« Repressive diversity | Main | Wages & profits »

July 27, 2014


Nick Rowe

Chris: OK. Fair points. But I was pretending to be an orthodox New Keynesian when I wrote those things!



Simon Wren Lewis has just a done piece that explains the New Keynsian consumption function, which is that consumption is a function of discounted future labour income.

Completely unobservable, not capable of refutation, and probably complete bollocks too.

Thornton Hall


Your example of the Higgs boson shows you don't know the difference between unobserved and unobservable. Mistakes like that mean you can't claim to practice the "philosophy" of anything.


@ Thornton - yes, there is a distinction between "unobservable" and "unobservable given current technology", but for my purposes this didn't matter.

Luis Enrique


would you claim that your current consumption would be completely unaffected if you expected your future income to halve (or double) from its current level?

also, there are ways of testing whether current consumption responds to expected future income (send me a post-dated cheque for £1000 for more details).

Luis Enrique

incidentally, the Euler equation fails to fit data, specifically relationship between c and interest rates, imho, because most of us do not using saving/borrowing to continually reallocate consumption over time - the sums involved and the resulting welfare gained are too small to come close to the threshold needed for our attention when we have limited mental space and lots of other things to think about (see here: http://sims.princeton.edu/yftp/Gerzensee/info.pdf) and the same goes for future inflation expectations etc. BUT this does not tell us 'expectations do not affect consumer spending much'. For example changing expectations such as 'I now believe there is a much higher chance of me losing my job' may have a large effect - not all expectations are equal.

Ralph Musgrave

“Love”, “hate”, “political left”, “joke” and numerous other commonly used words refer to things that are not “observable” or at least not measurable. Obviously we need to abandon half the English language.

Thornton Hall

No, it does matter. A lot.

Left Coast Bernard

A better example of an unobservable quantity in physics is the wave function. This function is the solution to the Schrodinger equation, which describes the non-relativistic evolution of a quantum system. The wave function contains everything that there is to know about the system. It is, however, a complex function of space and time. Therefore, it is unobservable in principle. Indeed, it can never be measured. All physical measurements result in rational numbers, and never a complex number. Researchers must "square" the wave function to produce a measurable quantity. Even that square, which is a real number cannot be measured. Rational numbers, which are results of any measurement are a tiny sub-set of real numbers.

Commenter Hall is correct that the Higgs boson was not an example of an object that is in principle unobservable. It was unobserved, and now it has been observed.


Interesting post, but I think you’d do well to distinguish this social-science/ policy-making issue from issues of science (and the philosophy of science). The commenter Thornton Hall above is quite right.

[Idealising somewhat] in hard science, all theory is expressed in terms of unobservables, in the sense that it seeks to explain the phenomena. Theory has to operate at a level of abstraction from the phenomena, otherwise it would merely describe (not explain) the phenomena.

Of course, this doesn’t apply to economics, because it isn’t a science. (I don’t mean this as an insult; merely a description. Economics has none of the deductive depth I’ve described above simply because there is no science of human behaviour: we have no idea why we act as we do: as far as we can see we simply have free will.) ‘Theory’ means something very different in science than in social science, whose goal is only to describe the data in a way that is useful for making policy.

Your point here simply seems to be that we have greater confidence in some metrics than others. This is an important point for policy making, but it has nothing to do with issues of scientific theory.


Earth to Steve,

"there is no science of human behaviour"


The basis of science (both historically and actually) is solving practical human problems, science is method not a ring fenced area, where everything outside is immune from science.

Jim Johnson

Read Dan Hausman's newish book Chocie, Value, Preference and Welfare (or some such concantenation) published by Cambridge. Doing without unobservables in economics is impossible. When talking about Philosophy of Economics it helps to know what actual philosophers of economics are saying these days.


"Unobservables" seems the wrong name for these two variables: "the natural rate of interest and expectations of government spending." Both of these could be at least loosely observed by tracking human behaviour and the limits within which people are operating. Most of the people I know, whether they say so or not, act in accordance with an expectation of low and falling government spending, especially if they have government jobs. The natural rate of interest for entry-level savers like me is slightly negative -- we do not put money in savings accounts, (I haven't had a traditional savings account in at least 15 years,) and are skeptically cautious in making any sort of investment. And that's not even including those for whom the point is moot -- people whose spare change disappeared in 2007-8 and who couldn't invest if they wanted to.



I have always felt that rational expectations were the hidden local variables of economics in that not only can't they be measured directly, they can be demonstrated not to exist.

