Content uploaded by Ayala Arad
Author content
All content in this area was uploaded by Ayala Arad on Dec 15, 2017
Content may be subject to copyright.
!
1!
The People’s Perspective on Libertarian-Paternalistic Policies
Ayala Arad
Coller School of Management, Tel Aviv University
Ariel Rubinstein
School of Economics, Tel Aviv University and Department of Economics, New York University
October 2017
Abstract
The non-material welfare effects of libertarian-paternalistic (soft) government interventions are
examined in a series of online experiments carried out in three countries. Both standard and new
methods are used to elicit attitudes towards soft interventions in various hypothetical scenarios.
The majority of the participants accepted such government interventions. However, a substantial
proportion opposed them and some refused to make the choice the government was encouraging,
even if they would have made that choice in the absence of the intervention. Furthermore, they
prefer that the government only provide information to help the public make the right choice rather
than a more effective choice architecture intervention. We find that the opposition to soft
interventions is primarily due to concerns about manipulation and the fear of a “slippery slope” to
interventions that lack a clear consensus among the public. Opposition to soft interventions is
reduced when they are carried out by employers rather than the government.
Keywords: choice architecture, libertarian paternalism, nudge, soft interventions
!
2!
1. Introduction
The goal of the paper is to provide a deeper understanding of the public’s attitude toward the soft
government interventions advocated by supporters of libertarian paternalism. We carry out a series
of online experiments to explore the existence of a negative attitude towards soft interventions,
examine the aspects of an intervention that lead to it and discuss its welfare implications.
Government interventions in personal decision making can be classified into three types:
(i) hard interventions - which change the choice set of the individuals (such as taxes or the
prohibition of particular choices); (ii) soft interventions - which preserve the set of choices but
manipulate its presentation in order to nudge the individual in a particular direction (such as setting
a default option); and (iii) informational interventions - which provide an individual with the
information needed to make an informed decision (such as by means of accessible smartphone
apps). Libertarian paternalism argues for the use of soft interventions by the government (or some
other institution) in an individual’s decision making in an attempt to improve public welfare (see,
for example, Camerer et al., 2003, Sunstein, 2014 and Thaler and Sunstein, 2003). This approach
seeks to preserve the principle of freedom of choice and to avoid coercing the public into adopting
some specific behavior. Nonetheless, soft interventions steer individuals toward the “right”
decisions by making use of insights from psychology, marketing and behavioral economics.1
The libertarian-paternalistic approach has been enthusiastically adopted by some
governments but has also attracted intense criticism from some corners on philosophical grounds.2
We do not intend to provide philosophical arguments for or against the libertarian paternalistic
approach, but rather we seek to experimentally determine whether a significant portion of the
population agree with the goals of a soft intervention, yet resist it nevertheless and which aspects
of such interventions bother the public the most.
We report the findings of three online experiments conducted among students at six
universities in Germany, Israel and the US. We present the results from the three countries side by
1 Frequently discussed examples of such policies include: (i) placing healthy foods at eye level and unhealthy foods
in less noticeable locations in cafeterias (see, for example, Rozin et al., 2011 and Thaler and Sunstein, 2009); (ii)
setting a particular rate of retirement saving as a “default option”, in order to increase the rate of saving (see, for
example, Choi et al., 2004 and Madrian and Shea, 2001).
2 See, in particular, Fumagalli (2016), Glaeser (2006), Grüne-Yanoff (2012), Hausman and Welch (2010),
Loewenstein and Haisley (2008), Rebonato (2012), Rizzo and Whitman (2009), Selinger and Whyte (2011), Vallgårda
(2012) and White (2013). An extensive response to these critics is provided by Sunstein (2015a).!
!
3!
side and focus on patterns they have in common rather than the differences between them. The
experiments elicited individuals’ choices and attitudes in hypothetical scenarios that involve
government interventions in the areas of saving and the consumption of healthy food. The
experiments involve three types of soft interventions: a default opt-in, changing the order of the
presentation of alternatives and the use of background music. The informational interventions
involved the provision of information through smartphone apps, educational campaigns and green
labeling of healthy products. This variety of interventions reinforces the robustness of the findings.
As explained in detail below, the study uses both the standard experimental approach and a novel
experimental design to measure individuals' attitudes towards interventions: given two
interventions with the same goal, we measure an individual’s preference for one intervention
method over another by means of the loss of effectiveness (in percent) that he is willing to tolerate
in order for his preferred type of intervention method to be adopted.
The results of the three experiments provide us with the following five insights:
(1) Negative attitude towards soft interventions. Study 1 and 3 utilized the standard approach
which involved asking participants to express their attitude (from very positive to very negative)
toward two soft interventions: setting a default saving rate with the goal of increasing personal
saving and using background music in cafeterias to encourage the consumption of healthy food. A
majority of participants regarded the soft government interventions positively. However, a
significant proportion, who were in favor of the interventions’ goals, nonetheless expressed a
negative attitude towards such interventions. The proportion varied from 26% to 53% depending
on the country and the specific intervention.
(2) Relative preference for informational interventions. The preference of one intervention
method over another was measured using a methodological approach that is novel to the literature
on behavior change and public policy. In Study 2 and 3, participants were asked to express their
subjective tradeoff between the effectiveness of an intervention and the desirability of its method.
That is, if a participant prefers method A to method B, we elicit the degree of effectiveness that
the participant is willing to sacrifice in order that A be adopted rather than B. In the comparison,
the interventions’ goals are fixed and their effectiveness is given, such that participants are induced
to focus on comparing the methods themselves. We find this tool to be more suitable for eliciting
attitudes towards public policies than other commonly used measures (such as
!
4!
approval/disapproval or rating of support according to some scale), since in real life the evaluation
of a policy is almost always in comparison to alternatives and is sensitive to the policy’s estimated
effectiveness.
In both studies, we found that a significant proportion of the participants prefer that the
government only provide information to the public rather than implementing a more effective soft
intervention (such as one that involves choice architecture): 21%-37% strictly prefer that the
government introduce an app that provides information on healthy nutrition over requiring that the
items on restaurant menus be ordered from healthy to less healthy; and 55%-68% strictly prefer
labeling healthy food over the use of background music in cafeterias to encourage the consumption
of healthy food.
(3) Psychological reactance. In Study 1, participants were asked to consider a hypothetical
arrangement in which instead of having to decide on one’s saving rate every month, there would
be a default rate. In one treatment, the arrangement is imposed by the government and participants
can opt-out; in another, participants choose whether or not to join the arrangement without the
government being involved. Although the arrangement itself is perceived as desirable by the vast
majority of participants, there is a not insignificant proportion of individuals who opt-out of the
arrangement even though they would have joined it had there not been any government
involvement. The results may reflect psychological reactance to the pressure to behave in a
particular manner (see Brehm, 1966; Wortman and Brehm, 1975). This result differs from those
of Loewenstein et al. (2015) who found that alerting the individual to the existence and intention
of a default option did not change his behavior.
Decisions in a hypothetical scenario do not necessarily predict what people will do in real-
life situations but they nonetheless capture their wishes and attitudes and hence are relevant for
the evaluation of their non-material welfare. If the hypothetical choices reflect real choice
behavior, then the results hint at a possibly significant material welfare loss for individuals who
contrarily do not make the choice being encouraged as a protest against the government
intervention.
!
5!
(4) Determinants of a negative attitude towards soft interventions. A number of reasons for
opposition to the “nudge approach” have been extensively discussed in the literature. We classify
these reasons into the following five categories:
1. Disagreement with the goal. The designers of libertarian-paternalistic policies attempt
to increase public welfare, or at least what they perceive it to be. However, even in cases where
one would expect unanimous support of the intervention, there might be people who disagree with
the intervention’s goal.
2. Private decisions are not the government’s business. People should have autonomy in
any private decision that does not involve externalities.3 Thus, issues such as nutrition are none of
the government’s business. Furthermore, policy makers are in the end only human and the
possibility that they will fail to make the correct decision on behalf of an individual is no less likely
than the individual failing to do so on his own.
