A misguided attack on Land Value Taxes

The idea of a Land Value Tax (LVT) is to tax the value of land independently of the value of improvements on that land (e.g. buildings, farms, or mines). Separating the value of a plot of land from the value of the structure built on top of it is a very difficult thing to do, since you can't usually observe the value of a piece of land both before and after the improvement is made. This implementation issue is the main problem with the LVT.

Periodically, people make criticisms of the LVT, and they usually boil down to this measurement problem. For example, Zac Gochenour and Bryan Caplan go to great lengths to show that a tax on the value of unimproved land reduces the incentive to search for better improvements. But under a true LVT, improvements would receive a tax credit, which would remove this problem entirely - if, of course, you can measure the value of the improvement. (Actually, in the case Gochenour and Caplan describe, the measurement of the value of the improvement would actually be easier than usual, since you could do a before/after observation.)

Adam Ozimek of Forbes has another argument against the LVT, which he claims doesn't boil down to the measurement problem. But I think his argument is mistaken. Adam writes:

[T]here are a significant amount of spillovers in local real estate investment. Land value is not just capitalized value of publicly provided public goods, but of nearby privately provided positive spillovers. It’s widely recognized that when individuals clean up a property, or open a popular business, there are often spillover values in the neighborhood. Urban economists recognize that the collective value of these spillovers is huge, and in fact makes up a significant amount of land value. 

The fact that private amenities have positive spillovers suggests that they will be underprovided by competitive markets. However, by allowing some of the value of spillovers to be captured, higher land values provide real estate developers, businesses, and even households with incentives to create them. 

The value of unimproved land does increase with improvements on neighboring land. But this does not mean that land value allows a landlord to capture the value of the spillovers created by his own investment. It does not.

Suppose there are two adjacent plots of initially undeveloped land, A and B. A is owned by Andy and B by Barbara. Andy pays for a nice house on plot A. This raises the value of plot B, and enriches Barbara. If no Coasean side payments are made, then Andy fails to capture the value of his investment in the nice house. Barbara gets a windfall from Andy's investment.

Now suppose there is a 100% LVT. When Andy builds his house, he pays no additional tax. But Barbara pays some tax - she pays the full value of the windfall she received from Andy's investment. Andy's incentive to build the house on plot A is unchanged under the LVT. And Barbara's incentive to build a house on plot B is likewise unchanged. 

So I think Adam's critique is just mistaken. 

But, you may ask, what if there is a spillover not to the value of Barbara's land, but to the value of her potential future improvements? Adam raises this possibility later in his post:

Real estate developers who move into neighborhoods with high vacancies, low demand, and high crime are often hoping that positive spillovers from their investment will spur additional investments from others, which will in turn make their investment more valuable.

This is easy to fit into the example above. Suppose the value of a house on plot B is 1.5 times as high if there is also a house on plot A. That's realistic, since a plot of undeveloped land may make a neighborhood less attractive. In this case, isn't Andy overtaxed by the LVT?

No. His incentive to build the house is exactly the same as it would be without the LVT, since without the LVT he would also fail to capture the spillover benefit on Barbara's improvements. There is an uncompensated positive externality, but it's no bigger with the LVT than without it.

In other words, the problem of neighborhood externalities is a thorny one, but the LVT does not make it worse (or better). The big problem with the LVT remains the measurement problem. Of course, that problem cannot be waved away.


Updates

Just to formalize the above intuition a little more, here's the Andy-Barbara example as a 2-person game. Define:

H_A = the value of a house on plot A when there is no house on plot B
C_A = the cost of building a house on plot A
L_A = the increase in land value of plot A when there is a house on plot B
N_A = the increase in the value of a house on plot A when there is a house on plot B
Without loss of generality let the value of a plot of land be 0 when there is no house on the other plot.

Each player decides whether to build a house or not. With a land value tax, L_A=L_B=0. Here are the games with and without a land value tax:

You can easily see that the condition for (Don't Build, Don't Build) to be a Nash equilibrium in the first game is the same as in the second game - namely, that H-C < 0 for both players.

You can also see that the condition for (Build, Build) to be a Nash equilibrium in the first game is the same as in the second game - namely, that H-C+N > 0 for both players.

