Write two one-page reflection reports

Anonymous
timer Asked: Nov 28th, 2018
account_balance_wallet $20

Question description

Please write referee reports for each of the paper i uploaded under. Both of them are one page with single spaced.

In the report, you should include the following sections:

• Motivation • Research question • Background and literature • Data • Methodology • Findings and policy implication • Your comments and critiques

Thank you

Dora L. Costa Matthew E. Kahn University of California, Los Angeles University of California, Los Angeles Abstract “Nudges” are being widely promoted to encourage energy conservation. We show that the popular electricity conservation “nudge” of providing feedback to households on own and peers’ home electricity usage in a home electricity report is two to four times more effective with political liberals than with conservatives. Political conservatives are more likely than liberals to opt out of receiving the home electricity report and to report disliking the report. Our results suggest that energy conservation nudges need to be targeted to be most effective. (JEL: Q41, D03, D72) 1. Introduction Europe, especially Scandinavia, has high taxes on electricity and gasoline to encourage conservation and counter global warming. Taxes in Denmark represent more than half of the cost of electricity to consumers.1 In contrast, the United States has low taxes and little political will to sacrifice for the sake of conservation. Congressional voting patterns highlight that conservative Representatives are highly unlikely to vote for carbon mitigation legislation (Cragg et al. 2013). Facing political gridlock in the Congress and concerned about the challenge of climate change, an ongoing policy agenda is seeking out alternative strategies for encouraging conservation. Recent psychology research suggests an alternative tool for changing household behavior is to focus on well crafted messages offering peer comparisons (see Griskevicius, Cialdini, and Goldstein 2008). Robert Cialdini and his coauthors have conducted a series of field experiments that have demonstrated that lowcost persuasion strategies or “nudges” can change an individual’s behavior by making The editor in charge of this paper was Stefano DellaVigna. Acknowledgments: We thank Maximilian Auffhammer, the participants at the 2010 POWER Conference, and seminar participants at Princeton and the University of Illinois for comments. We thank the UCLA Ziman Real Estate Center for funding. We thank the editor and five reviewers for their comments. Costa and Kahn are Research Associates at NBER. E-mail: costa@econ.ucla.edu (Costa); mkahn@ioe.ucla.edu (Kahn) 1. See Eurostat News Release, 75/2010, 28 May 2010. http://epp.eurostat.ec.europa.eu/cache/ITY_ PUBLIC/8-28052010-AP/EN/8-28052010-AP-EN.PDF Journal of the European Economic Association June 2013 c 2013 by the European Economic Association  11(3):680–702 DOI: 10.1111/jeea.12011 Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 ENERGY CONSERVATION “NUDGES” AND ENVIRONMENTALIST IDEOLOGY: EVIDENCE FROM A RANDOMIZED RESIDENTIAL ELECTRICITY FIELD EXPERIMENT Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 681 2. In the United Kingdom, where electricity taxes are low, David Cameron has touted the behavioral transformations of putting “the typical electricity bill for a house like theirs in a neighborhood like theirs” in front of households. Speech of 13 June, 2008. http://www.aletmanski.com/files/davidcameron-powerofsocialinnovation.doc Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 him aware of the actions of others who have been in a similar situation (Goldstein et al. 2008; Schultz et al. 2007). Nudges may be a low cost strategy for encouraging energy conservation (Allcott and Mullainathan 2010; Thaler and Sunstein 2008).2 We posit that liberal/environmentalists are more likely to respond to energy conservation nudges. A series of recent empirical papers has documented that environmentalists are more likely to engage in “voluntary restraint” than the average person (Kotchen and Moore 2008). Those who vote in favor of “green policies” and register for liberal/environmentalist political parties are more likely to have a smaller carbon footprint and to purchase green products such as the Toyota Prius (Kahn 2007, Kahn and Morris 2009). Such environmentalists consciously avoid free riding and voluntarily restrain their consumption of goods and services that generate a negative externality. Our evidence on the role of ideology in energy conservation “nudges” comes from a randomized field experiment carried out by a western utility district in which we can observe a household’s ideology (an unobservable in prior studies), its socioeconomic and demographic characteristics, and its behavioral responses. Starting in Spring 2008, this utility has been sending households in the treatment group a Home Energy Report (HER). The report provides household specific information on own monthly electricity usage over time and relative to neighbors’ usage over the same time period. The report provides energy saving tips. To examine the role that political ideology and environmentalism play in determining how randomly selected households respond to these reports, we have collected data on the customer’s political party of registration, household donations to environmental organizations and household participation in renewable energy programs, and data on the characteristics of the local residential communities where the households live. Households who are registered in liberal political parties and who live in residential communities with a large liberal share and who have previously signed up for energy from renewable resources and donate to environmental causes are arguably environmentalists. Our focus on ideology, an unobservable in previous studies, distinguishes our work from other research (e.g. Allcott 2011; Ayres, Raseman, and Shih 2009). We find that the effectiveness of energy conservation “nudges” depends on an individual’s ideology. In the United States, Democrat, Peace and Freedom, and Green party members (liberals in the US terminology) are more likely to vote for environmentalist causes than Republican, American Party, or Libertarian party members (conservatives in the US terminology). We measure ideology not just with registered political party, but also with indicators of living in a liberal or conservative community and willingness to pay for energy generated from renewable resources and to donate to environmental organizations. Although liberals and environmentalists are more energy efficient than conservatives (Costa and Kahn 2010), thus making it harder 682 Journal of the European Economic Association 2. The Energy Conservation “Nudge” The “nudge” that the electric utility company sends to treatment households in an ongoing randomized experiment to encourage reductions in electricity consumption is a two-page HER (see the Appendix for a sample). Similar reports have been used by other utilities in the United States. The front page compares the electricity consumption of the household with all neighbors with similar size homes and heat type and with neighbors who are in the bottom 20th percentile of electricity usage. The back page compares the household’s electricity usage in the current month relative to the same time month in the prior year and awards green stars in every month the household consumed less relative to the same month in the past year (panel not shown in the Appendix because it is not publicly available). It also provides three tips for saving energy, such as turning down the thermostat when using an electric blanket or purchasing an Energy Star durable, and indicates the dollar amount in energy savings per year (shown in the Appendix). Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 for them to reduce consumption further, we find that liberals and environmentalists are more responsive to these nudges than conservatives. We find that among political liberals who purchase electricity from renewable resources, who donate to environmental causes, and who live in a census block group where the share of liberals is in the top 75th percentile, receiving a HER led to reductions in electricity usage of 3.6%. In contrast, among political conservatives who do not pay for renewable electricity, who do not donate to environmental groups, and who live in a census block group where the share of liberals is in the bottom 25th percentile, receiving a HER led to reductions in electricity usage of 1.1%. Liberals are more likely to turn down the air-conditioning in the summer in response to the HER report. Political liberals were 15% less likely to opt out of receiving the report and, in a survey, political liberals are also less likely than conservatives to state that the reports were useless and to report disliking them. By documenting the role that ideology plays in determining the effectiveness of a specific “nudge”, this paper contributes to the growing literature on the consequences of political ideology. Much of this work has focused on the role of political ideology in shaping preferences for redistribution (e.g. Piketty 1995). Our work focuses on the role of political ideology in shaping responses to non-market mechanisms designed to reduce consumption. Recent work on the determinants of political ideology has examined the causal role of the media (DellaVigna and Kaplan 2007), property rights (Di Tella, Galiani, and Schargrodsky 2007), and historical circumstances (Alesina and Fuchs-Schundeln 2007; Giuliano and Spilimbergo 2009) in shaping a person’s ideological outlook. Both social psychologists and economists have argued that beliefs on how society and the economy work predominately are formed at ages 18–25 (see Giuliano and Spilimbergo 2009). In this study, we will take as given that a household either is or is not a liberal/environmentalist and we will study how these political and social views influence household response to the same randomized treatment. Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 683 2.1. The HER Experiment Between March 14 and May 9 2008, the electric utility sent the first Home Electricity Reports to a treatment group of approximately 35,000 households. By April 1, 43% of all treatment households had received the report and by April 15 the figure was 62%. Households are still receiving the report, either on a quarterly or monthly basis. A control group of roughly 49,000 households have never received a HER. The HER experiment selected households from 85 census tracts with a high density of single-family homes (see ADM Associates 2009). Both treatment and control households had to have a current account with the electric utility that had been active for at least one year, could not be living in apartment buildings, and had to be living in a house with square footage between 250 and 99,998 square feet. Groups of contiguous census blocks were randomly assigned to either the treatment or control group. A “block batch” of five contiguous census blocks was randomly assigned to the treatment group and then a contiguous census block batch was assigned to the control group. The process continued until roughly 35,000 households were assigned to both the treatment and control groups. The remaining census blocks (about 14,000 homes) were assigned to the control group. Contiguous block groups were used because the implementation contractor, Positive Energy (now OPOWER), believed that increased communication among people receiving the HERs in the same community would lead to greater energy savings.4 Allcott (2011), Ayers, Raseman, and Shih (2009), and Schultz et al. (2007) found that providing feedback to customers on home electricity and natural gas usage with a focus on peer comparisons decreased consumption by 1% to 2%, potentially saving 110 million kWh per year if feedback were provided to all of the utility’s customers (Ayers, Raseman, and Shih 2009). Additional evidence that social incentives can make a difference comes from California’s 2001 media campaign to promote voluntary conservation after rolling blackouts in 2000 and early 2001. Consumption in San Diego 3. We thank a reviewer for suggesting this research strategy. 4. In a 2009 Home Energy Use Survey conducted by the electric utility, households in the control group were more likely to report talking to friends and neighbors about their electricity bill than households in the treatment group, suggesting that receiving the HER did not inspire discussion and that any positive peer effects operate through implicit social pressure. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Each report contains two pieces of information: the household’s absolute level of consumption and how its consumption compares with that of 100 neighbors living in similar-sized homes. We also know whether a household in the treatment group received a message of “great”, “good”, or “room for improvement” in the first report. As we discuss in the Online Appendix, we use this information to implement a regression discontinuity design to test whether the change in electricity consumption for the treatment group differs depending on the normative message that the household receives in the first report.3 684 Journal of the European Economic Association declined by 7% during the initial two phases of the campaign, before rebounding (Reiss and White 2008). Within a household production framework, a household values electricity as an input in producing comfort (e.g. indoor temperature) and leisure and household production activities. Total household electricity consumption in any given period is the sum of electricity used in each of these activities. A household’s total electricity consumption depends on choices over (1) the attributes of the house, such as size; (2) the attributes of appliances; and (3) the intensity of utilization of appliances for leisure and household activities, indoor temperature control and illumination. These choices, in turn, depend on climate, prices and personal attributes, including ideology. We view environmental ideology as a set of prior beliefs including those about the importance of energy conservation The ideological divide on environmental issues between Democrats and Republicans could affect how a household responds to an energy conservation “nudge”. In the United States, conservatives consistently oppose environmental regulation and energy policies intended to further environmental aims, as seen in polling data on the belief in climate change (Dunlap and McCright 2008) and Congressional voting patterns (Cragg et al. 2013). Dunlap and McCright (2008) report that in 2008 there was 34 percentage point gap between Democrats and Republicans in their agreement with a statement that the effects of global warming have already begun, up from a four percentage point gap in 1997. The 2008 National Environmental Scorecard of the League of Conservation Voters gives the House Democratic leadership a score of 95 (out of a best score of 100) and the Republican leadership a score of 3.5 A 2009 Pew survey found a 23 percentage gap between Democrat and Republican agreement with the statement that people should be willing to pay higher prices to protect the environment. Republicans and Democrats respond differently to “carbon offsets” versus “carbon tax” (Hardistry, Johnson, and Weber 2010), suggesting that ideology moderates how individuals think of key words. European studies have highlighted the role that environmental ideology plays in determining the willingness to take voluntary actions to mitigate one’s carbon externality and the willingness to pay to purchase a “green product”. Thalmann (2004) and Halbheer, Niggli, and Schmutzler (2006) document that voters who are left-ofcenter or who are environmentalists were more likely to vote for Swiss environmental referenda. Brounen and Kok (2011) found that the price premium for residential homes that are certified as highly energy efficient in Holland is higher in “green” communities, that is communities where the Green Party and the Party for the Animals had received a larger fraction of the vote. Given this background, there are two main hypotheses that can be tested. Many households will read the report and respond to it for non-ideological reasons, such 5. See http://www.people-press.org/2009/05/21/section-9-the-environment-and-the-economy/ Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 3. Why Could Ideology Mediate the Response to this Nudge? Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 685 4. Data Our primary data set consists of residential billing data from January 2007 to October 2009. These data provide us with information on kilowatt hours purchased per billing cycle, the length of the billing cycle (measured in days), whether the house uses electric heat, and whether the household is enrolled in the electric utility’s program to purchase energy from renewable sources. We link each billing cycle to the mean temperature in that billing cycle.6 We link the billing data to the treatment and control data which contain information on when the household began to receive the HERs, as well as information on square footage of the house, information on whether the home heats with electricity or natural gas, and the age of the house. We cannot match 1,976 observations in the pilot and control data to the residential billing data. In our final data set, the treatment and control data therefore contain 81,722, with 48,058 households in the control group. Among the households in the treatment group, 24,028 received a monthly report and 9,636 received a quarterly report. We merge individual voter registration and marketing data for March 2009 to our data set.7 For registered voters we know party affiliation, and whether the individual donates to environmental organizations. We were able to link half of our sample to the voter registration data. We linked either the person whose name was on the utility bill or the first person on the utility bill.8 The individuals we could not link were living in smaller households and in census block groups with a low proportion of the college-educated, were more likely to receive a subsidy for electricity because of their 6. Two different households in the same calendar year and same month who are on different billing cycles will face different climate conditions. 7. We purchased the data from www.aristotle.com. 8. Only 5% of households were “mixed” between conservatives and liberals. