《实验经济学》课程教学资源（参考文献）Behind the veil of ignorance：Self-serving bias in climate change negotiations.pdf_大学文库

Judgment and Decision Making, Vol. 6, No. 7, October 2011, pp. 602–615 Behind the veil of ignorance: Self-serving bias in climate change negotiations Peter H. Kriss∗ George Loewenstein∗ Xianghong Wang† Roberto A. Weber‡ Abstract Slowing climate change will almost certainly require a reduction in greenhouse gas emissions, but agreement on who should reduce emissions by how much is difficult, in part because of the self-serving bias—the tendency to believe that what is beneficial to oneself is also fair. Conducting surveys among college students in the United States and China, we show that each of these groups displays a nationalistic self-serving bias in judgments of a fair distribution of economic burdens resulting from mitigation. Yet, we also show, by disguising the problem and the identity of the parties, that it is possible to elicit perceptions of fairness that are not influenced by national interests. Our research reveals that the self-serving bias plays a major role in the difficulty of obtaining agreement on how to implement emissions reductions. That is, the disagreement over what constitutes fair climate policy does not appear to be due to cross-national differences in what constitutes a fair distribution of burdens. Interventions to mitigate the self-serving bias may facilitate agreement. Keywords: self-serving bias, climate change, negotiation, fairness, veil of ignorance. 1 Introduction Although debate continues about technical and scientific issues surrounding climate change, the human dimensions of the problem pose significant impediments to developing and implementing solutions to mitigate its impact. Slowing climate change will almost certainly require a reduction in greenhouse gas emissions, but agreeing upon who should reduce emissions by how much has thus far proven an intractable problem. Despite longstanding agreement that combating anthropogenic climate change is a serious issue that necessitates international cooperation (United Nations, 1992), failures such as occurred in Copenhagen in 2009 show that nations are unable to agree on what constitutes a fair sharing of the emissions reduction burden (Lange et al., 2010). In this paper we show that the difficulty of reaching agreement is, in part, due to the self-serving bias—the pervasive tendency to believe that what is beneficial to oneself is also fair. The self-serving bias, and not intrinsic cross-cultural differences in what is judged a fair distribution of sacri- fices, may represent the primary obstacle standing in the We thank Jonathan Baron, Max Bazerman, John Beshears, Hadi Dowlatabadi, Keith Florig, Sonja Klinsky and two anonymous referees for thoughtful comments and suggestions at various stages of this work. We gratefully acknowledge support from the Hewlett and Rockefeller Foundations as well as the research priority program at the University of Zurich “Foundations of Human Social Behavior.” ∗Department of Social and Decision Sciences, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA, USA, 15213. Email: pkriss@cmu.edu. †Renmin University of China ‡University of Zurich way of agreement. Alternatively, one might consider the social dilemma structure of the problem (e.g., Tavoni et al., 2011) or other strategic concerns (e.g., Finus, 2001, chapter 9) to be the largest obstacles in combating climate change. This paper takes these structural issues as givens and implicitly assumes that they are the reason formal agreements are needed in the first place. In particular, a pre-condition for getting parties to agree to self-sacrifice in pursuit of a collective solution is to reach an agreement regarding the fair distribution of burdens. This paper focuses on the role that the self-serving bias, as opposed to differing fundamental beliefs about fairness, plays in impeding such agreement. The countries that take part in climate change negotiations differ on myriad dimensions of potential relevance to a fair assignment of burdens, including their levels of past and projected emissions, the costs of mitigation, and the benefits obtained by successfully addressing the problem. Research on the self-serving bias (Messick & Sentis, 1983; Bazerman et al., 1999; Diekmann et al., 1997; Rabin, 1995; Konow, 2000) shows that asymmetries between parties on dimensions relevant to a negotiation contribute to the bias, because people place greater weight on dimensions that support a settlement that is more beneficial to themselves (Thompson & Loewenstein, 1992; Wade-Benzoni et al., 1996; Di Tella et al., 2007). The self-serving bias has been documented and shown to play a central role in negotiation impasse both in laboratory studies (Loewenstein et al., 1993; Babcock et 602

Judgment and Decision Making.Vol.6.No.7.October 2011 Bchind the veil of ignorance 603 al..1995:Carlson Russo.2001:Farmer Pecorino Americans and Chinese provide evidence of subtle dis 2002)and field studies (Babcock et al.1996:Lange et similarities.While members of the two cultures have 1,2010 Moreover, cause negou are avers th ane als be they perceive to be fair stein et al 1980)and less likelv because the self-serving bias leads to incongruous per- sider the fair option to be the best option overall (Bian of lairn there is often ettlement upor Keller,1999).In the domain of price faimes Ch ailable agre large mt g nsider their own price more unfair if tions involved in climate change negotiations makes the a better price while Americans are less sensitive to this omplexity distinction (Bolton et al2010).Gao(2009)found that ver and Chinese were mo ope serving arguments To illustrate the self-serving selection of fairness dent than tho e of Americans.To the norms, consider the following study by van Avermaet judgments influence negotiation positions,disentangling d in 1985) out que of bias m give hints as to the sort most effectively facilitate agree hemselves a their work.If told that the other person was given half ar.give n th Importanc of agre he amount of tim to work and completed ha h hias and work done.But if the other person was given twice as by fundamental cross-cultural differences in what is per ong and completed twice the number of que tionnaires ceived as fair.is of great significance.By using a ve attempt to est sh th nstead divisi rolled experimen tended to keep more than half the m y in hoth the case has on their fairness judg an ar er person was given twice as much time and that draws on the concept of the veil of ignorance first s many ques onna es and the anyi and name by John Rawl: dicates that suhiects chose tir t or work a s the fair payment criterion depe ending on which was in nle know their their own favor. ge the terests bec they do not know what role they will be In the c context of climate cha e diffe ence in fai ssigned.A classi c application of the veil of ignorance s an indictr nt of laver the dens of developing versus developed nations(Klinsky& who atabadi.2009 for which pan We explore e faimess the 0) ed in ouse in olicy tend to corre pond to their material self-interes dividuals in situations in which they are aware of their urthermore,agents are more likely to belie eve that self wn national interests or not.We presented subjects witl er p ayed current and ed pe ve ith uld he alter or conscious self-serving arguments.as lange et al dynamically with a slider that was controlled by the sub. (2010)propose).Lange et al.'s.study cannot rule ou the ject(Figure 1).which led to shifts in per capita income pility that move th eement inden xtreme that ns on lead to disa dent of the role of d all bu the pther material self interest. ty.until they found the division of sacrifice that they Indeed.comparisons of fairness judgments between deemed most fair

