Critiques of mainstream macroeconomics are common among Austrian economists. In a new book, titled Macroeconomics as Systems Theory, Richard Wagner goes further. He starts with Erik Lindahl’s distinction between microeconomics as individual action and macroeconomics as interaction. He then offers a new approach to macroeconomics based on theories of complex systems.
Having taken courses with Wagner in grad schooI and attended seminars and workshops with him in the time since, I have learned that I often come to appreciate his thoughts long after first encountering them. I will leave you to judge whether that reflects the depth of his thinking or the shallowness of mine. Perhaps a bit of both. In any event, you should consider my thoughts provisional at this point.
I find myself in a strange position. On the one hand, I could not help but nod along while reading Wagner’s book. I fully embrace Lindahl’s distinction between microeconomics and macroeconomics. I think it is important that we focus on the plans of individuals; that we recognize those plans are often frustrated by chance or circumstance; and that we appreciate how plans are revised in response to such frustrations in the normal course of things. And I am similarly put off by the all-too-common suggestion that we can fix this or that outcome with crude policies like prohibitions on behaviors we find repugnant or government spending programs aimed at reducing the so-called output gap––where, more often than not, “we” refers to the modeler and not actually the individuals involved in the political process.
Indeed, much of my own work has tried to highlight these issues. I am a monetary economist. It is really hard to make sense of a commonly accepted medium of exchange if there is only one agent––or one agent type––in your model. Interaction is key! And interaction is all the more important when trying to launch an intrinsically worthless money. As I explain in a paper titled Getting Off the Ground: The Case of Bitcoin, individuals deliberately coordinated to make bitcoin valuable and thereby enable others to use it as a medium of exchange.
I have also been explicit about the planning and the revision of plans. In surveying the literature on Austrian labor theory, Peter Boettke, Solomon Stein, and I describe the entrepreneurial process in these terms. In another paper, on Mises’s evenly rotating economy, I suggest that the relevant choices in a macro model pertain to strategies, or plans, to achieve some outcome rather than choices over the outcomes themselves. Alex Salter and I make a similar argument in a paper titled The Optimal Austrian Business Cycle Theory.
Finally, we are perhaps most in agreement when it comes to our approach to evaluating policy. “To prohibit something that many people like to do will generally not be accepted passively by those who produce or consume the proscribed items,” Wagner (p. 25) writes. “Prohibition will increase the cost of doing business, leading to some decrease in supply and reduction in consumption. But it will also induce multiple changes in commercial organization and operation as people seek out alternative channels for fulfilling their desires without being detected and jailed.”
Along these lines, Josh Hendrickson and I have explained that banning cash would likely push those currently using cash in illicit markets to adopt alternatives with similar privacy-protecting features, like cryptocurrencies.
In some sense, it seems I have fully embraced the Wagnerian view. But I wouldn’t have written “on the one hand” if there were not another hand coming around.
On the other hand, my work looks very different from Wagner’s. I like deterministic models. While I often disagree with other macro- and monetary economists on important issues, I don’t think the models employed in these fields need to be scrapped; I don’t believe we need a radical rethink. Indeed, I believe much of Wagner’s work is compatible––if somewhat out of sync––with mainstream macroeconomic theorizing.
Naturally, while reading Wagner’s book, I found myself trying to resolve this tension. While I have not done so to my complete satisfaction, I think I have made some progress along those lines.
Wagner rejects the synchronous, universal approach to macroeconomics, of which he sees DSGE models as an exemplar, in favor of a diachronic, particular approach. By universal, he means “the presumption of deep-level homogeneity among people” (p. 44). By synchronous, he means involving an equilibrium, where “the plans of all the relevant entities are presumed to be synchronized with one another” (p. 45). His preferred approach, open-ended evolutionary (OEE) modeling, involves non-equilibrium theorizing about heterogeneous agents.
A simple two-by-two matrix of Wagner’s two dichotomies makes it clear, however, that (1) synchronous, universal and (2) diachronic, particular approaches are not the only options available. My preferred quadrant is the synchronous, particular approach, which involves equilibrium theorizing about heterogeneous agents. Indeed, I would argue that a lot of modern DSGE models fall into this quadrant.
|DSGE Models with Representative Agent|
|DSGE Models with Heterogeneous Agents||Wagner’s OEE|
A DSGE model, as you might know, is a dynamic stochastic general equilibrium model. It is, in fact, a large class of models. It is a large class of models because it is defined so broadly. It must be dynamic––that is, have some notion of how the economy changes over time. It must be stochastic––that is, have some notion of randomness. It must be general––that is, referring to the economy as a whole. And it must involve the sort of equilibrium theorizing, as discussed above.
DSGE models are not limited to Robinson Crusoe-type representative agent models. As Lawrence Christiano, Martin Eichenbaum, and Mathias Trabandt explain in the paper On DSGE Models, they have progressed well beyond the simple Real Business Cycle and New Keynesian models of the 1980s and 90s.
A lot of work in macroeconomics today makes use of heterogeneous agent DSGE models. I am most familiar with the work in monetary economics and, in particular, the work that has followed Kiyotaki and Wright’s seminal papers on a search-theoretic approach to money. Those models began with three types of agents encountering a double coincidence of wants problem as a result of randomly bumping into potential trading partners rather than meeting in a common, Walrasian market. They have been expanded to include an infinite number of agent types; endogenous matching, where agents choose their trading partners; and the coexistence of pairwise matching and common market exchange relationships.
Ricardo Reis provides further examples of heterogeneous agent DSGE models in a 2018 survey titled Is Something Really Wrong with Macroeconomics?
