Must-Read: Samuel Bowles, Alan Kirman, and Rajiv Sethi: Reflections on Hayek

Must-Read: Looking back at my archives, it is clear to me that I have written too little about the good and too much about the bad Friedrich von Hayek. Not, mind you, that I wrote anything wrong: when he was bad, he was horrid. But when he was good he was very, very good. And if I want to argue for interpretive charity for James Buchanan and Nancy MacLean—and I think I do—I should be willing to apply it to Friedrich von Hayek: Samuel Bowles, Alan Kirman, and Rajiv Sethi: Reflections on Hayek: “Hayek pioneered the informational view of markets in which prices are messages…

…his dynamic vision of the economy provides the basis of an alternative to the equilibrium methodology that today underpins the economics of information…. Hayek drew a sharp contrast between his approach and general equilibrium theory…. As he put it, the “argument in favour of competition does not rest on the conditions that would exist if it were perfect” (Hayek 1948: 104). Instead, his case for competitive markets rested on the idea that competition was a “procedure for discovering facts which, if the procedure did not exist, would remain unknown or at least would not be used” (Hayek 1968)….[But] the very usefulness of prices (and other economic variables) as informative messages–which is the centrepiece of Hayek’s economics–creates incentives to extract information from signals in ways that can be destabilising….

Hayek… saw the strength of the market economy as arising from the learning and diffusion of new information that it accomplishes in disequilibrium. He argued that “the modern theory of competition deals almost exclusively with a state… in which it is assumed that the data for the different individuals are fully adjusted to each other, while the problem which requires explanation is the nature of the process by which the data are thus adjusted.” In particular, “competition is by its nature a dynamic process whose essential characteristics are abstracted away under the assumptions underlying equilibrium analysis” (Hayek 1948). Hayek’s critique was aimed at the assumption of the passive, price-taking behaviour that is a defining feature of the Walrasian framework….

Is Hayek’s critique of Walrasian general equilibrium obsolete?… We think not. Economic analysis largely continues to be based on characterisations of equilibrium states, without attention to the processes through which such states might (or might not) be reached. Contemporary models of strategic competition and search are equilibrium models, characterised by mutually consistent plans…. There is a common understanding across all individuals regarding the structure of the economy in which they are embedded. Left unaddressed is the process through which such a common understanding might arise….

Hayek… did implicitly assume that the process of entrepreneurial discovery would be stabilising on average…. But the same problems of stability that have plagued general equilibrium theory also arise in the context of entrepreneurial discovery–individually profitable activities can be destabilising in the aggregate. In fact, the interpretation of prices as signals can itself give rise to destabilising feedbacks, especially through the linkage of financial and goods markets. Since changes in asset prices can lead to substantial short-term capital gains and losses, information relevant to changes in such valuations will be actively sought. To the extent that an asset price rise can be used to infer that this happened as a result of the reaction of informed individuals to a change in the conditions of demand or supply, other individuals may seek to profit by buying and hoarding the asset in anticipation of further increases in price. But this activity itself has price effects, which in turn may result in rational hoarding by others, amplifying the destabilising process.

Such effects can be captured by models of information cascades…. In financial markets, attempts to extract information from prices can give rise to prolonged departures from fundamentals in theoretical models (Hong and Stein 1999, Abreu and Brunnermeier 2003), the empirical counterpart of which is excess volatility in prices (LeRoy and Porter 1981, Shiller 1981). When leverage is significant, relatively small informational shocks can give rise to large asset revaluations as funding dries up and assets must be liquidated at fire sale prices (Brunnermeier and Pedersen 2009, Adrian and Shin 2010, Geanakoplos 2010). Since information is costly to acquire and process, assets that have sufficient seniority are considered safe under normal conditions. These can suddenly start to be perceived as risky and “information-sensitive” in crisis conditions, causing trading volume to collapse or markets to shut down entirely (Gorton 2012). Such phenomena do not remain confined to the financial sector….

While Hayek himself did not develop a mathematical formulation of his vision, there do exist models of the economy as a complex adaptive system in which aggregate outcomes are determined by the social interaction of agents with limited and local knowledge. This so-called agent-based literature makes intensive use of computational rather than analytical methods, and focuses on disequilibrium adjustment rather than the characterisation of equilibrium paths. Epstein (2007) calls the approach generative, while Tesfatsion (2006) calls it constructive. Its connection to Hayek’s thought has been previously noted (Vriend 2002, Rosser 2012, Axtell 2016). A key element in this literature is the absence of imposed coordination…. There is no assumption that individual plans are mutually consistent, or that subjectively perceived laws of motion coincide with the objectively realised laws of motion…. This does not, of course, rule out model-consistent expectations or market clearing as endogenous outcomes, arising through adaptation….

Agents may respond mechanically to inputs on the basis of physical laws or behavioural rules, or they may be sophisticated and forward-looking. They may be inter-temporal optimisers employing the same dynamic programming methods used in orthodox models, but subject to private beliefs rather than mutually consistent expectations. The key difference is that “events are driven solely by agent interactions once initial conditions have been specified… rather than focusing on the equilibrium states of a system, the idea is to watch and see if some sort of equilibrium develops over time” (Tesfatsion 2006)…. Leijonhufvud (2006) has argued that agent-based process analysis “will finally make it possible to tackle the central problem of macroeconomics, namely, the self-regulating capabilities of a capitalist economy,” but that the method remains in its “technical infancy.”… It is only a matter of time before the methodology will reach a level of development at which fundamental questions at the core of the discipline can be systematically explored…