The Choice Architecture of Choice Architecture: Toward a Non-paternalistic Nudge Policy

By

David Colander and Andrew Qi Lin Chong

October 2010

MIDDLEBURY COLLEGE ECONOMICS DISCUSSION PAPER NO. 10-36

DEPARTMENT OF ECONOMICS
MIDDLEBURY COLLEGE
MIDDLEBURY, VERMONT 05753

http://www.middlebury.edu/~econ
The Choice Architecture of Choice Architecture: Toward a Non-paternalistic Nudge
Policy

David Colander and Andrew Qi Lin Chong

Middlebury College

Abstract

The paper seeks to recast the goal of nudge policy from a goal of achieving a specific
result determined by government or by behavioral economists to a goal of giving
individuals as much power as is practical to decide the choice architecture they face. We
call a nudge with such a giving individuals “power over choice mechanisms” goal a non-
paternalistic nudge policy. The goal of non-paternalistic nudge policy is not to achieve a
better result as seen by government or by behavioral economists. The goal of non-
paternalistic nudge policy is to achieve a better result as seen by the agents being nudged
as revealed through their choices of choice architectures. We argue that non-paternalistic
nudge policy fits much better with the values inherent in Classical liberalism than does
libertarian paternalistic nudge policy.

Key words: libertarian, paternalism, nudge policy, choice architecture, behavioral

economics

JEL Code: D02, D60

The Choice Architecture of Choice Architecture: Toward a Non-paternalistic Nudge
Policy

David Colander and Andrew Qi Lin Chong

Behavioral economics adds a new dimension to economic policy; it allows for the
possibility of influencing people’s decisions through a change in the choice architecture
they face. The nudge policy work of Richard Thaler and Cass Sunstein (2003a, 2003b,
2008) is probably the best known, and in their book Nudge, they have popularized the
concept of nudge policy as a description for behavioral economic policy. They argue that
their nudge policies provide a way in using insights developed in behavioral economics
into policy in a minimally intrusive way, arguing that their nudge policies meet
libertarian paternalistic criteria. Nudge policies have won widespread popular acclaim
and have been discussed in popular outlets such as The New York Times (2009), Time
(Grunwald, 2008) and Newsweek (Will, 2009) among others, which all report that nudge
policies are being incorporated into government policy. 1

Nudge policies have not been supported by all economists; in fact, it has been
attacked by some traditional economists as being just another form of government
regulation and does not improve social welfare. Nudge policy advocates have responded
in two ways. First, they argue that there are ways of distinguishing true or “normative”
preferences, and thus of insuring that nudge policies improve social welfare. (Choi,
Laibson, Madrian and Metrick, 2003; Beshears, Choi, Laibson, and Madrian, 2008).
Second, they impose limits on choice architecture policy so that the policies only
minimally affect consumers’ and businesses’ freedom to choose. This underlies their
claim that such nudge policies meet a libertarian paternalism criterion and are welfare
improving even by libertarian criteria. 2

Neither response has been fully successful in convincing critics. While most
economists agree that behavioral economists have shown that people have preferences
that are influenced by choice architecture, many are not convinced that behavioral
economists can discern “true” preferences. (See, for example, Becker and Murphy, 1988).
Similarly, the libertarian paternalism justification for nudge policies has been
unconvincing to critics. Posner, for example calls “libertarian paternalism” an oxymoron,
and sees Thaler more as a “paternalist with a velvet glove.” The reality is that only a
small set of policies that behavioral economists advocate fall under a strict definition of a
libertarian paternalistic nudge, and those that do, don’t need an elaborate philosophical

1
Sunstein has also been appointed the head of the Office of Information and Regulatory Affairs, which
conducts analysis and oversees the development and implementation of government regulation.
2
They describe libertarian paternalistic nudge policies as policies that seek “to expand or maintain freedom
of choice as far as possible.”
 The Choice Architecture of Choice Architecture

justification involving paternalism; they can be justified by common sense and classical
liberal views alone.

