Explaining the Willingness to Pay for Environmental Protection rough draft – please do not cite without permission Michael Dorsch∗ January 28, 2011

Abstract This paper empirically investigates individuals’ willingness to pay for stronger environmental protection, analyzing nearly 42,000 survey responses across 41 countries from the 2005-2008 wave of the World Values Survey. Several interesting results emerge from the analysis: (i) within countries, there is support for the notion that environmental protection is a normal good, but not across countries, (ii) selfidentification as a “world citizen” is the strongest determinant of demand for greater environmental protection, (iii) there seems to be diminishing marginal utility from environmental protection, such that respondents from countries with high EPI scores were less likely to demand further environmental protection, (iv) democratic development increases the demand for environmental protection, and (v) the determining factors of the demand for environmental protection are significantly different in magnitude depending on the level of economic and institutional development. Keywords: Environmental protection policy, Political preferences, Global public goods, World Values Survey

∗

Department of Economics, The American University of Paris. Email:

1

1

Introduction

Introduce the idea, within the context of political economy National environmental policies, as any other policy, emerge from political processes that must take into account the policy preferences of citizens within the nation. In wellfunctioning democracies, it is believed that policy preferences are aggregated in such a way that the policy that gets implemented approximates the median preference within the society. Even in autocratic societies, citizens’ policy preferences constrain the policy direction of the leadership, in the sense that (sometimes violent) political transitions may occur if the leadership sets policy that is radically different from the plurality preference. Regardless of the political institution, preferences for environmental protection are similar to preferences for traditional public goods: the benefits of environmental protection are non-excludable and non-rival and the costs are generally foregone private market consumption. When citizens determine their preferences for environmental protection, their relative income in society is likely to be an important cost-side factor (either due to greater taxation or proportional adjustments in consumption patterns). This paper aims to explain the extent to which relative income can account for differences in individuals’ “willingness-to-pay” for greater environmental protection, and aims to identify other determining factors of willingness to pay. The data sources and the pooled sample I consider individual-level survey responses from the 2005-2008 wave of the World Values Survey (WVS) [World Values Survey Association (2009)]. The question of paramount interest, which approximates the respondents’ “willingness-to-pay” for greater environmental protection, was asked in 41 countries to nearly 42,000 respondents. The WVS questions also provide information on the respondents’ relative incomes, education levels, ages, attitudes about global warming, and attitudes about levels of citizenship. In addition, I have gathered various country-level characteristics from the myriad of data sources compiled in the Quality of Government database [Teorell et al. (2010)], including the Environmental Performance Index score [Etsy, D. et al. (2008)], national income per capita [United Nations Statistics Division, Economics Statistics Branch (2009)], and the widely-used pol i t y score of democratic development [Marshall and Jaggers (2002)]. The resulting sample pools WVS responses across countries and includes the aforementioned country-level characteristics.

2

Overview of empirical methodology The dependent variables in the regression analysis are converted into binary variables, so probit estimations are employed. The empirical analysis includes regressions on the full sample, as well as sub-sample regressions on the OECD-member countries and the non-OECD countries. The analysis reveals interesting differences between the OECD and non-OECD countries as to the determinants of preferences for environmental protection.1 Furthermore, I have preformed within-country estimations for each of the 41 countries and analyze how the factors that determine willingness to pay for environmental protection differ in their magnitude depending on the level of development.

2

Data description

2.1

Dependent variables

The dependent variables describe individuals’ “willingness to pay” for additional environmental protection in their country. The variables are taken from the 2005-2008 wave of the World Values Survey. In 41 countries, respondents were asked if they agree with the following statement: “I would give part of my income if I were certain that the money would be used to prevent environmental pollution.” The responses were used to create a binary dependent variable, w t pinc = 1 if the respondent strongly agreed or agreed and w t pinc = 0 if the respondent disagreed or strongly disagreed with the statement. Similarly, respondents were asked if they would support higher tax rates to finance environmental protection. Respondents were asked if the agree with the following statement: “I would be willing to pay higher taxes if I were certain that the money would be used to prevent environmental pollution.” The second binary dependent variable is similarly constructed, w t pt ax = 1 if the respondent strongly agreed or agreed and w t pt ax = 0 if the respondent disagreed or strongly disagreed with the statement. The questions seem to get at marginal willingness to pay, since there is already some level of environmental protection present, in all countries. The empirical results confirm that respondents are thinking in marginal terms, as the proportion of respondents answering “yes” is lower in the richer countries, after controlling for demographic and socio-economic respondent characteristics. This supports the notion that environmental quality is a normal good 1

Robustness checks include running the regressions using linear probability models, running ordered probit estimations on the raw, ordinal data, and re-running the probit estimations with alternative binary dependent variables, constructed using different cut-off points for the binary dependent variables.