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 603 al., 1995; Carlson & Russo, 2001; Farmer & Pecorino, 2002) and field studies (Babcock et al., 1996; Lange et al., 2010). Moreover, because negotiators are averse to accepting settlements that are even slightly below what they perceive to be fair (Loewenstein et al., 1989), and because the self-serving bias leads to incongruous perceptions of fairness, there is often no settlement upon which all parties will agree, even when large joint gains are available from agreement. The large number of nations involved in climate change negotiations makes the problem even worse, not only by increasing complexity over and above that which would exist with only two parties but also by multiplying the number of possible selfserving arguments. To illustrate the self-serving selection of fairness norms, consider the following study by van Avermaet (reported in Messick, 1985). Subjects filled out questionnaires and were then given money which was to pay themselves as well as another person in the study for their work. If told that the other person was given half the amount of time to work and completed half as many questionnaires, subjects kept on average two-thirds of the money, an allocation proportional to both the time spent and work done. But if the other person was given twice as long and completed twice the number of questionnaires, subjects instead chose the norm of equal division and kept half the money for themselves. And finally, subjects tended to keep more than half the money in both the case where the other person was given twice as much time and filled out half as many questionnaires and the opposite case (half the time and twice the questionnaires). This indicates that subjects chose time spent or work completed as the fair payment criterion depending on which was in their own favor. In the context of climate change, the difference in fairness norms is evidenced by the focus of some policy proposals on equal entitlement (equal per capita emission allotment) and of others on relative needs and burdens of developing versus developed nations (Klinsky & Dowlatabadi, 2009). As for which parties prefer which policies, Lange et al. (2010) show that selections of fairness criteria by agents involved in international climate policy tend to correspond to their material self-interest. Furthermore, agents are more likely to believe that selfinterest influenced other parties’ selections of fairness criteria than to believe they themselves were so influenced. Though this evidence is consistent with self-serving bias (or conscious self-serving arguments, as Lange et al. (2010) propose), Lange et al.’s, study cannot rule out the possibility that there are legitimate cross-cultural differences in perceptions of what is fair, and that these differences lead to disagreement independent of the role of material self-interest. Indeed, comparisons of fairness judgments between Americans and Chinese provide evidence of subtle dissimilarities. While members of the two cultures have similar judgments of fairness in health and safety risk scenarios, Chinese subjects tend to give other goals besides fairness more weight and are less likely to consider the fair option to be the best option overall (Bian & Keller, 1999). In the domain of price fairness, Chinese consumers consider their own price more unfair if a friend was given a better price than if a stranger was given a better price, while Americans are less sensitive to this distinction (Bolton et al., 2010). Gao (2009) found that Chinese were more tolerant of the use of market power than European cultures. One general pattern is that Chinese judgments of fairness tend to be more context dependent than those of Americans. To the extent that fairness judgments influence negotiation positions, disentangling the role of self-serving bias may give hints as to the sort of interventions that can most effectively facilitate agreement. In particular, given the importance of agreement on a suitable plan to mitigate climate change, establishing whether disagreement is driven by self-serving bias, or by fundamental cross-cultural differences in what is perceived as fair, is of great significance. By using a controlled experiment, we attempt to establish the role of self-serving bias. Because we cannot randomly assign individuals to nationality and assess the effect that this has on their fairness judgments, we employ an approach that draws on the concept of the veil of ignorance first proposed by John Harsanyi and named by John Rawls (Harsanyi, 1953; Rawls, 1971; Norton & Ariely, 2011). The veil of ignorance is a judgment of social justice (fairness) made by people who do not know their own interests because they do not know what role they will be assigned. A classic application of the veil of ignorance is an indictment of slavery based on the observation that few would endorse the institution if they were unaware of whether they were the ones who would be enslaved. We explore how judgments of the fairness of allocations of economic sacrifices resulting from reductions in greenhouse gas emissions are influenced by placing individuals in situations in which they are aware of their own national interests or not. We presented subjects with graphs that displayed current and projected per capita GDP (or income) of two countries (or farmers, in one scenario). The different possible divisions could be altered dynamically with a slider that was controlled by the subject (Figure 1), which led to shifts in per capita income over time for both parties. Subjects could move the slider between one extreme that placed all burdens on one party and the other extreme that placed all burdens on the other party, until they found the division of sacrifice that they deemed most fair