3. Concerns about manipulation. In many soft interventions, the planner manipulates the
individual without his knowledge and certainly without his consent. Even if people are informed
of the technical details of the intervention, they may not be aware of the relevant psychological
phenomena and might not internalize the potential effect of the intervention on their decision
making. It is inappropriate for the government to influence the decision making of individuals
without their knowledge.
4. Concerns about neglect of personal responsibility. The interventions both free
individuals from taking personal responsibility for their choices and encourage the development
of “fragmented selves” who become dependent on the authorities for guidance.
5. Concerns about a slippery slope. The success of soft interventions and their legitimacy
due to their liberal appearance may pave the way for the use of such interventions for non-
legitimate purposes or to achieve goals that lack a consensus.
In Study 3, we investigated the motives behind the negative attitude towards soft
interventions found among a significant proportion of the participants. Remarkably, there was
almost unanimity regarding the goal of the intervention yet there was considerable variation in the
3!Our experiments involve interventions that attempt to influence taste-based decisions that are not subject to any
direct externalities. When an intervention is aimed at preserving a well-accepted social norm (such as painted
footprints leading to trash cans), it is less likely to be perceived negatively and the welfare implications are less clear.
!
6!
extent of agreement with the four other factors listed above. After screening out the few individuals
who did not agree with the goal, we examined whether these four factors explain overall attitudes
towards the soft intervention. This was accomplished by looking at the correlations between the
extent of agreement with each factor and the overall attitude towards the intervention and by
estimating a regression of each factor’s effect on the participant’s attitude, while controlling for
the effect of the other factors.
The main findings indicate that concerns about manipulation4 and a slippery slope are
largely responsible for opposition to the interventions. These concerns are also present among
many of the participants who supported the specific soft intervention.
(5) Relative preferences for employer intervention. We focus primarily on the attitude towards
government interventions, but that attitude may change if someone other than the government,
such as an employer, carries it out. We therefore examined a soft intervention carried out by
employers to encourage their employees to eat healthier food. In Study 3, we found that both in
Israel and the US, the resistance to such a soft intervention is reduced if it is carried out by
employers rather than the government, whereas in Germany it made no difference.
We believe that an individual’s welfare is not determined solely by the material
consequences of an action but also by its non-material aspects. In particular, people often care
about the method used to achieve the material outcome.5 The results indicate that a narrow majority
of the participants accept soft government interventions but that nonetheless there is a considerable
portion that oppose soft interventions and would prefer the use of informational interventions.
Furthermore, they would be ready to sacrifice effectiveness in order for their preferences to be
realized. These findings should be taken into account when considering the overall welfare effect
of introducing a soft intervention and when choosing between informational and choice-
4 Concerns about manipulation were also found in Felsen et al. (2013) who considered interventions carried out by
employers in order to encourage some desirable behavior (such as exercising) among its employees. They find that
participants were more favorable to an intervention which activates a conscious process (such as directing the workers
to the exercise room) than one which activates an unconscious process (such as planning the building so that the
exercise room is located on the way to the cafeteria).
5 This point has received some experimental support from Chlaß et al. (2013), Eliaz and Rubinstein (2014) and Wailoo
and Anand (2005), among others.
!
7!
architecture interventions. In the concluding section, we return to the welfare implications of the
findings for public policy issues.
Comment on our methodological contribution. Several authors have investigated public
attitudes towards soft and informational interventions. Of particular interest are Hagman et al.
(2015), Sunstein (2015b), Tannenbaum et al. (2017), and Jung and Mellers (2016).6 Overall, our
results are not inconsistent with theirs though they differ in several key ways:
(1) We demonstrate that a significant proportion of participants act as if they protest against
the intervention by contrarily not making the choice being encouraged by the government.
(2) In other studies, participants’ attitudes towards a policy may have been affected by their
evaluation of the policy’s effectiveness. In contrast, we had participants compare between two
policies given various hypothetical specifications of effectiveness. This made it possible to
measure the attitude towards a policy using the tradeoff between the subjective desirability of the
intervention method and its effectiveness.
(3) Most surveys simply ask participants to state whether they support an intervention or
alternatively to what extent they support it. Often, the answers are likely to reflect the extent of
agreement with the intervention’s goal rather than its method. Furthermore, a method of
intervention that encourages positive behavior is likely to be perceived more favorably if presented
on its own rather than in comparison to another possible method of intervention with the same
goal.
It is the attitude towards the method (rather than the goals) that should be the focus in
evaluating libertarian paternalism. The approach adopted in Study 2 and 3, i.e. of comparing
between two interventions with the same goal, allows participants to express their attitude towards
6 Hagman et al. (2015) carried out surveys in Sweden and the US and report general acceptance of nudges; however,
a majority of respondents also found the nudges to be intrusive. Sunstein (2015b) surveyed a representative sample in
the US and elicited their views on informational and soft interventions. He found that a class of popular nudges was
supported by a majority of the subjects when asked whether they approve or disapprove of the intervention. The
proportion of subjects that opposed nudges in his survey is somewhat smaller than in our sample. Tannenbaum et al.
(2017) found that in the US the political affiliation of the initiator of an intervention (i.e. Republican or Democratic)
and whether its goal is part of a liberal or conservative agenda largely determines whether an individual supports it.
Jung and Mellers (2016) studied attitudes in the US toward a variety of interventions, some of which utilized System
I (such as setting defaults) and others which utilized System II (such as providing information). The results indicated
that people tend to view most System II nudges as being more effective and also more acceptable.
!
8!
an intervention independently of their support for its goal.
(4) We investigate the motives behind negative attitudes towards soft interventions and the
presence of these sentiments even among supporters of the interventions.
(5) We examine whether opposition to an intervention is reduced if it is carried out by an
employer rather than the government.
2. Study 1: Reactance to a soft intervention
2.1 Method
The experiment was carried out online among undergraduate students at six universities: two in
Germany (University of Hamburg and University of Manheim), two in Israel (Ben Gurion
University and Tel Aviv University) and two in the US (New York University and Ohio State
University). The proportions of men and women were similar. In order to encourage participation,
one out of every 20 participants was randomly selected to receive a fixed amount of money
(independent of their responses) that was roughly equivalent to $30. Invitations to participate were
sent by email and those who agreed were randomly assigned to one of the treatments T1, T2 or T3
described below. The number of participants in Germany was 345 (96, 117, and 132 in T1, T2 and
T3, respectively), in Israel 462 (159, 150 and 153) and in the US 310 (111, 102 and 97).
In each treatment, participants responded to a sequence of questions about their attitude
towards hypothetical government interventions. The participants were asked to imagine that they
work for a firm in which employees had the opportunity to deposit between 0% and 8% of their
salary in a special saving account that offers an attractive interest rate, with the restriction that the
money would only be available after 10 years. Initially, the employees had to choose their rate of
saving every month. A new arrangement was then offered which involves a default savings
deduction from one's salary, with the goal of increasing the employees’ rate of saving.7
In T1, which did not involve any government intervention, participants chose between
determining the saving rate each month and joining (opting-in to) the default saving arrangement
7!We use the case of depositing into a long-term saving account rather than the commonly discussed case of saving
for retirement since we felt the former would be easier for the students to understand.!
!
9!
in which 8% of one's salary is automatically deducted and deposited in the saving account, unless
the employee provides a one-time instruction to cancel the arrangement (and then goes back to
determining the rate again every month). In T3, the government requires employers to set 8% as a
default saving rate for their employees and the participants need to decide whether or not to opt-
out of the arrangement. In both T1 and T3, the second question examines how the participant feels
(on a scale of 1 - very positive to 4 - very negative) about the government imposing a default rate
of saving. T2 is similar to T3 except that the order of the questions is reversed. In all screens
presented to the participants, the order of the possible answers was determined randomly (see
Appendix A for the appearance of the screens). Table 1 summarizes the structure of Study 1:
T1
T2
T3
Screen 1
Would you opt-in independently?
