Economists: Do They Know Things? What Do They Know? Let's Find Out!
Do economists know anything about the economy?
A friend of mine once said: You know what the problem is with being an economist? Everyone has an opinion about the economy. Nobody goes up to a geologist and says, 'Igneous rocks are fucking bullshit.'
Popular adage in economics
Recently, all cryptocurrencies plunged in value for basically no apparent reason. I even wrote about it. But a more interesting question emerged: could anyone have seen it coming? “Economists can’t predict the next recession” is a common criticism. Economists, with all our knowledge of markets, finance, etc, find out about literally every major economic event alongside everyone else. Is this because economics is bad - or is there something more at play here?
What’s in a pencil?
A man is driving through the highway, listening to the radio, when he catches the news. In them, the speaker said “Caution all drivers, there is a madman driving in the opposite direction as everyone else. Please drive carefully”. The man, looking around, thinks “What do you mean just one? There’s thousands of them!”.
Common Spanish joke (in that it’s in Spanish and also racist against them).
The main player in the information economists use to analyze anything involving a specific market is prices. A price is, surprisingly to all readers, how much goods can be exchanged for a given amount of money (even if you wanted a peanut). Normally, prices are taught such that supply equals demand and they clear the market. But not so fast - what’s really going on?
Markets, and the price system in general, are a type of spontaneous order. Spontaneous order can best be summed up as “the result of human action but not of human design”. Without being guided by anyone, people just settle on certain courses of actions as an equilibrium - their actions being “decided” by cost and benefit, not by much else. Some examples include language or currencies, but traffic rules also work. The reason people follow them isn’t (just) because they’re law abiding, but rather because they expect that most everyone else will too, which means they’d be in jeopardy (legal or physical) if they didn’t.
People didn’t really settle on “money can be exhcanged for goods and services” because of some grand philosophical debate around social engineering - but, rather, someone came up with it, it seemed like it worked well enough, and it stuck around because barely anyone wanted to try something else. You’re always welcome to start a primitivist commune in the middle of the forest, but most people don’t want to.
So what do markets well, do? People do work, and they exchange that work for money. The baker knows how to make bread, the brewer knows how to make beer, the butcher how to uh make meat, and the only things that brings them together is their desire to make money. It is not from the benevolence of the butcher, the brewer, or the baker that we expect our dinner - greed, for lack of a better word, is good. And the knowledge of how to bring together the butcher, the brewer, and the baker to the supermarket is the kind that everyone has, but nobody does - they just buy and sell, and it all happens basically “spontaneously”.
But everyone holds specific knowledge that nobody else has, and there’s a lot of knowledge that nobody has but everyone does, how do people make decisions? This is especially relevant when looking at changes in circumstances: things change a lot, usually in genuinely unpredictable ways, and behaviors change a lot. It’s not really possible for a single entity to have all the specific information every individual does, let alone the information contained in “spontaneous phenomena”, and couldn’t keep up with all the changes that happen all the time. So how can people communicate information about changes in circumstances to each other?
The answer appears to be through prices. Prices ultimately reflect the scarcity of one good in relation to all others. Changes in circumstances are reflected as changes in scarcity, so things nobody wants have to get cheaper, and things everyone wants have to get more expensive. This provides signals about the relative scarcity of products, to which individuals and firms can respond by optimizing their actions. Thus, the main role of prices is to synthesize all available information on the relative abundance of something, regardless of the circumstances that produced that abundance.
Fool me once, fool me twice, can’t get fooled again
A hundred-dollar bill is lying on the ground. An economist walks past it. A friend asks: "Didn't you see the money there?" The economist replies: "I thought I saw something, but I must've imagined it. If there had been $100 on the ground, someone would've picked it up.”
Common joke in economics circles
What information do prices provide? I’ll start off with an example: pork. Like many other commodities, pork prices have a weird cycle, where prices increase and decrease cyclically over time - particularly, in three year periods, representing the time it takes a pig to reach maturity and therefore be fit to be turned into pork.
