The Henry George Tax is vulnerable to mismeasurement

1/23/2014 08:25:00 PM
High density residential buildings have value. But how much value does unattended scrub land have?
Noah Smith has a piece up advocating for the "Henry George Tax" which is like a property tax except that it is assessed not against the appraisal value of the property, but instead against the appraisal value of the land portion of the property, excluding the value of any development on it. Smith argues a common refrain:
"Unlike income taxes, sales taxes, or corporate taxes, the Henry George Tax has no chance of choking off economic activity; after all, the amount of land is fixed, so you can’t tax it out of existence. Also, unlike the property taxes we have now, a Henry George Tax actually encourages landlords to build useful, valuable stuff on top of the land they own. Conventional property tax pays people not to build things on their land, since doing so will mean having to pay more tax."
In theory, he's exactly right. Normally when you tax a good, producers respond by producing less of it, which results in deadweight loss that can't be redistributed. But land isn't produced or destroyed, so by taxing land you are merely redistributing from the fairly wealthy landowners to the less wealthy non-landowners, with no deadweight loss anywhere in between.

Unfortunately, the reality may not line up so well with the theory--this is a case where the empirics are considerably more important than the theory. In order to tax land value, the government first has to measure land's value. Unlike things like labor income, measuring land value is extraordinarily difficult because land is never traded on the markets in isolation--it is always part of a larger transaction including not just buildings constructed on that land, but also less-obvious kinds of development like the number of trees, quality of the soil, landscaping etc. What the government can observe is the sale price of the whole property--buildings and land all as one--but not the price of land it seeks to tax. What we'd need to do, then, is crank out hedonic pricing models to try to estimate what portion of each property's value comes from the land it sits on rather than the development it contains. A hedonic pricing model is just a regression of the total property price on a variety of characteristics about that property (floor space, building materials, neighborhood, number of bath rooms etc) one of which is the number of acres of land the property sits on:
In this equation, can be interpreted as the market value of a unit of land, meaning that the Henry George tax would be assessed percentage of this times the number of acres:
where is the percentage tax rate.

The problem here is that a hedonic pricing model like the one above is extraordinarily hard to estimate. I mean, any old hack can do an OLS regression on that data, but in reality the data are teeming with endogeneity, omitted variables, multicolinearity and the like that all bias the estimate of . And don't be thinking that there's any reason to suspect the OLS estimate will be "close enough"--I know of datasets where OLS will tell you that an acre of land has negative value!

To see why this matters, lets gather 'round for a parable: Suppose you are a real estate developer in the fictional town of HenryGeorgetown, where the currency is called Ducats. You spot a potential development site where you could build an apartment complex that would generate 200 Ducats in real present value at the current equilibrium rental rates. It would cost 100 Ducats to build this apartment complex, but first you must buy the one acre plot of land. Since the project generates 200 revenues of which 100 are spent, you'd be willing to pay up to 100 Ducats for the plot. Since you are the marginal investor in this market, that is now actually the true market value of an acre of land in this town. You could build the apartments, markets would clear, there'd be no deadweight loss, and everyone would be happy.

But no! Crazy economists working at HenryBrook University convinced the local authorities to adopt a land-value tax of 50 percent in order to finance HenryGeorgetown's government. Unfortunately, the Finance Professor at HenryBrook in charge of estimating the hedonic price model omitted a couple important (perhaps non-measurable) variables, and mistook his OLS coefficient of 300 Ducats as the true market value of an acre of land. The taxes assessed against that are 150 ducats, and as a result, your venture to build an apartment complex is no longer profitable--with taxes, you'd loose 50 Ducats! And remember, you were the marginal investor in this market, meaning yours was the most profitable venture possible to put on this plot. No other venture would generate more than 100 Ducats after construction costs, so simply put, no one can afford this land anymore. We now have unemployed land in HenryGeorgetown, and disappointingly low tax revenues.

The point is that even though it is true that if we had the true model we could calculate the true value of land and Noah Smith's tax story would work out, but when our estimates are wrong the whole story falls apart.

Also, as an aside it is worth noting that even with perfect measurement, it isn't totally true that taxing land value doesn't have any incentive effects on development--there are compound effects whereby having a bigger development on a plot of land makes the land itself more valuable, and taxing the portion of land's value that arises from compound effects would certainly discourage development. That means we must isolate only the portion of land's value that is totally orthogonal to all types of development--not just buildings but trees, landscaping, even top-soil quality--and it's just not clear to me that there's much value left there to tax.