Why use rational choice models?
A particular focus of criticism of mainstream economics is the use of rational choice in many orthodox models. Some critics are so extreme as to claim the use of rational choice should invalidate a model a priori. In reality there are many sensible reasons rational choice was and is still used commonly in economics:
People might be rational (on average over time) after all
Many results in behavioural economics come from small scale experiments in the form of simple games or challenges a small group of test subjects, often students, take part in. Some critics have challenged the external validity of these experiments and their application to the situations economists usually study, which might include seasoned entrepreneurs rather than nervous test participants unfamiliar to the ‘game’. Richard McKenzie emphasizes this point on his work defending the use of rationality. For instance, Richard McKenzie challenges the work of Kahneman for assuming that people favouring a ‘sure thing’ (option A: $800) over a gamble (option B: 85% chance of $1000, 15% chance of nothing) contradicts rational choice, he writes:
“What the behavioralists miss is that variance in outcomes is also consequential in assessing options. Option A has no variance; Option B has a substantial variance, with the outcome ranging from zero to $1,000. Hence, for many choosers, Option A can be more valuable than Option B. Indeed, if expected value were all that mattered, people would never buy insurance. Is the purchase of insurance irrational?”
He further points out that if the experiment is properly framed to see how the frequency of choices varies with price as conventional theory would predict, the participants behave in a consistently rational manner as his own experiments found:
They can, however, predict that if the payoff from Option A falls from $800 to a lower number, more people will choose Option B. Call this “predictable rationality.” My MBA students appear to be predictably rational, given that the percentage choosing Option A goes down progressively as the sure thing falls to $750 and below. This raises the question: Why did the experiment set the spread between the amount of the sure-thing option and the expected value of the gamble chosen so low — only $50? Was it to help the behavioralists demonstrate a high degree of “irrationality?”
Similarly, economists can predict that if the variance of Option B declines, holding Option A constant, a higher percent of people will choose Option B. I’ve also run an experiment in various classes by reducing the variance of Option B. For example, I’ve started the experiment by giving students one draw of one “coupon” from “Barrel A,” which contains only coupons with a redemption value of $800. Or they have ten draws from “Barrel B,” in which 85 percent of the “coupons” have redemption values of $100 and the rest have redemption values of zero. Sure enough, the percentage of students choosing Barrel B is higher than in my first experiment described above. When I tell them that they have 100 draws from Barrel B with 85 percent of the coupons worth $10, it increases yet again. The students are predictably rational, not predictably irrational.
He goes on to demonstrate how framing choice problems like these in a more relevant entrepreneurial/investment setting, with more realistic outcomes and pay-offs, allowing for repeats of the test to account for learning, and ensuring the students of a personal/financial stake in the choice they make causes the behaviour of the students to converge towards rational choice as predicted by conventional theory. He concludes that piece with a rather salient remark:
The moral of my sequence of classroom experiments is simple: You can easily prove that people are irrational if you tightly constrain the choice environment, barring choosers from knowing what others are doing, preventing choosers from correcting errant decisions, and ensuring that “bad” choices do not have economic (and monetary) consequences, with subsequent effects on people’s incentives to learn from and act on their own and others’ errant choices. In such environments, drawing out irrational decisions is like shooting fish in a barrel.
A more recent study from Francesco Cecchi and Erwin Bulte, which investigates using experimental evidence whether market experience promotes rational choice, make similar points, noting the first generation of experiments on rational choice typically consisted of a non random sample of students with limited external validity.
By making participants feel the losses associated with irrational behavior, markets may help individuals learn to express their preferences in ways more consistent with a rational model.
The authors develop a field experiment in Ethiopia involving local farmers and brokers without experience in a competitive market, split into control and treatment groups, with the treatment group participating in an auction exchange selling and buying sesame. The groups were asked to perform simple rational choice experiments before and after the trading session — these “Generalized Axiom of Revealed Preference (GARP)” experiments were deliberately distinct from simulated sesame market exchange so as to discount so called “learning by rules of thumb” effects. As expected, the initial test found frequent violations of rational choice. After the trading session however, the authors found a significant reduction (42%) in the number of violations of rational choice in the treated group, who were able to gain experience in a competitive market. Given a short trading session was able to generate such a significant increase in ‘rationality’ amongst the test participants, this should have major implications when studying actually existing markets and merchants, who have had years or decades of experience.
This of course corroborates previous studies that find experience is both a ‘catalyst’ for rationality and a ‘filter’ of irrationality, where aggregated market outcomes “quickly converge to neoclassical predictions.” To summarise, the results showing irrationality from first generation of economic experiments, often involving students in very limited scenarios with no personal stake, are not necessarily applicable to real world markets and scenarios — while there is strong evidence that real world market experience causes individuals to converge towards rational choice.
Self awareness, anomalies and Goodhart’s law
One alternative to using rational choice is to explicitly model irregularities and biases as if they are a consistent aspect of behaviour which can be used to make predictions. There is a challenge with this approach however: self awareness.
Many studies have shown that anomalies (divergences from what an efficient ‘rational’ market would produce) in asset pricing behaviour soon disappear once papers are published on their existence. This strongly suggests that behavioural irregularities in relevant economic scenarios cannot be relied upon to be persistent, and because of this may not be suitable for use in long term predictive models. This is somewhat similar to the concept of Goodhart’s Law, which states “any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes.” This lends itself to my third point:
Rational choice as a robust benchmark: don’t always assume people are idiots
Given what we know about the instability of anomalies in real world markets, one should question policies that rely on this behaviour. Ask yourself: how comfortable would you be with a set of policies that would completely break down if people were rational, following their own self interest? A policymaker should be strategic; is it better to overestimate or underestimate the intelligence of participants in the economic ‘game’? Making sure a policy or system design doesn’t collapse by people ‘cheating the system’ or simply behaving rationally is one use of models with rational agents. This is related to the concept of incentive compatibility. One could easily counter that policies that rely on people being rational, and that breakdown otherwise, are equally reckless. I would agree — which is why both rational agent and irrational agent approaches are useful, rather than ignoring one or both. And this is what modern economics actually does, it tends to analyse situations both in a rational framework, and one in where one or more axioms of rationality are relaxed.
In conclusion: rational choice is a useful abstraction that offers powerful bottom up modelled explanations of the emergence of many observed phenomena in economics, from the emergence of markets to corruption and rent seeking. It helps us analyse how robust a system or policy choice is, does not break down as soon as results are published, and may be of more relevance to real world markets with experienced participants than simplistic laboratory experiments using inexperienced students as recent empirical evidence shows. It should not be used comprehensively however, and in a future article I plan to detail when rational choice models might not be so useful (various financial asset markets, social and cultural phenomenon etc…), or not needed at all. Stay tuned!