Milton Friedman's Thermostat

http://worthwhile.typepad.com/worthwhile_canadian_initi/2012/07/why-are-almost-all-economists-unaware-of-milton-friedmans-thermostat.html

by cousin_it • 14 years ago

123 118 14 years ago

worthwhile.typepad.com

kevin_morrill• 14 years ago

Really interesting. Biggest complaint is the author takes a long time to explain the thermostat model. Here is the definition:

Everybody knows that if you press down on the gas pedal the car goes faster, other things equal, right? And everybody knows that if a car is going uphill the car goes slower, other things equal, right?

But suppose you were someone who didn't know those two things. And you were a passenger in a car watching the driver trying to keep a constant speed on a hilly road. You would see the gas pedal going up and down. You would see the car going downhill and uphill. But if the driver were skilled, and the car powerful enough, you would see the speed stay constant.

So, if you were simply looking at this particular "data generating process", you could easily conclude: "Look! The position of the gas pedal has no effect on the speed!"; and "Look! Whether the car is going uphill or downhill has no effect on the speed!"; and "All you guys who think that gas pedals and hills affect speed are wrong!"

1337biz2• 14 years ago

I'm really looking forward to the comments on this one. Overall it is a variation on the "correlation is not causation" theme and that defining variables in a social setting is super tricky, if not to say impossible.

cousin_it4• 14 years ago

It seems to me that when people say "correlation is not causation", they typically mean you can't learn about a causal relationship by just looking at the correlation. The article makes the converse statement: you can't infer the absence of a causal relationship from the absence of correlation, or equivalently (contrapositive), even when you know there is a causal relationship, it won't necessarily lead to a visible correlation in the data.

Symmetry4• 14 years ago

It's actually much more subtle than "correlation is not causation". More like "goal seeking behavior can remove all correlation from a causal relationship," which is very different since causation normally causes correlation, and normally is you don't see correlation you can be sure that there isn't any causation.

SeanLuke• 14 years ago

> And no, you can not get around this problem by doing a multivariate regression of speed on gas pedal and hill. That's because gas pedal and hill will be perfectly colinear.

Um, what? For a constant speed, which is this person's example, this claim is completely false.

drunkpotato• 14 years ago

When the incline increases, the driver pushes the gas pedal harder. Inline up correlates with gas pedal up. It's a statement of relation between gas pedal and hill incline. The claim is correct.

pdeuchler3• 14 years ago

I think the better question is why are (almost all) economists unaware of basic economic theory?

Not to go off topic, but it's hard to take any "economist" talking head seriously these days. I'm currently re-reading The Road to Serfdom by F.A. Hayek (highly recommend) and it's scary to see that even 60 years ago we understood cycles and events that are still being ignored today.

The sad fact remains that real economic policy is no longer in vogue in Washington or the L.A. TV sets. What is popular these days is hand wavy gestures that seem to please the most amount of people... the science be damned.

Rather scary stuff.

rvkennedy• 14 years ago

I think you underestimate the influence of Hayek, Friedman et al, whose moralistic market fetish has become the dominant ideology of Washington, London et al.

For a more evidence-based approach, I'll go with Keynes' General Theory, which works, but has few adherents among the major parties in the West.

loganfrederick• 14 years ago

Provide evidence that Hayek and Friedman are the dominant ideology in Washington.

If we wanted to try and accurately describe the economic policy by the government in Washington, you'd have to describe the monetary and fiscal policies. Both of these since 2007 have been Keynesian far more than the other two schools (Monetarist and Austrian).

This hasn't always been the case (Friedman was much more in vogue from the late 70s to approximately late 90s as measured by policies put in place).

Keynes General Theory could theoretically work as he describes in his book, but is unrealistic in practice because of the political dynamics in government.

rayiner8• 14 years ago

Empiricism and Kahneman et al. are going to tear down the house that Friedman built.

Symmetry4• 14 years ago

I'm not sure why you think that the lessons of The Road to Serfdom are being ignored. Does anybody in the halls of power seriously think that Soviet style central planning is a good idea anymore?