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 as wanting to lower their bill. These households, regardless of ideology, may reduce their consumption. The response of liberals is ambiguous. Liberals may reduce their consumption by more than conservatives because of their ideology and have been observed to consume less electricity (Kotchen and Moore 2008; Costa and Kahn 2010). However, because they have already invested more time and effort in monitoring their electricity bills and in engaging in voluntary restraint (i.e. lowering the air-conditioner in the summer), their response could be lower than that of conservatives. Secondly, and more controversially, it is possible that anti-environmentalist conservatives who receive a green looking report and learn that they consume more than their peers may refuse to decrease their consumption or even increase their consumption in an act of defiance. People find information more reliable when it conforms to their strong prior beliefs (e.g. Lord, Ross, and Leper 1979; Miller et al. 1993; Munro and Ditto 1997; Gentzkow and Shapiro 2006, 2010) and are influenced mainly by those in their network (Murphy and Shleifer 2004). 686 Journal of the European Economic Association 9. Relative to all homeowners in the same county these individuals were also more likely to be of Asian or other ancestry rather than of European ancestry, but were less likely to be Spanish speaking. They were also lower income. 10. The collected revenue is used by the electric utility to purchase and produce power from wind, water, and Sun. 11. Household income is available from credit bureau data. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 low income, and were more likely to have a household head above age 60.9 We also merge to these data, by the census block group, the share of registered voters who were liberal (Democrat, Green, or Peace and Freedom) in 2000 and the share of the college-educated in the block group. We expect that environmentalists are more likely to live in liberal, educated communities (Kahn 2007). We have access to two revealed preference measures of a household’s environmentalism. From the data base with voter registration information, we know whether a household has donated money to an environmental group. We also know whether the household has signed up for the company’s renewable power program prior to the treatment. This is the electric utility’s major program to increase the share of its customers who have signed up for renewable energy. Each household decides whether to opt in and pay a fixed cost of $3 a month to have 50% of its power generated by renewables or $6 a month to have 100% of its power generated by renewables.10 We also have access to an ancillary data set which we use to examine household attitudes about the HER by ideology. In 2009 the electric utility company surveyed 1,375 households who received the HER, asking them questions about the HER report. We restrict this sample to households for whom we have information on age and the fraction of liberals in the block group and to households who were not in minor parties we could not classify as liberal or conservatives. This leaves us with 1,061 observations in this ancillary data set. Table 1 shows that the treatment and control groups are roughly representative of all homeowners in the county in terms of household and neighborhood characteristics. But, there are some clear differences. The treatment and control groups consume roughly 10% more electricity than the average county homeowner as of 2007 (before the experiment). Relative to the average homeowner, the experiment homes are older and more likely to be electric homes. The households in the experiment group are roughly 10% richer than the average county homeowner.11 The geographical areas included in the experiment have a higher share of college graduates than the average county home owner’s community. The randomization of the HER across blocks was effective. Ayers, Raseman, and Shih (2009) reported that controlling for house characteristics, household demographics, and the number of cooling degree days and heating degree days, there was no systematic difference in energy usage between treatment and control groups prior to the treatment (also see Table 1 where we report no systematic differences between treatment and control groups adjusting for block batch group correlation). 17.710 1.159 14.967 43,252.950 682.611 20.619 0.162 0.105 0.492 0.498 0.330 0.045 0.499 0.275 0.371 0.283 27.930 2.111 56.582 66,484.710 1,709.447 1,976.764 0.283 0.460 0.412 0.461 0.124 0.002 0.474 0.082 0.165 0.088 285,717 S.D. 0.447 0.439 0.112 0.002 0.425 0.099 0.246 0.104 48,058 31.051 2.111 56.941 74,826.920 1,720.876 1,971.176 0.364 0.436 Mean S.D. 0.497 0.496 0.316 0.046 0.494 0.299 0.431 0.305 15.473 1.136 14.952 41,364.120 602.081 18.377 0.158 0.098 Control Group 0.438 0.449 0.112 0.002 0.430 0.097 0.264 0.103 33,664 30.801 2.103 56.594 74,312.590 1706.109 1972.618 0.363 0.438 Mean 0.496 0.497 0.315 0.043 0.495 0.296 0.441 0.304 14.727 1.137 15.085 41,546.370 578.287 18.547 0.162 0.097 S.D. Treatment Group Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Note: All variables listed after the block group variables are dummy variables. The treatment and controls are not statistically significant after adjusting for clustering at the block batch level. Data on all homeowners comes from the records of the utility company and includes households in both the treatment and control groups. Avg. Daily Electricity (kWh) in 2007 Household Size Age of Head Household Income Home Square Footage Home Year Built Block Group% College Block Group% Liberal Registered as Republican, American, Libertarian Democrat, Green, Peace and Freedom No party Other Not registered Donates to Environmental Causes Electric Heat Home Pays for Renewable Energy Observations Mean All Home Owners TABLE 1. Summary statistics, all homeowners, control and treatment group. Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 687 688 Journal of the European Economic Association TABLE 2. Summary statistics by political party registration. Conservatives Mean S.D. Mean S.D. 0.406 33.952 2.352 58.490 84279.550 1828.034 1973.234 0.380 0.408 0.087 0.247 0.070 0.491 16.236 1.192 14.104 43711.440 632.129 16.881 0.157 0.086 0.282 0.431 0.256 0.416 29.551*** 2.096*** 59.199*** 74806.960*** 1672.423*** 1968.815*** 0.375*** 0.454*** 0.114*** 0.237*** 0.141*** 0.493 14.248 1.099 13.413 40377.220 548.087 19.375 0.156 0.101 0.318 0.425 0.348 0.222 0.412 0.304*** 0.457 0.138 21,193 0.341 0.395*** 21,172 0.487 Note: A conservative is defined as Republican, American Party, or Libertarian. A liberal is defined as Democrat, Green, or Peace and Freedom. A “like-minded” community is defined for a conservative as a census block in the bottom quartile of fraction liberal. For a liberal, a “like-minded” community is defined as a census block in the top quartile of fraction liberal. The symbols ** and *** indicate that the differences between conservatives and liberals are statistically, significantly different at the 5 and 1 percent level (adjusting for block batch group), respectively. Households living in electric homes were more likely to receive a monthly rather than a quarterly report.12 Table 2 shows that registered conservatives are slightly younger, have higher household incomes, larger households, and larger homes than registered liberals. While both conservatives and liberals donate to environmental causes and pay for renewable energy, liberals are more likely to do so. We could explain only 2%–4% of the variance in registered conservative status with age, household income, house value, house size, electric heat, and renter status (see Online Appendix Table A.1 for details). Among the treated, liberals are more likely to receive the quarterly report, indicating that they use less electricity than conservatives. Liberals living in a liberal community (defined as a community in which the liberal share is in the top quartile) are almost three times as likely to receive the quarterly report as conservatives living in a conservative community (defined as a community in which the liberal share is in the bottom quartile). If political ideology does not affect the response to the HER, conservatives should respond more because they are treated more intensely. 12. Within a treated “block batch” households received either a monthly or quarterly report. Roughly 71% of households received the monthly report. Households with a low baseline electricity consumption received the quarterly report while households with a high baseline electricity consumption received the monthly report. Conditional that a household’s daily average electricity consumption was less than 20 kWh per day it had a 2.5% chance of receiving the monthly report. Households whose 2006 electricity consumption was greater than 23 kWh per day had a 99% chance of receiving the monthly report. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Fraction in Treatment Group Avg. Daily Electricity (kWh) Household Size Age of Head Household Income Home Square Footage Home Year Built Block Group% College Block Group% Liberal Donates to Environmental Causes Electric Heat Home Pays for Renewable Energy Treatment Group Receives Quarterly Report Receives Quarterly Report and in Like-minded community Observations Full Sample Liberals Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 689 5. Econometric Framework ln(kW h) = β0 + β1 (Household FE) + β2 (Month/Year FE) + β3 (Temp, Electric) + β4 TREAT + β6 (TREAT×Party Registration) + β7 (TREAT × Green Indicators) + β8 (TREAT × Individual Characteristics) + β9 (TREAT × Block Characteristics) + β10 (TREAT × House Characteristics) + β11 (Post TreatmentPeriod Dummy × All Characteristics) + ε, (1) where the unit of analysis is the household in a year and month and where TREAT is a dummy equal to one if the household received the Home Energy Report and where the different specifications use different subsets of the variables. TREAT thus is equal to 0 either if the household is never treated or is not yet treated. In all regressions we control for household and month/year fixed effects, a cubic in mean daily temperature within the billing cycle, and an interaction of the cubic mean daily temperature with a dummy indicator if the house is an electric house (Temp, Electric). We also interact party registration, green indicators, individual characteristics, block characteristics, and house characteristics with an indicator for whether the treatment period has started. Our political party indicators are liberal (Democrat, Green, or Peace and Freedom), other party (Reform, Conservative, Natural Law or Other), no party affiliation, and not registered, with conservative (Republican, American Party, and Libertarian) as the omitted dummy variable. Our green indicators are whether the household purchases energy from renewable sources and whether the household donates money to environmental causes. We cluster the standard errors on the block batch group to account for correlation within the block batch group. Our specifications differ in the number of included treatment interaction effects. We examine who accepts treatment by estimating, for the treatment group, a probit regression of the form OptOut = β0 + β1 High + β2 ln(Usage) + β3 Age + β4 Liberal +β5 Unregistered + ε, (2) where OptOut is a dummy variable equal to one if the household opts out of receiving the treatment, High is a dummy variable equal to one if the household’s consumption Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Previous studies have examined the HER report’s average treatment effect and have also examined how its effect varies by standard characteristics such as attributes of the home and socio-economic characteristics of the owner. The distinguishing feature of this study is our emphasis on household environmental ideology as an important determinant for how households respond to well-meaning new information. We estimate intent-to-treat effects (which we will simply refer to as treatment effects) of receiving the HER by seven specifications of the form 690 Journal of the European Economic Association Report Reaction = β0 + β1 High + β2 ln(Usage) + β3 Age + β4 Liberal + β5 Unregistered + ε, (3) where Report Reaction is either a dummy variable equal to one if the household found the reports of no value (responses of not at all or not very valuable) or a dummy variable equal to one if the household disliked the reports (responses of did not like or indifferent). 6. Results Own ideology, whether measured by political party affiliation, donations to environmental organizations, or the purchase of green energy, is associated with differential treatment effects (see regression 2 in Table 3).13 Although the mean overall treatment effect is –0.021 (see regression 1 in Table 3) when we do not allow for heterogeneous treatment effects, a registered conservative will decrease mean daily kWh by 1.7% in response to the treatment but a registered liberal will reduce consumption by 2.4%, all else equal (see regression 2 in Table 3). Those purchasing energy from renewable resources reduce their consumption by 0.9% in response to the treatment relative to those not purchasing green energy. Those donating to environmental organizations reduce their consumption by 1.1% more than households who do not donate to environmental organizations. The fraction of liberals and the fraction of college-educated in the census block group affects treatment response, independent of own characteristics (see regression 3 in Table 3). An increase of 0.1 in the fraction of liberals in the census block group reduces consumption by 0.6% in response to the treatment. Controlling for the fraction of liberals in the block group leads to statistically insignificant effects of own party affiliation but leaves the effects of ideology as measured by donations unchanged. Own ideology, donating to environmental groups, and paying for renewable energy remain jointly statistically significant. The higher the fraction of college graduates in a census block group, the lower consumption. 13. We also have information on household monthly expenditure on electricity. In the presence of an increasing block tariff structure, some recipients of the HER may reduce their expenditure by more than their electricity consumption. We have estimated regressions similar to equation (1) in which we use the log of household monthly electricity expenditure as the dependent variable. We obtained very similar results. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 is above the neighborhood average, Usage is the household’s electricity usage in 2006, Age is the age of the head of the household, Liberal is a dummy equal to one if the household head was registered as either a Democrat, Green, or Peace and Freedom party member, and Unregistered is a dummy equal to one of if the household was not registered. Using a small surveyed sample, we also examine who, in the treatment group, found the reports of no value or disliked the reports by estimating probit regressions of the form Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 691 TABLE 3. Treatment effects by ideology, education, and home structure. Dependent Variable: Log (Mean Daily kWh) (2) − 0.021*** − 0.017*** (0.003) (0.003) (Registered liberal) − 0.007** (0.003) (Registered other party) 0.031 (0.032) (No registered party) 0.004 (0.005) (Not in voter registration − 0.003 data) (0.004) (Donates to environmental − 0.011** (0.004) organizations) (Pays for renewable − 0.009* (0.005) energy) (Liberal share within block group) (College graduate share within block group) (Logarithm of age of house) (Logarithm of square footage of house) (Electric House) Treated (3) (4) (5) (6) 0.026** (0.011) − 0.004 (0.003) 0.028 (0.032) 0.004 (0.005) − 0.002 (0.004) − 0.011** (0.004) − 0.006 (0.005) − 0.062*** (0.023) − 0.045*** (0.017) 0.069 (0.063) − 0.004 (0.003) 0.028 (0.032) 0.004 (0.005) − 0.002 (0.004) − 0.010** (0.004) − 0.006 (0.005) − 0.038* (0.021) − 0.043** (0.018) − 0.009** (0.004) − 0.003 (0.007) − 0.014*** (0.005) 0.016 (0.073) − 0.004 (0.003) 0.030 (0.031) 0.003 (0.005) − 0.003 (0.004) − 0.009** (0.004) − 0.006 (0.005) − 0.036* (0.022) − 0.047** (0.019) − 0.007 (0.005) − 0.008 (0.007) − 0.014*** (0.005) − 0.001 (0.004) 0.008*** (0.003) − 0.008 (0.018) 0.008 (0.008) − 0.005 (0.003) 0.028 (0.032) 0.004 (0.005) − 0.002 (0.004) − 0.011** (0.004) − 0.007 (0.005) (Logarithm of household income) (Logarithm of home value) (Dummy = 1 if renter) (Second quintile of liberal share in block group) (Third quintile of liberal share in block group) (Fourth quintile of liberal share in block group) (Fifth quintile of liberal share in block group) Joint significance of registered liberal, donates to environmental organizations, and pays for renewable energy, F(3,956) Household fixed effects Month-Year fixed effects Y Y − 0.046*** (0.018) 5.68*** 3.81*** 3.30** 3.23** − 0.011 (0.007) − 0.018*** (0.007) − 0.010* (0.005) − 0.018* (0.010) 3.98*** Y Y Y Y Y Y Y Y Y Y Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 (1) 692 Journal of the European Economic Association TABLE 3. Continued Dependent Variable: Log (Mean Daily kWh) Observations R-squared (2) (3) (4) (5) (6) 2,760,175 2,760,175 2,760,175 2,760,141 2,754,232 2,760,175 0.804 0.804 0.804 0.805 0.805 0.804 Note: Each observation is a household-billing cycle. Standard errors, clustered on the block batch group, are in parentheses. * indicates significance at the 10% level, ** at the 5% level, and *** at the 1% level. Additional control variables are a cubic in mean daily (24 hr.) temperature, the cubic in daily temperature interacted with a dummy indicating whether the home is an electric home, household fixed effects, year-month fixed effects, and interactions between characteristics and a time dummy indicating the experiment has started. Mean daily kWh are 31.69. Conservative is the omitted category and is defined as Republican, American Party, or Libertarian. Liberal is defined as Democrat, Green Party, or Peace and Freedom. The fourth regression in Table 3 controls for the effect of house characteristics on treatment response. Those in older houses, in bigger homes, and in electric homes reduce their consumption more. Housing characteristics may reflect occupant characteristics. Liberals are more likely to be in older houses (but less likely to be in bigger homes). Controlling for housing characteristics, each increase of 0.1 in the fraction of liberals in the census block group reduces consumption by 0.4% in response to the treatment. The fifth regression in Table 3 adds controls for the effect of household income, home value, and renter status on treatment response. There is no statistically significant differential treatment effect of renter status. The treatment effect is increasing in home value. The addition of these and the previous control variables reduces the impact of liberal share within the block group but does not affect how liberals respond to the treatment. The sixth regression in Table 4 examines the linearity assumption on liberal share within census block group by substituting quintiles of the liberal share for liberal share within blocks. It shows that a greater liberal share relative to the lowest quintile is associated with households’ reducing their electricity consumption in response to the treatment. The quintiles are jointly significantly different from 0 at the 10% level (F(4,956) = 2.29). We observe a similar pattern when we examined quartiles instead of quintiles (see regression 1 in Online Appendix Table A.2). Our seventh and final regression (predicted values are graphed in Figure 1 and the regression is given in the last column of Online Appendix Table A.2) tests for persistence over time by adding indicators for whether the time period is during the first or second half of the experiment. It shows that registered conservatives who do not purchase renewable energy and who do not donate to environmental organizations reduce their consumption in response to the treatment by 0.01 in the first half of the experiment and by 0.02 in the second half of the experiment. Registered liberals who purchase renewable energy and who donate to environmental organizations reduce their consumption in response to the treatment by 0.04 in the first half of the experiment and by 0.05 in the second half of the experiment. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 (1) Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 693 TABLE 4. Predicted Treatment Effects by Ideology Std. Err. −0.036*** 0.006 −0.021*** −0.011*** 0.003 0.003 −0.048*** 0.010 −0.031*** −0.008*** 0.009 0.003 Note: Predicted treatment effects are estimated from Regressions 4 and 6 in Table 3. *** indicates statistical significance at the 1% level. Everyone in the treatment group is assigned the given characteristics while all other characteristics are kept at their median values. Conservative is defined as Republican, American Party, or Libertarian. Liberal is defined as Democrats, Green Party, and Peace and Freedom. When we restricted the sample to households whose electricity usage was above the median in 2006 (precisely those households a utility would want to target to reduce total electricity demand), we obtained a similar response for liberals but a larger response for households who pay for renewable energy (see Online Appendix Table A.3). The effects of being in an electric home are no longer as large. When we examined households with baseline electricity usage below the median we found a smaller treatment effect and a statistically insignificant effect of paying for renewable energy and for donating to environmental groups (see Online Appendix Table A.4). Table 4 examines the role that ideology plays in responding to receiving the HER. We use the regression results from Table 3’s columns (4) and (6). Evaluating all characteristics at the median and using the regression in column (4), the treatment effect for liberals who purchase energy from renewable resources, who donate to environmental causes, and who live in a block group where the share of liberals is at least in the 75th percentile (less than 1% of our sample) is –0.036. The treatment effect for registered liberal who live in a block group where the share of liberals is at least in the 75th percentile (26% of our sample) is –0.021. The treatment effect for a conservative who does not pay for renewable energy, does not donate to environmental groups, and is in bottom 25th percentile liberal block group (22% of our sample) is –0.011. When we use the regression results from Table 3’s column (6) we obtain a treatment effect of –0.008 for a conservative who does not pay for renewable energy, does not donate to environmental groups, and is in the bottom quintile liberal block group. The treatment effect for a liberal who pays for renewable energy, donates to environmental groups, and is in the top quintile liberal block group is –0.048. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Using Regression 4 (Table 3) Registered liberal, pays for renewable energy, donates to environmental groups, and in top 75th percentile liberal block group Registered liberals and in top 75th percentile liberal block group Registered conservative, does not pay for renewable energy, does not donate to environmental groups, and in bottom 25th percentile liberal block group Using Regression 6 (Table 3) Registered liberal, pays for renewable energy, donates to environmental groups, and in top 75th percentile liberal block group Registered liberals and in top quintile liberal block group Registered conservative, does not pay for renewable energy, does not donate to environmental groups, and in bottom quintile liberal block group Treatment Effect Journal of the European Economic Association 0 694 -.02 -.04 -.06 Registed liberal, pays for renewable energy, donates to environmental groups First Half Experiment Second Half Experiment Time F IGURE 1. Predicted treatment effects over time. Predicted treatment effects are estimated from Regression 4 in Web Appendix Table 2. Upper and lower bounds of a 95% confidence interval are given as dashed lines. Everyone in the treatment group is assigned the given characteristics while all other characteristics are kept at their median values. Conservative is defined as Republican, American Party, or Libertarian. Liberal is defined as Democrats, Green Party, and Peace and Freedom. “First Half Experiment” refers to all observations in 2008 and thus includes roughly 8 months of treatment. “Second Half Experiment” refers to households in 2009 and thus includes about 10 months of treatment. We further probed the robustness of our results by estimating equation (1) using quantile regressions (see Table 5). Estimating the impact of the HERs at the 10th, 25th, 50th, 75th, and 90th quantiles, our predicted results show that at the lower quantiles liberals who pay for renewable energy and donate to environmental organizations are three to four times as likely to reduce their consumption in response to the treatment as conservatives who do not pay for renewable energy and do not donate to environmental organizations. At the 90th quantile, differences between conservatives and liberals are smaller. 6.1. Persistence of the Report Effects Our examination of seasonal patterns of response to the treatment leads us to conclude that liberals are more likely to turn down the air-conditioning in the summer in response to the treatment. When we added to equation (4) in Table 3 an interaction Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Registered conservative, pays for renewable energy, donates to environmental groups Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 695 TABLE 5. Treatment effects by ideology, quantile regressions. Dependent Variable: Log (Mean Daily kWh) Quantiles Treated x (Registered liberal) (Registered other party) (No registered party) (Not in voter registration data) (Donates to environmental organizations) (Pays for renewable energy) Household fixed effects Month-Year fixed effects Observations Pseudo R-squared Predicted Treatment Effect Conservatives who do not donate to environmental organization or pay for renewable energy Liberals who donate to environmental organizations and pay for renewable energy 0.25 0.50 0.75 0.90 − 0.018*** (0.002) − 0.017*** (0.002) − 0.017 (0.021) − 0.017 (0.031) − 0.020*** (0.006) − 0.007*** (0.003) − 0.009 (0.034) 0.003 (0.005) − 0.003 (0.004) − 0.019*** (0.006) − 0.007*** (0.002) 0.013 (0.024) 0.000 (0.003) 0.000 (0.002) − 0.014*** (0.003) − 0.007 (0.039) 0.014 (0.101) 0.001 (0.061) 0.002 (0.011) − 0.009 (0.074) − 0.006 (0.043) 0.027 (0.048) − 0.001 (0.055) 0.004 (0.034) − 0.006 (0.104) − 0.005** (0.003) 0.073** (0.035) 0.005 (0.009) 0.007 (0.008) − 0.000 (0.008) − 0.002 (0.003) Y Y 2,760,175 0.112 − 0.010*** (0.003) Y Y 2,760,175 0.175 − 0.01 (0.066) Y Y 2,760,175 0.245 − 0.009 (0.033) Y Y 2,760,175 0.284 − 0.007 (0.006) Y Y 2,760,175 0.267 − 0.018 − 0.017 − 0.017 − 0.017 − 0.020 − 0.046 − 0.048 − 0.043 − 0.038 − 0.032 Note: Each observation is a household-billing cycle. Bootstrap standard errors, clustered on the block batch group, are in parentheses. * indicates significance at the 10% level, ** at the 5% level, and *** at the 1% level. Additional control variables are a cubic in mean daily (24 hr.) temperature, the cubic in daily temperature interacted with a dummy indicating whether the home is an electric home, household fixed effects, year-month fixed effects, and interactions between characteristics and a time dummy indicating the experiment has started . Mean daily kWh are 31.69. Conservative is the omitted category and is defined as Republican, American Party, or Libertarian. Liberal is defined as Democrat, Green Party, or Peace and Freedom. between treatment and summer months (May 1–October 31) and an interaction between treatment, summer months, and liberal, we obtained a coefficient on the interaction between treatment and summer months of –0.002 (σ = 0.006) and a coefficient on the interaction between treatment and summer months and liberal of –0.012 ( σ = 0.004) (see Online Appendix Table A.2, column 3). It is theoretically ambiguous whether the HERs will have a larger impact in the short run or the medium term. When a household first receives such a report this may be a salient event whose “new news” shocks the household and subsequent reports Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Treated 0.10 696 Journal of the European Economic Association 6.2. Opting out of the Treatment The decision to quit the HER treatment and consumer survey reactions to the HER provide additional evidence on which subgroups of the population disliked the treatment that are consistent with our identity story. Households could opt out of receiving the HER either by emailing, phoning, or mailing the utility. Although the information is free, 2% of households took action not to receive it and 36% of the survey group reported disliking the report. Our results have implications not just for our identity hypothesis but also for the long-term success of the HER program among different types of households. Households that opted out of the treatment were more likely to be high electricity consumers, both relative to their neighbors and in absolute levels, and they were less likely to be liberals than conservatives (see Table 6). At the mean opt out rate of 0.020, a liberal was 15% less likely to opt out. High electricity users relative to their neighbors were 65% more likely to opt out. In a subsample of 1,061 consumers interviewed about the home energy reports, high electricity users, both relative to their neighbors and in absolute levels, were more likely to claim that the reports were useless or that they disliked them. Liberals were less likely than conservatives to state that the reports were useless or that they disliked them. Being liberal decreased the probability of finding a report useless by 0.131, a decrease of 44% from the sample mean of 0.301. Being a liberal decreased the probability of disliking the report by 0.102, a decrease of 28% from the sample mean of 0.363. High electricity users relative to their neighbors were 27% more likely to find the report useless and were 38% more likely to report disliking the report. Liberals and conservatives did not report differential rates of spending less than 2 minutes reading the report (see the last column in Online Appendix Table A.5). High users were statistically, significantly more likely to spend less time reading the report. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 reinforce the original news. In this case, we might observe a large drop in consumption followed by a constant level (climate adjusted). Alternatively, in the medium term a household is more likely to adjust more of its durables stock and may make more energy efficient investments when it makes new investments in such durables. The evidence suggests that this strategy is being pursued. We found that households in the treatment group were more likely to obtain a rebate from the utility for purchasing an energy efficient durable. In a probit regression (see Online Appendix Table A.5) of the probability of obtaining a rebate on whether the household was in the treatment group, the household’s political affiliation, the age of the household head, and the household’s baseline electricity usage, we found that the derivative of the coefficient on the treatment dummy was a statistically significant 0.006 (σ = 0.002). At the sample mean of 0.056, this represents an 11% increase in the probability of obtaining a rebate. Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 697 TABLE 6. Decision to opt out of treatment and view of reports by ideology. Dependent Variable = 1 if Reports of Dislike Dummy = 1 if Above community mean electricity use Logarithm of 2006 electricity consumption Age of household head Dummy = 1 if registered Republican, American Party, Libertarian Democrat, Green, Peace and Freedom Dummy = 1 if not registered Pseudo R-squared Observations 0.013*** (0.002) 0.008*** (0.002) 0.000*** (0.000) − 0.003** (0.001) − 0.004** (0.001) 0.062 32,667 No Value Reports 0.082** (0.035) 0.159*** (0.038) 0.003*** (0.001) 0.138*** (0.036) 0.103*** (0.040) 0.001 (0.001) − 0.131*** (0.032) − 0.078** (0.032) 0.055 1,061 − 0.102*** (0.035) − 0.031 (0.036) 0.04 1,061 Note: The opt out decision is estimated for all treated households. Registered voters with no party affiliation or with an affiliation other than Republican, American Party, Libertarian, Democrat, Green, Peace, and Freedom are not included in the regressions. The mean opt out rate is 0.020. A subsample of the treatment group was interviewed about the home energy reports. 30.1% of the sample found the reports to be not at all or not very valuable. 36.3% of the sample reported not liking or being indifferent about receiving the reports. Standard errors are in parentheses. 6.3 A Regression Discontinuity Test of Differential Responses to Normative Messages This paper’s main focus has been to estimate the differential response of environmentalists and non-environmentalists to receiving a HER report. We have also examined whether, among those who received a HER report, there is a differential response between political liberals and conservatives to the normative message included with the first report. Households received one of three normative messages: “great”, “good”, or “room for improvement”. These normative messages were based on the household’s consumption compared to that of 100 neighbors living in similar-sized houses. For the first report only we observe both the normative message and the ratio of the household’s electricity consumption in the last month divided by the average consumption of 100 nearby neighbors. As discussed in the Online Appendix, we implemented a regression discontinuity design (see Lee and Lemieux 2010) to test whether political ideology influences how households respond to receiving sharp normative feedback. Although our point estimates suggest that a political conservative increases consumption by 5.7% in response to receiving a normative message of “good” versus “room for improvement” whereas a political liberal increases consumption by only 0.9%, the 95% point wise confidence intervals are large and we fail to reject the hypothesis that there is no “normative message” for either conservatives or liberals. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Opt Out 698 Journal of the European Economic Association 7. Conclusion Appendix. Sample Home Electricity Report 14. Allcott (2011) provides an excellent cost–benefit analysis of electricity nudges. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 “Nudges” can make us healthier, richer in our retirement (through opt out defaults), and better environmental citizens. Given that such information treatments are cheap to produce and distribute, these could be cost-effective policies especially for those subsets of households who are most responsive to these treatments.14 This paper exploited a unique data merge of information from an electricity information provision field experiment to study how liberal/environmental ideology mediates responses to peer comparison information. Liberal households are less likely to drop out of the experiment and more likely to report that they like receiving the report than political conservatives. In response to receiving the report, liberals reduce their electricity consumption by a larger percentage than conservatives. Our results suggest that environmental nudges are most effective in relatively liberal communities. What works in California may not work in Lubbock, Texas. And even in California, targeted messaging may be more cost-effective than random assignment of home energy reports. Future research should continue to test for what might be effective conservation messages with political conservatives. Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 699 Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Source: Residential Energy Use Behavior Change Pilot, OPOWER white paper, http://www.opower.com/LinkClick.aspx?fileticket=cLLj7p8LwGU%3d&tabid=7 700 Journal of the European Economic Association References Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 ADM Associates, Inc. (2009). The Impact of Home Electricity Reports. September. Alesina, Alberto and Nicola Fuchs-Schundeln (2007). “Good-Bye Lenin (or Not?): The Effect of Communism on People’s Preferences.” American Economic Review, 97, 1507–1528. Allcott, Hunt (2011). “Social Norms and Energy Conservation.” Journal of Public Economics, 95, 1082–1095. Allcott, Hunt and Sendhil Mullainathan (2010). “Behavior and Energy Policy.” Science, 327, 1204– 1205. Ayers, Ian, Sophie Raseman, and Alice Shih (2009). “Evidence from Two Large Field Experiments that Peer Comparison Feedback Can Reduce Residential Energy Usage.” National Bureau of Economic Research Working Paper 15386. Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan (2004). “How Much Should We Trust Differences-in-Differences Estimates?” Quarterly Journal of Economics, 119, 249–275. Brounen, Dirk and Nils Kok (2011). “On the Economics of Energy Labels in the Housing Market.” Journal of Environmental Economics and Management, 62, 166–179. Cialdini, Robert G. et. al. (2006), Managing Social Norms for Persuasive Impact, Social Influence, 1, 3–15. Costa, Dora L. and Matthew E. Kahn (2010). “Why Has California’s Residential Electricty Consumption Been So Flat Since the 1980s? A Microeconometric Approach.” National Bureau of Economic Research Working Paper 15978. Cragg, Michael, YuYu Zhou, Kevin Gurney, and Matthew E. Kahn (2013). “Carbon Geography: The Political Economy of Congressional Support for Legislation Intended to Mitigate Greenhouse Gas Production.” Economic Inquiry, 51, 1640–1650. DellaVigna, Stefano and Ethan Kaplan (2007). “The Fox News Effect: Media Bias and Voting.” Quarterly Journal of Economics, 122, 1187–1234. Di Tella, Rafael, Sebastian Galiani, and Ernesto Schargrodsky (2007). “The Formation of Beliefs: Evidence from the Allocation of Land Titles to Squatters.” Quarterly Journal of Economics, 122, 209–241. Dunlap, Riley E. and Aaron M. McCright (2008). “A Widening Gap: Republican and Democratic Views on Climate Change.” Environment, 50, 26–35. Gentzkow, Matthew and Jesse M. Shapiro (2006). “Media Bias and Reputation.” Journal of Political Economy, 114, 280–316. Gentzkow, Matthew and Jesse M. Shapiro (2010). “What Drives Media Slant? Evidence from U.S. Daily Newspapers.” Econometrica, 78, 35–71. Giuliano, Paola and Antonio Spilimbergo (2009). “Growing Up in a Recession: Beliefs and the Macroeconomy.” National Bureau of Economic Research Working Paper No. 15321. Goldstein, Noah, Robert B. Cialdini, and Vladas Griskevicius (2008). “A Room With a Viewpoint: Using Social Norms to Motivate Environmental Conservation in Hotels.” Journal of Consumer Research, 35, 472–482. Griskevicius, Vladas, Robert B. Cialdini, and Noah J. Goldstein (2008). “Social Norms: An Underestimated and Underemployed Lever for Managing Climate Change.” IJSC, 3, 5–13. Halbheer, Daniel, Sarah Niggli, and Armin Schmutzler (2006). “What Does it Take to Sell Environmental Policy? An Empirical Analysis of Referendum Data.” Environmental and Resource Economics, 33, 441–462. Hardistry, David, Eric Johnson, and Elke Weber (2010). “A Dirty Word or a Dirty World? Attribute Framing, Political Affiliation, and Query Theory.” Psychological Science, 2, 86–92. Kahn, Matthew E (2007). “Do Greens Drive Hummers or Hybrids? Environmental Ideology as a Determinant of Consumer Choice.” Journal of Environmental Economics and Management, 54, 129–145. Kahn, Matthew E. and Eric Morris (2009). “Walking the Walk: The Association Between Environmentalism and Green Transit Behavior.” Journal of the American Planning Association, 75, 389–405. Costa and Kahn Energy Conservation “Nudges” and Environmentalist Ideology 701 Supporting Information Additional Supporting Information may be found in the online version of this article at the publisher’s website: Table A.1. Correlates of being a registered conservative and of being a registered conservative who does not donate to environmental groups, does not pay for renewable energy, and lives in a conservative area. Table A.2. Treatment effects by ideology. Table A.3. Treatment effects by ideology, education, and structure for households above the median baseline usage. Table A.4. Treatment effects by ideology, education, and structure for households below the median baseline usage. Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Kahn, Matthew E. and Ryan K. Vaughn (2009). “Green Market Geography: The Spatial Clustering of Hybrid Vehicles and LEED Registered Buildings.” Contributions in Economic Policy, A Berkeley Electronic Press Journal. Kotchen, Matthew and Michael Moore (2007). “Private Provision of Environmental Public Goods: Household Participation in Green-Electricity Programs.” Journal of Environmental Economics and Management, 53, 1–16. Kotchen, Matthew and Michael Moore (2008). “Conservation: From Voluntary Restraint to a Voluntary Price Premium.” Environmental and Resource Economics, 40, 195–215. Lee, David S. and Thomas Lemieux (2010). “Regression Discontinuity Designs in Economics.” Journal of Economic Literature, 48, 281–355. Lee, David S., Enrico Moretti, and Matthew J. Butler (2004). “Do Voters Affect Or Elect Policies? Evidence from the U.S. House.” The Quarterly Journal of Economics, 119, 807–859. Lord, Charles G., Lee Ross, and Mark R. Lepper (1979). “Biased Assimilation and Attitude Polarization: The Effects of Prior Theories on Subsequently Considered Evidence.” Journal of Personality and Social Psychology, 37, 2098–2109. Miller, Arthur G., John W. McHoskey, Cynthia M. Bane, and Timothy G. Dowd (1993). “The Attitude Polarization Phenomenon: Role of Response Measure, Attitude Extremity, and Behavioral Consequences of Reported Attitude Change.” Journal of Personality and Social Psychology, 64, 561–574. Munro, Geoffrey D. and Peter H. Ditto (1997). “Biased Assimilation, Attitude Polarization, and Affect in Reactions to Stereotype-Relevant Scientific Information.” Personality and Social Psychology Bulletin, 23, 636–653. Murphy, Kevin and Andrei Shleifer (2004). “Persuasion in Politics.” American Economic Review, 94, 435–439. Piketty, Thomas (1995). “Social Mobility and Redistributive Policies.” Quarterly Journal of Economics, 110, 551–584. Reiss, Peter and Matthew White (2008). “What Changes Energy Consumption? Prices and Public Pressure.” RAND Journal of Economics, 39, 636–663. Schultz, Wesley, Jessica M. Nolan, Robert B. Cialdini, Noah J. Goldstein, and Vladas Griskevicius (2007). “The Constructive, Destructive, and Reconstructive Power of Social Norms.” Psychological Science, 18, 429–434. Thaler, Richard H. and Cass R. Sunstein (2008). Nudge: Improving Decisions about Health, Wealth, and Happiness. Yale University Press. Thalmann, Philippe (2004). “The Public Acceptance of Green Taxes: 2 Million Voters Express Their Opinion.” Public Choice, 119, 179–217. 702 Journal of the European Economic Association Downloaded from https://academic.oup.com/jeea/article-abstract/11/3/680/2300535 by State Univ of NY - Binghamton user on 05 November 2018 Table A.5. Correlates of probability of obtaining a rebate to purchase an energy efficient durable and of spending less than 2 minutes reading the home energy report. Table A.6. Regression discontinuity test of differential responses to a normative message. Figure A.1. Density of household/neighbor kWh, conservatives. Density of the ratio of average daily household kWh to average daily kWh of 100 nearest neighbors living in a similar sized house in the first HER report received. Conservative is defined as Republican, American Party, or Libertarian. Figure A.2. Density of household/neighbor kWh, liberals. Density of the ratio of average daily household kWh to average daily kWh of 100 nearest neighbors living in a similar sized house in the first HER report received. Liberal is defined as Democrat, Green Party, or Peace and Freedom. Figure A.3. Change in electricity consumption for conservatives who receive a her and are at the “Good” versus “Room for Improvement” Cutoff. Figure A.4. Change in electricity consumption for liberals who receive a her and are at the “Good” versus “Room for Improvement” Cutoff.
Global Environmental Change 36 (2016) 124–138 Contents lists available at ScienceDirect Global Environmental Change journal homepage: www.elsevier.com/locate/gloenvcha Social capital, trust, and adaptation to climate change: Evidence from rural Ethiopia Christopher J. Paula,* , Erika S. Weinthala , Marc F. Bellemareb , Marc A. Jeulandc,d a Nicholas School of the Environment, Duke University, Box 90328, Durham, NC 27708, USA Department of Applied Economics and Center for International Food and Agricultural Policy, University of Minnesota, 1994 Buford Avenue, Saint Paul, MN 55108, USA c Sanford School of Public Policy, Duke University, Box 90239, Durham, NC 27708, USA d Institute of Water Policy, Lee Kwan Yew School of Public Policy, National University of Singapore, Singapore b A R T I C L E I N F O A B S T R A C T Article history: Received 27 August 2015 Received in revised form 7 December 2015 Accepted 22 December 2015 Available online 11 January 2016 Climate change is expected to have particularly severe effects on poor agrarian populations. Rural households in developing countries adapt to the risks and impacts of climate change both individually and collectively. Empirical research has shown that access to capital—financial, human, physical, and social—is critical for building resilience and fostering adaptation to environmental stresses. Little attention, however, has been paid to how social capital generally might facilitate adaptation through trust and cooperation, particularly among rural households and communities. This paper addresses the question of how social capital affects adaptation to climate change by rural households by focusing on the relationship of household and collective adaptation behaviors. A mixed-methods approach allows us to better account for the complexity of social institutions—at the household, community, and government levels—which drive climate adaptation outcomes. We use data from interviews, household surveys, and field experiments conducted in 20 communities with 400 households in the Rift Valley of Ethiopia. Our results suggest that qualitative measures of trust predict contributions to public goods, a result that is consistent with the theorized role of social capital in collective action. Yet qualitative trust is negatively related to private household-level adaptation behaviors, which raises the possibility that social capital may, paradoxically, be detrimental to private adaptation. Policymakers should account for the potential difference in public and private adaptation behaviors in relation to trust and social capital when designing interventions for climate adaptation. ã 2015 Elsevier Ltd. All rights reserved. Keywords: Climate change Trust Social capital Adaptation Ethiopia 1. Introduction Climate change is expected to have a profound impact on livelihoods around the world by causing more severe weather events, rising sea levels, and higher average temperatures (IPCC, 2014). Building resilience to climate change depends upon improving existing options for adaptation, especially among vulnerable populations, such as poor rural households in developing countries. Those households adapt to the risks and impacts of climate change in many ways, both individually and collectively (Adger, 2003; Tompkins and Eakin, 2012). The ability of households and their communities to adapt, however, is conditioned by a myriad of factors that are often in short supply for rural households, including access to * Corresponding author. Fax: +919 681 7748. E-mail addresses: cjp2@duke.edu (C.J. Paul), weinthal@duke.edu (E.S. Weinthal), mbellema@umn.edu (M.F. Bellemare), marc.jeuland@duke.edu (M.A. Jeuland). http://dx.doi.org/10.1016/j.gloenvcha.2015.12.003 0959-3780/ ã 2015 Elsevier Ltd. All rights reserved. financial, human, physical, and social capital. While access to all types of capital is critical for building resilience and fostering adaptation to environmental stresses, little attention has been paid to the role of social capital which, following Ostrom and Ahn (2003), we define as the value of relationships that facilitates cooperation and collective action through trust. In the absence of other forms of capital, social capital may be particularly important for promoting adaptation to new threats from climate change by furthering cooperation and collective action. This paper addresses the question of how social capital affects adaptation at the household and community levels in poor rural communities in developing countries. Specifically, we (i) assess the role of social capital in poor, rural communities in the Ethiopian Rift Valley, (ii) test multiple survey and experimental measurements of social capital both qualitative and quantitative, and (iii) evaluate the relationship of our various measures of social capital to individual household and collective community adaptation behaviors. Although scholars have recognized the potential importance of social capital, most work on the role of social capital in adaptation C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 to date has been theoretical or based on case studies or small samples (Adger, 2003; Pelling and High, 2005; Wolf et al., 2010). The gap in applied research on this relationship is evident in reports from the Intergovernmental Panel on Climate Change that state “the capacity for collective action is a critical determinant of the capacity to adapt to climate politics,” yet only have evidence for the loss of social capital in conflict (Adger et al., 2014). Empirical work on this topic is particularly challenging because social capital and climate adaptation are both complex phenomena with a variety of potential mechanisms and effects (Adger et al., 2005; Ostrom and Ahn, 2003). Furthermore, there is limited work on how social capital relates to adaptation behaviors. Most of the empirical literature focuses on the positive benefits arising from social relationships and trust while ignoring the possibility that the complexity of mechanisms and scales of social capital may in some cases reveal a “dark side of social capital” (Bagnasco, 2008; Deth and van Zmerli, 2010; Portes, 1998) which promotes negative outcomes for certain groups. The levels and methods for measuring key variables may also have an influence on the relationships between trust, social capital, and outcomes. For example, trust can be measured using surveys (by asking about trust in general or about trust in the context of specific transactions) and experimental approaches (by simulating trust-based transactions and rewarding subjects accordingly). The measurements obtained using these different methods may not be consistent with one another, and may be differentially influenced by a variety of other characteristics that also affect adaptation (Bouma et al., 2008; Naef and Schupp, 2009; Ostrom, 2005). In addition, alternative measures may only be relevant to one or more specific levels—individual, household, and community—of social capital and adaptation (Smit and Wandel, 2006; Tompkins and Eakin, 2012). In order to better test empirically the relationship between social capital and trust at the household and community levels, and climate change adaptation, we combine multiple methods of data collection, including semi-structured interviews, surveys, and field experiments. This mixed-methods approach allows us to better account for the complexity and scales at which alternative institutions influence climate adaptation behaviors (Adger et al., 2005; Poteete et al., 2010; Vaccaro et al., 2010), and to consider the relevance of different measures of trust and social capital. Our analysis suggests that these different measures may be indicative of different constructs, and it reveals that survey measures of trust are more strongly related to observed behaviors. We also find evidence of a mixed effect of social capital in climate adaptation: social capital is associated with increased cooperative outcomes, but also with reduced private household-level adaptation.1 Though our analysis can obviously not identify causal relationships between various measures of social capital and adaptation to climate change, our results raise the possibility that social capital may, paradoxically, be detrimental to private adaptation, depending on which effect dominates in household behavior. The remainder of this paper is organized as follows. We begin in Section 2 by describing climate adaptation and defining the phenomenon of social capital and its mechanisms. In Section 3, we describe our study site in the rural Rift Valley of Ethiopia, and the specific methods used to evaluate social capital and adaptation. Next, we provide a description of the data in Section 4 and, in 1 Throughout the paper, we use the term “private adaptation” to refer to household-level adaptation, noting that some of these household-level behaviors do involve limited cooperation with other households (e.g., sharing of tools). Community-level or “public adaptation” behaviors represent contribution or participation in the provision of community-level public goods. 125 Section 5, we discuss our results in testing of each of the hypotheses, showing that while social capital is important in collective adaptation activities, it is negatively related to private household adaptation. We then conclude in Section 6 with potential policy implications and directions for future research. 2. Climate adaptation, social capital, and collective action Climate adaptation, “the process of adjustment to actual or expected climate and its effects” (IPCC, 2014), is a process that is both bio-physical and human. While humans have always needed to respond to a changing environment, the current period of global climate change strains human capacity for adaptation because of the combined rapidity and severity of the changes it entails. Individuals must make complex decisions about adaptation that determine the consequences of climate change for livelihoods under increasing uncertainty, for example that arising from changes in water availability, variability in crop yields, and greater extremes of natural disasters. By definition, constraints on adaptation, which are a function of financial, human, and other forms of capital, would appear highest for disadvantaged communities. Adaptation occurs at individual, household, community, and larger institutional scales (Adger et al., 2005). In this paper, we consider three potential levels of adaptation: household, community, and government. At the private household level, adaptation takes forms such as technology adoption, migration, or changes in livelihoods. Community level adaptation may occur through collective action, the ability of a group to achieve a common interest, and the provision of public goods (Olson, 1971; Poteete et al., 2010; Tompkins and Eakin, 2012). Collective action facilitates the pooling of resources, knowledge, and efforts for community responses. We treat collective action as a broad description of cooperative interaction. External interventions such as government programs and interventions can affect adaptation, with or without the input of households and communities, but do not always benefit rural areas because of the lack of infrastructure or state reach, the ability of governments to implement programming and exert power (e.g., Herbst, 2000). The degree of cooperation in rural areas is thus potentially more important in determining outcomes. Explanations for the emergence of collective action have focused on factors such as group size, leadership, and incentives (Olson, 1971), but the value of cooperative social relations and how precisely they emerge remains critical and unclear (Ostrom, 1994; Ostrom and Ahn, 2003). Theories of social capital arose out of work such as that investigating the resources of social networks and the function of social structures (Bourdieu, 1986; Coleman, 1988; Portes, 1998). These theories have been further developed and applied to diverse fields including economic activities, sustainable development, and natural resource management (Dale and Newman, 2010; Dale and Onyx, 2010; Fukuyama, 1995; Pretty and Ward, 2001). Ostrom and Ahn (2003), moreover, specify three components of social capital: institutions, social networks, and trustworthiness. Institutions are the social, economic, and political “rules of the game” that govern interactions (North, 1990); they mediate relationships, and thus influence the outcomes of individual and collective behavior (Agrawal, 2009). Opportunities for cooperation thus arise from the web of relationships that make up social networks (Ostrom and Ahn, 2003). The relationships in these networks are commonly classified as: bonding, the close ties within a group; bridging, the ties between groups; and linking, the vertical relationships across hierarchies (Szreter and Woolcock, 2004; Woolcock, 2001). Dense and stable networks facilitate generalized reciprocity and “trustworthiness,” which are all characteristics that facilitate trust (Putnam et al., 1993). 