Thought for the day: Horatio economists
from Merijn Knibbe
There are more things in heaven and earth, Horatio,
Than are dreamt of in your philosophy.
Recently, two Dutch economists, Johan de Jong and Jasper Verbruggen, published an analysis of the predictive power of economic models (http://www.tpedigitaal.nl/assets/static/1_Jong-Verbruggen-1-2011.pdf).
Their results are clear:
1. Models are to some extent able to predict the present and the recent past (i.e. ‘predicting’ present consumption or inflation before the official estimates are available), but they do lousy when it comes to predicting the future (even three months ahead is difficult).
2. Their analysis clearly shows the biased nature of ‘mainstream’ concepts and definitions (my translation):
“Available analyses of the effectivity of predictions of DSGE models are restricted:
to pre-credit-crisis models … these show that the number of variables which can be predicted by these models is limited …as, for instance, most … models exclude the possibility of involuntary unemployment”
They mention F. Smets and R. Wouters
3. A comparable statement can be made for leaving ‘flow of funds’ data out of traditional models. Including ‘flow of funds’ data into models (for instance: mortgage debt per house-owning household) clearly enhances the ability of economists to identify increasing risks (though it stays almost impossible to predict the exact time of an ‘economic earthquake’ or Minsky moment).
Models, by their nature, of course always leave out variables. That’s not the point. There is no need to include gravity into economic models. But is it wise to leave household debts and involuntary unemployment out of models? No, of course it isn’t. Variables like those are absolutely central to well being and the working of an economy. But it does come in handy, when you want to introduce austerity policies. It’s even a great thing when you want a government to cut back on wages and to dismiss tens of even hundreds of thousands of people as the banks have to get their money: according to the model, all these people are not really out of work, they just don’t want to find a new job! I can understand that people like that idea. But I don’t understand it when economists are not even aware of the characteristics of these kinds of models and really do think that the real world behaves more or less like them, as seems the case with many a DSGE economist. Let’s call the Horatio economists.
Awfully interesting conclusions despite the fact that all those weaknesses of the models were already known following their inability to predict the great economic crisis of 2007. However, the above research is useful because it concretises the ill points of the models and so there is some hope that the Economic Faculties, of all over the world, will abandon the model’s teaching as central item of the economic science.
“Models, by their nature, of course always leave out variables. That’s not the point. There is no need to include gravity into economic models.”
That’s almost beautiful. But is there a need to include earthquakes? Or earthquakes where tsunami’s tumble nuclear power plants….?
I would have thought so Dave – but there is the ongoing problem of models written by corporations for corporations that seem so pretty to the untrained eye (but in reality are pretty vacant).
The “vacancy” is the complement of what seems obviously not needed – the point about gravity, a “natural”, non-economic term. So the issue is identifying all the parts of reality that must be turned into economic terms and written into that great nomo-logical model in the sky. I’ve been having some fun trying to see how that’s down for those “value added” regression models in education – putatively robust enough to separate all the other factors which affect a child’s performance in school such that his/her teacher’s unique contribution can be measured. I tell people: Ya know, the economists who claim these models measure and predict are of the same discipline that gave the world the models that have trashed the economy, and I get either of three responses: Bemused concurrence, confusion, or denial.
We are obsessed with the need for buyers to decide what to buy in order for government’s to decide what to do.
WW II illustrated what the opposite obsession offers: government decides to protect democracy from tyrants. It uses full employment to maximize output of weapons for war — this would also work for the weapons of peace — if we were obsessed to make peace work to keep war at a proper great distance.
There are (at least) three parts to the art of modelling.
One is to include enough of the real world to gain useful insights from the models. Equilibrium models manifestly fail this test, because economies are far from equilibrium and their behaviour is far more complex than pendulums gently returning to the equilibrium point.
The second part is not to put too many variables in the models, because you may not be able to understand why the model behaves as it does. Therefore you will not gain insight into what drives the model, let alone the world.
The third part is keeping in mind what is not in the model. Indeed, earthquakes cannot be in models, because their timing cannot be predicted. Yet we know they will happen. Therefore there is no point in making models highly elaborate, because what is left out of the model (no matter how elaborate) will intervene and make the model irrelevant (for the time being).
Although the behaviour of economies is complicated, well-posed models may still give us useful insights into how they work, even though we can’t predict them in detail.
“Indeed, earthquakes cannot be in models, because their timing cannot be predicted.” So, when Ameren UE (here in Missouri, wants to collect money from rate payers to build a second nuke plant because they/it don’t want to ding their investors, or borrow capital) says (in paraphrase, but I’ll use quotes) “No worry, the chance of an earthquake leading to a catastrophe are so small as to be unrealistic for our consideration”, the reasonable response should be what?
What’s the expected utility of a utility’s catastrophe if the owners say some of what COULD affect that measure is unmeasurable?
