In Defense of Economics
I’ve generally been quite sympathetic to Jim’s recent criticisms of the science (or “science”, depending) of economics, but it seems to me that his demand for economic principles to be put through the gauntlet of “rigorous, replicated falsification trials” ends up setting the bar a bit too high. For one thing, unlike in the “hard” sciences, it’s very often unclear what such a process could be in the case of economics: we’re usually dealing with large-scale human societies, after all, so running experiments under sufficiently controlled or randomized conditions isn’t a possibility; and by analogy, we don’t want to end up saying, for example, that the theory of evolution is junk science because its primary objects of study (whole species, huge stretches of time, the very distant past, and so on) aren’t subject to tightly controlled experiment. Yes, the range of possible experiments is limited, but in economics as in any other science we do (or at least: try to do) the best we can with what we’ve got.
Secondly, and with reference to Jim’s demand for more precisely quantified predictions, it’s worth keeping in mind Aristotle’s dictum about not demanding more precision than the subject-matter admits of: human beings are tremendously complex, human societies tremendously more so, and it is the mark of an educated man to look for precision in each class of things just so far as the nature of the subject admits. Want to know the exact amount by which a given change in the minimum wage is going to affect employment levels? Economics can’t tell you, but that’s largely because a human society isn’t a closed system: there are too many background factors, both known and unknown, that will alter the outcomes in ways we’re not in a position to quantify. These problems could be mitigated, of course, if we could run large-scale randomized control trials – but as per the above, we can’t, so we do (or at least: try to do) the best we can with what we’ve got.
Thirdly and finally, it’s certainly not as if economic theories are entirely untested or non-predictive. For example, the claim that minimum wage increases tend to have negative effects on the rates of unemployment is a claim that runs counter to at least some intuitive ways of thinking about how human society works, and while there is (to say the least!) plenty of controversy over whether the available empirical data tend to support or falsify that theory, this is arguably (once again) more a matter of the inherent imperfection of the data than the unscientific character of macroeconomic theories. And it’s not as if the presence of controversy over the collection and interpretation of data is the sole property of economics or even the human sciences in general: there’s never such a thing as an experimentum crucis, and given our inherent human limits all that we can do is to do (or at least: try to do) the best we can with what we’ve got.
Perhaps Jim has in mind some more concrete ways in which economics could be set on the solid footing of a more rigorous science; if so, those ideas are worth exploring. And I’m certainly sympathetic to Jim’s demand that pro- (and, presumably, anti-) stimulus economists toss some predictions out there, so that we can use this (hopefully!) once-in-a-lifetime event to test how well their models work, as well as his more general caution against the perils of false certainty, so perhaps once we’ve hashed things out in the comments it will turn out that the two of us don’t really disagree after all. For now, though: economists aren’t perfect, and I’ll gladly admit that we shouldn’t actually let them run the world, but that doesn’t mean we shouldn’t do (or at least: try to do) the best we can with what we’ve got.
I seem to remember in the afterward for “Guns, Germs, and Steel” Jared Diamond had a few suggestions for how the soft sciences might be firmed up a little. Naturally I can’t recall much more than nodding my head and muttering “damn straight.”
But I don’t think we’d want to firm them up too much (he said, betraying his bias) lest the soft sciences become us unpopular as physics.
— Tony Comstock · Feb 19, 07:01 PM · #
John:
Thanks for the very thoughtful post.
A few quick things:
1. A narrow point is that I included the minimum wage proposition on my list of falsifiable statements. It, along with the rent control assertion, strike me as the most practically falsifiable items on Mankiw’s list. I think that actual validation would require a more specific version of the point along with more than one random assignment trial.
2. I get your point that decisions have to be made in realms where we do not have lab-bench chemistry type certainty. I’m not railing agianst that. I’m railing against the false assertion of the level of certainty that the economics profession asserts about their beliefs, and their desire to lay the mantle of scientific knowledge over these assertions as a public rhetorical tactic. The reason that I don’t think this is simply emotional pique on my part is that I believe recognizing the uncertainty of this knowldge tends to lead to different kinds of decisions. Becasue these different decisions are based on a firmer recognition of how little we know, I believe that they will tend to be better.
