Against forecasts

I said yesterday that forecasting isn't part of proper economics at all. I should expand.

There is pretty much no reputable profession which rests its reputation upon forecasts. You don't expect a doctor to tell you what your next ailment will, or a mechanic to tell you the next fault your car will have. Why then, expect economists to have powers of foresight not vouchsafed to other professionals?

The question is especially acute because there are several strands of economic thinking - I'm thinking of complexity theory as much as the efficient market hypothesis - which tell us that both financial markets and economies are unforecastable.

You might object here that if economics doesn't make forecasts, it can't test its hypotheses against the evidence. This is wrong. We must distinguish between forecasting and predicting. A prediction is a conditional statement, a hypothesis, whereas a forecast is an unconditional one. "If you cut prices, then ceteris paribus, demand will rise" is a prediction. "Demand will rise" is a forecast. And predictions can be tested against actually-existing past facts - for example the hypothesis, "value stocks are riskier than others and so should offer higher long-run average returns" can (only) be judged against existing evidence.

From this perspective, the fact that so many tried and failed to predict the crash doesn't discredit economics, but actually vindicates at least three of its predictions:

 - Complex emergent processes are almost impossible to forecast.

 - It's impossible for most people to forecast asset prices correctly, because if they see a crash coming they'll sell up in advance and so prices would never rise so much as to crash later.

 - People respond to incentives. If you give people incentives to do daft things - and economic forecasters are well-paid - then daft things is what they'll do.

It's worth noting here that some of those who are credited with forecasting the crash seem not to have actually done so; Steve Keen's precise forecasts for 2008 are hard to pin down, whilst Nouriel Roubini seems to have got a lot wrong as well as right.

Long before the crash - in 1990 - Deirdre McCloskey wrote: "The best economic scientists, of whatever school, have never believed in profitable casting of the fores." ( If You're So Smart, p109) Her comment stands.

As she points out, the belief that economists should forecast rests upon wishful thinking - of the same sort that's given us raindances, witchdoctors and ju-ju men. There's a hope that some arcane knowledge will bring us a fortune. But it can't. Worse still, there's a political urge here as well. The desire for economic forecasts is also a desire for a form of knowledge that will give us control over human events. As Alasdair MacIntyre noted in After Virtue, the claim that we can foresee and control the future is a claim to power. There's something deeply contradictory about the urge that economists both forecast the future and become more humble. You can't have both. 

And herein lies a paradox. Usually, I call myself a Marxist and am in a minority. In this context, though, I'm in a minority by being anti-Marxist. Marx famously said "the philosophers economists have only interpreted the world, in various ways: the point, however, is to change it." For me, this is wrong; the point is merely to interpret the world - something we can do without forecasting it. In this sense, those who want economic forecasts are being more Marxian than me.

Comments

[Luis Enrique]

I think there's room for some forecasting, with very modest goals. For example, forecasts based on fitting a model, perhaps based on some loose theory like an SVAR, to time series data and then projecting it forward, so long as you recognize that in doing so you are hoping the future will closely resemble the past and will by definition fail to predict events not also present in the historical data you are using (i.e. the ones you'd really like to predict). Some people might consume these forecasts, because they need a best guess, whilst realizing the forecast will inevitably be wrong.

[FromArseToElbow]

You're right that the problem stems in part from wishful thinking and the popular idolisation of metaphysical concepts such as the market, but the economics profession has aided this with its fondness for "iron laws", "mechanisms" and all the other phenomena of a clockwork universe, not to mention the privileging of mathematical models over the methods of social science.

But all of this is just a distraction. The question should not be "why are economists' forecasts sometimes wrong?", but "why are they encouraged to make them?" I think you addressed this pretty well last year:
http://stumblingandmumbling.typepad.com/stumbling_and_mumbling/2012/11/economists-as-dentists.html

[Enzo Rossi]

Any philosophy undergrad knows that scientific predictions take the form of conditionals. But economics’ problem is that it seems to throw too much into the ‘ceteribus paribus’ clause. Physics can’t predict the exact trajectory of a leaf falling from a tree because there are too many variables; but in the end physics gave us lots of useful technological innovations. In other words, physicists can get a grip on the ceteris paribus clause and manipulate the world effectively and reliably. Can economists do that?

[Enzo Rossi]

ceteris paribus, that is.

[Rich C]

I'm not unsympathetic to the point you're making, but I don't agree that a "forecast" is necessarily unconditional, or that a "prediction" is necessarily conditional in any terribly clear or useful sense. If, for instance, an economist heading a central bank were to urge elected officials not to introduce any prudential regulation affecting a rapidly growing market (like, say, for credit derivatives) and in so doing asserts that the spread of these instruments will reduce systemic risk, is this a conditional prediction? Is it a forecast? I have to say, it seems a lot like a forecast to me, and one (I'm referring to Alan Greenspan in the late 1990's and early 2000's) that turned out to be quite wrong. Do we really gain anything by trying to draw a distinction here? Do you think that Greenspan and other economists who encouraged financial deregulation and discouraged financial regulation were actually articulating a clear set of conditions under which their claims would be true, but in the absence of which would be false? I honestly don't think they were, and I think that this is what people think of when they say "economists failed to predict/forecast the crisis." They mean that economists were urging policy changes that had effects almost exactly the opposite of what those economists, and the economic literature they created, had anticipated. Whether you call this a failure of forecasting or prediction seems to me to be neither here nor there, and the issue of "conditions" something of a red herring.

