In a new working paper, “The Failed Appropriation of F.A. Hayek by Formalist Economics,” Peter Boettke and Kyle O’Donnell contrast Hayek’s emphasis on the economy of knowledge with the formalist economics of information. It follows a similar theme to some of Boettke’s other work, which accuses formalist economics of leaving things out which cannot be formalized, but are valuable nonetheless. I find myself somewhat divided, because I agree with the authors on what formalists tend to leave out, but I disagree that it’s formalism that’s to blame. Rather, my take is that what drives differences in conclusions is not formalism, but the questions being answered and the way they’re asked. Much of this post is a summary of how various economists have approached the problem of imperfect and decentralized knowledge, therefore I will bold the first sentence of where I start to make my point, contra the thesis presented in the paper (feel free to skip to that point).
Before tackling a summary, it should be clear that there is an explicit and implicit relationship between Hayek’s work and subsequent research in the field. This link is left out of my summary, both because of length considerations and because it’s hard to fit it in more naturally into this post, given my ultimate end — challenging Boettke’s and O’Donnell’s interpretation of why later economists diverged from Hayek in their interpretations. But, in any case, the economics of knowledge/information field stems directly from Hayek’s pioneering theoretical research on the problem of decentralized knowledge.
Hayek’s contribution to the problem of social coordination and knowledge focuses on how, despite the decentralization of knowledge, such that one person only knows so much and may have no idea on what she does not know, societies (economies, in this case) can coordinate. Formally, coordination refers to an expectational equilibrium, where the plans of all economic agents coincide over some period of time (Hayek ). Microeconomic theory gives us an extensive set of tools to model a coordinated economy, but it is, at best, a foil for understanding how coordination actually occurs in the real world. An equilibrium-based microeconomics must assume the coordination to have already taken place, so it arguably does not help us answer the more interesting question: how do economies achieve order?
In the context of the socialist calculation debate (“What is the problem we wish to solve when we try to construct a rational economic order?”), Hayek offered a preliminary answer to this last question in “The Use of Knowledge in Society.” Hayek’s answer is that forms of communication, such as language and money prices, spontaneously emerge, and these act as proxies which allow people to act as if they had the relevant set of information. The question that naturally follows from this premise is how do people search for, stumble upon, or, more generally, find the information they need?
Austrians are interested in the dissemination of information in a context where individual agents have some set of information, know of some sets of information they don’t have, but also know nothing (including the existence) of another, probably much larger, set of information. The latter aspect of the problem is frequently referred to as “radical ignorance,” or the unknown unknown — things we are unaware of not knowing. We do know that there is information out there we don’t know about, but the process of finding and interpreting this knowledge isn’t a straightforward maximization problem. It requires a competitive trial-and-error process, the outcome of which is, to one extent or another, indeterminate. But, seen in this light, economic loss is not necessarily something we should avoid. Oftentimes it is the byproduct of a very necessary process of communication that allows for social coordination. The alternative is the much more severe loss of the lack of this process. The broader context of the Austrian solution to the knowledge problem are the institutions which arise to help mitigate the problem of radical ignorance.
An early “post-Hayekian,” if you will, contributor to the economics of information was George Stigler, who I see as a brilliant microeconomist.1 I mention his work microeconomics work, because it’s within this context that his “The Economics of Information” is best interpreted. In this paper, Stigler tackled the problem of the need for information in an environment of ubiquitous ignorance. Take, for example, an agent who is interesting in buying some product and is interested in acquiring it for the lowest price. More likely than not, there is no single price and the various prices are dispersed, in the sense that the agent doesn’t have timeless, or automatic, access to them. Thus, the agent must commit to a search. This person can expect, with some probability, that a search will reveal lower prices, but we have to balance this gain with the opportunity cost — value lost by taking time, and whatever other resources, used to conduct the search — of and the diminishing returns to searching. As a result, there is some theoretical (i.e. formal) point where the expected gains of search become lower than the expected costs, implying that it is rational for agents to remain, to one extent or another, ignorant.
