Bias on Self-Control Bias

Since Eliezer is concerned about our ability to resist modern temptations, let me summarize recent economic analysis:  Paternalism does not help people who are aware of their self-control problems, and are able to make future commitments.   To argue for paternalism regarding self-control, one has to assume we are biased to underestimate our self-control problems.  From a 2004 paper in Quarterly Journal of Economics:

We analyze the profit-maximizing contract design of firms if consumers have time-inconsistent preferences and are partially naive about it. We consider … goods with immediate costs and delayed benefits (investment goods) such as health club attendance, and goods with immediate benefits and delayed costs (leisure goods) such as credit card-financed consumption. … The predictions of the theory match the empirical contract design in the credit card, gambling, health club, life insurance, mail order, mobile phone, and vacation time-sharing industries. … time inconsistency has adverse effects on consumer welfare only if consumers are naive.

From a 2002 paper by O’Donoghue and Rabin, who have many related papers:

We investigate the role that self-control problems – modeled as time-inconsistent, present-biased preferences – and a person’s awareness of those problems might play in leading people to develop and maintain harmful addictions. Present-biased preferences create a tendency to over-consume addictive products, and awareness of future selfcontrol problems can mitigate or exacerbate this over-consumption, depending on the environment. … For realistic environments self-control problems are a plausible source of severely harmful addictions only in conjunction with some unawareness of future self-control problems.

So, are we in fact biased to underestimate our self-control problems?   If so, why isn’t it easier to just tell us we have self-control problems?   Why wouldn’t we believe such advice?

Surely some of us do underestimate our self-control problems, but curiously I can’t seem to find any papers in this (large) economics literature that consider people who overestimate their self-control problems.  This ethics paper by Tyler Cowen, however, does consider it and convince me that such situations are common.   Are we biased to assume others are biased toward too little self-control?

Added 14Nov2014: A 2011 QJE theory paper finds:

When learning [about your own self-control abilities] fails at the limit, I find that it occurs in a particular direction, namely that individuals underestimate their self-control.

GD Star Rating
loading...
Tagged as: ,
Trackback URL:
  • Foolish Jordan

    You say “To argue for paternalism regarding self-control, one has to assume we are biased to underestimate our self-control problems.”

    Is this right? What if “we” are “unbiased”, but as with all things there is variance, and so a significant fraction of the people are biased towards underestimation. Can’t we now argue that the benefit the underestimators receive from paternalism outweighs the cost to the overestimators and correct-estimators?

  • William Newman

    Robin Hanson writes “paternalism does not help people who are aware of their self-control problems, and are able to make future commitments.” Is it really as ironclad as that? Or is it perhaps that even without paternalism, a lot of self-control problems tend to get controlled spontaneously in situations where people can make future commitments, so the net effect of self-control problems can be much smaller than (and, perhaps, not even of the same sign as) the first-order effect?

    Similarly, “government-provided housing does not increase the availability of housing to the poor” would be an awfully strong statement, and it’s pretty hard to believe it doesn’t at least have some important exceptions. But “in a free market in housing, low-end accommodations arise naturally” is fairly easy for many people to believe. The weaker statement suffices to infer agendas when you encounter a pattern of people hammering on syllogisms like “the current institutions are failing to distribute the vital service of housing to the poor, therefore [let’s not wonder about draconian restrictions on housing density and prefab and whatnot, but] it is our responsibility to expand government-provided housing.”

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Foolish, people who are uncertain about something they care about generally suffer from that uncertainty, making worse decisions. And someone else with info to reduce that uncertainty can be helpful. But limiting their actions doesn’t help them. People who are uncertain and understand that the cost of underestimation outweighs the cost of overestimation will be appropriately cautious as a result.

  • michael vassar

    Haven’t we just been discussing situations where the presentation of temptations saps a limited supply of willpower. In such situations even people who correctly estimate their future biases, and even people who don’t succumb to temptations, can be made worse off by the presentation of tempting options.

    Of course, the above argument is strongest if we assume that willpower is a strictly limited resource, like money, rather than a resource that can be depleted in the short term but which if regularly taxed will tend to rise over time, like muscular endurance. However, I think that most people’s model of willpower more closely matches the latter than the former. In this case, temptations could have positive externalities for some people and negative externalities for others depending upon a person’s existing willpower stores. It might even be the case that only fairly challenging temptations increase long-term willpower, in which case as a person’s willpower store grew a given temptation would go from very harmful to moderately helpful to slightly harmful.

    Amusingly, since we have been discussing computer games, the challenges that players are faced with in many video games, especially computer RPGs, follow the same progression of effects. For a newly minted RPG character, an ogre may be certain death. As they develop, the same ogre may become dangerous but potentially rewarding, then a slight drain on short term resources but slightly rewarding with high confidence, and finally a minute drain on short term resources with no corresponding reward. Fortunately for players, video-game worlds are usually spatially separated districted into an entire spectrum of “red light districts”, and each inhabited by a fairly narrow distribution of challenges. Characters can gradually move from low to high ability levels without ever having to face challenges which would place them at serious risk given their current ability, or can take somewhat larger risks in order to advance more quickly. In finance theory the risks associated with different investment classes are similarly organized.

