Nudge and the Subjectivity of Ends

I wanted to expand on one of the criticisms of Nudge that I made below.  Namely, I criticize Thaler and Sunstein’s inability to dispense with the fundamental fact of subjective human preferences.

The authors refer to their preferred program as “libertarian paternalism”.  I suspect this term is insincere, (perhaps merely an attempt to woo libertarians into more state-friendly territory?) since Thaler, for example, seems quite willing to endorse clearly non-libertarian paternalism.  But for argument’s sake I’ll use their nomenclature.  After all, the authors’ inability to stand by their professed principles is not a rebuttal of those principles per se (though it does undermine their pat dismissal of slippery-slope arguments).

Facially, libertarian paternalism attempts to account for subjective preferences in its definition.  Paternalism, Thaler and Sunstein say, “tries to influence choices in a way that will make choosers better off, as judged by themselves” (page 5).  This neat trick hangs subjective consumer preferences as the goal-point of all rigid government interventions.  But it gets us nowhere.

Any of the chapters could be used to demonstrate the inability of libertarian paternalism to choose a “correct” direction in which to point consumers.  But the discussion of retirement planning (Chapter 6 – “Save More Tomorrow”) struck me as the most obviously futile.

Musing whether people save enough, the authors proceed accordingly (pages 108-109):

This turns out to be a complex and controversial question.  For one thing, economists do not agree about how much saving is appropriate, because they do not agree on the right level of post-retirement income.  Some economists argue that people should aim to have retirement income that is at least as high as the income enjoyed when working, because retirement years offer the opportunity for such time-intensive expensive activities as travel.  Retired people also have to worry about growing health care costs.  Others claim that retirees can use their greater time to live a more economical lifestyle: saving the money once spent on business clothes, taking the time to shop carefully and prepare meals at home, and taking advantage of senior discounts.

We do not take a strong position on this debate, but consider a few points.  It seems clear that the costs of saving too little are greater than the costs of saving too much.  There are many ways to cope with having saved too much–from retiring earlier than expected, to taking up golf, to traveling to Europe, to spoiling the grandchildren.  Coping in the opposite direction is less pleasant.  Second, we can say for sure that some people in our society are definitely saving too little—namely, those employees who are not participating at all in their retirement plan….  These folks could clearly use a nudge.

It is worth noting that the authors’ language assumes that there are objective answers to the problem of savings (“how much saving is appropriate”, “people should aim”, “the costs of saving too little are greater”, “for sure some are saving too little”).  This might be waved aside as a linguistic concession to readers perhaps unfamiliar with subjective preferences, were it not for the paternalistic proposal that followed.

Instead of taking an absolute objective position (that people should save X amount), Thaler and Sunstein pretend to abstain from judgment (“We do not take a strong position on this debate”), then immediately change gears and take a relative objective position!  Instead of arguing that people should say a specific amount, they argue that people should be subjected to incentives that will induce them to save more than they are currently saving.

This conclusion is absurd.  The authors are rightly unwilling to specify some “correct amount” of retirement savings.  But absent any objective reference point, it is impossible to say whether people as a whole are spending too little.  Relative position is impossible to determine except in reference to an objective point.

Thaler and Sunstein attempt to pad their argument by referencing employer matching programs (pages 109-110):

For example, a common plan feature is that the employer will match 50 percent of the employee’s contributions up to some threshold, such as 6 percent of salary.

This match is virtually free money….

Some older American workers are also turning down “free money”….  For such employees, joining the plan is a sure profit opportunity because they can join, then immediately withdraw their contributions without any penalty, yet keep the employer match.  Nonetheless, a study finds that up to 40 percent of eligible workers either do not join the plan at all or do not save enough to get the full match.

The existence of this “free money” cannot, of course, justify further government intervention in savings.  Any aggregate windfall will vanish, like a mirage, if people attempt to capture it en masse.  As any economist will tell you, wages are determined by productivity.  Whether or not employees collect “free money” is a salient factor in wages.  If more people claim the “free money”, employers must lower wages for the elderly as a group to keep them equal to (an unchanged level of) productivity.

