Macro and Micro Economic Counterfactuals and Performativity

I am really interested in the idea of counterfactuals and their place in social scientific knowledge. Mapping out the space of possibilities* plays an important, explicit role in many kinds of modeling, and an implicit role in many historical accounts (such as an excellent history of economics by Yuval Yonay that I finished today and will likely write up for here soon). Today I started thinking about the connection between counterfactuals and the performativity of economics. Two examples.

In Harrison White’s famous and semi-whacky model of markets (e.g. White 1981), producers make decisions based on the observed production schedules of other firms and their own known cost functions. Producers do not, in any sense, rely on models of demand curves. Mark Mizruchi explains this modeling decision as a recognition that demand curves are “completely counterfactual” and thus not a solid basis for decision-making. A producer cannot know what the demand for a product would look like given a change in quality, cost, etc. So White’s producers don’t do that. They make decisions on observables.

I accept that demand curves are counterfacturals, but perhaps not all counterfactuals are unobservable. Perhaps the technologies (concepts, ideas, formulae, data, etc.) of modern economics allow producers to “see” demand in a curvy fashion?

Here’s a more concrete example. Right now, congress is considering another fiscal stimulus package (Brad DeLong’s “Keynesian fire alarm“). Today, Paul Krugman wrote up a justification for such a move and a suggestion for how to calculate its magnitude. Some quotes:

Wait, there’s more. Ben Bernanke can’t push on a string – but he can pull, if necessary. Suppose fiscal policy ends up being too expansionary, so that real GDP “wants” to come in 2 percent above potential. In that case the Fed can tighten a bit, and no harm is done. But if fiscal policy is too contractionary, and real GDP comes in below potential, there’s no potential monetary offset. That means that fiscal policy should take risks in the direction of boldness.

Krugman explicitly models a counterfactual – what real GDP “wants” to be, but can’t attain due to unemployment above the “natural rate”. Okun’s law tells Krugman that “every excess point of unemployment above 5 means a 2% output gap.” So, econometric data + economic theory = visible counterfactual on which to make policy. GDP “should” be a few percent higher than it will be next year, so government should make up the gap – and if it overshoots, the Fed can fix things by raising interest rates.

So, is this an example of performativity? Will economics have performed the “natural rate of unemployment” and the idea that GDP “wants” to be higher than it will be if, using justifications along Krugman’s lines, a large fiscal stimulus is passed that lowers unemployment?

*A term suggested to me by new UoM Prof. Elizabeth Bruch when we were talking about performativity, and the general problem of models that create things.



  1. Good post. I’ll be borrowing that idea of ‘mapping out the space of possibilities’ – it captures perfectly the whole rationale of the efficiency of environmental markets – a strange (and inherently contradictory) double movement: markets are necessarily the most efficient, yet we can (and must) discover in advance the cost of alternative approaches.

    This underpins claims like ‘the US Sulfur Dioxide Trading Scheme saved 50% of the alternative approaches that would have been implemented (command/control)’

  2. Dan,

    My question for this post is: what exactly do you mean by counterfactual? Frankly, most of economics is made of counterfactuals, if you define a counterfactual as a scenario delimiting events that are possible, but which will not happen. In fact, by that definition, every theory at all is implicitly a counterfactual (they all tell us what can and cannot happen).

    Also, is there a difference between a counterfactual and a latent-factual (obviously not a real word). By this, I mean to say, what is an “observable counterfactual” but a latent variable, which can be determined because of what has happened in the past, or from correlative movements between two things that are constrained by something that we cannot see.

    I tend to think of counterfactuals as being like: what if Hitler had won in Europe? This is a situation that I know could not happen (because it didn’t), but which it is interesting to think about, because, by laxing a few constraints on reality, it allows me to think about some interesting stuff I previously couldn’t have. But is the Krugman case one where we need to “lax constraints on reality”? Or is it one where we just need to find something we believe to be real, but cannot see?

  3. Jeff,

    Good points all around. I wish I had a clearer definition of counterfactual, maybe I can find one in the literature or invent one. What I mean here is something fairly specific, and indeed, specific enough that it could be used to make predictions, and yet is unobservable. So, while Harrison White’s model is a thoery that can be used to make predictions, and thus is in a sense constructing counterfactuals (which is what you are pointing to, I think), it draws on only observable data to the person presumably making the prediction. White’s critique of neoclassical econ is that demand curves are unobservable, and thus worthless for predicting changes (that is, constructing models whose counterfactuals are predictions).

    Krugman’s example of what gdp “should be” is another counterfactual that is very specific – but for the increaed unemployment caused by the financial crisis, GDP “should be” a few percent higher. Thus, we can restore it to where it ‘should be’ by activating the Keynesian fire alarm. But we can’t observe, for example, the independent effect of the financial crisis net other variables. But maybe Krugman’s model is more like White’s than like a demand curve, as the input (unemployment) is presumed to be visible.

    In any event, Krugman’s model constructs a counterfactual that is very specific and can be used as the basis for action. By framing the world in a particular way, we can then exercise a new kind of calculative agency (all in Callon’s (1998) terms), in this case, calculating the size of the bailout and then doing it. If we do, and the bailout restores GDP to what it ‘should be’, we could then argue that the model has been ‘performed’. No?

  4. “Perhaps the technologies (concepts, ideas, formulae, data, etc.) of modern economics allow producers to “see” demand in a curvy fashion?”

    Energy economics and the development of “demand response” (or “real-time demand-side infrastructure”) research and pricing schemes for electricity utility companies are a good example.