I disvalue at $1. Standard view notes that this is the same as Becker absent other differences.

One might also have the case of getting the same DWL either way, with or without revenue. Then a pollution tax is like a nudge plus a lump sum tax.

Finally, consider applying the standard view & subsidies. Say we can get the same behavior either by $100B/year in pension tax benefits or by changing the default. Nice to avoid the $100B handout that would need to be reversed through distortionary taxes.

OK, so what could be wrong with the standard view? To be discussed further at the session, but some initial points include the following:

(1) If the whole point is surprisingly high response, nudges may reduce not just revenue but also DWL. E.g., suppose that the Big Gulp tax would have high collection costs/ If a trivial nudge gets the same behavioral change & thus the same benefit, it reduces both revenues and DWL – possibly permitting use of a more efficient instrument.

Note also that big response may complicate the idea of utility deduced from preference satisfaction.

(2) Where surprisingly high response, we may question whether distributional effects are actually the same. E.g., hit smokers with a $1 tax or a horrible photo that gets the same behavior. Does this mean they stably disvalue it at $1? And even if so, the fact that it’s non-market rather than a cash cost means it won’t affect their ability to afford market goods.

One last point concerns income effects. The paper says we may want to give $$ to people who at the margin are producing positive externalities, take it away from those producing negative ones. But this only matters if we haven’t gotten the incentives right.

(3) Everyone knows there is more to the story. E.g., in the Becker example, incapacitation & reforming people in prison would be relevant. But this is context-specific, & it’s hard to generalize about nudges writ large.

(4) One of the biggest issues is targeting.  The "libertarian paternalism" argument that one can ignore the nudge and still get what one wants sounds backwards from a conventional tax policy perspective, as we want to raise revenue rather than lead people to incur deadweight loss by avoiding the tax.  But there can be a screening effect.  Consider two groups of people: Those whom we want to influence with the nudge (e.g., say it's about internalities), and those whom we will in fact influence because they don't opt out.  The greater the overlap between these two groups, and the less the overlap between the people who would pay the cash tax and the behavior that we'd want to discourage, the more attractive it may be to use the nudge instead of the cash instrument.  But the presence or absence of this effect presumably will always be context-specific, and it's unclear how one can generalize about it.

3. Internalities and the double dividend

Paper suggests that internalities may be very different than externalities. E.g., no double dividend problem with addressing them through tax instruments. Not sure I agree. But we need to define both internalities and the double dividend hypothesis.

Internality: per the paper, the only difference between it & a negative externality is who is being harmed. Instead of other people, it’s one’s own future self. Suppose the problem is hyperbolic discounting. You undervalue the harm to your future self when you drink Big Gulps or smoke cigarettes or don’t save for your own retirement. Yet yesterday you would have wanted to choose properly between today and the future.

Internalities differ from externalities in that, for them to matter, they require departing from a rational choice framework. Whereas externalities could be examined purely within such a framework. But when we use nudges to address externalities, we have likewise departed from that framework.

So, it’s not obvious to me that internalities & externalities should function as differently as Brian suggests. Either way, someone’s welfare is being ignored in the decision-maker’s calculus, & we’re trying to exploit individual irrationalities or discontinuities.

OK, on to the double dividend hypothesis. Say we have a carbon tax. We improve incentives AND raise revenue. Since the latter usually requires DWL, we’re doubly blessed. Does this suggest that, in today’s U.S. economy we can layer a carbon tax on top of the existing income tax and not discourage economic growth, etc.? (In the steady state – this is not about Keynes and the business cycle.)

If you have no taxes, need revenue, & can choose a Pigovian tax in lieu of enacting something distortionary, clearly a double dividend. So what’s missing?

Let’s switch subjects & then circle back. Optimal commodity tax literature, inverse elasticity rule. It turns out that if you are taxing all commodities except leisure, meaning that you have an “income” tax (no time in this model, so it’s the same as a consumption tax or a wage tax), and if you grant an assumption called separability of work & leisure in people’s utility functions, there is nothing to gain from commodity tax differentiation EXCEPT for leisure complements vs. work complements.

To offset work discouragement, you may want to raise the tax on leisure complements, lower it on work complements), thus reducing the work-leisure problem, albeit at the price of distorting commodity choice.

An example: tax food that’s cooked at home (leisure complement, since you have the time), subsidize restaurant food (work complement). Of course, that’s the reverse of what real world RSTs do.

OK, suppose we grant that not having a carbon tax is effectively a work subsidy (like exempting work complements). The same line of argument suggests that this is partly a good thing in efficiency terms.

Now we create a carbon tax at the right level. It’s a Pigouvian tax that improves incentives at that margin, but it also raises the tax rate on work. So that reduces efficiency. Still worth doing, but no longer free money in efficiency or economic growth terms.

What does this mean for internalities? Take the Big Gulp rule. In a rational choice scenario, the ban on Big Gulps does indeed make work (earning $$ to buy a 16 oz drink) less valuable. Unless we think the key is whether you have time for extra waiting on line, in which case it might be hitting leisure.

But why would the labor supply effects on you depend on whether our rationale is based on the internality or the externality? What matters is the relationship between the work-leisure choice and the item that’s being addressed.

Likewise, suppose we wanted to nudge people towards eating in restaurants more. Because that also happens to be a work complement, it would be double dividend plus.

Now suppose instead that we want to nudge them towards eating at home more. Then we would be facing the problem of increased work discouragement.

But either way, it wouldn’t depend on whether the nudge was motivated by an internality or an externality.

OK, if we weaken consistent rational choice, all bets are off. Now the key is how it plays through perception regarding work-leisure choices. But while this creates an imponderable, it’s still not about externality vs. internality as such.

SO: Is there a general link between internalities and double dividends? Or does it depend on other stuff, both for externalities and internalities?