Moreover, these “free money” programs likely only exist because governments have strong-armed businesses into offering them!  Thaler and Sunstein blithely accept this legislative bullying – and nearly every regulation mentioned in the book.  But this begs the very question that the authors leave unanswered in the first place – how much should people be saving?

Libertarian paternalism ought to reject government decisions to subsidize savers at the greater expense of consumers, if these decisions are based merely on the preferences of government bureaucrats (and we have not yet established other grounds for these programs).  The authors waste little time (for once) objecting to our government’s (non-libertarian paternalistic?) starting point in marriage laws, when they conflict with the obvious progressive goal of gay rights.  Yet more, not fewer, regulations are somehow warranted in retirement laws by consumers’ “irrational” decision not to adequately adjust to pre-existing costly (un-libertarian) government regulations that the authors never bother to justify in the first place.

The only reference Thaler and Sunstein make to consumers’ retirement preferences, beyond their central appeal to the emotionally “obvious”, is a halfhearted reference to survey data.  The authors elevate thin, aggregated poll responses into a demonstrated goal of all employees (page 108):

For what it’s worth, many employees say that they “should” be saving more….  It is easy to say that you “should” be doing many good things—dieting, exercising, spending more time with your children—and people’s actions may tell us more than their words….  But such statements are not meaningless or random.  Many people announce an intention to eat less and exercise more next year, but few say they hope to smoke more next year or watch more sitcom reruns.  We interpret the statement “I should be saving (or dieting, or exercising) more” to imply that people would be open to strategies that would help them achieve these goals.

Of course, the authors do much more than “offer strategies”.  They use consequence-free surveys conducted in artificial environments as a mandate for costly and inescapable regulation that will affect everyone.  And they could not do otherwise – nothing but thin survey data could serve as an alternative to human action in determining subjective human preferences.  And a program of paternalism must reject human action as a valid basis of demonstrated preference.

This is the fundamental dilemma facing libertarian paternalists.  The paternalist must figure out what people want without reference to what people actually choose to do.  What method could possibly suffice for this?  The planner could suppose himself to simply know all preferences by virtue of his common humanity.  I expect that any program of well-intentioned paternalism will eventually regress to this level – and the authors seem to concede this point.  But a shared human perspective does not, in itself, put a paternalist in a superior position to those he is regulating.  And, moreover, this sort of justification does not take seriously Thaler and Sunstein’s injunction that regulators make consumers better off as judged by themselves.

The only possible alternative to observed action (i.e., the market) is the survey.  It is perhaps not surprising that leftist intellectuals will favor programs that transfer authority from the market to the university psych study.  But whenever I see these proposals, I begin to suspect, as Megan McArdle likes to say, that somebody missed the socialist calculation debates.  The difficulties faced in determining preferences by poll are intractable.

As a warm-up, we have the problem of the contrapositive counterfactual.  Any group of retiring consumers evaluating their position for a survey are ignorant of the opinion that they would have in some conceivable alternative position.  For example, consumers who save for their retirement cannot know the relative life satisfaction they would have if they had not saved for retirement.  Conversely, those who do not save are ignorant of the satisfaction that they would experience in the hypothetical alternative where they had saved.  Even if consumers in both situations agreed that one of the situations was preferable (for example, if both the spendthrift and the miser agreed, upon retirement, that being miserly were better) this would simply be agreement between two ignorant parties.  Their agreement would lack the objectivity reserved for the impossible individual who simultaneously experienced both alternatives (whose own judgment could yet be assailed as dynamically inconsistent!).

In any event, if the choice is made by those who have already retired, it cannot represent the preferences of those who have not yet retired.  Thaler and Sunstein avoid this and the above problem by reporting prospective, rather than retrospective, survey data.  Individuals choosing for their future remain ignorant of their actual future satisfaction, but they at least avoid the bias of uneven ignorance of outcomes inherent to the retiring, retrospective choosers.  In this sense, the surveys are more like human action and the market.  But insurmountable difficulty remains.

