Is economics a science? Should we care?

I’ve always thought that the debate over whether economics is a science is actually a debate about the meaning of the term ‘science.’  If science is when people build models to better understand the world around us, then economics is a science.

Some people distinguish between soft sciences like economics, and hard sciences like physics.  I don’t understand that distinction.  Are physicists very good at predict earthquakes, tornadoes, heat waves, etc?  Obviously not.  Some people claim that those events belong in geology or meteorology, but not physics.  My response is that physicists claim that their models explain those phenomena, so we have just as much right to expect them to predict tsunamis as we have to ask economists to predict recessions.  Indeed even more of a right, as economics has a sort of Heisenberg uncertainty principle, which says that (demand-side) recessions should not occur if they are expected.

Economics can predict some things very well and other things not so well.  The more complex the system, the less well we predict–just like physics.  I predict that bond prices will be unusually volatile between 8:30 and 8:35am on the first Friday of each month over the next few years.  But our most useful predictions are conditional forecasts.

Tyler Cowen has an interesting post on this question, but ends up with a perplexing assertion:

I conclude that economics is not yet a science.  Economics is most like a science when people do not care about the outcome of the argument.

Tyler is much better read than I am, but I was under the impression that scientists often became emotionally attached to their theories, and that it was hard to dislodge them.  Didn’t Max Planck say that science progresses one funeral at a time?

There is also a widespread view that we economists should try to become more scientific.  I’m not sure what that means.  If economics is not a science, I don’t think it should try to be one; just as I don’t think music, law, and accounting should try to become sciences.  We should try to become more useful, not more scientific.  I happen to think we are a science, BTW, as we do build models and use them to try to explain cause and effect in the world around us.  But if I am wrong and we aren’t a science, then I presume there is a good reason for us not being a science.

To summarize:  I see us as being a science.  Those who disagree with me simply define science differently.  It makes no difference whether we are or are not a science; all that matters is whether we are useful.  Law and music are not sciences, but they are very useful.  I hope this blog is useful.


Tags:

 
 
 

55 Responses to “Is economics a science? Should we care?”

  1. Gravatar of marcus nunes marcus nunes
    19. March 2011 at 13:50

    On this topic here is a 22 year old article by Hal Varian:
    http://people.ischool.berkeley.edu/~hal/Papers/theory.pdf

  2. Gravatar of Scott Sumner Scott Sumner
    19. March 2011 at 14:14

    Thanks Marcus, That’s an interesting article.

  3. Gravatar of MW MW
    19. March 2011 at 14:51

    You said what you feel is and is not a science. But you have not said what is and is not useful. Surely there are some fields that are not useful or less useful than others.

    Given that we currently subsidize all degrees equally, can it be said that all fields of study are equally useful?

  4. Gravatar of Philo Philo
    19. March 2011 at 15:07

    Varian’s essay is fine. But he mangles the Bacon quote about error and confusion . . . and he attributes it to *Roger* Bacon rather than *Francis* (a 340-year error).

  5. Gravatar of Joe Joe
    19. March 2011 at 15:14

    Its not about the ability of making predictions. The difference is what sort of experiments are used to obtain knowledge.

    The chief difference between sciences and non sciences is the degree to which predictions can be verified by and accountable to “controlled experiments.” It is a spectrum. The closer you move towards the ability to make controlled studies and experiments the more like science you are. On the opposite side, the closer you are to the use of “natural experiments” the closer you are to philosophy.

    So, the two extremes are Philosophy (natural experiments) Science (controlled experiments)

    In college, I spent four years studying the history of western philosophy (ancient, medieval, modern, 20th century, etc.)

    Some examples of economics as philosophy
    1. Nick Rowe’s debates with Delong over whether recessions are caused by an “excess demand for the medium exchange” is entirely a philosophical debate, not a scientific. I can assure you that if you read the history of philosophy from say, Descartes, to Kant (200 year period) you will see absolutely no difference between the nature of the two arguments.
    2. Your debate with Krugman over the liquidity trap. It is philosophical, NOT scientific.
    3. The debates between Tobin and Friedman.
    4. Clower/Leijonhufvud and non-walrasian econ
    5. RBC versus sticky prices/wages

    I am continually amazed at how all the debates in econ are no different in any manner or nature from the major historical debates on metaphysics and epistemology.

    Metaphysics historically, from ancient to modern times, involved really really smart guys trying to create an objective model of reality from which we could get first principles and from which we could deduce everything else. Every era of debates in western philosophy was really just two or three sides debating over whose model of reality was better. Then, with the rise of the “sciences” in the 19th century, such questions were discarded by “scientists,” who instead focused on controlled experiments. Here are two more great examples.

    1. new classical economics said all unemployment was voluntary because they created a model of reality. Tobin and Solow jokingly dismissed such idea with phrases like “if it looks like its no voluntary, then its no voluntary!”

    2. DSGE versus you. They attempt to create very complex models of reality, in otherwise to obtain first principles. You ignore all that and are simply are an empiricist. You have never attempted to formally model your ideas. You do the best with what you can.

    It is the fundamental difference between rationalism and empiricism. Read the following LONG!!! article on the historical debate… you’ll see

    http://plato.stanford.edu/entries/rationalism-empiricism/

    Economics is simply another more modern branch of philosophy in which the debate between rationalism and empiricism continues. Modern Economics, like the 2500 years of western philosophy can only rely on natural experiments, ALMOST never on controlled ones, which is what science does.

    I hope that made sense….

  6. Gravatar of Joe Joe
    19. March 2011 at 15:22

    One more thing,

    Just take a look at Nick Rowe’s recent post.

    http://worthwhile.typepad.com/worthwhile_canadian_initi/2011/03/do-keynesians-understand-their-own-models.html

    There is rationalism -> “You could argue that the research program called RBC theory, or Classical Macroeconomics (a strange name for something that only started in the 1980’s), is an example of the Ricardian Vice. RBC theory is what happens when you insist on modelling everything formally. But who should we blame for the Ricardian Vice?”

    Empiricism -> “one of the complaints that the Keynesians like James Tobin had against Friedman was that Friedman didn’t write down his macroeconomic model in equations so they could see precisely where Friedman’s Monetarism differed from their Keynesianism.”

    Once again, its a spectrum, not black and white.

    Best wishes!

    Joe

  7. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 15:44

    Didn’t Adam Ozimek open this topic?