3. I went into some practical things that I think can be done to make social sciences more reliable (in certain narrow sub-fileds) here: http://theamericanscene.com/2009/02/03/popper-is-my-homeboy-a-manifesto
4. I went into what decisions I think this perspective would lead to concerning the current stimulus debate here: http://theamericanscene.com/2009/02/12/what-should-we-do-about-the-stimulus
— Jim Manzi · Feb 19, 07:11 PM · #
Thanks, Jim.
Re: 1, you did indeed count the minimum wage proposition as a falsifiable one. But most of the propositions you list as non-falsifiable are basically on that list because they’re normative (they say what we “should” do); it seems to me that if we extract the descriptive/predictive core from each of them (e.g.: outsourcing work tends to benefit general economic welfare; subsidies to professional sports franchises tend not to produce equivalent economic and social benefits; and so on), the task of falsification can get (and has gotten) off the ground.
Re: 2 and 3, I think that a central part of the difference between us is that you take Popperian falsification to be a normative demand on “scientific knowledge” as such, whereas I just see it as a (highly idealized!) description of what goes on in certain scientific fields, which then leaves it open that genuine scientific knowledge could also be garnered in different ways than that. You’re right that the methods of economics leave open levels of uncertainty that aren’t there in, say, physics or biology, but that doesn’t mean it’s not science at all.
— John · Feb 19, 07:21 PM · #
“I’m railing against the false assertion of the level of certainty that the economics profession asserts about their beliefs, and their desire to lay the mantle of scientific knowledge over these assertions as a public rhetorical tactic. The reason that I don’t think this is simply emotional pique on my part is that I believe recognizing the uncertainty of this knowldge tends to lead to different kinds of decisions.”
This could not have been said better.
— Rortybomb · Feb 19, 07:22 PM · #
Great post John. I’m glad you made the comparison with evolution. Jim appears to forget that all of science isn’t as simple as, say, measuring the mass of the electron. Many questions, even in the hard sciences, do not abide by Popperian falsification. How would you possibly falsify the atomic theory of matter? Even the Michelson-Morley experiment falsify the ether concept. Until Einstein explained the null result with relativity, people still clung to the idea.
I would argue that some claims in economics are at least as robust as some in space physics (my field), an allegedly “hard science.”
— PrajK · Feb 19, 07:29 PM · #
“ Many questions, even in the hard sciences, do not abide by Popperian falsification.”
Isn’t this Jim’s point? That the economic policy by expert consensus he’s cited is attempting to lay the “certitude of science” over the uncertain science of economics. At least in your field, it comes down to interested layman like myself as “this is what we think we know so far. we got here by making this obervations/experiments.”
That’s a far cry from the “four out of five dentist agree” approach to convincing people you’re right.
— Tony Comstock · Feb 19, 07:39 PM · #
John:
Thanks.
1. Fair enough, though these are different statements than the ones Mankiw lists, and these differences are not purely semantic (using that term non-technically). I agree that “People should do X” is about the same as saying that “X is good”, but only in that it punts on the meaning of “good”. Using “general welfare” is, in my view, about the same as saying “good” (recognizing that the most borderline case on the falsifiable list punted in exactly this way, and this is the reason that I called out that I was being charitable in certian cases). That is, by example, if a statememnt on this list had been “A US law making outsorucing illegal will reduce measured US GDP by more than 10% within 5 years”, it would have been falsifiable. But it would also have been a very different statement than either “The United States should not restrict employers from outsourcing work to foreign countries” (the one actually on his list)or “Restricting employers from outsourcing work to foreign countries will be bad for the general welfare of the United States” (your approximate restatement of the proposition).
2. I am comfortable punting on the semantic (again, using the term non-technically) discussion about “is it a science?”. Further, I am not arguing for a strict or naive falsificationist approach to science (though I accept it would be easy to draw that conclusion from reading the post). I do believe that the truth of a practical statement is its method of validation.
I was trying in that post to illustrate partially two beliefs that I do hold about academic economics: (1) that its fruits do not include many useful, non-obvious predictive rules, and that (2) it does not display a compelling mehtod of validation (i.e., a demonstration of “reliability”) for whatever predictive rules it does provide that might be useful if true.
— Jim Manzi · Feb 19, 07:40 PM · #
PrajK:
Thansk for the comment. I’ve done a decent amount of astrophysics, and get the point that this is necessarily “science without experiments”. I won’t try to go into in detail, but I think that as a practical matter, the reliability of astrophysics is subsidizied by the experimental knowledge of terrestrial physics. It is easier to generalize these physical laws into astronomical realms becasue of: (1) the relative simplicity of the interactions between astronomical bodies (driven by large distances plus the inverse-square law) as compared to human interactions in society, and (2) vast unobserved spaces that let us test theories against previously unibserved phenomena (e.g., Eddington’s confirmation of GR).
