Are default contributions sticky? An experimental analysis of defaults in public goods provision

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Journal of Economic Behavior & Organization journal homepage: www.elsevier.com/locate/jebo

Are default contributions sticky? An experimental analysis of defaults in public goods provision Dominique Cappelletti a , Luigi Mittone a,b , Matteo Ploner b,∗ a b

Fondazione Bruno Kessler, Trento, Italy DEM-CEEL, University of Trento, Italy

a r t i c l e

i n f o

Article history: Received 1 July 2013 Received in revised form 2 January 2014 Accepted 5 January 2014 Available online xxx

Keywords: Defaults Public goods Beliefs

a b s t r a c t Previous research provides compelling evidence that defaults affect individual behaviour in several domains. However, evidence of their influence in strategic interaction is scant. We experimentally investigate the effect of defaults on contributions to a public good and attempt to shed light on potential channels through which they operate. Our main experimental findings show that defaults influence contribution behaviour: preference for a suggested contribution significantly increases when it is presented as the default. However, this effect seems not to operate primarily through information conveyance or expectations about others’ behaviour. Default contributions, thus, appear to have an attractive power that goes beyond recommendation signals and expectation influences. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Our everyday economic decisions are pervaded by defaults. Defaults are predefined choices that become effective when decision makers do not take an action to change them. We encounter defaults when, for example, installing software, buying a flight ticket online, or ordering in a fast food. In many situations marketers, employers, and policymakers set default options that would not be problematic in a fully rational world, since people would not stay with defaults that do not correspond to the best option for them (Thaler and Sunstein, 2003). However, recent research has shown that defaults have the power to influence individual behaviour in domains as diverse as retirement savings (e.g., Madrian and Shea, 2001; Choi et al., 2004; Bütler and Teppa, 2007; Beshears et al., 2008), consumption (e.g., Johnson et al., 1993; Park et al., 2000; DellaVigna and Malmendier, 2006), organ donation (Johnson and Goldstein, 2003). More recently, experimental contributions provided evidence that defaults also affect choices in strategic situations, such as contributions to public goods (Altmann and Falk, 2009; Carlsson et al., 2011). Why do default effects occur in social dilemmas? Is it because of the information conveyed, i.e., defaults are interpreted as a suggestion? Is it because they serve as a stronger coordination device for those who cooperate conditionally? To answer these questions, we compare contributions to a public good in three different treatments in which the common-knowledge suggested contribution received by the members of a group assumes alternatively the form of simple advice given by a human participant, the form of default contribution set by a human participant, and the form of default contribution set by a computer with a certain probability. When it assumes the form of default, participants automatically contribute this amount unless they specify a different amount. When it assumes the form of simple advice, participants are asked to actively make their contribution, without any kind of automaticity. To check the effectiveness

∗ Corresponding author. Tel.: +39 0461 283139. E-mail address: [email protected] (M. Ploner). http://dx.doi.org/10.1016/j.jebo.2014.01.002 0167-2681/© 2014 Elsevier B.V. All rights reserved.

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model JEBO-3270; No. of Pages 12

ARTICLE IN PRESS D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

2

of the common-knowledge suggested contribution as coordination device, we elicit participants’ beliefs about their group members’ contributions. Our main experimental results show that defaults influence contribution behaviour: preference for the suggested contribution significantly increases when it is presented as the default. However, this seems to be explained by neither information conveyance nor impact on expectations about others’ behaviour alone. Default contributions, thus, appear to have an attractive power that goes beyond recommendation signals and expectation influences. The remainder of this paper is organized as follows. Section 2 reviews previous studies investigating default effects and discusses some contributions related to our study; Section 3 describes the experimental design and procedures and outlines the main predictions of the relative behaviour in the different treatments; Section 4 presents the results of the experiment; Section 5 summarizes and discusses the main experimental results and concludes.

2. Background One of the most robust findings in the behavioural and experimental literature is the existence of a default bias, i.e., an exaggerated preference for the default option. For example, Johnson et al. (2002) found that participants in a Web experiment were much more willing to be notified about subsequent surveys when the yes-response was checked by default than when it was not, and Park et al. (2000) reported that participants chose a car with a larger and more expensive set of options when the fully loaded model was presented as the default. In addition to their impact on consumption choices (e.g., Johnson et al., 1993; DellaVigna and Malmendier, 2006; Pichert and Katsikopoulos, 2008), the pervasive influence of defaults has been documented also in more consequential decisions such as organ donation and retirement savings. Johnson and Goldstein (2003) showed that the rates of consent for organ donation are dramatically higher in the presumed-consent than in the explicit-consent countries, and Madrian and Shea (2001) reported a dramatic increase in the participation in retirement plans following a switch from opt-in to opt-out participation default. More recently, experimental contributions provided some evidence that defaults play a role also in strategic situations. Altmann and Falk (2009) and Carlsson et al. (2011) showed that contributions to a public good (an experimental and a real one, respectively) are greatly influenced by the default contributions set by the experimenter. Why do default effects occur? Several explanations have been put forward in the literature. One set includes effort-based explanations. Making decisions requires cognitive effort and, often, also physical effort, such as completing a form, making phone calls, or going to an office. Some decisions may also involve questions that are likely to generate negative emotions. Sticking to the default may thus reflect an attempt to economize on cognitive, physical, and emotional effort. Effort increases with the complexity of the decision at hand. Complexity may stem from a variety of factors, for example the amount of time available for deciding (Dhar and Nowlis, 1999), the number of options to be evaluated (Iyengar and Lepper, 2000; Iyengar et al., 2004), the lack of familiarity or expertise, or the presence of decisional conflict, i.e., the lack of compelling reasons to choose one option over another (Shafir et al., 1993). The tendency to accept the available default option should thus be greater when facing difficult decisions (Tversky and Shafir, 1992; Mitchell and Utkus, 2006; Fleming et al., 2010). Another set of explanations focuses on cognitive biases related to the concept of loss aversion (Kahneman and Tversky, 1979), according to which the disutility associated with a loss is greater than the utility associated with a gain of the same magnitude. Evidence of loss aversion can be found in the well-known endowment effect (Thaler, 1980), i.e., the tendency for people to value an object more when they possess it than when they do not. In the case of defaults, an endowment effect may be at work: people may perceive the default option as something they possess and, thus, place more value on it. Loss aversion is also responsible for the status-quo bias (Samuelson and Zeckhauser, 1988), i.e., people’s tendency to prefer the current state of affairs over a change. The status-quo bias often occurs together with the omission bias (Ritov and Baron, 1992), i.e., people’s tendency to prefer inaction over action. Retention of the default option can be explained both in terms of status-quo bias—since the default can be perceived as the current state of affairs—and in terms of omission bias—since no action is required to accept the default, while an action is needed to change it. It has also been proposed that defaults matter because of the information they convey. Specifically, people may interpret the default as an implicit recommendation. If there are no conflicts of interest between those who set the default rule and the recipients, the default option is seen as a reasonable choice, since it can reflect what most people do or what informed people think is sensible to do (Johnson and Goldstein, 2003; Sunstein and Thaler, 2003; McKenzie et al., 2006). More recently, it has been proposed that defaults affect behaviour because they serve as a cue by which people construct their preferences (Dhingra et al., 2012). In this study we investigate default effects in strategic interaction, specifically in public goods provision. Indeed, while the influence of defaults in non-strategic environments has been object of extensive research, their influence in strategic settings has been under-investigated. The available evidence is limited to cases in which there are switching costs (although small) and the defaults are set by the experimenter. Defaults set by the experimenter have the potential to create a strong demand effect or, conversely, to make participants perceive a conflict of interest. We are interested in whether default effects are sticky also when there are no cognitive and physical costs for departing from the default and when defaults are not set by the experimenter. In addition, we aim to deepen our understanding of why defaults influence behaviour in strategic settings. It might be due to the information conveyed by default options, i.e., to the fact that defaults are perceived as a suggestion about what choice to make. Previous research has highlighted the effectiveness of suggestions in shaping contribution choices. For example, Croson and Marks (2001) investigated the role of suggestions in a Threshold Public Goods Game, where suggestions were posted in the instructions and were based on a symmetric cost-sharing rule, i.e., the cost Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

