Accepting market failure: Cultural worldviews and the opposition to corrective environmental policies

Journal of Environmental Economics and Management 85 (2017) 193–204

Contents lists available at ScienceDirect

Journal of Environmental Economics and Management journal homepage: www.elsevier.com/locate/jeem

Accepting market failure: Cultural worldviews and the opposition to corrective environmental policies Todd L. Cherry a,b,n, Steffen Kallbekken b, Stephan Kroll c a b c

Department of Economics, Appalachian State University, Boone, NC 28608-2051, USA CICERO Center for International Climate Research - Oslo, P.O. Box 1129 Blindern, 0318 Oslo, Norway Department of Agricultural and Resource Economics, Colorado State University, Fort Collins, CO 80523-1172, USA

a r t i c l e i n f o

abstract

Article history: Received 18 March 2016 Received in revised form 11 May 2017 Accepted 12 May 2017 Available online 24 May 2017

To explore whether and why people sometimes reject environmental policies that improve individual and collective outcomes, we create an experimental market in which transactions generate a negative externality. Market participants endogenously determine whether to implement corrective policies. We consider three policy instruments (Pigouvian taxes and subsidies, and quantity regulation) and two levels of policy efficiency (full and half). We then explore how individual cultural worldviews might contribute to the rejection of policies that correct the market failure. Our results indicate that people often oppose policies that improve their material outcomes, and we find that such opposition is significantly explained by cultural worldviews. Interesting connections emerge between individual worldviews and specific policy instruments. & 2017 Elsevier Inc. All rights reserved.

JEL code: C9 D62 H23 Q58 Keywords: Externality Pigouvian tax Policy aversion Worldviews Experiments

Introduction Many environmental problems persist not because of a lack of promising policies, but because of an inability to implement the policies. The issue is illustrated by the many failed attempts to address negative externalities with policy instruments that have broad support among experts—e.g., the rejection of a cap-and-trade system in the U.S. in 2009, the 2014 repeal of a controversial carbon tax in Australia, and the rejection of a revenue-neutral carbon tax in Washington state in 2016. An established literature on rent seeking and special-interest groups uses a rational-actor framework to explain the failure to implement first-best policies, but research from the behavioral sciences increasingly highlights the role that cognitive limits and biases play in undermining support for corrective policies. Social psychologists have long argued that dissonance arises when people are confronted with information that conflicts with their beliefs. Studies show that people tend to resolve the dissonance by rejecting information that conflicts with their beliefs and by selectively exposing themselves to information that reinforces their beliefs (Stroud, 2010; Hart et al., 2009).1 n

Corresponding author at: Department of Economics, Appalachian State University, Boone, NC 28608-2051, USA. E-mail address: 1 See Hart et al. (2009) for a review of the literature on selective exposure. Slater (2007) suggests that reinforcing spirals exist because the attitudinal

http://dx.doi.org/10.1016/j.jeem.2017.05.004 0095-0696/& 2017 Elsevier Inc. All rights reserved.

194

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

Research from political psychology shows that beliefs are often swayed by a need for cognitive closure—i.e., a desire to quickly formulate and firmly hold a clear opinion rather than being open to uncertainty and change (e.g., Golec de Zavala et al., 2010; Jost et al., 2003). Some argue that such behavioral tendencies and their influence on political decision-making have been amplified recently by the technology-led growth in differentiated and individually tailored information sources that facilitates the formation of fragmented “echo chambers” in which people only encounter self-reinforcing ideas (Brulle et al., 2012; Sunstein, 2001). The risk is that a marketplace of ideas becomes too specialized with too little trade.2 A potential consequence is that individual preferences over environmental policy prescriptions become defined more by the underlying worldviews that dictate the behavioural tendencies and less by the technical merits of the policy to achieve desired outcomes. In short, policy debates may be governed more by where people argue from rather than what people argue for. The notion that a person's worldview, the socially constructed orientation that dictates how one interprets and interacts with reality, can influence individual perceptions and actions around social and environmental risks originates from the cultural theory of risk pioneered by Douglas and Wildasky (1982). They argue that, in addition to economic interests and cognitive influences, social orientation has sway over people's perceptions and preferences of risk.3 Recent increases in political polarization and scientific denialism, which have impeded implementation of corrective policies, have revived researchers’ interest in cultural theory. Survey research provides empirical evidence that environmental risk perceptions are indeed skewed along cultural lines in ways theorized by Douglas and Wildavsky (e.g., Kahan et al., 2011). Moreover, given that cultural commitments are prior to factual beliefs, cultural worldviews shape how people access, process and assess information about environmental risk and policy (Kahan et al., 2007, 2010, 2011). Cultural worldviews effectively operate as a kind of heuristic that governs people's beliefs about the need and efficacy of public policies. A prominent illustration is the finding that worldviews, more than knowledge or education, explained people's willingness to accept or reject expert opinions on climate change and climate policy (Kahan et al., 2011; Smith and Leiserowitz, 2013). That worldviews can marginally sway opinions about social problems and corrective policies is not surprising, but it raises important questions about theory and policy if the influence of worldviews can move people enough to oppose polices that materially improve individual and collective outcomes. Observations from actual referenda suggest that people can indeed oppose policies that appear to serve their material self-interest, but the many confounding issues prohibit any definite claims.4 To gain clarity, researchers have turned to experimental markets to isolate the determinants of individual opposition to corrective policies. Using induced preferences and real payoffs in a market setting, studies show that market participants exhibit significant levels of opposition to policies that improve individual and collective material outcomes (e.g., Cherry et al., 2014; Kallbekken et al., 2011). More importantly, these efforts provide insights on the factors that underlie the opposition that obstructs the implementation of promising policies. Confusion does not appear to be the culprit because while learning more about the policy significantly improves understanding, it has no effect on acceptance (Kallbekken et al., 2011). Status-quo bias also does not seem to explain the lack of support; for example, Cherry et al. (2014) find that, even when policies are exogenously imposed as a trial run prior to a referendum on the policies, significant opposition to the corrective policies persists. Studies find that earmarking revenues from an environmental tax to environmental programs can significantly decrease opposition—potentially because it offers additional environmental action beyond the policy (Kallbekken et al., 2011).5 Recent work suggests that the timing of the policy's benefits also matters—when negative externalities are delayed, people are less receptive to current taxation (Tiezzi and Xiao, 2016). When benefits are immediate, opposition can be significantly diminished if people experience a trial run of the policy (Cherry et al., 2014). Framing, such as using the term fee instead of tax, can also have significant influence on people resisting policy (Blaufus and Möhlmann, 2014; Kallbekken et al., 2011).6 Herein we report results from, to our knowledge, the first laboratory market study on how cultural worldviews may explain individual opposition to welfare-enhancing policies in a controlled market setting. Specifically, we are asking three research questions: 1) Do market participants vote to support welfare-enhancing policies or do they exhibit policy aversion (footnote continued) outcomes from selected information in turn influence the selection of information. 2 Mutz and Martin (2001) argue “As the number of potential news sources multiplies, consumers must choose among them, and that exercise of choice may lead to less diversity of political exposure” (p. 111), and Stroud (2010) reports that partisan selective exposure leads to increased polarization. 3 This is related to the literature on the narrower concepts of social identity and political ideology (e.g., Unsworth and Fielding, 2014; Bliuc et al., 2015), as well as work on behavioral biases and heuristics (e.g., Aklin and Urpelainen, 2013; Hardisty et al., 2010). 4 In 2010, over 1.5 million (64%) Washington voters rejected a proposal (Initiative 1098) that would introduce an income tax to the approximately 38,400 returns that report income above $200,000 and direct the resulting $2 billion in annual revenue to education and health services (OFM, 2010). However, voters may believe the proposed tax will hurt them indirectly by negatively affecting the economy or cost them in the long-term as they expect to become richer over time. They may also suspect the initial policy will be an opening for more expansive tax policies in the future, or people simply may be confused or misinformed. 5 In reply to the question, why an environmental group would oppose a carbon tax (in Washington state in 2016), Jill Mangaliman, the Executve Director of Got Green, a community-based environmental group in Seattle, stated that “[w]ithout any kind of targeted revenue, business can continue as usual, and that's not what we want.” (http://www.pbs.org/newshour/bb/paying-carbon-pollution-environmentalists-dont-support-state-tax/) 6 Similar findings have been reported using hypothetical survey methods—providing more information about the policy effectiveness does not translate to greater support (Rhodes et al., 2014) and tax labels can affect support for the policy (McCaffery and Baron, 2006; Hardisty et al., 2010). In a conjoint experiment survey, Gampfer et al. (2014) find that institutional design features, not the underlying damage and emissions levels, dictate support for climate funding programs.