(without the intervention)
Attitude towards the soft
government intervention
Would you opt-out after the
soft government intervention?
Screen 2
Attitude towards the soft
government intervention
Would you opt-out after the
soft government intervention?
Attitude towards the soft
government intervention
Table 1. Description of the treatments in Study 1. The soft government intervention requires employers to set 8% as
a default saving rate for their employees.
In the real world, individuals are not necessarily aware that the choice architect is
exploiting their psychological tendencies in order to influence their behavior. The objective of
Study 1 was to understand how individuals react when they are made aware of the motive behind
the government soft intervention. Its design makes it possible to demonstrate two points:
(1)!Imposing a default (T2 and T3) may be less likely to achieve the intended result than when no
default is imposed (T1). This may be due to psychological reactance (Brehm, 1966) which
leads some participants to opt-out of the arrangement even though they would have joined it
had there been no intervention.
(2)!There may be a reduction of welfare among individuals who intended to choose the default
option in any case since they are cognizant of the attempt to influence their behavior and hence
will be less content. This will be reflected in a negative attitude towards the government
intervention among individuals who opt-in to the arrangement when it is simply offered to
them (T1).
!
10!
To conclude, we test (a) whether there is a significant proportion of participants who express
a negative attitude towards the soft government intervention in the various treatments, and (b)
whether the proportion of participants who opt-in in T1 is greater than the proportion of
participants who do not opt-out in T2 and T3.
2.2 Results
The results presented below exclude 5% of the participants who spent the least time reading the
first screen in the experiment, which included the description of the hypothetical scenario.
Result 1: Negative attitude towards the soft government intervention
The results indicate that although a majority of the participants (except in Germany) have a positive
attitude towards the soft intervention, a sizable proportion (28% in Israel, 42% in the US and 53%
in Germany) have a negative attitude.8
The proportions of participants with a negative attitude towards the automatic opt-in
intervention in the three treatments are presented in Table 2. The data for T2 and T3 was merged
since we did not find any order effect. We also pooled the participants who expressed “negative”
and “very negative” attitudes as well as those who expressed “positive” and “very positive”
attitudes since less than 20% of the participants expressed the two extreme positions.
Negative attitude
Germany
Israel
USA
Among those who opt-in independently in T1
59%
26%
42%
Among those who do not opt-in independently in T1
85%
64%
74%
Overall in T1
66%
34%
50%
Overall in T2 and T3
48%
25%
38%
Table 2. Proportions of participants with a negative attitude towards the government intervention in Study 1.
The attitudes in T1 are somewhat more negative than in T2 and T3, probably because in T1
participants were more aware that instead of the government soft intervention the individuals could
8 The 95% confidence intervals are (24%, 32%) in Israel, (36%, 48%) in the US and (48%, 58%) in Germany.
!
11!
have simply been asked whether they are interested in the default arrangement (as was done in the
first screen of T1).
As expected, a negative attitude is more common among participants who have stated in
T1 that they would not opt-in to the arrangement if offered to do so independently (without
government involvement) than among those who stated they would opt-in. However, even among
participants who said they would opt-in to the arrangement independently, a considerable
proportion felt negatively about the soft government intervention of an automatic opt-in. This
group was smallest in Israel (26%) and largest in Germany (59%).
Result 2: Government involvement reduces an intervention's effectiveness
The results suggest the existence of antagonism toward the government soft intervention among
some individuals, which leads them (perhaps unconsciously) to resist the intervention. In other
words, they opt-out of the arrangement even though they would have joined it had they been
offered to do so without government involvement. It is also possible that these individuals are
using their stated choice as a protest. This is supported by the finding that the percentage of
participants who chose not to opt-out when they are aware of the government intervention is
smaller than the percentage of those who chose to opt-in in the case that the government is not
involved (Table 3). The gap in Israel is small (9%-11%) but statistically significant (𝜒"(1)=4.37,
p<0.05). The gap in the US is larger (13%-18%) and that in Germany is the largest (19%-32%),
and in both cases highly statistically significant (𝜒"(1)=7.74, p<0.01 and 𝜒"(1)=16.52, p<0.01,
respectively).
Thus, the default arrangement is perceived as attractive when the government is not
involved, which is reflected in the high percentage of those choosing to opt-in in T1. Nevertheless,
for a not insignificant proportion of participants the arrangement becomes undesirable when the
government makes it the default.
Germany
Israel
USA
Opting-in (T1)
74%
79%
78%
Not opting-out (T2)
42%
70%
60%
Not opting-out (T3)
55%
68%
65%
Table 3. Opting-in with no government involvement vs. not opting-out with government involvement in Study 1.
!
12!
3. Study 2: The tradeoff between desirability and effectiveness of interventions
3.1 Method
Participants in this experiment consisted of undergraduate students from the same six universities
as in Study 1. The proportions of men and women were similar. The participants were recruited
and paid as in Study 1. The study consisted of five randomly assigned treatments T1, T2, T3, T4
and T5. The number of participants in Germany was 600 (120, 135, 96, 117 and 132 in the five
treatments, respectively), in Israel 740 (136, 142, 159, 150 and 153) and in the US 549 (121, 118,
111, 102 and 97). In this section, we report on the main treatments - T1 and T2, while T3, T4 and
T5, which check the robustness of the results, are reported in Appendix B.
In each treatment, participants were presented with a hypothetical scenario in which the
government considers various interventions intended to improve the public’s eating habits and in
particular decrease the consumption of fatty foods. We considered four government actions
(interventions) in T1 and T2:
(1) Prohibiting the serving of extremely fatty food in restaurants on Wednesdays.
(2) Imposing a Tax on extremely fatty food served in restaurants, which will be added to the price
of a meal (and passed on to the government).
(3) Requiring restaurants to Order the items on a menu from healthiest to unhealthiest.
(4) Providing information through a smartphone App to be created by the government, which will
make available information on the nutritional value of the items on every restaurant’s menu.
Prohibiting and Tax are considered to be hard interventions because they alter the choice
set that individuals face. Order is a soft intervention since it does not change the choice set but
rather its presentation in order to influence behavior. To enhance transparency, the text included
an explanation that this method is based on psychological research indicating that people tend to
choose items at the beginning of a list (see Dayan and Bar-Hillel, 2011). App is an educational
intervention which enables individuals to easily obtain relevant information without changing the
set of choices and without manipulating the presentation of the choice set.
In each of the treatments, a participant was asked to compare between different pairs of
government actions. Each comparison between actions X and Y consisted of two questions: The
first asked the participant to compare the two government actions, given that they are equally
!
13!
effective in improving the public’s eating habits, on a scale of five points: “greatly prefer X”,
“slightly prefer X”, “no preference”, “slightly prefer Y” and “greatly prefer Y” (see Screen 3 of
Study 2 in Appendix A). This question was included only in order to facilitate the participant’s
understanding of the second question, which provides finer information and is the core question
of this study. The results for the two questions are largely consistent and therefore we report only
the results of the second.
The second question was designed to elicit a participant’s subjective tradeoff between the
effectiveness and desirability of the intervention method. Effectiveness was defined in the
following manner: “Suppose that the World Health Organization has determined that consumption
of unhealthy food above a certain level is significantly harmful to human health. … The
improvement in public nutrition resulting from any government action is measured according to
the proportion of the population that drops to below this level as a result of the action.” Participants
were asked to indicate their preference between the two government actions X and Y in each row
of a table structured like Table 4 (i.e. given various relative levels of effectiveness of X and Y).
Improvement after
Action X
Improvement after
Action Y
I would choose -
%8
20%
Action X
Action Y
%12
20%
Action X
Action Y
%16
20%
Action X
Action Y
%20
20%
Action X
Action Y
%24
20%
Action X
Action Y
%28
20%
Action X
Action Y
%32
20%
Action X
Action Y
Table 4. The table used to elicit preferences between action X and action Y in Study 2.
