Tax Policy Colloquium, week 7: Theodore Seto's "Preference-Shifting and the Non-Falsifiability of Optimal Tax Theory"

Yesterday's paper, by Ted Seto, could be read more narrowly as concerning how advertising affects revealed preferences relative to utility, or more broadly as concerning the general relationship that price theory and optimal tax models often assume as between revealed preferences and utility.  Here is a brief overview of the main ideas that I had upon reading the paper:

The paper starts by giving us a couple of typical supply and demand curves. In the first, we get the standard equilibrium, given where they intersect.  In the second, a tax is being imposed on the demand side, so the equilibrium quantity declines, and the incidence is split between producers and consumers.

As the paper rightly notes, in a narrow reading this set-up merely how we should think about the equilibrium, with and without the tax.  There would not necessarily be any normative consequences. But a more ambitious reading, which captures how economic reasoning is commonly used these days, would assert that the shift in the equilibrium by reason of the tax creates deadweight loss (as shown by the triangle), resulting in lost utility to some set of individuals.

Economists for many decades - especially if they went to graduate school before, say, 1970 - were quite leery about making normative statements founded on claims about utility.  Hence, for example, Samuelson's famous revealed preference theory, expressly founded on the idea that we could bloody well do without "utility," thank you.  But today's less circumspect economists have, I think, a point.  First, there is absolutely no reason to care about the lost surplus (from transactions that don't take place even though willingness to pay exceeded the reservation price), unless there are people who are made worse off by their inability to share this surplus between themselves.  Second, the underlying psychological theory on which it is based (where I get utility from own consumption plus leisure, based on rational choice under a utility function that features non-satiation and declining marginal utility), while perhaps a bit simplistic and incomplete, has in my view enough merit to be worth using as one of one's tools.

But there are certainly a bunch of problems that one ought to keep in mind when using it, especially if it is being used in relation to large normative claims (not just micro-claims about particular markets).  For example, there are issues of rational choice, the implications of which include the internalities issues that I discussed with regard to our week 4 paper, by Donald Marron, entitled "Should We Tax Unhealthy Food and Drinks?"

Among the other issues that may be important, depending on the context, are the following two:

(1) Suppose you have two states of the world in which people's utility functions are different.  The method that Seto's paper critiques can be used to define equilibrium and tax-induced deadweight loss in each state of the world, but it can't be used to compare welfare as between the two states of the world.  So if something changes the preferences (and utility) that underlie the supply and demand curves, we'll need some other sort of tool to choose between them.

(2) Social context: the standard model looks at revealed preferences, based implicitly on utility from own consumption, but it's not addressed to the question of how utility functions might arise and change.  Since we are such social beings, it's clear that social context can affect such utility a lot, in ways that might matter but that, again, we'd need different sorts of tools to evaluate.  Here are two simple examples.

First, suppose positional goods (such as material consumption) in effect create negative externalities.  E.g., as in Robert Frank's work, suppose my having a bigger house means that you now have to work harder and sacrifice leisure just to get back to the same place (i.e., having as big a house as I do).  This might support taxing positional goods, relative to non-positional goods, and again we'd need different tools to evaluate this.

Second, and conceivably going in the opposite direction, suppose there are positive externalities to shared consumption experiences - reflecting, for example, imitation or influence or mutual reinforcement. I mention this here because advertising brings it to mind.  Think of the explosion of Beatlemania in the U.S. starting in 1964.  Or for an example that I find considerably less heartwarming, think of the collective meme that Coca Cola seems to be aiming for with ad slogans such as "Coke is it!," "It's the real thing," or "I'd like to buy the world a Coke."  Now, mind you, this definitely does NOT induce me to think that we should be subsidizing Coke (or for that matter, the 1960s "British Invasion"), but if we're asking, as this paper does, what exactly commercials are doing, in cases where they clearly are not making "rational" appeals to price and quality, then we should have this sort of thing in mind.

The paper offers a model for how to think about advertising that is basically the internalities story.  (If you download the paper, which again is available here, it's Figure 11 on p. 16.)  But I think that this is often the wrong way to think about what advertising is trying to do.  It gives a nice example where that might be a plausible explanation - describing, at pp. 7-8, a case in which including a small picture of an attractive woman in a commercial loan solicitation increased loan demand by about as much as a 25% interest rate reduction.  But "Coke is it!" doesn't strike me as being about internalities.  It's doing something else, whatever that is.

Another well-known ad campaign that the paper mentions is "The Ultimate Driving Machine," for BMW.  While we might think of this as making a very soft factual claim about BMW quality, or at least as referencing prior beliefs that BMWs are high-quality cars, it also seems to be about social context.  Only, whereas for Coke the message seems to be "I drink Coke and so do you," for BMW perhaps it is more along the lines of "I have a BMW and you don't."

Here's why this might matter, in relation to Figure 11 on p. 16.  The reason I think of that as telling an internalities story, is that it contrasts two demand curves.  The first is "true" demand, reflecting underlying utility.  The second, higher demand curve, is revealed preference after one has been manipulated by the advertising campaign.  In effect, one's utility from buying the product is the same as it was before, but one is buying more of it than previously by reason of the advertising manipulation.  So it is like an internalities story, where (by hypothesis) we know how the "true" preferences related to revealed preferences, and therefore in theory can correct the behavior via a corrective tax.  (As Marron noted, if one wants to be certain that the consumers who pay the tax will generally or collectively be better off overall, one may need to find a lump sum way to rebate the tax revenues to them.)

If the Coke and BMW ad campaigns end up increasing demand for those products, I don't think we can just say that people's utility functions are the same as they were before.  These campaigns are getting right in there and affecting our utility functions, by affecting the social meaning of these products.  Plus, there's no such thing as the "unpolluted prior," comprising the "true," pre-social meaning, utility value of a given commodity.

Now, I don't comprehensively subscribe to the Gary Becker of advertising as just (or even primarily) offering useful information - although that clearly is one part of what may be happening.  And I am skeptical that the Coke and BMW campaigns are increasing net social welfare to the same degree as they (if successful) increase the producers' profitability.  In short, I am open to further evidence and debate regarding how we should think about advertising, and what the policy implications might be.  But I would view the internalities story as just one of many stories that might be going on.

Nonetheless, it's a nice paper that raises interesting issues, and I am certainly in sympathy with its broader aim, which is to say that we need to think about looking behind supply and demand curves with regard to underlying utility (or welfare or wellbeing) claims, and should not just take them as canonically or always expressing the social optimum - even leaving aside distributional issues and standard-issue externalities.