Monday, November 09, 2009

New Paternalism on the Slippery Slopes, Part 3: Hyperbolic Discounting

New paternalists often rely on the phenomenon of “hyperbolic discounting” to justify their policies. Hyperbolic discounting is difficult to define in a non-mathematical way. It is sometimes summarized as excessive impatience, but that’s an over-simplification. A person with a high-but-consistent rate of time discounting would not be a hyperbolic discounter. What hyperbolic discounting really means is having inconsistent rates of time-discounting. One consequence is that a hyperbolic discounter may exhibit “time inconsistency,” a tendency to make choices and then reverse them. After explaining hyperbolic discounting (in more technical terms that I have here), Mario and I explain how paternalists have made unjustified leaps in their use of the concept (pp. 699-700):
In short, hyperbolic discounting means that people at first make long-term plans for saving or dieting but then, when the time comes to implement these plans, they succumb to the desire for short-term gratification. For the new paternalists, this type of behavior suggests an opening for paternalist intervention or correction. Examples include the previously mentioned proposal to automatically enroll people in savings plans, and to impose a sin tax (on unhealthy foods, cigarettes, and so forth) to provide additional incentive for impatient people to resist their temptations.

New paternalists claim that they are evaluating the observed behavior of the individual in terms of his own normative standard. This appears attractive until we realize that the individual has no unambiguous standard for the appropriate level of time discounting. The analytical “opening” for paternalist policy is created by the existence of an internal inconsistency of choice. But although an inconsistency does create a quandary for traditional rational-choice theory—which assumes that people have internally consistent preferences—it does not provide any grounds for choosing between the inconsistent preferences. The inconsistency of a hyperbolic discounter could be “fixed” by making him uniformly more patient ..., but it could also be “fixed” by making him uniformly less patient...

To craft new paternalist policies, it is necessary to decide the appropriate normative rate of time discounting. This matters because policies must specify the amount of money an individual is automatically signed up to save, the magnitude of a fat tax, etc. Which rate of discount is the correct one? Theory provides no answer, but the new paternalists have not hesitated to side with the more patient one. O’Donoghue and Rabin define “optimal behavior” as “that [which] maximizes long-run well-being,” where long-run well-being is associated with the more patient rate of discount. Gruber and Köszegi “take the agent’s long-run preferences as those relevant for social welfare maximization.”
To put the point slightly differently: the existence of an inconsistency does not give the new paternalists license to resolve that inconsistency however they please.

Hyperbolic discounting is difficult to work with mathematically. For that reason, behavioral economists have often used quasi-hyperbolic discounting instead. While hyperbolic discounting means a person has many (perhaps infinitely many) different rates of time discounting, quasi-hyperbolic discounting means a person has only two: one that applies when comparing any two periods in the future, and one that applies when comparing a present and future period. Once we recognize that real people tend to have a range of time discount rates, the slippery-slope potential becomes clear (p. 702):
Quasi-hyperbolic discounting makes it deceptively simple to choose the “correct” rate of discount, since there appear to be only two options. If real people actually engage in hyperbolic discounting, this implies a gradient or continuum of discount rates over time. If we assume, notwithstanding our earlier objections, that the immediate discount rate is impulsive or ill-considered, which of the longer-term rates is normatively preferable? There is nothing in the logic of new paternalism or behavioral economics that can provide an answer. We are faced with a continuum of normative possibilities. These arguments impel us to the conclusion that among the discount rates revealed in choice or planning behavior, none has a clear claim to normative superiority. Thus, the new paternalist is in a conceptual fog because his underlying standard of evaluation is unspecified. The notion of “excessive impatience” is both theoretically and empirically vague, and that means we have a gradient of possibilities. There is no clear line to resist the gradual creep of higher savings requirements, higher fat taxes, and the like.
Note the connection here to our earlier point, that gradients increase the slippery-slope risk. (As usual, full citations are available in the paper.)

Cross-posted at ThinkMarkets.

1 comment:

Eric Crampton said...

Some beautiful empirics on slippery slopes over at Chris Snowdon's blog where he shows the shifting position of Action on Smoking and Health over time. Each victory shifts the goalposts ever farther towards full prohibition, but only a bit at a time.