3

of the public good was equally divided among the group members.1 As underlined by the authors, this symmetric option is already a strong focal point, and suggesting it may promote common knowledge that this is a reasonable pattern of contributions. Different from Croson and Marks (2001), in our experiment suggested contributions assume (also) the form of default, i.e., what participants actually contribute if they do not take an action to change it; in addition, suggestions are made not by the experimenter but endogenously by other participants; further, suggestions are not strengthened by focal contribution patterns. Leaders’ contributions in sequential public goods (or public bad) provision can also be interpreted as a suggestion provided to the group members. In these studies (e.g., Güth et al., 2007; Levati et al., 2007; Gächter et al., 2012) a group leader decides first how much to contribute to a public good. This contribution is common knowledge: before simultaneously deciding their contributions, the other group members are informed about the leader’s contribution, and the leader knows, when deciding, that her contribution will be communicated to the other group members. Do the other group members follow this suggestion? Güth et al. (2007) found that group members condition their contribution decisions on their leader’s contribution, leading contributions to higher levels compared to the condition without a leader. However, although being highly correlated, leaders’ contributions are systematically higher than the other group members’ contributions. Unlike in leadership studies, in the present experiment suggestions take (also) the form of default contributions. In addition, while in leadership studies the participant who sends the suggestion (i.e., the leader) is a member of the group, in this experiment she is not a member, but has an interest in the performance of the group, as it will be explained in Section 3. Further, leaders’ suggestions are somehow reinforced by the example, since the leader actually contributes the suggested amount, while in this experiment the participant who sets the default contribution does not even make a contribution. Recently, Levy et al. (2011) analyzed the effectiveness of leaders’ suggestions when the leader does not have a first mover advantage and found that contributions are significantly influenced by suggestions made by an either elected or randomly chosen human leader, while suggestions from a computer have no impact. Another plausible reason for the influence of defaults on contributions could be that the default contribution serves as a sort of stronger coordination device. In this experiment, the suggested contribution is common knowledge: all members of a group know that all receive a common suggested contribution. It could influence participants’ expectations about others’ contributions more when it assumes the form of default contribution then when it is a simple advice and this could be particularly relevant for conditional cooperators, i.e., for those who cooperate more the more they know or expect others to cooperate (e.g., Fischbacher and Gächter, 2010). To investigate this facet of default effects, we elicit participants’ beliefs about the contributions of the other members of their group. Our experimental design and procedures are detailed in the next section. 3. The experimental design This study aims to investigate whether and to what extent default options influence choices in strategic settings. Specifically, we adopt a linear Public Goods Game, in which contributions are made on a voluntary base, and examine whether the presence of a default contribution affects behaviour. In addition, we aim to shed light on the reasons of default effect occurrence in public goods provision. In a standard linear Public Goods Game, each individual i in a group X composed of N individuals is provided with an endowment ei and chooses the amount ci she wants to contribute to the public good, such that 0 ≤ ci ≤ ei . The amount (ei − ci ) not contributed to the public good is kept in the individual’s private account, which earns a constant return of 1, and paid at the end of the experiment. The sum of all the contributions in a group is multiplied by a factor ˛ and equally split among the members of the group. Thus, the payoff function of each individual i ∈ X is given by the following function  i = e − ci + (1/N)˛( i∈X ci ), where (1/N) < (˛/N) < 1. Given the structure of the payoff function, the social optimum is reached when all the individuals i ∈ X contribute all their endowment, i.e., i∈X ci = Ne. However, a self-interested individual has no incentive to contribute to the public good, since for each unit contributed to the public good she earns only (˛/N). Thus, following conventional economic reasoning, the public good will not be produced and individuals will retain all their endowment in their private account. However, empirical evidence shows that in one-shot Public Goods Games and in the first round of finitely repeated Public Goods Games individuals generally contribute about half of their endowment to the public good (Marwell and Ames, 1979; Ledyard, 1995). In this experiment, participants played a linear Public Goods Game for three periods. Participants were assigned to 4person groups and informed that the group composition would not change across periods. Participants did not receive any feedback about the choices made by other members of their group until the end of the experiment. One of the three periods was randomly selected for payment at the end of the experiment. In each period of the game, each participant was endowed with 10 Experimental Currency Units (ECU) and asked to decide how much to contribute to a public good that yields a marginal per capita return of 0.4. Each period was composed of two stages: in the decision stage, participants made their contribution to the public good, while in the estimation stage

1 In the condition with heterogeneous preferences, i.e., group members receive different payoffs from the public good, a suggestion based on equal period payoffs was also implemented, but did not produce significantly different results from those obtained with the symmetric cost-sharing suggestion.