As Wagner and I agree that models should generally include agents with different wants and abilities, I will focus on our point of departure. In particular, Wagner prefers a nonequilibrium, or diachronic, approach whereas I am perfectly content with equilibrium theorizing.
Wagner’s primary complaint with equilibrium theorizing is its deterministic nature. They are “a robotic model of social systems,” he writes (p. 53). “It is not,” he says, “a model that assimilates to recognition of the emergent creativity that can arise through social interaction, or to recognition that different institutional frameworks can generate differing societal properties.”
It is true that DSGE models are deterministic. But they are perhaps better described as conditionally deterministic. Recall that the S in DSGE stands for stochastic! As a result, they are deterministic conditional on the realization of probabilistic events and any exogenous shocks encountered along the way.
When the agents in a model make their equilibrium plans, they do so with their best guesses regarding the likelihood of different states of the world. Over time, those uncertainties become certainties, as outcomes are realized. As this information becomes available, some agents will see their plans unraveled and they will have to revise those plans. Others will see their plans have paid off well, leaving them with a certain balance that they had only anticipated probabilistically before. These agents might then choose to let it ride or cash out or something in between. The ex-ante plans are synchronized, given the available information. But that information is uncertain. And the outcome is nonetheless emergent. Indeed, if we ran the model again we might reach a very different result, as the outcome depends, in part, on a dice roll.
Does the fact that agents cannot possibly calculate the relevant probabilities ex-ante “due in large measure to the combinatorial arithmetic of situations requiring agents to make choices” (p. 75) undermine the case for equilibrium theorizing? I don’t think so. Just as Armen Alchian explained we keep buffer stocks and leave resources unemployed to economize on information costs, so too might we adopt rough heuristics or flip a coin to get around the high costs of forecasting the relevant probabilities.
Does equilibrium theorizing “deny boredom, imagination, and creativity” as Wagner (p. 48) claims? Again, I don’t think so. In a paper titled Endogenous Matching and Money with Random Consumption Preferences, Thomas Hogan and I incorporate the idea of individuals acting on a whim in a deterministic general equilibrium model. Others have made similar advances along these lines.
To illustrate the ability of DSGE models, broadly understood, to capture the interesting characteristics of emergent phenomena, I want to briefly recast Wagner’s famous example of a crowd leaving a stadium. There are different types of agents. Some want to return home promptly to tuck their kids into bed or to get enough sleep before going to work the next morning. Some are impatient or hate feeling idle. For others, the night is just getting started. Some walk quickly. Others slowly. And no doubt a few will be doing well just to manage a stagger.
At some time, t, prior to the end of the game, each adopts a plan of action. They don’t usually discuss their plans with others outside their group, if at all. But their plans are nonetheless synchronized in some sense as they are based on their estimates of the relevant probabilities. Many of these folks have been to a game before and realize that the bathrooms tend to be crowded on the way out. “Best to go early,” they think. Others enjoy seeing the players exit the field or dislike close contact with strangers, so they plan to wait for a few minutes until the crowd thins out. Many understand the dynamics of crowds. They know that sometimes they bottleneck and sometimes they don’t. They don’t know who will come out of which stairwell at what time. That’s inconsequential to their plans. They need only think about things like the likely flow of the crowd and how they might adjust their steps as others merge in and out around them.
As the crowd starts filtering out, some of the unknowns become known. There was a chance that the hot dog vendor would leave his cart in the way, as happened once before. But that is not the case tonight, so those who were concerned maintain or perhaps even pick up speed––all according to plan. Others are surprised when someone cuts in front of them––surprised, but not totally unexpected, as that sometimes happens. They slow down according to plan.
Some might get to a gate that an usher forgot to open. They will have to revise their plan and head to another gate. This negative productivity shock will set them back a bit. And some will see a neat, vintage jersey that another fan is wearing and realize that they, too, would like to own a jersey like that. An exogenous preference shock, if you will.
This equilibrium theorizing might not tell us who will get to the gate first on any given night. But it yields pretty good predictions about the flow of traffic. Such a model might inform the vendors to pull their carts to the side before the bottom of the 9th, so as not to upset fans (and, as a consequence, their employer) by creating unnecessary congestion.
Most importantly, it would make sense of the beautiful mosaic that forms as people exit a stadium––all of which is the result of individual plans and not the design of an outside organizer but is nonetheless orderly in its own way. Our equilibrium model, in other words, does much the same thing as Wagner’s OEE. It just describes the situation using a different language.
I probably hold an idiosyncratic view of mainstream macroeconomics. No doubt many of the criticisms Wagner makes apply reasonably well to many practicing mainstream macroeconomists. My claim is a modest one: Wagner’s criticisms do not apply to all practicing mainstream macroeconomists. I would go further to add that they do not apply to the best practicing mainstream macroeconomists.
By pointing this out, I do not intend to reduce the importance of Wagner’s work. To the contrary, I hope his book will be widely read. I merely want to suggest to those already sympathetic to Wagner’s views––those who recognize the distinction between action and interaction and the importance of that distinction in spontaneous order theorizing; those who appreciate the entrepreneurial planning process and the creativity of human agents; those who are skeptical of top down solutions premised on the idea that some undefined we can act on society––it is to suggest to those sympathetic scholars and students that the solution is not to tear down all that has been built in macroeconomics, but to build bridges connecting Wagner’s interesting ideas to their counterparts in the profession.
Note: An earlier version of this essay was presented in a panel discussion of Richard Wagner’s new book hosted by the Mercatus Center at George Mason University. A recording of the event, including Wagner’s response, is available online.