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

195

—opposition to policies that materially improve their individual outcomes and market efficiency?7 2) If policy aversion is observed, does it vary systematically across different policy instruments? And the primary research question 3) (how) do individual cultural worldviews impact the levels of policy aversion? Our results indicate that people often oppose policies that unambiguously improve their material outcomes, and we find that such opposition is significantly explained by cultural worldviews. Results reveal that the influence of worldviews varies in predictable ways across worldview types and policy instruments. Findings also suggest that opposition is greater with more effective policies, indicating that policy-makers may face a trade-off between policy effectiveness and policy acceptance.

Experimental design The market Following Cherry et al. (2012), we construct an experimental market with a negative externality—akin to harmful emissions from market activity.8 The resulting market failure can be corrected with an environmental tax/subsidy (i.e., Pigouvian tax/subsidy) or a quantity restriction (e.g., environmental standards). The market consists of five buyers who make a choice on how many of up to eight units of a fictitious good to buy at a posted-offer price from an automated seller. All trading occurs in “tokens,” with 100 tokens equal to one USD. The buyers impose external costs on each other through their purchases, similar to a polluter negatively affecting others. They are informed about the market price of 38 tokens, which remains constant throughout the session, and their identical array of resale values—85, 70, 60, 55, 45, 40, 30, and 15 tokens. Fig. 1 illustrates the market demand curve, which offers a summary of the net payoffs (price minus resale value) for each of the eight units. Whenever a buyer purchases a unit of the good, it imposes an externality or external cost of 3 tokens on each of the four other buyers in their group (a buyer's own damage is internalized in the resale values). The marginal damage from each unit purchased is therefore 12 tokens, and the Social Demand Curve is 12 tokens below the Market Demand Curve in Fig. 1. In the unregulated market equilibrium all buyers purchase six units at a price of 38. In this equilibrium, each individual buyer realizes a profit of 55 tokens—127 from own purchases/resales less the 72 in external costs imposed by others’ purchases.9 This translates to an aggregate market equilibrium of 30 units being purchased at the price of 38 tokens, which provides an overall profit of 275 tokens. However, given the negative externality, this market equilibrium is not socially optimal because buyers do not account for the external costs and therefore engage in too much market activity from society's perspective. The socially optimal outcome is for buyers to purchase a total of 20 units (4 each), which gives an overall payoff of 350 tokens—an improvement of 75 tokens over the market equilibrium.10 This represents a 27 percent increase in overall payoffs. The shaded area in Fig. 1 illustrates the improvement in payoffs in the socially optimal outcome relative to the suboptimal market equilibrium. The policies We consider six policy options that vary across instrument type (tax, subsidy or quantity regulation) and efficiency level (full and half). Table 1 summaries the six combinations. For a full tax/subsidy, the rate is equal to the marginal external cost of 12, which is the theoretically optimal environmental tax/subsidy that fully corrects the market failure. For a half tax/ subsidy, the rate is equal to half the marginal external cost of 6, which only partially corrects the market failure. In the case of a full quantity restriction and a half quantity restriction subjects cannot buy more than four and five units, which represent a limit that fully and partially corrects the market failure, respectively. Note that the quantity limits for full and half quantity regulation correspond to the number of units that result (in equilibrium) from the full and half tax/subsidy. With a tax t, market participants pay for each unit they buy and the tax revenues are returned in lump-sum fashion to the entire group. With a subsidy s, the participants receive a subsidy for each of the eight units they do not buy (including the ones they would not buy anyway since their resale values are below the price), and the subsidy cost is then paid lumpsum by the entire group. Note that fundamentally there is no difference in this set-up between a tax and a subsidy that is set at the same rate: given a certain amount of purchases, participants realize the same payoffs independent of whether there is a tax or an equivalent subsidy.11 The quantity regulation imposes a perfectly enforced upper limit q on purchases, which 7 This definition of policy aversion is stronger than the definition of tax aversion some economists and psychologists provide; e.g., “[T]ax aversion bias [is] the phenomenon that people may perceive an additional burden associated with tax payments compared to economically equivalent payments labeled differently” (Blaufus and Möhlmann, 2014). 8 Experimental instructions are available upon request. 9 The 127 unit net revenue equals the resale price of each unit purchased [85 þ 70 þ 60 þ 55 þ 45 þ 40] less the common purchase price of each unit purchased (6*38). The 72 unit external cost equals the total units purchased by other buyers (24 ¼ four buyers purchasing six units each) multiplied by the per unit external cost (3). 10 Purchasing ten fewer units reduces aggregate external costs by 120 (12 tokens per purchase), but that gain is partly counteracted by a loss of 45 tokens in net return from purchases/resales (5*(7 þ 2)). 11 For example, imagine there is a full tax/subsidy and all participants buy five units except for participant A who buys four units. A's payoff would,

