“But you see we manage our land without such extreme measures,” said he, smiling: “Levin and I and this gentleman.”
He indicated the other landowner.
“Yes, the thing’s done at Mihail Petrovitch’s, but ask him how it’s done. Do you call that a rational system?” said the landowner, obviously rather proud of the word “rational.”
–Anna Karenina, Leo Tolstoy
Petrovitch is a wealthy landowner in early 19th century Russia, in the novel Anna Karenina. He’s a side character and exists mostly for one of the main characters—Levin—to argue with and explain his new idea on farming and labor. Tolstoy subtly crafts Petrovitch’s arguments to be stale, but with the right mix of economic terminology. Rationality means something important in their discussions, but isn’t well defined. In the book, Levin operates a farm that employs hundreds of peasants shortly after serfdom had been abolished. Like other farmers, he isn’t turning a profit. The peasants don’t have an incentive to work hard, it’s a simple principle-agent problem. And Levin crafts an idea to have them share in the profits with him to give them motivation. This was when economic reasoning was beginning to offer solutions in the abstract theory, but before much of it had been actually tested. Levin called his system rational farming, and it develops into his life’s work to implement it and teach it to the Russian economy.
Economists have always prided themselves on being scientifically more rigorous and rational than other academic fields. The field has grown without a rigorous empirical scientific method. Many of the greats, such as Samuelson, Wicksell, and Keynes, were viewed as mostly theoretical. It made sense at the time, and they were far smarter than I am, but following another half century of scientific knowledge, it’s time to realize that economics needs to be grounded completely in empirics. This doesn’t mean we have to start over though, it’s possible to reformulate their arguments within an empirical and testable framework. I won’t try to do that in this post, but I want to explain why it’s important. For example, it’s easy to assume an abstract and true model of the world, and then impose what rationality means. But it is challenging to rigorously explain what it means to call someone else who disagrees with you irrational.
In economic models it’s uncommon to consider that each individual is using his own different model of the world. It is more standard to have a single model that defines the action space, and have players within that world. Even in probability games, the players still are interacting within a shared structure. As a famous example, there is the prisoners’ dilemma. While originally viewed as theoretical, these models are absolutely empirical. This model is built on empirical observations ranging back before economics was a discipline. The concept of human betrayal is built into the written history of our species. Hobbes modeled our political system similar to a prisoner’s dilemma. Except in his model it was one where we were all caught in a bad government, with an optimal solution of cooperating.
That people are self-interested, and that people want more rather than less, are predictions based on observations and individual conceptions of history. Hobbes was a well-read philosopher, and as he studied history he saw that there was an equilibrium where even stable dictatorships were preferable to some idealistic democratic dream. This seemed to be a constant throughout different places and events. This reasoning was the type of prediction a computer can’t handle, and is based on our incomplete data of things we have seen and read throughout our lives. As well as the way we combine our empirical views with our own personal experiences to generate a prediction.
Throughout these observations, we have seen patterns that let us consistently look for a few key variables that are always present. Within this perspective, an economist is just someone who uses the scientific method when studying people, but who starts with a prior that a few variables regarding self-interest explain most of the variation. The reasoning behind these models distinguished economics as its own discipline. It wasn’t until later that economists went back and formalized these models in terms of rigorous assumptions and math.
The prisoners’ dilemma isn’t a theoretical model that needs to be specifically tested in a controlled environment to see if it’s true or not. In fact, the controlled environment will add experiment noise, seeing as how it will just appear to be a silly game. There are no rules for how ‘serious’ the punishments need to be for the game to be valid, but I think it’s reasonable to hypothesize that they should be way worse than simply gaining or losing $20. The way the model was originally created was by looking for a common structure among all historical events and human interaction. Following Ostrom’s work on how these models interact within society, in any given laboratory test of the model it will probably miss other important parameters unique to the time, location, and circumstance.
What would add to the field is an emphasis on studying what causes additional considerations and predicting them. Here a laboratory test could be useful, if it were standardized across different countries. Even if it’s only an approximation to a serious game, the differences between sample populations might still be meaningful. More importantly, the view that a formal test using error statistics and a computer is always the optimal tool for empirical analysis is wrong. Currently computers are not even close to being able to pour over disparate historical texts to search for a common structure of human interaction in certain situations.
Economics as a discipline is built on centuries of empirical research showing that simple measures of self-interest can predict people’s actions. But instead of letting sociologists handle the rest (and usually mess it up), economists should instead start studying the residuals. An empirical achievement would probably involve slightly less effort on extremely complicated mathematical equilibrium refinements (and I’m not just saying that because I’m jealous that I don’t understand them), and instead try to estimate how simple measures of self-interest in the form of interaction in tested games vary throughout different regions and cultures. These could be used to help predict policy success.
