Priming, since debunked
The more dramatic “talk about getting old to people and they’ll walk slower” examples were debunked. The more pedestrian examples, as with word-association, “appear to be well-established” (https://www.science.org/doi/10.1126/science.345.6196.523-b), judging by a few minutes’ examination of Wikipedia (which common-sense supports: trying to NON-word-associate, as with certain panel & improv comedy games, is strikingly difficult)
In the meantime, large language models were created by mass-producing epicycles and training them (What if intelligence is an emergent property of large numbers of epicycles in an evolutionary context?). What happens when macroeconomists mass-produce epicycles? You get DGSE models which would take thousands of years of data to train (https://arxiv.org/pdf/2210.16224.pdf). In the meantime, you can accommodate as many elephants as you wish, and they can wiggle their trunks and flap their ears!
TL;DR: economist erects glasshouse, installs trebuchet
What happens when macroeconomists mass-produce epicycles? You get DGSE models which would take thousands of years of data to train (https://arxiv.org/pdf/2210.16224.pdf).
Didn't Shalizi's paper you cite trying to school the economists turn out to be wrong and irreproducible due to source code bugs? He hasn't updated his post appendix on the matter despite saying 2 years ago that the fixes would be quick and he was sure the numerical results would still prove the point.
The problem with economics, however, is that while it’s got theories, they are, by and large, not theories about humans.
The discipline which was, at least, intended to provide the theoretical grounding for psychology as a whole was evolutionary psychology. The best summary of the motivation for, and conceptual basis of, evo-psych is the following, written by great cognitive psychologist Roger Shepard in his paper “The Perceptual Organization of Colors: An Adaptation to Regularities of the Terrestrial World?” (1992; this paper was included as a chapter in The Adapted Mind, probably the most import text in evo psych):
STRUCTURE IN HUMAN PERCEPTION AND COGNITION IN GENERAL
For over a century, psychological researchers have been probing the structures and processes of perception, memory, and thought that mediate the behaviors of humans and other animals. Typically, this probing has taken the form of behavioral experiments suggested by evidence from one or more of three sources: (a) introspections into one’s own experience and inner processes, (b) information gleaned about the anatomy or physiology of the underlying physical mechanisms, and (c) results obtained from previous behavioral studies. More recently, in seeking to understand not only the nature but also the origins of psychological principles, some of us have been turning to a fourth source for guidance—namely, to the ecological properties of the world in which we have evolved and to the advantages to be realized by individuals who have genetically internalized representations of those properties.
Taken by themselves, findings based on introspective, behavioral, and physiological evidence alike, however well established and mutually consistent they may be, remain as little more than “brute facts” about the human or animal subjects studied. What such findings reveal might be merely arbitrary or ad hoc properties of the particular collection of terrestrial species investigated. Even our own perceptual and cognitive capabilities, as much as our own bodily sizes and shapes, may be the products of a history of more or less accidental circumstances peculiar to just one among uncounted evolutionary lines. Certainly, these capabilities do not appear to be wholly dictated by what is physically possible.
The following are just a few of the easily stated and well known of our perceptual/cognitive limitations, as these have been demonstrated under highly controlled but nonnaturalistic laboratory conditions:
- Although a physical measuring instrument can reliably identify a vast number of absolute levels of a stimulus, we reliably identify only about seven (Miller, 1956).
- Although a physical recording instrument can register a vast number of dimensions of variation of the spectral composition of light, the colors we experience vary, as I have already noted, along only three independent dimensions (Helmholtz, 1856–1866; Young, 1807).
- Although the red and violet spectral colors differ the most widely in physical wavelength, these colors appear more similar to each other than either does to the green of an intermediate wavelength (leading, as noted, to Newton’s color circle).
- Although a camera can record and indefinitely preserve an entire scene in a millisecond blink of a shutter, the “iconic” image that our visual system retains from a single brief exposure decays in less than a second and, during this time, we are able to encode only about four or five items for more permanent storage (Sperling, 1960).
- Although a computer can store an essentially unlimited number of unrelated items for subsequent retrieval, following a single presentation, we can reliably recall a list of no more than about seven items (Miller, 1956).
- Although a computer could detect correlations between events separated by any specified time interval and in either order of occurrence, in virtually all animals with nervous systems, classical conditioning generally requires that the conditioned stimulus last for a short time and either be simultaneous with the unconditioned stimulus or precede it by no more than a few seconds (Pavlov, 1927, 1928).
- Although a computer can swiftly and errorlessly carry out indefinitely protracted sequences of abstract logical operations, we are subject to systematic errors in performing the simplest types of logical inferences (e.g., Tversky & Kahneman, 1974; Wason & Johnson-Laird, 1972; Woodworth & Sells, 1935)—at least when these inferences are not of the kind that were essential to the fitness of our hunter-gatherer ancestors during the Pleistocene era (Cosmides, 1989).