196

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

Fig. 1. Supply, Demand and Efficiency Gain.

Table 1 Experimental design: Policy treatments. Efficiency Instrument

Optimal (full correction)

Suboptimal (half correction)

Tax Subsidy Quantity Regulation

Full Tax (t ¼ 12) Full Subsidy (s ¼ 12) Full Qty Regulation (q ¼ 4)

Half Tax (t ¼ 6) Half Subsidy (s ¼ 6) Half Qty Regulation (q ¼ 5)

allows participants to deviate from the equilibrium in only one direction (fewer purchases). This differs from the tax/subsidy instruments that allow more or less purchases. Full tax, full subsidy and full quantity restriction are theoretically optimal (i.e., efficient) in equilibrium because the former two make the participants internalize the external costs of their purchases, while the latter does not allow the purchase of any units that have more costs than benefits for the group on aggregate. For each of the three schemes, the policy-induced equilibrium reduces aggregate quantity from 30 to the socially optimal 20 units, with individual buyers reducing the number of units purchased from 6 to 4 units. Half tax, half subsidy and half quantity restriction improve individual and collective benefits relative to the no-policy baseline, but the half measures are not socially optimal. In the equilibrium with half measures, buyers reduce their purchases by only one unit each, for a total reduction from 30 to 25. With a full tax/subsidy/quantity restriction, individual returns increase to 70 (from 55 without any policy); with a half tax/subsidy/quantity restriction, they increase only to 65. Experimental framework Table 2 summarizes the experimental framework. The experiment consisted of eight sessions, each with a series of 32 market periods. Each market period began with buyers, who are informed of resale values, the market price and possible external costs, indicating the number of units they wish to purchase. At the close of each market periods, buyers were informed of their total profit from the market period, including details about the net gain from each unit purchased, the external cost incurred by others’ purchases, and the payment or receipt of any tax or subsidy. Participants were not allowed to communicate with each other during the experiment. The 32 market periods occurred in four stages. The first stage had five market periods without any referenda or corrective policies. This stage serves as a baseline from which policies can be introduced and considered, paralleling actual (footnote continued) compared to her payoff in the equilibrium without any policy, go up by 12 (less external cost) – 9 (loss in consumer surplus) plus any change due to the policy: with a tax, A would pay 48 (4*12) but receive one fifth of the total tax revenues of 244 (her 48 plus 60 each from the other four). Her total payoff change is þ 12.6; the other participants all gained 10.6 each. With a subsidy scheme, she would receive 48 in subsidies (4*12) but had to pay one fifth of the total subsidy costs of 192 (her 48 plus 36 for each of the other four). Her total payoff change is again þ12.6, and that of the other participants is again 10.6.

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

197

Table 2 Timeline for the experiment. 2 ABA Tax Sessions (plus 2 BAB Tax Sessions not shown) Stage

1

Periods Vote# Instrument Efficiency

1–5 None None None

2 (A) 6–8 Vote 1 Tax F v. N

9–11 Vote 2 Tax H v. N

3 (B) 12–14 Vote 3 Tax F v. H

15–17 Vote 4 Qty F v. N

18–20 Vote 5 Qty H v. N

4 (A) 21–23 Vote 6 Qty F v. H

24–26 Vote 7 Tax F v. N

27–29 Vote 8 Tax H v. N

30–32 Vote 9 Tax F v. H

2 ABA Subsidy Sessions (plus 2 BAB Subsidy Sessions not shown) Stage

1

Periods Vote# Instrument Efficiency

1–5 None None None

2 (A) 6–8 Vote 1 Subsidy F v. N

9–11 Vote 2 Subsidy H v. N

3 (B) 12–14 Vote 3 Subsidy F v. H

15–17 Vote 4 Qty F v. N

18–20 Vote 5 Qty H v. N

4 (A) 21–23 Vote 6 Qty F v. H

24–26 Vote 7 Subsidy F v. N

27–29 Vote 8 Subsidy H v. N

30–32 Vote 9 Subsidy F v. H

Notes: Votes preceded the set of three market periods with the outcome of the referendum determining the policy for those three subsequent periods. “Tax”, “Subsidy” and “Qty” indicates whether the referendum proposes a tax, subsidy or quantity. “F” and “H” indicates the proposed instrument is either a full or half measure, and “N” indicates a no policy option.

referenda. The three subsequent stages had nine market periods each, and policies were endogenously determined by voting in a referendum prior to the stage's first, fourth and seventh market period, which enabled us to address research question 1. Each referendum presented a pairwise policy choice, and the five market participants determined (via a simple majority vote) which policy, if any, would be implemented in the subsequent three market periods. The referenda required an active choice between two options (no default), which is a design feature shown to mitigate default/status quo effects (Carroll et al., 2009; Kempf and Ruenzi, 2006). Participants were informed of the vote tally and the referendum outcome that would be implemented for the upcoming three market periods. To examine the second research question, we varied the pairwise choice in the referenda around two treatment variables: the instrument (tax, subsidy and regulation) and the efficiency (full measure, half measure and no policy). Treatments were applied within and between sessions with instruments varying across stages and efficiency varying within stages. Thus, referenda did not present choices between instruments, just between the efficiency levels of an instrument (full, half and no policy). To mitigate order effects, the design used an ABA/BAB design. Stage A presented either a tax or subsidy instrument, depending on whether it was a tax or subsidy session. Stage B always presented a quantity instrument. Within both A and B stages, the three referenda varied efficiency by presenting the following sequence of choices: full measure vs. no policy, half measure vs. no policy, and full measure vs. half measure.12

