There is a temptation for non-economists to answer any question that puzzles economists by simply declaring "people are irrational!"
"Why are prices so much higher in Canada?"
"PEOPLE ARE IRRATIONAL!"
"What explains the equity premium puzzle?"
"PEOPLE ARE IRRATIONAL!" etc.
- It works an awful lot of the time, in that it is a good predictor of human behaviour.
- It provides models which are falsifiable.
I will explain with the use of an example. Suppose consumers are choosing between five options: A, B, C, D, E. Given what we know about consumer preferences, costs, and benefits, we predict that consumers should always choose A. However, it turns out only 80% of the time they are choosing A, and 20% of the time they are choosing B.
If we assume consumers are rational, then there must be some cost to choosing A or some benefit to choosing B that we have not accounted for. We can find candidates for that cost/benefit and test if it makes a difference. If we believe there's a factor X that is a hidden cost to choosing A, we can compare situations where factor X exists and where it doesn't. If our theory is correct, we should expect to see consumers choose A when the factor is not in play, but choose B to a much greater degree when it is in play. This provides the falsifiability we need - we can actually test our theory. We can use a number of methods - we can run laboratory experiments or we can collect data from real-world situations where the cost is both present and not present.
Contrast this with simply saying "people are irrational!". There's no insight, plus there is no falsifiability - it "explains" everything.
If we want to introduce non-rational behaviour, we need to be specific and we need our explanation to be falsifiable. In our A, B, C, D, E example, there could be explanations such as:
- The number of options overwhelm consumers, and causes them to make a mistake.
- One or more of C, D, E is providing some form of framing effect that makes B appear more attractive than it actually is.
If we have a framing effect, replacing one of 'C', 'D' or 'E' with 'F' should reduce the frequency of people choosing 'B'. If this is not occuring, then we can conclude that it is not a framing effect. We could then try removing one of 'C', 'D', or 'E' and see what that does to the frequency of people choosing 'B'. This provides us with the falsifiability we require. We also have discovered not just that people are irrational, but something about the ways people make a decision.
Caveat: I happen to agree with my former Professor Jeff Smith when he states:
Fourth, it irritates me when people equate rational behavior with an assumption of costless information processing. It seems to me that the correct way to proceed is to incorporate a cognitive budget constraint into the optimization problem. It is hardly rational to spend huge amounts of costly cognitive resources to solve some problem when a quick, cheap but slightly wrong heuristic is available Our models should reflect this and, more broadly, we should not treat clearly irrational behavior as the benchmark of rationality. This requires learning a bit of psychology and/or neuroscience in order to get the budget constraint right.
A fair bit of what might be classified as 'irrational' behaviour is simply useful heuristics that occasionally fail. As such, I think the border between rational behaviour and irrational behaviour is very blurry.
When building our models, the distinction between rational and irrational is irrelevant. What matters is that are models are a good predictor of human behaviour and they are falsifiable.