Our performance in a natural setting is, however, a very different matter. There, our perceptual and cognitive capabilities vastly exceed the capabilities of even the most advanced artificial systems. We readily parse complex and changing visual scenes and auditory streams into spatially localized external objects and sound sources. We classify those objects and sources into natural kinds despite appreciable variation in the individual instances and their contexts, positions, or conditions of illumination. We infer the likely ensuing behaviors of such natural objects—including the recognition of animals and anticipation of their approach or retreat, the recognition of faces and interpretation of their expressions, and the identification of voices and interpretation of their meanings. We recode and transfer, from one individual to another, information about arbitrary or possible states of affairs by means of a finite set of symbols (phonemes or corresponding written characters). And we plan for future courses of action and devise creative solutions to an open class of real-world problems.
To the extent that psychological science fails to identify nonarbitrary reasons or sources for these perceptual/cognitive limitations and for these perceptual/cognitive capabilities, this science will remain a merely descriptive science of this or that particular terrestrial species. This is true even if we are able to show that these limitations and capabilities are consequences of the structures of underlying neurophysiological mechanisms. Those neurophysiological structures can themselves be deemed nonarbitrary only to the extent that they can be seen to derive from some ultimately nonarbitrary source.
Where, then, should we look for such a nonarbitrary source? The answer can only be, “In the world.” All niches capable of supporting the evolution and maintenance of intelligent life, though differing in numerous details, share some general—perhaps even universal—properties. It is to these properties that we must look for the ultimate, nonarbitrary sources of the regularities that we find in perception/cognition as well as in its underlying neurophysiological substrate.
Some of the properties that I have in mind here are the following (see Shepard, 1987a, 1987b, 1988, 1989): Space is three-dimensional, locally Euclidean, and endowed with a gravitationally conferred unique upward direction. Time is one-dimensional and endowed with a thermodynamically conferred unique forward direction. Periods of relative warmth and light (owing to the conservation of angular momentum of planetary rotation) regularly alternate with periods of relative coolness and darkness. And objects having an important consequence are of a particular natural kind and therefore correspond to a generally compact connected region in the space of possible objects—however much those objects may vary in their sensible properties (of size, shape, color, odor, motion, and so on).
Among the genes arising through random mutations, then, natural selection must have favored genes not only on the basis of how well they propagated under the special circumstances peculiar to the ecological niche currently occupied, but also, as I have argued previously (e.g., Shepard, 1987a), even more consistently in the long run, according to how well they propagate under the general circumstances common to ail ecological niches. For, as an evolutionary line branches into each new niche, the selective pressures on gene propagation that are guaranteed to remain unchanged are just those pressures that are common to all niches.
(Shepard then goes on to describe the deep questions which underlie his own work on color perception, one of which the rest of the paper is dedicated to examining and answering. I highly recommend reading the whole thing.)
The problem with economics, however, is that while it’s got theories, they are, by and large, not theories about humans.
To be more precise, economics does seem to have a lot of useful and powerfully predictive theories about groups of humans, but not so much about individuals.
There is a ton of variation in how any given person might act in a certain situation, but when you consider the economy as a whole (as in macroeconomics), the financial markets (as in... financial economics), or even just the market for a single good (as in microeconomics), the noise mostly cancels out and the overall effect can usually be modeled accurately and successfully by considering idealized notions such as purely utility-maximizing self-interested consumers.
But if you try to deeply study a single mind, you no longer benefit from this concentration of measure and your 1st order approximation will often be inaccurate and miscalibrated.
I'm not a fan of tribalism, and I hope you can acknowledge that economics and psychology are concerned with quite different slices and levels of decision-making. Economics handwaves a LOT of indiidual variance and choice into averages and generalities, and completely ignores all of the non-legible activities and behaviors people engage in. Economics is nearly useless in figuring out why I'd rather read trashy sci-fi than code for a few more hours, for instance.
I agree that the modeling of individual humans is WAY harder and psychology has fewer "hard" models than I wish it had. I often suspect economics has too many models, with more and more epicycles added, because they're failing to see some irreducible complexity, and because they're ignoring a lot of illegible motivational inputs.
Maximization is like surface tension: it can explain the local contours of society, but it cannot explain the overall shape of society. That instead requires interfacing with the things the economy attaches to in the external world.
In fairness, I think some Psychology is reasonably theory supported.
For example, there is that famous study where the economist added extra "decoy" options of ways to subscribe (https://thestrategystory.com/2020/10/02/economist-magazine-a-story-of-clever-decoy-pricing/), which apparently helped a lot. I have read about related studies following this idea where they got pictures of people, asked other people to pick the most attractive from a bunch, and found that decoy effects played a part in a similar way. If you take it as a theory that people are bad at comparing options objectively, and can be swayed into taking one option by being offered a strictly-worse alternative then you would be able to make all kinds of predictions, in various different contexts and studies, that you could maybe carry across.