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

4 Table 1 Experimental treatments.

(A) (B) (C)

Treatment

Form of suggested contribution

Source of suggested contribution

Default 50-50 Default 100 Advice 100

Default Default Advice

50-50 100 100

participants estimated the contributions to the public good made by the other three members of their group.2 The accuracy of estimations was rewarded with modest payoffs using the following payment scheme: if the participants estimated the sum of the contributions made by the other three members of their group precisely, then each of them earned 3 ECU; if they guessed wrongly, each of them earned 1.5 ECU divided by the absolute difference between the estimated sum of contributions and the sum of contributions actually made.3 In the first period, participants played a standard linear Public Goods Game. This allows us to gather information about the dispositional attitude of participants, i.e., what preferences exist in the absence of a default. In the second period, participants played one of the two different randomly assigned roles, referred to as role A and role B. In each session, there were five 4-person groups present in the laboratory. The members of four groups played role B, while the members of one group played role A. Each of the four participants A was associated with one of the groups composed of participants B. Participants B were asked to make a contribution to a public good and to estimate the contributions made by the other three members of their group. Participants A were asked to establish a suggested contribution that was communicated to the four participants B of the group associated with them. All the participants were asked to establish a suggested contribution before knowing their role. The contribution was effective only in case a participant was actually assigned the role A. The payoff of a participant A was calculated as the mean of the payoffs of participants B associated with her. In addition, participants A were asked to estimate the contributions made by the four members of the group associated with them. We ran three treatments, which differed only in the second period, in which we manipulated what form the contribution suggested by participants A assumes (either default or simple advice) and who set the suggested contribution (either a human participant or a computer). The three treatments are detailed in the next subsection. In the third period, participants played a standard linear Public Goods Game with no default contribution or advice. This allows us to check for carryover effects of default contributions. 3.1. Treatments This experiment was run under three different between-subjects treatments: Advice 100, Default 100, and Default 5050. These treatments differed only in the second period of the Public Goods Game. Table 1 summarizes the features of these treatments. In each treatment, 2 factors were manipulated: the form of the suggested contribution (form conditions) and the source of the suggested contribution (source conditions). The contribution suggested by the participant A associated with a group alternatively assumed the form of a default option and the form of simple advice. When it assumed the form of a default option (default condition), participants B in each group were informed about the contribution suggested by the participant A associated with their group and were told that they would automatically contribute this amount unless they specified a different amount. When the suggested contribution assumed the form of simple advice (advice condition), participants B in a group were informed about the contribution suggested by the participant A associated with their group and were required to make an active choice, without any kind of automaticity. The source of the suggested contribution, i.e., who set it, could be the participant A associated with a group or a computer randomly drawing a value from the set of available options. In the 100 condition, participants B in a group were informed about the suggested contribution and were told that it was established with 100% probability by the participant A associated with their group. In the 50-50 condition, participants B in a group were informed about the suggested contribution and were told that it was established with 50% probability by the participant A associated with their group and with 50% probability by the computer. 3.2. Behavioural predictions In this subsection, we outline our main predictions of the relative behaviour in the different treatments. To answer the question whether default contributions are sticky, i.e., whether people tend to stay with the default when a default is in place, we compare behaviour in treatment Default 100, in which the suggested contribution assumes the form of default, with behaviour in treatment Advice 100, in which the suggested contribution assumes the form of simple advice and participants are required to make an active choice. Given the mounting evidence that because of several psychological and

2 Participants were also asked to indicate, on a 5-point Likert scale ranging from “definitely unsure” to “definitely sure” how sure they were about their estimates. However, this task was not incentivized. 3 This payment scheme is the same used by Croson (2007).