But the real reason that nobody pays too much attention to Hayek is that the way countries recovered from the Great Depression worked entirely counter to Hayek's predictions. Going off the gold standard worked. It worked really well for every country that tried it! The way that the events of the 1970s conclusively disproved paleo-Keynesianism.

So now we have neo-Keyesianism as endorsed by people like Krugman and Monetarism from people like Friedman and they agree about a lot more stuff than Hayek and Keynes did. Now, they use different terms and have different ideas about the transmission mechanism by which monetary and fiscal policy influences aggregate demand, but everybody these days thinks of aggregate demand as the thing that drives fluctuations in the business cycle. That's progress, maybe in another 100 years economics will be a real science.

And please don't take the talking heads that show up on TV seriously. Mainstream reporting of economics is just as bad as it is of every other technical field.

kokey2• 14 years ago

I'm keen on what I loosely refer to John Kay's theory of rent seeking. I like it because it puts the issues with market and state influence into perspective, especially the issues people tend to blame the economic theory they don't favour for. I think a lot of the benefit from coming off the gold standard is that it disrupted an established rent seeking power structure around gold.

pdeuchler3• 14 years ago

Hayek addresses redistribution of wealth as well as central style planning. He even specifically notes that while they may differ at (sometimes important) points, the general cycles have very similar results. So while not many people still believe in central planning, they seem to have grasped the new theory du jour of "spread it around", and in the end that is just as dangerous. Even so, to ignore the lessons in the book simply because of a superficial difference in ideologies probably isn't wise.

I think you're rather over simplifying things. I'm no economist, but I make it my business to be at the least knowledgable about such things, and I feel recovery from the Great Depression was more of a combination of New Deal infrastructure, WWII profiteering and the resultant increase of wealth in US that was then transferred to Europe to aid in the rebuilding. But then again, that's probably one of the most complex economic time periods in modern history so I don't pretend to know everything.

To address the gold standard, there are indicators that removing the gold standard may have benefitted in the short term, but will hurt us in the long term.

nickik14• 14 years ago

> But the real reason that nobody pays too much attention to Hayek is that the way countries recovered from the Great Depression worked entirely counter to Hayek's predictions. Going off the gold standard worked. It worked really well for every country that tried it!

I would reconsider your argument here. (1) Hayek was not a huge gold fan as it is, he changed his views on money changed over the years

(2) The gold standard that exist at the time is something very diffrent then that gold standard, gold advocates actually wanted

(3) The third and I think most importend one is that if you inflate on a gold standard, deflating back to that standard is like driving over somebody and then backing up over that person. The right solution would have been to stop prudction of money and see the new price of gold and then fix the doller to that standard again.

(4) Hayek is actually still very relevant and even mainstream. What we have to understand here it is perfectly possible to reject hayeks macro but not his micro, or the other way around. Hayek is infact very respected for his devlopment of information economics in particular.

> So now we have neo-Keyesianism as endorsed by people like Krugman and Monetarism from people like Friedman and they agree about a lot more stuff than Hayek and Keynes did.

Here you are assuming that austrians did not change since the the 50s.

Modern austrians macro in fact are really close to modern monetarist. There is infact wide agreement between Market Monetarist and Austrian Free Bankers. I would even clame that the are much closer together then neo-keynsians are.

I would also argue that mainstream economics now is much closer to austrian economics then it was 50 years ago. Things like Public Choice Theory, Law&Economics, New Institutional Econoimcs are all much closer to austrian economics then the are to standard neoclassical economimcs.

Austrians have been talking about the role of the entrepreneur for a long time, modern development economics only in the last decade started agree that it is importend, movment like micro finance are all things that spring from this.

dbecker4• 14 years ago

I think the better question is why are (almost all) economists unaware of basic economic theory?

Economists are largely unaware of what you describe as "basic economic theory" because you've defined "basic economic theory" in an unconventional way that's incompatible with how most economists describe economic theory.

There are a couple of econ departments that subscribe to "Austrian economics," and they view Hayek as basic economic theory. But they are viewed ideological hacks by the larger community.

pdeuchler3• 14 years ago

You're right, but I was more using Hayek as an example.

