

Discover more from Overcoming Bias
Consider this critique of physics:
Once upon a time the universe was full of magic, mystery, and majesty, wherein humans lived organically and intuitively with nature. But then physicists (and their engineering minions) pretended to know far more than humans can ever know in an irreducibly complex universe. And they pretended to far more objectivity and neutrality in their inquiries than is possible for humans. Using impressive math, physicists rose in status, while other less mathy but more fluid and organic ways of thinking fell in status. Physics concepts became used more widely, displacing other useful and more human but now neglected ways of thinking.
Physicists are reductionist, and focus overwhelmingly on the simplest physical parameters of the smallest physical parts. So they ignore more interesting parameters and large scale organization. They study particular phenomena via vastly-over-simplified models that neglect most of the rich complexity of the real world. Worse, regarding the items that they do consider in their simple models, most of their assumptions are just wrong.
For example, standard models of mechanical systems assume that they sit in a flat space-time. Most materials are uniform, isotropic, solid with sharp boundaries, and uncharged. Ground anchors do not rotate or accelerate. Perfect vacuum sits between most adjacent parts, and between the other pairs is either an absolute bond or frictionless relative motion. Yet real mechanical systems sit in rotating, accelerating environments full of corrosive fluids and cosmic rays, at temperatures and pressures that often melt materials, and amid vibrations that often break them. And estimates of all physical parameters in such models are known to be wrong, i.e., not exactly correct. Physicists claim that such deviations make for only small errors in their final analysis, but how can they know that if they don’t model the full complexity?
Engineers who use physics tend to create system designs that are more like typical physics models, with a small number of simple parts having a few simple relations to one another. These systems are quite different from the fluid, complex, highly-interdependent rich-relation biological systems that we are, and once lived among. These physics-model-derived systems are harsh, ugly, fragile, uninspiring, and alienating. These systems may do well by simple physics metrics, but that neglects a vast space of better if less formal ways to evaluate systems.
The dominance of physics in engineer training and related government policy has unfairly neglected intuitive, magical, arty, and literary approaches to engineering system design. Approaches that look bad by physics metrics, but not by intuitive organic human ways to evaluate. Today the fields of “design” use better approaches, and are displacing the fields of “engineering”. It’s about time.
Here’s an obvious response:
For most products, few customers care much about how their systems are engineered, or the parameters by which they are described. So in a free competitive world, firms are free to offer products designed and evaluated via “intuitive, magical, arty, and literary approaches.” But few do. Yes, firms today also commonly use design as well as engineering, but mainly for a few relatively aesthetic choices close to the user experience. For at the vast majority of other choices, out of user sight, physics-based engineering dominates.
Physics winning this competition suggests that alternate approaches just aren’t as productive. Yes, there is often less free competition to woo government buyers, and physics-dominated regulations often demand that physics be used to prove that products are safe and effective. But consider that the world still has many competing nations, and engineering matters greatly in war, where simple physical parameters are quite meaningful. If a nation could build more effective weapons using other approaches to weapons design, they could win wars that way. The fact that few nations try is more evidence that physics-based approaches work better.
Yes, models greatly simplify. But for humans with some abstract understanding and greatly limited mental abilities of other sorts, approximation via simple modular models and designs is our main way to manage complexity. Nature faced different constraints, which is why her designs are different. Yes, simple modular designs can be harsh and alienating, but without them we could not create engineering designs nearly as capable. Humans just can’t do analysis without making a mass of simplifying, and thus wrong, assumptions. But the fact that our designs tend to work shows that our approximations tend to be appropriate. Yes of courses if we approximate badly, our models and designs will go badly. Which is why physicists and engineers pay so much attention to approximating well.
Now consider the many critiques of economics, which I’ve just spent many hours sampling. Most econ critiques are much like the above physics critique, making a similar response appropriate. But with one key difference, to be discussed at the end.
Before going into details, let’s review a few basics. Like physics, econ uses math to create a space of possible models. But instead of describing physical systems, econ models describe social systems. Economists have a standard set of assumptions that they see as most likely to be true, and other standard set of assumptions that seem easiest to analyze. Assumptions from the second set are often preferred, to allow entire models to be simple enough to analyze. Different economists explore different models, comparing them to each other and to data, and arguing about their relative accuracy as approximations. If you are arguing for different models in this topic area, but accepting that models are a reasonable way to think about social behavior, then you are doing econ. (And you might have a valid complaint re if your kind of econ gets a fair hearing.) Econ critics, in contrast, reject, or at lest minimize the value of, the whole econ approach to studying social behavior, and designing policy.