For example, consider the nudge concerning the default option that government
provides for people when they enroll in a government health insurance plan. Since some
choice architecture must be chosen by government, it seems reasonable that government
should make a decision it sees as best for consumers. Such nudges can be justified by
common sense. It is certainly better than the alternative of choosing a program that the
government does not believe is in the consumer’s interest. Alternatively, think of Austan
Goolsbee’s automatic filled out tax return proposal (2006). 3 This nudge is paternalistic
only in the sense that government is looking out for people’s welfare. Neither of these
nudges involve government expanding its domain over people’s choices. The same
reasoning holds in reference to nudge policies that firms can voluntarily choose to
implement. If a firm voluntarily wants to provide their employees with a savings default
option that behavioral economists believe will make its employees better off, then it
seems reasonable that it do so, and no libertarian paternalism justification is needed. 4

A much broader range of behavioral economist’s proposed policies go beyond

choice architecture involving government programs or nudges on employees imposed
voluntarily by businesses. These involve non-voluntary nudges on the part of firms, and
might be better classified as a “slight pushes.” In these cases, the firm, not the consumer
or the employee, is currently making the choice about choice architecture. Thaler and
Sunstein’s paternalistic nudge policy would require firms to change that choice
architecture to a choice architecture that behavioral economists believe will be better at
letting people’s “true” preferences be met. From the point of view of the firm, this is not
a nudge; this is a government regulation. It is in such “slight push” policies where the
issue of paternalism enters in because in order to undertake the push, government has to
assume that it, based on advice by behavioral economists, knows what is best for the
individual. Thaler and Sunstein see this paternalism as justified because the goal of the
policy is “to influence choice in a way that will make choosers better off, as judged by
themselves.” (Thaler and Sunstein, p. 8) They do not make it clear how those choices are
to be self-judged, and in a num of their proposed nudges Thaler and Sunstein have
assumed that government or behavioral economists are capable of making the judgment
about what makes the consumer better off.

For example, Thaler and Sunstein implicitly assume that people would be better
off with a choice architecture that encourages them to save more. By making this and
similar assumptions, they are replacing their views for the consumer’s views. Our
argument is that the explicit goal of nudge policy in this case should not be to encourage

3
This nudge involves the IRS providing citizens with a fully filled out form which can, but need not be
used.
4
In many cases, the firm’s interests in making a decision about choice architecture will likely differ from
that of the subgroup which is being affected by the default option, which means that they will not want to
make a change voluntarily. For example, a firm that provides with a matching grant for a savings program
might be better off if it chooses a choice architecture where fewer employees choose to enroll in it; since
the firm can seem beneficent, and those employees who strongly care can take advantage of the policy,
but the firm can hold costs down of being beneficent by choosing non-enrollment as the default option.

2
 The Choice Architecture of Choice Architecture

individuals to save more; rather it should be to give individuals the choice of whether
they want a choice architecture that is more likely to encourage to save more. This is a
subtle, but important, distinction that Thaler and Sunstein gloss over, and which underlies
the difference between our non-paternalistic and their paternalistic nudge policy.

Their glossing over it is why their nudge policy has been appropriately criticized
as not libertarian. Our solution to this is to remove the direct paternalism from the policy
by slightly modify the goal of nudge policy. Government’s goal in non-paternalistic
nudge policy is not to influence consumers to make a “right” choice, but instead is to
allow them to decide the choice architecture they desire. While this slight redefinition of
goal certainly raises issues of its own, and does not totally solve the problem, it is, in our
view, an approach more consistent with libertarianism than is the Thaler and Sunstein
exposition of nudge policy.

Nudge Policies, Classical Liberalism and Non-paternalistic Nudges

Nudge policies that are determined by government are by nature paternalistic, and
are subject to all the criticisms that can be mounted against any paternalistic policy. In a
Classical liberal framework, government interventions can only be justified as a last
resort. 5 This does, however, not mean that the classical liberal position would rule out all
government intervention; it only means that it would first search for architectures that
would allow people to make choices about choice architecture for themselves as far as is
practically possible, before it turns to government.