Imagine you are a farmer. You see prices are up. You breed more pigs, and wait 3-4 years for them to grow big. But everyone else thought the same thing, so you go to market and it’s overrun. You say “okay, I’m going to not breed any pigs this year, because prices are low”. 3 years go by, and boom, there’s barely any pigs. The cycle repeats - in what’s known as the cobweb model. In it, supply is determined by past prices, which means there’s many opportunities to make a lot of money for nothing - just breed a lot of pigs when the market’s “loose”. The demand cycle is the same, but the reverse: consumers expect low prices to remain low, and high prices to not go down, generating a reverse pattern that produces cyclical prices.
In the world of the cobweb model, where people are stupid (let’s be frank here), two things happen. The first is that “you can fool all of the people all of the time”: if you wait long enough, you always get to pull away the football, and you always get to tease they might kick it again, and they always run back. The second one is that economists always know best: unlike the uncultured yokels wrangling disgusting animals, us smart geniuses can always provide advice that’s better than what they already have. Abitrage is possible - there are hundred dollar bills on the sidewalk. For the macro inclined, this means that businesses invest too much, they make too little money and go belly up, they invest too little, there’s a recession, they invest more because the marginal benefit is high and rates are low, etc. The market economy is inherently unstable. It also means recessions are predictable and financial crashes are forecast-able, if you know the signs and can spot them.
But wait: why do pork farmers never realize they’re playing themselves? To quote George W Bush “fool me once, shame on you. Fool me twice, can’t get fooled again.” It would be realistic to assume people are a bit stupid, but it would be very unrealistic too assume they’re that stupid. Pork farmers might be yokels but even they can tell bankruptcy from solvency. The key issue here is the nature of expectations: the cobweb model relies on backwards looking expectations (“adaptative”), which means you are always running around based on old information.
The alternative is the much reviled and equally misunderstood rational expectations view. The adaptive expectations position is that expectations are formed by looking at the past state of the economy by actors who are as close to real human individuals as possible - i.e. not very bright. The rational expectations view is that individuals are very smart, and that they operate with a full working model of the economy (whichever one is being used, in fact) that has complete information. If individuals have a full working model of the economy in their heads, and they have all the information necessary, that means that they can make predictions about the future that always come true - unless something unexpected happens. You can’t fool all of the people all of the time, and there are no hundred dollar bills in the sidewalk. When a farmer sees pork prices, he knows what all the other farmers are going to do, and so cannot benefit by unilaterally deviating from that pattern- if he did, he’d lose money.
… if expectations were not moderately rational there would be opportunities for economists to make profits in commodity speculation, running a firm, or selling the information to present owners.
John Muth (1961), “Rational Expectations and the Theory of Price Movements”
The rational expectations view sounds super wacky, but it’s pretty insightful. Instead of assuming that prices convey basically no relevant information, it states that prices convey all information1; instead of concluding that individuals systematically leave out tons of information, it states that they use all of it. Economists can’t make infinite money by just barging into commodity markets and going against the grain because the people who do it for a living and have always done it for a living would have come up with it at some point. You don’t need a PhD in real analysis to notice something that happens in your full time job every six years, believe it or not.
Rational expectations, however, does not say that people literally act like that, or that it’s reasonable to think so. Contrary to what whatever training in economics you’ve had, and also the preceding paragraphs, the assumption isn’t that actual people literally behave like that. Rather, assuming that people are minimally rational (i.e. pursue profit and kind of understand price signals), behaviors that are patently irrational get rooted out, and therefore the remaining actors will seem as though behaving rationally regardless of whether they actually do. The assumption isn’t placed on individuals, but rather behind them: the economic system works as if individuals were perfectly rational, because they can learn from mistakes that cost them money. The pig loving yokels know more about pigs that some guy from Cambridge University, and the latter can’t profit from their knowledge, because they understand pigs well enough: anyone who tried to breed pigs during the low price period to make arbitrage profits went out of business. Economists, regardless of whether we understand the economy, don’t understand your job better than you.
