I haven't written much on behavioural, partly because dissertation work overshadowed my class last semester and partly because there are much more interesting things to spend your time on. This critique, however, is quite startling and seriously challenges some solutions to the neat biases behavioural economists refer to.
Now, some background: behavioural economics is this trendy wing of economics (bordering on psychology) that assess some of the foundational microeconomic assumptions economists often make about human behaviour (optimisation, expected utility, consistency etc). Since the 1970s, distinguished economist Daniel Kahneman and cognitive psychologist Amos Tversky have performed countless of experiments where their subjects failed to act in accordance with such assumptions: they acted "irrationally". Kahneman & Tversky gradually incorporated those many experiments into a coherent theory able to explain such behaviour, nowadays known as Prospect Theory – an alternative approach to the assumptions underlying your average microeconomic class. Kahneman eventually received the Nobel Prize in 2002, for "having integrated insights from psychological research into economic science".
Behavioural economics has been very influential in justifying various kinds of state interventions and limitation based on the sub-optimal behaviour of individuals. The idea is to use insights from behavioural econ to improve such sub-optimal outcomes, tweaking individual behaviour by resticting, inducing, nudging or otherwise influence behaviour. If people don't act in accordance with their own good, we can make them do so; in other words, paternalism found great defenders in the work of behavioural economists.
Along comes Klick & Mitchell with their superb and comprehensive critique of the blind spots of such thinking. They raise a whole range of problems. Firstly, it is uncertain to what extent experiments designed to trick people generalise into actual behaviour in the real world – not to mention that the test subjects are probably the least representative group in society (Western, Educated, Industrialised, Rich, Democratic, = WEIRD). But the most substantive point comes in realising learning, informational costs, cognitive efforts and moral hazards. Let me explain this by doing what made Prospect Theory successful: refuting economic thinking in its own language – equations (they don't bite, ok!)
The naive paternalist/politician reading Kahneman sees that by making people act in a certain way, we can improve overall outcomes. If we nudge, force, coerce, regulate or tax people into line, we're all better off, sometimes even Pareto-style better off. But after reading Klick & Mitchell's brilliant response one is much less likely to make that mental mistake; informational gathering is costly, even for allegedly benevolent governments, costs that must be weighted against the marginal "gain" by moving people from sub-optimal to optimal positions. Moreover, there's a time-dimension here: people gain information by making mistakes, "learning-by-doing", which means that governmental action to protect or restrict action, reduces the informational intake in one period, making the distance from "optimal" even further in the next. In other words, there are unintended consequences with government interventions. Big surprise.
This is how they did it:
where e = effort, UH and UL are high and low utility (where we are and where paternalists ideally want us to be), p = probability of choosing one or the other, and C = cost of obtaining information and exerting effort.
Standard optimisation implies that the usefulness of paternalist policies depends on three things, rather than simply observing that people act irrationally, ála Kahneman and Tversky:
- The productivity of effort, that is by how much p (=probability of choosing the better outcome) increases through effort; in other words, if very costly informational search only slightly improves my ability of choosing the higher utility, it may not be worth it.
- The spread between high and low outcomes, i.e. the difference between UH and UL: if we're not that far away from an "optimal outcome", correcting that behaviour may be more costly than the benefit it yields.
- if the cost of effort, C, is prohibitively high, then none of the other variables matter: It's simply not worth correcting for suboptimality.
As long as search effort is costly, an individual will likely accept an expected outcome in which she sometimes chooses incorrectly because the marginal value of increased search, after some point, will not exceed the marginal cost of the effort required by the search. (p. 1643)
Moreover, insofar as spending C in this period teaches me valuable lessons (information about this decision, insights that may generalise to other cases), there is a "invest now-yield benefits later" dynamic going on. If governments restrict or prevent certain behaviours in accordance with the belief that we're off-optimality, that investment cannot take place, and the problem is likely worse the next time around. Moral hazards, and, Klick & Mitchell adds, Cognitive Hazard play a role.
The fact that we can point to isolated observations of irrationality in WEIRD student groups doesn't have as dire consequences for orthodox microeconomics as proponents of behavioural would like to point out. Costs matter, the value of information gathering and feedback mechanisms matters. And sometimes suboptimality can be optimal. Or at least as close as we can practically get.
No comments:
Post a Comment