    “Another question is, if economics weren’t a science, then would previous paradigms so have been done in by empirical outcomes? The old Keynesian Phillips Curve held that there was a tradeoff between inflation and unemployment. When that relationship broke down during the stagflation of the 70s, the Phillips Curve was invalidated, and this helped shift macro away from old Keynesianism and towards the new classical paradigm. Real Business Cycle models of the 80s were also invalidated by reality: it was clear that money mattered, and in the real world it was hard to find technology shocks to explain actual recessions.”

    http://modeledbehavior.com/2011/03/17/is-economics-a-science/

  8. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 15:47

    Krugman responded:

    “Um, there’s a problem here. Yes, the old Keynesian Phillips curve was abandoned in the face of evidence. But while real business cycle theory has indeed been “invalidated by reality”, as far as I can tell it’s still going strong in freshwater departments.”

    http://krugman.blogs.nytimes.com/2011/03/17/economics-as-a-science-a-bad-example/

    My own department (Delaware) is hard to classify as “freshwater” but it is so Plosser dominated in influence that I share Krugman’s skepticism. As a pointed out in a prevous post RBC is still taken very seriously by my department despite the fact it has absolutely no correspondence with economic reality.

  9. Gravatar of Scott Sumner Scott Sumner
    19. March 2011 at 16:13

    MW, We don’t subsidize all fields equally, but even if we did it wouldn’t imply they were equally useful.

    Joe, I have lots of issues with your comment. First, economics does do controlled experiments, so the entire premise seems wrong. Second, fields like astronomy typically are considered science, and yet rely on natural experiments.

    I don’t see much connection between economics and philosophy, although there are philosophical aspects to economics, but the same is true of physics (Should string theory be taken seriously if it can’t be tested? Etc.)

    You said;

    “You ignore all that and are simply are an empiricist. You have never attempted to formally model your ideas. You do the best with what you can.”

    Nothing could be further from the truth. My studies are very much informed by theoretical models, which I take very seriously.

    Don’t pay too much attention to debates over terms like “voluntary unemployment.” The only debate that matters is over the issue of how unemployed workers respond to various changes in the incentives they face, like 99 week UI. The term “voluntary” is so vague as to be almost useless. It’s like debating whether union workers are “greedy” in asking for wage increases.

    Mark, I was taught that starting in the late 80s the real business cycle models began adding monetary frictions. Is that not true? And I would say they are not completely divorced from data. Wasn’t one motivation for RBC theory the failure of econometricians to find trend reversion in RGDP? RBC economics is really just applied microeconomics. The question is not whether it is true, but how much it can explain. I happen to think it doesn’t explain much of the more severe contractions.

  10. Gravatar of Benjamin Cole Benjamin Cole
    19. March 2011 at 16:14

    Well, science. Since we cannot set up replicable experiments, I don’t think economics will ever be a science. For example, what if we today abolished the USDA, including the large food stamp component. Maybe it would lead to just a few months of starvation, and then the population would adjust. Or maybe some people would starve to death every year for the next 30 years.

    I would like to abolish the USDA and the food stamp program–but I confess, the thought of seeing starving people makes me worry. So maybe we never kill off the food stamp program, meaning we can never truly conduct an experiment, let alone replicate it. Oh sure, small-scale pilot studies can be done, but small-scale studies seem to prove whatever the conductors of the study want to prove.

    Still, is economics useful? I contend it is, on basic levels. Does more currency in circulation stimulate the economy, and to what point? At some point it will lead to inflation. I think common sense and economics can lead to roughly right answers–we can stimulate until capacity is roughly reached. I contend that in economics it is fine to be roughly right, better than to be exactly wrong.

    Sometimes I compare economics to driving a car. You make a left hand turn. There is some play in the wheel (especially in my 1990 Isuzu Trooper). You do not turn the wheel to exactly 91.6 degrees for exactly 3.123 seconds, etc. There is no need to be so precise, nor can such precision mean much with varying terrain, road conditions, sudden emergence of speeding motorcycles etc.

    This is one reason I am suspect of overly complex mathematical models of economies. It assumes a precision in measurement and results that is not there–or that even everything that should be measured is being measured.

    But we know roughly that if we turn the wheel left the car will go left. If we need more gas, we push the pedal down. We know if we feed more gas in, and the engine is not flooded, we will go faster. If we put more impediments in the system, we will get lower output. Airbags may be good; but they will slow the car down a just a bit, leading to lower performance (the heavier weight of the vehicle).

    You know, this car analogy is not all bad.

  11. Gravatar of rhmurphy rhmurphy
    19. March 2011 at 16:15

    Snarky answer: economics is not falsifiable because we can always fall back on saying ceteris wasn’t paribus. Therefore, it is not a science.

    I don’t like to get bent out of shape over debating an essentialism like whether the category of study we call economics belongs in the category of knowledge we call science. But I have problems coming up with a useful definition of science that economics would fit under.

    Your definition of science – model building and cause and effect is a common positivist position in the economics profession, but positivism has been dead in philosophy of science for 40 years.

  12. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 16:28

    Scott wrote:
    “Mark, I was taught that starting in the late 80s the real business cycle models began adding monetary frictions. Is that not true? And I would say they are not completely divorced from data. Wasn’t one motivation for RBC theory the failure of econometricians to find trend reversion in RGDP? RBC economics is really just applied microeconomics. The question is not whether it is true, but how much it can explain. I happen to think it doesn’t explain much of the more severe contractions.”

    True, RBC began adding monetary frictions but only obligingly. Having to share the same air that they breath I can assure you that they think money is total nonsense. And absolutely none of the RBC-centric research seminars I have observed at the UD recently (that’s pretty much all there is here) take into account money.

    IMO the most that applied micro can explain in terms of business cycles is 30% and that was all during the Great Moderation. Micro explains almost nothing of the current contraction (as I know you agree), but that doesn’t prevent the RBCers from living in a self imposed bubble that blocks out all interaction with reality.

  13. Gravatar of David Pearson David Pearson
    19. March 2011 at 16:59

    Economics is most useful in improving our grasp of conditional probabilities. First, if the Fed raises expectations of future NGDP levels, economic theory tells us the price of inelastic goods in tight supply/demand would be most affected. Unfortunately, it does very little to tell us how much the resulting change in price will be due to easing, and how much to real effects.

    Second, the portfolio balances thesis tells us that a Fed-induced change in the mix of privately held assets will prompt a re-balancing towards an optimal asset mix. Conditional on QE, we would expect actors to allocate into commodities funds, and, all else equal, for commodities prices to rise as a result.

    Third, we would expect that same portfolio balances thesis to drive funds into higher yield emerging markets liabilities. The resulting flow of funds might, in a fixed-exchange rate regime, drive up the money supply of that country. All else equal, this would raise nominal demand for commodity imports, and raise commodity prices.

    Fourth, as a result of Fed easing, we would expect import prices to be disproportionately affected.

    In short, our knowledge of economics sheds light on conditional probabilities, which in turn tell us that Fed easing should result in the relative prices of commodities and import prices rising. In fact they did! Who says economics is not a science?

    Sometimes economics is less useful. For instance, will relative price distortions cancel out the positive effects of higher NGDP expectations? For instance, if a certain group (say, middle class consumers) sees its real wages fall disproportionately, and their real wealth (housing equity)also fall, will they quickly pull back on spending, thus offsetting a potential rise in aggregate income? I’ve not seen a piece of economics research that attempts to answer that question.