A better example would be the proposed increase on capital gains/investment tax. Almost all economists, regardless of ideology, would agree that this will certainly be detrimental. And yet we continue to see "Professors" and "Analysts" explain how sticking it to the rich will solve everyones problems... through science!

josephlord2• 14 years ago

Try Steve Keen for an economist who predicted the credit crunch. http://www.debtdeflation.com/blogs/

His book basically declares war on most other economists too: http://www.amazon.co.uk/Debunking-Economics-Revised-Expanded...

I'm very impressed but I'm always looking for counter arguments to my current understanding so if you know of any good critiques of his stuff I would be very interested.

bubbleRefuge2• 14 years ago

Steve Keen is a heterodox economist(Circuit-ist I think) with close ties to MMT Modern Monetary Theory (born and raised on blogs). MMT intellectuals predicted the great recession, the European Crisis, and the Recession of 2001. I've become a believer. They are basically hyper-keynesians. The basic Tennats are 1) As a point of logic Government must spend dollars into existence(or loan out) before they can tax them back. 2) Therefore Fed Government funding is independent of taxation. 2) Taxation creates demand for currency and unemployment in the currency of record by definition. 3) Government is not funded by borrowing either. 4) Because of 1) 2) 3) Policy should be to proactively spend counter-cyclicly until economy reaches full capacity. 5) Banks create money ex-nihilio 6) Banks are effectively agents of government because of deposit insurance and membership in the Federal Reserve System.

Its pretty fascination stuff b/c MMT says that the mainstream has everything backwards.

http://moslereconomics.com http://neweconomicperspectives.org http://www.amazon.com/Currency-Economics-Modern-Monetary-ebo...

nhebb2• 14 years ago

It sounds like the problem boils down to the difference between understanding a scalar (speed) and a vector (velocity). So economics would need a way to represent economic forces similar to the way a free body diagram is used in engineering and physics. And in order to do that, you would need an economics coordinate system. And therein lies the problem of economists building models without known or absolute reference points. Economics != Physics, QED.

cousin_it4• 14 years ago

I think I can make up a variant of the problem that doesn't boil down to scalar vs vector. Imagine the driver is still trying to keep a constant speed, but the road is now horizontal, and the car gets slowed down by mud instead of hills.

Evbn2• 14 years ago

That's exactly the same. Vector refers to directionality, not number of dimensions.

ajuc2• 14 years ago

Why? The analogy works just as well for temperature of water in a boiler, and the thermostat that modifies heating power inversely to the current temperature of the water.

Both are scalars in that example, and the basic problem remains.

nhebb2• 14 years ago

You have a point there. But going with your example, the ideal gas law (PV = nRT) still models a system where all inputs are known. While not a perfect model, it's much better for predicting the performance of a thermodynamic system than most economic models are at predicting economic systems.

alanctgardner22• 14 years ago

Nope, absolutely not. The problem is that they have a feedback system which is attempting to maintain a constant value. This could be a driver pressing the gas, a thermostat (as listed below), or the Fed trying to modify interest rates to achieve a result. The problem is, if you do everything right, and there's no economic collapse, people wonder why you need to exist. If there is, you're clearly useless.

I fall into this a lot, but I really hate the attitude of engineers and scientists believing that everything can be solved by approaching a problem a certain way. We have this idealized, fetishized approach that if you define all your axioms and have a rigid framework in place, you can solve anything. Of course, as soon as you try it in real life, especially in a field like sociology or economics (which, of course, we deride as not being scientific enough), it all goes to hell.

I would recommend that anyone who is of this opinion - that economics, psychology, linguistics, et al are not scientific enough - actually take a course in that area. What you find are: people who are experts in a field are really smart, yes, they have thought of that already, no, it didn't work. It's a very humbling experience.

cousin_it4• 14 years ago

> The problem is, if you do everything right, and there's no economic collapse, people wonder why you need to exist. If there is, you're clearly useless.

