Jones v. Harris and Some Ramblings on Burdens of Proof, Empirical Evidence, and Behavioral Law and Economics

Cite this Article
Joshua D. Wright, Jones v. Harris and Some Ramblings on Burdens of Proof, Empirical Evidence, and Behavioral Law and Economics, Truth on the Market (August 21, 2009), https://truthonthemarket.com/2009/08/21/jones-v-harris-and-some-ramblings-on-burdens-of-proof-empirical-evidence-and-behavioral-law-and-economics/

Much has been made about the importance of Jones v. Harris as a battle in the ongoing war between behavioral economics  and rational choice/neoclassical framework (see, e.g. the NYT).   If the case if to be about the appropriate economic methodology or model for assessing legal questions, it is definitely an interesting turn to have Judge Easterbrook representing the rational choice economists while Judge Posner (who is simultaneously taking some flack for fast and loose and incorrect uses of macroeconomics) defends the behavioral view, considering that the latter wrote an important critique of the behavioral law and economics literature (here is an excellent summary of Posner’s opinion from Professor Birdthistle).  Professor Ribstein frames the issue of Jones v. Harris and the New Paternalism nicely with a prediction:

I suspect that in this day and age the Supreme Court will side with Posner. Such a decision would be a symptom and signal of our sharp turn toward paternalism in everything from complex finance to corporate governance to the simplest products.

I’m no expert on Gartenberg or any other particular legal issues arising in Jones v. Harris.  For commentary from the real experts, see Professor Bainbridge, Birdthistle, Ribstein, or Oesterle.  But what I’m interested in more generally is the law and economics angle here.  More specifically, I’m interested in both the arguments about how the relative merits of behavioral and standard “vanilla” neoclassical economics play out in the legal sphere as well as the how these debates play out from an empirical perspective.   I’ve written on the relative performance of behavioral and “vanilla” neoclassical economics in the context of consumer product markets and found the claims supporting the former (at least in the behavioral law and economics literature) to be overstated.  In particular, I believe it is incredibly common practice in that literature to jump from the identification of a behavioral or cognitive bias identified in the experimental literature to accepting that some regulation must be appropriate,  and shifting the discussion to the design of that regulation.  Infrequently are the relative social costs of the cognitive quirk and the regulation discussed — much less unintended consequences, error costs and the sort of dynamic learning costs imposed by the new paternalism on incentives to learn and mitigate biases.

These latter types of dynamic effects are discussed by Klick and Mitchell and in Ed Glaeser’s essay on Paternalism and Psychology and are important — perhaps critical — to the accuracy of any cost-benefit analysis of regulatory proposals / legal rules aimed at “solving” cognitive bias because there is a danger that any given rule / regulation will increase the rate of errors.  This point goes directly to the appropriateness of the “libertarian” modifier for this type of paternalism when its proponents describe it as “libertarian paternalism.”  For example,  Sunstein & Thaler argue that liberty is maintained because these proposals encourage choice rather than coercion.  But the libertarian case also rests on the presumption that allowing individuals to bear the costs of their errors leads to better and more competent choices in the future.

Admittedly, estimating the true social costs and benefits of changes in legal rules is quite difficult in practice.  But the all to common formulaic approach in the behavioral law and economics literature of: (1) cite experimental evidence that identifies bias, (2) do not discuss whether this bias correlates with other biases that may be offsetting, (3) argue that the bias undercuts all of rational choice economics and its predictions, and (4) design appropriate regulation without regard for the true social costs it imposes (including error and dynamic costs) — is problematic.  (3) and (4) are problems that law and economics scholars utilizing behavioral economics are responsible for — not the folks in behavioral finance and economics actually generating theory and evidence.

Indeed, one classic example of this is the reaction of the behavioral law and economics community to the Zeiler & Plott (2005, American Economic Review) analysis demonstrating that the common asymmetry between people’s valuation of gains and losses can be made to grow or to disappear by manipulating the context or circumstances under which the valuations are made.  I blogged about what I described as the Endowment Effect’s Disappearing Act here.  In the meantime, I noted back at the end of 2005 a search of the JLR database had 541 hits for “endowment effect.”  Post-2006, that number is already 255, with 210 of those also including the word “regulation.”  How many of those cite (much less discuss) Zeiler & Plott (2005)?  Sixteen.  Sixteen out of 255! That’s an incredibly disappointing figure and, at a minimum, suggests that taking empirical evidence seriously is not a top priority in the portion of this literature appearing in JLR.