  14. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 17:24

    Am I the only one to notice his absence? Where is the Morgan? I need a good argument.

  15. Gravatar of OGT OGT
    19. March 2011 at 17:57

    Joe good comment, that’s the dichotomy that came to my mind when this debate began to make the rounds.

    It also brought to mind this interview with Ronald Coase:

    …WN: You mentioned many times that you do not like the term, “Coasean economics”, and prefer to call it simply the “right economics” or “good economics”. What separates the good from bad, the right from wrong?

    RC: The bad or wrong economics is what I called the “blackboard economics”. It does not study the real world economy. Instead, its efforts are on an imaginary world that exists only in the mind of economists, for example, the zero-transaction cost world.

    Ideas and imaginations are terribly important in economic research or any pursuit of science. But the subject of study has to be real.

    http://economistsview.typepad.com/economistsview/2011/01/coase-does-not-like-the-term-coasean-economics.html

    There seems to be too much bad “blackboard economics,” much of which is not useful, and perhaps harmful.

  16. Gravatar of W. Peden W. Peden
    19. March 2011 at 18:01

    Joe,

    “Modern Economics, like the 2500 years of western philosophy can only rely on natural experiments, ALMOST never on controlled ones, which is what science does.”

    ‘Control’ is an ambiguous term. If it just means controlling for certain variables, then natural experiments in economics can easily be made into controlled experiments in most cases.

    If it means controlling for every variable, then that’s an impossible ideal. No experiment is controlled for every variable, because no two experiments can take place under literally identical conditions.

    The only ideal we can attain with controls is to control for those variables which our current theoretical position suggests are important. In this respect, economics is in the same boat as physics, chemistry, biology and so on.

  17. Gravatar of W. Peden W. Peden
    19. March 2011 at 18:21

    Prof. Sumner,

    I recently wrote an essay about this topic for my masters applications. Most of what I wrote was a critique of a specific philosopher, but one thing I did note was that predictions in economics often take place in a very competitive context as part of the rhetoric between two rival theories.

    For example, Phillips used the apparent retrodictive power of his link between money wages and employment to bolster his theory’s standing, because that power suggested that the Phillips Curve was a causal link. Then Friedman and Phelps presented theoretical arguments for why this correlation was not a case of causation and their arguments were vindicated by their prediction of the mess of the early 1970s. The next result of this competition has been that the idea of a long-term trade-off between inflation and unemployment came into fashion and passed out of fashion.

    It is notable that it tends to be big singular predictions (and retrodictions) that tend to turn the tides, rather than econometric work. I think this is because straightforward prose is more effective at arguing for causal links than econometric work, mainly because we all know that a good (or, rather, clever) econometrician can demonstrate just about anything with any data set given enough statistical contortions.

    As I see it right now, the following are the key problems in the methodology of economics:

    (1) How to develop a set of regularly used, generally accepted and sufficiently challenging empirical methods. I actually suspect that economic historians are doing better in this task than econometricians.

    (2) How to move beyond Homo Economicus, particularly in respect of methodological psychological egoism.

    (3) A more systematic mapping of what economists do and do not take literally. No-one (not even Friedman at his most pragmatic) thinks that economics is a purely instrumental predictive affair and no-one (I hope) takes everything to be descriptive. The ‘spray test’ (“If you can spray it, then it’s real”) needs to be applied to work out just what we should be realists about.

    (4) Utility really needs sorting out. Just as Marxian economics is largely based on a very confused and unempirical notion of “socially useful labour”, so economics since Menger has had utility rotting at its core. How empirical can we make it? What does it mean? Is it worth having?

    (5) The idea that macroeconomics needs to be reduced to microeconomics must be stamped-out.

    Broadly speaking, my ideal methodology of economics would be not-quite-history and not-quite-physics, but would bear resemblances to both. I think that the theoretical clarity of physics is admirable, but economics is dealing with far less robust regularities and searching after “laws” will leave economists with willow-the-wisp laws like psychological egoism that are either tautological or false. I think that the historical method is the best test of causality that is currently available to economists, but I find that economic history often lacks the theoretical clarity necessary for harsh tests.

  18. Gravatar of W. Peden W. Peden
    19. March 2011 at 18:26

    The rationalist-empiricist distinction is easy to make in the philosophy of economics, though I find it’s harder to maintain in the history of economics. For example, Friedman’s professed methodology is extremely empiricist and Von Mises’s professed methodology is extremely rationalistic, but their actual practice was far less different. I like aspects of both; Von Mises’s regression theorem and version of the economic calculation problem are both fascinating, even if they’re distant from modern mainstream economics.

  19. Gravatar of Dave Dave
    19. March 2011 at 18:31

    Peden,

    I would agree with Joe in that (I believe) he means control which might affect the outcome of the experiment. It’s true that literally identical conditions can never be achieved for any two experiments (if nothing else they must happen at different places or different times) but in “hard” sciences every variable which could matter is controlled for (generally speaking). In chemistry or physics the outcome of many experiments can be predicted with 100% certainty in many cases…

    My own offbeat view is that science is descriptive rather than predictive. That is, it tells us about physical phenomena which are measurable and repeatable. I don’t for example, view evolution as a science (I definitely believe in it, to be clear) but view it is a predictive theory which has grown from scientific observation about biology, botany, geology, etc.

    I realize that this considerably narrows the definition of science, but it’s the only definition of “science” I’ve heard or come up with that did not strike me as ultimately rather arbitrary. Scott makes great points above about exactly some of the difficulties defining science using other definitions, and as he correctly points out, just because something might not be properly considered “science” does not at all mean it isn’t useful…

  20. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 18:34

    W. Peden,
    You wrote:
    “I recently wrote an essay about this topic for my masters applications.”

    Was that in the subject of Philosophy? Just wondering. I always like to keep track of our expertise at this site.

  21. Gravatar of W. Peden W. Peden
    19. March 2011 at 18:37

    Apropos nothing in particular, I’d be interested in how various people on here would handle the US budget:

    http://www.nytimes.com/interactive/2010/11/13/weekinreview/deficits-graphic.html

    This is how I handled it-

    http://www.nytimes.com/interactive/2010/11/13/weekinreview/deficits-graphic.html?choices=hcbcp0k0

    Of course, there are more complex games out there, but this is good if you only have half an hour (or less) spare.

    Also, for those with too much time who are after a challenge and seeking to test interest rates to their limit, Belgium has given us this little game:

    http://www.nbbmuseum.be/2010/12/economia.htm

  22. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 18:41

    Dave,
    You wrote:
    “My own offbeat view is that science is descriptive rather than predictive.”

    Not so offbeat. That’s my view. I suspect it is also Scott’s view.