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

5

behavioural reasons defaults are sticky in numerous domains (e.g., Johnson and Goldstein, 2003; DellaVigna and Malmendier, 2006; Beshears et al., 2008), we expect to observe a higher fraction of participants following (i.e., not changing) the suggested contribution when a default is in place. Hypothesis 1. The fraction of participants following (i.e., not changing) the suggested contribution in treatment Default 100 is significantly higher than the fraction of participants following (i.e., actively choosing) the suggested contribution in treatment Advice 100. To answer the question whether default effect occurrence is at least partly explained by the informational value conveyed by the default, i.e., to the fact that the default can be interpreted as a recommendation of what amount to contribute, we compare behaviour in treatment Default 100, in which the default contribution is established by a human participant, with behaviour in treatment Default 50-50, in which the default contribution is possibly set by a computer randomly. More precisely, we aim to check whether default effects are stronger, or even emerge only, when the default contribution is set by a human, rather than by a computer randomly drawing a value from the set of available options. Indeed, a default contribution set by a human can be interpreted as being more meaningful than a random message sent by a machine.4 This is in line with recent experimental evidence provided by Levy et al. (2011), who showed that suggestions from humans, either elected or randomly chosen, affect subjects’ contribution decisions, while suggestions from a computer have no impact. The authors say that these results support the view that “information provided by human leaders is uniquely able to establish welfare-enhancing norms”. Thus, if the default effect is partly explained by information conveyance, we expect to find it less pronounced when the default contribution can be a randomly drawn number over the support of choices in the game. Hypothesis 2. The fraction of participants sticking to the default contribution is significantly higher in treatment Default 100 than in treatment Default 50-50. One might wonder why participants should follow recommendations from other participants. There is a large amount of evidence showing that recommendations provided by other participants influence the behaviour of those receiving them, both when recommenders have prior experience (e.g., see for example intergenerational experiments on coordination games (Schotter and Sopher, 2003), trust games (Schotter and Sopher, 2006), ultimatum games (Schotter and Sopher, 2007), public goods games (Chaudhuri et al., 2006)) and even when, as in our experiment, recommenders do not have any prior experience or other kinds of informational advantage (e.g., Kuang et al., 2007; Levy et al., 2011). 3.3. Participants and procedures 120 students at the University of Trento (Italy) participated in the experiment. They were randomly assigned to one of the 3 treatments described in the previous section. Participants were recruited by poster advertising placed at the University. The experiment was conducted at the Cognitive and Experimental Economics Laboratory of the University of Trento (Italy).5 Upon entering the laboratory, participants were randomly assigned to computer-equipped cubicles that did not allow visual interaction among the participants. Participants received on-screen written instructions,6 which emphasized that the identity of interacting partners would have never been revealed to the participants. Questions were answered individually by the experimenter at the participants’ seats. Sessions lasted for about 45 min, and participants earned, on average, about D 9.74 (including a show-up fee of D 3).7 4. Results Fig. 1 presents a summary description, for each experimental treatment, of contribution choices in the three independent periods (ti ) of the Public Goods Game. The boxplots provide usual information on relevant quantiles of the distribution of choices, while the X dots identify the average contribution level in each period of the game. Across periods, contributions to the public good tend to concentrate on values comprised between 2 and 6. The lowest average contribution (3.056) is observed in period 3 among those who reject the suggested contribution in treatment Default 100. The highest average contribution (6.000) is observed in period 2 among those who accept the suggested contribution in treatment Advice 100. Concerning the correlation of choices across periods, the following pattern emerges. According to a Spearman’s rank correlation test, a positive but not statistically significant correlation between choices in the first and second period is observed among those who accept the suggested contribution (all p-values ≥0.107). In contrast, among those rejecting the suggested contribution the correlation is positive and statistically significant (all p-values ≤0.003). When

4 An anonymous referee noticed that in the real world, unlike in our experiment, recommendations given by computers could be very valuable. As an example, think of the algorithms backing “auto-suggestions” provided by many web-based services. 5 Marco Tecilla is acknowledged for developing the software and for support in the recruiting of participants and in the management of the experimental sessions. 6 Experimental instructions are provided in Appendix. 7 Each ECU was converted in D 0.50.

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

6

0 10 t3

6

X 4

X

X

0

t2

t3

8

X

t1

t2

t3

6 4

X

X X

X

0

0

2

X

Contribution

6

8

10

t1

2

4

t3

8 t2

10

t1

Contribution

t2

2

X

Contribution

8 6 4

Contribution

X

2

X

0

Default_100

X

t1

10

t1

Advice_100

6

8 t3

X

4

t2

X

2

X

Contribution

6 4

Contribution

X

2

X

0

Default_50−50

8

10

Sugg. Contrib. ACCEPT

10

Sugg. Contrib. REJECT

t1

t2

t3

Fig. 1. Choices.

comparing choices in the second and the third period, the positive correlation in choices is confirmed for those rejecting the suggested contributions (all p-values ≤0.001). Among those accepting the suggested contribution, a marginally significant positive correlation is observed only in treatment Default 50-50 (p-value = 0.065). Overall, following the suggested contribution seems to weaken the consistency of choices across periods. However, the impact of the suggested contribution weakly reverberates to the third period. In the analysis below, we further elaborate on these findings and their determinants. We now check whether there are default effects, i.e., whether preference for the suggested contribution increases when it is presented as the default contribution. To this end, we first compare the frequencies of participants choosing the suggested contribution in treatment Default 100, i.e., when the suggested contribution assumes the form of default, and in treatment Advice 100, i.e., when it assumes the form of simple advice. In each of the two treatments Default 50-50 and Default 100, 43.75% of the participants choose to follow the suggested contribution, while only 21.87% follow the suggested contribution Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

7

Table 2 Regression analysis (Logit). SC.accepted∼

Model 1

Model 2

Model 3

(Intercept) Source Default SC Ch1 SC.dist Bel SC.dist SC.higher Default×Bel SC.dist Source×Bel SC.dist Default×Ch1 SC.dist Source×Ch1 SC.dist SC.higher×Ch1 SC.dist

−1.273 (0.661)o 0.000 (0.504) 1.022 (0.557)o

−0.941 (1.178) 0.018 (0.611) 1.482 (0.697)* 0.212 (0.140) −0.237 (0.135)o −0.386 (0.193)* −0.641 (0.624)

2.153 (2.456) −1.442 (1.252) −0.741 (1.369) 0.257 (0.168) −0.556 (0.640) −2.911 (1.144)* −1.220 (1.182) 2.498 (1.109)* 0.198 (0.367) 0.077 (0.447) 0.343 (0.365) 0.034 (0.360)

BIC Log likelihood Num. obs.