That said, let us now consider some common econ criticisms.
Macro Forecasting – Economists can predict many things with high accuracy, but have low accuracy when predicting some big important aggregate changes, like the timing and size of recessions, wars, and asset price falls. Physicists are also unable to predict some large aggregates, such as the weather, and other places where turbulence dominates. In both cases our best theories go a long way to explaining why such things are generally hard to predict. Yet this lack of ability is held against economists, but not against physicists. Even though no other social analysts can predict these macro changes substantially better than do economists. Seems an unfair double standard to me.
Objectivity Is A Lie – Like many other academic disciplines, such as physics and engineering, economists endorse norms of neutrality and objectivity, and adopt many social practices with an eye to encouraging such tendencies. However, some say that such attempts are mistaken; social analysis goes better when analysts announce and embrace their passions and partisan leanings. But would this really make physics or engineering go better? Should legal judges and teachers no longer aspire to treat all plaintiffs and students neutrally? If you accept that neutrality is often a useful ideal, but reject it for economic analysis, then what key difference in economics justifies this different treatment? Yes, there are many specific differences, but I can’t see one that works. We should of course be honest about how much objectivity we actually achieve, which may be disappointingly small. But it still seems an ideal worth working toward.
Unfair Privilege – Compared to other social scientists and analysts, those associated with economists get more respect and attention, in business, government, and the media. As a result, they tend to be paid more, and their concepts are more widely known and used. Others who do social analysis complain that this is unfair. And it would be if this dominance were caused by a cabal of powerful insiders who control social analysis in these areas of life, and favor themselves. But there is in fact a great deal of competition for social analysis in these areas. And there’s no reason different kinds of social analysts should get equal attention, if there’s no reason to think them equally productive.
Market Fetish – Economists’ models of markets are especially simple, and also especially robust as approximations of more complex variations. This leads economists to often recommend markets. Critics call this a bias, part of a larger economists’ bias toward stark, mechanistic, inhuman, inorganic social institutions. But if so, it is the same sort of “bias” that makes physical engineers prefer device designs that are easier to analyze with tractable physics models. While there may well be better designs out there in the vast space of designs too complex to analyze, we just don’t know which ones they are. Similarly, without a better way to evaluate the vast space of complex institution alternatives to markets, we must choose among the ones we can evaluate. (Some say to trust their intuitions about the virtues of particular complex candidates they like; economists aren’t very trusting here.)
Commodification – This is actually a complaint about markets, not economists. The idea is that long ago, when people lived in small communities, they each had far more negotiating power re their deals with each other, which made each person feel valued. As communities got larger, and markets got thicker, negotiating power fell, closer substitutes appeared for whatever any one person had to offer. Some say that we feel less valued as a result, more like “commodities”, and they suggest banning some kinds of markets, forcing people into less formal more fragmented deal-making, where they have more negotiating power. Standard economic theory, however, suggests we are worse off with such fragmented deal-making, and the fact that people tend to choose large over small communities, when they have a choice, suggests that most people agree. When both sides get more negotiating power, that just increases variance in deal value to each side, not the average.
Reductionism – Like physicists, economists build models of big worlds made out of many small parts, in their case the many particular choices of particular people. Some object that social analysis works better when it considers large aggregates, like classes, ideologies, nations, regions, and professions, as the most basic units of analysis. But economists don’t object to thinking in terms of such units. It is just that, like physicists, we are much more comfortable when when know how to describe such larger in terms of many small parts. Physics concepts like pressure, temperature, entropy, and much more are useful aggregate concepts, but helpfully understood in terms of interactions between small parts.
Rationality – A common econ assumption is that people act in ways that get them what they want. This turns out to be robust to noise; assuming that we only tend to act to get what we want usually gives pretty similar results. And many economists explore models where various other factors influence action choices. Even so, critics say that this usual assumption undermines economist authority, and implies that others should not listen to them. But this seems to reject the very idea of using simplifying models.
Selfishness – Another common econ assumption is that people want more personal material gains for themselves. (Sometimes they even assume people want only money.) This assumption is usually robust to people merely putting more weight on themselves than on others, and to their caring a great deal about a small number of close associates. And many economists explore other alternate assumptions. Even so, critics suggest that this usual assumption undermines economist authority. Again, this seems to reject the very idea of simplifying models.