  23. Gravatar of W. Peden W. Peden
    19. March 2011 at 18:58

    Mark A. Sadowski,

    Yes. My interest in economics up until this point has been purely non-academic.

  24. Gravatar of W. Peden W. Peden
    19. March 2011 at 19:06

    Dave,

    I agree that one can control for variables that (we think) matter, but I think natural experiments can be controlled in this way e.g. one can exclude particular historical incidents on the basis of controls.

    “In chemistry or physics the outcome of many experiments can be predicted with 100% certainty in many cases…”

    Indeed. My (very controversial and undeveloped) view is that some sciences deal with what we seem to be universal regularities, while others deal with more mundane regularities. Economics falls into the latter category. However, the debate in philosophy between those who believe in necessary laws of nature (either ordained by God or as some sort of causal forces) and those who believe that the ‘laws of nature’ are just regularities which apply in every case is an ongoing one and neither side has a fully satisfying position.

  25. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 19:10

    W. Peden,
    Well I declare. You misunderstood my question.

    You are a Philosopher.

  26. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 19:13

    W. Peden,
    I don’t mean that in a bad way but in a good way.

  27. Gravatar of Kevin Dick Kevin Dick
    19. March 2011 at 19:45

    Controlled experiments are a red herring. They provide one possible source of evidence. And in fact, all experiments are on a spectrum of control. None is perfectly controlled.

    My opinion is that any discipline that is trying to create a predictive model of some aspect of reality is a science. More specifically, any discipline that uses Bayesian updating–proposing a model, observing events, updating the model–is a science.

    By this definition, Scott certainly treats economics like a science.

  28. Gravatar of W. Peden W. Peden
    19. March 2011 at 19:47

    Mark A. Sadowski,

    I think the fact that it is 3:45 at the end of a hard Saturday night is a more likely cause of that misunderstanding.

    The essay is a philosophy of science essay, a philosophy of science/history of science essay, or a history of science essay, depending on the particular application.

  29. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 19:57

    W. Peden,
    “I think the fact that it is 3:45 at the end of a hard Saturday night is a more likely cause of that misunderstanding.”

    Speak for yourself. You’re several hours ahead of me.

    Congratulations (even if premature). I always like to hear about the academic success of our loyal members.

  30. Gravatar of Dave Dave
    19. March 2011 at 20:54

    W. Peden,

    Thanks for the reply. I do understand the idea of controlling for variables in what might be described as the “softer” sciences (economics, or some social sciences).

    The problem – at least in my experience – is that these controls are practically impossible to implement completely. I believe this is why any question of any complexity in these sciences is the subject of such vigorous debate (as indeed are the particulars of evolution, like I mentioned). In any predictive model made by economics or political science factors like culture and current events will have a significant effect, and that effect will be difficult to describe and quantify and nearly impossible to replicate later.

    To be honest, I’m not sure exactly what you mean by universal versus mundane regularities, so I’m not really able to comment on the second part of your reply. Thanks again.

    Mark,

    Thanks for the second. Nice not to feel like an island every once in a while.

  31. Gravatar of Mark A. Sadowski Mark A. Sadowski
    19. March 2011 at 21:10

    Dave,
    You wrote:
    “Nice not to feel like an island every once in a while.”

    Yes I know the feeling.

  32. Gravatar of Mike Sandifer Mike Sandifer
    19. March 2011 at 21:15

    Science is just a method and nothing more. It can be applied to anything, including law, music, and accounting.

  33. Gravatar of Jon Jon
    19. March 2011 at 23:18

    In physics, five sigma is the gold-standard. Go to arxiv.org and look at any experimental paper claiming to have discovered anything: a novel discovery has to meet a three sigma standard to get published, and it takes a five sigma result for the community to treat as proven even something that ‘makes sense’.

    Conversely, well trod questions in economics, such as the effect of minimum wage increases, are routinely republished with two-sigma results. That so many of these papers contradict each other should not surprise.

    Why? Consider, if the “signal to noise ratio” is a priori distributed over several orders of magnitude, then for the ratio to be between 2 and 3 sigma, the likelihood of that is only 3%, but the journals are primarily comprised of 2-3 sigma results. This is exceedingly unlikely to also be true in a world where the journals were filled with correct results. Most of what is published in economics must necessarily be wrong.

    Economics as a field is much more sophisticated about the subject of data-mining than the other ‘soft sciences’ but given that knowledge, two sigma results are still regarded as worthy of publication and are still treated as strong demonstrations. That suggests to me that economics still has not matured. The methods and seriousness simply hasn’t reached the same level as, say, particle physics.

    So yes, economics is a science but it still feels like a cargo cult science. Everyone goes through the motions with their VAR studies, but there seems to be very little meta-cognition that the standards of proof are rather low, and the work is consequently rather fatuous.

  34. Gravatar of Greg Ransom Greg Ransom
    20. March 2011 at 01:03

    There are no model in Darwin.

    There is are problem raising design-like patterns and there is an explanatory causal mechanism to account for those patterns — just like Smith / Hayek economic science.

    You are stipulating a bogus identify, that “science” = models.

  35. Gravatar of Scott Sumner Scott Sumner
    20. March 2011 at 04:43

    Benjamin, You said:

    “Well, science. Since we cannot set up replicable experiments, I don’t think economics will ever be a science.”

    First, it makes little difference whether we are or are not a science, it’s just a question of semantics. Second, we can set up replicable experiments, so by your definition we are a science.

    The car analogy does work well for certain aspects of economics.

    rhmurphy, You said;

    “I don’t like to get bent out of shape over debating an essentialism like whether the category of study we call economics belongs in the category of knowledge we call science. But I have problems coming up with a useful definition of science that economics would fit under.”

    The first sentence was the point of my post. The second I don’t agree with, as I think it’s easy to come up with plausible definitions of science that include economics: I.e., building models to try to explain the world around us. That’s science and that’s economics.

    You said:

    “Your definition of science – model building and cause and effect is a common positivist position in the economics profession, but positivism has been dead in philosophy of science for 40 years.”

    No, I am certainly not making a positivist argument, and I don’t think you have correctly characterized positivism. I am a pragmatist, and pragmatism is not dead within philosophy. I favor using all sorts of methods, not just those approved of by the positivists.

    Mark, I’ll defer to your expertise, as I don’t keep up with recent trends in the profession. When I read the major journals, I certainly don’t get the impression that RBC is a particularly popular approach to macro. Most of the articles seem new Keynesian.

    David Pearson, You said;

    “Sometimes economics is less useful. For instance, will relative price distortions cancel out the positive effects of higher NGDP expectations? For instance, if a certain group (say, middle class consumers) sees its real wages fall disproportionately, and their real wealth (housing equity)also fall, will they quickly pull back on spending, thus offsetting a potential rise in aggregate income? I’ve not seen a piece of economics research that attempts to answer that question.”