135.034 −60.670 96

137.498 −52.774 96

148.885 −47.057 96

*** p < 0.001; ** p < 0.01; * p < 0.05, o < 0.1.

when it assumes the form of simple advice (treatment Advice 100). A binomial test shows that in treatment Advice 100 it is less likely to stick with the default than to move away from it (p-value = 0.002). In contrast, in Default 100 the two alternatives have the same likelihood (p-value = 0.597). When comparing Default 100 and Advice 100, a Fisher’s exact test shows that there is a marginally significant statistical difference in the percentage of participants following the suggested contribution (pvalue = 0.055, one-tailed).8 The Fisher’s test, while being widely employed, is known to be overly conservative (e.g., Merotra et al., 2003). When adopting a more powerful unconditional alternative, i.e., Boschloo’s exact test, a statistically significant difference between the two conditions is observed (p-value = 0.034, one-tailed). The suggested contributions provided by participants in the Default 100 are generally lower (mean = 4.000) than those provided in Advice 100 (mean = 6.375). This may potentially impact on the likelihood of sticking with a default. To control for this, the regression analysis of Table 2, we introduce the level of the suggested contribution observed when assessing the likelihood of accepting a suggested contribution via a logistic regression. The dependent variable (SC.accepted) in the regression estimates is a dummy variable taking value 1 when the suggested contribution is accepted and value 0 otherwise and the explanatory variables are the two treatment variables. Concerning the explanatory variables, three alternative specifications are considered. In Model 1, only the two treatment variables are taken into account. The variable Source is set equal to 1 when the suggested contribution comes with certainty from a human participant and equal to zero when it comes with 50% probability from a computer randomly drawing a value from the set of available options. The variable Default is set equal to 1 when the suggested contribution assumes the form of default and equal to 0 when it assumes the form of a simple advice. In Model 2, the following explanatory variables are added to those already included in Model 1: SC captures the level of the suggested contribution; Ch1 SC.dist measures the distance, i.e., the difference in absolute value, between choices in first period and the suggested contribution; Bel SC.dist measures the distance between beliefs about others’ behaviour and suggested contribution; SC.higher measures whether the suggested contribution observed is higher (SC . higher = 1) or lower (SC . higher = 0) than own contribution choice in the first period. Finally, in Model 3 a few relevant interactions (×) between explanatory variables are introduced in the analysis. The regression outcomes of Model 1 show that the impact of Default is positive and marginally significant. However, when controlling for the characteristics of the suggested contribution (Model 2), the impact of Default is statistically significant at the conventional 5% level: participants are more likely to accept a suggested contribution when it assumes the form of default than when it assumes the form of simple advice. Thus, our Hypothesis 1 is supported. The results of the comparison of behaviour across different default conditions are summarized in Result 1. Result 1: There is a default effect in contribution behaviour: preference for a suggested contribution significantly increases when it is presented as the default. Model 2 also shows that the higher is the distance between suggested contributions and beliefs, the less likely is that the suggested contribution is followed. The same holds for own dispositions, as captured by choices in the first period, even though the effect is only marginally significant. Result 2: Preference for a suggested contribution significantly increases when the suggestion is closer to beliefs about others’ behaviour and to own dispositional attitudes. We now examine whether default effect occurrence is partly due to the informational value conveyed by the default, i.e., to the fact that the default can be interpreted as a recommendation of what amount to contribute. To this end, we first compare the frequencies of participants following (i.e., not changing) the default contribution across different Source

8

The adoption of a Fisher’s exact conditional test is motivated by the few individuals following the suggested contribution in Advice 100.

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

8 Table 3 Suggested contribution acceptance.

Default 100

Accept (Period 2) Reject (Period 2)

Accept (Period 3)

Reject (Period 3)

2 2

12 16 Default 50-50

Accept (Period 2) Reject (Period 2)

Accept (Period 3)

Reject (Period 3)

1 1

13 17 Advice 100

Accept (Period 2) Reject (Period 2)

Accept (Period 3)

Reject (Period 3)

2 5

5 20

conditions, i.e., when the default contribution is set with certainty by a human participant and when it is set with 50% probability by a computer randomly drawing a value from the set of available options. As already reported above, the percentage of participants choosing the default contribution is 43.75% in each of the two treatments Default 50-50 and Default 100. Thus, the source of the default contribution seems to have no impact on the likelihood of it being followed. The ineffectiveness of the source of default contribution is also evidenced by the regression outcomes of Table 1, which show that the source of the default contribution (Source) has no impact on the likelihood of following the suggested contribution. Thus, our Hypothesis 2 is not supported. The results of the comparison of behaviour across different source conditions are summarized in Result 3. Result 3: The value as a recommendation of the default does not explain default effect in contribution behaviour: preference for the default contribution does not decrease when it is possibly set by the computer. We now check whether there is a carryover effect of defaults, i.e., whether participants sticking with the default contribution continue to contribute the same amount when the default is removed. To this end, we compare contribution choices made in the second period (where suggested contributions are provided) with contribution choices made in the third period (where no suggested contributions are provided). If there is a carryover effect, we should not observe a significant reduction in the percentage of participants choosing in the third period the amount suggested in the second period. Table 3 provides a description of the joint distribution of suggested contribution acceptance in the second and in the third period. Table 3 clearly shows that there is a shift from acceptance to rejection of the suggested contribution in each treatment. Of those accepting the suggested contribution in the second period, only 14.29%, 7.14%, and 28.5% in treatments Default 100, Default 50-50, and Advice 100, respectively, chose the same amount also in the third period. With reference to this figures, binomial tests show that it is less likely to choose the default than not to choose it in treatments Default 100, Default 50-50 (p-value = 0.013 and p-value = 0.002, respectively.) In treatment Advice 100, in which very few acceptances are observed, there is not a significant difference between choosing or not the suggested contribution that was observed in the second period (p-value = 0.453). Similar conclusions are reached when testing whether there is a significant change in acceptance between periods. A McNemar’s test shows that there is a significant drop in the acceptance of the suggested contributions in treatments Default 100 (p-value = 0.016) and Default 50-50 (p-value = 0.003), but not in Advice 100 (p-value = 1). Result 4: The influence of default contributions on behaviour is not long-lasting: preference for the default contribution disappears when the default is removed. Next, we consider the beliefs of participants about the other group members’ contributions to the public good.9 This helps us shed more light on potential mechanisms underlying default effects. An important channel for suggested contributions to affect behaviour is via beliefs about how others are going to behave. Before assessing the impact of suggested contributions on beliefs, we check whether beliefs positively correlate with observed behaviour. A Spearman’s rank correlation test shows that there is a strong positive correlation between individual-level average beliefs and average contributions across the three periods (rho = 0.678, p-value < 0.001). Thus, participants in our sample are heavily influenced by what they expect from the other group members and behave as conditional cooperators. Concerning the relation between beliefs and the suggested contribution observed in the second period, a Spearman’s rank correlation test shows that there is a strong positive correlation between these two variables (rho = 0.409, p-value < 0.001). However, there is a difference between the correlation coefficients of those accepting and those rejecting the suggested contribution. For those accepting it, the rho is equal to 0.651 (p-value < 0.001) and for those rejecting it is equal to 0.286

9 As a measure of individual beliefs, in each period we compute the average of the three values (one for each other member of the group) provided by each participant.