Interesting. Can we come up with some general method of answering questions like "does institution X really play a part in keeping parameter Y stable?"

rayiner8• 14 years ago

The level of discourse in policy-oriented economics papers is shockingly low. My pet policy issue is spectrum management (allocation of radio frequency ranges). Its a subject that lends itself to "neat" economic arguments. Ronald Coase wrote an instrumental paper in 1959 suggesting we should propertize spectrum and allow trading, which would lead to efficient allocations. It saddens me how many policy papers I see from "think tanks" that completely ignore the second part of that paper, as well as Coase's earlier work: the efficient allocation assumes that trading spectrum has negligible transaction costs. This works for say big TV stations in a city, not so much for say allocating spectrum for personal wireless devices. In my research for a seminar paper, I came across people who really should know better who ignored crucial features of the problem domain, ignored transaction costs, ignored considerations of lock-in, etc.

I'm tremendously skeptical of economics-based policy papers. A biologist works on a paper for years, and concludes that this kind of gene expression might be linked to this kind of nutrient deficiency. An economist at a "think tank" works on a paper for a fraction of that time and gives advice on how to structure a tens of billions of $ industry...

1337biz2• 14 years ago

The basic problem is here in my perspective, that Think Tanks almost always subscribe to a specific line of thought and exist to promote this cause. While on the other hand less political motivated researchers are too much in love with constructing models, that nobody outside of their peer group understands and often come with questionable value for real life implications.

rayiner8• 14 years ago

Possibly. I read some absolute drek as well in a U of Chicago symposium on the subject...

Re: think tanks, isn't it something that even in supposedly less rigorous fields like, oh English literature, such upfront results-orientation would be considered the badge of unserious scholarship?

nickik14• 14 years ago

I like the saying, the economy does not need economist, just like evolution does not need biologists.

The things we talk about today are not really that diffrent from what Adam Smith argued for/against a couple hundert years ago. Today you have people like Stigliz arguing against markets with much more sophisticated models then the people that argued against Adam Smith. The people that argue against Stigliz use much better models too. So where does that leave us?

I conclude that we should let the market be the market and not try to influence it, its a far more complex thing then economist understand. Trying to influence in to one direction will probebly just cause ten more things that are unexpected. The market is not perfect, not even close, read Stigliz as a good example of this but I highly doute that the goverment can improve the end result in most cases. If you look at the pridictions of TOP economist at every time X, the are almost always wrong and you get some other people that get it right, these people are then wrong the next time.

I barly trust the goverment to do the easy things write, tweek a huge economy is far from easy.

rayiner8• 14 years ago

I agree we shouldn't try to tweek the economy with government, but the corrolary to this is that we shouldn't then raise up economic arguments to keep the government from solving real problems.

Also, the knee-jerk cynicism about the effectiveness of government is misplaced. E.g. government funds about 1/3 of all R&D in this country (and it was as high as 2/3 in 1964). It is definitely not the least productive third of the R&D expenditures. We crow about all these advances in medical technology, etc, but many of those advances are the direct result of enormous NIH funding. Our world-leading universities get well over half their R&D funding from the government.

dllthomas3• 14 years ago

This argument applies equally well to most things.

Defending a country is a hugely complicated mess, and moving battleships over here is going to have ramifications over there. Policing society is difficult and complicated - we try to cut down on crime by targeting something over here, and it causes problems over there.

Clearly, we should call the whole thing off, and allow international affairs and society take care of themselves - all the way through invasions and murder?

Evolution doesn't need biologists, but people who need to make decisions dealing with the consequences of evolution on epidemiology sure do!

kokey2• 14 years ago

I suspect, just like spectrum allocation, the issues regarding peering/routing agreements between telcos/ISPs are also not new and it's probably been proven from a number of angles that everyone is better off if they find a way to cooperate.

aidenn0• 14 years ago

Nah, Comcast is better off not cooperating. The cost for a consumer oriented site of being cut off from 19M customers in the US is huge, so they have a very strong bargaining position.

eavc• 14 years ago

>And no, you can not get around this problem by doing a multivariate regression of speed on gas pedal and hill.

I'm no expert on statistics, but this seems like a pretty basic scenario for a statistical analysis to provide insight into. These two phenomena would be perfectly correlated in both the time of appearance and the degree.

You wouldn't be able to derive anything about the effect of hills or gas pedals on speed, but you'd be given a powerful clue as to what's going on.

