Taxes vs subsidies, flows net and gross

A straightforward objection to the style of policy I proposed as “market dirigisme” in the previous post is that it’s too expensive for the central government to pay to subsidize all the behavior it wants from the public. As Peter Dorman writes in the comments

With so many objectives to pursue, if policy mainly took the form of bribery — excuse me, incentives — the state would be overwhelmed. It’s necessary to have negative incentives, taxes, as well as positive ones, feebate systems for instance.

The taxation-based complement of my market dirigisme is just Pigouvian taxation. Instead of purchasing from the citizenry the behavior the central government desires, the central government can penalize citizens who fail to behave as the central government prefers. That seems “cheaper”, but also meaner and more contentious. People generally don’t mind being offered an opportunity — “Hey! I’ll pay you if you do this thing. You don’t have to, but if it’s worth the money to you you can!” People object vociferously to, and organize politically against, state coercion that takes the form of “You should do this thing, and if you don’t we will punish and fine you.”

So it’s usually politically wise, I think, to prefer the former. Subsidize the behavior you want.

Revenue-motivated taxation should be broad-based and imposed gently on things mostly coded as good (like income). Pigouvian taxes, to be politically sustainable I think, require a strong consensus among the influence-weighted citizenry that the thing to be taxed is “bad”, that people, at an individual level, deserve to be punished or at least to compensate society for the behavior. Matt Yglesias has been talking up alcohol taxes, and that’s probably fine as a Pigouvian tax, because most of us have been persuaded at some level that drinking is a vice and even if we ourselves enjoy it responsibly, it’s fair that we should make some compensation to society for indulging. Gas taxes or carbon taxes are not fine. It is not a vice, at an individual level, for a person to drive to the grocery store or commute to work, even if they live far away. It’s not a vice for a person to live in an exurban or rural place where ordinary life requires long drives. On the contrary, many of us consider rural living an outright virtue. Yes, these choices have social costs, “externalities” in the language of Pigouish economists. But unless those social costs have been pervasively internalized as sin by the broad public (no, your own politically engaged friend group isn’t enough), Pigouvian taxation is “partial equilibrium“. A polity won’t allow large, enfranchised publics to be selectively penalized for behavior they think is fine. Michael Bloomberg can get a soda tax passed maybe in snobby elitist towns whose enfranchised publics look down paternalistically on frequent pop drinkers. But that’s pushing the limits. I’m all for a (refunded!) carbon tax, higher gas taxes, etc. I, um, voted for the proposed soda tax in San Francisco. But, realistically, it’s not a smart approach. You’re not going to save the planet by punishing large, influential blocks of citizens for behavior intrinsic to what they see as legitimate ways of living.

So, what about the expense? Are subsidies to encourage desirable changes by the citizenry affordable? Yes, absolutely. There are two broad points to make here. First, the real cost of such subsidies is lower than it appears. As always, we should be wary of a dollar-for-dollar accounting frame when considering the costs of public policy. Public expenditures are not alike in their effect on price stability, distribution, and other social desiderata. Secondly, to the degree that we are forced to ration net outlays, balancing conditional subsidies with broad-based taxes can be made equivalent to Pigouvian taxes fiscally, while retaining the benefits of relying on subsidy politically.

Subsides of citizen behavior are cheaper than they look for a bunch of reasons. Very narrowly, subsidies are not expenses at all, they are only transfers. They do not directly recruit or put pressure on real resources. To the degree the state is conditioning subsidies in order to “purchase” something from the public, that isn’t quite right, but usually the behavior that will be purchased displaces minimal alternative uses of real resources, or even increases the availability of such resources, making them negative expenditures in a sense.

Let’s unpack that. If the state hires people to build dense housing, that imposes a direct real-resource cost. People who otherwise would do different work are diverted to building these homes. Brick and timber and steel that otherwise would go elsewhere go here. In a broad sense, we can argue that the dense housing net-frees resources, because the people who move in might otherwise have lived in resource-costly exurbs. But that’s a complicated, very contestable, calculation. However, if the state subsidizes people who live in densifying neighborhoods, and the voters in some communities encourage construction they otherwise would have forbidden, very little of the recruitment of real resources should be attributed to the state. The people the state pays with its subsidy are not hired away from their old jobs or withdrawn from the labor market. The choices about which real resources will be recruited and deployed to the now popular densification will remain with the local governments, communities, and developers, displacing their own alternative uses. Mostly what will have happened is we’ll have (i) gotten the policy outcome we might have wanted but failed to achieve if we tried to induce local governments to change their laws despite the resistance of their citizenry; and (ii) made a very-broad based, rather than targeted, transfer to citizens of that community.