Worldview measures After the market sessions were completed, participants answered a questionnaire used to elicit their cultural worldviews. This allowed us to address research question three. We followed the literature and employ the (short-form) cultural worldview questions from Kahan et al. (2011). The survey presents a set of statements that can be used to measure an individual's worldview across two dimensions—hierarchy-egalitarianism and individualism-communitarianism. Six hierarchy-egalitarianism statements capture the “attitudes toward social orderings that connect authority to stratified social roles based on highly conspicuous and largely fixed characteristics such as gender, race, and class” (p. 51, Kahan et al., 2011). Six individualism-communitarianism statements (individualism-communitarianism) focus on the “attitudes toward social orderings that expect individuals to secure their own well-being without assistance or interference from society versus those that assign society the obligation to secure collective welfare and the power to override competing individual interests” (p. 51, Kahan et al., 2011). Using a six-level Likert scale, subjects indicate the extent to which they agree with each of the hierarchy and individualism statements, which is translated to a score of 1 to 6 that measures the level of agreement (1 ¼ strongly disagree to 6 ¼ strongly agree). The sum of scores for the six hierarchy statements places the individual's views along the hierarchyegalitarianism spectrum (6 to 36), while the sum of scores for the individualism statements focuses on the individualismcommunitarianism dimensions (6 to 36). The two worldview measures, hierarchy and individualism, will provide the basis for examining the influence of worldviews on individual preferences for alternative welfare-enhancing policies. 12 We note that, despite the fixed order of proposed efficiency levels within each stage, potential order effects are mitigated because actual efficiency levels experienced by participants vary considerably because of mixed outcomes of previous referenda. Data show that participants had a wide range of experiences with efficiency levels at each point within and across stages, and this experience is controlled for in the upcoming conditional analysis.

198

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

Table 3 Mean market efficiency and individual payoffs by policy instrument. No

Market Efficiency (in percent) Individualpayoffs (in tokens)

Full measure

Half measure

policy

Tax

Subsidy

Quantity

Tax

82.0[180] 57.4(10.14) [900]

94.7[78] 66.3(7.81) [390]

94.8[111] 66.3(8.68) [555]

98.7[162] 91.0[84] 69.1(8.37)[810] 63.7(8.04) [420]

Subsidy

Quantity

88.3[63] 61.8(10.37) [315]

93.4[186] 65.4(9.55)[930]

Note: number of observations reported in brackets and standard errors reported in parentheses.

Procedures The experiment was conducted at Colorado State University with a total of 160 student participants. We ran eight sessions, each with 20 participants and four independent markets. Sessions were split between ABA and BAB formats with A stages being split between tax and subsidy instruments while B stages always presented quantity instruments. This resulted in 160 subjects participating in 1024 market periods and casting a total of 1440 votes. People with experience in similar experiments were not invited to participate. To ensure understanding, participants were required to answer questions (with subsequent explanations of correct answers) prior to the start of the market periods. Each session lasted about 90 min, including reading the instructions and the questionnaire at the end of the session. Participants earned an average of about 28 USD, which corresponds to the literature. The experiment was programmed and conducted with the software z-Tree (Fischbacher, 2007). Hypotheses The experimental design generates primary data to examine the three main research questions on policy aversion. We first determine whether policy aversion appears in our setting. Hypothesis 1. states that market participants vote to support corrective policies over no policy. The null hypothesis follows standard economic theory, which predicts support for any welfare-enhancing policy. Rejecting the null indicates the presence of policy aversion, which we expect based on previous studies. Second, we explore whether any policy aversion varies across instrument types by testing for instrument-specific policy aversion. Hypothesis 2. states that, for a given efficiency level (full or half), the level of policy aversion is equal across instrument types (tax, subsidy and quantity). Rejecting the null indicates that policy aversion is systematically influenced by the policy instrument. By identifying instrument-specific policy aversion, findings may contribute to understanding the underlying barriers to implementing corrective policies. Though standard economic theory is consistent with the null, previous reports suggest that people may exhibit instrument-specific opposition based on perceptions of coercion and cost (e.g., Heres et al., 2015; Baron and Jurney, 1993). As an extension, we consider whether any policy aversion is affected by the stringency of the instrument (i.e., full versus half efficiency), with an interest in a possible efficiency-acceptability tradeoff. Third, we turn to the primary research question that addresses how individual cultural worldviews explain policy aversion. Standard theory is silent on worldviews, but the literature provides a basis for expecting that worldviews matter. We therefore test for the relationships between worldviews and policy aversion. Hypothesis 3. contends that the extent of any policy aversion is equal across individual worldviews. Rejecting the null would indicate that worldviews systematically explain policy aversion. However, the experimental design enables a more nuanced investigation of how worldviews affect policy aversion. We draw from the literature for our a priori expectations. Kahan et al. (2011) find that “egalitarians and solidarists [communitarians] perceived business regulation to be conducive to economic prosperity, hierarchists and individualists destructive of it.” From this, an alternative hypothesis is that hierarchical and individualistic people will exhibit greater policy aversion than egalitarians and communitarians. For possible instrument-specific policy aversion, we consider that egalitarians and communitarians tend to surrender individual autonomy when forming beliefs, while hierarchical and individualistic people often form beliefs around individual responsibility and free choice. Thus, a second alternative hypothesis is that hierarchical and individualistic people will be relatively more averse to the quantity regulation instrument because it explicitly impedes choices. Similarly, egalitarians and communitarians will be relatively less offended by, and therefore more supportive of, redistributive instruments (taxes and subsidies).

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

199

Table 4 Support and implementation of welfare-enhancing polices.

All Referenda Stage First Stage Second Stage Third Stage Instrument Tax Subsidy Quantity Efficiency Full vs. No Half vs. No Full vs. Half

n

% Votes

% Passed

1440

61.1

68.4

480 480 480

55.2 61.0 67.1

60.4 67.7 77.1

360 360 720

57.5 67.8 59.6

61.1 76.4 68.1

240 240 240

55.0 62.1 70.1

58.3 75.0 72.9

Note: % Votes is the proportion of votes in favor of the efficiency-enhancing policy; % Passed is the proportion of referenda that approved the efficiencyenhancing policy.