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

9

Table 4 Regression analysis (OLS). Ch2 1.diff∼ (Intercept) Source Default SC Ch1.diff SC Bel.diff SC.higher SC Default×SC Ch1.diff Source×SC Ch1.diff SC.higher ×SC Ch1.diff AIC Log likelihood Num. obs.

−0.632 (1.464) 0.042 (0.586) 0.387 (0.687) 0.803 (0.266)** −0.306 (0.147)* 0.392 (0.870) 0.041 (0.148) −0.052 (0.144) −0.071 (0.136) −0.332 (0.228) 427.398 −202.699 96

*** p < 0.001; ** p < 0.01; * p < 0.05; o p < 0.1.

(p-value = 0.025). The fact that beliefs of those rejecting and those accepting the suggested contribution are differently affected by the suggested contribution is confirmed also by the distance between beliefs and the suggested contribution observed. The average distance is equal to 2.399 for those rejecting and to 1.181 for those accepting (Wilcoxon Rank Sum test, p-value = 0.002). Thus, those accepting the suggested contribution are more likely to believe that other group members adjust their contributions to the suggested one. The regression outcomes of Model 3 in Table 2 provide us with some insights on the possible mechanisms underlying the higher acceptance of the suggested contribution observed when it is presented as the default. Specifically, the negative and statistically significant coefficient of Bel SC.dist indicates that the higher the distance between own beliefs about others’ contributions and the suggested contribution observed, the lower the likelihood of accepting the suggested contribution. However, this does not hold when the suggested contribution has the form of a default, as indicated by the positive and statistically significant coefficient of Default×Bel SC.dist.10 This suggests that when the suggested contribution is a simple advice, considerations about others’ behaviour are more relevant than when the suggested contribution has the form of a default. Thus, the explanation that default effects occur because a default may be a stronger coordination device than a simple advice for conditional cooperators does not seem to find support in our data. These results are summarized in Result 5. Result 5: Considerations about others’ behaviour seem to be less relevant in the presence of a default. So far, we considered default effects in terms of strict adherence to the suggested contribution value, i.e., we considered frequencies and likelihood of choosing exactly the value suggested. However, defaults can influence behaviour also by making people adjust their contribution in the direction of the default contribution. The OLS regression model reported in Table 4 assesses the impact of a set of explanatory variables on the change of the contribution choice from the first period to the second period. Accordingly, the dependent variable Ch2 1.diff is given by the difference between contribution choices in the second and in the first period. In terms of explanatory variables, the main difference between the set adopted here and that adopted in Table 2 is given by the fact that, rather than on distances, we focus on differences between suggested contributions and choices in the first period (SC Ch1.diff) and beliefs about others’ behaviour (SC Bel.diff). The interaction terms in Table 2 test whether the impact of the SC Ch1.diff differs in alternative treatments. Moreover, we also check whether the impact of SC Ch1.diff is affected by the fact that the suggested contribution is higher or lower than own choice in the first period SC.higher×SC Ch1.diff. The positive and statistically significant coefficient of the variable SC Ch1.diff indicates that the higher the difference between the suggested contribution observed in the second period and the dispositional preference (contribution in the first period), the higher the adjustment of contributions towards the suggested value. Thus, a generalized convergence towards the observed suggested contributions is observed, irrespective of the fact that the suggested contribution is the default or a simple advice. However, as shown by the negative coefficient of the variable SC Bel.diff, the adjustment is more moderate when beliefs about others largely differ from the suggested contribution. These results are summarized in Result 6. Result 6: The suggested contribution does not need to be a default to be effective in moving contributions closer to the suggested value. However, the effectiveness of the suggestion is higher when the suggestion is in line with expectations about others’ behaviour. We conclude the analysis with a brief description of the suggested contributions (behaviour of participants A). As already noticed above, observed suggested contributions do not significantly differ across experimental treatments. Since participants were asked to provide their suggestions before knowing their role, the test can be extended by taking into account suggested contributions provided by all the participants in the experiment. Average suggested contributions are very close

10 A Linear hypothesis test indicates that we cannot reject the null hypothesis that the sum of the estimated coefficients of Bel SC.dist and Default×Bel SC.dist is equal to zero.

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model JEBO-3270; No. of Pages 12

10

ARTICLE IN PRESS D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

to the median value of five in each treatment and a series of Wilcoxon Rank Sum tests confirms that they do not differ across treatments (all p-values ≥ 0.781). Furthermore, a Spearman’s rank correlation test on pooled data shows that own choices in the first period are positively correlated with the suggested contribution provided (rho = 0.447, p-values < 0.001). A similar coefficient of positive correlation is observed also when considering treatments Default 100 and Advice 100 in isolation, but not in treatment Default 50-50 in which the correlation is not statistically significant (p-values = 0.102). Thus, suggested contributions seem to originate in preferences of the individual, even though the link is weaker when participants are aware that the likelihood that their suggested contributions are shown to the others is lower.