So, what does the rational expectations view imply? As stated above, rational expectations means that no information reflected in prices is wasted, so that the market works as if everyone had a complete understanding of it. In consequence, as long as information is properly reflected in prices, then there are no opportunities for arbitrage - economists don’t know better than everyone else. This doesn’t mean that economists can’t say anything useful, because markets don’t always function correctly, plus people don’t actually understand the market, it works as if they did - which are entirely different things. This view also never implies that market outcomes are always optimal, since incentives might be inadequate or information improperly distributed. Lastly, that predictions are rationally informed doesn’t mean they’re always right - just that they aren’t systematically wrong. Pork breeders might occasionally miss a bad harvest or a plague or a new fad for some pork product, but they don’t fail at their only job every single time.
In a very “Kafka and his Precursors” (Spanish version) kind of way, John Muth’s 1961 paper establishing the rational expectations assumption is only really relevant because of posterior developments it inspired that made it relevant: the Lucas Critique and the Efficient Markets Hypothesis, both of which happened nearly 20 years later.
Monkeys throwing darts at newspapers
Do stock market analysts have the same problem as pork farmers? Stock crashes, bubbles, currency runs - those happen all the time, and everyone (professionals and economists alike) are always caught with their pants down. Are they good at their job? Yes, actually, they are very good. That financial crashes are unpredictable is true by design.
Let’s start with the basics: under the weakest possible version of RatEx, asset prices reflect publicly available information - about past values, present values, and expected future value at each point in time. For example, Tesla might be worth 10 million dollars, but if everyone thought they would make a breakthrough that tripled their value very soon, the company would be worth 30 million instead. Under competition, low barriers to entry, and low cost of acquiring information, almost all public information is reflected in prices - and markets are informationally efficient.
The Efficient Markets Hypothesis (EMH) simply states that, in informationally efficient markets, three things happen: past prices can’t predict future prices, new (public) information is incorporated into prices as soon as possible, and no trading rules can consistently overperform the market. That future prices are esentially unpredictable also means that future crashes in prices, even across many stocks, are unpredictable as well - because past prices only tell you anything about future prices ceteris paribus, and the real world is mutatis mutandis. Additionally, this means that forecasts made on public information won’t move the needle, because they’re already priced in - if I go on tv and say Tesla stock will crash, nobody would care on average. But if Elon Musk did it, it would move the market, because he has private information.
Are these implications true? The first one is true in the short term, because long-term patterns can actually be understood by changes in technology, demand, risk preferences, etc. - Tesla might say “we’re about to make a breakthrough” but, if the breakthrough doesn’t come after a while, it’s safe to assume it’s never coming. New information being incorporated into prices quickly is hard to test too, but stuff like Elon Musk’s semi-credible bid to buy Twitter pushing stock prices up to a level somewhat similar to the one he offered to buy the company for would seem to at least not contradict it. The hardest one to prove is definitely the last one - professional stock pickers couldn’t consistently outperform “monkeys with a dartboard”. The evidence is highly convoluted, but in general, index funds (i.e. funds that just buy up a lot of random stock in a way that reduces risk) tend to be the best long-term performers, precisely because they don’t try to guess which stocks will go up. As you can see in the (very funny) example above, beauty influencers picking random stocks overperformed Norway’s best fund managers. Ouch. Now, this doesn’t mean that the EMH is always literally true; it means that it’s a better approximation of facts than assuming information is just lying around, waiting to be used.