Judgment and Decision Making,Vol.6.No.7,October 2011 Behind the veil of ignorance 604 We randomly assigned students from the United States and China e of three scenari t varied the cruited either at the end cy were ware own na One he the cations at ther respe ceriodesigned to show the impact of the self-serving bias.made the domain (climate change)explicit as well subjects to a computer terminal.where they completed the online survey. Each subject randomly assigned to one of also altered the nature of the environmental problem so share the of whic with th subjects would be unlikely to realize that it was about cli- problem.One context involved the U.S.A.and China mate change.A third scenario disguised the problem even the second involved two generic countries("Country A" her castingt and"Country B"),and the third involved two neighbor ing farme "and B )Ine key sec ing burdens between the parties were identicalinal three. economic state of the two parties was summarized by per capita GDP(USA/China and CountryA/B contexts) ng sc were pos: or income i o）and asso projecte strate simple ental p e that s two partes the effect of self-serving bias from the possible effects identical for each of the three contexts of cultural differences in perceptions of fairess.Doing After seeing this information.subiects were presented nelps us cle ry esta that sel serving bias plays with graphs that displayed current and projected per s to lea D The tions may be of tremendous value.Finally.our study could be atered with the slider.As the slider dentifies subpopulations that may be more was moved,the lines in the graphs indicating both coun ries (or farmers) “with sacrifice”per capita GDP(oi ncome the sh er ur nost fair To tert and c anchoring effects the 2 Method starting position of the slider was randomized across sub jects to be all the way to the left or all the way to the right The as a ran er pos te an popul were th nese students who completed the survey in in which they made a choice)or that the slider position American students who completed the survey in English. was randomized. reasons. First,students rep After submitting their decision,subjects were told omc across differen each A emograp what percentage of the burden they had a tative of the population of ither China or the U.S the party and similarity of their social and economic position enhances ome foreach of the tw the validity of comparing and averaging across their re. 2030 and 2040)The same g raph and slider were pre- co s are that were posed by the elicitation hod,and they have hey had spec he n th: the n n the table as the lines on the graph changed.After submitting the found to slider position this second time.subjects answered sev- eral attitudinal and demographic questions

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 604 We randomly assigned students from the United States and China to one of three scenarios that varied the degree to which they were aware of their own national interests—i.e., were behind the veil of ignorance. One scenario, designed to show the impact of the self-serving bias, made the domain (climate change) explicit as well as the two countries involved (China and the United States). A second scenario disguised the countries (using the generic labels “Country A” and “Country B”), and also altered the nature of the environmental problem so subjects would be unlikely to realize that it was about climate change. A third scenario disguised the problem even further, casting it as a negotiation between two neighboring farmers. Although the scenarios differed on multiple dimensions, the relative economic consequences of shifting burdens between the parties were identical in all three. Our goal in this research is not to suggest implementing a veil of ignorance in real climate negotiations as a policy recommendation (though if doing so were possible, we would support it). Our contribution is to demonstrate a simple experimental procedure that can isolate the effect of self-serving bias from the possible effects of cultural differences in perceptions of fairness. Doing so helps us clearly establish that self-serving bias plays a role in the inability of parties to reach agreement. Our findings also suggest that interventions that eliminate the self-serving bias in international climate policy negotiations may be of tremendous value.1 Finally, our study identifies subpopulations that may be more susceptible to self-serving bias, which may facilitate targeted interventions to influence opinion and allow agreement. 2 Method The survey was conducted as a randomized experiment with a 2 (Population) X 3 (Context) X 2 (Slider position) between subjects design. The two populations were Chinese students who completed the survey in Mandarin and American students who completed the survey in English. We used students for two reasons. First, students represent a relatively similar demographic across different countries. Although college students are not representative of the population of either China or the U.S., the similarity of their social and economic position enhances the validity of comparing and averaging across their responses. Second, college students are sufficiently intelligent and well educated to think about the kinds of issues that were posed by the elicitation method, and they have 1We recognize that developing such interventions is extremely challenging. A variety of methods, such as encouraging the parties to consider the problem from one another’s perspective, have been found to backfire (Caruso et al., 2006; Epley et al., 2006). credibility as a nation’s likely future policy makers. Subjects were recruited either at the end of classes or from public locations at their respective universities. We describe the recruitment procedures and monetary incentive below. Following recruitment, an experimenter guided subjects to a computer terminal, where they completed the online survey. Each subject was randomly assigned to one of three contexts, all of which involved two parties that needed to share the economic burden associated with solving a joint problem. One context involved the U.S.A. and China, the second involved two generic countries (“Country A” and “Country B”), and the third involved two neighboring farmers (“Farmer A” and “Farmer B”). The key sections of the description for each of the three contexts are presented in Table 1. In addition to this context, the economic state of the two parties was summarized by per capita GDP (USA/China and CountryA/B contexts) or income (FarmerA/B context) and associated projected growth rates for each of the two parties. This information was presented numerically as well as graphically and was identical for each of the three contexts. After seeing this information, subjects were presented with graphs that displayed current and projected per capita GDP (or income), under different possible divisions of the economic burden associated with solving the problem (Figure 1). The different possible divisions could be altered dynamically with the slider. As the slider was moved, the lines in the graphs indicating both countries’ (or farmers’) “with sacrifice” per capita GDP (or income) shifted, and subjects could move the slider until they found the division of sacrifice that they deemed most fair. To test and control for anchoring effects, the starting position of the slider was randomized across subjects to be all the way to the left or all the way to the right. Subjects were not aware that the study involved Chinese and American populations, that there were three different contexts (they were only aware of the particular context in which they made a choice) or that the slider position was randomized. After submitting their decision, subjects were told what percentage of the burden they had allocated to each party and were presented with a table indicating the numerical implications of this decision (per capita GDP or income for each of the two parties in the year 2010, 2020, 2030 and 2040). The same graph and slider were presented on this page (with the slider starting at the position they had specified on the previous page), and subjects had the opportunity to revise their decision. As they moved the slider on this page, the numbers in the table as well as the lines on the graph changed. After submitting the slider position this second time, subjects answered several attitudinal and demographic questions