[Lars Syll]

Chris, if you only had written "forecasting SHOULDN'T be (instead of ISN'T) part of proper economics," I -- and i guess most other heterodox economists as well -- would agree. But IT IS very common for orthodox economists to actually argue that the ability to forecast is important virtue of their theory/models, and even more important than explanation in economics (often giving reference to M. Friedman's instrumentalism). They are wrong -- on that we do agree -- and I have a couple of blog posts about that, so instead of recapitulating the arguments here, read e. g. these posts:
http://larspsyll.wordpress.com/2013/04/04/forecasting-errors/
http://larspsyll.wordpress.com/2013/08/18/why-we-are-better-at-forecasting-drizzle-than-financial-crises/

[Deviation From the Mean]

Will capitalism incur periodic overproduction? Absolutely no doubt that it will. Can we pinpoint the time and date that this will happen, not exactly.

We can look at the mechanics of the system and draw certain inferences and make certain statements.

Also, planning has to be based on a certain amount of forecasting, modelling of future expectations etc. I work in budget forecasting and we develop models that for a given input produces a given output. The inputs are based on current conditions but various what if scenario's are included in the process. Most businesses today have a planning regime and forecasting is part of the package.

This is a case where economics says one thing and business does another. Richard Wolff makes this point, at university there are 2 departments, the fantasy masturbatory economics department and the business department that has to deal with the real world.

From a left wing point of view I would critique how planning is carried out under capitalism and how society is organised. And the same constraints that appear in economic models also affect business planning models.

[Deviation From the Mean]

Incidentally, a further use of forecasting is to compare what happened to what was forecasted, and from this learn future lessons.

So in isolation, forecasting pretty useless but combined with other things, quite powerful.

[Dipper]

if not fortune tellers, or even less so magicians, then perhaps economists should recast themselves as a couple of other professions.

i. Systems engineers. This is an old point, but an economy is an interaction between a number of players with input and output signals. Various participants apply forces, and others respond. It's engineering. Economists can point out these links so we can understand what outcomes are possible given the prevailing forces

ii. Priests. Economic policies reward some behaviours and punish others. Hence our society becomes a reflection of cumulative economic polices. Economists can point out the moral consequences of economic policies.

[Martin Connelly]

You say "You don't expect a doctor to tell you what your next ailment will (be)" but you are not right. Every year my doctor tells me, based on a number of measurements and tests what my chances of certain deadly diseases will be. While there is obviously a disease, or two, that are not on this doctor's list he is forecasting the chances of certain common ailments. Indeed a doctor that did not tell a patient that to continue smoking was to increase the risk of lung cancer, would not be doing his or job properly. Now to mechanics and engineers ... they too forecast part failure and replace bits an d pieces before they fail. That is why so few planes fall out of the sky.

[kio]

"There is pretty much no reputable profession which rests its reputation upon forecasts. You don't expect a doctor to tell you what your next ailment will, or a mechanic to tell you the next fault your car will have ..."

How do you fly plane, drive car, take strong medicine, use elevators, mobiles ..? They are all based on robust professional forecast of reliable performance (never 100%). After any major treatment doctor gives a prognisis, which is based on broad statistics and specific terms. All reputable professions do rest only on relibale forecasts. Otherwise reputation cannot be defined. Economics is not a reputable profession because it does not provide reliable forecasts.

[Leslie48]

Articulating your Marxist position is right on. With a sociologist & history background & college teacher I have long wrestled with this conundrum; Am I a Marxist? The 'Milliband dad affair' put it in some sort of perspective too. Analytically, Marxist interpretations of history -that its primarily about struggles between economic groups/class/elites & the role of capitalism, the unremitting search for profits ( e.g. globalisation/immiseration) ,the vital role of bourgeois ideology/false consciousness(e.g. D/Mail) & political power fits the bill in many instances. Althusser & Gramsci provide much insight. But that does not mean we are Marxists in any religious or prophetic sense or that it could delivers us from all evil, injustice & inequality.

[chris]

@kio - I think you're stretching the definition of forecasting a little. When a doctor says "this treatment will make you better" what he means is "it has worked with [a % of?] other patients, and there's no good reason to think you're different." That's a description of the past and present, more than a forecast.
@Martin - Yes, doctors do give probabilities and warn of risks. But so do economists; I do it very often in my day job.

[Floormaster Squeeze]

Trained in economics I was once asked to predict a fairly specific but large and highly variable number. I used a reasonable method that had not been used before (previous methods were also reasonable) and forecasted that number with remarkable accuracy. At that point, I was briefly feted and then asked to forecast all kinds of things and to teach my method to all kinds of people.

