II Basic notions of Harberger taxation
III The academic discussion around Harberger taxation and the COST
IV Attribute-dependent taxation
IV.1 Heterogenous evaluation cost
IV.2 Heterogenous degree of loss aversion
The Common-Ownership Self-Assessed Tax (COST) and the current state of Harberger taxation
Seminar “New Approaches to Economic and Public Policy”
Ulrich Roschitsch* June 2019
Abstract. In their book “Radical markets: Uprooting capitalism and democracy for a just society”, based on a series of papers by the authors and affiliates, Eric Posner and Glen Weyl introduce the idea of partial common ownership: every asset in the economy is constantly auctioned through a mechanism that has been existent in a small strand of literature for quite some time—Harberger taxation. The mechanism has been endorsed by some scholars for its combination of simplicity and incentive compatibility (regarding truthful revelation of valuations), enabling the economy to move to a state of higher allocative efficiency. Other scholars have pointed to some practical, but fundamental issues that might cripple a system of universal Harberger taxation (such as failure to assess value).
This paper aims at providing a concise description of the current state of the discussion around the mechanism of Harberger taxation, and Posner & Weyl’s proposal to universally implement it. In a second part, some thoughts on the consequences of a particularly relevant concern are explored: the failure of agents to assess the value of their property. This might lead to the Harberger tax placing potentially rather different effective tax rates on asset owners, depending on their personal characteristics.
At the time when Eric Posner and Glen Weyl published their book “Radical Markets”1, in which they propose (among other things) a system for redefining property rights and how physical property should be taxed, they could already refer to a rich literature that dealt both with the inefficiencies alledgedly caused by our current understanding of property rights, and the suitability of self-assessment-based property-taxation (SAPT) schemes for a broad number of applications. While Henry George stressed the importance of taxing land already in the 19th century2 and a self-valuation-based system for land taxation was introduced by Sun Yat-Sen in China in the early 20th century3, academic interest in SAPT systems began to concentrate after Arnold Harberger’s tax proposal (Harberger, 1965). Since then, researchers from both mechanism-design (Plassmann and Tideman, 2008, 2019) and property rights theory (Levmore, 1982; Epstein, 2014; Fennell, 2019a,b) worked on questions of applicability and modification of SAPTs. Yet, as some scholars lament (Fennell, 2019a), SAPTs failed to spark broader discussions that went beyond the community of the field—until the recent work on the so-called “Common-ownership self-assessed tax” (COST) (Posner and Weyl, 2017, 2018; Weyl and Zhang, 2018).
While the COST, that is essentially a universal “Harberger tax” (see section II), is endorsed by many scholars for ameliorating allocative efficiency through depressing excessively high prices (in theory), it is not uncontroversial. In this paper, we will therefore (1) bring together the literature’s main points of discussion on the (universal) Harberger tax; and (2) develop some intuition on the complications that might arise under imperfectly introspecting agents (Thompson and Leyton-Brown, 2007) that could render a Harberger tax undesirable from a normative standpoint. These points mainly revolve around what we will refer to as “attribute-dependent taxation”: the COST might put potentially very different effective tax rates on different people, depending not on characteristics that should be relevant for a tax (e.g. wealth), but on idiosyncratic, personal attributes. Specifically, we shall consider the effect of differing cost to evaluate one’s property, as well as a differing degree of loss aversion—a concept primarily known from prospect theory (Kahneman and Tversky, 2013) that could prove relevant in situations where average households face decisions with large chunks of hard-to-diversify risk. While it is clear that our current tax systems are not perfect in this respect either, it seems important to still think about the distortions that a Harberger tax might lead to, before one can talk about implementation.
Before considering the above points, we will first take a brief look at how we can formalize Posner & Weyl’s proposal in a concise way, for better understanding and later reference.