If we step away from the world of testing models, we can look at how terms like ‘rational’ are casually misused by Nobel Prize winning economists. Krugman is a good example here of how even smart economists often have a tenuous grip on the rigors of economic scientific inference (I also get to be like my academic role model John Cochrane, and bash on Krugman). Or they somehow think the rules don’t apply when not formally making a model for an academic paper, which is ridiculous, since the entire point of scientific inference isn’t just to publish, but to actually meaningfully understand the world. Krugman’s scientific formalization of the world starts by imposing preferences on others. Of course, since he is not other people, he needs to try to infer their preferences and their own model of how their actions would affect the world, and then imagine how he would have acted given their preferences but using his model of the world. Imagine a conversation between Paul Krugman and a conservative voter:
Krugman: I was looking at your economic information. It says that you work as a janitor for a local school system, and make $28,000 a year. The Republicans don’t work as hard to fund education as Democrats. They also believe and will support high quality health insurance for you, and are in favor of unions that could help you negotiate higher wages.
Voter: I don’t need a hand-out from anyone. I know a janitor doesn’t make much money, but it’s my job and I work for what I earn. I don’t know as much about funding, but the school looks alright. My wife and I work hard to pay for our health insurance, and so far have not had any issues. But we really don’t like the condescending attitude most Democrats have about how we choose to live. We are part of a Christian community, and we think a great society needs to follow the right values.
Krugman: By ‘right values’ do you mean you want to defund Planned Parenthood and force the poor to suffer from high unemployment and low opportunities because the Republicans will shoot down any economic policy that tries to lower inequality?
Voter: We work hard and are proud of our values. It’s not our job to fund welfare programs and abortion clinics. If liberals want those, fine, they can buy them.
We know Krugman views these voters are part of, as he calls them, the “irrational right.” To come to this conclusion he tries to infer what they care about and what they want from their country, then he thinks to himself “If these are the things I wanted, how would I go about getting them?”, then he looks at how they go about trying to get them. If these two things don’t match, he calls them irrational. It’s not just Krugman, it’s how most economists use the word. He is making the implicit statement here that he knows what they want, and knows what they should do to get it better than they know themselves. And he could be right, after all he knows more about economics and policy than most people, but it would be hard to know.
It gets weird when we think about how we would test his hypothesis that they are irrational. We could split the world into two timelines, and then simulate ones where their policies go through and where Krugman’s go through. Imagine it is a game show, and Krugman and the voter are each in separate rooms. Then we can have the results of any variable—no matter how intangible or vague—quantified and given to our players. After pouring over the results, they both meet. They each say “Hah! Looks like I was right!” Krugman looked at all the policies he thought they would care about. But the voter looked at other things, like the excitement of celebrating an election victory with his community, the excitement and feeling of accomplishment he shared with his family, and the belief he had that the world would be better for his children.
This is an area I’m still trying to work out myself. It’s very hard to properly be scientific when talking and disagreeing with someone else. When starting it’s possible to share information and try to understand the evidence and model being used by whoever you disagree with. But if you two still disagree it is not clear how to resolve the issue scientifically (if that is even possible). From all this follows that rationality is at its core about scientific inference and how it is used to infer what people want given their preferences. And even this is unsatisfactory, as preferences are based on our model of the world, so it’s not accurate to consider them as separate or conditional. At its best it’s a useful term because it lets us embed scientific inference when we talk, and still represents an important idea about disagreement on models and conceptions of the world. But at its worse, it is a term that is associated with our over-confidence in our ability to predict how others view the world.
In Anna Karenina, Levin’s rational farming system didn’t work how it was supposed to. He incentivized the peasants, taught them how profit-sharing works, and tried to inspire an excitement at the prospect of mutual riches. But the system was too different. They had no generational or cultural history of entrepreneurship. Their conception of work and life was one that had developed throughout serfdom, and was antagonistic at its very core. They were fundamentally suspicious of land-owners, uneducated, and viewed every elaborate new plan as one that was zero-sum.
He guessed that if they were rational, they would want to participate in his rational-farming system. But his prediction of the peasants’ model was wrong. He didn’t know what they wanted or how they would use their prior knowledge of the world to interpret his economic view of the world. Unfortunately, Russian agriculture after serfdom wasn’t able to bring an entrepreneurial system to the peasants. By the end of the 19th century it was the worst agricultural system throughout all of Europe. And by the 1930s the centrally planned Soviet model of agriculture resulted in the deaths of around 6 million people.