Well, Grasselli has responded to my previous post. His response, full as it is with personal attacks, bile and insinuations is pretty embarrassing; Grasselli hasn’t yet learned that a rapid, heated exchange on Facebook is a little different from a blog post that is supposed to be carefully thought through. But I won’t engage Grasselli’s clear seething dislike of me because I think such would be pointless — infinitely amusing, especially his chess club insults (small brain etc.), but pointless.
The best way to approach this is by pulling out the substantive points in an orderly manner.
(1) Grasselli says that I am a hypocrite because I say that economists should make predictions but that they should not engage in crass games comparing various models. This is apparently a contradiction because, according to Grasselli, you need models to make predictions. Well, I make economic forecasts all the time and I never use models. This appears to blow Grasselli’s mind but this is only because, as discussed in the first post, he is a mathematician and not an economist so he doesn’t really know the trade.
Let me state this point clearly so that everyone understands what I’m saying: I do not use models for prediction; I do not believe that models are a good means by which to make predictions; and ultimately I think that models are only really didactic tools, “classroom gadgets” as John Hicks once said. (I have, by the way, also built models, so I’m not saying this from the standpoint of someone who eschews them; I just think them to be no more than didactic tools). Grasselli will likely not understand this point but that is on him, not on me.
(2) More importantly Grasselli has misunderstood what I meant when I said that his models cannot be used in meaningful empirical work. He says that his models can make precise predictions. He then discusses frequentist and Bayesian statistics. (He also, in a feat of admirable narcissism, seems to think that I refer to this as “the Grasselli approach”, although a close scrutiny of the text reveals no such statement).
Look, I’m not going to get into a debate over what Bayesian theory is or isn’t. There’s only two ways that you can use a model to generate a prediction. One way is just to assume that a model represents reality. For example, a New Classical economist might use their model to say that an increase in government spending will lead to inflation in the medium-to-long run. The other approach is to feed historical data into the model and try to extrapolate based on this — again there is the assumption that the model is correct, but it is being used to “process” data and project past empirical trends forward rather than make a priori predictions. The latter approach, which as Grasselli notes is the “frequentist” approach, fails due to economic data being non-ergodic.
Grasselli instead does something more so in line with the former approach. As he writes:
In Bayesian statistics, the modeler is free (in fact encouraged) to come up with her own priors, based on a combination of past experience, theoretical understanding, and personal judgment.
I was well aware of this when I wrote up my criticisms. This is precisely what I meant when I said that Grasselli’s aim was not what I consider real empirical work but rather the testing of his model. He comes up with a model based on what he calls “priors” and then tests this against the data. Why do I not think that this is real empirical work? Because it is ass-backwards. The point of Grasselli’s approach is not to discover new novel facts that might tell you how the economy is evolving through time but instead to confirm or disprove already held knowledge. Since I do a good deal of empirical work I know that this approach is basically useless; the scope is much to general.
A good applied economist is in the process of always discovering new facts buried in the data; not banging his model against the wall until he breaks through. Grasselli will, of course, not understand this criticism because, as we have already seen, he thinks that you need a model to make predictions. But I cannot help him here. It is not his field. This is why I don’t think that mathematicians should be invited into the tent. I don’t pretend to know how to do good mathematical theory, but for some reason a lot of mathematicians seem to think they know how to do applied work in economics. And then their jaws drop to the floor when I tell them that the approach to empirical work which I adhere to is not model-based but rather, to again quote Keynes (I’m not just doing this for fun…), to have an “organised and orderly method of thinking out particular problems” with which we then scrutinise the relevant facts and data.