    There is no doubt in my mind that monetary stimulus raises both NGDP and RGDP, at least when the economy is at less than full employment. Most of the relative price effects encourage more production, for instance real wages tend to fall.

    Mark, He’s back.

    OGT, Yes, there is lots of bad economics, but of course that’s unrelated to the question of whether economics is a science. The term ‘science’ is not a goal that should be aspired to, but a description of a field, carrying no normative implications. There’s also lots of bad music, but that doesn’t mean music is not an art. There is lots of bad accounting, but that doesn’t mean accounting is not a craft.

    W. Peden, Great comments, indeed your comment is far better than my post. I’d recommend readers look at it instead of my post. I especially agree about utility being the rotten core of economics, no one seems to have a clue as to what it means. But I agree with all of your points.

    Dave and Mark, I think economics is both descriptive and predictive. But the best predictions are usually conditional. Still, here’s an unconditional prediction:

    “Real GDP in China will grow faster than in the US over the next 20 years.”

    Not very exciting, is it? That’s why I say that generally only the conditional predictions are interesting:

    “Legalizing drugs would reduce the murder rate in the US.”

    That’s the sort of conditional prediction that the average person might find more interesting.

    Kevin, I agree, and would add that economists do use controlled experiments, albeit not often. But astronomers also don’t do them very often.

    Mike, You said;

    “Science is just a method and nothing more. It can be applied to anything, including law, music, and accounting.”

    Sure, but I was considering the bulk of the activity that occurs in each field.

    On the other hand I have seen even broader definitions of science than the one I use. I seem to recall that in early history the term ‘science’ referred to any field of study. I was relying more on modern usage.

    Jon, I completely agree. Indeed in this blog I have frequently ridiculed what you call 2 sigma econometrics. I think it’s mostly worthless.

    Greg, No, I never said science equals models, I said scientists use models, as did Darwin.

  36. Gravatar of flow5 flow5
    20. March 2011 at 05:07

    Economics is as much a science as the math & accounting that explains it. Take Wikipedia’s definitions of (1) accountancy & (2) mathematics:

    (1) Accountancy is the process of communicating financial information about a business entity to users such as shareholders and managers.[1] The communication is generally in the form of financial statements that show in money terms the economic resources under the control of management; the art lies in selecting the information that is relevant to the user and is reliable.[2] Accountancy is a branch of mathematical science that is useful in discovering the causes of success and failure in business.

    (2) Mathematics is the study of quantity, structure, space, and change. Mathematicians seek out patterns,[2][3] formulate new conjectures, and establish truth by rigorous deduction from appropriately chosen axioms and definitions.[4]

    Beyond that let me say that Tyler Cowen does not understand economics, but most professional economists don’t either. In the area of economics, no one wants to end up like the seer Joeseph Granville, when after he acquired so many followers that his market calls became well anticipated, and subscribers would front-run his forcasts (maket turns) (e.g., Sept. 1981).

    Thus the deep, dark, secrets in economics are explained away.

  37. Gravatar of W. Peden W. Peden
    20. March 2011 at 05:35

    Dave,

    But now we seem to be moving back to “complete” controls, which makes the supposition that (in all sciences) we know for what we need to control. I think we generally do know, but we can only control on the basis of our existing beliefs and obviously these need not be correct. I think that, in economic inquiry, we can at least reach the standard of control that exists in most sciences: we can control for what we believe at the time to be important.

    The difficulty of replication is a real one in economics; however, it is by no means unique to economics. In fact, it’s a ubiquitous feature in sciences that engage in complex experimentation-

    http://www.newyorker.com/reporting/2010/12/13/101213fa_fact_lehrer

    Jon,

    I agree that the standards of correlation in economics are low, relative to other sciences. However, I don’t agree that economics should aspire to finding deterministic universal laws of society, because I see no reason to assume that such laws exist in human behaviour and I can think of a number of reasons to think that they don’t. I expect little more from economics than the postulating and testing of systems and regualarities in the transactive/productive aspect of society. If it can do that, it’s already the Queen of the Social Sciences.

    My greatest fear is that, in the pursuit of such physics-like laws, economists will end up sacrificing either truth or relevance. Any time “law” is used in economics, it seems to be either affixed to a relatively weak regularity (like Okun’s Law) or a proposition that is either tautological* or wrong (like the law that people seek to maximise utility).

    * Tautologies aren’t always to be sniffed at, provided one doesn’t confuse them with testable claims. As Friedman pointed out, M*V = P*Q is a tautology, but it is still very useful because it clarifies the relations between a number of our ideas and can stimulate empirical inquiry, just like E = M*C2.

    Of course, since V doesn’t had an independent measurement, the quantity theory equation is not testable in the way that E = M*C2 is testable, but one can still make it testable e.g. making the fairly reasonable assumption that there is a demand function for M that determines V and then testing versions of this demand function for consistency with NGDP. That is an example of a tautologous equation being used to create testable hypotheses; so there’s not a big problem provided that we don’t confuse economic equations and hypotheses with their counterparts in physics. Insofar as the “cargo cult” criticism is sound, it is in the dangerous tendency (sometimes unfairly called the Samuelsonian tendency) to confuse tautologies with universal regularities.

  38. Gravatar of David Pearson David Pearson
    20. March 2011 at 07:22

    Scott,

    So policy works by taking down real wages so aggregate income can rise (through employment). This assumes no disproportionate decline in spending due to future real wage expectations. In other words, it assumes most Walmart shoppers will think, “My incredible shrinking real paycheck is only temporary; I trust the Fed when I set future income expectations!” I would argue this was true in 2003-2007 when housing — the main store of wealth for the middle class — was on a tear. Not so much now…

  39. Gravatar of Ted Ted
    20. March 2011 at 09:04

    Of course economics is a science. What people are actually debating is whether economics is successful as a science. While I could get into a complicated philosophy of science, I think a pretty good definition of science is that it is the study of any natural phenomena (physical or social) where hypotheses are made that can be falsified by empirical observation. I think it’s fairly clear that physics is a more successful science than economics, in the sense that the models are able to match observation better (though obviously there is still a lot to do), but I think it’s pretty clear why this is so. Physicists can run a lot of controlled experiments, with humans the best you have is very broad observational studies and quasi-natural experiments. Let’s take economic growth theory as an example. If an economic researcher had the ability to take millions of people and control everything about their population and government: political institutions; public policies; historical trajectories; average intelligence; cultural values; geography; resource endowment etc. I’m fairly confident economists would have wonderfully accurate theoretical models of economic growth. The social sciences are at a massive disadvantage relative to the physical because for obvious ethical and practical reasons truly controlled experimentation is nearly impossible. Also, I’d like to make another observation. Ever notice that the physical sciences where controlled experimentation is possible have much more successful theories than those areas of science where observation is only possible (see quantum mechanics vs. cosmology)? I think this is because controlled experimentation is the key to both weeding out bad ideas and for spurring new ideas.