Subsequent research on the economics of knowledge follows a similar theme. But, formalization of these theories led to models where, to be able to judge the net benefits of knowledge acquisition, agents become aware of the information they don’t know. Boettke and O’Donnell mention a paradox, which is the problem of accurately judging net benefits if agents don’t know the content of some information set — thus, there is some assumption of omniscience in a model meant to help explain knowledge acquisition (which implies some level of ignorance). But, more importantly, what Boettke and O’Donnell are getting at — I think — is that by modeling the problem in this way, the economist leaves out an important aspect of the process: specifically, the consequences of not being able to accurately judge net benefits.
To illustrate how such a framing could misguide economists, Boettke and O’Donnell briefly explore the work of Joseph Stiglitz and Stanford Grossman. To summarize their contribution, we’ll look at one of their more interesting conclusions, explained in “Information and Competitive Price Systems.” Starting with the premise that prices are the communication device, they show that prices are least reliable when they are most needed. Suppose, for instance, that an agent is interested in acquiring information that will help to better predict the expected gain from buying some risky asset. The search cost will be highest when the least amount of people know the relevant information. Following Stigler, we know that the higher search costs are, the less information a rational agent will be willing and able to acquire. Thus, the smaller the fraction of people who know the relevant information, the more costly acquisition will be, therefore the less likely this information will be acquired. But, if we’re worried about efficient allocation, these are the situations where prices, being proxies for decentralized knowledge, are needed the most. Yet, because this knowledge is less likely to be sought, and because comparatively few people hold this knowledge, the less likely it is that prices will reflect it. Based on this finding, Stiglitz and Grossman put into doubt the informational efficiency of the pricing process. Stiglitz and Grossman mention one more implication, but I will return to this below.
Boettke and O’Donnell summarize the facets of the formalized theory of the economics of information as follows,
- Knowledge is defined objectively — that is, we can define a known set of knowledge, and then assume that different agents hold different combinations of the elements within this set. For example, assume three agents and a known, objective knowledge set (a, b, c, d, e). Then we can assume that agent 1 knows (a, b), agent 2 knows (b, c), and agent 3 knows (d, e). Boettke and O’Donnell mention that this method leaves no room for the problem of interpretation (
suggesting that there is no such thing as objective knowledge[Update: suggesting that some subset of knowledge/information is not objective]);
- This knowledge set remains the same under alternative institutional frameworks;
- Agents have objective functions which they maximize, implying the mechanization or automation of human action.
There is a relatively clear divide between Hayek and “mainstream” information economists in the conclusions derived from their respective work. My interpretation of Boettke and O’Donnell is that their argument is an attempt to explain the divergences in conclusions that stem from an incomplete understanding of Hayek’s work.
The authors mention that some, such as Grossman and Stiglitz, put too much emphasis on the notion that prices are efficient at informational dissemination. Therefore, the argument in favor of the market process does not rely on the belief in an efficient pricing process. Another reason that many informational economists diverge is because their formal models necessarily leave out important concepts and ideas that cannot be formalized. Thus, later, formal models leave out some of the nuance that is present in Hayek’s work, especially “Competition as a Discovery Procedure.” Boettke and O’Donnell go on to argue that formal models, as a result, leave out various imperfections, and therefore drive focus away from institutional solutions to these imperfections.
I agree that there are differences in interpretation of the problem. The differences are perhaps most starkly contrasted in the final part of Grossman and Stiglitz (1976). They explicitly link their research (summarized above) on the inefficiencies of market prices to the socialist calculation debate of the 1930s. They, I think rightly, interpret the debate as being a comparative analysis of information problems under different institutional frameworks (i.e. capitalism v. socialism/communism). They argue that their findings suggest that the real debate is a comparative analysis between the costs of an imperfect process of price arbitrage and the costs of monitoring bureaucrats. I find this conclusion remarkable, because I don’t see how it follows from their research.
I disagree that these differences are caused, at least to a large extent, by the emphasis on formalization (see also my post, “Intuition and Math“). Rather, I think the problem is one of intuition. Like Boettke and O’Donnell emphasize, one problem in informational economics is that of interpretation. Even if we assume one objective set of information, knowledge will still be asymmetric between agents, because information has to be interpreted. Where different agents have different sets of priors, and because of the creative dynamic of the human mind, two people are unlikely to interpret something in the same way. This is as much a problem in broader society as it is in the sciences. Grossman’s and Stiglitz’ interpretation of the central issue of economic calculation and resource allocation is clearly different to that of Hayek and Mises. They fail to recognize that information decentralization will exist in socialist societies as much as it does in capitalist societies, and therefore the real issue is which set of institutions allows for a better use of resources, given the similar constraint.