Back to Jones v. Harris and mutual funds . My sense is that some of that style of argument, and in particular failing to distinguish between the observation of a defect and a thorough analysis of its consequence in actual markets, is working its way into the Jones v Harris debate (at least in a milder form in the Posner opinion and the Litan, et al. brief (they do rely on some empirical evidence and favor litigation over excessive fees to other forms of more invasive regulation such as price controls)).

One of the most important issues here is to distinguish between behavioral quirks and competitive outcomes.  What do the theory and evidence say?

Much of the empirical debate here appears to turn on this Coates & Hubbard study showing that the mutual fund industry tends to be structurally competitive with low entry barriers, low concentration, unstable and shifting market share. Here’s how Coates and Hubbard characterize the evidence:

In sum, the market structure and performance of the mutual fund industry is consistent with strong competition among funds. New entry is common, and for decades has been a constant feature of the industry. Barriers to entry are evidently low, and funds are distributed through multiple distribution channels that themselves reflect a second layer of competition for investor assets. While our survey of evidence of the industry’s market structure is necessarily general, and thus it is possible that there are subsectors of the mutual fund industry where the market is more concentrated, barriers to entry are high, or distribution channels are few, the general survey suggests that the burden of proof should be to establish that such potentially uncompetitive subsectors exist, rather than for to critics to presume, as they have since the 1960s, that competition is generally weak among mutual funds. This general conclusion is only reinforced by a review of evidence of the performance of the fund industry. Fee reductions are common, fees have shown no dominant long-term trend, and market shares are unstable. All of this evidence – admittedly indirect – suggests that competition among funds and fund complexes is robust and, if anything, has been growing in intensity over the past decades.

Easterbrook appeals to the study while Posner and the amicus from Litan, Mason and Ayres offer various rebuttals or counter-evidence.  E.g. Litan, Mason and Ayres point out that Coates & Hubbard can’t identify the impact of changes in fees on fund market shares econometrically because individual fees are not sufficiently variable (it’s true that this weakens the strength of the evidence, correlation not being causation and all of that — but I read Coates & Hubbard as appropriately circumspect with regard to what their evidence shows and doesn’t show. Meanwhile it is true, in my view, that the body of evidence they point to is consistent with competitive conditions) and offer evidence from two other studies (Javier Gil-Bazo & Pablo Ruiz- Verdu, When Cheaper is Better: Fee Determination in the Market for Equity Mutual Funds, 67 J. Econ. Behav. & Org. 871, 883 (2008) and Guo Ying Luo, Mutual Fund Fee-Setting, Market Structure and Mark-ups, 69 Economica 245, 245 (2002)) to conclude: “thus, the overwhelming evidence is that competition in the mutual fund industry has not produced competitive outcomes.”  This is a rather loose use of the term “overwhelming” in my view — not because these studies are poorly done but because there are two of them and there are others with conflicting evidence.

There’s an important theoretical discontinuity going on here in terms of the economics.  It’s one thing to point out these behavioral anomalies.  But think for a moment about the evidence of very low market concentration and low barriers to entry.  Evidence of behavioral anomalies is not sufficient to suggest that there is not competition.  Competition is multi-dimensional: Ex ante v. ex post, price, quality, service, innovation, etc.  Imagine that a fraction of investors exhibit these cognitive biases and will be profitable to exploit ex post because they will not switch funds after poor performance and will continue to pay high fees.  If sellers anticipate this ex post profitability, and there is competition among sellers, one expects these rents to be dissipated by ex ante competition across these different dimensions.   This is a key point about switching costs or other ex post holdup (or in this case the presence of behavioral quirks that make for opportunities to holdup consumers).  Where ex ante competition is vigorous and such opportunities are anticipated — and nobody seems to dispute that both of these conditions are satisfied by the evidence presented by Coates & Hubbard — the likelihood of supranormal returns is dubious.  In other words, showing that such holdup occurs or that switching costs actually deter some switching on the margin (of course) or that behavioral quirks are real and not imagined is quite different than showing that the mutual fund industry is not competitive.  To be sure, the competitive equilibrium might look different where these biases exist, in that one might see economic rents dissipated on different margins that consumers value (if not fees, something else), but that is not the same as saying the market is not competitive.