Participants reveal a strong preference for the action X by choosing it even in cases where
the action Y is expected to be more effective, and vice versa. This type of preference elicitation
provides more information than an abstract question such as “How much do you support the new
!
14!
policy?” since it forces the participants to express their subjective tradeoff between the
effectiveness of an intervention and the desirability of its method.
In T1 and T2, participants were asked to compare Order (a soft intervention) to App and
also either Tax or Prohibition (both of which are hard interventions) to App. This aspect of the
design makes it possible not only to assess the resistance of participants to soft and hard
interventions but also to examine the correlations between their attitudes towards the two types of
interventions.
In order to determine the extent to which participants do not support any intervention of
the sort discussed in this paper, we preceded T1 and T2 with background information on the issue
of consumption of unhealthy food and asked participants whether they think the government
should intervene in order to improve the public’s eating habits (see Screen 2 of Study 2 in the
Appendix A). Those who answered “No” were not asked to compare between pairs of government
actions and instead were asked to explain their response (see Screen 3 after “No” in Appendix A).
Those who answered “Yes” continued on and were asked to compare between pairs of policies, as
described above. A summary of the two treatments appears in Table 5.
Table 5. Description of T1 and T2 in Study 2.
To conclude, we measure (a) the proportion of participants who object to any government
intervention in the public's eating habits (Screen 1), (b) the proportion of participants in T1 and T2
who have strong preferences for information and are ready to sacrifice effectiveness so that App
(rather than Order) will be adopted, and (c) the correlation between “not opposing” Order (not
willing to sacrifice effectives) and “not opposing” Tax or Prohibition (in T1 and T2, respectively).
T1
T2
Screen 1: Should the government intervene?
Continue if “Yes” (otherwise, explain why)
Screen 2: Comparison when equally effective
Tax vs. App
Order vs. App
Screen 3: Tradeoff comparison
Tax vs. App
Order vs. App
Screen 4: Comparison when equally effective
Order vs. App
Prohibition vs. App
Screen 5: Tradeoff comparison
Order vs. App
Prohibition vs. App
!
15!
3.2 Results of Study 2
In the presentation of the results, we have eliminated the fastest responding 5% of the participants
(as in Study 1) and those whose answers to questions structured like those in Table 4 were not
monotonic.
Result 1: Objection to any type of government intervention
There exists a small but not insignificant group of participants who object to any form of
government intervention.
In total, 14%-25% of the participants stated in Screen 1 that the government should not
intervene in any way.9 Most of them (76%-80%) provided the explanation that it is not the
government’s business to intervene in the private domain (Table 6). The results of T1 and T2
presented in the rest of the section relate only to those participants who think that some form of
government action is justified in this context.
Germany
Israel
USA
Government should not intervene at all
19%
14%
25%
No place for interventions in the private domain
76%
79%
80%
Table 6. Percentage of participants in T1 and T2 who believe that the government should not intervene at all in this
context and the proportion of them whose reason is opposition to intervention of any kind in the private domain.
Result 2: Negative attitude towards the soft intervention (relative to information provision)
The results indicate that a majority of the participants consistently prefer the intervention which is
more effective. However, 21%-37% of the participants strictly prefer App to Order. In other words,
they prefer that App be implemented even if it is significantly less effective than Order.10 Similarly,
9 The 95% confidence intervals are (14%, 24%) in Germany, (10%, 18%) in Israel and (20%, 30%) in the US.
10 The 95% confidence intervals are (33%, 44%) in Germany, (29%, 41%) in Israel and (15%, 27%) in the US.
!
16!
19%-54% of the participants strictly prefer government information campaign over Order even at
the price of reduced effectiveness (T5; see Appendix B).11
If a participant prefers action X to action Y even in cases where action Y is more effective
(as in rows 1-3 in Table 4), we label him as strictly prefers action X and vice versa. If he always
chooses the more effective intervention, we label him as exhibiting no tradeoff (i.e. between the
subjective desirability of the intervention’s method and its effectiveness). Table 7 presents a
summary of the preferences in T1 and T2 for the comparison between Order and App.
Note that the category strictly prefers App includes individuals who are willing to sacrifice
4%, 8% or “12% or more” in effectiveness in order for their preferred intervention to be adopted.
In fact, almost half of the individuals in this group were willing to sacrifice 12% or more in
effectiveness.
Germany
(n=206)
Israel
(n=239)
USA
(n=180)
Strictly prefer Order
13%
9%
13%
No tradeoff
50%
55%
66%
Strictly prefer App
37%
35%
21%
Table 7. Comparison between Order and App in T1 and T2 of Study 2.
Result 3: Attitude towards hard interventions
The responses regarding hard interventions provide two additional insights (see Appendix B for
more details):
(1) The proportion of subjects who chose solely according to intervention effectiveness
when comparing informational interventions to hard interventions is much smaller than when
comparing them to soft interventions. When comparing App and Tax, a larger proportion of
individuals prefer the former (51%, 63% and 41% in Germany, Israel and the US, respectively)
11 In the main treatments that included Order, i.e. T1 and T2, we added to the description of the intervention a short
summary of the typical arguments for and against soft interventions in general. In T5, this summary was omitted. This
did not qualitatively affect the results.
!
17!
and similarly when comparing App and Prohibition (84%, 73% and 66%, respectively). Thus, the
resistance to a soft intervention is much less than that to a hard intervention in this context.
(2) We find that in all three countries individuals who don’t mind Order when it is more
effective than App are substantially more likely than the rest (57% vs. 19%) to also accept the use
of Tax to improve people’s eating habits. A similar pattern (38% vs. 5%) was found when Tax is
replaced by Prohibition.
4. Study 3: Determinants of a negative attitude towards soft interventions
4.1 Method
The experiment was carried out online among undergraduate students in Germany (University of
Hamburg), Israel (Tel Aviv University) and the US (Ohio State University). The proportions of
men and women were similar. They were recruited and paid as in Study 1 except that the amount
was roughly equivalent to $20. Participants who agreed to participate were randomly assigned to
one of two treatments, TG or TE, which are described below. The number of participants in
Germany was 229 (121 in TG), in Israel 163 (76 in TG) and in the US 227 (120 in TG).
In TG, participants responded to a sequence of questions about their attitude towards
hypothetical government interventions that seek to encourage healthy eating habits in the
workplace. In TE, participants responded to identical questions, except that government
intervention was replaced by intervention by employers.
Each treatment consisted of four screens (see Appendix A):
First screen - Attitude towards green labeling intervention: Participants were told that there is
growing awareness that most of the population consumes too much unhealthy food, which leads
to major health problems. They were presented with a scenario in which the government has
decided to take action to reduce the consumption of unhealthy food by requiring workplace
cafeterias to mark healthy foods with a green label. In TE, the scenario was identical except that
the intervention was carried out by employers. The participants rated their attitude to such an
intervention on a scale of 1 (very positive) to 5 (very negative).
!
18!
Second screen - Attitude towards a background music intervention: Participants were told that
psychological studies have shown that background music unconsciously affects people’s
consumption habits (see, for example, Hansen and Melzner, 2014). They were then asked to
assume that background music of a certain type induces people to refrain from eating unhealthy
food. The government in TG has decided to reduce the population’s consumption of unhealthy food
by requiring workplace cafeterias to play this type of background music. In TE, the scenario was
identical except that the intervention was carried out by their employer. Again, the participants
rated their attitude to such an intervention on a scale of 1 (very positive) to 5 (very negative).
This type of intervention is not farfetched given that the use of background music is a
commonly used marketing tool (see Bruner, 1990). According to North and Hangreaves (2009):
“The range of the effects that music can have on consumer behavior is striking and different types
of music very clearly have different effects.”