There is no formalization–intuition trade off. Models are meant to formalize intuition as a means of: (a) reducing the probability of misinterpretation, and (b) reducing the probability of error in reasoning. But, intuition is still relevant. Formal models are built on intuition, and therefore it’s the intuitive framework that precedes the model, not the other way around. For example, we can be fairly certain that if Hayek and Stiglitz were told to formally model the same problem, their models would be different — and Hayek’s model would better internalize the nuances that Stiglitz doesn’t consider. Stiglitz’ interpretation of the problem is path-dependent on his training, which includes his interpretation of economic theory developed before him. We should expect divergences in opinion, caused by differences in how different economists perceive the problem.
Differences in perception also allows for differences in what problems economists see as relevant. More broadly, there are different issues that different economists can explore, and models will be shaped around the problem the economist is attempting to solve. Stigler, for example, may not have been necessarily interested in explaining the process of macroeconomic coordination. Instead, knowing that people are interested in acquiring information to improve their decisions, he built a theory of rational ignorance. This theory also solves the problem of explaining things that economics otherwise wouldn’t be able to explain: the institutions, techniques, and organizations that help mitigate the costs of information searches. Likewise, George Akerlof used (“The Market for Lemons;” my summary) a premise of asymmetric distribution of knowledge to explain institutions like reputation and product guarantees. One advantage of interpreting differences in models and frameworks in this way is that it helps to interpret these contributions not as mutually contradictory, but as complements that, together, help provide a more complete picture of social institutions.
This latter outlook may be more difficult to reconcile with other cases that do seem contradictory, such as differences in opinion on what the socialist calculation debate is really about. But, maybe this is just an issue of differences in the what is being explained. One point that Grossman and Stiglitz (1976) make is that central planners with a certain set of information may be able to improve on markets, especially if prices are imperfect representatives of knowledge. Within the Hayekian framework, we might reinterpret this as arguing that if bureaucrats have local knowledge, it behooves markets to gain access to it. This is not necessarily inconsistent with the lessons of the socialist calculation debate, especially if we interpret the point more narrowly — not as arguing for complete planning, but for selective planning.
We have reason to believe that markets will oftentimes significantly oversupply or undersupply some good or service, implying a misallocation of inputs. Institutions arise to help constrain market agents to minimize the costs of, what are called, market failures. These institutions often include rules that help consider knowledge sets held by government, which are acquired through various means (i.e. voting). Under certain conditions, government, bureaucrats, or whatever else you want to call central planning, will have a comparative advantage in providing some good or service. This point is made, for example, in The Calculus of Consent. Likewise, organizations that act like bureaucracies will arise because of inefficiencies in the pricing process. This was the basis for Coase’s theory of the firm. Interpreted in this way, Grossman’s and Stiglitz’ conclusions seem to be less at odds with Hayekian coordination theory than one might interpret at first.2
To summarize my criticism of Boettke’s and O’Donnell’s argument, I don’t think there is any problem with formalism. If there is a problem it’s with intuition, because it’s natural to assume that there will be differences in priors, interpretations, and therefore opinions. Formal models will represent the idiosyncrasies of their authors, not the other way around. Also, to argue that formalization causes a rift in conclusions takes away from the fact that these conclusions may not be so contradictory after all, depending on one’s interpretation. More broadly, maybe each economist are exploring different sides of the same shape. A suitable comparison between Hayek and others may be that the former explores why agents don’t need perfect knowledge, while the latter explore the knowledge agents do need. But, Hayek was aware that agents need knowledge to be able to coordinate with each other, and so these problems aren’t necessarily separate. They’re just different angles of approach.
This is not to say, of course, that perhaps there are aspects of Hayek’s arguments that have been missed and that should be re-emphasized. But, this is a natural consequence of the knowledge problem that economists face.
1. While not relevant to the topic of this post, since we’re discussing Stigler, I can’t recommend his work in the field of the history of economic thought enough. See, for example, “The Economics of Carl Menger” and “Sraffa’s Ricardo.”
2. I still hold, however, that Grossman and Stiglitz inexplicably overlook the knowledge problem that central planners face, even if they do have local knowledge that others do not.