Economics' Next Top Model
Economists: do we know things? How do we know? Let's find out!
Economics is, as you might be aware by now, one of my passions. It has many things worth discussing within: macro, labor, urban, growth, development, education, etc. But let’s take a step back. Is economics even a science? And how are social and natural sciences different from each other? How do economists claim to know things in the first place? And why do economics papers make such weird assumptions anyways? All of these questions fall under the purview of epistemology, the philosophy of science (and/or knowledge, it’s a whole thing).
Butchering epistemology 101
Rene Descartes’ “Meditations on First Philosophy” is considered to be the first work of modern philosophy. Rejecting the medieval tradition of appealing to the authority of the Bible (and/or Aristotle and/or Augustine and/or Aquinas), Descartes intended to question everything, down to whether reality itself existed, to find something he could be certain of and then build from his way up. He says that everything can be questioned: the senses cannot be trusted, because they can sometimes deceive (is the dress blue or white, anyone?). He also says that people can sometimes be unnable to tell apart dreams and reality, so he doesn’t know if reality is real (a very good short story on this is “The Night Upside Down” or, in the original Spanish, “La Noche Boca Arriba” by Julio Cortazar). And reality itself might be questioned too - wouldn’t it be possible for an evil demon to fool you into thinking reality existed, when it did not? The only thing you cannot doubt is that you yourself exist, since if you are questioning everything that means you cannot be false too. After (very badly) proving the existence of God, Descartes adds that the senses, etc must not be deceiving him, because a good God that exists wouldn’t allow for any of that. Not very convincing stuff. But Descartes’ whole intent was: he proved everything existing using reason and reason alone. The senses couldn’t be trusted, but logic could - thus rationalism was born. The rationalists’ whole deal is that they believed pure reason and logic could be used to derive absolutely certainly true knowledge.
David Hume and the empiricists had some thoughts. Hume is also relevant to economic thought because he also wrote extensively about economics and was friends with Adam Smith, but that’s neither here nor there. Hume wrote that pure logical knowledge was, of course, true, but it also wasn’t very useful. “The square of the hypothenuse equals the sum of the squares of the other two sides” is, of course, always true, but it’s not useful in the same way as “metals expand when heated” is useful. Empiricism is the belief that knowledge must be derived from experience to be considered valuable. But the problem with statements like “metals expand when heated” is that they cannot be proven with accumulations of experience - finding every single metal and heating it to every single possible temperature under every single possible circumstance is not especially easy. So claims through induction, which is the type of reasoning that uses particular examples to justify general statements, are simply not logically sound. Similarly, causation cannot be proven to occur empirically: it’s just an inductive reasoning. A pool player, even after a million games, cannot prove that it is the ball hitting the other ball that causes it to move instead of magical fairies, or the wind, or sudden localized earthquakes, or jumping cows.
So far, we could be wrapping up the post with a discussion on theoretical and empirical models, and the folly of insisting that only one kind can provide valid knowledge. But instead, another player comes to the round: Karl Popper. Popper is famous as a political philosopher as well, but his contributions to epistemology are both deceptively simple and incredibily insightful. Rationalism and empiricism both fail to grasp, according to Popper, that true knowledge, at least for science, is impossible. It will never be possible to prove anything by induction. And pure logic cannot say that much about the real world. But while we can’t prove anything is true, do not despair: we can prove something is false. “Metals expand when heated” cannot be proven true by heating metals and seeing them expand, but can be proven false by the converse: a metal not heating. If metals expand, then the hypothesis is simply considered not proven false until further evidence can be discovered. If they don’t, it has to either be abandoned or reworked in a way that replaces key assumptions. Therefore, the line separating science and opinion is making empirically testable statements that can be proven wrong. Assumptions work to generate the groundwork for empirical testing. Something relevant is that testing a hypothesis doesn’t just mean going out and measuring things, or doing a controlled lab experiment - using theory to disprove that the conclusion follows the premises works just fine. Popper claims this has a variety of implications for political philosophy, but so far so good enough.
“A dismal science” or just dismal science?