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 605 Table 1: Three survey contexts. USA/China The two countries must cut greenhouse gas emissions to avoid the negative effects of climate change. Cutting emissions will necessarily reduce per capita GDP now and in the future. There are many ways that the sacrifice can be divided between the two countries. Your task will be to specify how the countries should share the substantial economic sacrifices that they need to make. Assume that the sacrifice has to be made: the question is of how it should be divided. CountryA/B On the border shared by the two countries lies a river. The two countries must build a dam on this river to avoid serious floods. Building the dam will reduce per capita GDP now and in the future. There are many ways that the sacrifice can be divided between the two countries. Your task will be to specify how the countries should share the substantial economic sacrifices that they need to make. Assume that the sacrifice has to be made: the question is of how it should be divided. FarmerA/B Uphill from the two farmers lies a stream. The two farmers must divert this stream to prevent it from running through their land, flooding their residences, and damaging their crops. Diverting the stream will cost a significant amount of money now and in the future, but will save their farms from very severe damage. There are many ways that the sacrifice can be divided between the two farmers. Your task will be to specify how the farmers should share the substantial economic sacrifices that they need to make. Assume that the sacrifice has to be made: the question is of how it should be divided. Figure 1: Graphical fairness elicitation. 2.1 Construction of Stylized Model of Relative Burdens Our goal in presenting per capita GDP for the U.S. and China both with and without the necessary sacrifices needed to combat climate change was not to develop a model that was as realistic as possible; nor was it to capture the underlying uncertainty that is unavoidable in such an endeavor. Rather, our goal was to create a simplified version of the problem that captured a few of its most

Judgment and Decision Making,Vol.6.No.7,October 2011 Behind the veil of ignorance 606 important features unchanged.It is also important to note that the only nu I de mally given to subj ects were the cu d to each party The model of relative burden te we used the historic 30-vear average accord only graphically and the potentially confusing numerica ing to data from the U.S.Bureau of Economic Analysis were not p ed to subject GD growth I very unli 、of their choice after ould have been in riate instead we referred to ing their initial decision.At this point they could revise a Goldman Sachs Global Economics Paper (Wilson their choice if they so desired. Stupnytska,2007)for projections of Chin s GDP gre of this infon ble for 2.2 Recruitment procedures subjects,we used the average projected growth rate over his time frame,and we expre ed all numbers in terms er than rea Popula Caegie Mellon University in Pitts rgh,Pennsylvania cnmin Univer were re The most controversial assumption of the model is that the onnortunity to narticinate in a survey for a fixed cash the necessary eco omic sacrifice change payment of 15 Yuan(approximately US$2.25).Interested students were takent a computer lab wh they com ple eats were not and s ernment.Because our s cenario involved only two parties. we use 1%of joint GDP as the require to particular individuals.Furthermore.dividers between d sacrifice ith these figu omputers prevented between subjects were not i ed of the of the survey i ed to subiects (Figure 1) d linearly on thi bsolute dollar amount 15 minutes nd had no imnact That is.if located one third of on their standing in the class from which they were recruited the way from the left. the U.S.A (or Country/Farme the a Approximately half the Camegie Mellon University subjects ere recruite the s To reach implications of this division on per canita gDe for in as I the come)with sacrifice would then be immediately reflected tudent center.which is centrally located on campus and n the graph Ve that this is inte d only as requented by most students.a researcher was statione a laptop computers on no attempt to incorporate other features that are relevant to quivalent in purchasing power to the 15 Yuan given to such as how Ray's b ets).One minor difference in procedure at ha vas tha ere n llow for a hical elicitatic that d require multiple dependent inputs.our study simplifies ey and a subiects complied with this instruction the negotiation process to include only tw rties.Tha the inh erves as a proof of concept that the relative roles 3 Results of self-serving bias and cultural differences in faimess udg ents can be There were 292 Renmin University subjects and 429 h. ekey property gTiondcnmnaencgotia However, vealthier hut growing more slowly and the other oorer but growing more quickly.Even substantial devi- to the self-se ations in the parameters used would leave these features

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 606 important features. We based the model on parameter estimates taken from maximally authoritative sources. Because the U.S. GDP growth is not expected to have dramatic long-term changes, we used the historic 30-year average according to data from the U.S. Bureau of Economic Analysis. Because China’s GDP growth is very unlikely to remain at recent historic levels, such an assumption for China would have been inappropriate. Instead, we referred to a Goldman Sachs Global Economics Paper (Wilson & Stupnytska, 2007) for projections of China’s GDP growth over the next 30 years. To make the presentation and interpretation of this information as simple as possible for subjects, we used the average projected growth rate over this time frame, and we expressed all numbers in terms of nominal rather than real GDP. Population growth rates for both countries were assumed to remain at their 2009 levels. The most controversial assumption of the model is that the necessary economic sacrifice required to combat climate change is 1% of world GDP. This estimate was taken from the Stern Review on the Economics of Climate Change (Stern, 2006) written for the British government. Because our scenario involved only two parties, we used 1% of joint GDP as the required sacrifice. With these figures, we calculated the absolute dollar sacrifice that must be made by the two parties. The slider presented to subjects (Figure 1) operated linearly on this absolute dollar amount. That is, if located one third of the way from the left, the U.S.A. (or Country/Farmer A) would bear two-thirds of the absolute dollar cost and China (or Country/Farmer B) would bear one-third. The implications of this division on per capita GDP (or income) with sacrifice would then be immediately reflected in the graph. We reiterate that this model is intended only as a qualitatively appropriate simplification of the economics underlying the problem. For example, our design does not attempt to incorporate other features that are relevant to fairness, such as how each party’s historic behavior has contributed to the problem (Rive et al., 2006). Also, to allow for a graphical elicitation procedure that does not require multiple dependent inputs, our study simplifies the negotiation process to include only two parties. That is, rather than attempting to capture all the inherent complexity of multi-national climate change negotiations, our design serves as a proof of concept that the relative roles of self-serving bias and cultural differences in fairness judgments can be disentangled. The key property of international climate negotiations that is reflected in our design is the fact that one party is wealthier but growing more slowly and the other party is poorer but growing more quickly. Even substantial deviations in the parameters used would leave these features unchanged. It is also important to note that the only numerical details initially given to subjects were the current per capita GDP (or income) and the growth rate for each party. The model of relative burdens was presented only graphically and the potentially confusing numerical details were not presented to subjects. To avoid overwhelming them with information, we showed subjects numerical implications of their choices only after making their initial decision. At this point they could revise their choice if they so desired. 2.2 Recruitment procedures The survey was conducted with subjects recruited from Renmin University of China in Beijing, China and from Carnegie Mellon University in Pittsburgh, Pennsylvania, U.S.A. The Renmin University subjects were recruited from large class lectures with a brief announcement of the opportunity to participate in a survey for a fixed cash payment of 15 Yuan (approximately US$2.25). Interested students were taken to a computer lab where they completed the survey online. Seats were not assigned and so it was clear to subjects that responses could not be traced to particular individuals. Furthermore, dividers between computers prevented communication between subjects. Subjects were not informed of the topic of the survey in advance, but were told that it would take approximately 15 minutes and had no impact on their standing in the class from which they were recruited. Approximately half the Carnegie Mellon University subjects were recruited by the same method. To reach the required number of subjects, however, it was necessary to use an additional recruitment procedure. In the student center, which is centrally located on campus and frequented by most students, a researcher was stationed with several laptop computers. All Carnegie Mellon University subjects were given a cash payment of $5 (roughly equivalent in purchasing power to the 15 Yuan given to Chinese subjects). One minor difference in procedure at Carnegie Mellon University was that there were no dividers between computers. However, subjects were instructed not to communicate during the survey and all subjects complied with this instruction. 3 Results There were 292 Renmin University subjects and 429 Carnegie Mellon University subjects. However, 97 of the Carnegie Mellon University subjects were not U.S. citizens. Because country of citizenship is directly relevant to the self-serving bias in these contexts, the main analysis is restricted to the 332 Carnegie Mellon University