Third screen – Reaction to five statements about the background music intervention: This
screen constitutes the core of Study 3. Participants were asked to what extent they agree with each
of the five statements below on a scale of 1 to 5 (1 - strongly agree, 2 - tend to agree, 3 - neutral,
4 - tend to disagree, 5 - strongly disagree):
3.1 People should significantly reduce their consumption of unhealthy food.
3.2 People's nutrition is not the government’s (employer’s) business.
3.3 It is inappropriate for the government (employers) to influence the personal decisions of
citizens (employees) without their knowledge.
3.4 There is a concern that this sort of intervention may lead citizens (employees) to absolve
themselves from their personal responsibility for their health.
3.5 Successful intervention is likely to lead the government (employers) to try to similarly
influence citizens’ (employees’) decisions in areas where there is no consensus on the appropriate
choice.
Fourth screen - Tradeoff between the desirability and effectiveness of the background music
and green labeling interventions: We used the same type of question as in Study 2 (see Table 4).
In this case, the participant was asked to compare between the green labeling and background
music interventions, assuming various differences in their effectiveness.
!
19!
The experiment was intended to elicit the attitude of participants towards a soft intervention
that uses background music to induce people to eat healthier without them being aware of the
intervention. As in Study 1 and 2, the goal of the intervention is likely to be viewed positively
which makes it possible to investigate the attitudes towards the intervention method. However, the
main motivation for Study 3 was to investigate the following:
(1) The reasons for negative attitudes towards some soft interventions: This is
accomplished by examining the responses to questions 3.1-3.5 and the connection between those
responses (i.e. participants' extent of agreement with the statements) and their attitude towards the
soft intervention. Statement 3.1 is intended to screen out the participants who did not agree with
the goal of the intervention. The other four statements correspond to the four other potential
reasons for opposing soft interventions mentioned in the introduction: opposition to any
government (employer) intervention, opposition to being manipulated, fear of people neglecting
their personal responsibility and the fear of a “slippery slope”.
(2) The difference in attitude towards government interventions as opposed to interventions
carried out by someone else (in this case, one's employer): A comparison of the two treatments
allowed us to examine how people's attitude towards a soft intervention depends on whether it is
carried out by the government or by employers.
To conclude, we perform the following analysis:
(a) We measure the proportion of the participants who have a negative attitude towards the
background music intervention in TG (Screen 2).
(b) We also measure the proportion of the participants who strongly prefer the green labeling
intervention over the background music intervention in TG (Screen 4).
(c) We then screen out the participants who do not agree with the intervention's goal by means of
3.1 and use two tests to determine which of the four other statements in Screen 3 are related to the
attitude towards the soft intervention in TG:
- We measure, for each statement in Screen 3 and each country, the correlation between the extent
of agreement with the statement and a negative attitude toward the background music intervention
in TG.
!
20!
- We run a regression with the attitude towards the government background music intervention as
the dependent variable and the extent of agreement with the four statements as explanatory
variables in order to determine the relative importance of the potential reasons.
(d) We test the whether the participants’ attitude towards the soft intervention in TE is more
positive than that in TG by comparing the responses to Screen 2 in the two treatments.
(e) We repeat the analysis described in (c) for TE.
4.2 Results
As in studies 1 and 2, the results presented below exclude 5% of the participants who spent the
least time reading the first screen in the experiment.
Result 1: Attitude towards the government background music intervention (TG)
The results indicate that a majority of participants viewed this soft government intervention
positively; nonetheless, a considerable proportion of the participants viewed such an intervention
negatively. The proportion of participants who expressed a negative attitude (response 4 or 5 to
Question 2) towards the soft government intervention was 35% in Germany, 45% in Israel and
31% in the US.12
In contrast, almost the entire sample viewed the green labeling informational intervention
(Question 1) positively (only 6% in Germany and 4% in the US viewed it negatively, while none
of the participants in Israel viewed it negatively). The preference for the green labeling
informational intervention over the background music intervention is also reflected in the results
of Question 4, i.e. the tradeoff question, which appear in Table 8. As in Study 2, we define a
preference for one intervention over another as strict if the participant is willing to sacrifice
effectiveness in order for his preferred intervention to be adopted. The table indicates that 55%-
67% of the participants strictly prefer the green labeling intervention over the background music
intervention.13
12!The 95% confidence intervals are (26%, 44%) in Germany, (34%, 56%) in Israel and (22%, 40%) in the US.!
13 As in Study 2, we have eliminated the participants whose answers were not monotonic.
!
21!
Germany
(n=106)
Israel
(n=72)
USA
(n=107)
Strictly prefer Music
5%
0%
1%
No tradeoff
34%
33%
44%
Strictly prefer Labels
61%
67%
55%
Table 8. The proportion of participants who strictly prefer each of the interventions in TG.
While there was considerable opposition to the background music intervention, there was
almost unanimous support for the goal of the two interventions, i.e. reducing the consumption of
unhealthy food. Only 3% of the participants in Germany and 4% in the US disagreed with the goal,
according to the responses to Question 3.1, while none of the participants in Israel disagreed. We
can now proceed to the main goal of the study: to determine the reasons for opposition to soft
interventions.
Result 2: Determinants of the attitude towards the government background music
intervention (TG)
The main reasons for opposition to the government intervention varied from country to country.
In Germany, it was the resistance to being manipulated; in Israel, the resistance to being
manipulated and fear of a slippery slope; and in the US, concerns about the neglect of personal
responsibility and fear of a slippery slope.
In order to gain further insight into the reasons for opposition to the background music soft
intervention, despite agreeing with its goal, we eliminated the (small number of) individuals who
do not support the goal and test the other potential reasons among the remainder of the participants.
In particular, we examine the link between the attitude towards the intervention and various views
reflected in the responses to questions 3.2-3.5 regarding the intervention’s method and
consequences. Table 9 compares these views between two groups: those who hold a negative
attitude towards the intervention (answered 4 or 5 to Question 2) and the rest. The table suggests
that in all three countries, people who hold a negative attitude toward the intervention tend to agree
more with all four statements, although in some cases this tendency is not very pronounced. (Recall
that smaller numbers reflect stronger agreement with the statement.)
!
22!
Germany
Israel
USA
Negative
(n=37)
The rest
(n=74)
Negative
(n=33)
The rest
(n=40)
Negative
(n=34)
The rest
(n=74)
3.2 It is not the government’s business
3.03
(1.26)
3.61
(1.04)
3.27
(1.1)
3.80
(0.85)
3.00
(1.28)
3.11
(0.95)
3.3 Concerns about manipulation
1.65
(0.92)
2.74
(1.09)
1.64
(0.93)
2.53
(0.78)
1.91
(1.08)
2.42
(1.05)
3.4 Personal responsibility concerns
2.68
(1.00)
3.07
(1.01)
2.91
(1.26)
3.43
(1.06)
2.35
(1.13)
2.99
(1.01)
3.5 Slippery slope concerns
2.08
(0.95)
2.58
(0.92)
1.58
(0.75)
2.55
(1.13)
1.91
(0.87)
2.43
(0.94)
Table 9. Average level of agreement with statements about the background music intervention in TG of Study 3
(standard deviations in parentheses).
The correlation between a negative attitude to the intervention and the level of agreement
with each of the statements suggests the following connections: In Germany, a negative attitude is
correlated with agreement with all the statements except for personal responsibility concerns,
where the strongest correlation is with concerns about manipulation (Pearson’s r=0.45, p<0.001).
In Israel, a negative attitude is correlated with all four statements, where the strongest correlations
are with concerns about manipulation (Pearson’s r=0.47, p<0.001) and slippery slope concerns
(Pearson’s r=0.45, p<0.001). In the US, a negative attitude is correlated with concerns about
manipulation (Pearson’s r=0.22, p<0.05), personal responsibility concerns (Pearson’s r=0.27,
p<0.01) and slippery slope concerns (Pearson’s r=0.26, p<0.01).
