Economics doesn’t seem able to predict anything of great import. It certainly didn’t seem able to predict or even, after the fact, explain the recent banking and financial crisis; the repercussions of which will be felt by millions for decades to come. Why was this? What are the limits of economic prediction? Here are just a few thoughts.
Although I’m a simple economist and not a physicist – who do tend to think, pithily though incorrectly, that anything other than physics is stamp collecting – I’d like to venture the opinion that even in physical mechanics there are two primary ways by which we can make predictions. They have to do with what we can call “momentum” and “cyclicality”.
If you’re driving along a straight motorway and someone asks you to predict where you will be in a few seconds time, the most likely outcome, and thus possibly the best prediction, is that you’ll still be on the motorway but just a bit further on. The car has momentum and so, even if the engine suddenly packs up, you’ll probably still be on the road in the near future – momentum will tend to see to that. Of course lots of other things might happen. I could suddenly decide to take the exit road that has appeared, fall asleep at the wheel or hit an on-coming lorry that has crashed through the central barriers into my path. But all these things require human volition – that means me or others making choices and taking decisions – so they are not purely mechanical. Without volition, mechanical momentum, even angular momentum, will generally determine where I’ll end up in the very near future. Over much longer periods of time it’s also possible to predict with great accuracy where a moving asteroid will be quite far out into the future.
In economics it’s the same thing. If the rate inflation in the first quarter of any year is, let’s say, 2.5%, it’s highly unlikely that in the subsequent quarter it will be 20%. It’s possible but very improbable. If the inflation rate were a random walk (which it isn’t) the best forecast of next quarter’s rate would be what it was in the last quarter i.e. 2.5%. However economists can incorporate various leading indicators into their models and might thus predict that the second quarter inflation rate will be 2.7%. Basically unless something very drastic happens -a huge meteor strike or a nuclear war for example – I think it’s clear that the economy, and thus the inflation rate, have both inertia and momentum.
Whether an economist predicts that next quarter’s inflation will be 2.5%, 2.4% or 2.7%, the prediction will be reasonably accurate (depending on our definition of accurate), but only because of the inertia and momentum in the system. If you ask an economist to forecast what the inflation rate will be in 10 years time, he or she will either wisely refuse to make a prediction or if he or she does offer one it will, after the elapsed period of time, tend to prove to have been wildly inaccurate.
All physical scientists and engineers know this.
Meteorology is a science; yet even with all their sophisticated and mathematically elaborate models, meteorologists generally limit their weather forecasts to a few days out. If they get ambitious then they may extend this to a few weeks. They have no idea what the weather will really be like in London next June. Perhaps because it’s summer they could predict it’s not likely to be freezing, a point to which I will return in an oblique way later on. But even in the short term they can get it spectacularly wrong. Those who live in Britain might remember, as I do, the BBC’s weatherman, Michael Fish, confidently telling us in 1987 that: “Earlier on today, apparently, a woman rang the BBC and said she heard there was a hurricane on the way… well, if you’re watching, don’t worry, there isn’t!”. That evening, the worst storm since 1703 hit South East England. It caused record damage and killed 18 people. It also ripped off my roof! So even meteorology can only hope to make reasonably accurate weather predictions for very short periods – and then not always with complete success. This is because weather is a very complex dynamic system, and so its predictions tend to rely primarily on inertia and momentum.
The second, and related, way in which we can make predictions is by taking account of cyclicality – events or patterns that repeat or reoccur. Our mechanical models can forecast the next appearance of Haley’s comet with great accuracy, as well as the times of the daily ebbs and flows of the tides at any location on the planet. Although these predictions are not strictly 100% accurate, they are very nearly so, and thus they are extremely useful. Newtonian classical mechanics can and does predict a hell of a lot, and with enough certainty to have made it the bedrock of many of the most useful technological advances over the last couple of hundred years or so.
Unlike economics, the physical and other natural sciences have moved on from classical mechanics. This doesn’t mean they’ve abandoned it but just that mechanics and a “reductionist” method can’t explain everything. Einstein’s theories of relativity were a revelation. They told us truths about the nature of Space-Time that we could have hardly conceived before. Without knowing that time goes slower the faster we travel the world’s GPS system would become fairly inaccurate within a day or so. The two theories of relativity have been experimentally confirmed on numerous occasions since Einstein proposed them and they have enabled scientists to push forward the limits of physical prediction.
Similarly, and please excuse my economist’s ignorance of the subject, while the world awaits a grand unifying theory, it does seem to be the case that the sub-atomic world operates very differently to the macro world. You can’t simultaneously predict the position and momentum of particles; all you can do is work out some probabilities. Not only that but it seems that particles can disappear from one place and then reappear in another, seemingly without having been anywhere else in the interim! But theoretical quantum mechanics, despite its probabilistic nature, can also make predictions which are very often confirmed by experimental physicists. Even predictions of the likely existence of things – such as Higgs’ Boson.