Judgment and Decision Making.Vol.6.No.7.October 2011 Behind the veil of ignorance 607 Table 2:Final ratio (percent burden on China/B). 3.2 Fairness Judgments US China t-test The american and chine 41.69 in a manner consistent with the self-serving bias.How USA/China t= -6.17 p<.0001 r Judgments of fair divisic when 330 3357 the a mild =-015 CountryA/B (18.85 in the usa/china context did not result from inheren n=125 m=106 cultura differences.In the USA/China 3791 34.76 subje China s FarmerA/B (20.63) t=-1.04 (20.27 p=.30 ear 26 36 of the cost (t n=99 =84 ecs6.17, <000.How ever in the CountryA/B context,American subjects spec Standard errors in parentheses that country B ch was equivalent to Chir he nst 336 Table 3:Nominal per capita GDP (or income)in 2040. These mean results are presented in Table 2.while the US mplications of these judgments on nominal pe capita China e)in USA/China$162,924/S39,854 $150.680/S44.378 (on NOVAr 204 are present ulation and co Country A/B$156.730/s39.854$156.451/s42.246 t to be highly significant F(1,437)=18.49.p<.0001 Farmer A/B$159.910/S40.967$157.399/s41.895 As for the Carnegie Mellon University subjects who were not U.S citize mean judgment of den o Uny t =2 0205)and higher than that of chinese citi ens (t 2.05.D<.05) Their mean judgments were he (38.12 3.1 Demographic and Attitudinal Data =-104=30 andt= -0.91,p=37,e each lation wer tivelv). A more detailed breakdown of the distribution of re. in the U.S.).Details are provided in Tables 4 and 5 in the sented in Fig hat the diffe n addition to de ographic informtion,subjects were can and Chinese subiects is driven not by a small number differences between the two populations are in their re sponses to the environmental questions Chinese sub To examine the data in more depth and to control s are le th the c y of air,the e vanous de s between the or area in which they live.Furthermore.we on the second party (China.Country Bor Farmer Bas Chinese subjects to be hesitant to report low patriotism. the dependent va riable.The main insight from these re- s re n,we uded er q that we s litie ssions is t rving bias is very robu hoth ga st in the bling and pre-marital ugh both are in Furthermore there is evidence of what could be called fact commonplace. Though opposing the Chines motivated anchoring"on the part of Chinese subjects t vie nese ttitudinal a ions are resented in Tables 6 and 7 of when they did not have to move the slider very far to the Appendix. do so.This effect is not present for American subjects