II BASIC NOTIONS OF HARBERGER TAXATION
Building on the assertion that the notion of full private ownership of certain assets in the economy will generally inhibit the constant flow of assets to their most productive use (allocative efficiency), Posner & Weyl propose to partly eliminate private ownership. In short, their main argument is that our current property rights system allows owners to hold on to their assets indefinitely, causing many asset transactions to higher-valuing owners to fail because the current owners announce higher reservation prices than their own valuations of the assets. Posner & Weyl thus propose to mitigate this individual “ability to exclude” by introducing a Harberger tax: Every owner of an asset has to publicly declare a monetary valuation for it, in exchange for which she must be ready to cede the asset to any individual that offers to pay this amount. Simultaneously, the owner is taxed at a fixed (and asset-dependent) rate that is levied on the declared valuation (henceforth the price).
Formally, this mechanism can be modeled in a variety of ways (Weyl and Zhang, 2018; Winn et al., 2018; Plassmann and Tideman, 2019). The simplest, and for our analysis sufficent, framework is the one that Weyl and Zhang (2018) use in one of their models: Consider the owner of an asset who lives one period and thinks about which reservation price to announce. The individual has a Bernoulli-utility function U(•): U' > 0, U" < 0 that assigns utility to money values.5 The decision on the announced price is captured by
Abbildung in dieser Leseprobe nicht enthalten
where p is the announced price, F(p) is the distribution function of valuations in the population (continuous by assumption), F(p) := 1 - F(p) is the probability that by the end of the period an individual with a higher valuation than p will materialize and buy the asset at p, v is the owner’s monetary valuation of holding the asset, and t is the Harberger tax rate (the full private property scheme can be throught of as t = 0). As with any expense, the tax payments tp (that are effectively lease payments to society for using the asset) might be subject to a budget constraint tp < m (for our considerations, thought, liquidity / income constraints will not be relevant). The owner’s optimal price is thus characterized by
Abbildung in dieser Leseprobe nicht enthalten
and if the government sets the tax rate equal to the “natural turnover rate” (NTR) of the owner (i.e. the probability that some other individual has a higher valuation than the current owner), t = F(v), then by inspection of (2) we can see that it is optimal for the owner to announce her true valuation: p* = v (formally, p* ^ v both yield a contradiction).
We can see what happens if the tax rate does not match the NTR of the owner, by total differentiation of (2) at t = F(v):
Abbildung in dieser Leseprobe nicht enthalten
where C = (1 - F(v))(1 - t)2 + t2 F(v) > 0 is a positive constant. By the negativity of the derivative we can say that for t < F(v), the owner will overvalue, p* > v and vice versa for t > F(v).5 Generally, there will be many owners of assets that are very similar and in practice one can hardly prescribe an individual tax rate for every owner, depending on her individual valuation. So Posner & Weyl’s proposal of the Harberger tax closes off by gauging the average valuation of owners in an asset class (e.g. small compact cars) through its empirical turnover rate and then setting the tax rate at about one half of that rate (Posner & Weyl aim for t = 1/2 • F(v) to give incentives to invest in the asset). If the distribution of valuations is not too dispersed, the amount and magnitudes of misvaluations will be small (in our further analysis, we concentrate on continuous, unimodal distributions).