(3) With regards to the next substantive point I think that I should quote Grasselli in full because whereas I am fairly confident that on the last point Grasselli really did not understand what I was saying, on this point I think that he is being evasive:
Pilkington seems to think that the only way to measure something is to go out with an instrument (a ruler, for example) and take a measurement. The problem is that risk, almost by definition, is a property if future events, and you cannot take a measurement in the future. ALL you can do is to create a model of the future and then “measure” the risk of something within the model. As Lady Gaga would say “oh there ain’t no other way”. For example, when you drive along the Pacific Coast Highway and read a sign on the side of the road that says “the risk of forest fire today is high”, all it means is that someone has a model (based on previous data, the theory of fire propagation, simulations and judgment) that takes as inputs the measurements of observed quantities (temperature, humidity, etc) and calculates probabilities of scenarios in which a forest fire arises. As time goes by and the future turns into the present you then observe the actual occurrence of forest fires and see how well the model performs according to the accuracy of the predictions, at which point you update the model (or a combination of models) based on, you guessed it, Bayes’s theorem.
Again, Grasselli is telling me things that I already know. Of course you cannot measure the future. What you can do, however, is make predictions about the future based on an analysis of past probability distributions (provided that the data is ergodic, which in this case it is not, but I won’t get into that right now). My criticism of Grasselli was simple: he does not have such a measure of risk. The article title that I quoted said that Grasselli had devised a “better way to measure systemic risk” but when pressed on it Grasselli could not give me any such measure. Indeed, Grasselli later claimed that one should “estimate nothing”.
The title of the article was completely misleading in this regard. When an investment professional hears someone claim that they can measure systemic risk they will obviously come to think that this measure can be applied in some concrete way. But Grasselli’s cannot. All he can do, as we have seen, is continuously test his model over and over again to prove or disprove it. Anything that Grasselli says about the level of systemic risk in the real world will simply be based, as we have seen, on his “own priors, based on a combination of past experience, theoretical understanding, and personal judgment”. But given that Grasselli is not actually an economist and clearly does not understand how to do robust empirical work I do not see why we should trust such “priors”.
This gets to the heart of what I think a lot of this modelling does. I think it provides a “wow” factor. A modeler flashes their model in the face of an investment professional or a politician and they are entranced by its complexity. The modeller then proceeds to give them advice. The politician or investment professional then believes that the advice is coming out of the model — which is now imbued with a mystical aura — when in fact the advice is coming from the modeller’s “own priors, based on a combination of past experience, theoretical understanding, and personal judgment”. This process, one that has been noted many times before by Post-Keynesians, strikes me personally as being extremely phony.
(4) Let’s wrap this up with Grasselli’s discussion of the Keynes quote I put forward. I laid out the Keynes quote because it is a very well-known quote in Post-Keynesian circles regarding methodology. I fear that Grasselli has missed the key part of the quote where Keynes says “the object of our analysis is, not to provide a machine, or method of blind manipulation, which will furnish an infallible answer, but to provide ourselves with an organised and orderly method of thinking out particular problems”. This is not a criticism of mathematics, as Grasselli seems to think, but a criticism of modelling as such. It is a recognition that models are limited didactic tools that cannot be applied directly to the data. When we approach the data we do so without models but instead with an “organised and orderly method of thinking out particular problems”.
What Grasselli does is approach the data purely through the model. The model becomes a sort of stand-in or prosthetic limb with which he handles the data. This is not the generally accepted method of empirical economic work among Post-Keynesians (although there are a couple of people who do this).
What frustrates me so much about Grasselli is that he doesn’t really understand what Post-Keynesian economics is all about. He has raided it for a few ideas but he has not engaged it in any deep and meaningful way. Every Post-Keynesian who reads this post will understand precisely what I am talking about (even though some, for whatever reason, might side with Grasselli) but there is a very good chance that Grasselli will not because he is simply not familiar with the way we do economics.
At the beginning of his post Grasselli says that I fear that he will destroy Post-Keynesian economics. I fear no such thing. I think that his work will fade rather quickly. What I do fear is that he will mislead younger aspiring Post-Keynesians and waste their time when they could be learning how to do economics. And what I also fear is that he might become a flashy, mathematicised representative of the discipline making claims that I think grandiose and applying ideas poorly. That’s what I fear.