Let’s Move Past the Hayek-Keynes Debate

The debate between Hayek and Keynes on the question of depressions still looms large in the economics profession, at least in the way it’s taught and communicated, and – in some corners – still in the way it’s conducted. Formative as that debate was, being several decades prior to the ‘expectations revolution’, neither one had the right tools to even pose the question adequately. And as great as the rap battle is – and it is truly great – returning to this debate in the modern era keeps these ill-posed framings on life support, framings that should at this point be relegated to history of thought.

I

Consider two stories about intertemporal decisions. Consumers face a choice between consumption today and consumption later. In practice, they effectuate this choice by either consuming today, or…

• Hayek: investing in capital goods.
• Keynes: holding onto their money to spend later.

The demand for capital goods is primarily determined by…

• Hayek: the availability of funds to finance their purchase.³⁶
• Keynes: the expectation of profitable returns, which is primarily informed by current demand conditions.³⁷

So if consumers shift their preferences to reduce consumption today in favor of consumption later, the demand for capital goods…

• Hayek: rises
• Keynes: falls

which affects the economy by…

• Hayek: falling interest rates. The lengthening of production processes picks up the slack in the volume of spending (Prices and Production, ch. 2).
• Keynes: compounding the initial fall in demand. This triggers even more saving, and economy enters a deflationary spiral.

Both of these are pretty inadequate stories, for different reasons.

II

Notice how many assumptions are being made in both of these stories.

Critics of Keynes frequently note that Keynes neglected the role of the interest rate in coordinating intertemporal decisions. In fact, Keynes seems to have the upper hand here. It’s difficult to imagine a world where an autonomous fall in the demand for a particular consumer good would lead the owner of the capital goods further up the supply chain (whether durable or circulating) to anticipate an increase in future demand for that good, as would be necessary for him to increase investment in that line (though Garrison tries to do exactly this with the Saving-Up-For-Something notion in Time and Money, ch. 3). As Keynes points out, a fall in the interest rate does not by itself send such a signal.

On the other hand, it’s also hard to imagine expectations reacting so mechanically to current conditions as in Keynes’ story. Friedman’s critique of Keynes with the Permanent Income Hypothesis argued that consumption-smoothing on the consumer side prevents the volume of spending from reacting wildly to transitory income shocks. Similarly on the producer side, “production smoothing” (partly of choice because the shock is seen as transitory, partly of necessity due to capital immobility) means that an increased demand for future income at the expense of present income should not lead to the spiraling depression that Keynes feared.

The whole discussion seems to rest on the contrived example. Just as we do comparative statics exercises where the demand for a good rises, can we not imagine the demand for future income in general rising as well?

Likely not. While preferences over consumer goods make sense to think of as exogenous, the preference for future consumption over present is endogenous to expectations of the size of future income. As with Scott Sumner’s dictum to “Never reason from a price change”, positing an endogenous variable changing autonomously leads us to tell causally confused stories. In this case, never reason from a change in time preference.

In fact, the demand for future consumption must be distinguished from both the demand for capital goods and the demand for money. And doing so rigorously requires tools to think about expectations that had yet to be developed at the time.

III

What’s missing in both of these stories is the way banks and other financial institutions link together expectations with the volume of spending, and how this coordinates intertemporal consumption decisions.

Because the ability to finance capital goods played such a crucial role in Hayek’s story of how intertemporal decisions get coordinated, he looked at banks’ ability to mark up their balance sheet in order to finance capital projects very skeptically. If banks are willing to finance any profitable project at the going rate, there’s nothing to nail down the demand for capital goods. Money lending is a “loose joint” in the economic system, as he put it.

Keynes, on the other hand, was skeptical that any amount of liquidity would lead banks to invest in capital projects during a downturn. In his story it is not consumers who hoard money balances in base money, but banks who fail to lend the money that consumers have deposited, on the basis of self-fulfilling expectations of poor demand for final goods. Even Hicks’ later elaboration with the idea of an “elasticity of expectations” (Value and Capital, ch. 16) is similarly mechanical, hence his similar pessimism about the stability of the dynamic capitalist system.

Though much attention has been paid to their opposing remedies, it’s striking that both of them conclude that the capitalist system is inherently unstable, for quite opposite reasons that nevertheless center around the role of expectations in credit creation. A more rigorous theory of expectations, however – take Lachmann or Lucas, as you prefer – puts both of these fears to rest. It is not merely that current prices are determined by expectations, which both Hayek and Keynes appreciated, but that expectations themselves are determined by looking forward, and not merely mechanically.

If this is the case, if banks are able to increase the volume of circulation in response to improved expectations, but if those expectations are informed not merely by current demand conditions but also by long-term forecasts, fractional-reserve banking systems will not be prone either to inflationary spirals (as Hayek feared) nor to deflationary spirals (as Keynes feared). The interest rate is an important brake on investment, but quantitatively, the availability of expected profitable investments seems to be the more important limitation, especially in the era of excess reserves.