To see our argument, it is helpful to think of the general policy problem that
behavioral economists have posed to the theory of economic policy. By explicitly
recognizing that choice architecture influences choices, they have recognized that, when
revealed choices are influenced by choice architecture, economic policy must make
decisions about the choice architecture that best brings out people’s “correct” or “true”
preferences. Only the “correct” choice architecture will reveal the “correct” or “true”
revealed preferences, so one cannot be separated from the other. This problem of the
interconnection between choice architecture and “true” preferences has not been
addressed by economic theorists before because they have assumed away ambiguous
preferences, and thus have studied a model in which people’s revealed preferences are
their true preferences.

5
Terminology such as libertarian and paternalism are value-laden. We find it more useful to discuss
policies in relation to Classical liberalism, by which we mean the approach of economists such a John
Stuart Mill. In On Liberty, John Stuart Mill argues that “with respect to his own feelings, the most
ordinary man or woman has means of knowledge immeasurably surpassing those that can be possessed by
anyone else,” and that since government decisions “must be grounded on general assumptions,” “which
may altogether be wrong” or if right “misapplied to individual cases” by outsiders, governments should
keep out of trying to make decisions for people where these decisions do not impinge on the liberties of
others (Mill 1869). This view, in our view, captures the essence of classical liberalism—that people know
their own feelings and preferences better than anyone else, and that people have both a responsibility and
a right to make these decisions for themselves It is that policy view that we believe captures the policy
view of traditional economists today.

3
 The Choice Architecture of Choice Architecture

The natural Classical liberal theoretical answer to the problem of ambiguous

choice is to let people choose the choice architecture in which they make choices. The
problem with that answer is that those metadecisions on choice architectures would,
themselves, be made in some context, which means that people also have to make
decisions on the choice architecture of the choice architecture as well. In fact, the same
argument can be carried through ad infinitum, making the theoretical problem of who is
to determine what way to nudge an infinite regress. This infinite regress cannot be
avoided, and does not prevent a government that believes in Classical liberal ideas from
trying to give individuals as much control over it can over the choice architecture they
face. Doing precisely that is the policy that we believe is most consistent with the
philosophy Thaler and Sunstein suggest is behind nudge policy. But that policy is not
paternalistic. It is as minimally paternalistic as is practically possible, and we believe that
a policy consistent with that should be called a non-paternalistic classical liberal nudge
policy. It is achieved by making the explicit goal of the nudge policy giving people as
much choice over choice architecture as is practically possible, and not as making them
more likely to arrive at a particular preference that is determined by some outside group,
however well meaning that group may be. Only this “giving people choice over choice
architecture” goal will meet classical libertarian standards.

While implementing a non-paternalistic nudge policy will likely involve

regulation of firms, since, through inertia and current laws, firms have exerted a much
greater control over choice architecture than is consistent with a classical liberal
conception of the way a market economy should work. Implementing a non-paternalistic
nudge policy would not involve the government choosing a choice architecture. Instead,
it would involve indirect regulation of the decision process rather than direct regulation
of outcomes. For example, a non-paternalistic nudge policy would not require firms to
structure choice architecture in a certain way. Instead it would require parties relevant to
the decisions, such as firms, customers and employees, to collectively develop
architectures that take all groups preferences about architecture into account in setting up
choice architecture that directly affects them.

For firms, this policy offers more maneuverability than direct government
regulation and also allows them to use their innovation and resources to satisfy
consumers, and to differentiate themselves, in what can be regarded as competition for a
market for a new good. Consumers, of course, are the big winners. In effect, the policy
encourages a bottom-up regulation where consumers have more choice about the choice
architecture they face. From a Classical liberal perspective such “choice of choice
architecture” regulation and be justified by the lesser evil argument, since it will likely
involve less pressure for paternalistic top-down regulation of market architecture. By
giving the individual more explicit responsibility for the choice architecture he or she
faces, one reduces the pressure for paternalistic measures. It is extending Friedman’s
“free to choose” argument to freedom of choice over choice mechanisms.