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 607 Table 2: Final ratio (percent burden on China/B). US China t-test USA/China 41.69 (20.53) n = 108 26.33 (14.88) n = 102 t = −6.17 p < .0001 CountryA/B 33.92 (18.85) n = 125 33.57 (18.04) n = 106 t = −0.15 p = .88 FarmerA/B 37.91 (20.27) n = 99 34.76 (20.63) n = 84 t = −1.04 p = .30 Standard errors in parentheses Table 3: Nominal per capita GDP (or income) in 2040. US China USA / China $162,924 / $39,854 $150,680 / $44,378 Country A / B $156,730 / $39,854 $156,451 / $42,246 Farmer A / B $159,910 / $40,967 $157,399 / $41,895 subjects who are also U.S. citizens and the 292 Renmin University subjects (all of whom were Chinese citizens). 3.1 Demographic and Attitudinal Data The vast majority of subjects from each population were between 18 and 20 years old (76.7% in China and 77.4% in the U.S.). Details are provided in Tables 4 and 5 in the Appendix. In addition to demographic information, subjects were asked several questions relating to global warming, the environment, patriotism and social issues. The biggest differences between the two populations are in their responses to the environmental questions. Chinese subjects are less satisfied with the quality of air, the quality of water and efforts to preserve the environment in the city or area in which they live. Furthermore, we expected Chinese subjects to be hesitant to report low patriotism. For this reason, we included other questions that we suspected would correlate with their political attitudes. In particular, the Chinese government opposes both gambling and pre-marital sexual activity, though both are in fact commonplace. Though opposing the Chinese government directly is forbidden, holding divergent views on these topics is more acceptable. Detailed results for these attitudinal questions are presented in Tables 6 and 7 of the Appendix. 3.2 Fairness Judgments The American and Chinese subjects strongly disagreed on the fair division of burden in the USA/China context, in a manner consistent with the self-serving bias. However, their judgments of what constitutes a fair division were virtually identical when behind a mild veil of ignorance. Thus, the different perceptions of what is fair in the USA/China context did not result from inherent cultural differences. In the USA/China context, American subjects proposed that China should bear 41.7% of the cost while Chinese subjects specified that they should bear 26.3% of the cost (t = −6.17, p < .0001). However in the CountryA/B context, American subjects specified that country B (which was equivalent to China in the first scenario) should bear 33.9% of costs and Chinese subjects specified that country B should bear 33.6%. These mean results are presented in Table 2, while the implications of these judgments on nominal per capita GDP (or income) in the year 2040 are presented in Table 3. A two-way ANOVA reveals this interaction between population and context to be highly significant, F(1, 437) = 18.49, p < .0001. As for the Carnegie Mellon University subjects who were not U.S. citizens, their mean judgment of the fair burden on China in the USA/China context was 33.23% (n = 31), which is lower than that of U.S. citizens (t = 2.02, p < .05) and higher than that of Chinese citizens (t = 2.05, p < .05). Their mean judgments were not significantly different in the CountryA/B (38.12%, n = 33) or FarmerA/B (38.67%, n = 33) contexts (t = −1.04, p = .30 and t = −0.91, p = .37, respectively). A more detailed breakdown of the distribution of responses of U.S. and Chinese citizens is presented in Figure 2. Note that the difference between the USA/China context and the CountryA/B context for both the American and Chinese subjects is driven not by a small number going to extremes, but was produced by a more incremental mean-shift of each population. To examine the data in more depth and to control for the various demographic differences between the two populations, we ran regressions with the percent burden on the second party (China, Country B or Farmer B) as the dependent variable. The main insight from these regressions is that the self-serving bias is very robust to statistical controls in the USA/China scenario, and the lack of difference is equally robust in the other two scenarios. Furthermore, there is evidence of what could be called “motivated anchoring” on the part of Chinese subjects. That is, only in the USA/China context, Chinese subjects were more likely to propose divisions that favored China when they did not have to move the slider very far to do so. This effect is not present for American subjects

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 608 Figure 2: Distribution of fairness judgments by context and population. F(1, 194) = 0.01, p = 0.92. The fact that this effect is found only for the USA/China scenario supports the idea that the bias is motivated. That is, subjects did not exhibit such anchoring when behind the veil of ignorance. Table 8 in the Appendix presents these models broken down by context and also for all three contexts combined. For a more detailed examination of the predictors of responses in the USA/China scenario, where we observe self-serving bias, we conducted regressions separately for each population and including all of the attitudinal variables (Tables 9 and 10 in the Appendix). The main insights gained from the regressions of U.S. citizen responses are that Democrats are less self-serving and those who are very patriotic are more self-serving. As for Chinese citizens, the self-serving bias is stronger for males and stronger for those who study natural sciences. Also, the tendency for Chinese subjects to give more selfserving judgments when the slider begins on the end of the scale favoring China is robust to controls. Finally, we ask whether agreement between the two countries increases when subjects are placed behind the veil of ignorance. A statistic to measure potential agreement is the proportion of pairs of U.S. and Chinese subjects whose judgments are compatible. That is, we calculate the probability that a random U.S. subject and a random Chinese subject in our sample give fairness judgments that would require no additional concessions by either side for an agreement to be achieved. Assuming two continuous and identical distributions of responses—i.e., the judgments of what is fair are identically distributed across countries—then the maximum possible agreement frequency would be .5. We therefore construct an index of agreement relative to this theoretical maximum and express the result on a zero to one scale—an index score of zero indicates agreement between the two populations never occurs, while a score of 1 indicates that subjects from one country are as likely to agree with subjects from the other country as with those from their own (agreement frequency of .5). In the USA/China context, the index of agreement between US-China pairs is a modest .54. However, in the CountryA/B context, the index of agreement between US-China pairs is .95. The index of agreement for the FarmerA/B context is also high (.87). This improvement in the veil of ignorance context is quite dramatic and suggests that, once we remove the nationalistic self-serving bias, agreement may occur with very high frequency. 4 Discussion We first demonstrated that differing perceptions of fairness pose a problem for obtaining international agreement on the distribution of the economic burden of mitigating climate change. In judging what is “fair,” subjects from the U.S.A. and from China provided very different responses. Both groups’ ideas of fairness tended towards their own national interests. The likely resulting impasse mirrors the outcome of recent high-level negotiations, such as in Copenhagen in 2009