The basic economic point is that demonstrating that some consumers are systematically irrational alone says nothing about whether fees are likely anticompetitive or supra-competitive or about the strength of the economic logic that says you’d still get competitive outcomes. A key question is whether mutual funds are earning supra-competitive rents.  And this is simply not likely in the face of low entry barriers, dynamically changing shares, plenty of entry and exit, and low concentration.    That doesn’t mean it’s impossible.  But much of this discussion is about assigning the correct burdens in the empirical debate.  It’s difficult to know when fees are high relative to some competitive benchmark precisely because we don’t observe the counter-factual, but-for world.  That’s why understanding the role of ex ante competition, even where there are ex post profit opportunities deriving from behavioral quirks or switching costs, is an important part of resolving that issue.  So is the other structural evidence that suggests that there is plenty of competition between funds.  But alas, much of the behavioral literature (and, unfortunately, Posner and the Litan, et al. brief) engage insufficiently with the nuances of this crucial analysis. As a result, their claims are significantly weakened.

Coates and Hubbard make a similar point in their paper.  As does Judge Easterbrook in offering a related point about how competition benefits the infra-marginal consumers even if they are not sophisticated (though this point is slightly different in economic substance than saying that the consumers are systematically irrational):

It won’t do to reply that most investors are unsophisticated and don’t compare prices. The sophisticated investors who do shop create a competitive pressure that protects the rest. See Alan Schwartz & Louis Wilde, Imperfect Information in Markets for Contract Terms, 69 Va. L.Rev. 1387 (1983). As it happens, the most substantial and sophisticated investors choose to pay substantially more for investment advice than advisers subject to § 36(b) receive. * * * When persons who have the most to invest, and who act through professional advisers, place their assets in pools whose managers receive more than Harris Associates, it is hard to conclude that Harris’s fees must be excessive.

Litan, et al. do respond to this point in their brief (as does Posner), so I do not mean to imply that they are unaware of the general argument.  The issue I’m interested in is figuring out is what quantity and quality of empirical evidence is necessary, when coupled with an argument about behavioral economics that consumers are sometimes irrational, to justify legal change to mitigate those biases. I don’t believe the findings of these anomalies in the literature are artificial — though some evidence is better than other evidence — but it strikes me as reasonable to believe that the burden of proof is much broader than generally assumed by proponents of libertarian paternalism in the behavioral law and economics literature.  The Litan, et al. brief and Posner opinion, in my view, both fall short of an interpretation of the existing data that grapples with the structural evidence.  I most definitely do not find the empirical case, at least from what I’ve seen so far, “overwhelming.”

The more interesting point to me is the implicit claim that citations to the behavioral literature are presumed to change or even shift the burden.  To their credit, both the Posner and Litan, et al. actually deal with some of the empirics (though not in great detail — nor do I here in the blog post).  My prior is that a mixed body of evidence and arguments about behavioral quirks does not shift the burden.  Others disagree, I’m sure.  The point of this particular post is not to quibble about the details of any specific empirical studies or make a comprehensive review of the literature.

Which brings me back to the questions which started me down this road and on this post: how should we weight behavioral / experimental evidence in these arguments?   What about evidence that the biases are mitigated over time in markets?  What about when the predictions conflict with the structural evidence?  How we we reconcile those conflicts?   And most importantly, how large is the theoretical and empirical gap between demonstration of cognitive biases in the laboratory or even real markets and satisfying the burden to show both (1) that markets do not mitigate these biases and generate competitive outcomes, and (2) if they do not, that proposed regulation will help more than it hurts (or that its benefits will exceed costs including those discussed above).

This is my tentative view, and I’d like to look more closely at the existing empirical literature before saying anything more concrete, but the more I think about this the more I think that the really important issue in Jones v. Harris is not about Easterbrook v. Posner, or even classical v. behavioral economics (though this is also important), but about how the Supreme Court assigns the empirical burden and evaluates the existing econometric literature.  Similar issues arise in antitrust, and there is a movement in antitrust (wrongheaded I believe) to integrate the insights of behavioral economics into policy and analysis,  so I have some interest in how the Roberts Court resolves both of those issues.