5. Discussion and concluding remarks This study examined the effect of defaults in public goods provision and attempted to shed more light on some potential mechanisms through which they work. Specifically, we tested whether defaults are effective because they are interpreted as a recommendation about what amount to contribute and because they may be a stronger coordination device than a simple advice for conditional cooperators. In the experiment, all the participants in a group received a common-knowledge suggestion about the amount to contribute to the public good. In three different treatments, the suggested contribution could be alternatively the default contribution, i.e., the amount contributed if a different amount was not explicitly specified, or a simple advice, thus requiring participants to make an active choice. In addition, the suggested contribution was provided alternatively by a human participant or by the computer with a certain probability. We also elicited participants’ beliefs about other group members’ contributions in order to check whether defaults operate indirectly through expectations. Our findings show that there is a default effect in contribution decisions: preference for the suggested contribution significantly increases when it is presented as the default. This result adds to the scant existing evidence of default effects in public goods provision by showing that defaults have an impact that goes beyond a simple advice and that defaults influence behaviour even when there are no physical or cognitive costs incurred in changing the default. In fact, participants made their decisions in the laboratory, seated in front of a computer. In addition, participants already took the same decision in the previous round, thus the marginal cognitive cost was minimal. Since they previously made the same decision, participants had already formed their preference when they were exposed to the default. We may expect to find an even stronger default effect when participants are not previously confronted with the same decision task. Indeed, Löfgren et al. (2012) found, in a study related to carbon offsetting, that default attenuate with experience. In addition, as proposed by Dhingra et al. (2012), default options may influence decision-makers in constructing their preferences. Albeit sizable, in our experiment the default effect did not result in a permanent change in contributions: preference for the default contribution dramatically decreased when the default was removed. Our results also show that default effects in our setting are not accounted for by information conveyance, i.e., by the fact that a default option signals a recommended action. In fact, the preference for the default contribution did not decrease when it could not undoubtedly be interpreted as a recommendation, i.e., when it was possibly chosen randomly by a computer. We do not find support for the view that defaults effectively influence contributions by exerting a stronger impact on expectations about others’ behaviour than a simple advice. Our results, in fact, show that considerations about others’ behaviour seem to be less relevant in the presence of defaults. Finally, looking at the influence of defaults in promoting adjustment towards suggested values, our results show that defaults are as effective as simple pieces of advice in moving contributions towards the suggested value. We found a number of interesting results. Nevertheless, much remains to be understood. In our experiment, the default contribution was established by a randomly chosen participant. Future research may investigate the effect of different sources of the default contribution, such as an experienced participant that performed well in previous public goods experiments, a participant that performed better than the others in a task, or a participant elected by the others. One might expect to find an even stronger default effect, since these sources have a certain degree of “legitimacy” relative to a simple random appointment. In addition, in our experiment the default contribution was set by a participant that had a monetary stake in the outcome of the game. Kuang et al. (2007) found that a simple advice about what action to take in a pure coordination game is more effective when the adviser has no monetary stakes in the outcome of the game than when she has. One might expect to find a stronger default effect when defaults are set by uninterested participants. Another line of research may also investigate the impact of default contributions in Public Goods Games with different characteristics, such as the presence of a provision point, or the presence of heterogeneity in the initial endowments or in the return on the public good, features that create a more complex environment. It would also be interesting to explore how default effects interact with individual characteristics. A recent and fast growing line of research investigates how cognitive skills are related to preferences and behavioural biases (e.g., Burks et al., 2009; Burnham et al., 2009; Dohmen et al., 2010; Jones, 2008; Oechssler et al., 2009). For example, Oechssler et al. (2009) found lower incidences of certain biases, such as the conjunction fallacy and overconfidence, in individuals with greater cognitive skills. Altmann and Falk (2009) found a negative relationship between default bias and cognitive ability. Knowing the relationships between individual characteristics and the occurrence of default effects may help deep our understanding of the mechanisms underlying default effects. Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model