Judgment and Decision Making.Vol.6.No.7.October 2011 Bchind the veil of ignorance 609 In light of the large disagreement between Americar reflect a self-serving bias.Ouarterly Journal of Eco and Chinese subjects in the USA/China context,the most nomics.111.1-19. re.D..&Gillespie.JJ.(199 Country B"and changin the problem from climate change to a different but analogous situation.We argue hat th f-serving bias as the driving force opp anticipated.our results indicate that common ground in Risk analesis cating the burden of mitigating climate change is pos- Bolton,L.E.,Keh,H.T.Alba,J.W.(2010).How do can somehow be immunized agains price The fact that no de nhics affected ments of A fairness in the two veil of ignorance conditions (with the tation of evidence by mock iur ossible exception of gender for American subjects,bu mental Psychology:Applied.7.91-103. only in the neigh f thg fa ext)further po M.(0)The Caruso.E. were identified and the context was familiar did varia it c respons an tions in demographics matter,and they mattered in a fash ocial Psychology.91.857-871. on consistent with th e groups (e.g..those Di Tella,R.,Galiani,S .Schargrodsky,E.(2007) are more pat 10 of beliefs:Evide n the alloc land Qua serving bias also raises the possibility that these groups could be targeted in any intervention designed to allevi- Diekmann,K.A.,Samuels.S.M..Ross,L..Bazerman ate the bi not h M.H.(1997).Self-interest and fairness in problems of of this ha n to e fo pose the veil of 1074 ignorance as a feasible intervention in ral cl ate negoti- Epley.N..Caruso,E.,Bazerman,M.H.(2006).When tions.Ins oudbe a broa Moreover.by 91.872 9 al时PeTs using an experiment in which our manipulated treatment Farmer.A.&Pecoring.P (2002).Pretrial bargain- the presenc of sel ing with self-serving bias and asymmetric information. of Economic Behavior and Organization,48 bevond existing research in more cleanly demonstrat Finus.M.(2001).Game the and int ing the impact of self-serving bias as an impediment to mental cooperation (w E Oates H Folmer Eds ing agre study gves and Resoure Economics.Chelenham. UK h mpton. A:Edward Elgar red can o in Chin nternational agreement on greenhouse gas emissions re Switzerland and Ca duction and climate policy. and gender matter?Judgment and Decision Making.4. 214-22 3).0 References welf EcOm0m6L.434435. Babcock,L.Loewenstein.G.Issacharoff.S.& Klinsky.S..&Dowlatabadi.H.(2009).Conceptualiza Camerer,C.( 995 Bi ed judgn 0 juic in climate policy. (2000)Eair Babcock.L..Wang.X..Loewenstein.G.(1996). tive dissonance in allocation deci Choosing the wrong pond:Social comparisons that nomic Review,90.1072-1091

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 609 In light of the large disagreement between American and Chinese subjects in the USA/China context, the most dramatic finding is that this difference is completely eliminated by merely labeling the parties “Country A” and “Country B” and changing the problem from climate change to a different but analogous situation. We argue that this points to self-serving bias as the driving force behind disagreement, as opposed to fundamental cultural differences in beliefs about what is fair. Moreover, as we anticipated, our results indicate that common ground in allocating the burden of mitigating climate change is possible if negotiators can somehow be immunized against the self-serving bias. The fact that no demographics affected assessments of fairness in the two veil of ignorance conditions (with the possible exception of gender for American subjects, but only in the neighboring farmers context) further points to the robustness of the methods. Only where countries were identified and the context was familiar did variations in demographics matter, and they mattered in a fashion consistent with the values of those groups (e.g., those who are more patriotic). The fact that certain subpopulations exhibit greater tendency to be influenced by selfserving bias also raises the possibility that these groups could be targeted in any intervention designed to alleviate the bias. The point of this exercise has not been to estimate the fair distribution of greenhouse gas mitigation burden for the U.S. and China. Nor has it been to propose the veil of ignorance as a feasible intervention in real climate negotiations. Instead, it is to suggest that there could be a broad consensus about the fair distribution of burdens between different parties in this kind of negotiation. Moreover, by using an experiment in which our manipulated treatment variable affects the presence of self-serving judgments, we show that disagreement is not due to ingrained cultural differences in beliefs about fairness. We therefore go beyond existing research in more cleanly demonstrating the impact of self-serving bias as an impediment to reaching agreement. Our study gives reason for optimism that interventions to mitigate self-serving bias, if discovered, can greatly improve our prospects of reaching an international agreement on greenhouse gas emissions reduction and climate policy. References Babcock, L., Loewenstein, G., Issacharoff, S., & Camerer, C. (1995). Biased judgments of fairness in bargaining. American Economic Review, 85, 1337– 1343. Babcock, L., Wang, X., & Loewenstein, G. (1996). Choosing the wrong pond: Social comparisons that reflect a self-serving bias. Quarterly Journal of Economics, 111, 1–19. Bazerman, M. H., Moore, D. A., & Gillespie, J. J. (1999). The human mind as a barrier to wiser environmental agreements. American Behavioral Scientist, 42, 1277– 1300. Bian, W. Q., & Keller, L. R. (1999). Chinese and Americans agree on what is fair, but disagree on what is best in societal decisions affecting health and safety risks. Risk Analysis, 19, 439–452. Bolton, L. E., Keh, H. T., & Alba, J. W. (2010). How do price fairness perceptions differ across culture? Journal of Marketing Research, XLVII, 564–576. Carlson, K. A., & Russo, J. E. (2001). Biased interpretation of evidence by mock jurors. Journal of Experimental Psychology: Applied, 7, 91–103. Caruso, E., Epley, N., & Bazerman, M. H. (2006). The costs and benefits of undoing egocentric responsibility assessments in groups. Journal of Personality and Social Psychology, 91, 857–871. Di Tella, R., Galiani, S., & Schargrodsky, E. (2007). The formation of beliefs: Evidence from the allocation of land titles to squatters. Quarterly Journal of Economics, 122, 209–241. Diekmann, K. A., Samuels, S. M., Ross, L., & Bazerman, M. H. (1997). Self-interest and fairness in problems of resource allocation: Allocators versus recipients. Journal of Personality and Social Psychology, 72, 1061– 1074. Epley, N., Caruso, E., & Bazerman, M. H. (2006). When perspective taking increases taking: Reactive egoism in social interaction. Journal of Personality and Social Psychology, 91, 872–889. Farmer, A., & Pecorino, P. (2002). Pretrial bargaining with self-serving bias and asymmetric information. Journal of Economic Behavior and Organization, 48, 163–176. Finus, M. (2001). Game theory and international environmental cooperation. (W. E. Oates & H. Folmer, Eds.). Environmental and Resource Economics. Cheltenham, UK and Northhampton, MA: Edward Elgar. Gao, Y. (2009). A study of fairness judgments in China, Switzerland and Canada: Do culture, being a student, and gender matter? Judgment and Decision Making, 4, 214–226. Harsanyi, J. (1953). Cardinal utility in welfare economics and in the theory of risk-taking. Journal of Political Economy, 61, 434–435. Klinsky, S., & Dowlatabadi, H. (2009). Conceptualizations of justice in climate policy. Climate Policy, 9, 88–108. Konow, J. (2000). Fair shares: Accountability and cognitive dissonance in allocation decisions. American Economic Review, 90, 1072–1091