Maybe THAT is the obvious point I am stumbling toward. But when we all concur that models always miss something, I want to know who benefits from what’s been missed.
Dave – not sure where you’re going here. My point is that models can be useful guides, but we must keep in mind all the things not in the model. So there has to be a judgement about low-probability, high-consequence events like earthquake-nuclear meltdown. My own judgement is that nuclear is never worth it, because there’s a cheaper, quicker, far less dangerous path, which is to dramatically improve energy efficiency so renewables will be sufficient.
Geoff,
examples of what Dave might mean:
- The concept of capital used in many economic models, includes ‘land’. This makes it impossible to see windfall profits and rents of people owning land and non renewable natural resources as all these rents and windfalls are called ‘profit’ (this is, of course, a Georgist critique). The whole idea of a finite earth is, when one thinks of it, alien to neo-classical economics.
- In many DSGE models, involuntary or ‘cyclical’ unemployment is, by definition, excluded. The implication of this is that according to these models unemployment is voluntary/structural (it takes time to sell your house and move to another job). This makes it, according to the model, unnecessary for governments to pursue anti-cyclical policies (not just fiscal policies but also, as might be usefull in the case of Denmark, ending the euro peg of the currency). For the record: unemployment in Spain increased from about 8% in 2008 to about 20% in 2010, i.e. the amount of jobs just diminished and history teaches us that no matter how hard people search the economy has to grow with at least 20% and mybe even 40% to solve such a level of unemployment. Old school economic models do admit this, many DSGE models don’t – the idea is that odern economies do not have to grow to produce jobs (an idea com-ple-te-ly refuted by the economic history of the twentieth century). Who benefits from this idea?
- Another one: many models do not contain money and prices. ‘Prices’ do show up as non-monetary relative values of one product compared with another, but not as the seperate monetary entities we know in real life. In such models, monetary crises and even debts can’t exist. Who benefits of such ideas?
I completely agree with you about how to use models, but many economic models are not just simple to keep things simple, but to keep things out of our thinking – as that threatens vested interests, or long cherised ideas or whatever. For economists, concepts and models are sometimes not just tools, but almost religious statements about the way the world is supposed to be.
You impute too much finesse and detail to my thinking – I’m the guy trained in philosophy who never worked there and reads economics because … I’m a glutton for punishment….
My meaning is/was just about as literal as my question – well, what about catastrophes? My philo of science is still in that Hemple-Popper tradition. So I still think of models as axiomatic inventions, and when their predictions fail in confirmation something must change – the axioms, the data collection – something.
So, where do catastrophes fit in models? I said that dismissing “gravity” as a natural phenomenon not needed in an economic model was “almost beautiful” because it starkly pointed that we do know SOMETHING about this. Shouldn’t models be able to at least suggest a range of bad consequences should something – whatever, a tsunami, a plane crash – disrupt a power grid that sustains auto parts needed for manufacturing half a world away? And if we then had such a range, we could then argue how wise it would be to build economic schemes so interlocked?
I know there is a more literature than I could ever consume on the difference between “uncertainty” and risk. Is that really where the solution’s to be found? Or do we already know from 10,000 years of collective wisdom that even the most robust basket can be tumbled and all the eggs lost ….
Whose interest is that there be just the one basket?
Right Dave! How do we apply that to the possibility/likelihood of catastrophic global warming?
I don’t know. I look for the ways to talk about it. I can only keep hinting a glimpse of my general point by pointing at the contradiction of some particulars, and in this case, nuclear power plants. Ignore the politics that lets Ameren UE bribe the Mo Legislature and the Rural Elective Co-ops for its purposes. Look at Ameren UE’s arguments: So many jobs, so much power available at such and such a cost to consumers, so much reduction in CO2 (cynical assholes) – all the upside, all putatively measurable. But when skeptics and heretics ask for the downside, we get the mumble-tongued double speak “Well the full effect of an accident is immeasurable, too many variables and too few known assignable values.” That then becomes the accusation that the critics have no hard data for “bad” effects to counter the alleged hard data for good effects….
And all I can do in response is point to Japan and say: Have you seen this?
I keep getting, in my own mind, back to the implication I drew from Derek Parfit’s book “Reasons and Persons” – which is not to say it is the implication he wants people to draw (I am much more libertarian about vices, regardless of my future self’s interest): There are different “languages” for talking about reality. They are irreconcilable, in that, there is no one language into which all of them can be reduced by translation. Some languages are economical, some languages are moral. When a language cannot talk about some aspect of reality about which it arrogates confidence – that is, it cannot talk about some part of reality for which it claims (its speakers claim) to have confidence – well, time to switch languages, find one that gives the means to talk about what the other cannot….
What are economical “lacunae” about reality are only filled by moral discourse?
When/how do we know to switch languages, even though, I bet, we all know we do it?
So I fumble along.