It is true as well that the availability of financing can be an important hindrance during a crisis too. Where banks are ordinarily able to take advantage of profitable opportunities, a fall in the volume of spending can dry up liquidity and make it impossible to finance what would otherwise be profitable projects. In this case, the problem is not self-fulfilling expectations (per Keynes), but neither is it simply the relative price of capital goods to consumer goods reasserting itself (per Hayek). Rather, a collapse of liquidity has made it more difficult to effectuate positive long-run expectations. People do, after all, generally expect recoveries to follow recessions.

This does, indeed, seem to be the way central banks are looking at the problem of business cycles today: the best response is to provide sufficient liquidity so that expectations can reassert themselves, and in particular that interest rates are low enough that gains from beyond the recovery can be profitably capitalized into the present. Whether they have been successful or not, whether they even have the right tools to do so, is a separate question entirely. But by continuing to cast monetary policy in the light of the Keynes-Hayek debate, we miss entirely the logic behind what monetary policy is trying to accomplish, let alone whether it’s been successful.

In fact, I may even go so far as to say that nothing written prior to the expectations revolution (1950s-70s) on the topic of business cycles is worth reading today. Formative as this debate (along with many others) have been, there is no reason not to avail ourselves fully of the tools and mental models available to us today that were not available at the time.

Footnotes

1. See especially Hayek (1941: 47), where he argues that continuity with Ricardo and Mill is a point in favor of Austrian capital theory and against contemporary Anglo-American capital theory. Hayek (1931; 1933) also express debt to Ricardo’s conceptualization of monetary and capital problems. Gordon (1973) argues that the Ricardo-Mill wage fund construct was essentially dead by the end of the 19th century until Böhm-Bawerk ([1891] 1930) revived it in a modified ‘subsistence fund’ form and moved it to the realm of capital theory rather than wage analysis. That very division, of course, is a Ricardian legacy as well.
2. Other uses in economics, such as ‘social capital’, are more analogies than proper instances of capital, as they do not generally have a rental or sale value.
3. Hayek (1941) is most notable in this respect for attempting to assimilate in this direction in full view of the difficulties of doing so.
4. Hayek (1936) at various points both asserts that capital can only be sensibly measured in value terms and not physical, and criticizes Knight’s use of value units to homogenize capital into a fund. Yeager (1976) retains the “attractive quasi-homogeneity” of capital by retreating to pure theory and defining it quasi-tautologically as the thing (whatever that may be) which embodies waiting as a factor of production. It is not clear that this is more operationalizable than the standard Ricardian setup, or even that it suggests a paradigmatic capital good, though Yeager does seem attached to the circulating capital metaphor.
5. The translator of Hayek (1929) from its original German brings attention to his rendering of Kapitalintensität (an obvious cognate to “capital intensity” in English, though Hayek [1939: 17] thought this rendering “somewhat too literal”) as “roundaboutness” (Ibid.: 234). This does not match Böhm-Bawerk’s ([1891] 1930) terminology, for whom “roundabout production” was translated from Productionsumwegen (“detours in production”). Hayek did, however, use “roundabout processes” in his English work (e.g. 1931).
6. The Cambridge (UK) antagonists to the debate, on the other hand, saw the basic incompatibility between marginalism and the Ricardian income-determination project more clearly, but preferred to preserve the latter by rejecting the former. See Cohen & Harcourt (2003).
7. Ironically, O’Driscoll & Rizzo (1985), the most influential book explicitly advancing Lachmann’s research agenda, contains in its chapter on business cycles (ch. 8) a markedly Hayekian and pre-Lachmannite exposition of ABCT which would later be developed into Garrison (2001). Ebeling (2015), similarly, even states that Lachmann “remain[ed] true to the Böhm-Bawerkian emphasis on capital goods as intermediate goods within time-structures of production”, despite the fact that the lack of this emphasis is precisely what separates Lachmann from preceding Austrian capital theorists.
8. Hayek (1936) in arguing against Knight’s conception of capital as intrinsically perpetual emphasizes that capital maintenance is not routine; that capital equipment is not necessarily replaced with identical items, and that active decisions are involved at each step. This seems like an overstatement. For our purposes, we can say that planned replacement with an identical item is a routine extension of the original plan, and replacement with different equipment would be a relevant time-consuming change to the capital structure in our sense.
9. Interestingly, Hayek (1936, fn. 33) suggests avoiding the use of the term ‘capital’ altogether.
10. Analytically, jointness in production poses the same capital-theoretic problems as the durability of capital. The effect of both is to make it impossible to attribute particular units of output to units of input.
11. Wagner’s (2010) process-focused and non-equilibrium ‘neo-Mengerian’ approach to economics, for example, appears quite comfortable with this sort of simultaneity:

    Suppose that 95 percent of enterprises are operating within their execution phases, leaving five percent of enterprises at nodal positions where they are either creating or revising plans. This kind of situation would generate observations that would fit with the reasonably predictive properties of models of static equilibrium. An established furniture manufacturer that also owned its forests would confront the world in pretty much simultaneous fashion. During any year, or other time span, it would be planting trees, harvesting trees, buying and repairing equipment, and making furniture, all of it appearing to be simultaneous. . . . The source of the motion [in an economy] . . . is the five percent of enterprises not in stasis at any particular instant that are eroding the static reposes of the other enterprises.

12. Cachanosky & Lewin (2016) argue that investments of long duration “do not correlate coherently” with what were understood by previous Austrian authors as higher-order stages of production. Salter & Luther (2016) argue that, in the context of a rational expectations model, whether the boom occurs in higher-order industries or somewhere else has little relevance to the Austrian story. Luther & Cohen (2014; 2016) discuss some of the difficulties in operationalizing the stages of production concept in the context of empirical work.
13. Even such a small set of failures could, in principle, trigger a recession if these failures snowball by forcing financial firms to contract their issues of broad monies and financial assets (see Harwick 2019a). This can cause a decline in real money balances, slow spending until prices fall, and a rise in interest rates due to a scarcity of liquidity. This transmission mechanism is significantly different than the standard ABCT story, however: it would place the primary emphasis on what Hayek (1933) and later writers have largely dismissed as “secondary” deflation (though Horwitz [2000] does place somewhat more importance on it). Very different practical conclusions for stabilization policy follow.
14. Given the role of real business cycle theory in reviving the neoclassical production function following the Cambridge capital controversies (Cohen & Harcourt 2003), a richer capital-theoretic approach has the potential to elucidate the plausibility of its causal claims, as opposed to its statistical fit.
15. Superstition also supports separating equilibria via a similar mechanism in Iannaccone (1992) and Leeson (2013a), where visible commitment to a burdensome superstition filters out noncooperators. These examples also result in separate inside and outside perspectives, and the remarks below on costless signaling will also apply.
16. This assumption is not quite taken for granted: Leeson (2013b) for example builds a Bayesian model of belief, so believers are not totally credulous in the face of crass manipulation, and derives an equilibrium quantity of manipulation.
17. In this sense we are considering belief not as a Nash equilibrium, but as an evolutionarily stable strategy. Similar considerations also militate against the theory that language evolved for the purpose of manipulation or deception (e.g. Dawkins and Krebs 1978). It must be incentive-compatible not only to send a signal, but also to receive and act upon a signal (Fitch & Hauser 2002; Searcy and Nowicki 2005: 8). Knight (1998) takes these considerations and comes to a similar conclusion to this paper, with trust in the veracity of language vouchsafed by the costly rituals implicit in a normative community.
18. Sociality in this sense is distinct from gregariousness (e.g. herding behavior), which is incentive-compatible under certain conditions. Sociality generally depends on a favorable mix of coordination games and social dilemmas in the environment of the cooperating group (Bear et al. 2017), but – as this section shows – the dilemma aspect is irreducible. Because coordination games have stable cooperative equilibria, we leave those to the side and focus on social dilemmas as the more difficult impediment to social behavior.
19. E.g. the classic simulation in Axelrod (1984) which showed the dominance of tit-for-tat when paired against other strategies for some number of periods. Hardin (1985) criticizes his generalization to N-person games. Kandori (1992), similarly, shows that cooperation-sustaining strategies exist for repeated pairwise games where the pairings are sampled randomly from a population, but not for non-pairwise interactions. Alger & Weibull (2013) examine the divergence between preferences and payoffs in a similar spirit to the present paper, but only for pairwise interactions with positive assortativity.
20. For the signaling game in particular, if c_i is an unobservable or imperfectly observable cost that one may bear for the benefit of the group (say, refraining from crime), and the cost of signaling compliance is known to be zero (say, enthusiastically assenting to undergo an ordeal), then the signal’s value as an indicator of c_i will be a commons which free riders will be motivated to deplete by falsifying the signal.
21. Or, equivalently, the probability that any agent assesses another agent to have failed to contribute when he in fact did not, or vice versa.
22. The same argument also applies to inclusive fitness explanations for cooperation, i.e. that altruistic genes can proliferate on the basis of kin selection. As Bowles and Gintis (2011: 60) note, relatedness enters into the structure of payoffs from a gene’s perspective in exactly the same way as the probability of a repeated interaction, which is to say that the relatedness coefficient within human groups must be implausibly high in order for kin selection to support altruistic behavior.