The second there's even a tiny imbalance between the two, you're given the relationship to speed.

simonster• 14 years ago

Well, in this contrived scenario, you can't tell what would happen if the driver _didn't_ push down on the gas pedal while going up the hill. The relationship between the gas pedal and the speed of the car is entirely obscured.

However, in a real world system, the chance that there is zero imbalance is also nearly infinitesimal. You could take the most skilled driver in the world and you would still be able to pull out the relationship between gas pedal and speed out of his driving pattern. Similarly, with the right variables and enough data, you could pull a relationship like the one described here out of economic data.

(As a side note, I think the way in which this article is written is pretty obnoxious. The first 500 words are dedicated to the author congratulating himself for being aware of an idea that he hasn't even explained yet.)

ratsbane• 14 years ago

Exactly. This whole article could have been distilled down to http://en.wikipedia.org/wiki/Multicollinearity (but I still thought it was a good read)

klochner3• 14 years ago

Am I the only one who read Friedman's paper? [1] The answer to OP's question [2] - "why are economists unaware of milton friedman's thermostat" - is because the OP is misinterpreting Friedman's paper, and therefore has misattributed this nonexistent "theory".

Friedman just said that the Fed acts like a thermostat, and in the linked paper, claimed that the period subsequent to 1985 had increased price stability owing to improvements in their understanding of inflation and it's relation to the quantity of money, hence their "better thermostat".

AFAICT, the OP has made up this "Friedman's Thermostat" insofar as it relates to economic forecasting, and now he wants to know why no one has heard of it.

As an aside, "exception that proves the rule" is an almost universally misapplied phrase, that should only be used if you're really really sure you know what it means (kind of like "begs the question").

[1] http://online.wsj.com/article/0,,SB106125694925954100,00.htm...

[2] OP = Nick Rowe - http://worthwhile.typepad.com/worthwhile_canadian_initi/abou...

taylorwc• 14 years ago

Agreed. An analogy made by Friedman in a WSJ op-ed != established theory.

Symmetry4• 14 years ago

Right, if you find yourself asking "Why has nobody else noticed this" you should always consider that the answer is "because it isn't there." But on the other hand, the idea that intentional behavior can remove correlation from causal relationships is important and deserves more thought than people give it. And since economics has something of a history of naming things for the wrong person, calling this "Friedman's Thermostat" and spreading it around might be a decent idea after all.

j2labs• 14 years ago

The delivery is so emotionally charged too. It took several paragraphs of "I can't believe no one gets this!" before we even got to Milton's theory.

gweinberg• 14 years ago

It's worse than that. He rambles on, and then immediately shifts the metaphor to a more complicated and less clear one. He should have just stuck with the thermostat: if the temperature inside a house is determined by a thermostat, the outside temperature has no correlation with the inside temperature.

marris5• 14 years ago

Sorry, what do you mean "OP is misinterpreting Friedman's paper"? I don't think there's much "interpretation" here. He's basically saying: here is a paper about X. It is titled "X".

And I'm also confused by what you mean by "the OP has made up this 'Friedman's Thermostat'"... seriously, what does that mean? Friedman wrote a paper on Friedman's Thermostat. So how could OP [whoever that is, I assume you mean Nick Rowe] have "made it up"?

chiquitabacana2• 14 years ago

the article is "the FED's thermostat" as an analogy about the Fed's. There is no "Thermostat Theory" it's just a title, not a general theory.

pretoriusB• 14 years ago

>Friedman wrote a paper on Friedman's Thermostat.

No he didn't.

He just used the WORD "thermostat" once, in an article.

And his use of the word was unrelated to the concept presented in the blog post (which is not Friedman's but older and which nobody calls "Friedman's thermostat" except the author of the blog post).

Here's an analogy for the situation:

David Pogue writes and article titled "The perfect gadget for Christmas", which is simply a review of 10 hot current gadgets.

Then, 30 years later, some blogger writes "How come nobody knows "Pogue's Perfect Gadget" theory? You know the one that says that when you put a cat inside a gadget's enclosure with a quantum based poisoning device the cat's state is both dead and alive?"

He then goes to admit that idea: is not Pogues, it was found earlier, and nobody calls it that -- all the while complaining why people don't know it by his made up name "whereas they do some other Pogue's articles".