Stable Preferences – Econ models of what people want usually contain a limited number of contributing factors, and limited ways in which wants change with context. Critics, however, are often impressed with how what people want seems to change over their lifetimes, and has differed across in societies in history. Critics often conclude that there’s little point in analyzing how to get people what they want, as society can change what people want. However, any dependence of what people want on social choices is just another form of context dependence of wants, and fits directly into standard model frameworks. It is fine for economists to make such models and to argue for them. But that’s a way of doing economics, not of rejecting it.
Other Assumptions – Critics tend to complain about econ model assumption unrealism, but usually fail to mention many examples. Let’s help them out. Econ models focus on “small worlds,” leaving out larger outside connected world. Econ models tend to model only limited kinds of uncertainty, yet we actually can be uncertain about almost everything. Different games are analyzed separately, but in fact uncertainty about what game we play merges them into one big game. Models also usually ignore uncertainty about which computational strategies, and decision and game theories, apply. Models use finite and infinite state spaces when the other is more plausible. And econ models often make unrealistic assumptions about actor variety, action timing, action durations, action knowledge, commitments, externalities, insurance, public announcements, and motivations. We usually assume we are near an interior local maximum with concave payoffs.
Wealth Maximization – This seems the most common complaint, so I’ll discuss it in more detail. Economists’ gold standard for evaluating social policy is “Pareto improvement” [PI]; a policy change is good if everyone expects to get more of what they want, or at least would so expect if they fully aggregated their info. A common approximation to PI uses what people expect to gain, without info their aggregating. When transfers cancel losses that observers can estimate, a good approximation to PI is repeated application of cost-benefit analysis, which in essence seeks to max economic (not just financial) “wealth.” Being a net loser here requires implausibly large correlations across many errors in observer loss estimates.
Some call it immoral to give people what they want; policy should instead do what is moral, about which econ knows little. But others (e.g. preference utilitarians) see big moral gains from people getting what they want. And econ using a wealth metric can suggest deals for negotiators to consider, which seems a useful social role for economists to fill; policy advice need not be about morality. Also, just as firms typically use artsy design to set a few surface features, but then use engineering to minimize costs along thousands of orthogonal design dimensions, it can make sense to set overall policy via a few key salient moral and aesthetic criteria, such as various kinds of inequality, and yet set thousands of less visible orthogonal policy parameters via wealth maximization. If a parameter doesn’t much influence key overall dimensions, why not set it to minimize costs?
The above are the most common econ critiques I see. This last one is what I think critics should say:
Little Outcome Competition – Physics-based engineers often compete to make devices that are shown to work or not relatively quickly. But the advice from economist-affiliated policy advisors is not tested as strongly, often, or directly as that of physics-based engineers. So while we might say that econ has won a competition of some sort, it is less clear that this is a competition to be useful in policy design. Perhaps economists are more often just telling customers what they want to hear, or letting them affiliate with economists’ impressive analysis abilities. It is worth noting that the econ advice embodied in the most prestigious econ publications often differs substantially from the advice that the same economists give in non-economist forums. Of course if other social analysts seem similarly disconnected from real world tests, this isn’t a reason to favor them over economists. Nor to favor your own intuitions, which also haven’t been much tested.
As an econ insider, I can see better than most outsiders how well econ models and analysis fits data on human behavior. And having once been deeply immersed in physics and engineering, I can see roughly how econ practices compare to those. I see great insight in common econ model assumptions and practices of approximation, and a substantial continuity with similar practices in physics and engineering (who also have great insight). Economists actually know a lot about the social world, and have tools that can usefully inform social policy. But I can also see how the key institutions that organize economists too often fail to induce us to tell what we know to outsiders. I know, as an outsider you might well expect me to say that about my field even if it were not true. But still, I feel that I should say it.
Common Econ Critiques
Job's genius came from the fact that he was not a propellor-head. He didn't understand a lot about how computers work. He didn't want something that only geeks know how to work properly. Whenever they showed him a new design that required some geekiness to make it work, he told them to try again. He knew that if he, with his limited computer knowledge, could make the device work, then almost anyone could. Just as Coca-Cola is built on the word "Euphoria," Jobs wanted "Delight." Yes, that's an arty concept, backed up by buildings full of engineers who knew how to translate it from abstract into something we can touch and see.
If Facebook can do it...