    Intra-national federalism and the division of the world into a number of states, to a limited degree, and licensing, to a greater degree, can replicate this organizational arrangement in the real world. People who want to avoid temptation can live in Singapore, Orthodox Jewish communities, or Salt Lake City. Those who wish to be surrounded by temptations and to cultivate balance (and Aristotelian Virtue) with occasional indulgence while attempting to carefully maintain long-term goal stability can move to New Orleans, Thailand, or Eastern Europe, or can seek admission to elite universities and acceptance in various finals clubs and “secret societies”. In this case, temptations can still threaten those who underestimate their self-control problems and are in fact only a threat to such people, (although systematic governmental campaigns of unrealistic propaganda transparently aimed at deceiving the public regarding such problems may greatly aggravate this problem by encouraging such underestimation) but how closely does this situation resemble the existing world?

    Might part of the “culture war” in the US be reasonably modeled by assuming that on the right, Americans typically advocate a temptation management strategy of minimizing temptation absolutely (at least for non-elites) while on the left a strategy of temptation acclimatization is preferred? If both sides are highly concerned with the dangers to themselves associated with their counterparts following the opposite strategy, with, for instance, left-wing Americans fearing that those naive to temptation constitute easy prey for demagogues and right-wing Americans fearing libertine abandon and degeneration as in Yeltsin’s Russia or in nightmare visions of the late counter-culture.

  • http://profile.typekey.com/sentience/ Eliezer Yudkowsky

    I think Vassar’s comment would make a good post.

  • Foolish Jordan

    Reading Robin’s reply to my comment, I guess I didn’t explain what I meant very well. Let me try again.

    “Surely some of us do underestimate our self-control problems”. Even if we grant that most people do not underestimate their problem, one can still argue that paternalism is necessary to save those few underestimators. Obviously paternalism has a cost to the over and correct-estimators, but who’s to say whether that cost to the many(?) outweights the benefit to the few(?).

    In other words, even if the population is correctly tuned, or even biased towards over-estimation as a whole, the fraction of the population which under-estimates might still need help.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Michael, within a city, neighborhood, or even a house one can make choices to limit one’s exposure to temptation. People who realize they have self-control problems can make the right such choices, and so have no need of others to force them to make good choices.

    Foolish, consider the choice to carry an umbrella with you. The risk of not carrying one is that you might be caught outside in the rain and get wet. We might pass a law requiring everyone to carry an umbrella, for the benefit of those poor people who get wet in the rain, arguing “who’s to say whether that cost to the many outweights the benefit to the few.” But a person who understands the chances of rain and the costs and benefits of carrying or not is in a good position to make the choice for themself.

  • TGGP

    I did not find Cowen’s paper very convincing. It is often asserted that self-control leads to neuroses, but I have never heard any evidence for it (and the Caplan/Szasz definition of mental illness as extreme socially unpopular but self-indulgent preferences would seem to argue against it). There is no evidence that Kant was unhappy because of his odd lifestyle, unlike the alcoholic or overeater who voices regret. Cowen seems to simply assume that because Kant went to such extreme lengths he must have been miserable.

  • http://profile.typekey.com/bayesian/ Peter McCluskey

    Robin, you admit that the ability to make future commitments is needed for your conclusion, and then proceed as if we didn’t need to ask whether that ability exists.
    There appear to be commitment mechanisms that I want but don’t get due to high transaction costs. E.g. I would like to be able to ask certain nearby stores to make it hard for me to buy ice cream from them without advance notice. Yet I suspect the costs of negotiating such an agreement are prohibitive. Did I misunderstand what you meant by “able to make future commitments”, or are there limits to this ability which complicate the analysis?

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Peter, in many places people who have gambling problems can put their name on a list of people not allowed into casinos. Similar commitment mechanisms could make it hard for you to buy ice cream at nearby stores. So, yes, my claim had a clause about needing commitment mechanisms, and yes it is possible for transaction costs to make it hard to commit. But government bans require transaction costs as well.

  • http://profile.typekey.com/sentience/ Eliezer Yudkowsky

    Second McCluskey’s objection. When it comes to explaining the current real world – rather than arguing for a policy alternative to regulation – the lack of commitment mechanisms or expensiveness of commitment mechanisms is a major factor in willpower failure. It costs money to drive to a mall that doesn’t have ice cream, and no such mall may exist.

  • Paul Gowder

    I have to agree with TGGP on Cowen’s paper — it’s really unconvincing. There’s a big difference between self-control and neurosis, and extreme examples like Kant not only are huge outliers, but are also kind of double-edged — under many standards, Kant’s life — the life of arguably the greatest philosopher since Plato was a pretty good one.