In order to further understand the relative importance of each factor in determining the
attitude towards the intervention, a linear regression was run with attitude (1-5) as the dependent
variable and the agreement with the four statements (3.2-3.5) as explanatory variables.14 The
results appear in Table 10. It was found that in Germany only manipulation concerns affect
attitudes both strongly and negatively (!=-0.48, p<0.001). In Israel, concerns about manipulation
negatively affect attitudes (!=-0.44, p<0.05) more than concerns about a slippery slope (!=-0.27,
14 Adding the level of agreement with the intervention's goal (statement 3.1) does not qualitatively change the results.
!
23!
p<0.10). In the US, concerns about personal responsibility negatively affect attitude (!=-0.21,
p<0.05) as do slippery slope concerns (!=-0.20, p<0.1).
Dependent variable: Attitude towards
the government music intervention
Germany
Israel
USA
Not the government’s business
-0.07
(0.09)
0.02
(0.15)
0.03
(0.11)
Manipulation concerns
-.48***
(0.09)
-0.44**
(0.17)
-0.10
(0.11)
Personal responsibility concerns
0.04
(0.10)
-0.08
(0.12)
-0.21**
(0.11)
Slippery slope concerns
-0.11
(0.11)
-.27*
(0.14)
-0.20*
(0.12)
R2
0.28
0.24
0.11
N
111
73
108
*p<0.1, ** p<0.05, *** p<0.01
Table 10. Coefficients in a linear regression where the dependent variable is attitude towards the government
background music intervention.
Interestingly, about half of the participants who do not oppose the background music
intervention have concerns about manipulation (43%, 50% and 58% in the three countries) and the
slippery slope (43%, 63% and 60%). The agreement among this group with the statement that “it
is not the government's business” was much lower (18%, 8% and 31%) and personal responsibility
concerns were not particularly common either (35%, 25% and 35%).
To conclude, the results shed light on the reasons for opposition to soft government
interventions. Among the four main reasons, there are two that appear to dominate: manipulation
concerns and slippery slope concerns. These exist not only among those who oppose the
intervention but are also present among many of those who overall are in favor of the intervention.
The regression analysis shows that each of these two concerns is also a major factor in explaining
the negative attitude in two of the three countries included in the study.
!
24!
Result 3: Attitudes toward interventions by an employer (TE)
The results indicate that the attitude towards the background music intervention is more positive
when it is carried out by one's employer than by the government. Manipulation concerns are found
to be the most common factor explaining a negative attitude towards soft intervention by an
employer. The attitude towards the informational green labeling intervention does not appear to
depend on who carries out the intervention.
The proportion of participants that expressed a negative attitude towards a background
music intervention by an employer was 25% in Germany, 18% in Israel and 14% in the US.15 In
all three countries, the proportion is lower than in the case that the government carries out the same
intervention (𝜒"(1)=2.77, p<0.1 in Germany, 𝜒"(1)=13.09, p<0.001 in Israel and 𝜒"(1)=9.07,
p<0.01 in the US). Similarly, in both Israel and the US, the distribution of attitudes (between 1 and
5) is significantly more positive (p<0.001 and p<0.01, respectively) according to a Mann Whitney
U test, though in Germany there is no significant difference (p=0.32). Thus, overall we find that
the attitude towards the background music intervention was more positive when the employer
carries it out than when the government does, though this is true to a lesser extent in Germany.
Regarding the green labeling intervention, in all three countries, there are no significant
differences in attitude whether it is the employer who carries it out or the government (p=0.23 in
Germany, p=0.27 in Israel and p=0.51 in the US).
There are small differences between the two treatments in the results of Question 4
concerning the tradeoff between the desirability of the intervention's method and its effectiveness.
As in the results for TG, a majority of the participants in TE (58%-59%) prefer the informational
intervention even at the cost of effectiveness, except in the US where the proportion is only 36%
(see Table 11 which corresponds to Table 8 for TG).
Germany
(n=101)
Israel
(n=81)
USA
(n=95)
Strictly prefer Music
7%
0%
8%
No tradeoff
35%
41%
56%
Strictly prefer Labels
58%
59%
36%
Table 11. The proportion of participants who strictly prefer each of the interventions in TE.
15!The 95% confidence intervals are (16%, 34%) in Germany, (10%, 26%) in Israel and (7%, 21%) in the US.!!!
!
25!
Table 12 presents the average extent of agreement with each of the four statements 3.2-3.5
among participants in TE and TG (after eliminating those who do not agree with the goal of the
intervention).
Germany
Israel
USA
TE
(n=91)
TG
(n=111)
TE
(n=79)
TG
(n=73)
TE
(n=94)
TG
(n=108)
3.2 It is not the employer’s business
2.80
(1.06)
3.41
(1.15)
2.97
(0.92)
3.56
(1.00)
2.97
(1.02)
3.07
(1.06)
3.3 Concerns about manipulation
2.55
(1.23)
2.38
(1.15)
2.39
(1.04)
2.12
(0.96)
2.68
(1.09)
2.26
(1.08)
3.4 Personal responsibility concerns
3.18
(1.08)
2.94
(1.02)
3.67
(1.10)
3.19
(1.17)
3.34
(0.90)
2.79
(1.09)
3.5 Slippery slope concerns
2.34
(1.00)
2.41
(0.96)
2.49
(1.04)
2.11
(1.09)
2.74
(1.03)
2.27
(0.94)
Table 12. Average level of agreement with statements concerning the background music intervention in TE and TG in
Study 3 (standard deviations in parentheses).
We find that in Israel and Germany, there is more agreement with the statement that it is
not the employer’s business than with the statement that it is not the government’s business
(p<0.001) according to a Mann-Whitney U test. In contrast, the agreement with this statement in
the US is not significantly different between the treatments (p=0.53) and there is less agreement
with the other three statements in the employer scenario than in the government scenario (p<0.01).
This again indicates that the participants in the US are more positive overall towards interventions
by an employer relative to those by the government. In Israel, there is also less agreement with the
slippery slope and personal responsibility concerns in the employer scenario (p<0.05), while in
Germany the level of agreement with these concerns does not differ significantly between the two
treatments (p=0.50 and p=0.13).
With regard to the link between a negative attitude towards the background music
intervention by an employer and statements 3.2-3.5, we find that in Germany and Israel a negative
attitude is correlated with all the statements except for concerns about personal responsibility,
whereas in the US it is correlated with all the statements except the belief that it is not the
!
26!
employer’s business. Furthermore, a linear regression analysis with attitude towards the
background music intervention as the dependent variable and agreement with the four statements
as explanatory variables (like that estimated in TG) suggests that in the US manipulation concerns
and personal responsibility concerns are the main factors affecting the attitude to the intervention
(!=-0.28, and !=-0.36, respectively, p<0.01), whereas in Germany and Israel the concern about
manipulation alone is the main factor (!=-0.29 and !=-0.63, respectively, p<0.001).
Overall, manipulation concerns are found to be the most common reason for opposition to
the soft intervention in TE, with other factors having a smaller effect.
5. Discussion
We carried out a series of experiments to shed light on the attitudes towards the intervention
methods advocated by libertarian paternalism. The main innovation of the experiments lies in the
method of eliciting the participants’ attitudes towards soft interventions (in comparison to other
interventions) and the investigation of the reasons for opposition to soft interventions among some
of the population.
The responses provide several indications of a negative attitude towards soft interventions:
In Study 1 and 3, we observe that negativity towards the soft interventions is at a fairly high level.
Of particular interest is the fact that a significant number of participants in Study 1 declared that
they would have made the choice desired by the government even if there had been no intervention
but that they nevertheless have a negative attitude towards the intervention. Furthermore, the
participants’ choices are consistent with the phenomenon of psychological reactance. Thus, the
default arrangement in Study 1 is quite attractive to the participants and the vast majority of
participants choose to opt-in when the government is not involved. However, when they are told
that the government is using choice architecture (automatic opt-in) to increase savings, some of
them choose not to use the arrangement, possibly as a protest against the government intervention
or simply because it induces antagonism.
In Study 2 and 3, it was found that a significant number of participants prefer an
informational intervention over a soft intervention that attempts to affect a participant's choice
without him being aware (for example, by altering the presentation of the choice set). The method
!