126 C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 Trust, the confidence that others will act on commitments reliably and with reciprocity, is a core mechanism of social capital for collective action (Ostrom and Ahn, 2003; Putnam, 2001). Trust is dependent upon characteristics of individuals and their setting, including institutions, the nature and extent of social networks, and individual characteristics. Trust may also be related to an individual’s tolerance for risk, since trusting another individual may in many cases carry risks (Schechter, 2007). In the face of threats from climate change, trust affects households’ confidence that they can rely on others for resource sharing, conflict resolution, and sustained cooperation. As a result, higher trust may enhance opportunities for adaptation (Adger, 2003). As an example, the sharing of farm equipment can help illustrate the nature of relationships between trust and social capital. In many situations, farmers may have short-term demand for more physical capital than they personally own. Individuals who are well endowed with such physical capital must decide whether to loan or lease farming equipment to their neighbors, a decision that is influenced by multiple aspects of social capital, including bonding social capital and trust. Bonding social capital helps groups leverage their resources more effectively by sharing risk and cost (Woolcock and Narayan, 2000). Here, trust functions to aid the lender or lessor to make a decision based on a history of past interactions with the potential borrower (i.e., his or her trustworthiness) (Fafchamps, 2004; Platteau, 2000, 1994a, 1994b), on his or her own perceptions, or on other community members’ perceptions of the borrower’s trustworthiness. Thus, social networks may contribute information about agents’ trustworthiness, and may provide recourse in the event that the terms of the transaction are violated. Meanwhile, existing institutions structure transactions, for example by specifying the time over which a loan is allowed, the conditions of enforcement of the agreement (e.g., returning the equipment on time), or the terms of reciprocity. When faced with environmental threats, such as worsening growing conditions, farmers must draw upon social capital, among other forms of capital, to cope. For example, if additional labor and tools are needed for terracing a field to cope with stronger rainstorms, a farmer must determine information about best practices and find other people to contribute, by hiring them or leveraging social relationships. In the example, these network relationships mediate opportunities for information, such as new or best practices. Bridging social capital allows information to be shared between groups. Confidence in information can be determined by linking social capital across vertical levels to government or outside agencies. Ultimately, groups of high bonding social capital can act upon this knowledge. The sharing of information promotes adaptation by combining the human capital of knowledge with the social capital of networks (Falco and Veronesi, 2013a). Trust is, moreover, essential for assessing and acting upon shared information (Creech and Willard, 2001). Finally, adaptation occurs through cooperation and collective action supported by social capital. Community-level adaptation may also depend upon external factors, such as government institutions and programming, which could complement or offset the effect of social capital. Social networks specifically serve multiple types of functions for adaptation and collective action, as networks can be horizontal between peers within a community, or vertical across hierarchies (Putnam et al., 1993). Particularly when higher-level (e.g. statelevel) institutions are absent, the networking function of social capital supports local institutions and collective action responses that are needed for addressing community challenges, including those arising from shocks or crises (Adger, 2003; Bratton, 1989; Platteau, 1994a,b). Thus, the value of social networks is in both facilitating trustworthiness and contributing to the possibility of accessing different (and perhaps collective) resources through multiple venues (Woolcock and Narayan, 2000). Social capital does not necessarily have universally positive effects or serve as insurance mechanisms against adverse shocks. Social capital may be ineffective if there is a general lack of resources or knowledge of effective solutions. In this sense, the community may be the inappropriate scale of action necessary to adapt; rather, adaptation could depend primarily on choices made by the individual household (e.g., migration) or by the state (e.g., aid programming). Trust may not be enough to overcome the transaction costs for collective action. There also may be a “dark side of social capital” (Deth and van Zmerli, 2010), in which strong social institutions can generate negative outcomes or overpower formal legal institutions, as in the case of the mafia (Gambetta, 1988). Groups may be isolated and made less diverse by a process of homophily, the tendency to become more similar, reducing valuable bridging social capital between groups (Newman and Dale, 2007). Decisions involving trust and social capital may also be governed by other decision-making characteristics such as risk preferences (Schechter, 2007). In other words, an individual’s propensity to trust may be partially governed by her willingness to take risks. Finally, it is important to distinguish between social capital of households within communities, and bridging (vertical linking) of social ties beyond communities. Bridging social capital can help link individuals and households to new ideas and resources beyond their community, by either substituting or complementing the role of the state (Adger, 2003). In adaptation, communities that organize and cooperate can better access external support (Karlsson and Hovelsrud, 2015). These types of links can enhance connection with outside organizations and government officials, generating better provision of resources. We hypothesize that social capital influences the ability of households to respond to change. This is because the constituent parts of social capital, and access to collective action processes, influence the quality and set of options (or constraints) that households face when threatened by climate change. Specifically, trust should be associated with collective action and increased adaptation behaviors, perhaps due to information sharing, knowledge mobilization, and resource coordination. Households with higher levels of trust are likely to possess more social capital and are hypothesized to undertake more adaptation activities. 3. Study site and methods 3.1. Study location Ethiopia is one of the fastest growing economies in the world, averaging over 10% annual growth in gross domestic product since 2004. Yet, Ethiopia remains a predominantly poor and rural country, with a national average per capita income of $470 and a population that is 84% rural (World Bank, 2014a, 2014b). Throughout rural Ethiopia, farmers typically use labor-intensive agricultural methods and practice subsistence farming, as most farms are rain-fed and yields are accordingly low (Mengistu, 2006). These factors, coupled with extreme and increasing climate variability, suggest a high degree of potential vulnerability to climate change, especially in drought-prone rural areas such as the Rift Valley (Notre Dame Global Adaptation Index, 2014). The effect of climate change on water supplies in this region could be quite significant (Legesse et al., 2003). The 13 million people living in this region are primarily smallholder and herder households, and have minimal access to financial capital and outside resources for coping with such disruptions. C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 This study spans 20 villages located in four woredas (i.e., districts) of the Ziway-Shala lake basin; communities in this zone share common hydrological and agricultural conditions. The study communities are small and relatively isolated, and mostly homogenous in religious and ethnic make-up. The final sample consists of 20 randomly selected households from each of these communities, for a total of 400 households. We chose the Ethiopian Rift Valley to study adaptation because the region’s rural population is already experiencing (and responding to) significant stresses due to extreme climate variability, the effects of which are likely to become more pronounced in the future (Kassie et al., 2013). In rural Ethiopia, climate change has been linked to reduced income; in the absence of adaptation, further decline in household income is likely to occur (Deressa and Hassan, 2009). Moreover, the semi-arid lowlands of the Rift Valley are similar to many other sub-Saharan locations that are facing or expected to face new threats from climate change (Niang et al., 2014). 3.2. Data collection strategy The villages in our sample were selected using a stratified method. Half of the villages were selected from all 5936 villages within the study area, and half were randomly selected from a list of 50 sites with known poor water quality. (This sampling process comports with a separate study on water quality and health in this region and builds upon prior research (Kravchenko et al., 2014; Rango et al., 2012). Within each sample community, we used a structured field counting randomization method to select 20 households within a two-kilometer radius of the community center. In each selected household, we interviewed both the male and female household head whenever possible. Data collection occurred primarily during the month of February, which is immediately post-harvest for teff and maize, the primary crops in the area. The first element of the data collection was semi-structured interviews, which establish the institutional conditions for adaptation and resource management and also provide context for understanding the survey and experimental data. Specifically, we interviewed community representatives in each of our 20 villages, as well as regional government officials located at the woreda-level, with a set of guiding questions and allowing open-ended responses. Local interviews were complemented by interviews with officials from the central government, foreign donors, and nongovernmental organizations (NGOs). The second data collection component comprises surveys conducted with the 400 selected households. The third component of data collection 127 consisted of field experiments. Specifically, following Schechter (2007) and Tanaka et al. (2010), we played investment and risk games developed to measure individual trust and risk preferences with a male and female member of each household in our sample. These are described in more detail in Section 3.3. The Institutional Review Board approved the study and experimental game protocols. All respondents provided informed consent prior to participation in the study. The confidentiality and anonymity of survey respondents has been maintained. 3.3. Measures of social capital, trust, collective action, and adaptation The multiple data collection activities provide us with a rich set of empirical measures of trust and social capital constructs, and of collective action and agricultural adaptation outcomes. Survey questions corresponding to our key variables are listed in Table 1. 3.3.1. Trust We measure trust through surveys and experimental games. In the survey, we use standardized questions from the General Social Survey (Smith et al., 2011). The specific wording of our survey questions is listed in Table 1. Previous research has indicated that this survey measure of trust, albeit imperfect, is relatively stable and comparable with real world behavior (Glaeser et al., 2000). The experimental games’ measure of trust relies on the investment game, a tool that has been widely used and tested in field experiments (Berg et al., 1995; Ostrom and Walker, 2003). Our design closely follows the model of Schechter (2007), combining a risk game with a trust game in which participants invest and entrust a sum of actual money with another participant. A key advantage of this field experimental methodology is that participants have the potential to earn real money, which is thought to induce truthful revelation of preferences and beliefs as compared to hypothetical payoffs. The trust experimental game is played in a group with household heads who participated in the survey. Approximately three quarters of household heads participated in the game. As described further below, there were no systematic demographic differences between those who participated and those who did not. Each participant is randomly assigned to an anonymous partner, and both partners play the role of sender and receiver. Groups for the trust game were separated by gender because small-scale financial transactions in the study communities, such as interpersonal loans, are often segregated by gender. To play the first role (sender), each participant is given an initial endowment of 10 Birr (approximately 0.50 USD, or 25% of a day’s wage in this region). The sender is then told that she can choose to Table 1 Survey questions for key variable. Key independent variables Variable/question text General trust (survey) Community participation Total adaptation changes “In general, would you say that most people in your village can be trusted or that you cannot trust people in your village?” “Do you or any members of your household participate in any activities for improving your community (outside the immediate limits of your house)?” An index of responses to “In the last 10 years, if you have changed [PRACTICE], for what reason did you do so? (Mark all that apply)” Covariates Player male Player age Player education level Player married Household size Total land area (Ha) Dependency ratio Number of bovine owned Income HH assets What is your gender? What is your age? What is your highest level of schooling? Are you married? Number of members listed on detailed roster Sum of “What is the area of [each] plot you own or rent?” Ratio of number of dependents under 16 on roster to household size How many cows, bulls, oxen, and calves do you own? “Please estimate the total amount of money your household receives in an average year” The sum of total value of ten key asset types (e.g. furniture, technology, transportation) 128 C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 anonymously send none, some, or all of the initial endowment to another anonymous receiver in the group. She is also told that any amount sent will be tripled, and that some of the money sent could then be returned, depending on what the receiver decides to do with it. The sent amount is placed in an envelope with a facilitator, who triples the sum in view of the sender. In the second stage, these envelopes are randomly mixed, and each participant is given a different envelope (not her own). At this point, each individual plays the receiver role, and decides what proportion to return to the original sender. Envelopes with the returned sum are then given back to the original sender. We also implemented an experimental procedure following Schechter (2007) and Tanaka et al. (2010) to elicit risk preferences with the same individuals who participated in the trust games. In the risk experiment, respondents choose between binary lotteries (of known probability and outcomes) to measure parameters establishing the shape and concavity of the individual’s value function, and the degree of loss aversion (Liu, 2013; Tanaka et al., 2010). By including these risk measures in our analyses, we can ensure that behavior we ascribe to trust is not the result of underlying risk preferences. 3.3.2. Community adaptation We evaluate community adaptation activities by asking households in the surveys about their participation and contribution to community improvements and public goods, such as water source maintenance and construction of community water harvesting. In interviews, we also asked about communal activities and organization. In our regression analysis, the dependent variable “community participation” is a binary variable from the survey question “Do you or any members of your household participate in any activities for improving your community (outside the immediate limits of your house)?” The dependent variable “community contribution” is a binary variable from the survey question “Did your household contribute to village activities or services with money or other donations in the past year?” 3.3.3. Private adaptation Private adaptation behaviors were measured directly and indirectly in the survey. We asked households detailed questions about specific changes made to agricultural practices and crops in recent years. Adaptation across multiple behaviors is likely critical for success (Falco and Veronesi, 2013b). Given that there are a variety of different adaptation behaviors reported by households, we constructed indices of these, using a simple count method and principal component analysis (PCA) on polychoric correlations (Kolenikov and Angeles, 2004). In the main regression analysis on private adaptation, we use two specifications of the dependent variable. The “adaptation index” is a simple count (sum) index of twelve adaptation behavior categories, including: Proportion of different crops; Type of seed (traditional vs. improved); Timing of planting; Timing of harvest; Method of farming; Number of livestock; Amount of crops; Farm equipment/assets; Work for income outside the farm; Change total area harvested; Fertilizer use; and, Other. The “adaptation PCA index” is a principal component analysis index of these twelve adaptation behavior categories. The result of the principal component analysis is also presented in the Appendix (Tables A7–A8). 3.3.4. Control variables A number of socioeconomic and demographic characteristics are likely to be important in adaptation behaviors, namely wealth in the form of assets, animals (i.e. livestock), and land; income; household size and the ratio of dependents; and individual characteristics of the household head, including gender, age, education, and marital status. We thus control for these variables in our regressions. 