The whole blog post smells crank to me...

cynicalkane4• 14 years ago

Am I the only one who read the blog article? It's an old idea that shows up in the literature of various smart economists, and is in fact just a name for a common phenomenon in statistics, with the twist that human behavior causes it. He's not criticizing people for not having heard of the name (which he seems to have made up in 2010), he's criticizing people for being unaware of the idea despite it being an old idea and, he alleges, a very important one when you start studying human behavior. All of this is in the blog post.

klochner3• 14 years ago

cynicalkane: "He's not criticizing people for not having heard of the name"

Nick Rowe: "Google seems to tell me I'm right. I'm the first link, which is really pathetic for such an important idea."

Not to mention that googling for "Fed Thermostat" brings up a lot of references. Friedman's is in the top page, Nick's is not:

https://www.google.com/search?q=fed+thermostat

MaysonL• 14 years ago

Am I the only one who read the Casey Mulligan blog post which kicked off Professor Rowe's rant? Actually I didn't read the whole thing; I couldn't force myself to push past the egregious statement that "the Federal Reserve’s monetary policy has little effect on a number of financial markets, let alone the wider economy." Anyone who lived through the recessions of the early '80s in the US and believes that is sadly unable to learn from experience.

eykanal• 14 years ago

OK, I think I'm missing something. To quote the article:

> And no, you can not get around this problem by doing a multivariate regression of speed on gas pedal and hill. That's because gas pedal and hill will be perfectly colinear.

This is one of the reasons why you check for multicollinearity [1] when performing multivariate linear regressions; aside from introducing significant instability into the model, if your predictors are correlated, then you're essentially measuring the same thing twice. The stated problem is actually a good example as to why you should avoid having highly correlated predictors in a multivariate model; by blindly pursuing the regression despite the inputs being correlated, we miss out on the fact that there is a relationship present (i.e., the dependent variable (speed) is actually a factor of both of the independent variables (pedal height and hill slope)).

[1] http://en.wikipedia.org/wiki/Multicollinearity

Irregardless• 14 years ago

Sounds like the nail in the coffin on this topic.

It always sucks when you accuse an entire profession of overlooking a fundamental principle, only to find out you've committed the same error in a more egregious fashion.

cynicalkane4• 14 years ago

You aren't missing anything. The author's article isn't about the optimal ways to do statistics, it's a criticism of the informal economic discussion that dominates policy, blogs, and pop-economics--and, according to the post, an unfortunate number of actual econ papers--that says things like "economic plan X doesn't work because there hasn't been a change".

The criticism generalizes to any statistical methods, though. You always need to think about the underlying model before doing any kind of statistics in any science.

chimeracoder2• 14 years ago

Yes, but I wouldn't say that "almost all" economists don't understand the principle (as in the title). I would be more likely to say that "almost all" reporters don't understand basic economics & statistics.

chimeracoder2• 14 years ago

For anybody who's interested: perfect multicollinearity (which is rare, except in case of human oversight) causes a matrix of less-than-full rank, which essentially means that it's not invertible (the matrix analog of "divide-by-zero" error).

More commonly, you'll end up with imperfect multicollinearity, which means that two variables are almost the same, but slightly different. The danger with perfect multicollinearity is that (a) you'll end up with very high variances associated with those predictors, but furthermore (b) the predictor reported as being the 'strongest' may actually not be the strongest. In fact, it may not even be relevant; the model just "happens to be associating most of the variation in the response variable with that predictor. It's a more complicated version of this principle: http://xkcd.com/882/

I mention this because I see people - including people who should know better! - falling into this trap all the time. It's easy to overlook, particularly when you see the results you were expecting.

marris5• 14 years ago

No, that's not the point. The point is that when velocity is constant, there's no indicator that your pedal manipulation is good as opposed to bad. For that to happen, you would need a period of time where you messed up with the pedal, and saw that you messed up. Then you could make an argument like: Hey, when we let hillAngle-pedalAngle [or whatever] drift away from zero, things go bad [velocity will fall].

Now it is true that you can measure collinearity between the hillAngle and the pedalAngle and show that they are correlated... but we sorta knew this already, didn't we? After all, we're setting the pedalAngle based on the hillAngle. We created that functional relationship. There's no need to empirically discover it.