In distributional terms, broad based-transfers are often scored as “expensive” relative to counterfactual programs targeting the needy (although when the cost of attrition is taken into account, they shouldn’t be). But relative to ordinary expenditures, broad-based transfers are desirably progressive. When the government hires a contractor to do something, some of the funds may go to modest wage workers, but lots also go to well-paid professionals or flow through as capital income to business owners. None go to people who are neither investors or workers. Business revenue from government expenditures is distributed like market income generally. To the degree a more egalitarian distribution is desirable, broad-based subsidies are “cheaper” than ordinary expenditure in the sense that they incur less of the social cost that comes with making the rich richer.

Finally, on a per capita basis, broad-based subsidies intended to alter citizen preferences with respect to local governance can be cheap even in accounting terms. Most voters are not rich, a little bit of money can make a big difference relative to the costs and benefits they perceive from local policy and the convenience costs of voting. Often when you want to get something done, you have to purchase things from rich people who control access to scarce resources or capabilities, for whom it takes a lot of money to move the needle. But when the state wants to purchase changes in behavior or political preference from the citizenry, most people are very far from rich, and small amounts of money can go a long way. (The current vaccine lotteries are a great example. The per-person cost of the lotteries is pretty tiny, yet they have seem to be effective.)

However inexpensive subsidies are in the senses I’ve described, in some contexts and for some purposes, fiscal limits in an accounting sense will sometimes bind, politically or institutionally. These cases should be infrequent. Well-arranged governments don’t face purely financial constraints on socially valuable investments, just like firms in a well-arranged financial system can always issue securities to fund high NPV investments. But sometimes, pathologically, firms are subject to hard capital rationing, and sometimes governments may have to tax in order to subsidize. In these cases, you use the ordinary tax system to ensure you have the revenue you will need to purchase changes of behavior from the public, just like any other public expenditure. We have property taxes, income taxes, sales taxes. If we want to pay citizens of neighborhoods that tolerate densification and our expenditures are revenue constrained, we can raise those other taxes. The tax increases should not take the form of matched “pay-fors”. That’s a very bad style of public policy. But financially constrained governments should use the tax system to ensure the revenue they require for public purposes, including purchasing behavior from the general public.

A certain kind of wonk will object and say that this is really dumb. Why tax everybody and pay some people a subsidy when you could more “efficiently” impose the same net financial flows by levying a tax on the people who don’t do what you want?

This is a point I’ve made before and I suspect I will make again. There is social meaning, and political effect, in gross financial flows. An income-tax-financed UBI is not the same thing as a negative income tax even when the net financial effect of the two would be the same. Besides the stuff wonks will deign to recognize (like that paying first and taxing back later ensures more certain and prompt payment to those who will ultimately be entitled), different program forms create different social facts. A UBI informs the public in the straightforward language of cash that we are all together receiving a benefit, for which we are paying according to our means. A negative income tax shouts that the poor get a benefit for being poor that the rich must pay for. How the public understands a policy is part of the policy, inseparable from its effects on behavior, its political sustainability, and its ultimate effectiveness. Gross flows condition the meaning of a policy to the public, in addition to net outcomes.

Sure, increasing general taxes in part to pay for conditional subsidies to the public can be equivalent in net terms to just imposing a conditional tax. But imposing a conditional tax has the social meaning that the payer is being punished or asked to compensate for doing wrong, and is politically unsupportable when influential publics think the taxed behavior is legitimate or virtuous. Offering a subsidy communicates no such judgment. I don’t feel judged by the state for not driving a Tesla, even though had I purchased one, I’d probably have been eligible for some government cash. An offer of a subsidy is a market bid, an opportunity. You can take it or leave it, but the option is no injury. The general schedules of taxes on property, income, and sales are… schedules of taxes that apply broadly to everyone and carry no judgment about the lifestyles or choices of segments of the public. Combining those two elements, a general tax schedule and a lucrative offer, yields an entirely different social animal than a tax targeted and conditioned on behavior, even when from a high level you could argue the net financial flows would be similar.


6 Responses to “Taxes vs subsidies, flows net and gross”

  1. Bolt writes:

    So what’s the market dirigisme version of the soda tax, a subsidy for diet sodas? For housing there is a dimension (density) of a single good (housing) that we want to incentivize but for the alcohol/soda taxes there isn’t a clear substitute good to prefer. If there is something extremely specific to penalize, a Pigouvian tax makes more sense than a broad subsidy of every alternative.