    On a similar note, earthquake prediction now has a nice analogy to economic prediction. Part of the reason earthquake science is hard is because we cannot run controlled experiments – just like economics. Furthermore, our current technology does not permit us to accurately collect a lot of data – just like economics. Developing rigorous theory is difficult without ample, accurate data. For any given physical phenomena it’s possible to develop a lot of explanations (with varying degrees of plausibility) only good data can weed out the winners and losers, and with earthquakes good data is hard to come by. On the issue of weather, we’ve known the problem with weather prediction for a long time. Lorenz showed these were chaotic systems and so really small measurement errors in our models can lead to really inaccurate predictions. So, this isn’t really a flaw of weather science. Weather science is actually very successful in terms of explaining observed phenomenon, prediction on specific events is harmed by insufficient measurement.

    Also, the ability to predict is not the hallmark of science anymore, at least in the way most people think about prediction. Let’s take the most basic experiment in quantum theory – the famous double-slit experiment. Quantum mechanical models can only predict the distribution of emissions, and it will do so with amazing accuracy. It can not predict the behavior of any particular emissions with much accuracy. In fact, this is been shown to be a fundamental property of nature – not a model flaw.

    And this is just random, but a lot of times people will say the social sciences are different because humans are unpredictable and not governed by physical laws. But this is just wrong. Obviously humans are governed by physical law – how else are we here? What people really mean is that our understanding for what makes humans tick sucks. This observation actually means that social scientists need to develop some sort of quantum theory of cognition and consciousness and then use this to make predictions about individual behavior and aggregate. Obviously this is very challenging, and no one is even slightly close (it’s one of those problems that’s so mind-boggling I can’t imagine even seeing a toy model in my lifetime – but who knows). However, the key difference between social and physical sciences isn’t that “humans are different,” it’s that natural experiments with humans are basically impossible which makes theory difficult to develop.

  40. Gravatar of W. Peden W. Peden
    20. March 2011 at 09:49

    Ted,

    “This observation actually means that social scientists need to develop some sort of quantum theory of cognition and consciousness and then use this to make predictions about individual behavior and aggregate.”

    Social scientists are neither qualified to do this, nor is it clear why such a “ground-up” approach is worth pursuing, unless one makes the supposition that there are deterministic mental laws to be discovered.

    The grand programme of reducing social sciences to psychology, psychology to biology, biology to chemistry and chemistry to physics (in some cases plus some extra step like reducing physics to sentences describing private experience) is a very old one that goes back at least to Auguste Comte in the 19th century. It has also been an entirely unsuccessful one, suggesting either that it’s very difficult or (more likely in my opinion) that the assumption that there HAS to be some fundamental level to which all science reduces is wrong.

    The idea of government by laws of nature is a very dodgy one and it’s plain bizarre if one doesn’t believe that these laws of nature are ordained by God. However one thinks these laws of nature come about, though, there is a good reason for thinking that talking about “government by laws of nature” is the wrong way to go-

    Let’s start with the plausible idea that for a proposition (e.g. “Nothing can travel faster than the speed of light”) to be true is for it to correspond with the universe that we live in. So, if I say that “Nothing can travel faster than the speed of light” and the universe we live in is one in which nothing travels faster than the speed of light, then I have said something that is true.

    We can sum this relation up like this: “The proposition ‘P’ is true if the universe is such that P.” In other words, propositions are made true in virtue of the way things are.

    The idea that the world is governed by laws of nature seems to be based on a perversion of this relation: instead of propositions (the laws of nature) being made true by the facts of how nature is, the truth of these facts is made true by their correspondence with propositions. In other words, the relation has turned into: “The fact P is true if it accords with the proposition P”.

    A trivial distinction? Not at all. With our common-sense understanding of the way truth works (the first case) the laws of nature are nothing more than descriptions of a particular kind of regularity. In the laws of nature-as-governors conception, these laws of nature are taken to have a certain kind of power, such that the world cannot fail to correspond to them.

    I side with common-sense in this case and regard the idea of laws of nature as governors of some sort of natural necessity to be a hang-over from an age where the Word of God assumed a big role in the way we understood the world. There is no natural necessity; there are only regularities of differing degrees.

    So one doesn’t have to appeal to qualitative differences between human beings and the rest of the universe to get to the idea that the social sciences differ fundamentally from the natural sciences. One only has to think that the regularities involved in social science are much weaker than, say, the laws of physics which are universal regularities (regularities that hold in every case and in all times) which is to say that there is a quantitative difference between the results that can be obtained in social science and the results that can be obtained in some natural sciences.

    Of course, as almost always in philosophy, the regularity view has its own negative consequences: counterfactuals become difficult; the distinction between doom and misfortune breaks down; the free will vs. determinism debate becomes inexpressible, which you might think is a good thing but it’s a terrible thing for the thousands of philosophers who get paid to write about it; and more besides.

    In sum, though, I think that this vestigial idea (that the world is the way it is because of the laws of nature rather than vice versa) is the main reason why people have generally either overestimated the prospects of social science (“We MUST be able to find deterministic laws of human society!”) or underestimated these prospects (“Social sciences are completely different from natural sciences!”).

  41. Gravatar of Dave Dave
    20. March 2011 at 10:51

    W. Peden,

    You make fair points, and I don’t think I would disagree with most of what you wrote. The only point I would make is that I don’t think the controls which can be implemented in many of the “interesting” economic questions that Scott refers to are rigorous enough to be described as complete.

    With that said, you correctly point out that controlling for variables in many of what are considered the “hard” sciences is also difficult/impossible. Which takes us back to my original statement that my own definition of science is probably narrower than most. (Though not, it appears, than my kindred sceptic, Mark).

    Our only disagreement (I think) is the epistemological one of at what level a control could be considered “complete”. Perhaps I have deep seated trust issues stemming from my childhood… Generally I find myself looking for basically 100% certainty before I will refer to something as “scientific”. Though once again I will say that something being “not scientific” is very different from something being “not useful”.

    Thanks again for the interesting conversation.

  42. Gravatar of Jason Jason
    20. March 2011 at 11:21

    Some random interjections:

    1. Actually, the tsunami prediction was pretty good, down to the little wisp that broke off and nailed Crescent City, CA, sinking a bunch of boats in the harbor.

    http://www.winlink.org/images/2011JapanTsunamiForecast.jpg

    2. The prediction of pretty much everything people here say can’t be predicted by “physicists” have all gotten much better at being predicted in my lifetime. Earthquake aftershock distributions, tornadoes, volcanoes, weather … especially in geographically constrained areas … this is almost always correct:

    http://www.atmos.washington.edu/~ens/uwme.cgi

    The chief impasse behind better predictions in “physics” is the collection of data (and basically the funding of collecting that data). Accurate enough initial conditions and enough computation time and you could predict anything — and no one doubts that, they doubt whether it is possible to collect all the necessary data (or worth it) or possible to compute the prediction (or worth it).