3

with diminishing marginal utility.

2.2 2.2.1

Independent variables Individual-level explanatory variables

All individual-level explanatory variables are taken from the 2005-2008 wave of the World Values Survey. The individual=level variables include relative income, four attitudinal variables, two education variables, the respondent’s age, and the respondent’s political ideology (where available).2 The exercise begins by looking at willingness to pay and a(n imperfect) measure of relative personal income within a country. The survey question in WVS asks respondents for their perceived income decile, which is reported as 1-10 in the variable incdec.3 All regressions were ran with non-linear income decile effects, but no significant non-linear effects were found. Ex ante, higher income deciles should, on average, be more willing to pay for environmental protection.4 The attitudinal variables are all binary. First, globalwar m = 1 if the respondent felt that global warming was a “serious” or “very serious” problem. Second, there are two citizenship questions: wor l dci t = 1 if the respondent agreed or strongly agreed that they see themselves as a world citizen and nat ionci t = if the respondent strongly agreed that they see themselves as a citizen of their nation. The final attitudinal question asks respondents if it is justifiable to avoid fare on public transport on a scale from 1 to 10, where 1 is “never justifiable” and 10 is “always justifiable”. Respondents who answered less than 5 to this question are coded as never f r ee = 1. To control for education, I consider whether or not the respondent completed high school (hs = 1 if they finished high school) and whether or not the respondent gets information about the world from books (books = 1 if the respondent used a book to learn about the world in the week prior to the interview). I also control for the respondent’s age in years (ag e) and whether or not they consider themselves politically liberal 2

Several authors have found a significant difference between the sexes, such as Israel and Levinson (2004) and others. Also, urban residence seems to be in a lot of studies, e.g., Marquart-Pyatt (2008). I may need these variables too. 3 There appears to be a systematic under-reporting of income decile (cumulatively, most countries have less than 10% reporting to be in the top two deciles of the income distribution). 4 A result derived in a theoretical companion paper [Dorsch (2011)] is that most-preferred levels of environmental protection are increasing in individual income levels due to the diminishing marginal utility of consumption and the constant marginal utility of emissions reduction.

4

(l i ber al = 1 if, on a scale from 1 (left) to 10 (right), they consider themselves between 1 and 4).5 2.2.2

Country-level explanatory variables

At the country level, I control for per capita income, the level of environmental protection, and the level of democratic development. Per capita GDP from 2002 is expressed in terms of PPP-adjusted 1990 US dollars, taken from the United Nations. l gd pc is the natural logarithm of the UN per capita GDP data. The Environmental Protection Index (epi) measures “how well countries succeed in reducing environmental stress on human health and promoting ecosystem vitality and sound natural resource management.” The index ranges between 0 and 100 and is increasing in environmental performance. All of the regressions were run with non-linear epi terms, which are reported in any cases where epi was found to have significant non-linear effects. Finally, poli t y is the “revised combined polity score,” a measure of democratic development, which ranges between −10 (strongly autocratic) and 10 (strongly democratic).

3

Empirical analysis

3.1

Preliminaries

As motivation for the more detailed statistical analysis to follow, consider Figure 1, which shows the fraction of respondents who agreed they would pay part of their income for greater environmental protection (w t pinc = 1), by income decile, in Germany, the United States, China and India. Two notions emerge from the Figure 1: (i) willingness to pay for further environmental protection does appear to increase with income in all four countries (the effect is clearly monotonic in Germany, less so for the others) and (ii) on average, higher willingness to pay for further environmental protection in China and India compared to Germany and the United States. The first notion is as expected. The second notion, however, deserves further attention, as it contradicts well-known theories which predict that the desire to protect the environment strengthens along the development path.6 5

Non-OECD countries were far less likely to ask this question, so it was only used as a control variable for the OECD countries. 6 In the environmental economics literature, environmental protection has been modeled as a normal good, the demand for which increases with income level [Baumol and Oates (1979)]. In the environmental

5

Figure 1: Proportion of individuals willing to pay part of their income for further environmental protection, by income decile, in Germany (top left), the United States (top right), China (bottom left) and India (bottom right).