An economist who was a big fan of Karl Popper was Milton Friedman, down to a (rumored) meeting between the two in the University of Chicago, where Friedman was said to have quizzed Popper on a hallway and was immediately taken by his contributions to epistemology. His 1953 “The Methodology of Positive Economics” is very clearly influenced by Popper, since it obviously begins by stating that true knowledge is impossible and knowledge can only be falsified, not proven. We’ll return to the implications for how economics deals with knowledge in a bit, but first something more important: is the dismal science even a science?1
According to Friedman, economics can be divided in two: normative, the kind that makes statements of intent (“poverty should be abolished”, etc) and positive, meaning the one that makes actual predictions (“more welfare would abolish poverty”). While politicians love to play up disagreements as normative (“the enemies of freedom” vs “the enemies of the poor”), actual economic debate mostly focuses on positive aspects - whether or not something will actually fulfill its objectives, whether it has any unwanted drawbacks, whether these drawbacks are too bad for the policy to be worthwhile, etc. The road to hell is paved with good intentions precisely because bad economic policies are designed by well-meaning policymakers, not because they aren’t. Positive economics deals with the “is”, normative economics with the “ought” - and positive economics can be as scientific as physics, medicine, or astronomy, since it also follows Popper’s rule of making testable statements.
Now, the positive vs normative distinction plays a much bigger role in economics than in the “hard” sciences, since mathematicians don’t disagree in principle about whether integers can be written as the sum of four cubes. Normative disagreements do play a role, especially in policy, but it’s a much smaller role than perceived. For instance, two doctors can disagree about abortion, and it’s a normative disagreement: one thinks a fetus is a person, ergo abortion is murder, and another thinks it’s not, ergo abortion is just a medical procedure. Whether a fetus is or is not actually a person can’t be “proven” to the other side by talking about fetal sentience in the first trimester, or viability, or how mothers are happier than non-mothers - it’s a matter of profound philosophical disagreement. While a lot of economists do disagree on the normative side, the most common disagreements (on stuff like free trade, the minimum wage, etc) is normally positive: does this accomplish its goals, and are its second-order effects worse than the first-order effects - without questioning the underlying motivations. Economists who think raising the minimum wage is bad don’t hate workers, they use a different model of the labor market (that can be correct or not) that yields very negative results from raising the minimum wage - and economists who disagree use other, different models with different conclusions. If the economists are approaching this issue with the “right” mindset, whether their models are correct can be verified through empirical evidence or rigorous theoretical probing, and they should eventually either change the way they see the issue or change their model altogether - otherwise, they’re just ideologues.
Friedman’s 1976 Nobel lecture “Inflation and Unemployment” elaborates on the “natural vs social sciences” debate. While many claim that social sciences, such as economics, cannot aspire to the same level of rigor as the natural sciences due to fundamental differences in subject matter, Friedman disagrees. Both have a fundamental incapability of providing true knowledge, but both can root out false claims through experimentation, logical reasoning, inquiry, whatever - both have tentative knowledge built on not-yet disproven claims, both have limits to experimentation (astronomers cannot conduct any experiments whatsoever, after all, and rely on observation), and both have experimental results biased by the presence of an observer (the Heisenberg Unceirtanty Principle’s implications wouldn’t be too far out in a microeconomics class). While both have different sources of data and ways of testing hypotheses, the differences between economics and physics might be as large as the differences between physics and biology.
While you normally hear that “economics isn’t a real science” from weird STEM types that don’t like it out of some weird superiority complex, it also very frequently comes from left wing people who don’t like the (positive) conclusions reached by economists. Their criticisms are frequently “economic models are so unrealistic, they assume perfect rationality and reduce important decisions to equations, how can we trust their conclusions?”. This line of criticism is so old it comes from John Stuart Mill opposing the original neoclassical economists for trying to make economics (back then, political economy) more science-y and less opinion-y. This talking point is so long-lived it still remains today.