— Jim Manzi · Feb 19, 07:48 PM · #
That’s true, but then the problem is that unless it included a big, fat ceteris paribus clause it would almost certainly have been false, too, due as I said to the interference of unaccountable background conditions. Those statements do have predictive equivalents, but they simply can’t be precise in the way you’re demanding – again, though, this sort of thing holds true of perhaps every science other than fundamental physics.
— John · Feb 19, 08:13 PM · #
John:
Fair enough (and I think the ceteris paribus clause is what I was try to get at in the paragraph of my post where I say that they would really need a big econometric model to account for this, in order to have useful predictive rules that could be generalized from a particular instance).
I don’t think that useful, relaible, non-obvious predictive rules are only created by physics. I can think of mnay in biology and chemistry, for example. I’m not really trying to prove that economics isn’t physics, or isn’t a science in some abstract way. I am however responding to economists assertions that we should do X or Y with a simple demand: prove it.
The problem I have with their response (obviously, speaking in cartoons here) is that it collapses when faced with serious scrutiny. This doesn’t mean I think they’re dumb or charlatans, but it does mean that when faced with a counter-intutive prediction I am not rationally prepared to defer to them the way I am when a all physics professors tell me that unequally weighted bodies fall at the same rate in a vacuum. It also means that I am very inclined to hedge my bets when dealing with economics in a way that I am not when it comes to, say, designing aircraft.
— Jim Manzi · Feb 19, 08:40 PM · #
“ the relative simplicity of the interactions between astronomical bodies (driven by large distances plus the inverse-square law) as compared to human interactions in society”
Cutting back, isn’t the political process the mechanism we have to deal with decision making when even the best good faith advice is contradictory and, given the complexity of human interactions, unprovable?
I’m inclined to find the “four out of five economists agree” approach suspect, because it smacks of marketing, but in the end I can’t think of a better way to resolve these sorts of questions than to have various parties advocate for their respective positions, even if the results sound more a like a Dentine commercial than what my 5th grade civics course taught be to regard as political discourse.
I still think we wouldn’t hurt by having a better educated electorate, especially in math and science. I regard these disciplines as less subject to the whims of fashion, political and otherwise, and thus more effective at developing critical thinking skills; especially at (what ought to be) a lay person’s level of expertise.
— Tony Comstock · Feb 19, 09:06 PM · #
Better papers are out there, but this one, accessible for free, goes a good deal of the way toward showing why economics has a problem with nailing down cause and effect.
Humility and respect are the cardinal virtues here. The human economy just might the most complex organism in the universe.
— JA · Feb 19, 09:34 PM · #
Economics is the study of competition, and competitive situations are inherently unpredictable because the competitors are constantly trying to figure out new ways to win. People don’t keep on making the same old mistakes — they make new mistakes.
When competitive situations become easily predictable — e.g., Will the average price of homes sold next year be higher in Beverly Hills or in Compton? — we lose interest in the question because the answer is boringly obvious. The social sciences in general can tell us much about the future with great accuracy (e.g., school test scores will be higher in Beverly Hills than in Compton), but we tend to find these answers boring and, indeed, depressing, so much policy (e.g., No Child Left Behind) is based on ignoring everything we honestly know.
— Steve Sailer · Feb 19, 09:48 PM · #
One possible response to the “realities” of the subject that you describe is to set the bar lower. This allows us to continue making the more exciting claims to which one is accustomed. However, there’s an alternative, namely becoming much more conservative and incremental in the entire enterprise. You see something like this in analytic philosophy. Philosophy in the English-speaking world is a much different affair than it once was, mostly because certain philosophers early in this century wrestled with this dilemma and came the the latter conclusion. Thus analytic philosophy is one with relatively small ambitions, advancing only as far as its stewards can “verifiably” take it (in this case through careful logical deduction rather than experimentation). In economics, a similar approach would seem possible. Give up the notion that we’re going to solve the big questions anytime soon, and trust instead that the sum of many years (even generations) of incremental but verified claims will eventually add up to a more robust picture.
As another commenter notes, this approach is pretty boring. But it’s the more “scientific” one.
— Joel Martin · Feb 21, 09:26 PM · #