    The laws are known to more detail than is needed to simulate physical processes of everyday life, so it is a computational problem. But it may be intractable:

    http://www.cs.virginia.edu/~robins/Breaking_Intractability.pdf

    3. String theory is a framework, like quantum field theory. It is used to build models that make predictions. String theory does some pretty neat stuff and makes some pretty good predictions, e.g.:

    http://arxiv.org/abs/hep-ph/0501128

    A new-ish (mid-2000s) direction in the study of QCD is using AdS/CFT to do the non-perturbative low energy QCD calculations in a place where a string theory is perturbative.

    It is fundamental string theory that has a limited number of accurate testable predictions. However, working from what is known about string theory (using it more as philosophical motivation), you can get neat results like why we are here:

    http://arxiv.org/abs/1001.1155

    4. I concur with the “two-sigma” vs “five-sigma” discussion, but I have never seen an macroeconomics paper that reaches that “five-sigma” standard. It is possible that some exist. In their absence, it seems you should consider economics to be groping in the dark still. I think that is what separates “soft” sciences from “hard” sciences. Physics was a very soft science until Galileo (or around there).

    5. A second place where I see economics as being un-physics-like is that orders of magnitude tend to be ignored. This describes an effect on the order of 0.01% of the economy:

    http://gregmankiw.blogspot.com/2010/08/challenge-to-extreme-keynesians.html

    Yet the stimulus effect wasn’t predicted by anyone to within 10% or so (no one even dared!). It is like worrying about the Lamb shift before anyone wrote down the Schroedinger equation.

    6. A third place I see economics as being un-physics-like is in a seeming rejection of what we call in high energy physics “effective field theory”:

    http://en.wikipedia.org/wiki/Effective_field_theory

    It is related to the Lamb shift comment above, but is more general. The idea is that, in a way, you build models with all the effects anyone says could possibly be there. Your favorite effect. That other person’s favorite effect you think is ridiculous. All of them. Terms in your model that get coefficients of order 1 are natural. Things with big coefficients or small coefficients are unnatural and need an explanation with a strong mechanism.

    Example: there is a possible symmetry-violating term allowed in the model of the strong force. We don’t know why the term is small — so a particle, the axion, was invented to make it small.

    In economics, I see this problem arising in the Keynesian/monetarist/etc divides. You should include all of the measures — especially since even anti-stimulus theory proponents say Keynesian stimulus can have positive effects at the local levels. It may work out that the Keynesian multiplier/coefficient of government spending is unnatural (very small relative to 1) and the reason is that monetary policy offsets it. I don’t think current models are good enough to say it is exactly zero since they seem to be in the 10% area. Arguing against doing one or the other (or both) seems unscientific at this point.

    7. I always like that Max Planck quote.

  43. Gravatar of Morgan Warstler Morgan Warstler
    20. March 2011 at 14:20

    1. I 110% believe economics is a science.

    2. What people like DeKrugman do is not economics.

    3. For economic policy to be more scientific, we would have to behave in far more, “try it your way” tests.

    I’d LOVE for the progressives / Obama to just turn to Ben, and say, “it’s all on you”

    “we agree to stop arguing for any kind of stimulus, don’t fail!”

    ——

    Then we’d have science.

    Here’s another rule: anyone who says it is a science SHOULD AND OUGHT support state rights.

    What could be better for science than lots of labs?

  44. Gravatar of mbk mbk
    20. March 2011 at 18:14

    “Here’s another rule: anyone who says it is a science SHOULD AND OUGHT support state rights.

    What could be better for science than lots of labs?”

    It exists already, it’s called the European Union. But people find squabbles with that too.

    On science: much has been said already and I especially like W Peden’s posts here. I only want to add some footnotes to the whole stream of arguments.

    As a natural scientist myself, I’d venture that most working scientists go with Popper’s model in that it’s science if it has hypotheses that are falsifiable at least in-principle. Whether these falsification experiments are actually carried out is another matter. Beyond that most working scientists don’t usually spend much time on epistemology and also allow for plenty of leeway for borderline cases (burden of falsifiability is higher for well established theories, cosmology is not hounded for falsification experiments…). E.T. Jaynes in his “Logic of Science” has found a nice way to use Bayesian thinking to express falsification as a matter of probability of belief, rather than a black and white affair.

    In order to be falsifiable, hypotheses have to make predictions. The important distinction people do not make well enough in economics is not one of “descriptive vs. predictive”, but one of “descriptive (and predictive!) vs. prescriptive”. And I believe much confusion in economics comes from pervasive injections of the prescriptive, and of the political aspect into the scientific aspect.

    So economics must of course make predictions both to be useful and to be a science. Beyond this, I have made the point before that complex sciences are all suffering from a prediction deficit because of emergent phenomena (which is why macro is more than aggregate micro). Sometimes physics manages to get around that hurdle (statistical mechanics), sometimes it does not (chaotic phenomena). Nancy Cartwright’s “How the laws of physics lie” builds a nice argument around this issue in physics (prediction is easy only in oversimplified systems, even in physics). In the social sciences we not only have complex adaptive systems with poor “regularities” sensu W Peden, and emergent phenomena, but we also have agency / volition. Here, the regularities become probability distributions rather than tight laws with single solutions.

    Still, economics is a science to me, and I find it a very sophisticated one. It just happens to suffer from a complex field of study and chronic injections of political belief and personal goals. Unfortunately the prescriptive is hard to extricate from economics because it is built into it from the start under the disguise of “utility”.

  45. Gravatar of Greg Ransom Greg Ransom
    20. March 2011 at 21:52

    Show me the “model”.

    ” I never said science equals models, I said scientists use models, as did Darwin.”

    And let’s not kid ourselves, when economists say model, they are talking about a math construct.

    Have you read Darwin.

    He identifies connected problem raising patterns and he proposes a causal mechansim that can account for those patterns. He doesn’t construct a model. Neither does Hayek. They both identify problems and locate an causally open ended mechanism to account for those patterns.

    No models of the sort economists are talking about when they speak of models.

  46. Gravatar of Greg Ransom Greg Ransom
    20. March 2011 at 21:55

    mbk, show me the predictions in _The Origin of Species_.

    Hayek forced Popper to concede tnat Darwin is science, and Popper’s falsifiability criterion was falsified by the examples of Darwinian biology and economics andother essentially complex sciences.

  47. Gravatar of mbk mbk
    21. March 2011 at 04:51

    Greg, I don’t follow your question. I didn’t mention Darwin and I don’t understand in which way “Popper’s falsifiability criterion was falsified”.