Results To confirm that markets and policies generate outcomes as expected, we first review how the policy instruments affected market efficiency and individual payoffs. Table 3 reports the observed market outcomes by policy condition. As expected, the numbers show two things: first, the market underperforms without any intervention, and second, market outcomes improve once policies are implemented. Market efficiency, measured as the actual total group payoffs relative to theoretical efficient group payoffs (i.e., the payoffs if everybody buys the optimal four units), improves from 82 percent when no policy is implemented to about 91 and 96 percent with half- and full-policy measures in place. This corresponds to the predicted market efficiency of 78.6, 92.8 and 100.0 percent with no policy, half measures and full measures. Individual payoffs improve in a similar fashion. This verifies the market performs as expected and that the market participants do indeed benefit materially from the implementation of a policy. Now we turn to the aggregate voting behavior to see if people vote in their material self-interest. Aggregate numbers Policy aversion Table 4 provides summary results of the policy referenda by instrument and efficiency. The first column of results (% Votes) reports the overall percentage of market participants that voted to implement the welfare-enhancing policy. Aggregating across all referenda, 56 percent of participants voted to implement the more efficient policy, leaving 44 percent contradicting expectations and exhibiting policy aversion. Considering voting by stage, we reveal that the level of policy aversion declines over time. This is consistent with previous reports that experience with policies can increase support when the policy provides immediate benefits (Cherry et al., 2014).13 The question that follows is how the observed policy aversion affected the passage of a welfare-enhancing policy. Table 4 reports the percentage of referenda in which a majority of voters supported and implemented the policy. The numbers show that individual policy aversion was often sufficient to block the passage of welfare-enhancing policies, with the policy failing to pass in over 68.4 percent of cases. As expected, as policy aversion diminished over time, the rate of passage increased— from 60.4 percent in stage one to 77.1 percent in stage three. Instrument-specific policy aversion Given the presence of policy aversion, we consider how this varies across policy instruments. Table 4 stratifies the numbers by the three policy instruments. The summary numbers are consistent with previous reports of tax aversion and provide evidence that such behavior may extend to other policies (e.g., Cherry et al., 2014; Kallbekken et al., 2011). Even so, the numbers indicate that participants exhibited different preferences across the equivalent instruments. The subsidy instrument received more support (67.8%) than the tax and quantity regulation instruments (57.5% and 59.6%). This finding is consistent with arguments that people dislike policies that appear more coercive (Baron and Jurney, 1993; Schuitema et al., 2011). The corresponding opposition was sufficient to block implementation of the corrective policy in 23.6 to 39.9 percent of cases. The bottom half of Table 4 provides a glimpse at how efficiency (full or half measures) may affect policy aversion. In the two referenda with a no-policy baseline, market participants exhibited somewhat more support for the half measure than 13 Cherry et al. (2014) point out that the positive effect from experience on support likely depends on the immediacy and salience of the policy effects. While people may drop their opposition to a congestion charge after experiencing less traffic, people may continue opposing a new climate change policy because the benefits are neither immediate nor clearly apparent.

200

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

Fig. 2. Scatterplot of Hierarchy and Individualism Worldview Scores. Note: plots are slightly jittered using STATA 13 to better illustrate the distribution of data.

Table 5 Support for efficiency-enhancing policy by referendum and worldview (% vote). Hierarchy

All Referenda Stage First Stage Second Stage Third Stage Instrument Tax Subsidy Quantity Efficiency Full vs. No Half vs. No Full vs. Half

Individualism

Hierarchical

Egalitarian

Individualistic

Communitarian

53.9

66.9

56.0

65.5

47.8 58.7 55.1

54.0 67.5 79.4

48.2 60.3 59.6

60.2 66.7 69.6

46.9 62.5 55.9

69.8 76.9 60.1

51.9 61.8 55.4

61.8 70.2 65.7

39.7 52.1 64.4

71.7 71.7 80.0

51.4 52.9 65.7

56.5 69.6 69.6

the full measure (62.1% vs. 55.0%). Though findings may be limited due to potential order effects, the numbers are consistent with an efficiency-acceptability tradeoff. However, considering that a majority of participants in the full versus half referenda prefers the full measure to the half measure, this finding appears dependent on the choice set or baseline. Worldviews and policy aversion We now turn to the primary research question: how do individual worldviews explain policy aversion? Fig. 2 provides a scatterplot of the individual scores for the two worldview dimensions described in the experimental design section— hierarchy-egalitarianism and individualism-communitarianism. Each worldview score ranges from 6 to 36, with a higher score indicating a more hierarchical (less egalitarian) worldview and a more individualistic (less communitarian) worldview. The hierarchy measure has a mean of 19.2 and standard deviation of 6.3, while the individualism measure has a mean and standard deviation is 23.3 and 5.1. To allow preliminary comparisons of aggregate voting behavior across worldview types, we assign subjects to worldview categories based on their scores. Participants that scored in the top quartile of the hierarchy and individualism measures are defined as hierarchists and individualists, respectively, while those that scored in the bottom quartile of each measure are defined as egalitarians and communitarians. For the hierarchy dimension, this yields 46 hierarchical types casting 414 votes and 42 egalitarian types casting 378 votes. And for the individualism dimension, we have 47 individualistic types casting 423 votes and 46 communitarian types casting 414 votes. Thirty-four people (and their 306 votes) fell in the interquartile range in either dimension, and therefore were not included in either set of categories. Table 5 reports the percent of voters that supported the (more) efficient policy by stage, instrument, efficiency and worldview. The summary numbers indicate that policy aversion is a behavioral phenomenon across all worldview categories, but the extent of policy aversion varies systematically across both hierarchical and individualist dimensions. The

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

201

Table 6 Breakdown of voting patterns for worldview types. Prefers Policy

Prefers No Policy

Mixed Preferences

Pooled

YYY

YYN

NNN

NNY

NYY

NYN

YNY

YNN

Hierarchical Egalitarian

22.5 39.0

7.9 13.5

17.4 8.7

13.0 7.9

16.7 11.1

10.9 8.7

9.4 7.1

2.2 4.0

100.0 100.0

Individualist Communitarian

26.2 35.5

7.8 10.9

14.2 11.6

14.2 2.9

12.8 17.4

10.6 11.6

9.2 8.7

5.0 1.4

100.0 100.0

Note: The letters Y and N refer to votes within a stage. For example, YYN indicates that a subject voted for the full policy over the no-policy option in the first referendum of a stage (Y), for the half policy over the no-policy option in the second referendum (Y), and for the half policy over the full policy in the third stage (N).