Before turning back to economics, perhaps we can mention the weather again and even, though with some trepidation, the climate. Regarding the weather, I guess we’ve all heard how the flapping of a butterfly’s wings might cause a hurricane in India and about chaos theory in general. The problem for meteorologists when trying to predict the weather, over any time period other than the very short term, is that even if we can fully understand all mechanical and physical processes in their entirety, and thus could theoretically build comprehensive mathematical models that are fully deterministic – by this I mean for example that the weather system could be treated as a closed system uninfluenced by such outside “stochastic” shocks as periodic solar flares and the like – unless you can specify with absolute precision the exact starting positions, mass and energy, the “initial conditions”, of every atom that is involved in the dynamic development of weather, you’ll never be able to predict the weather somewhere in the world more than a few days out. Meteorologists know this and thus tend not to make such predictions.
As I mentioned, I hesitate to touch on climatology and climate change, about which I know even less than I do about meteorology. Is there such a thing as global warming? Is it man-made? Such questions also touch on the nature of scientific prediction and on issues of cause and effect. From a personal point of view all I can say is that it does seem from the overwhelming bulk of the evidence provided by science that the earth is warming and that this doesn’t seem to be just another of the normal and natural cycles of changes in the climate witnessed on the planet. Rather it does appear to be caused by mankind. Even with my limited knowledge of physics and chemistry, I think it stands to reason that if you pump lots of gases with known properties into the atmosphere, it would be strange if they didn’t produce the normal physical and chemical effects – i.e. warming.
The point I would like to make here as it relates to the nature and limits of prediction is this: Assuming that the conventional scientific wisdom on climate change is correct, it’s still quite unlikely that the trend in rising temperature will be a neat linear one – that the cumulative build-up and concentration in the atmosphere of gases such as CO2 will lead to a parallel incremental rise in the world’s average temperature. The study of the deep history of the earth’s climate plus an understanding of biological evolution and eco-systems, has quite clearly demonstrated that physical and biological systems often jump almost instantaneously from one condition to another (from one equilibrium to another if you prefer). These jumps are often referred to as “tipping points”. We see similar “phase transitions” in physics.
Potential tipping points are understandably one of the fears of climate change scientists – not to speak of the rest of us. In terms of prediction, although we know that such tipping points have happened and do happen, because the climate is undisputedly a complex system, and I would argue a complex adaptive system at that, we might probabilistically suggest that such jumps, or tipping points, are quite possible in the future due to global warming. But I don’t think we are in a position to predict with any certainty or accuracy when such a jump will happen nor its extent and nor what the precise trigger for it might be.
I want to use these examples from the physical sciences as analogies for economics. Not, I hasten to add, as strict equivalences.
In the example I gave of driving along a straight motorway, the reason we might be able to predict where we will be in a few seconds is momentum. And so, as I suggested, short term predictions of key economic variables such as inflation, employment, growth etc also depend on momentum, coupled with the data provided by some key available leading indicators.
But think about it. Driving a car is not an exclusively physical act; it is also a social act. There is an actor – the driver. This actor, or “agent” as we might call him or her, has a whole raft of characteristics. Just two of these are that he has volition to choose or to decide at any moment what to do, and he can also adapt to the circumstances he finds himself in. If a motorway exit suddenly appears the driver can decide either to turn off the road or stay on it. If he turns off he will make our simple mechanistic prediction based on momentum wrong; even though he won’t break any physical laws.
The economy and anything remotely economic is like the situation of the driver writ large. There are millions of individual, collective and institutional agents continually making choices and decisions – i.e. exercising their volition. They are also adapting their actions according to the actions of others and the collective result of all these actions. That’s life. The economy is very decidedly a complex adaptive system. From the point of view of economic prediction it doesn’t really matter if the “rules” used by the actors to help them make their decisions are completely “deterministic” or if there are also a stochastic “shocks” as well. Even with completely deterministic rules, the longer term dynamics of any complex adaptive system are, in the real world as well as in modelling, pretty much non-predictable. Why so? Surely if, like with the weather, we knew all the rules and all the initial conditions of every actor/agent, we could predict the outcome? The answer to this question is that “Yes” in theory we could. Unfortunately for the purposes of prediction the physical, social and economic worlds don’t work like that. Not only do we not know all the rules (I’ll leave that to one side) but if you change the initial conditions of any single actor/agent even in the tiniest degree then the collective result, after a bit of interaction, will not just be slightly different, it will be completely different – and this you can’t predict ex ante, before the event.