Put it another way. Suppose v = w1 * hillAngle + w2 * pedalAngle + v0, and you want to find w1 and w2.

If your dataset shows pedalAngle = -hillAngle, then you have

v = w1 * (-pedalAngle) + w2 * pedalAngle + v0 = (w2 - w1) * 0 + v0

So you're dataset tells you nothing about w1 and w2.

mathattack• 14 years ago

A similar problem happened at Harvard Business School. In the mid 80s they found no correlation between GMAT scores and success at the school, so they stopped requiring applicants to submit scores. This was a thermostat-unaware view.

The real answer was that they were giving it the right amount of weight. If they were overweighting the GMAT, unqualified people with high GMATs would get in. If they were not underweighting it, the opposite.

Approximately 10 years later they did a correlation of student GMATs (which weren't used for admissions purposes) and school success, and found that there was indeed now a correlation.

The lesson was that if you're properly managing a factor, you won't see it in the outcome. Or in the thermostat example, if you're properly managing the thermostat, you won't see the changes.

vavoida• 14 years ago

find that explanation shorter & faster, although polsci

http://themonkeycage.org/blog/2012/07/31/milton-friedmans-th...

tomrod• 14 years ago

In addition to klochner's comment, this is why economists rely on exogenous variation. Any applied econometrician worth his or her salt would not say "the gas pedal has no effect", but that the data doesn't rule out the hypothesis that it does.

dbecker4• 14 years ago

I am an econometrician... exactly the sub-population that the author faults most for not knowing about Friedman's thermostat.

We haven't heard of the thermostat because it does not introduce new ideas to us. If the data includes hill height, this is an example of multicollinearity. If your data does not include hill height, this is an example of endogeneity.

Both concepts are taught in considerable depth in an undergraduates first course in econometrics. It's a nice examples, but the author's claim that these phenomena are outside are awareness shows a striking ignorance of basic econometrics.

ScottBurson• 14 years ago

Okay, fine, but then what do you think of the Casey Mulligan column that the author links to (third paragraph)? It certainly seems at first glance to contain instances of the error the author is describing.

(I am not an economist and have no position on this; just curious.)

dbecker4• 14 years ago

That's a great question, and I'm afraid I'm not going to do it justice. But the place where I think you see the thermostat concern is in the paragraph:

A 1983 study by Lars Peter Hansen of the University of Chicago and Kenneth Singleton of Stanford showed that short-term rates on Treasury bills and short-term returns on stocks traded on the New York Stock Exchange had very little correlation with consumer spending. Many empirical studies have confirmed this sort of result (this comparison of inflation-adjusted Treasury bill returns and business sector profitability is a recent example).

This summarizes "many empirical studies" in a single sentence. Unfortunately, it doesn't include citations.

There are a number of ways to address the "thermostat" problem. Each potential solution only works in specific circumstances. I'd hope the underlying research adequately addresses this concern, but it's hard to say without seeing the research.

In general, summarizing "many" 40 page papers in a paragraph is inherently difficult, and Mulligan seems to be especially vague here. If the claims here have any credibility, it would only be through an "appeal to authority," which most of us won't find very compelling.

So to answer your question, I'm unimpressed with the Mulligan column... in part because we have no indication what problems there are in the underlying analysis.

DannyBee• 14 years ago

As a meta-comment, this article dances around and only after an entire page of text, does it get to the actual idea.

danielweber• 14 years ago

Why does it 11 paragraphs repeating himself about how important this idea is? I thought this was some set-up for a joke in which he never tells us about the thermostat.

pseut3• 14 years ago

I have no idea if the author describes "friedman's thermostat" correctly, but to say that Economists are unaware of the phenomenon he describes is beyond absurd (disclosure: I'm an Economist!).

There's another way to find the relationship between the gas pedal, slope, and speed that the author did not mention (but is somewhat related to some of his suggestions): write down a model of the driver that assumes he or she uses the gas pedal to try to keep speed constant:

E_{t-1} speed(gas_t, slope_t, parameters) = const (eq1)

where E_{t-1} is the conditional expectation of the period-t term, given the information available in period t-1. Maybe speed should be nonparametric, but that will introduce some new problems in estimation; maybe you know enough about physics to write down a parametric formula for speed; whatever, the exact details of speed(.,.) are kind of beside the point for the author's argument.