numbers show that hierarchical types consistently exhibit greater policy aversion than the egalitarian types, while individualists generally show more policy aversion than communitarians. A few items from the aggregate numbers are worth mentioning. First, the differences in worldviews persist across all three stages. Second, the instrument-specific policy aversion appears to vary considerably more for the hierarchical-egalitarian dimension than the individualist-communitarian dimension. And third, the efficiency-specific policy aversion is much starker across the hierarchical-egalitarian dimension than the individualist-communitarian dimension, which may indicate a efficiency-acceptability tradeoff may be worldviewspecific. Moving beyond single votes, we consider how voting patterns may differ across worldviews. Each stage has three referenda with pairwise choices, which translates to eight possible voting patterns. Table 6 reports the frequency of each voting pattern for each worldview. Each three-vote pattern is denoted by a sequence of three Y or N that indicate yes and no votes in the order of the three referenda. For example, YYN indicates a yes vote in the first and second referenda, and a no vote in the third referendum. For clarity, the patterns are organized around three revealed preference groups: prefers some policy over no policy, prefers no policy over some policy, and mixed preferences for policy. The numbers indicate that people with hierarchical worldviews consistently exhibit greater policy aversion than those with egalitarian views. Also, people with individualist worldviews display more policy aversion than those with communitarian views. From Table 6, the NNN pattern is much more frequent among hierarchical types than egalitarian types (17.4% vs. 8.7%), while the YYY voting pattern is much more frequent among egalitarian types than hierarchical types (39.0% vs. 22.5%). A similar, though less stark, finding emerges when comparing individualist and communitarian types.14 These findings match the inferences from the individual votes (Table 5), but the voting patterns offer additional evidence of the systematic nature of the influence of individual worldviews on voting behavior. Conditional analysis We follow the aggregate numbers with a conditional analysis by estimating the probability of voting in favor of a proposed policy instrument with the following linear probability model:

Yij = α + δWorldviewi + ψEfficiencyij + ϖExperienceij + θInstrumentij + γi + ϕj + μi + εij , where Yij is a binary variable that indicates whether the ith subject voted in favor of the proposed instrument in referendum j ( ¼1 if yes; ¼0 otherwise); Worldviewi contains the two continuous worldview measures for the hierarchy and individualism dimensions for subject i; Efficiencyij is a vector of two indicator variables that signify the stringency of the instrument proposed to subject i in referendum j (full omitted); Experienceij is the ith subject's cumulative instances of prior markets with a policy intervention in referendum j; Instrumentij is a vector of indicator variables that signifies which policy instrument is proposed to subject i in referendum j (quantity regulation omitted; tax and subsidy included); γi is a vector of control variables that capture individual attributes; ϕj is a set of J  1 dummies that capture potential non-linear period effects; ui are random effects that control for unobservable individual characteristics; α is the estimated intercept and εit is the well-behaved error term.15 Three aspects of the conditional estimates are worth noting. First, we estimate four models—a pooled model that examines voting behavior across all policy instruments, and three instrument models that examine voting behavior specific to each policy instrument. Second, due to the different baseline in the referenda, we estimate two sets of the four models—one that uses the two referenda with a no-policy option (i.e., excluding full vs half referenda) and another that uses only the full 14

The influence of worldviews was consistent across stages. While the frequency of YYY increased over time, this shift was similar across worldview

types. 15

Results are consistent across probit and logit specifications. To allow for estimates of the time-invariant worldview measures, we utilize a randomeffects specification to control for subject heterogeneity. Fixed period effects drop out in cases of collinearity with treatments. Estimates are robust to alternative measures of experience, such as instrument-specific experience and time-specific experience.

202

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

vs half referenda.16 And third, we employ clustered-robust standard errors to account for non-independence and withingroup correlation in the panel. Table 6 reports the estimates of voting models.17 We first consider the referenda with a no policy baseline. We begin by reporting the findings related to possible instrument-specific policy aversion. Results from the pooled model indicate that policy aversion exists across the three instruments at statistically equivalent levels—i.e., there is no evidence for instrumentspecific policy aversion. Thus, while the aggregate numbers showed marginal differences, the conditional estimates fail to reject the null of equivalent policy aversion across instruments. Moreover, the estimates indicate that voters are significantly more likely to support a half measure than a full measure when the alternative is no policy, but the instrument-specific models show this result is not common to all instruments. Therefore, conditional estimates indicate that instrument-specific policy aversion may interact with the efficiency of the instrument. We now turn to the primary research question and review the estimates for the two worldview measures. The results from the pooled model indicate that worldviews significantly explain policy aversion—rejecting the third null hypothesis. Hierarchy and individualism measures significantly affect the likelihood to support or oppose efficiency-enhancing policies. Consistent with the aggregate numbers, the estimated negative relationship implies that when a person's worldview is more hierarchical (less egalitarian), support for a corrective policy is less likely. Similarly, support is less likely as worldviews become more individualistic (less communitarian). The three instrument models provide estimates specific to each policy instrument, and the results reveal that the influence of worldviews varies across instruments. A clear distinction emerges when comparing the two price instruments (tax and subsidy) and the quantity instrument. While hierarchical/egalitarian views significantly explain the likelihood of individual support for tax and subsidy instruments, they have no significant effect on the support of the quantity instrument. Conversely, individualistic/communitarian views are significant determinants of individual support for the quantity instrument, but do not matter in the case of the two price instruments. Evidence therefore suggests the influence of worldviews depends on the interactions between an individual's cultural orientation and an instrument's attributes. For instance, more coercive quantity instruments may be particularly offensive to individualist types despite the gains from correcting the market failure, while price instruments may be more attractive to egalitarian types because of the redistributive nature of the instruments. A different story emerges from the full vs half referenda models in Table 7. Across this set of models, estimates show that individual worldviews have no significant effect on the likelihood of supporting (or opposing) the more efficient full measure. Therefore, while worldviews explain policy aversion when the referendum has a no-policy option, they do not matter when the referendum is a choice between full and half measures. This finding suggests that worldviews have a role in people having dominant and lexicographic preferences for alternative policies (e.g., Gelso and Peterson, 2005; Scott, 2002). We briefly mention a few other findings. In each voting model, estimates control for any past experience that each market participant has with corrective instruments. Specifically, experience is defined as the number of times the individual had previously engaged in a market with a corrective policy, and as such, depends on the cumulative outcomes of the referenda. Individual experience varied within and between referenda. The reported findings were not affected when experience was disaggregated by instrument type, efficiency level, or voting sequence. Estimates generally find that past experience with efficiency-enhancing instruments increased the likelihood of support for an instrument in the current referenda. The noteworthy exception is experience having no significant effect in the tax models. Concerning individual attributes, estimates from the full/half vs no referenda show the likelihood of support for corrective policies is lower among men, and somewhat unsettling, people that have taken at least one economics course.

