Could climate change liability help to strike a climate agreement?: An economic experiment

Transcription

1 Could climate change liability help to strike a climate agreement?: An economic experiment Elisabeth Gsottbauer USF, Osnabrück & EURAC, Bozen Joint work with Robert Gampfer, Elizabeth Bernold and Anna-Mateja Delas (ETH Zurich) 13th May, 2015

2 Loss & Damage: the actual and/or potential manifestation of impacts associated with climate change in developing countries that negatively affect human and natural systems, including impacts from extreme events (for example: heatwaves, flooding and drought) and slow- onset events (including sea-level rise and glacial retreat) Decision 2/CP.19: Warsaw International Mechanism for Loss and Damage Associated with Climate Change Impacts FCCC/CP/2013/10/ Add.1 (UNFCCC, 2013)

3 Claims about responsibility for past greenhouse gas emissions are one of the central reasons for the failure of recent climate negotiations Claims of responsibility are based on legal and ethical principles concerning liability fro wrongdoing. They are, in essence, tort claims. Weisbach (2012, p522), Iowa Law Review The possibility that a small island state, or another injured party, would bring a liability claim against states responsible for climate change no longer is a topic for fiction or a theoretical prospect. There is a rise in plans for litigation worldwide for consequences of global warming. Faure and Nollkaemper (2007, p ), Stanford Environmental Law Journal

4 Motivation: Loss & Damage from CC Loss (non-recoverable) and damage (recoverable) from extreme weather or slow-onset events Natural catastrophes nearly doubled from around 400 in 1980s up to 880 in in 2013 economic loss and damage from disasters totaled around US$ 125 billion (Munich Re) At least 60 trillion USD between 2000 and 2020 even with adaptation (PAGE model estimate, Hope 2009) Originate in the idea that rich nations can be held liable for climate change-related damages based on their historical responsibility for carbon emission climate justice

5 Motivation: Loss & Damage in the negotiations Developing countries are pushing negotiations for a mechanism to deal with L&D including compensation rules Highly controversial, treated with great caution First mentioned in Cancún 2010; decision to establish institution in Doha 2012; 2013: Warsaw International Mechanism for Loss and Damage associated with Climate Change Impacts Enhance knowledge, risk management Strengthen dialogue, coordination Provide support, including finance, technology and capacity-building

6 Climate Change Litigation based on Tort Law Torts are wrongdoings that are done by one party against another The injured person may take civil action against the other party to recoup compensation Usually involves restoring the injured party monetarily

7 Liability Rules for Climate Damages Framework for allocation of responsibility (based on tort and civil liability regimes): Strict liability Countries responsible for damages (high emitters) have to compensate victim countries for damages suffered Negligence liability Compensation only if no due care by responsible country (e.g., no mitigation efforts) Contributory negligence Compensation only if due care taken also by potential victim countries (through adaptation or mitigation efforts)

8 Previous Experimental Research Some experiments on the efficiency of different liability rules on preventing damage showing mixed results Kornhauser and Schrotter (1990) show that negligence outperforms strict liability Angelova et al. (2012) finds that strict and negligence are equally effective Only one experiment testing the effect of liability in public goods setting (social dilemma) Eisenberg and Engel (2014) show that liability rules are able to stop downward trend of cooperation in a repeated public goods game Extensive experimental literature on the deterrent effect of legal institutions on the provision of public goods (e.g., Galbiati and Vertova, 2014); yet none focusing on effectiveness of liability

9 Aims & Research Questions To investigate the effects of liability rules on cooperative behavior in a realistic setting with (environmental) damages How do liability rules per-se effect cooperation? To explore a behavioral interpretation (social preferences, beliefs, trust, etc.) of the possible effect of liability on cooperation What are the channels through which liability incentives affect behavior? To examine the robustness of our findings in a setting with study participants from countries reflecting modeled wealth and vulnerability differences Are individuals with heterogeneous background more or less likely to cooperate under set liability rules?

10 Experimental Framework: Prisoner s Dilemma Benchmark case (control): A Rich (R) with high and a Poor (P) country with low endowment P can be potentially hit by climate change and loose 180 Both can invest (R) more and (P) less to reduce probability of occurrence If no one invests: 75% P invests: 65% R invests: 25% Both invest: 5% Poor Not invest (defect) Invest (cooperat) Not invest (defect) Rich Invest (cooperate) Risk 5% Risk 25% Risk 65% Risk 75% PNE={D, D}, SO={C, C}

11 Introducing Liability Rules Strict liability: rich player fully compensates poor player s damages if climate catastrophe occurs Negligence: rich player compensates poor player only if rich player has not invested in climate protection Contributory negligence: rich player compensates poor player only if poor player has invested in climate protection

12 Treatments & Expectations Poor a Baseline (BL) game Not Invest (defect) Invest (cooperate) Not Invest (defect) Rich Invest (cooperate) 65 / / / / 180 b Strict liability (SL) game Not Invest (defect) Invest (cooperate) Not Invest (defect) Rich Invest (cooperate) 200 / / / / 171 PNE={D, D}, SO={C, C} PNE={D, D}, SO={C, C} Poor c Negligence (N) game Not Invest (defect) Invest (cooperate) Not Invest (defect) Rich Invest (cooperate) 200 / / / d Contributory negligence (CN) game Not Invest (defect) Invest (cooperate) Not Invest (defect) Rich Invest (cooperate) 65 / / / PNE=SO={C, C} PNE=SO={C, C}