Therefore, the presence of an LVT won't affect the outcome of which houses get built. This outcome also doesn't change if you make the game sequential.

On a related note, people on Twitter read this post and started bugging me to cite empirical work, which I had previously failed to locate. But I looked again, and this time I found a couple things. For instance, there was a 1997 study in National Tax Journal that examined Pittsburgh's experiment with an LVT in 1979-1980. The study found that the LVT increased building activity. A 2010 study in the Journal of Urban Economics found similar results when examining a number of LVTs implemented in cities in Pennsylvania; not only did LVTs increase the supply of housing, they also increased density.

Affirmative action for conservatives?

Yesterday it was my pleasure to hang out with Jonathan Haidt, a social psychologist working at NYU Stern. Many interesting things were discussed. Much yummy Japanese food was eaten.

One thing we briefly discussed was Haidt's complaint that social psychology has been hijacked by political interests. This is interesting, because a lot of people say that about economics, but in social psych the political types seem to have made much more headway (though politicization probably matters a lot less in social psych, because the fates of millions of jobs and trillions of dollars don't hinge on psych policies the way they hinge on economic policies).

Anyway, the question is what to do about it. Haidt recommends "affirmative action for conservatives":

I'd like us to set a goal for [the Society for Personality and Social Psychology] that we become 10% conservative by 2020. Yes, I am actually recommending affirmative action for conservatives. Set aside any moral arguments; my claim is that it would be good for us. 

Just Imagine if we had a true diversity of perspectives in social psychology. Imagine if conservative students felt free enough to challenge our dominant ideas, and bold enough to pull us out of our deepest ideological ruts. That is my vision for our bright post-partisan future.

This is an interesting idea. But I have a couple of problems with it:

1. Unlike, say, race, political affiliation is a matter of choice. If we start giving preferential treatment to people who say they're conservative, won't people just pretend to be conservative in order to get a leg up in the brutal academic job market? Incentives matter.

2. Affirmative action type programs never perfectly cancel out bias. Instead, they partially counteract bias in some ways and create bias in others. If you start giving jobs preferentially to conservatives, it seems like you could end up with a lot of low-skill conservatives. Conservative researchers might be quietly ignored and disrespected, with the assumption that "he checked the box to get in". This is one of the big problems with race-based affirmative action, and it seems like it would work for political affiliation just as strongly.

3. What are "conservative" ideas anyway? In econ, "conservatives" (or "libertarians", as economic conservatives insist you call them) want to cut government intervention in the economy. In social psych, it seems like "conservative" means something totally different. What if the "conservative" ideas in a field just suck? Shouldn't we be afraid of permanently enshrining bad ideas?

Academia is about ideas. If you treat a package of ideas as if it were an identity group like race or gender, and offer it permanent shelter within academia, I feel like you're restricting the ability of ideas to improve.

But that leaves the question of how to fight against political hijacking of an academic field. Maybe the best way to do it is simply to fight ideas with ideas. If conservative ideas aren't getting enough play in social psych, start giving them play. Write some papers on conservative topics - if you're famous, who cares if no one publishes them, just post them as working papers on your website. Or start a blog, like Scott Sumner did in econ. Gather like-minded academics, using tools like the internet and human networks. Eventually, people will read your ideas and join your movement. If that group reaches critical mass, you can start new conventions, new societies, new journals, etc. As Gandhi said, "Be the change you wish to see in the world."

In fact, this is exactly why academia has tenure in the first place. It's so you can speak out against consensus and not be afraid for your career. The system isn't broken - just use it!

A case where RBC works

I am a fan of John Cochrane because of his intellectual honesty. He's always very up-front and clear about what his priors and his politics are. But he almost never lets that make him tendentious (the one exception being when he is talking directly or indirectly about Paul Krugman). He goes out of his way to acknowledge alternative interpretations and the limits of knowledge.

