On Friday, November 3rd, Professor Rodrik is in Room for Discussion to discuss Trade and Globalization, a topic he has discussed at length in The Globalization Paradox (2011). What I appreciate about Rodrik is that he is highly self-critical and openly admits that he, too, was overoptimistic in the run-up to the financial crisis. As he puts it: ‘Along with the rest of the economics profession I too was ready to believe that prudential regulations and central bank policies had erected sufficiently strong barriers against financial panics and meltdowns in the advanced economies, and that the remaining problem was to bring similar arrangements to developing countries. My subplots may have been somewhat different, but I was following the same grand narrative’ (2011: xii). According to Rodrik, however, the dismal failure of this grand narrative is not necessarily a sign of economics failing as a science. It merely shows we were placing too much faith in the wrong models at the wrong time. The problem thus lies in model selection, rather than being caused by flaws in these models themselves.
He explains his position further in his later book Economics Rules (2015). In it, he argues that we need to distinguish between critical and non-critical assumptions. In his view ‘an assumption is critical if its modification in an arguably more realistic direction would produce a substantive difference in the conclusion produced by the model’ (2015: 27). These critical assumptions have to correspond to reality, while the non-critical ones need not (and may well appear absurd to the uninitiated). The critical assumptions also guide our selection of models. So when confronted with some empirical problem, like – say – disappointing growth figures in a particular country, we examine the many possible impediments to growth (such as lack of trade, training, infrastructure, property rights enforcement and what not) and the models that map out the consequences of these assumptions and check whether the growth barrier assumed in a specific model corresponds to that country’s reality. To verify that this is the right model to use, one can further check whether the model’s many (direct and incidental) implications are also borne out in reality. If so, we can use the model to see what it predicts will happen if that growth barrier is removed. If the results conform to the goal that the politicians in that country have set, we can safely advise them to go on and start working on removing the barrier in question. Models thus are best viewed as diagnostic tools, and as such they can be properly or improperly used, but cannot themselves be right or wrong.
Now this sounds like a prudent way of applying models to reality, but the problem is that many assumptions that are critical in Rodrik’s sense cannot actually be modified ‘in an arguably more realistic direction’ at all, because doing so, would invalidate a whole class of models. If households are not identical for instance, models invoking individual rationality cannot generate (predict) a unique market outcome at all, and would thus be useless as diagnostic tools. (If you follow my columns (1 and 2), you may be getting a bit weary with me talking about the SMD conditions by now, but its implications are so pervasive and so often downplayed, that I feel they can never be overexposed.) Similarly, People like Minsky (1992) and Keen (e.g. 2017) have shown that when the idea of loanable funds is replaced by the much more realistic idea that investment and concomitant money creation is a function of prospective profits, the dynamics of the system change completely. Yet growth diagnostics (a field that Rodrik proudly hails as being a prime example of the correct use of models) almost always hinges crucially on this doubtful idea of loanable funds. So, although Rodrik doesn’t say it, some critical assumptions remain in force, not because they are reasonable non-disruptive simplifications of the underlying problematic, but because neoclassical ideology forbids contemplating alternatives.
As to the latter, Rodrik is crystal clear that the model library that economists are to select their diagnostic tools from, has to be neoclassical, as he writes in yet another book on globalization: ‘At the core of neoclassical economics lies the following methodological predisposition: social phenomena can best be understood by considering them to be an aggregation of purposeful behavior by individuals—in their roles as consumer, producer, investor, politician, and so on—interacting with each other and acting under the constraints that their environment imposes. This I find to be not just a powerful discipline for organizing our thoughts on economic affairs, but the only sensible way of thinking about them’ (2007: 3).
From a modeling perspective, however, this is neither the most powerful, nor the only way to think about social phenomena. Models can only function as diagnostic tools if the entities and transmission mechanisms critical to getting the diagnosis right are stable while diagnosing and correspond to real world stability and it just so happens, that individual predispositions and actions are much more volatile than aggregate ones. In fact, it was the resilience of suicide rates among specific social groups (Durkheim 1979 [1897]), that spawned the emancipation of sociology as a scientific field. Apparently, social phenomena are stabilized more by social pressures in certain groups than by stable, predictable individual inclinations. Durkheim showed, for instance, that soldiers have higher suicide rates than civilians. So if civilians become soldiers, the overall suicide rate rises. Similarly, consumption of certain positional goods (goods that showcase your status) probably depends more on the relative size of high status groups, than on individual inclinations (cf. Bourdieu 2010 [1984], Ilmonen 2011, Miles 1998). But this implies that in many cases ‘purposeful behavior by individuals’ is a far less powerful discipline for organizing our thoughts on economic affairs than are ‘social pressures emanating from social groups’.
All in all then, Rodrik’s defense of economics fails, because his predisposition towards neoclassical economics precludes him from modifying many critical assumptions in a more realistic direction. It was a good try, but it will take an even more open minded economist to take the discipline into a truly fruitful direction and remove the barriers that currently prevent economists from producing models that have something to do with reality. But this will only happen if all models, no matter whether they originated in Marxist, institutional, post-Keynesian, Austrian or other thinking, are accepted as being on a par with neoclassical models. If that happens, Rodrik’s methodological ideas just might work.