3.4. Hypotheses and analytical methods We make the following hypotheses: H1. If communities and households have limited access to and support from government institutions, then there should be more household and community-level adaptation than government-led adaptation. H2. There is a positive association between survey and experimental game measures of trust. H3. There is a positive association between trust and adaptation for households because trust increases exposure to new and tested adaptation options, and for communities because it facilitates collective action. We analyze H1 using qualitative data from local interviews with community representatives of the institutions relevant to social capital and climate adaptation. This analysis provides context for the household-level and community-level adaptation hypotheses. We evaluate transcribed interviews for dominant themes, which include community concerns, climate change awareness, conflict related to environmental factors, and a particular focus on water. We also study the village interviews to deepen insights on the types of changes from year to year within and between villages. We analyze H2 and H3 using linear regression with household survey and field-experimental data. We control for individual characteristics, socioeconomic covariates at the household level and cluster the standard errors of all estimates at the village level. Village fixed effects control for unobserved heterogeneity among villages. In the trust experiment analyses, we additionally control for the risk preference parameters. To test H2, i.e., that there is a positive association between experimental measures of trust—in terms of proportion of money sent by individual i (Ai)—and the binary survey measures of individual trust (X1,i), we estimate the model in Eq. (1), where the other controls include individual risk preferences (X 2;i ), individual characteristics (X3,i), a vector of controls for household j (W j ), and fixed effects for each distinct village k (Zk): Ai ¼ a þ b1 X 1;i þ b2 X 2;i þ b3 X 3;i þ dW j þ g Z k þ ei ð1Þ For H3, i.e., that there is a positive association between adaptation at both the community and household levels (Y) and trust (X1,i), we evaluate the model shown in Eq. (2), where we again control for individual risk preferences (X 2;i ), individual characteristics (X3,i), household level characteristics (W j ), and include village fixed effects (Zk): Y i ¼ a þ u1 X 1;i þ u 2 X 2;i þ u 3 X 3;i þ zW j þ yZ k þ hi ei ð2Þ We reiterate, however, that our analysis of observational and field-experimental data can only estimate partial correlations between these parameters and adaptation outcomes. In other words, the usual sources of statistical endogeneity, viz. reverse causality or simultaneity, measurement error, and unobserved heterogeneity, are all likely to compromise the causal identification of the parameters of interest in Eqs. 1 and 2. 4. Data For our regression analysis, we use household survey data from 400 households, and risk and trust experiments with 614 male and female household heads from households who participated. The qualitative data used in this paper includes interviews with local representatives in each of the 20 villages across the three waves (2012–2014), for a total of 51 interviews (nine villages are missing one of the waves because a representative was unable to be contacted; but all villages have at least two waves of interviews). C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 129 Table 2 Descriptive statistics of household survey data. Variable Key independent variables General trust (survey) Experimental trust (proportion sent) Community participation Community contribution Household adaptation changes (simple index) Household adaptation changes (PCA index) Covariates Risk alpha parameter Risk sigma parameter Risk lambda parameter Player male Player age Player education level Player married Household Size Total land area (Ha) Dependency ratio Number of bovine owned Income per capita (birr) Log of income per capita Household asset per capita Log of household asset per capita Obs Mean Std. Dev. Min Max 360 360 360 360 360 360 0.42 0.45 0.94 0.43 4.50 0.20 0.49 0.24 0.27 0.50 3.26 1.54 0 0 0 0 0 2.32 1 1 1 1 12 3.43 360 360 360 360 360 360 360 360 360 360 360 360 360 360 360 0.70 1.23 1.53 0.82 40.84 1.60 0.84 6.37 3.49 0.46 4.77 2645 7.0 261.27 4.39 0.19 0.34 1.93 0.38 16.75 1.57 0.36 2.36 12.13 0.23 6.79 18289 1.06 745.38 1.84 0.05 0.05 0.12 0 14 0 0 1 0 0 0 0 2.01 0 -2.30 1.45 1.50 7.85 1 101 6 1 15 201.5 1.50 81 345015 12.75 11576.67 9.36 Villages in our study had an average population of approximately 2000 people. Households had on average of 6.3 members (adults and children), with a mean per capita income of 2623 Birr (USD 137), far below the Ethiopian average of 8995 Birr (USD470) ( World Bank, 2014b). Eighty-nine percent of study households are Oromo, the most populous language group in Ethiopia, and 51% of households are Muslim (as compared to 34% nation-wide (CIA, 2015)). Households cultivate on average 3.4 hectares of land. The primary crops were maize, wheat, and teff. A summary of key statistics is reported in Table 2. As described above, we use a variety of survey questions to assess social capital characteristics. Respondents indicated high levels of community participation, especially in collective activities: 93% of households said they regularly participated in activities to improve the community. Nearly 80% of households reported participating in a community meeting within the two weeks preceding the survey. Fourteen percent of households indicated being active members of religious groups, with about equal participation among Christians and Muslims. Another measure for assessing cooperative behavior is the sharing of resources or labor. Just over half (52%) of households share farming equipment. About 78% of respondents indicate that they expect a loan to be repaid when it is given to others. Fig. 1. Trust game proportion sent and returned. 130 C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 Regarding trust, 40% of all individual male and female respondents reported that most people in their village can be trusted, while 23% reported that people cannot be trusted (and the remaining proportion indicate that “it depends”). This level of trust is higher than a 2007 World Values Survey in Ethiopia, which found 21% of respondents stating that most people can be trusted, as well as a global meta-analysis estimate of 32% (Johnson and Mislin, 2012; WVS, 2014). In our sample, as with the World Values Survey results, males were slightly more likely than females to report trusting others in their village (44% vs. 37%, t-test p < 0.05). In addition, most respondents (61%) said that it was not at all likely that a neighbor would return lost money (e.g. 100 Birr), but only 40% said it was not at all likely a neighbor would return a lost hen (an animal having similar value), which may indicate varying norms for different items, or may reflect challenges in monitoring ownership of less identifiable assets such as money. These rates did not vary substantially by religion. The sizable group of respondents reporting low trust of others may indicate a significant challenge to collective action and social insurance in these communities. Our field-experimental approach allows for a different way to evaluate trust among study participants. The main results from the investment game are presented in Fig. 1. The average proportion of the initial endowment that was sent by players in the first stage was 0.43. This was somewhat higher than the average proportion returned, 0.34, though the average cash amount returned was similar due to the tripling of the amount sent. Men, on average, sent and returned slightly higher amounts than women (p < 0.01). The average proportion initially sent was 0.47 for males and 0.39 for females, while the average proportion returned was 0.37 for males and 0.31 for women (p < 0.01). Senders in the first round also showed a strong tendency to send half of the sum, a common anchor identified in these types of games. The correlation of the proportion sent in the first round to the proportion returned in the second round is 0.38. These results are similar to those found by Schechter (2007) in Paraguay. We would expect there to be a relationship between the “trust” behavior in sending the initial investment, and the “trustworthy” behavior of the altruistic return, but should not expect these parameters to be fully correlated. 4.1. Construction of the final analytical sample To avoid analyzing outcomes across different samples, our final analytical sample for regression includes 360 household heads of the 400 households in the original study for whom we have the full set of experimental measures of trust and all other covariates (as described above and listed in Table 2). We estimated a regression with all households, including those with missing data for experimental measures (n = 400), to see if these households are systematically different on other covariates from those with full experimental data, and no covariates were significant at the p < .05 level. Regressions were also tested for sensitivity to variables with outliers, and there was no significant change in the regression results. 5. Results 5.1. Community-level social capital To evaluate the importance of community-level social capital, we test H1, i.e., that the communities in our study have limited access to government or outside institutions and, in the example of water, depend primarily on local mobilization of resources to respond to hardship. Our interview data suggest that a large proportion (45%) of the communities have limited access to government officials (visits by officials once a month or rarer). Yet, agricultural and health extension, however, are prominent institutions in Ethiopia including in these communities, with significant expansion in reach over the prior decade (Banteyerga, 2011; Spielman et al., 2014). Forty percent of communities in the study have a full time agricultural extension agent (known as a development agent), but even villages without an agent are visited at least once per week. Village representatives reported that development agents may distribute or sell subsidized inputs if they have them available, provide guidance on government recommendations or instructions, and provide training on agricultural topics. Forty-five percent of study communities have a full-time community health worker, and only two (10%) receive less than weekly health worker visits. In the interviews with community representatives, they most frequently complained about poor attention from the water bureau. Only three villages (15%) received even monthly visits from water bureau representatives, and one stated that while “the water bureau comes to teach skills . . . there has not been a meeting this year” (Authors’ Interview, December 2011). One village representative noted that though the water “bureau takes samples, but they do not report” the results to the community” (Authors’ Interview, January 2012). Another said “we have communicated [our concerns] with the woreda water bureau, but the bureau does not give any response, so we have had no further communication. We do not expect a positive response” Table 3 Survey and experimental trust measures. Variables (1) (2) (3) (4) Experimental trust Survey trust Experimental trust Experimental trust Risk alpha Risk sigma Risk lambda Survey trust Constant Observations R-squared Controls Village fixed effects Village clustered standard error 0.0279 (0.0259) 0.330*** (0.0109) 360 0.211 No Yes Yes 0.173 (0.183) 0.00369 (0.0977) 0.00258 (0.0146) 0.0245 (0.0573) -0.00665 (0.0389) 0.00840 (0.00638) 0.425* (0.224) 0.345*** (0.107) 360 0.118 Yes Yes Yes 360 0.254 Yes Yes Yes 0.0298 (0.0570) 0.00654 (0.0402) 0.00848 (0.00641) 0.0304 (0.0251) 0.332*** (0.109) 360 0.257 Yes Yes Yes Notes: ***p < 0.01, **p < 0.05, *p < 0.1 Robust standard errors in parentheses. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio, Livestock, Log Income Per Capita, Log Household Assets. C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 131 Table 4 Trust and community adaptation. (1) (2) (3) (4) (5) (6) Variables Community participation Community participation Community participation Community pontribution Community pontribution Community contribution Survey trust 0.0482** (0.0187) 0.161** (0.0678) 0.0828 (0.0795) 0.0628 (0.0633) 0.00875 (0.00614) 0.720*** (0.196) 0.0492 (0.0404) 0.0899 (0.0847) 0.0623 (0.0621) 0.00904 (0.00626) 0.758*** (0.187) 0.0499** (0.0189) 0.0568 (0.0429) 0.0811 (0.0810) 0.0624 (0.0630) 0.00923 (0.00625) 0.739*** (0.193) 0.0425 (0.131) 0.0654 (0.108) 0.0245* (0.0136) 0.433 (0.258) 0.187 (0.124) 0.0749 (0.140) 0.0660 (0.106) 0.0225 (0.0141) 0.429 (0.254) 0.156** (0.0678) 0.163 (0.135) 0.0474 (0.130) 0.0665 (0.105) 0.0231 (0.0135) 0.487* (0.253) 360 0.220 Yes Yes Yes 360 0.213 Yes Yes Yes 360 0.222 Yes Yes Yes 360 0.170 Yes Yes Yes 360 0.154 Yes Yes Yes 360 0.175 Yes Yes Yes Experimental trust Risk alpha Risk sigma Risk lambda Constant Observations R-squared Controls Village fixed Effects Village clustered standard error Notes: ***p < 0.01, **p < 0.05, *p < 0.1 Robust standard errors in parentheses. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio. Livestock, Log Income Per Capita, Log Household Assets. (Authors’ Interview, February 2014). The lack of communication is important because the water bureau representatives both perform maintenance activities and determine priorities or allocation of government resources for water source development and improvement. Some communities reported that they were successful in reporting problems to the water bureau, but that repairs took multiple months, as there are not enough technicians (Authors’ Interviews, January 2012). Yet, many community representatives complained in interviews about not receiving any support or communication from the water bureau, even when the community initiated an inquiry. One community leader said, “the government does not assist them, and the water bureau ‘knows nothing.’ The water bureau does not matter; it is just a symbolic organization” (Authors’ Interview, January 2012). Another community said they were told to stop drinking the water due to poor quality, “but were not offered an alternative” (Authors’ Interview, February 2013). Another community, whose well had not been working for eight months asked “the water bureau for help, but they have not,” and so the community representatives also “talked to an NGO seven months back, which said they would help, but the NGO has not come” (Authors’ Interview, February 2013). With regards to water supply (a primary concern of these communities), of the 12 communities that had a well, four had service interruptions during the three years of the study period, and of those, two reported having trouble getting assistance from the relevant agency (either the water bureau or an NGO) to restore water supplies. Seven of the 20 communities reported having trouble getting assistance from the Water Bureau more generally. Many communities reported that they must raise all of the funds needed for repairs locally; such repairs are often costly and thus require strong collective action. One community said they had “reported the problem to the Water Bureau, and someone has come twice, but has not fixed it. The community will try to gather money to fix ourselves: this is our personal problem” (Authors’ Interview, February 2013). Some communities mentioned getting occasional help from NGOs when faced with water supply problems, but most were heavily reliant on their own savings, which typically resulted in delayed repairs and consequent water shortages. In one community, villagers experiencing months of water shortage due to a broken pump would have to travel for three hours for drinking water and nine hours for livestock to get water (Authors’ Interview, January 2012). Representatives from communities who sought help from the water bureau described having to wait for days in the woreda seat for the water bureau to respond, and then have to compensate the technician for his travel and per diem. One community “sent 2 people to the water bureau office for 6 days to petition for help” (Authors’ Interview, February 2013). Another water manager described that when a technician comes, “we have to pay the per diem he asks for. If the technician asks for 500B, we pay it as we can’t argue” (Authors’ Interview, December 2011). Notably, the water bureau officials also emphasized their extremely limited resources for responding to community needs, including a problem of insufficient vehicles (motorbikes) and money for fuel needed to reach remote villages. Another measurement of engagement with government institutions is how the community deals with conflict. For less serious violations, communities rely on elders and social ostracizing to punish those held responsible for a conflict. If a conflict was too serious or unable to be resolved, then community leaders said that they would seek the assistance of the police and formal justice system. Across the twenty communities, 15 (75%) had community elders who were noted as an authority for resolving conflict, as compared to only 11 (55%) mentioning official government (kebele and woreda) leaders. The household-level survey data are consistent with the villagelevel data indicating low levels of government involvement. When confronted with worsening conditions, such as water and food supply, less than 10% of individual households surveyed said they sought help from the local or national government, and mostly endured greater hardship. Notably, few households indicated seeking help from the community when affected by poor environmental conditions, and mostly indicated self-reliance and hardship. Only 29% of households had direct interaction with government officials apart from health and development agents, yet 60% percent of households reported having received some form of government assistance, however, primarily healthcare, education, and food assistance. Fifteen percent of households had received food and nutrition aid, and 17% had received government training. These results suggest that government is neither absent 132 C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 Table 5 Trust and private adaptation. (1) (2) (3) (4) (5) (6) Variables Adaptation index Adaptation index Adaptation index Adaptation PCA index Adaptation PCA index Adaptation PCA index Survey trust 0.973** (0.377) 0.456** (0.179) 0.0154 (0.996) 0.493 (0.598) 0.187** (0.0721) 0.672 (2.519) 0.687 (0.752) 0.200 (0.984) 0.492 (0.592) 0.196** (0.0718) 0.495 (2.578) 0.956** (0.369) 0.542 (0.758) 0.0315 (0.983) 0.489 (0.599) 0.192** (0.0708) 0.852 (2.639) 0.00729 (0.476) 0.229 (0.278) 0.0850** (0.0340) 1.948 (1.181) 0.315 (0.358) 0.0937 (0.471) 0.228 (0.275) 0.0888** (0.0339) 2.033 (1.210) 0.448** (0.175) 0.247 (0.358) 0.0146 (0.469) 0.227 (0.279) 0.0871** (0.0334) 1.866 (1.239) 360 0.182 Yes Yes Yes 360 0.164 Yes Yes Yes 360 0.183 Yes Yes Yes 360 0.177 Yes Yes Yes 360 0.160 Yes Yes Yes 360 0.178 Yes Yes Yes Experimental trust Risk alpha Risk sigma Risk lambda Constant Observations R-squared Controls Village fixed effects Village clustered standard error Notes: ***p < 0.01, **p < 0.05, *p < 0.1 Robust standard errors in parentheses. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio, Livestock, Log Income Per Capita, Log Household Assets. nor prominent in the lives of the study households. Overall, our qualitative results from the interviews and surveys support the hypothesis that communities in our study have limited access and support from government institutions for climate-related adaptation, especially as it relates to management of water resources, and therefore must draw on internal resources to meet many of their needs. 5.2. Trust measurement results (hypothesis 2) For H2, we test the null hypothesis of no association between survey and experimental measures of trust (Table 3). In this case, a rejection of the null due to a positive coefficient would provide evidence in favor of H2. The coefficient of experimental trust regressed on survey trust is positive, as expected, but it is modest in size and not statistically significant. The lack of a statistically significant relationship between the experimental and survey measures of trust may indicate that the experiment was not well understood, that the experimental results primarily reflect risk preferences (given that respondents may consider the investment sent to an anonymous community member to be risky), or that the survey and experimental trust variables relate to different constructs. For example, the experimental measure may be specific to shared financial investments, whereas the survey measures may measure more general trust in other people. Alternatively, the lack of relationship may indicate that the survey questions do not yield reliable measures of trust, given that they are filtered through subjective perceptions and are possibly influenced by respondent-enumerator interactions or differing relative conceptions of what the word “trust” actually means. The division between private and community benefits of social capital may also affect the results. The survey questions are about generalized trust and community activities, whereas the experimental game outcome results in a private gain from trusting behavior, these tools may measure somewhat different characteristics. Interestingly, neither survey nor experimental measures of trust appear significantly related to risk preferences. 5.3. Trust and adaptation results (hypothesis 3) For hypothesis 3, we test whether there is a positive association between trust and both communal and private adaptation activities. Our main results for communal adaptation are displayed in Table 4. These results suggest that observational measures of trust significantly predict contribution to public goods, and collective action that may support adaptation, both in terms of participation and contributions of money. This is consistent with the hypothesized role of social capital in collective action. We do not find evidence of a similar relationship between our experimental measure of trust and participation in community adaptation or monetary contributions to public goods. In the Appendix A (Tables A1–A3), we present additional results that test the robustness of the relationships identified in Table 4 using alternative measures of community participation from different survey years, or based on an index of participation, and alternative measures of contributions (from different survey years). We also test whether results are sensitive to the inclusion of controls and village fixed effects. Generally speaking, we find that the results are insensitive to the inclusion of the latter variables, but that survey trust from 2013 is only weakly related to participation and contributions in 2014. Moreover, survey trust has a weak negative relationship with the participation index, which is largely driven by a relatively small number of households who report both participation in many activities and low trust. The experimental trust measures remain insignificant across all of these additional analyses. Also noteworthy is the fact that the R-squared values are low; this is not uncommon in cross-sectional analyses of heterogeneous socio-economic variables, but it nevertheless suggests that our models explain relatively little of the variance in our outcome variables. This is consistent with the idea that there remain unobserved factors, such as confidence in information about adaptation, that inform decisions to undertake adaptation behaviors besides the social capital and other control variables included in our model specifications. The qualitative interviews suggest that a variety of sources of information, from government, traditional knowledge, and peers influence adaptation decision making by individual households. Despite the positive link between survey trust and participation and contributions, survey trust is negatively related to private adaptation behaviors as measured through the two indices of these behaviors (Table 5). This result suggests that social capital may be detrimental to private adaptation. The experimental trust measures are again not significantly related to these outcomes, but also have negative signs for both private and communal adaptation C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 activities. In additional robustness checks (Appendix A Tables A4–A6), we find some consistency with the patterns described above for community participation and contributions, in that the results do not appear sensitive to inclusion of controls. Yet the relationships are weakly positive between longer-term measures of adaptation and survey trust. This suggests the need for additional longitudinal work to better understand the relationships between these variables over time. Finally, we note that in the analyses, the risk loss parameter (lambda in Table 5 and in Tables A4–A6) is consistently and significantly related to different measures of private adaptation: the higher the loss aversion, the greater the engagement in private adaptation behaviors. Socioeconomic factors such as education and wealth are positively, if insignificantly, associated with both forms of adaptation. Land area, however, is negatively associated with private adaptation, which may reflect the high implementation cost of adaptation, or may indicate that larger land ownership provides greater intrinsic diversification or buffer against shocks. 6. Conclusions We investigated the relationship between social capital and climate adaptation-relevant behaviors using survey and experimental measures of trust. In so doing, we found that a survey measure of trust is positively and significantly related to engagement in community-improving activities, but negatively and significantly related to private household-level adaptation activities. This may stem from the fact that people who can rely on community members are less likely to try new activities as individuals, or that individuals who engage in individual adaptation are less likely to engage with the community, even when controlling for wealth. A negative relationship between household adaptation and trust is surprising, as it suggests the possibility that social capital is unhelpful or even detrimental to adaptation. Conversely, trust and social capital could be considered to substitute for private adaptation. As a third alternative, and because our study cannot establish causal relationships, the negative relationship of household adaptation and trust may reflect an erosion of trust due to private adaptation, or the influence of other omitted variables that are positively correlated with trust that also impede adaptation. Our interviews with community representatives suggest that government programs and instructions are considered important when they occur, though they do not occur at a high frequency. If government instructions dictate activities related to climate adaptation, this might alter patterns of private adaptation. With regards to measurement of household behavior and characteristics, we found a statistically insignificant relationship between survey and experimental measures of trust, suggesting that further development of theory linking social capital and trust, and additional empirical tools to measure these constructs, may be necessary. Simulations and interactive activities similar to the trust experiment may have the dual effect of indicating the level of social capital and providing an opportunity to enhance cooperation. An alternative possible critique of the underlying theoretical link between social capital and adaptation arises from concerns over the effectiveness of social capital for achieving better outcomes at the household level. It is useful to consider social capital as a form of value, because the ability to share knowledge, cooperate, and share other forms of capital remains useful, even if all households are negatively affected by a climate-related stresses. Another possibility is that social capital presents an analytical paradox if smaller and more homogenous groups have greater social capital or possibility of collective action. The effect of heterogeneity within a group is complex, however, and inconsistent (Olson, 1971; Poteete and Ostrom, 2004). Social capital may 133 have value for explaining adaptive capacity, but a true test of the causal impact of social capital requires exogenous modification of social relationships, which is difficult to do in any setting. An empirical implication of this work for the Ethiopian context is that policy makers should be aware of the potential heterogeneity in social capital and its effects: social capital is not necessarily uniformly good, yet neither is it unimportant. Social capital may be useful in some settings, but not useful or even detrimental in others. From the interview and survey data, it is clear that rural Ethiopians in communities similar to those in this study still have limited support from the state, particularly as documented in the case of water supplies. Given our main finding that suggests a difference between those engaged in household adaptation and those engaged in community adaptation, Ethiopian policymakers should be aware of the impacts of different forms of adaptation being promoted. While it is unclear if policy should or can be used to increase social capital with regards to adaptation, some research suggests useful interventions in this arena, such as institutional design for participation and community building activities (Aldrich, 2012; Ostrom, 1992; Putnam, 2001). Future research involving multiple qualitative and quantitative methods, as used this in this paper, can better identify the relevant variables influencing climate adaptation behavior. Further, using mixed methods at multiple scales, though intensive in time and resources, generates more relevant policy prescriptions. Local-level policymaking is the appropriate scale at which to integrate social capital into climate adaptation, yet it is important to draw upon a comparative perspective of experiences of adaptation in other locations and at different policy scales (Vogel and Henstra, 2015). Policymakers may need to account for multiple scales and forms of adaptation, for the individual, household, and community, when designing interventions. Acknowledgements This paper was completed with support from a USAID Conflict Management and Mitigation grant (#AID-OAA-A-12-00068), the Duke University Global Health Institute, and the Nicholas School of the Environment. Courtney Harrison, Tewodros Rango, Eshetu Lemma, and all our colleagues in Ethiopia were essential for this project. This study is made possible by the support of the American people through the United States Agency for International Development (USAID). The opinions expressed herein are those of the authors and do not necessarily reflect the views of USAID. Appendix A. See Tables A1–A8 References Adger, W.N., 2003. Social capital, collective action, and adaptation to climate change. Econ. Geogr. 387–404. doi:http://dx.doi.org/10.1111/j.1944-8287.2003. tb00220.x. Adger, W.N., Arnell, N.W., Tompkins, E.L., 2005. Successful adaptation to climate change across scales. Glob. Environ. Change 15, 77–86. doi:http://dx.doi.org/ 10.1016/j.gloenvcha.2004.12.005. Adger, W.N., Pulhin, J.M., Barnett, J., Dabelko, G.D., Hovelsrud, G.K., Levy, M., Ú. Oswald, S, Vogel, C.H., 2014. Climate change 2014: impacts, adaptation, and vulnerability. Part A: global and sectoral aspects. contribution of working Group II to the fifth assessment report of the intergovernmental panel of climate change. In: Field, C.B., Barros, Dokken, V.R., Mach, D.J., Mastrandrea, K.J., Bilir, M. D., Chatterjee, T.E., Ebi, M., Estrada, K.L., Genova, Y.O., Girma, R.C., Kissel, B., Levy, E.S., MacCracken, A.N., Mastrandrea, S., PR, White, L.L. (Eds.), Human Security. Cambridge University Press Cambridge, United Kingdom and New York, NY,USA pp. 755–791. Agrawal, A., 2009. The Role of local institutions in adaptation to climate change. Social Dimensions of Climate Change: Equity and Vulnerability in a Warming World. World Bank Publications, Washington, DC. 134 C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 Table A1 Community participation. (1) (2) (3) (4) (5) (6) Variables Community participation Community participation Community participation Community participation Community participation 2014 Community participation 2014 Survey trust 0.0709*** (0.0219) -0.0102 (0.0424) 0.0482** (0.0187) 0.0828 (0.0795) 0.0628 (0.0633) 0.00875 (0.00614) 0.720*** (0.196) 0.0492 (0.0404) 0.0899 (0.0847) 0.0623 (0.0621) 0.00904 (0.00626) 0.758*** (0.187) 0.0584* (0.0304) 0.0122 (0.0836) 0.974*** (0.0175) 0.0499** (0.0189) 0.0568 (0.0429) 0.0811 (0.0810) 0.0624 (0.0630) 0.00923 (0.00625) 0.739*** (0.193) 0.914*** (0.0317) 0.0338 (0.0291) 0.0397 (0.0723) 0.0630 (0.0582) 0.0206 (0.0521) 0.00613 (0.0116) 0.624*** (0.156) 360 0.070 No Yes Yes 360 0.222 Yes Yes Yes 360 0.220 Yes Yes Yes 360 0.213 Yes Yes Yes 347 0.109 No Yes Yes 347 0.238 Yes Yes Yes Experimental trust Risk alpha Risk sigma Risk lambda Constant – Observations R-squared Controls Village fixed effects Village clustered standard error Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio, Livestock, Log Income Per Capita, Log Household Assets. The dependent variable “community participation” is a binary variable from the survey question “Do you or any members of your household participate in any activities for improving your community (outside the immediate limits of your house) ?”. Regressions 1–4 –use the 2013 survey, regressions 5–6 use the 2014 survey question regressed on the 2013 covariates. Table A2 Community participation index. (1) Variables – Survey trust (2) Constant 5.715*** (0.199) 0.315* (0.169) 0.859 (0.597) -0.196 (0.537) 0.240 (0.531) 0.0136 (0.0985) 0.715 (1.624) Observations R-squared Controls Village fixed effects Village clustered standard error 360 0.056 No Yes Yes 360 0.209 Yes Yes Yes Experimental trust (3) (4) (5) (6) Participation index Participation index Participation index Participation index Participation PCA index Participation PCA index 0.133 (0.157) 0.540 (0.509) Risk alpha Risk sigma Risk lambda 0.341* (0.182) 0.221 (0.552) 0.235 (0.523) 0.00631 (0.0979) 1.000 (1.537) 0.907 (0.602) 0.140 (0.552) 0.239 (0.537) 0.0148 (0.0983) 0.832 (1.661) 360 0.204 Yes Yes Yes 360 0.205 Yes Yes Yes 0.0221 (0.0837) 0.431 (0.311) 0.556*** (0.117) 0.114 (0.102) 0.521 (0.344) 0.119 (0.372) 0.0738 (0.289) 0.0280 (0.0555) 1.965** (0.797) 360 0.059 Yes Yes Yes 360 0.157 Yes Yes Yes Robust standard errors in parentheses ***p < 0.01, **p < 0.05, *p < 0.1. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio, Livestock, Log Income Per Capita, Log Household Assets. The dependent variable “participation index” is a simple count (sum) index of twelve adaptation behavior categories, including: Sweeping public streets; Cleaning drains; Cleaning water sources; Cleaning school area; Removing garbage; Planting trees; Cleaning community latrines; Well maintenance; Security patrols; Terracing; Bridge or road maintenance; Water Conservation; Construct water harvesting. “Participation PCA index” is a polychoric component analysis (PCA) index of these categories. C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 135 Table A3 Community contribution. (1) (2) (3) (4) (5) (6) Variables Community contribution Community contribution Community contribution Community contribution Community Contribution 2014 Community contribution 2014 Survey trust 0.181*** (0.0625) 0.185 (0.124) 0.161** (0.0678) 0.00340 (0.0425) 0.0755 (0.0976) 0.0425 (0.131) 0.0654 (0.108) 0.0245* (0.0136) 0.433 (0.258) 0.187 (0.124) 0.0749 (0.140) 0.0660 (0.106) 0.0225 (0.0141) 0.429 (0.254) 0.00926 (0.0434) 0.126 (0.101) 0.397*** (0.0392) 0.0847 (0.313) 360 0.170 Yes Yes Yes 360 0.154 Yes Yes Yes 347 0.033 No Yes Yes 347 0.079 Yes Yes Yes Constant 0.123*** (0.0419) 0.156** (0.0678) 0.163 (0.135) 0.0474 (0.130) 0.0665 (0.105) 0.0231 (0.0135) 0.487* (0.253) Observations R-squared Controls Village fixed effects Village clustered standard error 360 0.113 No Yes Yes 360 0.175 Yes Yes Yes Experimental trust Risk alpha Risk sigma Risk lambda Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio, Livestock, Log Income Per Capita, Log Household Assets. The dependent variable “community contribution” is a binary variable from the survey question “Does your household contribute to village activities or services with money or other donations in the past year? Regressions 1–4 use the 2013 survey, regressions 5–6 use the 2014 survey question regressed on the 2013 covariates. Table A4 Household adaptation simple index. (1) (2) (3) (4) (5) (6) Variables Adaptation index Adaptation index Adaptation index Adaptation index Adaptation 2014 index Adaptation 2014 Index Survey trust 0.851** (0.352) 0.758 (0.844) 0.973** (0.377) 0.0154 (0.996) 0.493 (0.598) 0.187** (0.0721) 0.672 (2.519) 0.687 (0.752) 0.200 (0.984) 0.492 (0.592) 0.196** (0.0718) 0.495 (2.578) 0.0649 (0.115) 0.108 (0.189) 0.304*** (0.0839) 0.0413 (0.116) 0.0586 (0.156) 0.00142 (0.255) 0.325** (0.132) 0.0269 (0.0322) 1.072** (0.483) 360 0.182 Yes Yes Yes 360 0.164 Yes Yes Yes 360 0.069 No Yes Yes 360 0.124 Yes Yes Yes Constant 4.828*** (0.339) 0.956** (0.369) 0.542 (0.758) 0.0315 (0.983) 0.489 (0.599) 0.192** (0.0708) 0.852 (2.639) Observations R-squared Controls Village fixed effects Village clustered standard error 360 0.123 No Yes Yes 360 0.183 Yes Yes Yes Experimental trust Risk alpha Risk sigma Risk lambda Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio, Livestock, Log Income Per Capita, Log Household Assets. The dependent variable ‘adaptation index’ is a simple count (sum) index of twelve adaptation behavior categories, including: Proportion of different crops; Type of seed (traditional vs. improved); Timing of planting; Timing of harvest; Method of farming; Number of livestock; Amount of crops; Farm equipment/assets; Work for income outside the farm; Change total area harvested; Fertilizer use; and, Other. For regressions 1–4, this index was generated from the 2013 survey data for adaptations done over the prior 10 years. For regression 5–6, the index was generated from 2014 survey data for adaptations over the prior 5 years. 