A good counter is Paul Krugman’s 1996 “The Dismal Science”. In it, Krugman draws out an analogy between biology and economics: much like economists think Robert Reich is an uniformed hack, many biologists hold the same contempt for pop biology writers, who get plenty wrong. This is because models in both economics and biology are largely mathematical, which strips away a lot of nuance from the reasonings: “the first derivative is positive, ergo inflation is caused by printing too much money” is not a convincing argument. Pop econ/bio gurus try to bridge the gap, but some things cannot be taught with just words. Krugman states:
There are important ideas in both fields that can be expressed in plain English, and there are plenty of fools doing fancy mathematical models. But there are also important ideas that are crystal clear if you can stand algebra, and very difficult to grasp if you can't. (…) Alas, there is probably no way to resolve this conflict peacefully. It is possible for a very skillful writer to convey in plain English a sense of what serious economics is about, to hide the algebraic skeleton behind a more appealing facade. But that won't appease the critics; they don't want economics with a literary facade, they want economics with a literary core. (…) The literati truly cannot be satisfied unless they get economics back from the nerds. But they can't have it, because we nerds have the better claim.
Realism vs reality
Friedman’s central thesis, besides that distinction, can be boiled down pretty easily: a model cannot be judged by the realism of its assumptions, but by the strenght of its conclusions to evidence. Imagine you asked a physicist how a soccer player should choose the optimal strategy to kick the ball into the . They’d the kicker would have to know his velocity, the weight of the ball and his shoes, the density of grass, how moist it is, wind resistance, humidity, etc, then solve a differential equation in his head and kick - that’s without considering the goalkeeper or opposing players. Is this a realistic account of how football players behave? No. Would it tell you whether Rodrigo Palacio would have missed this shot in the 2014 World Cup Final? Yes.
The same goes for economics: perfect competition, profit maximization, rational agents, perfect foresight, whatever assumption you don’t like - they’re all as good as the conclusions of the models they build. Perfect competition is reasonable to assume in some cases, and unreasonable in others. And asking people in surveys how they actually run a company isn’t especially useful - nobody is doing profit functions or Cobb-Douglas production functions because they think they’re literally true, but rather, because they think that they’re a useful shorthand for the complicated reality of actually running a firm. Acting as though the soccer player solves a derivative might not be reflective of their actual behavior, which is mostly rules of thumb and past trial and error, but it might actually be a way to make stylized predictions about otherwise too complicated to describe phenomena.
For example, many labor economists consider raising the minimum wage would not reduce employment, since the labor market suffers from what they call “monopsony” - a monopoly, but for the buyer. Because there’s only one company hiring, that company can push down wages - and imposing a minimum wage simply pushes them back up to equilibrium levels (or pushes prices up, or cuts back fringe benefit, or whatever the hell depending on your initial conditions). Problem is, it’s clearly laughable to assume there’s only one company, and improving the realism of the monopsony mode by assuming that there are many employers actually does yield disemployment effects. And the empirical evidence, although mixed, is generally in the “small to no effects” camp. So which is more real: realism or reality?
The rational, self-interested, utility-maximizing moth
Nowhere is this clearer than in the criticism at “representative agent” models and utility maximization. “Companies don’t maximize profits, they apply fixed markups to costs, surveys say” and “people don’t really solve intertemporal integrals of the utlility function over an infinite horizons” are common criticisms, beaten to the death in such tropes as the homo economicus, who brings out an intermediate calculus textbook to figure out which cereal to have in the morning. As mentioned above, nobody is saying that firms literally do choose how much to produce by solving a derivative (they might, in some weird cases), it’s just a very useful shorthand. Alchian (1950) “Uncertainty, Evolution, and Economic Theory” offers a very good explanation of this. But I’ll start with an example from biology and then go over Alchian’s paper.
In Manchester, there is a species of moth known as biston betularia. Traditionally, and since time immemorial, biston betularia was white with dark dots, which is why it’s called the “peppered” moth coloquially. But during the Industrial Revolution, Manchester was filled with coal-fired factory that covered the city in black dust. After a while, biston betularia moths changed color, from predominantly white to predominantly black. How would an economist explain this?2
There’s two ways of seeing this:
A group of heterogenously colored moths, with extremely incomplete information, non-singular objectives, under bounded rationality, and subject to high uncertainty are faced with an exogenous change. The moths are completely incapable of adapting. Consequently, the ones of a lighter color die out, and the ones of a darker color survive. This changes the color of the “representative moth” from white to black.