Now,

speed(gas_t, slope_t, parameters) - E_{t-1} speed(gas_t, slope_t, parameters) (eq2)

is a martingale difference sequence when "parameters" is set to their true value, so the sequence equals zero in expectation for all t, assuming the model is true, and that can be the basis for estimation through, say, Generalized Method of Moments. Because combining (eq1) and (eq2) gives us

E speed(gas_t, slope_t, parameters) = const,

so you can estimate the parameters as

\hat parameters = argmin (average_over_t speed(gas_t, slope_t, parameters) - const)^2

Everything I laid out is an extreme simplification of the DOMINANT STRATEGY in applied macro (with, in all likelihood, some errors due to sloppiness). Now, for any realistic economy, it's going to be hard as hell to write down a sensible formula for "speed(.,.,.)" and since there aren't that many years since WWII (which is kind of seen as the beginning of the "modern" economy) there isn't a lot of data to estimate the model, but neither of those issues have anything to do with the author's "critique" and are basically the direction of almost all research in macro.

ignostic2• 14 years ago

At least the author admits that a lot of people know more about this stuff than he does. It's tough to make big assertions when you're relatively naive.

Nick_Rowe2• 14 years ago

pseut: "so you can estimate the parameters as

\hat parameters = argmin (average_over_t speed(gas_t, slope_t, parameters) - const)^2"

No you can't.

Think about it.

1. If the driver has as much information as the econometrician, setting both parameters on slope and gas equal to 0 will fit the data equally well.

2. If the econometrician observes slope and gas and speed with error (as will almost always be the case), then GMM will estimate both parameters on slope and gas as zero (if those 3 errors are independent of each other), even if the true parameters aren't.

3. If the driver observes gas or slope with error, and the econometrician doesn't, then the econometrician can indeed actually estimate the parameters.

How likely is 3? Only a stupidly arrogant econometrician would assume a priori that he knows better how to drive the car than the guy actually driving it. (OK, maybe the econometrician has final revised data, and the driver has only real time data, and the final revised data is better than the real time data. But even then the driver will be observing other indicators that the econometrician doesn't have data on, so the econometrician will interpret the driver's response to those omitted variables as "gas pedal shocks", and will screw up the estimation royally.

Yep, all that Sims VAR stuff is wrong. Here, read this.

http://worthwhile.typepad.com/worthwhile_canadian_initi/2011...

You have just proved my point: economists don't understand Milton Friedman's thermostat.

Forget your fancy stuff. Just STOP AND THINK about what you are really doing when you try to estimate parameters.

pseut3• 14 years ago

I should have been a little more specific; the approach I described assumes that the driver makes some errors, so the speed is not actually constant in practice. That can be because the driver slightly errs in seeing the slope or gas, or because of other sources of error (gusts of wind, maybe). Without that variation, you're right, the system is unidentified. The larger the variation, the better the identification. But none of this is news to economists, and none fundamentally requires the econometrician to have better information than the driver (although measurement error can matter in some circumstances, it matters in a different way).

As I'm sure you know, these arguments (i.e. mainstream Economics is wrong and ignorant!) are more convincing when they're accompanied by a model. DSGE models have a lot of limitations, but they're pretty good for demonstrating failure of identification. Monetary policy is tricky to identify, and I'd be sympathetic to an argument showing that deviations from a Taylor-rule are bad for identifying the effects of monetary policy shocks (lots of people would agree with this, the interesting question is whether they're bad in empirically important ways or just conceptually bad) but your car+driver analogy seems like it's aiming to be broader than that.

ignostic2• 14 years ago

I thought he was never going to get to the point. This entire post should be about 1/5th the size.

wissler• 14 years ago

This is a subset of David Hume's analysis of causality, see his "An Enquiry Concerning Human Understanding."

baddox• 14 years ago

There are exactly 500 words before a single bit of substance in this article.

Milton Friedman's Thermostat

Related Stories