Unfortunately, when I explained that while there may have been reason to believe my method had some small impact on its accuracy, the biggest thing going on in the accuracy was luck (or a whole bunch of factors that I could not forecast cancelling each other out). Unfortunately too many people wanted to believe that some magical power resided in me or my model. Despite never claiming to have any special powers, everyone soon became disenchanted with me and my approach and moved on to try yet another magical approach (without learning anything about the limits of forecasting).

[John Williams]

You are wrong that workers in other sciences are not expected to make forecasts. Those of us in environmental sciences are asked regularly to make forecasts, for example with environmental impact statements, or with forecasts of global warming.

And, if you knew much about real estate, the housing bust was not hard to predict; rental housing was selling for far more than the rents would justify, and people were borrowing on the inflated value of their houses and spending the money. This could no go on. However, I agree that predicting the timing of the bust could not be done precisely.

[joe]

Engineers are one profession that do have to make forecasts that will be tested and usually pass those tests. The airport, railway of skyscraper will appear in the future - it has never existed the same way before - it may cost a little more or take a little longer but it will work. It builds on the past but above all uses the whole wisdom of current crowds in not being dependent on one persons' view of the World but thousands.

[Magpie]

@Chris

"Why then, expect economists to have powers of foresight not vouchsafed to other professionals?"

Maybe because economists demand those powers of foresight from those they don't like.

For instance, Noah Smith crucified Steve Keen for failing to fulfill the following criteria (check in Smith's post):

(1) what is “the crisis” Keen predicted (“there are different things that get labeled ‘the crisis’ ”); Smith includes an explanation here, causal relations;
(2) “what it means to ‘predict’ something”: the timing of an event, the size or severity of an event, the duration of an event, particular characteristics of an event;
(3) “what degree of confidence you make a prediction”;
(4) “how far in advance a prediction was made”;
(5) “whether the prediction was made by a model or by a human”.

You will notice that in that blog, Smith never used the word "forecast"; at all times he spoke of "prediction" (as Keen himself, btw). Therefore, it's clear Smith considers prediction a little more than simply "a conditional statement", as you define it here.

----------

Incidentally, I'm not taking sides in the Keen vs Smith feud.

But the claim that "Steve Keen's precise forecasts for 2008 are hard to pin down", which you make above, is, as far as I know, unfair in two ways:

(1) Keen never claimed to have forecasted anything, precisely or not; the word he used was prediction;

(2) and Dirk Bezemer did not have much difficulty to pin Keen's claims down:

'No one saw this coming' – or did they?
Dirk Bezemer, 30 September 2009
http://www.voxeu.org/article/no-one-saw-coming-or-did-they

You'll notice Bezemer did not have much difficulty in finding the causal mechanisms Smith could not find.

Cheers

[Magpie]

“There is pretty much no reputable profession which rests its reputation upon forecasts.”

This is an odd claim. Are you saying astronomy, physics, chemistry… are not reputable?

What use is meteorology if it's not making forecasts, like the ones you see on TV, for instance? And they are getting better and better, btw.

“You don't expect a doctor to tell you what your next ailment will be”

True. But doctors and actuaries can calculate the probabilities of diverse diseases, can they not?

Doctors recommend vaccines and changes in diet and lifestyle on the basis of statistics: "You need to quit smoking, because it doubles your chances of getting a stroke".

How can insurance firms calculate their budgets without those forecasts?

In general, how do firms, households and the government budget without forecasts? "We'll need more investment in age care, because the population is growing old".

[dr oz recommended diet pills]

new: Weight loss is perhaps the most sought after treatments today, with increasing number of people suffering from excess weight due to wrong eating, lack of exercise and hectic lifestyles v3 diet pills augusta ga To lose the fat even faster you need to eat 5-7 small meals every 2-3 hours throughout the day to supply the energy you need to be able to do justice to your exercise program.

[Lee Malatesta]

I'd like some clarification on the distinction between a prediction and a forecast. By a "prediction is a conditional statement, a hypothesis, whereas a forecast is an unconditional one" could mean a few different things and the given examples don't clarify the matters very much.

First, the conditional/unconditional distinction could refer to the material condition operator in logic which is usually translated into English as an if/then statement. If X, then Y.

Second the conditional/unconditional distinction could refer to a non-absolute/absolute distinction where predictions are not absolute (it is likely that x will happen) where forecasts are absolute (x will happen).

Many fields that make forecasts (medicine, epidemology, meteorology, astronomy, finance, economics) do so in probabilistic terms. For example, weather forecasts often call for an "80% chance of rain" and when the federal reserve decides what to set various interest rates at, they do so on the basis of what they think financial conditions are likely to be in the near future. These sorts of forecasts would not be forecasts if the distinction between predictions and forecasts is an non-absolute/absolute one. But they do remain forecasts if the distinction is more akin to the material condition in logic.