  48. Gravatar of flow5 flow5
    21. March 2011 at 05:10

    “physics is a more successful science than economics, in the sense that the models are able to match observation better”

    Those are the words that come out of the mouth of the ignorant. Politicians & capitalists are 2 constituents that economists have to deal with. But math models on the other hand, are truthful, accurate, & realistic.

    Contrary to economic theory, & Nobel laureate, Dr. Milton Friedman, monetary lags are not “long and variable”. The lags for monetary flows (MVt), i.e. the proxies for (1) real-growth, and for (2) inflation indices, are historically, always, fixed in length:

    2010 jan ,,,,,,, 0.13 ,,,,,,, 0.538
    feb ,,,,,,, 0.056 ,,,,,,, 0.507
    mar ,,,,,,, 0.072 ,,,,,,, 0.558
    apr ,,,,,,, 0.056 ,,,,,,, 0.552
    may ,,,,,,, 0.067 ,,,,,,, 0.477
    jun ,,,,,,, 0.038 ,,,,,,, 0.474
    jul ,,,,,,, 0.078 ,,,,,,, 0.499
    aug ,,,,,,, 0.031 ,,,,,,, 0.49
    sep ,,,,,,, 0.045 ,,,,,,, 0.542
    oct ,,,,,,, -0.01 ,,,,,,, 0.386
    nov ,,,,,,, 0.041 ,,,,,,, 0.321
    dec ,,,,,,, 0.099 ,,,,,,, 0.32
    2011 jan ,,,,,,, 0.089 ,,,,,,, 0.155
    feb ,,,,,,, 0.089 ,,,,,,, 0.23
    mar ,,,,,,, 0.139 ,,,,,,, 0.332
    apr ,,,,,,, 0.109 ,,,,,,, 0.242
    may ,,,,,,, 0.12 ,,,,,,, 0.213

    First column is real-output. Second column is inflation. Inflation bottomed in JAN (Core cpi). The roc in real-output is rising – & at an increasing rate.

    This metric is not meant to theoretically work, but does given certain conditions.

  49. Gravatar of Scott Sumner Scott Sumner
    21. March 2011 at 05:44

    flow5, The comment about Cowen was pretty silly.

    David, you are confusing hourly real incomes (which fall) with total real incomes (which rise.)

    Ted, No, economics can do controlled experiments.

    I can’t understand how someone can say physics is more successful than economics. It’s like saying math is more successful than poetry. You are comparing apples and oranges. Successful at what? Explaining the difference in living standards between North and South Korea? How would a physicist explain that? Yes, physics is important for some areas of life, and economics for others.

    Your mistake is to look at the sort of problems that physicists address, in which case of course they’ll be more successful.

    Jason. I meant predict tsunamis a week ahead.

    I meant scientists can’t agree on whether string theory is science, I wasn’t addressing the usefulness of it.

    You said;

    “4. I concur with the “two-sigma” vs “five-sigma” discussion, but I have never seen an macroeconomics paper that reaches that “five-sigma” standard. It is possible that some exist. In their absence, it seems you should consider economics to be groping in the dark still. I think that is what separates “soft” sciences from “hard” sciences. Physics was a very soft science until Galileo (or around there).”

    Then you need to read some of my papers. Indeed 5 sigma results are very common in economics (if by 5 sigma you mean 4 standard deviations.)

    Many of your points are quite good.

    Morgan, I agree with your last comment about states rights.

    Greg, Models don’t have to be mathematical.

    flow5, Lags can be very short–check out Brazil in the 1980s.

  50. Gravatar of David Pearson David Pearson
    21. March 2011 at 06:16

    Scott,

    I mentioned aggregate incomes. They will have a hard time rising if the consumer pulls back on spending as a result of a steep drop in real wages (for the middle class) accompanied by renewed drops in the price of his main asset (btw, the housing data today was awful).

  51. Gravatar of Jason Jason
    21. March 2011 at 12:10

    Scott, I’d love read some of your papers. Unfortunately I only have access to “hard science” papers through work — just kidding. I think I can do a proxy thing through my alumni library access. Which paper?

    I realized the Reinhart and Rogoff book (This Time Is Different) has some 1% significance effects which is a little above two sigma. There is another half of the five sigma criterion which is the evidence that your errors are from a known distribution … which I don’t think Reinhart and Rogoff meet (but you have to start somewhere).

    I was also just trying to emphasize that it is a very particular sense that people don’t consider (fundamental) string theory a science. The experiments can be imagined, but may be impractical since 10^19 GeV is much much bigger than 7000 GeV. So string theory may be limited to astrophysical observations like the cosmic microwave background (http://arxiv.org/abs/0704.0647), but there are some people who don’t think astronomy is a science since it can only observe. And there aren’t many (any?) astrophysical predictions that can only be arrived at via string theory. String theory is math (and useful math; it was used to deal with some singularities in the Poincare conjecture). Some people don’t think math is science.

  52. Gravatar of Bababooey Bababooey
    21. March 2011 at 12:48

    Isn’t your blog (and first 6 entries in the FAQ) a memorial that testifies to hownot useful economics has been? Have economists been a net contributor or despoiler of economic progress? Stated differently, how much more influential was/is Keynes and the 1920s Harvard’s economics department than Hayek? (Or, if asked of Paul Krugman (or his wife): “how much more influential is Uncle Milt and the University of Chicago than Keynes and Dani Rodrick?”).

    Might as well call lawyers “useful”….

  53. Gravatar of ssumner ssumner
    22. March 2011 at 07:27

    David, I meant that you seemed to assume that lower hourly wages meant lower aggregate incomes. But just the opposite is true. Monetary stimulus reduced real hourly wages, but raises real aggregate income.

    Jason. The problem in economics is not in finding highly signficant results, especially in macro time series, it’s avoiding spurious correlations due to some missing third variable.

    I don’t know about math, but it seems rather incredible to argue that astronomy is not a science. By that definition only physics (and perhaps chemistry) would be sciences. That’s a pretty arbitrary defintion. (Although I suppose in a sense all sciences are applied physics.

    Bababooey, It’s one of those glass half empty/half full things. Some days economics seems very productive, other days I am more pessimistic.

  54. Gravatar of Greg Ransom Greg Ransom
    28. September 2011 at 10:09

    Scott, show me Darwin’s model .. you can’t. He didn’t construct anything like what economist construct when they are talking about a “model”. If you simply want to play games with the word “model” like you accuse others of playing games with the word “science”, help yourself.

    Scott wrote,

    “Greg, No, I never said science equals models, I said scientists use models, as did Darwin.”

  55. Gravatar of ssumner ssumner
    29. September 2011 at 05:32

    Greg, Origin of the Species contains Darwin’s model. It involves traits being passed from one generation to the next, and natural selection assuring that the traits that tend to survive are the most adaptive. When various groups within one species get separted geographically, and evolution takes place, eventually a new species will form. It’s a myth that “models” must be expressed in math.

Leave a Reply