Indeed, it seems that the survey question is getting at individuals’ marginal willingness to pay for further environmental protection (MWTP). Individuals in some advanced economies, such as Germany, may feel that their country has achieved a satisfactory level of environmental protection and they are not willing to pay for marginal improvements. In other words, the marginal utility from environmental protection may be diminishing, which would mean that willingness to pay for additional environmental protection should be lower for countries that have already achieved a high level of environmental quality. As such, it will be important to control for country-level fixed effects, such as the achieved level of environmental protection in the respondents’ home countries. Table 1 presents the proportion of respondents willing to pay for further environmental protection by country, where the countries are sorted in ascending order according to per capita sociology literature, the post-materialism hypotheses describes how individuals in affluent societies are more free to pursue post-materialistic goals (including environmental protection) as they become less preoccupied with the economic struggle to survive [Ingelhart (1995) and Ingelhart (1997)].

6

GDP. The last column of Table 1 presents the Environmental Protection Index score.

3.2

Pooled sample regression analysis

I run probit specifications on the pooled sample of individual responses with country fixed effects for 41 countries to explain the variation in the willingness to pay binary variables controlling for the individual and country-level variables described above. The baseline specification that I consider is the following: � � prob(W T Pi j = 1|Xi , Xj ) = Φ α� Xi + β � Xj + ui j ,

(1)

where W T Pi j ∈ {0, 1} are the binary responses variable of individual i from country j, Φ

represents the standard normal cumulative density function, Xi is a vector of individuallevel explanatory variables, Xj is a vector of country-level explanatory variables, and α and β are vectors of coefficients to be estimated using probit. The regression output in Tables 2 and 3 report the average marginal effects for each variable. Table 2 contains the estimated marginal effects on the probability that respondents were willing to pay part of their income for further environmental protection and Table 3 is the analogue for the willingness to pay higher taxes. Both tables report the estimates for the full sample of 41 countries in the first column, only OECD member countries in the second column and only non-OECD countries in the third column, As measures of fit, the last two rows of each table report the pseudo−R2 and the percentage of within-sample predictions that the specifications get correct. Generally speaking, the ability to predict more than half of the binary outcomes correctly is seen as support for the specification. As the regression output tables indicate, 68% of responses to the WTPinc question were correctly predicted and 62% of the WTPtax responses were correctly predicted, which gives support to the specifications, despite the relatively low pseudo−R2 statistics. As a robustness check, I also estimated linear probability models, which are presented in Tables 4 and 5. 3.2.1

Full pooled sample results

For both the income and tax questions, there are several individual-level variables that are highly significant covariates with W T P. Individuals that believe that global warming is a serious problem are more likely to be willing to pay for environmental protection, as are those that view themselves as world citizens. Identification as a national citizen had a significantly positive impact on the probability that W T P inc = 1, but it was in7

significant in the W T P t ax regression. Of the attitudinal variables, viewing oneself as a world citizen had the largest marginal impact on the probability that the respondent was willing to pay for further environmental protection. For example, respondents who identified themselves as a world citizen were about 15 percentage points more likely to respond that they were willing to pay for further environmental protection (for each question). As expected, education has a significantly positive impact on willingness to pay for environmental protection. Finally, those with higher incomes (relative to their countrymen) were more likely to agree to making economic sacrifice to protect the environment, with the probability an individual had W T P = 1 increasing 1.5% points for each income decile. Non-linear income decile terms were included in previous specifications that I ran, but found to be insignificant, so are not included in the specifications presented in Tables 2–5. As for the country-level variables, individuals from countries with higher EPI scores were less willing to pay for further environmental protection, though the negative effect was diminishing. This result supports the notion that there is diminishing marginal utility from environmental protection (or, that the demand for environmental protection may become saturated). For both W T P questions, respondents from more democratic countries were more likely to be willing to pay for further environmental protection, indicating that perhaps contributions to public goods were viewed as a waste of money in more autocratic regimes. Surprisingly, respondents from higher income countries were no more likely to be willing to pay for further environmental protection, which poses a serious challenge to theories of post-materialism. Tables 4 and 5 demonstrate the estimated marginal effects are highly robust to estimation with the linear probability model rather than probit. 3.2.2

Differences between OECD and non-OECD pooled sample results

Tables 2–5 also provide sub-sample estimations, breaking the pooled sample into OECD and non-OECD pooled sub-samples.7 For both the income and the tax question, many of the estimated marginal effects are similar between the pooled sub-samples (incdec, wor ld ci t, never f r ee, and l g d pc). There are some interesting differences, however. First, the impact of believing that global warming is a serious problem is much stronger (8% points for the income question and nearly 10% points for the tax ques7

Note that there are far more observations from non-OECD countries (24,345 from non-OECD countries and 15,470 from OECD countries). This alone seems to make the paper a contribution as many of the papers I have reviewed only consider the developed world.