The most controversial conclusion is, clearly, that it’s impossible to predict future stock prices - and thus also future financial crises and speculative bubbles. The EMH doesn’t really say that they are impossible - it’s fairly plausible for expectations of a company or a currency’s future value to just be very wrong due to inadequate public information - but rather, that they are impossible to predict reliably. Robert Shiller, a big opponent of the EMH, uses the 2009 housing crash2 as an example of a predictable bubble - Fama’s response: “Right. For example, Shiller was saying that since 1996”. Accurately predicting a crash is actually perfectly in line with Fama’s view, as long as you don’t have inside information, because, while you can actually predict that a bubble popping or a market crashing are imminent, you can’t systematically predict when they’re going to happen. Bubbles happen but nobody can reliably predict when they pop3 - the people in The Big Short lost millions while waiting for the housing market to crash, and made millions when it did - all in line with the EMH.
If there were, we couldn’t afford them
Can economists predict recessions? Can economists predict the exact results of economic policy changes? These seem like fundamental questions, and very easy ones to answer: recessions still happen, and bad policies get implemented anyways, so no. But why is it this way?
The key place to start with is that forecasts and predictions are not the same. A forecast is a claim about the future based around present trends continuing: “since it’s cloudy, cold, and humid, I predict it will rain”. Meanwhile, a prediction is a conditional statement about a potential state of affairs - “if you print too much money, you get inflation”. The key difference is in the how: to make a forecast, you only need to know statistics, but absolutely nothing about the subject in question. If you rate all of the soccer players in each team, and then all the teams, you can forecast who is likeliest to win - without knowing anything about soccer. A forecast is right or wrong, whereas a prediction is true or false. Meanwhile, you actually need to know anything about the subject to make a prediction: “based on current polling, Trump is going to win Florida” is not the same as “if crime rates increase, Trump would improve his performance among Latinos and win Florida”.
The Lucas Critique states that the same techniques being used to make economic forecasts are worthless at making economic predictions. To make a forecast, say of the tradeoff between inflation and unemployment, you run a regression linking the variables in question. Then, you plug in the observed values of the independent variable, and get an estimate for the effect - “decreasing inflation by 1% would require a 10% contraction in GDP”. You could even use this for policy: how much should the Fed raise rates to get inflation back to 2%? Just run the numbers and know for sure!
The problem, if you assume that expectations are rational, or that people react to new information, is that this is untenable: imagine you assume people consume and save fixed proportions of their income, and that investment and savings are always equal. You could say “raising interest rates would harm the economy by reducing consumption and investment, through lower savings”. Now, higher interest rates raises savings, so it’s possible that a big enough hike both simultaneously lowers consumption and raises savings so much investment offsets it. But a fixed coefficient regression would never produce this outcome, because it does not allow for changes in patterns of behavior. The result of your regression is a forecast; the result of an economic model dynamically linking the key variables is a prediction.
So, for example, imagine you forecast that raising inflation by one point would lower unemployment by two. You then raise inflation as much, and since nobody expects it, it works. The forecast was right. Now, if you try it every time, people catch on (“can’t get fooled again”), and it stops working - suddenly inflation increases with no change in unemployment. You can try raising it by two, but it only works because its unexpected - in the long run, people catch on. Some of the time is not all of the time, and the long run ends up coming around. Forecasting the result of policy changes is useless when those policy changes also result in incentive changes - once the new information is “priced in”, everyone will behave differently.
That forecasts are useless does not mean economists can’t say anything about policies. Lucas was an avowed “Friedmanite in the issue of methodology”. Let’s go back to predictions. Imagine a proposed tax hike, and a prediction saying “if you raise taxes, growth will slow down”. This proposition might be estimated empirically, but the prediction itself can only be adequately analyzed by people who know anything about economics. To answer the qeustion, you need to know if the model is reasonable, how it reaches its conclusions, what other similar models claim, and if there is empirical evidence about similar proposals. Predictions require skill, and those skills are not easily learned - what separates good and bad research is not trivial.