Some Thoughts on Designing a Workable Choice Architecture for Choice

Architecture:

4
 The Choice Architecture of Choice Architecture

To provide an example of what we mean by non-paternalistic nudge policy and

how it contrasts with libertarian paternalistic nudge policy, let us consider the example of
cafeteria food placement that Thaler and Sunstein use to introduce the concept of nudge
policy. In their example, Carolyn, a hypothetical director of food services for a large city
school system, is making a choice about choice architecture for the students. They
present five possible options that Carolyn received as suggestions from her friends and
coworkers. These options were:

1. Arrange the food to make the students best off, all things considered.

2. Choose the food order at random.

3. Try to arrange the food to get the kids to pick the same foods they would choose
on their own.

4. Maximize the sales of the items from the suppliers that are willing to offer the
largest bribes.

5. Maximize profits, period. (Thaler and Sunstein 2008, Pg. 2)

Notice that all the options involve Carolyn making some decision for the students. They
argue that if you choose option 1 you are supporting their nudge policy. We do not
support option 1, and do not believe it meets a classical liberal criterion. Instead, our non-
paternalistic nudge policy would involve choosing option 6:

6. Provide students with relevant information, and empower or burdened them,

depending on your perspective, with the responsibility of explicitly making the
choice about choice architecture themselves.

This option conveys the essence of a non-paternalistic nudge policy; it leaves the
consumer the choice over choice architecture but designs institutional structures to get
him or her the information, and establishes an architecture that provides the student the
opportunity to reflect seriously on that choice. 6

Our new preferred option 6 is very similar to their option 3, and if option 3
included the addendum “if architectures were in place that provided relevant information
to students and provided a framework which encouraged the students to think seriously
about it”, it would be identical. It recognizes that many choices are ambiguous so without
specifying a choice architecture of the choice architecture within which students choose
on their own, option 3 is meaningless. Option 6 makes it meaningful; the goal of a non-
paternalistic nudge policy is to design choice architecture to make people think seriously
about the choices they make, not to lead them to make any specific decision.

6
Obviously, there is a paternalistic outside judgment here—the judgment that thinking seriously about
choice is good, and in principle, this judgment should in principle also be subject to consumer choice. In
practice it would probably be impossible; because of the infinite regress nature of the problem, there is no
fully neutral position.

5
 The Choice Architecture of Choice Architecture

While our proposed modification changes the explicitly stated goal of nudge
policy, we see it as a friendly amendment to nudge policy and to applied behavioral
economics work. It primarily involves a change in the way nudge policies are thought
about and presented. Many of Thaler and Sunstein’s examples of nudge policy already
are non-paternalistic nudge policies; they just don’t always make it explicit that giving
people greater freedom of choice over choice mechanisms is their primary goal. 7 Many
of their suggestions involve giving consumers greater choice and awareness over choice
architecture, especially where choice architecture can be tailored towards the individual.
That’s what makes them so appealing.

But they blend such non-paternalistic nudges with others that go beyond giving
consumers greater choice and awareness of choice architecture. Their example of food
placement was one example of a nudge that does not meet our non-paternalistic criteria.
We suspect that one reason why they blend the two is that the decision about cafeteria
food is complicated by the fact that a collective decision has to be made by students since
it is impossible to have a separate choice architecture for each student in this example.
Thus, the choice architecture has to be designed for the entire group. Determining a
workable collective choice architecture to adequately reflect the views of a group
involves all the problems of collective choice that have been discussed in the work of
economists such as Amayta Sen (1970) and Albert Hirschman (1970). Such architectures
are difficult to create, and involve serious problems and compromises by individuals
within the collective. We argue that these are precisely the type of problems that
behavioral economic policy should be concerning itself with when dealing with choice
architecture that cannot be tailored for the individual.