This post on news shocks is a good example of what I mean. Cochrane reports on a paper by Arezki, Ramey, and Sheng that uses a very simple macro model to explain the economic response to big oil discoveries. Cochrane notes that the paper doesn't need a lot of the fancy friction-mining and utility-mining that are common in macro these days:

My comment was something to the effect of "this paper is much more important than you think. You match the dynamic response of economies to this large and very well identified shock with a standard, transparent and intuitive neoclassical model. Here's a list of some of the ingredients you didn't need: Sticky prices, sticky wages, money, monetary policy, (i.e. interest rates that respond via a policy rule to output and inflation or zero bounds that stop them from doing so), home bias, segmented financial markets, credit constraints, liquidity constraints, hand-to-mouth consumers, financial intermediation, liquidity spirals, fire sales, leverage, sudden stops, hot money, collateral constraints, incomplete markets, idiosyncratic risks, strange preferences including habits, nonexpected utility, ambiguity aversion, and so forth, behavioral biases, nonexpected utility, or rare disasters. If those ingredients are really there, they ought to matter for explaining the response to your shocks too. After all, there is only one economic structure, which is hit by many shocks. So your paper calls into question just how many of those ingredients are really there at all."

Cochrane himself has done a little utility-mining, in his famous habit formation model of asset pricing with John Campbell. But in general, as an opponent of government intervention in the economy, he would (I am guessing) probably rather that the economy work according to a simple RBC-style model where there are no big market failures that would necessitate countercyclical policy.

The Arezki et al. paper is a victory for that kind of simple RBC-type model. But it's a limited victory, since the fluctuations produced by oil news shocks don't look like most business cycles, and because simple models like this don't explain things like the Great Recession. Cochrane, unlike someone making a lawyerly case, goes out of his way to point this out:

Valerie, presenting the paper, was a bit discouraged. This "news shock" doesn't generate a pattern that looks like standard recessions, because GDP and employment go in the opposite direction... 

Thomas Philippon, whose previous paper had a pretty masterful collection of [complex elements], quickly pointed out my overstatement. One needs not need every ingredient to understand every shock. Constraint variables are inequalities. A positive news shock may not cause credit constraints etc. to bind, while a negative shock may reveal them. 

Good point. And really, the proof is in the pudding. If those ingredients are not necessary, then I should produce a model without them that produces events like 2008. But we've been debating the ingredients and shock necessary to explain 1932 for 82 years, so that approach, though correct, might take a while.

Quite true. And many bloggers or op-ed writers would not go out of their way to point this out.

Anyway, to touch on Cochrane's actual point, it's very interesting that simple RBC-type models should be so good at explaining something like an oil shock and so bad at explaining things like big recessions. This fact could lead economists toward something incredibly valuable: an understanding of the scope conditions of RBC-type models.

Scope conditions are the conditions under which a model works well. (**Physics analogy alert**) For example, we know that a model of frictionless motion works pretty well on an ice skating rink and pretty badly under the ocean. And we know exactly why. In decision theory, I personally think that experiments are starting to teach us the scope conditions of super-basic econ 101 demand theory: it works well for one-shot decisions, and not very well for dynamic situations with lots of uncertainty.

But for macro, it's inherently very hard to identify scope conditions, because there's so much going on at once that you can't get a clean comparison between the cases when a model works and the cases when it fails. That's what makes this Arezki et al. paper so interesting - it gives us a clear case (oil discovery shocks) when a mostly frictionless, very forward-looking, perfectly rational representative agent model with Econ 101 type preferences really works. That, in turn, lets us look at cases where RBC models don't work, and ask "How is this case different from an oil discovery shock?" For example, it might be a negative shock as a opposed to a positive one. It might be a shock to a different sector of the economy. It might be an immediate productivity shock instead of a news shock. Etc. Having a case where RBC models actually work helps us narrow down the list of possible reasons why they usually fail.

There will inevitably be many such differences, but they narrow down the types of models we want to consider. If a model fits the Great Recession but doesn't reduce to the Arezki et al. result when applied to an oil discovery shock, we should be skeptical that that is the right model of the Great Recession. As we accumulate more clear-cut cases like the one in Arezki et al., we increase our list of limiting cases that macro models should reduce to in well-defined limits. That in turn moves us closer to what we really want - a model that really explains why big recessions happen, and what can be done to prevent or combat them.

In other words, having a bunch of limiting cases like Arezki et al. lets us throw away macro models. I personally think that's the real problem in the macro literature - the profession lets a thousand flowers bloom, but the flowers never get cut. Clear results like this one give macroeconomists a pair of scissors.