The optimal policy in basically any realm of economics, under RatEx, is to just create stable rules for when you’ll do what, so people know what to expect at all times and aren’t just caught blindsided every time. If you want to keep inflation down, or help the unemployed during recessions, having stable, credible rules you commit to in advance makes your commitments more effective - otherwise you’ll have to overdo it every single time, with the necessary drawbacks it produces. If policymakers try committing to a rule and then breaking it for maximum effect, what happens is people won’t trust any of their other rules until they make up for it - at a cost.
This has a very interesting implications: you can’t anticipate recessions if policy is systematic. Absent policy blunders, and with stable rules of the game known by all, then equilibrium outcomes are self reinforcing: the economy doesn’t go into a recession on its own, but rather, something pushes it in that direction. It can be a policy mistake, such as a bit cut to government spending or excessively tight monetary policy, or it can be some unanticipated event, like OPEC raising oil prices or a virus forcing everyone indoors for a year. The Federal Reserve would find out a recession is happening alongside the rest of us - and even when they cause the recession themselves, they wouldn’t know until it’s too late (assuming they didn’t intentionally cause it on their own), because it’s either completely unpredictable (how could anyone expect the FFR to determine Putin’s military agenda?) or because it’s based on decisions with unforeseen consequences (popping the 1929 stock bubble).
Conclusions
What have we learned about rationality in economics?
Rational expectations means that individuals understand the markets they participate in. Unlike in other frameworks, information is not consistently wasted in the RatEx model, and agents do not systematically make the same mistakes over and over again. Fool me twice, can’t get fooled again.
Rational expectations implies that there are not many arbitrage possibilities that haven’t been exploited, and none at all that only economists could exploit.
The Efficient Market Hypothesis finds that asset prices are fundamentally impossible to forecast, so that financial analysis is nearly all guesswork and luck, and financial crashes can never be foreseen. If all (publicly available) information is already priced in, there is no way to forecast future values, which means only individuals with private information may do so.
Forecasts are only useful insofar as there is reason to expect that changes in outcomes don’t change incentives. A forecast assumes a continuation of current trends, and does not require any insight onto economics. A prediction is a conditional statement, and does. Forecasting is not accurate because forecasters must assume stable relationships between variables, even when policy changes might change those variables.
Sources
Previous posts on the epistemology of economics, and on the economics of Bioshock
Prices and spontaneous order
Hayek (1935), “The Use of Knowledge in Society”
Rational expectations
Kaldor (1934), “A Classificatory Note on the Determinateness of Equilibrium”
Muth (1961), “Rational Expectations and the Theory of Price Movements”
Alchian (1950), “Uncertainty, Evolution, and Economic Theory”
Josh Hendrickson “Rational Frameworks, Not Rational People”, Economic Forces
The Efficient Markets Hypothesis
John Cochrane, “Gene Fama's Nobel”, The Grumpy Economist, 2013 (summary)
Fama (1980), “Efficient Capital Markets: A Review of Theory and Empirical Work”
The Lucas Critique
Lucas (1976), “Econometric Policy Evaluation: A Critique”
Kydland & Prescott (1977), “Rules Rather than Discretion: The Inconsistency of Optimal Plans.”
If you’re wondering how this doesn’t contradict the previous claim that prices contain only the bare minimum information, you are in for a treat from one of my favorite substacks.
I’m still of the opinion there wasn’t really a crash because there wasn’t anything irrational about expecting housing prices to grow indefinitely in large cities: major economic centers that allowed nearly no new housing construction were going to get more expensive. The Fed’s response to a commodity price shock crashed the economy and it nearly brought the entire financial system down because of banks’ extraordinarily irresponsible practices.
The big policy implication isn’t that policymakers can’t or shouldn’t regulate the financial market, but rather that regulations should be about creating rules for “fair play” and efficiency, and that they shouldn’t try to “pop bubbles” or predict upcoming crashes. The Fed trying to pop an asset bubble was, after all, a major trigger of the 1929 crash, and therefore an inciting incident for the Great Depression. With a track record like that…
I wanna say... Elijah Wood?