Conclusions We confirm findings from previous studies that people oppose policies that correct market failure—i.e., improve individual outcomes and market efficiency. Many referenda and environmental policy debates outside the lab suggest that such opposition is real, and our results corroborate such policy aversion in controlled laboratory markets with policy referenda. This finding suggests there is an inherent barrier that impedes elected policy-makers to effectively address and solve market failures, and more specifically, environmental problems. Previous work finds that policy aversion is not explained by confusion about the policy (Kallbekken et al., 2011), and that elements of the policy design can mitigate the aversion (Cherry et al., 2014). We extend this research by exploring how policy aversion may be explained by individual cultural worldviews. Our results indicate that worldviews significantly explain policy aversion. In our laboratory referenda, people with different worldviews exhibit substantially different levels of policy aversion—by more than 25 percentage points in some cases. 16 Tests indicate that voting behavior differs across the two baselines, so while a fully specified pooled model (i.e., one including interactions) would yield the similar results, we report the separate estimates over the combined estimates for succinctness and clarity. 17 The results show that male participants are less supportive of corrective policies (significant for the pooled model, as well as for subsidy and quantity regulation). This is consistent with finding that men are more overconfident and more competitive than women (e.g. Niederle and Vesterlund, 2007), which may translate to free-riding behavior in social dilemmas, and the somewhat more mixed evidence that women are more cooperative than men (e.g., Frank et al., 1993).

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

203

Table 7 Estimates of voting models. Full/Half vs No

Hierarchy Individualism Half Measure Male Economics Age Experience Tax Subsidy

χ2 # of clusters N

Full vs Half

Pooled

Tax

Subsidy

Quantity

Pooled

 0.008** (0.003)  0.009** (0.004) 0.122*** (0.029)  0.100*** (0.037)  0.022* (0.012)  0.001 (0.004) 0.026*** (0.009)  0.025 (0.049)  0.009 (0.049)

 0.015*** (0.005)  0.005 (0.007) 0.169*** (0.050)  0.122 (0.089)  0.009 (0.019)  0.0002 (0.008)  0.005 (0.021) –

 0.014** (0.007)  0.003 (0.010)  0.046 (0.062)  0.125** (0.060)  0.032* (0.017)  0.003 (0.005) 0.032** (0.016) –

 0.002 (0.005)  0.012*** (0.004) 0.179*** (0.035)  0.084** (0.039)  0.026*** (0.010) 0.0004 (0.005) 0.041*** (0.008) –

0.002 (0.004)  0.004 (0.005) –

68.28 (0.000) 32 960

60.93 (0.000) 16 240

–

–

37.27 (0.000) 16 240

–

88.29 (0.000) 32 480

Tax

Subsidy

Quantity

 0.007 (0.008) 0.003 (0.008) –

 0.002 (0.008)  0.008 (0.014) –

0.008 (0.005)  0.005 (0.006) –

 0.045 (0.059)  0.007 (0.008) 0.002 (0.004) 0.043*** (0.010)  0.051 (0.064)  0.243 (0.080)

 0.038 (0.112)  0.014 (0.028)  0.006 (0.008) 0.041 (0.026) –

 0.089 (0.120) 0.004 (0.016)  0.012* (0.007) 0.002 (0.010) –

 0.026 (0.083)  0.004 (0.011) 0.009 (0.008) 0.058*** (0.012) –

33.80 (0.000) 32 480

7.07 (0.315) 16 120

–

–

5.05 (0.538) 16 120

–

31.01 (0.000) 32 240

Notes: dependent variable is the individual vote (1¼ vote yes; 0¼ vote no); full/half vs no estimation use referenda with no policy baseline; full vs half estimates only use referenda without a no policy option (i.e., full vs half referenda); robust standard errors are reported in parentheses.

Differences in opposition are systematic and appear to follow patterns consistent with the principles associated with the different worldviews. We find that the influence of worldviews is instrument-specific with individualist/communitarian worldviews affecting support for the coercive quantity instruments, and hierarchical/egalitarian affecting support for the redistributive instruments. Additional results suggest that proposing less ambitious policies (in a choice between a policy and no policy) may enhance acceptability, but this tradeoff might not be universal. That worldviews matter in people's judgement of environmental policies is not surprising, but our controlled investigation allows the revelation that they can matter a great deal—more than material interest. The implication is that improving the prospects for enacting promising environmental policy requires moving beyond arguing the material merits of specific policies. To improve the prospects for implementation of welfare-enhancing policies, experts must recognize that policy design and communication requires an understanding and consideration of individual cultural worldviews (see Kahan et al. 2012).

Acknowledgements We appreciate helpful comments received from David McEvoy, Michael McKee, Jim Murphy, JShogren, Håkon Sælen and Nicholas Janusch, and seminar participants at Duke University, Emory University, European University Institute, University of Potsdam, University of Kiel, and Virginia Tech. Two referees and the editor also provided helpful comments. All errors remain our own. Cherry acknowledges and appreciates his time working on this research as Rasmuson Chair at the University of Alaska Anchorage. Funding for this research was provided by the Research Council of Norway, NORKLIMA program, project 190768.

Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/ 10.1016/j.jeem. 2017.05.004.

References Aklin, M., Urpelainen, J., 2013. Debating clean energy: frames, counter frames and audiences. Glob. Environ. Change 23 (5), 1225–1232. Baron, J., Jurney, J., 1993. Norms against voting for coerced reform. J. Pers. Social. Psychol. 64 (3), 347–355.