13 Experimental Protocol Online experiment via Amazon Mechanical Turk - marketplace for work increasingly used for research Sample (N) 758 US participants (Study 1) Random assignment to the role of a Rich or Poor player All subjects participate in baseline game and consequently play one of the treatments (253 in SL; 256 in NL; 249 in CN) Players indicate choice (invest or do not invest) as well as expectations about their partners investment Subjects also completed an incentivized social value orientation test (SVO) and a short socio-economic survey Monetary incentives: random game payout, payoffs between 1.40 and 2.65 USD including fixed show-up fee

14 Cooperation by Treatments BL SL N CN As expected, cooperation highest in N and CN

15 Behavioral Channels We find that both beliefs and preferences as stable predictors of cooperation across treatments Generalized trust towards other individuals emerged as an additional predictor for participants response, particular in in SL Sensitivity analysis of social preference data from SVO slider measure shows that cooperators consistently have higher scores that non-cooperators across all treatments BL SL N CN Non-Cooperators Cooperators Social preferences support cooperation irrespective of which liability regime being in place

16

17 Ad-on experiment: India vs. US participants Sample (N) 373: 186 US and 187 Indian participants (Study 2) Recruitment through online labor market AMT Indian participants: Poor players; US participants: Rich players Players informed about the nationality of their counterparts Role assignment remained fixed throughout the experiment; all participants made decision in a baseline game and one treatment Does difference in national/cultural background affect decisions?

18 Results Study US only in blue, Study US&India in red ** BL SL N CN BL SL N CN Rich Poor Differences in background effect cooperation in SL Significantly fewer poor players (India) are willing to exploit unconditional compensation entitlement, even if negligence is not sanctioned

19 Behavioral Channels Again, beliefs and social preferences among the strongest predictors of cooperation across all treatments In contrast to Study 1, in Study 2 we find evidence between the positive effects of social preferences being different across liability regimes BL* SL* N CN Non-Cooperators Cooperators Note: while non-cooperators and cooperators significantly differ in the degree of social preference in the US population (14.38 vs , p=0.00,ttest), Indian players don t (14.76 vs , p=0.23)

20 Summary Liability mechanism can improve climate cooperation and climate protection ambition Design of liability matters: negligence and contributory negligence most successful need incentives for both parties (Barrett, 2004) Often no actual compensation, because climate risk successfully reduced -> might increase political feasibility Institutionalizing rules for L&D could aid to break negotiation deadlock

21 Appendix

22 Table S1. Clustered probit regression on cooperation (investment) in Study 1 Variable Cooperate (invest) in Study 1 (1) (2) Strict liability 0.404** 0.443* (0.152) (0.185) Negligence 1.069** 1.139*** (0.168) (0.192) Contributory negligence 0.983*** 1.167*** (0.164) (0.188) Poor player (=1) 0.846*** (0.134) Belief partner ± (1=invest) 0.629*** (0.141) Social preference (angle) *** ( ) Trust ** (0.0267) Loss aversion (0.0446) Age (y) ( ) Female(=1) (0.133) Education dummies Yes Failed comprehension (1=yes) (0.543) Constant 0.423*** (0.0743) (1.030) No Observations * p<0.05, ** p<0.01, *** p<0.001; standard errors are in parentheses

23 Table S2: Clustered probit regression on cooperation (investment) in Study 2 Variable Cooperate (invest) in Study 2 (1) (2) Strict liability 0.474* 0.685* (0.223) (0.271) Negligence 0.633** 0.754* (0.245) (0.296) Contributory negligence 0.844*** 0.964*** (0.237) (0.264) Indian or poor player (=1) 1.220*** (0.233) Belief partner 0.513* (0.220) Social preference (angle) *** ( ) Trust (0.0359) Loss aversion (0.0571) Age (y) * ( ) Female (=1) (0.169) Education dummies Yes Failed comprehension (1=yes) (0.426) Constant 0.532*** *** (0.107) (0.766) No. observations * p<0.05, ** p<0.01, *** p<0.001; standard errors are in parentheses

24 Table S3: Clustered probit regression on cooperation (investment) by treatment for (Study 1 and 2) Variabel Cooperate (invest) in Study 1 and 2 Baseline Strict liability Negligence all subjects Contributory negligence (1) (2) (3) (4) (5) (6) (7) (8) Indian or poor player (=1) 1.498*** 2.069*** *** *** * 1.198*** 1.327*** (0.161) (0.204) (0.302) (0.342) (0.326) (0.342) (0.347) (0.377) Study 2(=1) (0.181) (0.186) (0.379) (0.409) (0.436) (0.509) (0.353) (0.371) Player x Study * 1.897** (0.294) (0.359) (0.503) (0.620) (0.558) (0.680) (0.579) (0.677) Belief partner 0.905*** 1.166*** 1.393*** 1.010** (0.183) (0.292) (0.309) (0.329) Social preference (angle) *** *** ** ** ( ) ( ) ( ) ( ) Trust * (0.0325) (0.0564) (0.0672) (0.0566) Loss aversion (0.0507) (0.0911) (0.0973) (0.108) Age (y) ** ( ) (0.0117) (0.0137) (0.0142) Female (=1) (0.150) (0.266) (0.300) (0.284) Education dummies *** 10.90*** Failed comprehension (1=yes) * (0.553) (0.766) (0.701) (0.750) Constant 0.274** * 1.639*** *** 11.14*** 0.899*** 10.67*** (0.104) (1.107) (0.245) (1.246) (0.243) (1.060) (0.201) (1.445) No. of observations * p<0.05, ** p<0.01, *** p<0.001; standard errors are in parentheses

25 Study 1 (US; N=754) UNFCCC (N=18) 0.05 Proportion SVO angel from slider measure Study 2 (US; N=186) Study 2 (India; N=187).05 Proportion Proportion SVO angel from slider measure.05 Proportion SVO angel from slider measure SVO angel from slider measure Figure S4: Distribution of SVO scores for different samples