8

tion). As global warming is more of a global public bad, it seems that being motivated to alleviate it would represent more of a “post-materialist” intention. Following the “objective problems and subjective values” hypothesis of Ingelhart (1995), this difference can be rationalized by noting that the developing world is likely to be motivated by more local environmental problems than the developed world. A second difference between the sub-sample estimates is less easily rationalized. For the income question, identification as a national citizen is significantly positive for the non-OECD respondents, but insignificant for OECD respondents. For the tax question, the opposite is true. This may indicate that in the OECD, respondents may balk at the income question because they are weary of the free-rider problem among less nationallyminded citizens, whereas in the non-OECD countries, nationalist respondents are more weary about the efficacy of a government program to solve tough problems.8 As for education, in the OECD countries the effects of formal education (hs) and informal education (books) have significantly positive effects on willingness to pay for environmental protection of roughly the same magnitude. In the non-=OECD countries, however, informal education has a strong significantly positive effect on willingness to pay, whereas formal education has no effect. There are also interesting differences between the sub-samples in regards to the country-level variables. For one, willingness to pay for environmental protection is higher in countries where the government is already doing a better job protecting the environment. In the non-OECD sub-sample, the epi score has no effect on willingness to pay, however. This result is difficult to explain, as it seems to contradict the notion that there is diminishing marginal utility to environmental protection. Finally, quality of democratic institutions is shown to reduce willingness to pay in the OECD countries and increase it in the non-OECD countries. All of the similarities and differences between the OECD and non-OECD sub-samples are robust to estimation with the linear probability model rather than probit.

3.3

Comparison of within-country estimates

In addition to the pooled sample results, I have also preformed within-country probit estimations for each of the 41 countries in which the willingness to pay part of income for environmental protection question was asked in the 2005-2008 wave. The benefit 8

Indeed, the relevant government agency may not even exist in many non-OECD countries and nationalist respondents may have better information about the capacity of their national government.

9

Figure 2: Estimated marginal effects of believing global warming is serious problem on willingness to pay for greater environmental protection and country characteristics. The estimates are plotted against log of per capita GDP (top left), Environmental Performance Index (top right), Polity score (bottom left), and Economist Intelligence Unit’s Functioning of Government score (bottom right).

Notes: Log of per capita GDP linear fit has R2 = 0.259. EPI score linear fit has R2 = 0.171. Polity score linear fit has R2 = 0.190. EIU functioning of government score linear fit has R2 = 0.146.

of such an exercise is that we can learn more details as to the differential impacts on willingness to pay of relative income, perception of the global warming problem, and attitudes about world citizenship. The within-country estimates of the marginal effect of believing that global warming is a serious problem are presented in the first column of results in Table 6. As indicated in the table, the sign on the estimated marginal effect is positive in 40/41 countries (significantly positive in 35/41 countries). The magnitude of the effect varies, however, and is shown to depend positively on the level of development. Figure 2 plots these estimates against several country characteristics: (i) log of per capita GDP, (ii) EPI score, (iii) Polity score, and (iv) EIU Functioning of Government score. The marginal effect of believing 10

Figure 3: Estimated marginal effects of self-identification as a world citizen on willingness to pay for greater environmental protection and country characteristics. The estimates are plotted against log of per capita GDP (top left), Environmental Performance Index (top right), Polity score (bottom left), and Economist Intelligence Unit’s Functioning of Government score (bottom right).