27!
used in the intervention is important to this group and they are also willing to pay a price in
effectiveness in order to avoid an undesirable method of intervention.
Study 3 also confirmed that a large number of people have concerns about the manipulative
nature of soft interventions. Furthermore, they fear that the acceptance of the approach will lead
to further interventions, carried out by a government that “knows what's good for its citizens” and
justifies it by arguing that freedom of choice is not being violated. This can be seen as another
indication that information provision might be a socially preferable type of intervention, even if
less effective than a choice architecture intervention.
The aforementioned indications of opposition to soft interventions are in addition to the
existence of a group (14%-25% of participants depending on the country) who feel that
governments should not intervene at all in the private domain (Study 2). For them, any intervention
has a negative effect on their non-material welfare.
The findings provide some practical insights for the implementation of soft interventions:
(i) An intervention involving a default opt-in might be less effective than simply suggesting
to people that they opt-in. The findings on reactance to manipulation suggest that simply asking
people whether they wish to change the choice architecture in their own decision problems might
improve the physical outcome, as well as diminish the negative emotional reaction to soft
interventions.
(ii) Although informational interventions may be less effective than other types of soft
intervention, many people are willing to sacrifice effectiveness in order to avoid a less desirable
method and therefore they may be preferable in cases where the loss in effectiveness is small.
(iii) Public decision makers should find ways to reduce the concerns of the public with
regard to soft interventions. Thus, it may be possible to improve welfare by explicitly informing
and reminding people that their choices are being manipulated and by constraining the government
not to intervene in issues that lack a broad consensus among the public.
(iv) Since hard interventions are usually more effective than soft ones, the fact that a
majority of those supporting a soft intervention also supported the hard intervention of imposing
a tax suggests that a government policy of refraining from altering individuals' choice sets
produces less of a welfare gain than one might have thought.
!
28!
(v) The reactance to soft interventions is reduced if they are carried out by an employer
rather than the government. This suggests that the government should encourage employers to
implement soft interventions on their behalf.
To conclude, the results confirm that a narrow majority of the participants accept soft
interventions. However, the results raise doubts as to the claim that soft interventions are
unambiguously welfare-improving. We believe that when considering the welfare effects of an
intervention, account should be taken of the reaction to the method used, rather than just the direct
material consequences of the intervention. In other words, policy makers should consider both the
support for an intervention's goal and the opposition to its method.
References
Jack Brehm. 1966. A Theory of Psychological Reactance, New York: Academic Press.
Gordon C, Bruner. 1990. “Music, Mood, and Marketing”. Journal of Marketing, 54(4), 94–104.
Colin Camerer, Samuel Issacharoff, George Loewenstein, Ted O’Donoghue and Matthew Rabin.
2003. “Regulation for Conservatives: Behavioral Economics and the Case for ‘Asymmetric
Paternalism’,” 151 University of Pennsylvania Law Review, 101, 1211–1254.
Nadine Chlaß, Werner Güth and Topi Miettinen. 2013. “Purely Procedural Preferences - Beyond
Procedural Equity and Reciprocity”, working paper.
James Choi, David Laibson, Brigitte Madrian, and Andrew Metrick. 2004. “For Better or For
Worse: Default Effects and 401(k) Savings Behavior”, in David Wise, ed., Perspectives in the
Economics of Aging. Chicago: University of Chicago Press.
Eran Dayan and Maya Bar-Hillel. 2011. “Nudge to Nobesity II: Menu Positions Influence Food
Orders”, Judgment and Decision Making, 6(4), 333–342.
Kfir Eliaz and Ariel Rubinstein. 2014. “On Fairness of Random Procedures”, Economic Letters,
123, 168–170.
Gidon Felsen, Noah Castelo and Peter Reiner. 2013. “Decisional Enhancement and Autonomy:
Public Attitudes towards Overt and Covert Nudges”, Judgment and Decision Making, 8 (3),
202-213.
!
29!
Roberto Fumagalli. 2016. “Decision Sciences and the New Case for Paternalism: Three Welfare-
Related Justificatory Challenges”. Social Choice & Welfare, 47 (2), 459-480.
Edward L. Glaeser. 2006. “Paternalism and Psychology”. University of Chicago Law Review, 73
(1), 133-155.
Till Grüne-Yanoff. 2012. “Old Wine in New Casks: Libertarian Paternalism Still Violates Liberal
Principles”. Social Choice and Welfare, 28 (4), 635-645.
William Hagman , David Andersson, Daniel Västfjäll and Gustav Tinghög. 2015. “Public Views
on Policies involving Nudges”. Review of Philosophy and Psychology, 6 (3), 439-453.
Jochim Hansen Johann Melzner, 2014. "What you hear shapes how you think: Sound patterns
change level of construal." Journal of Experimental Social Psychology, 54, 131-138.
Daniel M. Hausman and Brynn Welch. 2010. “Debate: To Nudge or Not to Nudge”. Journal of
Political Philosophy, 8 (1), 123-136.
Simon Hedlin and Cass R. Sunstein, 2016. “Does Active Choosing Promote Green Energy Use?
Experimental Evidence”, Ecology Law Quarterly, 43 (1), Article 3.
Jance Y. Jung and Barbara A. Mellers, 2016. “American attitude toward nudges”. Judgment and
Decision Making, 11, 262-74.
George Loewenstein and Emily Haisley. 2008. “The Economist as Therapist: Methodological
Ramifications of ‘Light Paternalism’,” in A. Caplin and A. Schotter (Eds.), Perspectives on
the Future of Economics: Positive and Normative Foundations, volume 1 in the Handbook of
Economic Methodologies, Oxford, England: Oxford University Press.
George Loewenstein, Cindy Bryce, David Hagmann and Sachin Rajpal. 2015. “Warning: You are
about to be Nudged,” Behavioral Science and Policy, 1 (1), 35-42.
Brigitte C. Madrian and Dennis F. Shea. 2001. "The Power Of Suggestion: Inertia In 401(k)
Participation And Savings Behavior," Quarterly Journal of Economics, 116(4), 1149-1187.
Adrian C. North, and David J. Hargreaves. 2009. “Music and Marketing”, in Patrik N. Juslin (ed),
Handbook of Music and Emotion: Theory, Research, Applications, 909-930. Oxford: Oxford
University Press.
Riccardo Rebonato. 2012. “Taking Liberties: A Critical Examination of Libertarian Paternalism”.
Palgrave Macmillan.
!
30!
Mario J. Rizzo and Douglas Glen Whitman. 2009. “Little Brother Is Watching You: New
Paternalism on the Slippery Slopes,” Arizona Law Review, 51 (3), 685-739.
Paul Rozin, Sydney Scott, Megan Dingley, Joanna K. Urbanek, Hong Jiang and Mark Kaltenbach.
2011. “Nudge to Nobesity I: Minor Changes in Accessibility Decrease Food Intake”.
Judgment and Decision Making, 6(4), 323–332.
Evan Selinger and Kyle Whyte. 2011. “Is There a Right Way to Nudge? The Practice and Ethics
of Choice Architecture”. Sociology Compass, 5 (10),923-935.
Cass R. Sunstein. 2014. “Why Nudge? The politics of Libertarian Paternalism”. Yale University
Press.
Cass R. Sunstein. 2015a. “Nudging and Choice Architecture: Ethical Considerations”, Yale
Journal on Regulation, Forthcoming.
Cass R. Sunstein. 2015b. “Do People like Nudges?”. Working paper.
http://ssrn.com/abstract=2604084
David, Tannenbaum, Craig. R. Fox and Todd Rogers. 2017. “On the Misplaced Politics of
Behavioural Policy Interventions”. Nature Human Behaviour, 1, Article 0130.
Richard Thaler and Cass R. Sunstein. 2003. “Libertarian Paternalism”, The American Economic
Review, 93, 175–179.