We don’t know what choice architecture is best in this cafeteria example, and our
goal here is not to argue for any particular choice architecture; it is simply to argue that
this is what a Classical liberal behavioral policy would involve. A classical liberal policy
would advocate that students should be allowed to make decisions about policies that
would affect them as far as is practically possible. 8

As a second example of where non-paternalistic nudge policy will differ from

Thaler’s and Sunstein’s paternalistic nudge policy, consider the U.S. Consumer Financial
Protection Agency Act of 2009. That act incorporates a variety of nudges that behavioral
economists believe would better serve customers. For example, one nudge would offer
sellers of financial instrument safe harbor from being sued if they offered a “plain
vanilla” product “designed to be read in less than three minutes.” The goal of this nudge
is to eliminate much of the fine print that currently is in financial products. A non-
paternalistic nudge policy would not start with such a specific goal. Instead, its goal

7
Similarly, some of the methods that Beshears, Choi, Laibson and Madrian (2008) present as methods by
which behavioral economists can discover people’s “true” preferences can be seen as methods of giving
people choice over choice architectures.
8
The fact that students are doing the deciding here raises the question of at what age paternalism should
end and individual freedom of choice should begin. While our view is that even young students could
benefit from being involved in the choice, the specific example is not one that we want to defend in this
paper. We use this as an example only.

6
 The Choice Architecture of Choice Architecture

would be to set up architectures through which financial consumers could work with
firms to decide what the nature of offered products would be.

Conclusion

A non-paternalistic nudge policy is not a panacea; it involves many of the

complications and difficult judgments that paternalistic nudge policies have avoided by
the assumption that government and behavioral economists can, and will have incentives
to, discern people’s “true” preferences, and will be able to set up choice architecture to
direct people to make decisions that arrive at those true preferences. Non-paternalistic
nudge policy holds that that assumption is unwarranted. There is no guarantee that
government or behavioral economists can discover people’s true preferences, and thus the
goal of non-paternalistic nudge policy is to develop choice architectures that will be more
likely to bring out people’s true preferences than the current architecture. By changing
the goal of nudge policy to empowering the consumer rather than doing what the
government thinks is best for the consumer, it makes nudge power more compatible with
classical liberalism, and thus more likely to find support among traditional economists.

References

Becker, Gary S. and Kevin M. Murphy. 1988. A Theory of Rational Addiction. The
Journal of Political Economy, 96(4): 675-700

Beshears, John, James Choi, David Laibson, and Brigitte Madrian. “How are Preferences
Revealed? NBER Working Paper 13976 Cambridge, Mass. 2008

Choi, James J. David Laibson, Brigitte C. Madrian and Andrew Metrick. The American
Economic Review, Vol. 93, No. 2, (May, 2003), pp. 180-185

Glaeser, Edward L. Paternalism and Psychology. 2006. University of Chicago Law

Review 73(2006): 133-56

Goolsbee, Austan. “The Simple Return: Reducing Americaa’s Tax Burden Through
Return0-Free Filings. Brookings Institution Web site, July 2006. ,

Grunwald, Michael. “How Obama is Using the Science of Change,” Time, April 2, 2009.

Hirschman, Albert O. 1970. Exit, Voice and Loyalty: Responses to Decline in Firms,
Organizations, and States. Cambridge, Massachusetts: Harvard University Press.

Mill, John Stuart. “Of the Limits to the Authority of Society over the Individual,” in On
Liberty, 1869, Reprinted as Chapter IV in The Collected Works of John Stuart Mill,
Volume XVIII - Essays on Politics and Society Part I. Edited by John M. Robson.
University of Toronto Press, 1977.

Posner, Richard. “Treating Financial Consumers as Consenting Adults” WSJ July 23,
2009

7
 The Choice Architecture of Choice Architecture

Sen, A.K., 1970, Collective Choice and Social Welfare, San Francisco: Holden Day

Sommer, Jeff. “When Human Beings Need a Nudge Towards Rationality,” New York
Times, February 8, 2009.

Thaler, Richard H. and Cass R. Sunstein. 2003a. Libertarian Paternalism. American

Economic Review (Papers and Proceedings) 93(2) Fall: 175-179.

Thaler, Richard H. and Cass R. Sunstein. 2003b. Libertarian Paternalism is not an

Oxymoron. The University of Chicago Law Review 70(4): 1159-1202

Thaler, Richard H. and Cass R. Sunstein. 2008. Nudge: Improving Decisions about
Health, Wealth, and Happiness. New Haven: Yale University Press.

Will, George F. “Nudge against the Fudge,” Newsweek, June 30, 2008.