There are things we can observe and measure indirectly. A thermometer doesn't measure temperature, but rather than effect of temperature on the size of some material or the conduction of some junction of two materials. In this sense, we can measure love or hate or ambition or inertia. In these latter cases, we do our measurements by watching past behavior.

The problem is that economics tends to have its forces only measurable in retrospect. If someone buys a second pizza oven, economists can then argue that they had expected a larger demand for pizza. This is different from the more obvious approach of measuring the rising sales of pizza and comparing them to the peak possible output of the existing oven and then making a forward argument that pizzerias would soon be fixing to buy additional ovens.


theOnlySanePersonOnPlanetEarth, thanks for the link. I was in fact aware that Psychology existed. We seem to agree that the social and behavioural ‘sciences’ are about solving practical problems (policy), and that this is valuable. They do this by *describing* phenomena in ways that equip us act upon them.

It’s just that this isn’t the same thing as hard science. The hard sciences aren’t about equipping us to solve practical problems by describing phenomena; they’re about *explaining* phenomena. They do this by positing abstract ‘unobservables’ that make deductive predictions of the phenomena, which can be tested with data. Surely you can see that this is nothing like the social and behavioural ‘sciences’? I agree that in principle there could be a science of human behaviour, but at present we have no idea what such a thing would look like. What could deductively predict what I’m going to do in the next five minutes? We have no idea: we do appear simply to have free will. Thus, at present, there is no science of human behaviour.


Even within the 'hard' sciences there is a 'pure' and 'applied' side. So if we were to talk of a difference between 'hard' and 'soft'? sciences we would also have to talk of differences within 'hard' sciences. And then we may have applied 'hard' sceince having more in common with applied 'soft' science etc etc etc etc.

"The hard sciences aren’t about equipping us to solve practical problems"

That is exactly what they are about.

"The hard sciences aren’t about equipping us to solve practical problems by describing phenomena; they’re about *explaining* phenomena."

Don't 'soft' sceinces also try to explain phenomena?

Having said all of that, I recognise that there is a difference between, say Physics and Economics, and economic theory must bear in mind the laws of physics!


> Don't 'soft' sciences also try to explain phenomena?

No, that’s precisely my point. The ‘soft sciences’ merely describe data. The theories of the ‘hard sciences’ seek to explain the phenomena; that’s what marks them out as different. Note that the ‘explanation’ we are talking about is a deep one. These theories are not in the terms of the phenomena, otherwise they would still be mere descriptions. That means a theory operates at a level of abstraction from the phenomena (‘unobservables’). In other words, a theory demonstrates the underlying unity behind the various phenomena being explained by coming up with exotic new categories. The theories deductively imply predictions for the phenomena, which means we can test the theories empirically by testing their implied hypotheses against data.

This is the difference you recognise between physics and economics, and it's a profound one, which goes to the heart of what the hard sciences are. In the sense we've been discussing, ‘theory’ simply does not exist in the soft/behavioural/social ‘sciences’: there is no explanation, only generalised description. And that has its useful role in enabling policy making. One day someone clever might come up with a theory about human behaviour, creating a new science. But until then, we simply appear to have free will.


Sorry Steve I beg to differ. I don't think this is the difference between physics and economics. 'Soft' sceinces also attempt to describe phenomena.

Thornton Hall

The only sane person on earth may be sane, but he's wrong about the goal of science. Solving practical problems? Seriously? That's your claim?

But his larger mistake, seeing no fundamental distinction between economics and physics, apart from subject matter, is a colorable claim, if only because it is so widely held.

It is, however, quite wrong as well. The reason it is so widely held is that the implications of the truth are radically destabilizing to the field of academic economists. But those are the people having the discussion. What's the line about persuading someone of something when their livelihood depends on believing the opposite?

I recently saw an economist, not DeLong, but on that blog, use the word "indeterminate" to describe a parameter of a model that could be changed by the modeler (Nash equilibrium need not be assumed, but some other assumption instead) It is a revealing mistake. Will command and control carbon regs slow GDP growth? Economists "know" the answer, which is insane! The string of causes and effects set off by a change in regulations is indeterminate. That is, some of the circumstances that form the deductive chain don't actually exist yet. It's not that we don't know the mechanics. It's that the mechanics aren't a thing. It's a non-thing.

When economics discovers that it's filled with claims to know about non-things, things are gonna get ugly.

The comments to this entry are closed.

blogs I like

Blog powered by Typepad