23. The importance of joint production is that it forecloses the possibility of paying by marginal product, as product exhaustion will not hold where each agent’s marginal product is not independent of the effort of other agents. In this situation, compensation on the basis of inputs (i.e. effort) can be more feasible than compensation on the basis of value added. The fact that this requires monitoring is what creates the incentive gap.
24. Ostrom’s (2005: 259) famous design principles for the management of common pool resources, particularly the ones relating to monitoring, sanctions, and punishment, presuppose altruistic preferences of some form or another. In this crucial respect, Ostromian agents depart from the standard homines œconomici. See the following section.
25. This argument would apply to any mismatch between the level of selection and the locus of decisionmaking, including inclusive fitness explanations (see above, Footnote 8), where altruism is selected for at the gene level. Thus eusocial insects are excluded, not because their altruism arose from a different selective process, but only because they do not make deliberate and conscious decisions.
26. Bear, Kagan, and Rand (2017) show that deliberation (the process of rationally assessing one’s interests) leads to lower levels of cooperation, and that cooperative strategies are nevertheless pervasive, but generally non-deliberative. Alger, et al. (2018) offer a formal model showing that exactly such a divergence can sustain social behavior.
27. Miller is concerned here to show that the incentive gap in the firm can be closed by a non-maximizing “company culture” which allows credible commitments. This constitutes the divergence necessary to approximate Pareto-optimality in joint production.
28. See below, Section 4.1.
29. This suggests a novel interpretation of the rise of scientific rationalism (in the Weberian sense) in the West as a transition from belief-based cooperation to preference-based cooperation. This interpretation is supported by the facts that (1) scientific rationalism has been accompanied from the beginning by persistent worries of social decay, (2) that decay has so far failed to materialize, at least in terms of organizational capacity, and (3) that people from Weberian-rationalist cultures do seem to have a stronger preference for altruism (specifically, they are far more generous in one-shot dictator and ultimatum games than those from more traditional cultures – see Henrich et al. 2010). I will not pursue this line of thought here, however.
30. Krugman (1993) is a particularly self-aware example. The real point of classical trade theory, Krugman argues, is not that tariffs can never be welfare enhancing, but to obscure opportunities for rent-seeking that an “optimal tariff” policy would illuminate. Buchanan and Wagner (1977) lament the eclipse of classical public finance principles by Keynesian aggregate demand management on the same basis.
31. Buchanan, at least, seems to have been self-aware on his assumed role as Noble Liar: “Our normative role, as social philosophers, is to shape this civic religion” (Brennan & Buchanan 1985: 166). See also Brennan and Buchanan (1988) and Leeson (2018).
32. From the perspective of this paper, it could be argued that the solution to Hobbes’ dilemma is not an overawing Leviathan – which, per above, poses its own dilemmas – but the fact that humans are apt to generate and internalize Locke-esque ideologies. Locke’s work is not itself an effective answer to Hobbes, but the existence of Locke’s work is.
33. Foucault (1978) is one of the ur-texts for this expansive understanding of power, one that attacks the inside-outside perspective distinction more directly than previous conceptions which have often focused on coercive power (which, per above, is exercised relatively infrequently on the equilibrium path of play). Foucault himself can be read as having some appreciation for the functional role of the exercise of power in his sense, but subsequent literature has been predominantly liberationist. Critical theory should be distinguished from orthodox Marxism in its rejection of dialectical materialism: power, for the Marxist, is epiphenomenal to modes of production.
34. As an illustration, in recent years critical theory has been aimed at the institutions of science, even the scientific method itself, as upholding certain power structures. The “objectivity” of science is derided as a self-serving myth. It is of course true that science does not grant the scientist an Archimedian vantage point from which to view the world. It is, rather, a process of replacing descriptions of objects in terms of our senses with descriptions in terms of other objects (Hayek 1952: 3) – a process which can in principle never reach perfection, and one which benefits some interests over others. The benefits of science, like the benefits of society in general, are vast, but predicated on a prosocial myth – in this case, objectivity.
35. A large proportion of Islamic fundamentalist leaders, for example, are Western educated (Devarajan et al. 2016), not traditionalists in any meaningful sense. Similarly, scientists and engineers (whom we may take as exemplars of education in the Western scientific-rationalist tradition) are dramatically overrepresented among extreme Hindu nationalists in India (Lutz 2007: 151). The West has also seen its intellectuals swept by waves of ideological extremism, for example in the attraction of fascism and communism.
36. Hayek acknowledges that expected returns are important, but takes it for granted that an increase in the availability of funds signals that the future demand will be there.
37. Keynes acknowledges that financing is important, but takes it for granted that savings “will always keep pace” with investment (“Alternative Theories of the Rate of Interest”).