Table 1.
DUTCH measures range on a scale from one to five and measure individual conflict handling tendencies.
NEP-R score can range from one to five. Scores closer to five imply a more pro-environmental attitude. Content refers to the cumulative number of statements made by a subject during communication sessions in the experiment and coded to the listed categories. A randomization check was conducted for the observable controls (“Demographics;” “DUTCH Measures;” “Environmental Preferences”). Bold variable names indicate a statistically significant difference between treatment groups (p ≤ 0.05). We conclude that, on average, there is no statistical difference between our treatment groups, except in the number of previous experiments. No t-tests were conducted for the “Content” variables as these measures are not based on the randomization.
Fig 1.
This figure describes the main stages of the experiment.
We randomly assign groups to either interests or position treatments. Groups are given instructions and then provided with a 20 minute cheap talk communication session. Then, groups engage in a binding non-verbal offer and acceptance session in which a group decision is unanimously signed onto, or the controller picks the outcome. Last, the points allocated during negotiations are turned into lottery tickets and the lottery drawing occurs and subjects are paid accordingly.
Table 2.
List of interests vs. positions were provided to subjects who were prompted to use the terms in negotiations during communication sessions.
Each term was printed on a card to help facilitate the use of these frames.
Table 3.
List of the experimental payouts for each group option by subject assigned role.
The numbers reported here directly transform into probabilities of winning the lottery cash prize at the end of the experiment. Subjects were provided with information only about their own payouts.
Fig 2.
Fig 2a shows economic efficiency of group decisions by treatment. Fig 2b shows the average side payments made by role by treatment. Standard errors are shown at 5% with 27 observations in each treatment in each role.
Fig 3.
This figure shows the relative frequency of the content of various communication used by subjects during the communication phase of the experiment.
Fig 4.
Effect of communication on value creation.
The estimates from an OLS model are estimated and the marginal effects of subject’s use of communication is measured against bargaining efficiency, defined in Eq 1. N = 162 and standard errors are clustered at the group level. The confidence intervals on the figure are calculated at 95%. These results are estimated without additional controls, models with additional controls are available in Table J in S1 File.
Fig 5.
Effect of communication on side payments.
The estimates from a linear OLS model are estimated and the marginal effects of subject’s use of communication is measured against the associated side payment of the subject. N = 162 and standard errors are clustered at the group level. The confidence intervals on the figure are calculated at 95%. These results are estimated without additional controls, models with additional controls are available in Table C in S1 File.
Table 4.
Mediating role of language on side payments.
Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01.