136 C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 Table A5 Household adaptation PCA index. (1) (2) (3) (4) Variables Adaptation PCA index Adaptation PCA index Adaptation PCA index Adaptation PCA index Survey trust 0.401** (0.166) 0.354 (0.397) 0.456** (0.179) 0.00729 (0.476) 0.229 (0.278) 0.0850** (0.0340) 1.948 (1.181) -0.315 (0.358) 0.0937 (0.471) 0.228 (0.275) 0.0888** (0.0339) 2.033 (1.210) 360 0.177 Yes Yes Yes 360 0.160 Yes Yes Yes Constant 0.0722 (0.161) 0.448** (0.175) 0.247 (0.358) 0.0146 (0.469) 0.227 (0.279) 0.0871** (0.0334) 1.866 (1.239) Observations R-squared Controls Village fixed effects Village clustered standard error 360 0.120 No Yes Yes 360 0.178 Yes Yes Yes Experimental trust Risk alpha Risk sigma Risk lambda Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio, Livestock, Log Income Per Capita, Log Household Assets. The dependent variable “adaptation PCA index” is a polychoric component analysis (PCA) index of twelve adaptation behavior categories, including: Proportion of different crops; Type of seed (traditional vs. improved); Timing of planting; Timing of harvest; Method of farming; Number of livestock; Amount of crops; Farm equipment/assets; Work for income outside the farm; Change total area harvested; Fertilizer use; and, Other. This index was generated from the 2013 survey data for adaptations done over the prior 10 years. Table A6 Household adaptation binary. (1) (2) (3) (4) (5) (6) (7) Variables Adaptation 10years binary Adaptation 10years binary Adaptation 10years binary Adaptation 10years binary Adaptation 2014 5years Binary Adaptation 2014 5years Binary Adaptation 1year Adaptation 1year 2014 Binary 2014 Binary Survey trust 0.0205 (0.0502) 0.0558 0.00159 (0.0510) 0.0214 0.00224 (0.0501) 0.0216 0.124** (0.0537) 0.209 0.0963* (0.0480) 0.185 0.0842* (0.0447) 0.133* 0.0614 (0.0418) 0.0875 (0.0897) (0.0900) 0.134 (0.0989) 0.120* (0.0654) 0.0317*** (0.00929) 0.228 (0.249) (0.122) 0.350*** (0.0483) (0.116) 0.0132 (0.162) 0.189** (0.0793) 0.0207 (0.0215) 0.444 (0.281) (0.0724) 0.135 (0.0971) 0.120* (0.0649) 0.0319*** (0.00924) 0.235 (0.248) 0.182*** (0.0326) (0.0868) 0.0319 (0.133) 0.158*** (0.0517) 0.0188 (0.0133) 0.286 (0.223) Experimental trust (8) Constant 0.709*** (0.0314) (0.0923) 0.134 (0.0961) 0.120* (0.0654) 0.0317*** (0.00924) 0.228 (0.249) Observations R-squared Controls Village fixed effects Village clustered standard error 360 0.069 No Yes 360 0.137 Yes Yes 360 0.137 Yes Yes 360 0.137 Yes Yes 360 0.100 No Yes 360 0.187 Yes Yes 360 0.076 No Yes 360 0.147 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Risk alpha Risk sigma Risk lambda Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Ordinary Least Squares (OLS) Model. Controls: Male, Age, Education, Marital Status, Household Size, Log Land Area, Dependency Ratio, Livestock, Log Income Per Capita, Log Household Assets. The dependent variable “adaptation binary” is a binary variable indicating action of at least one of twelve adaptation behavior categories, including: Proportion of different crops; Type of seed (traditional vs. improved); Timing of planting; Timing of harvest; Method of farming; Number of livestock; Amount of crops; Farm equipment/assets; Work for income outside the farm; Change total area harvested; Fertilizer use; and, Other. For regressions 1–4, this index was generated from the 2013 survey data for adaptations done over the prior 10 years. C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 137 0.734 0.725 0.595 0.638 0.631 0.718 0.942 0.365 0.563 0.681 0.965 1 0.763 0.973 0.359 0.330 0.409 0.583 0.622 0.486 0.668 0.568 1 0.697 0.651 0.328 0.116 0.150 0.781 0.345 0.695 0.530 0.982 0.697 0.572 0.401 0.691 0.605 0.561 0.531 0.893 0.535 0.131 -0.004 0.706 0.766 0.547 0.302 0.606 0.742 1 1 0.484 1 0.843 0.467 1 0.549 0.697 0.324 1 0.502 0.699 0.640 0.731 1 0.771 0.557 0.853 0.837 0.724 Area farmed Farm equipment/assets Amount of crops Fertilizer Use Timing of harvest Work for income outside of the farm Number of livestock Method of farming Other changes Proportion of different crops Type of seed (traditional vs. improved) Timing of planting Farm equipment/ assets Area farmed Adaptation changes Table A7 Household adaptation polychoric correlation matrix. Proportion of different crops Fertilizer use Timing of harvest Work for income outside of the farm Number of livestock 1 0.741 0.340 Method of farming 1 0.969 Other changes 1 Type of seed (traditional vs. improved) Proportion of different crops 1 Timing of planting Table A8 Household adaptation principal component analysis. k Eigenvalues Proportion explained Cumulative explained 1 2 3 4 5 6 7 8 9 10 11 12 7.796 1.604 0.999 0.747 0.514 0.325 0.260 0.174 0.091 0.028 0.141 0.342 0.650 0.134 0.083 0.062 0.043 0.027 0.022 0.014 0.008 0.002 0.012 0.028 0.650 0.783 0.867 0.929 0.972 0.999 1.020 1.035 1.043 1.040 1.028 1.000 Aldrich, D., 2012. Building Resilience. University of Chicago, Chicago. Bagnasco, A., 2008. The blackwell companion to political sociology. In: Nash, K., Scott, A. (Eds.), Trust and Social Capital. John Wiley & Sons. Banteyerga, H., 2011. Ethiopia’s health extension program: improving health through community involvement. MEDICC Rev. 13, 46–49. doi:http://dx.doi. org/10.1590/s1555-79602011000300011. Berg, J., Dickhaut, J., McCabe, K., 1995. Trust, reciprocity, and social history. Games Econ. Behav. 10, 122–142. doi:http://dx.doi.org/10.1006/game.1995.1027. Bouma, J., Bulte, E., van Soest, D., 2008. Trust and cooperation: Social capital and community resource management. J. Environ. Econ. Manag. 56, 155–166. doi: http://dx.doi.org/10.1016/j.jeem.2008.03.004. Bourdieu, P., 1986. The Forms of Capital. In: Richardson, J. (Ed.), Handbook of Theory and Research for the Sociology of Education. Greenwood, New York. Bratton, M., 1989. Beyond the state: civil society and associational life in africa. World Polit. 41, 407–430. doi:http://dx.doi.org/10.2307/2010506. CIA, 2015. The World Factbook: Ethiopia [WWW Document]. URL https://www.cia. gov/library/publications/the-world-factbook/geos/et.html. Coleman, J.S., 1988. Social capital in the creation of human capital. Am. J. Sociol. S95– 120. Creech, H., Willard, T., 2001. Strategic Intentions: Managing Knowledge Networks for Sustainable Development. International Institute for Sustainable Development, Winnipeg, Manitoba. Dale, A., Newman, L., 2010. Social capital: a necessary and sufficient condition for sustainable community development? Community Dev. J. 45, 5–21. Dale, A., Onyx, J., 2010. A Dynamic Balance: Social Capital and Sustainable Community Development. UBC Press Vancouver, British Columbia. Deressa, T.T., Hassan, R.M., 2009. Economic impact of climate change on crop production in ethiopia: evidence from cross-section measures. J. Afr. Econ. doi: http://dx.doi.org/10.1093/jae/ejp002 ejp002. Deth, J.W., van Zmerli, S., 2010. Introduction: civicness, equality, and democracy—a dark side of social capital? Am. Behav. Sci. 53, 631–639. doi:http://dx.doi.org/ 10.1177/0002764209350827. Fafchamps, M., 2004. Market Institutions in Sub-saharan Africa: Theory and Evidence. MIT Press, Cambridge, MA. Falco, S.D., Veronesi, M., 2013a. Managing environmental risk in presence of climate change: the role of adaptation in the Nile basin of Ethiopia. Environ. Resour. Econ. 57, 553–577. doi:http://dx.doi.org/10.1007/s10640-013-9696-1. Falco, S.D., Veronesi, M., 2013b. How can african agriculture adapt to climate change? A counterfactual analysis from Ethiopia. Land Econ. 89, 743–766. doi: http://dx.doi.org/10.3368/le.89.4.743. Fukuyama, F., 1995. Social capital and the global economy. Foreign Aff. 74, 89–103. doi:http://dx.doi.org/10.2307/20047302. Gambetta, D., 1988. Trust: Making and Breaking Cooperative Relations. B. Blackwell, New York, NY,USA. Glaeser, E.L., Laibson, D.I., Scheinkman, J.A., Soutter, C.L., 2000. Measuring trust. Q. J Econ. 115, 811–846. doi:http://dx.doi.org/10.1162/003355300554926. Herbst, J.I., 2000. States and Power in Africa: Comparative Lessons in Authority and Control. Princeton University Press, Princeton, N. J. IPCC, 2014. Climate change 2014: impacts, adaptation, and vulnerability. Part A: global and sectoral aspects. contribution of working Group II to the fifth assessment report of the intergovernmental panel on climate change. In: Field, C.B., Barros, V.R., Dokken, D.J., Mach, K.J., Mastrandrea, M.D., Bilir, T.E., Chatterjee, M., Ebi, K.L., Estrada, Y.O., Genova, R.C., Girma, B., Kissel, E.S., Levy, A. N., MacCracken, S., Mastrandrea, P.R., White, L.L. (Eds.), Summary for Policymakers. Cambridge University Press Cambridge, United Kingdom, and New York, NY,USA pp. 1–32. Johnson, N.D., Mislin, A., 2012. How much should we trust the world values survey trust question? Econ. Lett. 116, 210–212. doi:http://dx.doi.org/10.1016/j. econlet.2012.02.010. Karlsson, M., Hovelsrud, G.K., 2015. Local collective action: adaptation to coastal erosion in the monkey river village, Belize. Glob. Environ. Change 32, 96–107. doi:http://dx.doi.org/10.1016/j.gloenvcha.2015.03.002. Kassie, B.T., Hengsdijk, H., Rötter, R., Kahiluoto, H., Asseng, S., Ittersum, M.V., 2013. Adapting to climate variability and change: experiences from cereal-based farming in the central rift and Kobo Valleys, Ethiopia. Environ. Manage. 52, 1115–1131. doi:http://dx.doi.org/10.1007/s00267-013-0145-2. 138 C.J. Paul et al. / Global Environmental Change 36 (2016) 124–138 Kolenikov, S., Angeles, G., 2004. The use of discrete data in PCA: theory simulations, and applications to socioeconomic indices. Measure. Measure Evaluation (Publication). Kravchenko, J., Rango, T., Akushevich, I., Atlaw, B., McCornick, P.G., Merola, R.B., Paul, C., Weinthal, E., Harrison, C., Vengosh, A., 2014. The effect of non-fluoride factors on risk of dental fluorosis: evidence from rural populations of the Main Ethiopian Rift. Sci. Total Environ. 488, 595–606. Legesse, D., Vallet-Coulomb, C., Gasse, F., 2003. Hydrological response of a catchment to climate and land use changes in Tropical Africa: case study South Central Ethiopia. J. Hydrol. 275, 67–85. doi:http://dx.doi.org/10.1016/s00221694(03)00019-2. Liu, E.M., 2013. Time to change what to sow: risk preferences and technology adoption decisions of cotton farmers in China. Rev. Econ. Stat. 95, 1386–1403. doi:http://dx.doi.org/10.1162/rest_a_00295. Mengistu, A., 2006. Country Pasture/Forage Resource Profile: Ethiopia. Food and Agricultural Organization, Rome. Naef, Michael, Schupp, Jürgen, Measuring Trust: Experiments and Surveys in Contrast and Combination (March 23, 2009). SOEP paper No. 167 Available at SSRN: http://ssrn.com/abstract=1367375 or http://dx.doi.org/10.2139/ ssrn.1367375 Institute for the Study of Labor IZA DP No. 4087; Berlin, Germany. Newman, L., Dale, A., 2007. Homophily and agency: creating effective sustainable development networks. Environ. Dev. Sustain. 9, 79–90. Niang, I., Ruppel, O.C., Abdrabo, M.A., Essel, A., Lennard, C., Padgham, J., Urquhart, P., 2014. Climate change 2014: impacts, adaptation, and vulnerability. Part B: regional aspects. Contribution of working group ii to the fifth assessment report of the intergovernmental panel of climate change. In: Barros, V.R., Field, C.B., Dokken, D.J., Mastrandrea, M.D., Mach, K.J., Bilir, T.E., Chatterjee, M., Ebi, K.L., Estrada, Y.O., Genova, R.C., Girma, B., Kissel, E.S., Levy, A.N., MacCracken, S., Mastrandrea, P.R., White, L.L. (Eds.), Africa. Cambridge University Press, Cambridge, United Kingdom and New York, NY,USA, pp. 1199–1265. North, D.C., 1990. Institutions, Institutional Change and Economic Performance. Cambridge University Press, Cambridge; New York. Notre Dame Global Adaptation Index, 2014. Notre Dame Global Adaptation Index. Notre Dame, South Bend, IN. Olson, M., 1971. The Logic of Collective Action Public Goods and the Theory of Groups. Harvard University Press, Cambridge, MA. Ostrom, E., 2005. Understanding Institutional Diversity. Princeton University Press. Ostrom, E., 1994. Constituting Social Capital and Collective Action. J. Theor. Polit. 6, 527–562. doi:http://dx.doi.org/10.1177/0951692894006004006. Ostrom, E., 1992. Crafting institutions for self-governing irrigation systems. Center for Self-Governance. San Francisco, California. Ostrom, E., Ahn, T.K., 2003. Foundations of Social Capital. Edward Elgar Pub., Northampton, MA. Ostrom, E., Walker, J., 2003. Trust and Reciprocity: Interdisciplinary Lessons for Experimental Research. Russell Sage Foundation. Pelling, M., High, C., 2005. Understanding adaptation: what can social capital offer assessments of adaptive capacity? Glob. Environ. Change 15, 308–319. doi: http://dx.doi.org/10.1016/j.gloenvcha.2005.02.001. Platteau, J.-P., 2000. Institutions, Social Norms, and Economic Development. Psychology Press. Platteau, J.-P., 1994a. Behind the market stage where real societies exist-part I: the role of public and private order institutions. J. Dev. Stud. 30, 533–577. doi:http:// dx.doi.org/10.1080/00220389408422328. Platteau, J.-P., 1994b. Behind the market stage where real societies exist-part II: the role of moral norms. J. Dev. Stud. 30, 753–817. doi:http://dx.doi.org/10.1080/ 00220389408422338. Portes, A., 1998. Social capital: its Origins and Applications in Modern Sociology. Annu. Rev. Sociol. 24, 1–24. Poteete, A., Janssen, M., Ostrom, E., 2010. Working Together: Collective Action, the Commons and Multiple Methods in Practice. Princeton University Press, Princeton, N. J. Poteete, A., Ostrom, E., 2004. Heterogeneity, group size and collective action: the role of institutions in forest management. Dev. Change 35, 435–461. doi:http:// dx.doi.org/10.1111/j. 1467-7660.2004.00360. x. Pretty, J., Ward, H., 2001. Social capital and the environment. World Dev. 29, 209–227. Putnam, R.D., 2001. Bowling Alone, 1st ed. Simon & Schuster. Putnam, R.D., Leonardi, R., Nanetti, R., 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton University Press, Princeton, N. J. Rango, T., Kravchenko, J., Atlaw, B., McCornick, P.G., Jeuland, M., Merola, B., Vengosh, A., 2012. Groundwater quality and its health impact: an assessment of dental fluorosis in rural inhabitants of the Main Ethiopian Rift. Environ. Int. 43, 37–47. doi:http://dx.doi.org/10.1016/j.envint.2012.03.002. Traditional trust measurement and the risk confound: an experiment in rural Paraguay. J. Econ. Behav. Organ. 62, 272–292. doi:http://dx.doi.org/10.1016/j. jebo.2005.03.006. Smit, B., Wandel, J., 2006. Adaptation, adaptive capacity and vulnerability. Glob. Environ. Change 16, 282–292. doi:http://dx.doi.org/10.1016/j. gloenvcha.2006.03.008. Smith, T., Marsden, P., Hout, M., Kim, J., 2011. General Social Surveys, 1972–2010: Cumulative Codebook (National Data Program for the Social Sciences Series No. 21). National Opinion Research Center, Chicago. Spielman, D.J., Kelemwork, D., Alemu, D., 2014. Seed, fertilizer, and agricultural extension in Ethiopia. In: Dorosh, P., Rashid, S. (Eds.), Food and Agriculture in Ethiopia: Progress and Policy Challenges. IFPRI/University of Pennsylvania Press, Philadelphia. Szreter, S., Woolcock, M., 2004. Health by association? Social capital, social theory, and the political economy of public healthth by association? Social capital, social theory, and the political economy of public health. Int. J. Epidemiol. 33, 650–667. Tanaka, T., Camerer, C., Nguyen, Q., 2010. Risk and time preferences: linking experimental and household survey data from Vietnam. Am. Econ. Rev. 100, 557–571. doi:http://dx.doi.org/10.1257/aer.100.1.557. Tompkins, E.L., Eakin, H., 2012. Managing private and public adaptation to climate change. Glob. Environ. Change 22, 3–11. doi:http://dx.doi.org/10.1016/j. gloenvcha.2011.09.010. Vaccaro, I., Smith, E.A., Aswani, S., 2010. Environmental Social Sciences: Methods and Research Design. Cambridge University Press. Vogel, B., Henstra, D., 2015. Studying local climate adaptation: A heuristic research framework for comparative policy analysis. Glob. Environ. Change 31, 110–120. doi:http://dx.doi.org/10.1016/j.gloenvcha.2015.01.001. Wolf, J., Adger, W.N., Lorenzoni, I., Abrahamson, V., Raine, R., 2010. Social capital, individual responses to heat waves and climate change adaptation: an empirical study of two UK cities. Glob. Environ. Change 20, 44–52. doi:http://dx.doi.org/ 10.1016/j.gloenvcha.2009.09.004. Woolcock, M., 2001. The place of social capital in understanding social and economic outcomes. Can. J. Policy Res. 2, 11–17. Woolcock, M., Narayan, D., 2000. Social capital: implications for development theory, research, and policy. World Bank Res. Obs. 15, 225–249. doi:http://dx. doi.org/10.1093/wbro/15.2.225. World Bank, 2014a. Ethiopia Overview [WWW Document]. URL http://www. worldbank.org/en/country/ethiopia/overview. World Bank, 2014b. World Bank Databank [WWW Document]. URL. http:// databank.worldbank.org/data/home.aspx. WVS, 2014. World Values Survey: Wave 5 2005-2008 OFFICIAL AGGREGATE v.20140429. (www.worldvaluessurvey.org). Aggregate File Producer: Asep/JDS. World Values Survey Association, Madrid, Spain.

Tutor Answer

DrAtticus
School: University of Virginia

...

flag Report DMCA
Review

Anonymous
Outstanding Job!!!!

Similar Questions
Hot Questions
Related Tags

Brown University





1271 Tutors

California Institute of Technology




2131 Tutors

Carnegie Mellon University




982 Tutors

Columbia University





1256 Tutors

Dartmouth University





2113 Tutors

Emory University





2279 Tutors

Harvard University





599 Tutors

Massachusetts Institute of Technology



2319 Tutors

New York University





1645 Tutors

Notre Dam University





1911 Tutors

Oklahoma University





2122 Tutors

Pennsylvania State University





932 Tutors

Princeton University





1211 Tutors

Stanford University





983 Tutors

University of California





1282 Tutors

Oxford University





123 Tutors

Yale University





2325 Tutors