Nudge recognizes that the answers that individuals give in a survey are strongly influenced by the circumstances in which the survey is conducted.  The authors’ paternalistic program is, in fact, supposedly justified by this inconsistency.  On page 24, the authors note that survey respondents rated their overall happiness lower if they were first asked about their dating lives.  On page 36, they note that decisions and assessments are strongly influenced by “framing effects” – e.g. whether a glass is portrayed as half-full or half-empty.

So what environment is most appropriate for conducting surveys?  Should they be result-oriented?  Should the government ask questions about savings in environments aimed at encouraging support for retirement programs, as the authors encourage corporations to do throughout the book?  But this would simply assume as objective the preference that surveys are supposed to determine.  Should surveys occur in some sort of “neutral” setting, with the pollee isolated, listing and ranking preferences as they occur to him, without any input from the pollers?  Aside from the difficulty there would be in aggregating this sort of data, why would we expect a complete and careful result from this low-pressure survey?  Why would we even expect a lack of bias – wouldn’t the individual simply bring in with him whatever biases that clung to him the particular day of the survey?

Is the idea of a “neutral” polling methodology coherent?  Thaler and Sunstein give no indication that it is.  So what exactly are they trying to accomplish?  Do mutable survey results give them license to simply cherry pick the one whose results they like and then “nudge” people in a direction that was never really anything other than their own subjective preference?  If surveys ever did manage to become truly objective and comprehensive, they would likely need to be as long and onerous as those that interwar socialists of the economic calculation debates imagined would replace markets as a preference transmitting device in the Soviets’ communist paradise.

Lest this piece become a novella, I want to comment on only one more problematic feature of surveys.  There is one inescapable difference between decisions made for a psych study and decisions made in the market.  You might call it a “framing effect”, and it is one that seriously undermines the authority of all survey results.  Survey respondents do not conceive of their answers to surveys as being actual life choices. The most fundamental difference between surveys and action makes the former an unacceptable substitute for the latter.  Paternalists can only make their surveys like actual choice by discarding the surveys altogether and just permitting the market to run its course.  Instead, Thaler and Sunstein seem ready to tilt the playing field whenever a survey yields a result different from the market, as if unaware that this might simply reveal a deficiency in surveys.

A final, perhaps ill-placed, note.  Some readers may object to my use of the terms “interference” and “regulation” throughout this piece.  Thaler and Sunstein pretend or imply, for most of Nudge, that their interventions are “costless” or “nearly costless” (“nearly” being another arbitrary term open to interpretation by – who else? – the paternalist).  But they confess as early as page 8, in a footnote, that their interventions are not actually costless:

Alert readers will notice that incentives can come in different forms.  If steps are taken to increase people’s cognitive effort—as by placing fruit at eye level and candy in a more obscure place—it might be said that the “cost” of choosing candy is increased.

This puts the matter too tentatively.  It must be said that increasing cognitive effort increases costs!  All costs – from hard physical labor to selecting food in a cafeteria are ultimately experienced by individuals in terms of “cognitive effort”.  Imposing costs in terms that are not monetized obscures but does not reduce them.  It does, of course, deny citizens a quite useful measuring heuristic: prices.  In World War II, for example, governments hid the monetary costs of maintaining large standing armies by drafting millions of men, instead of hiring them at the market wage.  In advocating for the imposition of non-monetary costs, Thaler and Sunstein merely join a storied bureaucratic tradition of dissimulation.


  • Jacob

    Yes, human preferences are subjective. And yes, construction of a choice architecture implies some manipulation of preferences. But the point is that a choice architecture exists anyway–you cannot avoid having some sort of imposition. As such, we might as well think carefully about how we structure choices.

    Furthermore, in a libertarian paternalist framework there is still the same amount of freedom to choose. Say somebody sets the default so that you participate in your firm’s retirement plan. You can always opt out if you want.

    In other words, this subjectivity is a reason paternalism is bad, but the libertarian aspect renders it unimportant for this analysis.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.