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 610 Lange, A., Löschel, A., Vogt, C., & Ziegler, A. (2010). On the self-interested use of equity in international climate negotiations. European Economic Review, 54, 359–375. Loewenstein, G., Issacharoff, S., Camerer, C., & Babcock, L. (1993). Self-serving assessments of fairness and pretrial bargaining. Journal of Legal Studies, XXII, 135–159. Loewenstein, G., Thompson, L., & Bazerman, M. (1989). Social utility and decision making in interpersonal contexts. Journal of Personality and Social Psychology, 57, 426–441. Messick, D., & Sentis, K. (1983). Fairness, preferences, and fairness biases. In D. M. Messick, Cook, K.S. (Ed.), Equity Theory. New York: Praeger. Messick, D. M. (1985). Social interdependence and decision making. In G. Wright (Ed.), Behavioral decision making. New York: Plenum Press. Norton, M., & Ariely, D. (2011). Building a better America — one wealth quintile at a time. Perspectives on Psychological Science, 6, 9–12. Rabin, M. (1995). Moral preferences, moral constraints, and self-serving biases. University of California at Berkeley Working Paper No. 95–241. Rawls, J. (1971). A Theory of Justice. Cambridge MA: Harvard University Press. Rive, N., Torvanger, A., & Fuglestvedt, J. S. (2006). Climate agreements based on responsibility for global warming: Periodic updating, policy choices, and regional costs. Global Environmental Change, 16, 182– 194. Stern, N. (2006). Stern Review on the Economics of Climate Change. Cambridge UK: Cambridge University Press. Tavoni, A., Dannenberg, A., Kallis, G., & Löschel, A. (2011). Inequality, communication, and the avoidance of disastrous climate change in a public goods game. Proceedings of the National Academy of Sciences of the United States of America, 108, 11825–11829. Thompson, L., & Loewenstein, G. (1992). Egocentric interpretations of fairness and interpersonal conflict. Organizational Behavior and Human Decision Processes, 51, 176–197. United Nations. (1992). United Nations Framework Convention on Climate Change. Retrieved from http://unfccc.int/resource/docs/convkp/conveng.pdf Wade-Benzoni, K. A., Tenbrunsel, A. E., & Bazerman, M. H. (1996). Egocentric interpretations of fairness in asymmetric, environmental social dilemmas: Explaining harvesting behavior and the role of communication. Organizational Behavior and Human Decision Processes, 67, 111–126. Wilson, D., & Stupnytska, A. (2007). The N-11: More than an acronym. Global Economics Paper, 153. New York: Goldman Sachs Group. Appendix Table 4: Race by population. USA China White 43.7% Han 92.1% Black/African American 6.9% Zhuang 0.7% Asian 36.1% Manchu 1.4% Indian/Pacific Islander 5.7% Hui 0.7% Hispanic 3.9% Miao 1.0% Multiracial 3.3% Uyghurs 0.3% Other 0.3% Tujia 0.3% Yi 0.3% Mongols 1.4% Other 1.7%

Judgment and Decision Making, Vol. 6, No. 7, October 2011 Behind the veil of ignorance 611 Table 5: Demographics by population. USA China Difference between distributions Age 18–20 77.4% 76.7% χ 2 = 9.3 21–25 19.0% 23.0% p < .05 26–35 2.7% 0.3% 35+ 0.9% 0.0% Sex Male 56.3% 42.8% χ 2 = 11.4 Female 43.7% 57.2% p = .001 Hometown Urban 29.8% 64.0% χ 2 = 97.5 Semi-urban 58.4% 20.2% p < .001 Rural 11.8% 15.8% Major Humanities 8.5% 4.8% χ 2 = 104.9 Social Sciences 25.7% 66.6% p < .001 Natural Sciences 65.9% 28.6% Political affiliation∗ Democrat 37.4% Republican 12.4% Independent/Other 20.8% None 29.5% ∗ Due to the lack of political freedom in China, political affiliation is not an applicable question. Table 6: Patriotism and social issues items. Some people talk about patriotism being “love for a country”. How patriotic do you yourself feel toward [the USA/China]? USA China Not at all patriotic 6.1% 0.4% χ 2 = 87.2 Not very patriotic 25.6% 4.5% p < .0005 Fairly patriotic 54.6% 60.8% Very patriotic 13.7% 34.4% Which of the following best describes your feelings toward gambling? USA China It should be illegal 2.1% 36.9% χ 2 = 188.3 It should be discouraged but legal 39.2% 49.8% p < .0005 It should be neither encouraged nor discouraged 55.3% 13.2% It should be encouraged 3.3% 0.0% Which of the following best describes your feelings toward pre-marital sexual activity? USA China It should be illegal 0.9% 4.2% χ 2 = 82.5 It should be discouraged but legal 22.7% 54.9% p < .0005 It should be neither encouraged nor discouraged 68.1% 38.4% It should be encouraged 8.3% 2.5% To what extent do you believe that governments should strive for equality, that is they should attempt to make everyone in a society equally well off? USA China It is not something that government should try to do 28.6% 7.9% χ 2 = 84.8 It is somewhat important 46.7% 42.8% p < .0005 It is very important 20.5% 48.3% Don’t know 4.2% 1.0% Note: The patriotism question was taken from a survey commissioned by the BBC. (http://www.icmresearch.co.uk/pdfs/2002_december_bbc_radio_4_patronism_poll.pdf)