Richard Thaler and Cass Sunstein. 2009. “Nudge: Improving Decisions about Health, Wealth and
Happiness”. 2d edition. New York: Penguin Books.
Signild Vallgårda. 2012. “Nudge—A New and Better Way to Improve Health?” Health Policy,
104 (2), 200–203.
Allan Wailoo and Paul Anand. 2005 “The nature of Procedural Preferences for health-care
Rationing Decisions”, Social Science and Medicine, 60 (2), 223-236.
Mark White. 2013. “The Manipulation of Choice: Ethics and Libertarian Paternalism”, Palgrave
Macmillian.
Camille B. Wortman and Jack W. Brehm. 1975. “Responses to Uncontrollable Outcomes: An
Integration of Reactance Theory and the Learned Helplessness Model”, Advances in
Experimental Social Psychology, 8, 277-336.
!
31!
Appendix A: The experiments
The screens shown to participants in Study 1 and Study 2 in the US are presented below. In Germany, the text was
in German and the hypothetical scenarios took place in Germany. In Israel, the text was in Hebrew and the
hypothetical scenarios took place in Israel.
Study 1
Following are the two screens presented to participants in Treatment 1:
!
32!
In the first screen of Treatment 2 and 3, the beginning of the third paragraph was replaced by the following: “Assume
now that the government in USA has decided to try to increase the public’s rate of saving and is proposing a new law:
Employers will be required to remove the question about the percentage of saving from the form and set as a default
that 8% of each employee’s salary will be deducted and deposited in the special personal account.”
The rest of the text remained the same as in Treatment 1, except for the question that appeared at the end:
- In Treatment 3, participants were asked the following at the end of the first screen: “Would you cancel the default
arrangement, so that the question about savings would reappear in your monthly form?” (The possible answers were:
I would probably cancel/not cancel the arrangement.) In the second screen they were asked “How do you feel about
the government decision?”, as in Treatment 1.
- In Treatment 2, the order of the two questions was reversed.
After answering these two questions, the participants in all three treatments continued on to answer the
questions in Treatments 3-5 of Study 2.
!
33!
The following is the last screen that these participants were presented with:
* In Germany and Israel, Screen 6 included two additional questions requesting the participant’s age and gender.
!
34!
Study 2
The following two screens were presented to participants in Treatment 1:
!
35!
Those who answered “No” to the last question were presented with the following screen (and then Screen 7):
Those who answered “Yes” were presented with the following screens:
!
36!
!
37!
!
38!
* In Germany and Israel, Screen 7 included two additional questions that requested the participant’s age and gender.
!
39!
•! In Treatment 2, which had the same structure as Treatment 1, the first comparison was between ordering
menu items according to healthiness and development of an application. The second comparison was
between development of an application and a government prohibition against serving fatty foods in
restaurants on Wednesdays.
•! In Treatment 3-5, participants were not presented with Screen 2 (which contained the following question:
“In principal, are you in favor of the government taking measures to influence the public to eat healthier?”)
In those treatments, each participant made only one comparison between a pair of policies:
In Treatment 3 - between a tax on high-fat meals and an information campaign.*
In Treatment 4 – between an information campaign and development of an application.
In Treatment 5 - between ordering menu items according to healthiness and an information campaign.
In this treatment, we eliminated the two paragraphs on the advantages and disadvantages of soft
interventions that appear in Screen 5 of Treatment 1 (which starts with: “The government's proposal
is part of a general approach…”)
* Following is the description of the information campaign intervention that appeared in Treatment 3 (and similarly
in Treatments 4 and 5): “The government is also considering an alternative action: launching an extensive information
campaign in the media that would explain which foods are rich in fat and how they are harmful to our health. “
!
40!
Study 3
The following screens were presented to participants in TG (parallel questions replacing the government in their
employer were presented in TE). In Israel, the text was in Hebrew.
!
41!
!
42!
!
43!
Appendix B: Additional results of Study 2
In this section, we provide some of the additional results obtained in Study 2 which are not directly
related to the focus of the paper.
Participants were assigned to T3, T4 or T5 in Study 2 immediately after participating in
Study 1, which elicited their attitude towards soft government interventions in the saving decision.
In these treatments, we used a different informational intervention that involves an extensive
Information campaign in the media to explain which foods are high in fat and how harmful they
are to one's health. In each treatment, participants were asked to make only one comparison: In T3,
Tax to Information; in T4, App to Information and in T5 Order to Information. In each of the
treatments, the participants were first asked to compare the interventions on the assumption that
they are equally effective and then were asked to express their subjective tradeoff between
effectiveness and the method’s desirability in the same manner as in T1 and T2.
Additional comparisons: Information vs. App and Order vs. Information (T4 and T5)
Tables 13 and 14 present the preferences of participants between Information and App and between
Order and Information, respectively.
Germany
(n=117)
Israel
(n=159)
USA
(n=102)
Strictly prefer Information
15%
8%
7%
No tradeoff
61%
75%
69%
Strictly prefer App
23%
17%
24%
Table 13. Comparison between Information and App in T4.
Note that in all three countries, App was somewhat preferred but most of the participants chose
solely according to the intervention's effectiveness.
Recall that in T5, the summary of the advantages and disadvantages of libertarian-
paternalistic methods, which appeared in T1 and T2, was omitted. This did not qualitatively affect
the result that a significant proportion of the participants strictly prefer Information to Order.
!
44!
Germany
(n=132)
Israel
(n=153)
USA
(n=97)
Strictly prefer Order
8%
11%
12%
No tradeoff
38%
60%
68%
Strictly prefer Information
54%
30%
19%
Table 14. Comparison of Order to Information in T5 of Study 2.
Attitude towards hard interventions (T1, T2 and T3)
Tables 15, 16 and 17 present the preferences of participants between a hard intervention (Tax or
Prohibition) and an informational intervention (App or Information).
Germany
(n=90)
Israel
(n=108)
USA
(n=89)
Strictly prefer Tax
9%
1%
8%
No tradeoff
41%
36%
49%
Strictly prefer App
51%
63%
41%
Table 15. Comparison between Tax and App in T1 of Study 2.
Germany
(n=116)
Israel
(n=131)
USA
(n=91)
Strictly prefer Prohibition
4%
4%
3%
No tradeoff
13%
23%
31%
Strictly prefer App
84%
73%
66%
Table 16. Comparison between Prohibition and App in T2 of Study 2.
Germany
(n=96)
Israel
(n=166)
USA
(n=111)
Strictly prefer Tax
4%
6%
4%
No tradeoff
37%
28%
49%
Strictly prefer Information
59%
65%
48%
Table 17. Comparison between Tax and Information in T3 of Study 2.
!
45!
Correlation in the support for soft and hard interventions
We examined the relation between a participant’s attitudes towards hard and soft interventions
with the same goal (i.e., within-subject analysis). Only the data from T1 and T2 was used since
only in those treatments were participants asked about their attitude towards both a soft and a hard
intervention. It was found that in all three countries individuals who don’t mind Order (a soft
intervention) when it is more effective than App are substantially more likely to accept Tax (a hard
intervention) in order to improve people’s eating habits. According to Table 18, which presents
the aggregated results for T1, the difference in attitude towards Tax between those who “oppose”
Order and those who do not is statistically significant (n=287, 𝜒"(1)=28.9, p<0.01).
“Oppose” Tax
Do not “oppose” Tax
“Oppose” Order (n=69)
81%
19%
Do not “Oppose” Order (n=218)
43%
57%
Table 18. Attitude towards Tax as a function of the participant’s attitude towards Order in T1.
Table 19 presents the aggregate results for T2. The difference in attitude towards Prohibition
between those who “oppose” Order and those who do not is statistically significant (n=338,
𝜒"(1)=44.82, p<0.01).
“Oppose” Prohibition
Do not “oppose” Prohibition
“Oppose” Order (n=129)
95%
5%
Do not “oppose” Order (n=209)
62%
38%
Table 19. Attitude towards Prohibition as a function of the participant’s attitude towards Order in T2.