III THE ACADEMIC DISCUSSION AROUND HARBERGER TAXATION AND THE COST
The literature of mechanism design and Law & Economics has produced many schemes for (re-) organizing several specific asset markets out of necessity to overcome monopoly- induced issues, such as holdouts, externalities, etc. Eric Posner, Glen Weyl and Anthony Zhang, though, claim two main advantages of their COST-proposal6
1. Universality: the Harberger tax provides a flexible and powerful tool to apply to many assets—the central claim here is that by appropriate choice of the tax rate, many context-specific issues, like the importance of investment-incentives, can be balanced out
2. Self-enforcement: many property-taxation schemes traditionally rely on juries, government agencies or other professionals to gauge a property’s value, thus creating inefficiencies (because the owner, they claim, knows best the value of the asset) and discretionary power—a Harberger tax would enforce itself through appropriate choice of the tax rate
Indeed, the Harberger tax appears to be a particularly convenient, i.e. practical, SAPT scheme among the previously proposed ones. Plassmann and Tideman (2019) compare several SAPTs in terms of revelation incentives, allocative efficiency and practicability. One of their most important findings is that the true valuation for an asset is co-determined by the tax schedule that is used—a fact that can have adverse effects on allocative efficiency. Specifically, if the tax liability of each person depends on her natural turnover rate (as in the original proposal from Harberger (1965), or in the ones from Plassmann and Tideman (2008) or Becker et al. (1964)), then it can happen that the following scenario arises: suppose person A owns the asset and values using it at va, while having some conjecture about the probability of selling it in the future (which determines its tax liability together with va); suppose further that Person B values using the asset at vb > va, but has so much better odds at selling the asset once it is acquired (perhaps because of her personal contacts) that its after-tax value is below that of A. Allocative efficiency would still require B to obtain the asset, but because B has a lower after-tax-valuation (because her tax rate is so much higher), the asset will remain in A’s possession (note that central to this argument is that valuation is composed of use- and potential-sale-valuations; the tax is determined according to both). This example captures the main intuition of theorem 3 in Plassmann and Tideman (2019): having a tax rate uniform across individuals (as in Posner & Weyl’s COST) may actually be better for allocative efficiency than an elaborate mechanism.
Unfortunately, there has not been too much empirical work on SAPTs along the factors just outlined. One interesting first attempt is by Winn et al. (2018): they test the efficiency of the mechanism proposed by Plassmann and Tideman (2008) (a SAPT on real estate with a more complicated tax structure that still guarantees that marginal tax and sale probability are equal at the owner’s valuation) and the Harberger tax with a low uniform rate in a laboratory experiment on real-estate sales with 195 participants. Overall, they find that under both mechanisms, participants overstated their valuations compared to the control group (which operated under full private ownership). The rate of land assemblies (the situation when a buyer manages to assemble all properties needed for a certain project), however, was higher compared to the baseline scenario—they ascribe this to the fact that under both SAPTs, an offer above the declared price instantly leads to a transaction, while under full private property some potential buyers might get frustrated by the long negotiations and leave the market. Regarding the comparison of SAPTs, one of their key observations is that while for the Harberger tax a slight overstatement of valuations was to be expected, the same is not true for the Plassmann-Tideman-mechanism. They take the fact that still both mechanisms perform statistically identical as indication that the Harberger tax might be preferrable in terms of simplicity and clarity to the agents.
Despite the encouraging work especially on uniform-rate Harberger taxation, there are also critical voices. Frequently made arguments include the complexity of Harberger taxation relative to full private property, the concern that owners subject to liquidity constraints might be ’bought out by rich people’ and the handling of complementary asset groups. While many of these points can be accomodated by proper modification of the Harberger system, one point, that is often picked up by scholars in the Austrian tradition, attacks the fundamental notion of the necessity of truthful revelation itself. In an explicit reply to Posner & Weyl, Rallo (2019) claims that monopolistic markups are in fact needed to incentivize agents to invest effort in finding out an asset’s true value to them; without this incentive (i.e. if a Harberger tax forces them to announce their valuation), agents— anticipating less opportunities for arbitrage7 —would produce less knowledge, leading to more failed assessments and thus misallocations.
* M.Sc. Economics candidate,
1 See Posner and Weyl (2018).
2 Cf. Posner and Weyl (2017), p. 52f.
3 Cf. Niou and Tan (1994).
4 This is slightly different from Weyl and Zhang (2018), since they directly work with money values as utility (U : c ^ c); we will follow Winn et al. (2018) in choosing a more general BUF.
5 Actually we can only make this observation for a neighborhood of t = F(v), however for sufficiently ’nice’ Bernuoulli-utility-functions, our analysis is w.l.o.g.
6 In many contexts, the phrase “COST” and “Harberger tax” are used interchangeably, since the COST is a universal HT; it is important to keep in mind, though, that it is naturally possible to introduce a HT for only certain asset markets—thus creating room for arguments against a COST but not against a HT.
7 Rallo essentially just points to the Grossman-Stiglitz-paradox as a rationale, see Grossman and Stiglitz (1980).