    Also, the two selves thing is really more of a metaphor than a model. And it’s a really damaging metaphor, as Cowen’s paper demonstrates. Sooner or later, economics is going to have to give way to psychology and then neuroscience on the topic of addictive behavior. Whatever it is that we may think our “preferences” or our “impulsive selves” are, there’s a mess of dopamine firing around the anterior cingulate cortex that says that we’re going to repeat activities with immediate perceived rewards, and that mess of dopamine doesn’t say that we’re going to repeat activities with long-term rewards that are disconnected from actions and stimulus. This isn’t an objection to the first two papers, but I think it is an objection to Cowen’s — you can’t inflict externalities on the impulsive self — the impulsive self is a compulsion, not a utility function! And it’s not responsible for “bursts of creativity and innovation!” It’s a reinforcement learning mechanism, not the seat of creativity! And — heaven help us — the nonexistent impulsive self engages in strategic behavior too?

    The notion of “liberating” one’s compulsive behavior is totally absurd.

    Incidentally, Robin, what incentive do casinos (or ice cream stores) have to create lists of people who aren’t allowed to give them money? (Or is the state paternalistically supplying this incentive through regulation?)

  • http://profile.typekey.com/hollerith/ Richard Hollerith

    If willpower can be trained like muscles can, then decades ago I seem to have reached the point where the muscle is as strong as biologically possible and further exertion no long increases my willpower. That’s not to say I no longer learn tricks to make my willpower go a little further. Yet it sure seems I would be more useful to the world if I had more willpower.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Paul, when and where we can see that the market would fail to provide sufficient commitment mechanisms, we could regulate to make commitment easier. For consumers who do not systematically underestimate their self-control problems, this would be preferable to paternalism limiting their options.

    Don’t you all think it would be very odd if most everyone estimated that most everyone else underestimated their self-control problems?

  • http://profile.typekey.com/hollerith/ Richard Hollerith

    Doesn’t Robin think that some of the complex functional adaptations of the human mind are very odd?

  • Paul Gowder

    Robin,

    Systematic underestimation of self-control problems doesn’t seem necessary to have a market that fails to supply sufficient commitment mechanisms. Even if all consumers accurately estimated their self-control problems, the technology might just not be there for the consumer demand to be met in a profitable way.

    Consider the ice cream stores — how can they meet consumer demand for commitment mechanisms? They could start a no customer list, but then they’re just throwing away profit. I suppose they could charge people to be on the no customer list, but the transaction costs (they’d have to implement a system to recognize customers on the list, train their employees to use it, etc.) would probably be very high, high enough that the prices would price a lot of people out of the market for self control.

    They could try and take advantage through the price structure, as per the first paper you posted. But that’s a really ad hoc (and kind of dubious) model — as far as I can tell from a quick skim, the authors consider investment goods under perfect competiton, but not leisure goods. And if any market is perfectly competitive, the ice cream store market is. So here’s the scenario for the ice cream goods market with all sophisticated consumers. Store A charges a very high price per cone. The sophisticated consumers are very happy, because they have a self-control mechanism. Store B charges a lower price per cone. The consumers, because they have low self-control, shop at store B. They know they’ll shop at store B — they have perfect knowledge of their tendency to shop at store B, but they can’t help themselves. Store A goes broke. Poor consumers.

    As far as I can tell, that first paper you posted above only works in markets where there’s a real chance of imposing low initial costs and high time-delayed/per-unit costs. Which sure ain’t the ice cream industry.

    There’s also another problem with the sophisticated addict model. It’s possible to be a sophisticated addict and still not know what to do about it. Imagine a customer who is addicted to cell phone usage but still has an instinctive and irrational aversion to high per-unit prices. He knows he’ll over-use the cellphone, but won’t demand the commitment mechanisms outlined in that paper.

    That being said, I don’t think it would be that odd for most everyone to believe most everyone else underestimates their self-control problems. Irrational, yes. Unusual? In a world where things like overconfidence bias appear in study after study? Not odd.

  • http://profile.typekey.com/robinhanson/ Robin Hanson

    Paul, I just agreed that “systematic underestimation of self-control problems [is not] necessary to have a market that fails to supply sufficient commitment mechanisms.” As I said, regulation could induce more mechanisms if the market failed to provide them.

  • Foolish Jordan

    Robin,

    (Re: a law requiring all people to carry umbrellas) “But a person who understands the chances of rain and the costs and benefits of carrying or not is in a good position to make the choice for themself.”

    I totally agree. The problem is all the dumb people who can’t figure out to carry an umbrella or not. Now we can argue whether a paternalistic regulation would be good. I suspect most people would agree that it was a bad idea since having everyone lug around umbrellas on sunny days would be ridiculous.

    All I want to say is that your one sentence: “To argue for paternalism regarding self-control, one has to assume we are biased to underestimate our self-control problems.” doesn’t seem to be strictly correct; you need to replace “we” with “some of us”. One can make a logically consistent argument in favor of paternalism based on the few dummies.