204

T.L. Cherry et al. / Journal of Environmental Economics and Management 85 (2017) 193–204

Blaufus, K., Möhlmann, A., 2014. Security returns and tax aversion bias: behavioral responses to tax labels. J. Behav. Financ. 15 (1), 56–69. Bliuc, A., McGarty, C., Thomas, E.F., Lala, G., Berndsen, M., Misajon, R., 2015. Public division about climate change rooted in conflicting socio-political identities. Nat. Clim. Change 5, 226–229. Brulle, R.J., Carmichael, J., Jenkins, J.C., 2012. Shifting public opinion on climate change: an empirical assessment of factors influencing concern over climate change in the U.S., 2002–2010. Clim. Change 114 (2), 169–188. Carroll, G., Choi, J., Laibson, D., Madrian, B, B., Metrick, A., 2009. Optimal defaults and active decisions. Q. J. Econ. 124, 1639–1676. Cherry, T.L., Kallbekken, S., Kroll, S., 2014. The impact of trial runs on the acceptability of Pigouvian taxes: experimental evidence. Resour. Energy Econ. 38, 84–95. Cherry, T.L., Kallbekken, S., Kroll, S., 2012. The acceptability of efficiency-enhancing environmental taxes, subsidies and regulation: an experimental investigation. Environ. Sci. Policy 16 (1), 90–96. Douglas, M., Wildasky, A., 1982. Risk and Culture: An Essay on the Selection of Technical and Environmental Dangers. University of California Press, Berkeley. Fischbacher, U., 2007. Z-Tree: zurich toolbox for ready-made economic experiments. Exp. Econ. 10, 171–178. Frank, R.H., Gilovich, T., Regan, D.T., 1993. Does studying economics inhibit cooperation? J. Econ. Perspect. 7 (2), 159–171. Gampfer, R., Bernauer, T., Kachi, A., 2014. Obtaining public support for North-South climate funding: evidence from conjoint experiments in donor countries. Glob. Clim. Change 29, 118–126. Gelso, B.R., Peterson, J.M., 2005. The influence of ethical attitudes on the demand for environmental recreation: incorporating lexicographic preferences. Ecol. Econ. 53, 35–45. Golec de Zavala, A., Cislak, A., Wesolowska, E., 2010. Political conservatism, need for cognitive closure, and intergroup hostility. Polit. Psychol. 31, 521–541. Hardisty, D.J., Johnson, E.J., Weber, E.U., 2010. A dirty word or a dirty world? Attribute framing, political affiliation, and query theory. Psychol. Sci. 21, 86–92. Hart, W., Albarracin, D., Eagly, A.H., Brechan, I., Lindberg, M., Lee, K., Merrill, L., 2009. Feeling validated versus being correct: a meta-analysis of selective exposure to information. Psychol. Bull. 135, 555–588. Heres, D., Kallbekken, S., Galarraga, I., 2017. The role of budgetary information in the preference for externality-correcting subsidies over taxes: a lab experiment on public support. Environ. Resour. Econ. 66 (1), 1–15. Jost, J.T., Glaser, J., Kruglanski, A.W., Sulloway, F., 2003. Political conservatism as motivated social cognition. Psychol. Bull. 129, 339–375. Kahan, D.M., Braman, D., Gastil, J., Slovic, P., Mertz, C.K., 2007. Culture and identity-protective cognition: explaining the white-male effect in risk perception. J. Empir. Leg. Stud. 4 (3), 465–505. Kahan, D.M., Braman, D., Cohen, G.L., Gastil, J., Slovic, P., 2010. Who fears the HPV vaccine, who doesn’t, and why? An experimental study of the mechanisms of cultural cognition. Law Hum. Behav. 34 (6), 501–516. Kahan, D.M., Jenkins-Smith, H., Braman, D., 2011. Cultural cognition of scientific consensus. J. Risk Res. 14 (2), 147–174. Kahan, D.M., Peters, E., Wittlin, M., Slovic, P., Ouellette, L.L., Braman, D., Mandel, G., 2012. The polarizing impact of science literacy and numeracy on perceived climate change risks. Nat. Clim. Change 2, 732–735. Kallbekken, S., Kroll, S., Cherry, T.L., 2011. Do you not like Pigou or do you not understand him? Tax aversion and earmarking in the lab. J. Environ. Econ. Manag. 62 (1), 53–64. Kempf, A., Ruenzi, S., 2006. Status quo bias and the number of alternatives: an empirical illustration from the mutual fund industry. J. Behav. Financ. 7 (4), 204–213. McCaffery, E.J., Baron, J., 2006. Thinking about tax. Psychol. Public Policy Law 12, 106–135. Mutz, D.C., Martin, P.S., 2001. Facilitating communication across lines of political difference: the role of mass media. Am. Polit. Sci. Rev. 95 (1), 97–114. Niederle, M., Vesterlund, L., 2007. Do women shy away from competition? Do men compete too much?. Q. J. Econ. 122 (3), 1067–1101. OFM, 2010. Office of Financial Management, State of Washington. Fiscal impact for Initiative 1098 August 2010; 〈http://www.ofm.wa.gov/ballot/2010/1098. pdf〉 (Accessed 15 October 2015). Rhodes, E., Axsen, J., Jaccard, M., 2014. Does effective climate policy require well-informed citizen support? Glob. Environ. Change 29, 92–104. Schuitema, G., Steg, L., van Kruining, M., 2011. When are transport pricing policies fair and acceptable? Social. Justice Res. 24, 66–84. Scott, A., 2002. Identifying and analysing dominant preferences in discrete choice experiments: an application in health care. J. Econ. Psychol. 23 (3), 383–398. Slater, M.D., 2007. Media selectivity and media effects and their impact on individual behaviour and social identity. Commun. Theory 17, 281–303. Smith, N., Leiserowitz, A., 2013. American evangelicals and global warming. Glob. Environ. Change 23 (5), 1009–1017. Stroud, N.J., 2010. Polarization and partisan selective exposure. J. Commun. 60, 556–576. Sunstein, C., 2001. Republic.com. Princeton University Press. Tiezzi, S., Xiao, E., 2016. Time delay, complexity and support for taxation. J. Environ. Econ. Manag. 77, 117–141. Unsworth, K.L., Fielding, K.S., 2014. It's political: how the salience of one's political identity changes climate change beliefs and policy support. Glob. Environ. Change 27, 131–137.