    It’s a similar situation with a carbon tax versus a subsidy for alternative energy – on one hand the government is picking winners, on the other they are just saying “find something else, and it will be cheaper”.

  2. T writes:

    > The tax increases should not take the form of matched “pay-fors”. That’s a very bad style of public policy.

    Can you clarify what you mean by this? Are you referring to earmarking taxes? It seems intuitive that earmarks could be used to tie together incentives and penalties in a way that is acceptable to be public – soda taxes towards healthcare, carbon tax towards green energy subsidies, income taxes to basic income, etc. Why is it a bad style of public policy?

  3. Alfonso writes:

    Thanks for the post Steve :)

    People who study animal learning think a lot about shaping behaviour using positive versus negative reinforcement. Positive reinforcement just works better (i.e., it is more successful in shaping behavior).

    I think an interesting counter-example to this pattern is young children. Of course positive feedback is useful with kids too :), but one is forced to use punishments sometimes because what they are trying to learn is the limits of acceptable behavior, i.e., the meaning of “social safety”. In this case, unless actively prevented to do so, they will keep exploring the action space to see how far one can get.

    The key difference is whether behavior can be construed as “homeostatic” (i.e., geared towards maximizing/satisficing some small set of parameters). In this case reward is generally more useful than punishment because, whereas both provide information about the cost function, reward promotes motivation and does not promote fear (in humans, the kind fo punishments we’re discussing will, as Steve notes, promote anger if the agent does not feel like she “deserves” the punishment).
    With kids, actions during this kind of exploration are not means to a (homeostatic) end. The end is understanding which actions will derive punishment, so negative feedback is (also) needed.

  4. Geoff writes:

    It may be that one of the benefits of this approach is to wean us from very narrowly-focused (and hence apparently punitive) taxes such as the soda tax. The Pigouvian tax on soda was presumably not a full representation of the beliefs of its designers, it was merely the simplest implementation of a the belief that processed sugar has negative health care externalities or addictive properties. As such, it both appears punitive towards soda drinkers and invokes boundary cases around e.g. extremely sweet coffee drinks that have the same or worse negative impact. Perhaps more importantly, it incentivizes border consumers to purchase outside of the region that introduces the law, and thus has an unintended negative effect on small businesses that happen to fall within the law’s boundaries.

    We see this very directly in Chicago: “We all know what happened five minutes after this went into effect, consumers started taking pictures of their receipts, started sending them to news outlets like yours, and they started writing letters to the editor, they started calling their elected officials all on their own, they didn’t need any prompting,” said Rob Karr, president of the Illinois Retail Merchants Association.

    In a rewards-based model designers should restart from their original beliefs. Subsidizing healthy foods is a goal on par with penalizing unhealthy ones – of course there will be edge cases, some of which can be blurred by scaled subsidies. The big win is that both consumers and businesses in the law’s area are natural allies rather than natural enemies of this law; Chicago businesses near the borders of the affected area are likely to see an increase rather than a decrease in their sales, and those on the outside of the border will be asking to get in rather than asking to get out.

  5. N writes:

    I don’t feel like you really answered the original objection. Of course it’s more pleasant for the dirigiste if they can offer incentives and positive feedback, but that’s the easy part of governing. Is there any evidence that this is sufficient? What happens when people don’t pay their Pigouvian taxes? As a local counterexample, “positive feedback only” is a cherished belief among dog owners in SF, with negative externalities for the whole city.
    It sounds like you’re squeamish about policing, which is understandable in this day and age. What sort of incentive are you going to use to keep Walgreens open? Do you compensate them for their shoplifting losses because prosecuting shoplifters is regressive?

  6. Detroit Dan writes:

    This is a thought-provoking post and that is reflected in the comments. My favorite tool with regard to fiscal deficits and control of inflation is the use of automatic fiscal stabilizers, such as income taxes and welfare benefits. Income taxes collected automatically rise in times of inflation. Welfare spending automatically rises in times of high unemployment (and vice versa).

    I’m generally in favor of Pigouvian taxes although good countervailing points are made in both the original post and the comments. A carbon tax seems reasonable given the clear and overwhelming danger posed by the climate change based upon the burning of fossil fuels. However, the case against the soda tax as described by Geoff is convincing (as is the case for incentives for healthy food in Chicago). I also think Alfonso and N are onto something in arguing that in some cases negative incentives (e.g. targeted taxes) are useful / necessary.

    Again, good constructive food for thought…