However, I do have a lot of sympathy for your general position. A simple theory that is mostly-right is gold, a staggering textbook of complexity that is 100% right is just overfitting. Economics has a lot of the former, psychology more of the latter.
i. The problem with psychology
Psychology and economics are both attempts to study human behavior. While the precise focus may differ, we want to be able to make precise and accurate predictions about how humans respond to stimuli. We can use reasoning to make theoretical predictions about what will occur, and use empirical results to confirm or disconfirm them, or use empirical results to guide what our theory should be. You must have both, however.
I argue that economics is far better than psychology because it has a theory. Psychology does not. It is the fitting of epicycles to match observed empirical phenomena. If published research were an unbiased reflection of reality, this would be a distinction without a difference, but published research is distorted and obviously distorted. Psychology is a science without priors. They are left chasing the newest paper, without the slightest idea why it should be true or not true. Its only test of plausibility is whether it can tell a plausible story rationalizing the results. The funny thing about stories, though, is that you can spin a story to fit any particular set of facts. It is, moreover, in principle not possible to say what causes what without theory when looking at data sets after the fact. A randomized controlled trial (where people are assigned some treatment at random) can sort out that some treatment caused an effect, but for many interesting questions we can only observe data sets after the fact. As this paper proves, you cannot establish what causes what without a theory, without a prior.
Consider the example of priming. Priming, since debunked, is the idea that making people think about certain things will unconsciously affect their behavior. People who read a text about being old would supposedly walk slower after reading it. It is striking that it would be perfectly plausible to rationalize the exact opposite effect. You find that they walk faster – perhaps because they were reminded of their mortality, and walk faster to remind themselves of youth. You don’t have a sound sense of what would be confirming evidence, and what would be surprising evidence.Research will go on until a significant, publishable result is found – and so you end up with spurious claims on pointless questions.
This is not merely my complaint. Psychologists, surveying their own field with disappointment, agree. Muthukrishna and Henrich (2019)write, “Rather than building up principles that flow from overarching theoretical frameworks, psychology textbooks are largely a potpourri of disconnected empirical findings on topics that have been popular at some point in the discipline’s history.” They go on, “outside of psychology, useful theoretical frameworks tell scientists not only what to expect, but what not to expect.” To quote from Poincare, “Science is built up of facts, as a house is built up of stones; but an accumulation of facts is no more a science than a heap of stones is a house.”
I suspect that psychology is like this because it has its intellectual origins as clinical practice, in an era where medicine had only the scantest idea what to do, and knew still less why things worked. Reading Paul Meehl’s polemic against case study conferences, one is struck by how much of a psychologist’s work is still one to one. He, a professor with a sound footing in statistics, still gave ten to twelve hours a week in private psychoanalysis. Psychology even now has not fully shifted over to being statisticians.
ii. Why economics?
Economics has a core set of theoretical claims that can stand on their own. We can explicitly state our assumptions (generally, what people or firms are trying to maximize) and then show precisely how they can maximize it. Our work on how to optimize an auction requires no experimental proof whatsoever – it stands on its own. So too does much of microeconomic theory. How a monopoly could maximize their profits is simply proven – there’s no other method which maximizes revenue.
And this theory can meaningfully guide what we research. A paper that comes to mind is Ben Bridgeman’s “Competition, Work Rules, and Productivity”, as it could never have been made without a sound theoretical foothold. Suppose that there is a firm with some degree of market power – while not perfectly a monopoly, they are able to get rents. (Rents, in the context of economics, are profits in excess of what would be earned in a perfectly competitive market). Labor unionizes because they wish to divide the rents between them and the firm in some way. Crucially, the union is able to control wages and the number of people hired, but is not able to control output. Under these assumptions, it is optimal for the union to insist upon hiring some people who add nothing to production – what is called “featherbedding” – rather than simply maximizing their wages. A change in wage changes the marginal cost of producing a good. Firms would be incentivized to reduce their total production, in order to claim more rents for themselves. Requiring a certain number of people to be hired changes it to a fixed cost. The marginal cost of producing additional outputs is kept nearer the competitive outcome.
That’s it! That’s all you need! You do not need to appeal to theories of solidarity across workers (which is not so much an explanation as acknowledgement of the facts – you cannot make meaningful predictions about where and why “worker solidarity” would vary). Everything follows from conventional microeconomic theory taught to every undergrad. All you need to explain featherbedding is simple maximizing behavior. You can then make strong predictions about the degree of featherbedding as competition increases, and what will happen to wage rates. As the market becomes more competitive, workers will be willing to reduce wages first, before they reduce the number of people required to be hired – as he supports on page 13. The theory leads to meaningful, testable predictions. The two work hand in hand.
I am an economist. I may be biased by tribalism, but I should hope I have the independent will to choose my tribe. I do fundamentally believe that the ethos of economics is simply a better way to study the world. I like economics for its serious concern for proper statistical inference, for its intolerance of stupidity, and for venturing out into poorly handled fields and setting them right. I hope we never lose this.