A single homogenous moth, with complete information, maximizing its utility, perfectly rational, and facing no uncertainty is faced with the same exogenous change. Since it is completely able of changing its color, the moth rationally chooses to become dark instead of pale. This changes the color of the “representative moth” from white to black.
Which model is better? Think carefully. The first one is much more realistic, true, but it’s also far more complicated. The second one isn’t very realistic, but it’s fairly simple. And both make the exact same predictions, that can be falsified in the exact same ways, and corroborated under the same circumstances. So even though the more realistic model seems better, the unrealistic one is just as good instrumentally while also being easier to work with in general.
The same happens with the representative agent paradigm. Firms do behave more like the first model, being not very able to change, facing the same constraints, and choosing prices and outputs following many rules other than “max profit subject to demand”. Instead of maximizing profits, firms take a positive profit as a signal that they’re making the right choices, and take losses as a signal that they’re doing things wrong. Firms that make too many wrong choices get wiped out eventually, and firms that take adequate courses of action make it. This process does involve talent, analysis, calculation, strategy, etc; but since exogenous shocks are basically random, it can be understood that eventually any firm with bad enough luck will go under. However, the average firm can be understood to behave as in the second model with no loss of explanatory power: when “bad” firms go under, that leaves the “good” firms intact, which means that the average firm is “better” than before - the representative firm model is just as good as predicting actual generalized behavior.
While this obviously stems from a change in the composition of the sample of existing firms, it is way simpler to just claim that the average firm changed behavior than to go on and on about how winning many coin flips doesn’t make you good at calling them and doesn’t mean you have a good strategy for the game. This means that talking about it as if a representative firm did X, Y, or Z instead of “the market selected firms that did X, Y, or Z at random” is perfectly valid, as long as the predictions remain as good as in an alternative, less complicated model. Obviously this can be just plain wrong: Le Corbusier proposed that, for architecture and urban design, the representative agent was 6 feet. That’s just a bad model! 87% of men are shorter!
Conclusion
So, what can we take away from this? I think each section has a major takeaway for economists, or just for people who like talking and reading about it:
Empirical or theoretical methods are not inherently superior to each other. As long as a theory can make strong, testable predictions, it’s the same whether they’re tested by experience or by probing its own internal consistency.
Economics is a science because it follows a scientific method. If sciences are understood to be disciplines that make claims that can be proven wrong, economics is one, since its methods are not too different from other sciences.
Economic disagreements are often, and should mainly be, about the implications of decisions, not about values. Whilst most economic policy debates are framed as value distinctions, the only way for economics to remain scientific in scope is to focus on disagreements about the implications of certain modeling choices and their consequences on conclusions.
Models cannot be judged by the realism of their assumptions. Whether or not an assumption is close to reality or not is not enough to rule out any given model, since the durability of its conclusions is what matters - and those conclusions don’t even need the assumptions to be true.
Unrealistic models can be just as good as realistic ones. An unrealistic model that is possible to work with is better than a realistic one that’s a pain in the ass to use - as long as they’re equally good at making non-false predictions.
Sources
Friedman (1953), “The Methodology of Positive Economics”
Friedman (1976), “Nobel Lecture: Inflation and Unemployment”
Paul Krugman, “Economic Culture Wars”, 1996, Slate
Alchian (1950), “Uncertainty, Evolution, and Economic Theory”
The nickname “the dismal science” doesn’t come, as commonly believed, from Malthus’s gloomy predictions about economic growth. Instead, it seems to have originated from slavers criticizing economists for being against slavery.
Karl Popper at some point considered the theory of evolution selection non-scientific, since it could never be disproven that “selection of the fittest” wasn’t how organisms reached their current state - until he changed his mind, because the theory was actually testable (through natural experiments or bacterial cultures, for instance).
Very good post, by the way. I'm a fan of yours.
Why isn't the Pythagorean theorem considered True Knowledge?