ARTICLE IN PRESS

JEBO-3270; No. of Pages 12

D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

11

In general, further research is needed to understand why default effects occur in public goods provision. Default contributions seem to have an attractive power that goes beyond recommendation signals and expectation influences. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jebo. 2014.01.002. References Altmann, S., Falk, A., 2009. The impact of cooperation defaults on voluntary contributions to public goods. Mimeo. Beshears, J., Choi, J.J., Laibson, D., Madrian, B.C., 2008. The importance of default options for retirement savings outcomes: evidence from the United States. In: Kay, S.J., Sinha, T. (Eds.), Lessons from Pension Reform in the Americas. Oxford University Press, Oxford, pp. 59–87. Burks, S.V., Carpenter, J.P., Götte, L., Rustichini, A., 2009. Cognitive skills explain economic preferences, strategic behavior, and job attachment. Proceedings of the National Academy of Sciences of the United States of America 106 (19), 7745–7750. Burnham, T.C., Cesarini, D., Johannesson, M., Lichtenstein, P., Wallace, B., 2009. Higher cognitive ability is associated with lower entries in a p-beauty contest. Journal of Economic Behavior & Organization 72 (1), 171–175. Bütler, M., Teppa, F., 2007. The choice between an annuity and a lump sum: results from Swiss pension funds. Journal of Public Economics 91, 1944–1966. Carlsson, F., Johansson-Stenman, O., Khanh, N.P., 2011. Funding a new bridge in rural Vietnam: a field experiment on conditional cooperation and default contributions. Working Papers in Economics No. 503. University of Gothenburg. Chaudhuri, A., Graziano, S., Maitra, P., 2006. Social learning and norms in a public goods experiment with inter-generational advice. Review of Economic Studies 73, 357–380. Choi, J.J., Laibson, D., Madrian, B.C., Metrick, A., 2004. For better or for worse: default effects and 401(k) savings behavior. In: Wise, D.A. (Ed.), Perspectives on the Economics of Aging. University of Chicago Press, Chicago, pp. 81–121. Croson, R., Marks, M., 2001. The effect of recommended contributions in the voluntary provision of public goods. Economic Inquiry 39 (2), 238–249. Croson, R.T.A., 2007. Theories of commitment, altruism and reciprocity: evidence from linear public goods games. Economic Inquiry 45 (2), 199–216. DellaVigna, S., Malmendier, U., 2006. Paying not to go to the gym. American Economic Review 96, 694–719. Dhar, R., Nowlis, S.M., 1999. The effect of time pressure on consumer choice deferral. Journal of Consumer Research 25, 369–384. Dhingra, N., Gorn, Z., Kener, A., Dana, J., 2012. The default pull: an experimental demonstration of subtle default effects on preferences. Judgment and Decision Making 7 (1), 69–76. Dohmen, T., Falk, A., Huffman, D., Sunde, U., 2010. Are risk aversion and impatience related to cognitive ability? American Economic Review 100 (3), 1238–1260. Fischbacher, U., Gächter, S., 2010. Social preferences, beliefs, and the dynamics of free riding in public goods experiments. American Economic Review 100 (1), 541–556. Fleming, S.M., Thomas, C.L., Dolan, R.J., 2010. Overcoming status quo bias in the human brain. Proceedings of the National Academy of Sciences of the United States of America 107 (13), 6005–6009. Gächter, S., Nosenzo, D., Renner, E., Sefton, M., 2012. Who makes a good leader? Cooperativeness, optimism, and leading-by-example. Economic Inquiry 50 (4), 953–967. Güth, W., Levati, M.V., Sutter, M., van der Heijden, E., 2007. Leading by example with and without exclusion power in voluntary contribution experiments. Journal of Public Economics 91, 1023–1042. Iyengar, S.S., Lepper, M.R., 2000. When choice is demotivating: can one desire too much of a good thing? Journal of Personality and Social Psychology 79 (6), 995–1006. Iyengar, S.S., Huberman, G., Jiang, W., 2004. How much choice is too much? Contributions to 401(k) retirement plans. In: Mitchell, O.S., Utkus, S.P. (Eds.), Pension Design and Structure: New Lessons from Behavioral Finance. Oxford University Press, New York, pp. 83–95. Johnson, E.J., Goldstein, D., 2003. Do defaults save lives? Science 302, 1338–1339. Johnson, E.J., Hershey, J., Meszaros, J., Kunreuther, H., 1993. Framing, probability distortions, and insurance decisions. Journal of Risk and Uncertainty 7, 35–51. Johnson, E.J., Bellman, S., Lohse, G.L., 2002. Defaults, framing and privacy: why opting in-opting out. Marketing Letters 13 (1), 5–15. Jones, G., 2008. Are smarter groups more cooperative? Evidence from prisoner’s dilemma experiments, 1959–2003. Journal of Economic Behavior and Organization 68 (3–4), 489–497. Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision under risk. Econometrica 47 (2), 263–292. Kuang, X.J., Weber, R.A., Dana, J., 2007. How effective is advice from interested parties? An experimental test using a pure coordination game. Journal of Economic Behavior & Organization 62, 591–604. Ledyard, J., 1995. Public goods: a survey of experimental research. In: Kagel, J., Roth, A. (Eds.), The Handbook of Experimental Economics. Princeton University Press, Princeton, pp. 111–194. Levati, M.V., Sutter, M., van der Heijden, E., 2007. Leading by example in a public goods experiment with heterogeneity and incomplete information. Journal of Conflict Resolution 51, 793–818. Levy, D.M., Padgitt, K., Peart, S.J., Houser, D., Xiao, E., 2011. Leadership, cheap talk and really cheap talk. Journal of Economic Behavior & Organization 77, 40–52. Löfgren, A., PeterMartinsson, MagnusHennlock, Sterner, T., 2012. Are experienced people affected by a pre-set default option – results from a field experiment. Journal of Environmental Economics and Management 63, 66–72. Madrian, B.C., Shea, D.F., 2001. The power of suggestion: inertia in 401(k) participation and savings behavior. Quarterly Journal of Economics 116 (4), 1149–1187. Marwell, G., Ames, R., 1979. Experiments on the provision of public goods. I: Resources, interest, group size and the free-rider problem. American Journal of Sociology 84, 1335–1360. McKenzie, C.R., Liersch, M.J., Finkelstein, S.R., 2006. Recommendations implicit in policy defaults. Psychological Science 17 (5), 414–420. Merotra, D.V., Chan, I.S.F., Berger, R.L., 2003. A cautionary note on exact unconditional inference for a difference between two independent binomial proportions. Biometrics 59, 441–450. Mitchell, O.S., Utkus, S.P., 2006. How behavioral finance can inform retirement plan design. Journal of Applied Corporate Finance 18 (1), 82–94. Oechssler, J., Roider, A., Schmitz, P.W., 2009. Cognitive abilities and behavioral biases. Journal of Economic Behavior & Organization 72 (1), 147–152. Park, C.W., Jun, S.Y., MacInnis, D.J., 2000. Choosing what I want versus rejecting what I do not want: an application of decision framing to product option choice decisions. Journal of Marketing Research 37 (2), 187–202. Pichert, D., Katsikopoulos, K.V., 2008. Green defaults: information presentation and pro-environmental behaviour. Journal of Environmental Psychology 28, 63–73. Ritov, I., Baron, J., 1992. Status-quo and omission biases. Journal of Risk and Uncertainty 5, 49–61. Samuelson, W., Zeckhauser, R., 1988. Status quo bias in decision making. Journal of Risk and Uncertainty 1, 7–59.

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002

G Model JEBO-3270; No. of Pages 12

12

ARTICLE IN PRESS D. Cappelletti et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

Schotter, A., Sopher, B., 2003. Social learning and coordination conventions in intergenerational games: an experimental study. Journal of Political Economy 111 (3), 498–529. Schotter, A., Sopher, B., 2006. Trust and trustworthiness in games: an experimental study of intergenerational advice. Experimental Economics 9 (2), 123–145. Schotter, A., Sopher, B., 2007. Advice and behavior in intergenerational ultimatum games: an experimental approach. Games and Economic Behavior 58, 365–393. Shafir, E., Simonson, I., Tversky, A., 1993. Reason-based choice. Cognition 49 (1–2), 11–36. Sunstein, C.R., Thaler, R.H., 2003. Libertarian paternalism is not an oxymoron. The University of Chicago Law Review 70 (4), 1159–1202. Thaler, R., 1980. Toward a positive theory of consumer choice. Journal of Economic Behavior & Organization 1, 39–60. Thaler, R.H., Sunstein, C.R., 2003. Libertarian paternalism. The American Economic Review 93 (2), 175–179. Tversky, A., Shafir, E., 1992. Choice under conflict: the dynamics of deferred decision. Psychological Science 3 (6), 358–361.

Please cite this article in press as: Cappelletti, D., et al., Are default contributions sticky? An experimental analysis of defaults in public goods provision. J. Econ. Behav. Organ. (2014), http://dx.doi.org/10.1016/j.jebo.2014.01.002