Quite categorically this means that economic predictions, even if they could be based on fully deterministic mechanistic rules and modelling (or maybe any type of modelling), will never be able to forecast the timing and magnitude of economic events with any accuracy over more than the short term. They decidedly can’t predict the future positioning and situation of individual agents. This has actually been the experience of economics. It has rarely been able to predict anything of any great macro importance by using conventional neoclassical theories and associated modelling techniques. Occasionally a prediction might get lucky and will no doubt be much trumpeted as a great success for the predictor and for his/her economic model. Yet I do tend to think that these rare “successes” might better be explained in terms of “Black Swans” and the like, so persuasively discussed by Nicholas Taleb Nassim.
Where does that leave us? Strange though it might sound, I do actually believe that economics can make some very insightful and useful predictions or forecasts; but they are not of the type I’ve discussed thus far. Rather they have to do with cyclicality or, more generally, with regularly repeating patterns.
Take the latest banking and financial crisis; a crisis that will inevitably lead to the gradual impoverishment of countless ordinary people throughout the world over the next decades, despite the fact that it wasn’t in any way their fault.
We know that most conventional economics (and unfortunately that means the neoclassical economics I studied for so many years) failed to predict this crisis. Economists and central bankers (not to mention ideologically driven politicians and greedy investment bankers) continually asserted that all was well in the world. When the crisis hit it was rather amusing, though a little sad as well, to witness the spectacle of Queen Elizabeth, when she visited my alma mater, the London School of Economics, asking why economists hadn’t seen the crisis coming. There was lots of embarrassed huffing and puffing and subsequently quite a bit of blather about how economics should take more account of “systemic risk” and such like. But there was no real answer. Actually most economists and even policy-makers have by now admitted that “there was something wrong with our models”. This crisis simply shouldn’t have been able to happen. Even central bankers such as Alan Greenspan and Mervyn King admitted this; not to speak of some of the more honest archpriests of the intellectually and empirically vacuous bunkum that is called “Modern Finance Theory”. In economic circles what then happened was that economists dashed around like chickens without heads (as a very senior central banker once told me), trying both to stick on a few patches and retrofit their theories to the experience of events. One commentator on something I recently wrote to this effect even said that it would be better to say that they “retrofitted the facts or denied them completely”. That’s probably fair though a little hard.
In true sciences if a theory is shown by experiment to be unable to make accurate predictions, or even not to be able to explain what has happened after the event then, despite many entrenched reputations, it is eventually abandoned. This doesn’t seem to be the case in economics. I think the reasons have to do with power, but that’s for another time.
Yet certain economists did actually predict that there would be a banking crisis and a subsequent economic meltdown. Perhaps not that many; but there were quite a few. Let me just mention the Australian economist Steve Keen in passing – I’ll quote more such if need be. Such economists knew, either explicitly or implicitly, that the economy and the banking system were both complex adaptive systems. Thus they never dreamed of making point estimates as to the timing of the onset of the financial crisis nor its magnitude nor, indeed, what might be the triggering event. No, what they did, with an understanding of some very simple economic principals, was to see that the developing economic dynamics (for example of the build-up of the levels of absolute debt) were displaying a known historical pattern. We had seen this before and we had seen the results. How could it not happen again? Something was bound to crack. We can term such events Minsky moments or something else but the fact is that their predictions were predicated on an understanding of “cyclicality” in the economy – an understanding of historical dynamics, repeating economic patterns and how and why these patterns arise.
On the other hand, neoclassical economists do tend to have a sort of collective amnesia when it comes to the past. They don’t seem to have learned anything from what’s gone before. To be sure these cyclicalities and patterns of economic dynamics aren’t in any way like the regularity of the amplitude and frequency of physical waves or the regular ebbs and flows of the tides. They’re not even like the cycles of various economic “trade cycle” theories once so beloved of economists. How could they be! Anyone who has ever looked at the dynamic properties of complex adaptive systems will know that sometimes a system will start to oscillate. These oscillations might be explosive or they might die down, only to appear once again, possibly after a long period of relative stability, but in any case certainly totally unpredictably. But what economics can try to do, and is increasingly doing, is not to predict precisely when a particular event or “crisis” will occur, or even to suggest what might trigger it, which is unknowable. Why not Northern Rock or Bear Stearns? Why Lehman Brothers? Rather what they try to do is look at the prevailing economic conditions and how these are developing through time (and space) and predict, with a decent knowledge of real economics and economic history, what seems likely will happen. It’s not much, but that’s what the economics profession can do in terms of prediction if it puts its mind to it.
Economic prediction will never be a precise science, if it’s a science at all. But, I would like to suggest, useful predictions can be made if we use what we know about how economies really work and tend to evolve (and we know quite a bit) coupled with a greater historical awareness.
In terms of economic modelling and the use of mathematics in economics, both of which I support, the main questions to ask of models are: What are we using this model for? And what are we trying to predict with it? A topic for another day.
Ultimately I think it’s the explanatory power of economics more than its limited ability to predict that is of crucial importance. If we understand how and why something has happened it will go a long way towards enabling us to create a system that might limit the deleterious things we wish to avoid.