Notes: Log of per capita GDP linear fit has R2 = 0.068. EPI score linear fit has R2 = 0.022. Polity score linear fit has R2 = 0.050. EIU functioning of government score linear fit has R2 = 0.025.

that global warming is a serious problem is greater for higher levels of development, for any of the four proxies for development. The within-country estimates of the marginal effect of self-identification as a world citizen are presented in the third column of results in Table 6. The table indicates that the sign on the estimated marginal effect is positive in 38/41 countries (significantly positive in 33/41 countries). The magnitude of the effect, again, is shown to depend positively on the level of development. Figure 3 plots the estimates against the same battery of country characteristics as in Figure 2. Again, the estimated marginal effect of self-identification as a world citizen is of a greater magnitude in countries where the four development indicators are higher. 11

Figure 4: National Income and the Within-Country Effect of Higher Income. Per Capita GDP on the left and Natural Log of Per Capita GDP on the right.

Notes: Linear fit on Per Capita has R2 = 0.003. Linear fit on Log of Per Capita has R2 = 0.005.

The within-country estimates of the marginal effect of higher relative income are presented in the second column of results in Table 6. The table indicates that the sign on the estimated marginal effect is positive in 38/41 countries (significantly positive in 24/41 countries). In contrast to the two previous series of estimated coefficients, however, the relative income estimates do not vary systematically with any of the development indicators. Figure 4 plots the estimates against per capita GDP and the log of per capita GDP. There is essentially no relation. There is no relation with the other development indicators either, so the plots are omitted.9 In the last column of Table 6, I calculate the predicted probability that an individual in the median income decile is willing to pay part of his income for greater environmental protection. As the table indicates, the predicted probability is greater than 0.5 in 37/41 countries (significantly greater than 0.5 in 33/41).10 Figure 5 plots the predicted probabilities against the same battery of development indicators and finds a negative relation between the predicted probabilities and all four development indicators.11 9

That within-country relative income does not have a stronger effect in richer countries is in line with the results presented in Israel and Levinson (2004). 10 So, why aren’t more governments doing more about this problem?? 11 In the analog to the top left panel of Figure 6, Israel and Levinson (2004) find “after controlling for age, sex, household income, education, and city size . . . low income countries display no pattern with respect to GDP, while willingness to pay appears to decline for high-income countries.” I find a more uniform decrease in WTP with GDP across my country-level fitted values.

12

Figure 5: Country-level fitted values (for the median income decile) of the probability respondent is willing to pay part of his income for greater environmental protection plotted against country characteristics. Log of per capita GDP (top left), Environmental Protection Index (top right), Polity score (bottom left), and Economist Intelligence Unit Functioning of Government score (bottom right).

Notes: Log of GDPC linear fit has R2 = 0.129. EPI linear fit has R2 = 0.106. Polity score has linear fit R2 = 0.053. EIU linear fit has R2 = 0.089.

3.4

Do preferences affect environmental outcomes or do bad environmental outcomes affect preferences?

The top right panel of Figure 5, where the fitted probability is plotted against the EPI score, provides a hint. It seems that individuals in the middle of the income distribution are more willing to pay for environmental protection in societies that do not have strong environmental protection already in place. Figure 6 provides further support of this idea. Figure 7 plots carbon dioxide emissions in tons per 1000 US$ PPP-adjusted GDP against the willingness to pay for greater environmental protection. The left-hand panel plots emissions against the proportion respondents in the median-income decile who are

13

Figure 6: C02 emissions divided by GDP (Y-axis) and the percentage of respondents in the median income decile willing to pay part of their income (on left) and the percentage of respondents in the median income decile willing to pay higher taxes (on right).

Notes: WTPincM fit has R2 = 0.124. WTPtaxM linear fit has R2 = 0.121.

willing to pay part of their income and the right-hand panel plots emissions against the proportion of median-income decile respondents who are willing to pay higher taxes. The positive relationship implies that effective carbon emissions are higher in societies that have a greater demand for environmental protection in the middle of the income distribution. Therefore, it appears that bad environmental outcomes drive the demand for environmental protection, rather than the demand for environmental protection driving policy that results in good environmental outcomes.

3.5

Miscellaneous empirical result

Finally, I investigate why in some countries there may be a greater willingness to pay a part of their income, rather than higher taxes, to protect the environment. To this end, I subtract the proportion of respondents that are willing to pay higher taxes from the proportion willing to pay part of their income. Figure 7 plots these differences against a measure of the quality of government. The left-hand panel considers the difference among the proportions in the whole population, while the right-hand panel considers the differences among the middle income decile. In both panels, it appears that the difference is larger in societies where the government does not function as well.

14

Figure 7: Difference between willingness to pay higher income and willingness to pay higher taxes against EIU functioning of government index. Entire population (on left) and median income deciles (on right).

Notes: Linear fit on entire population has R2 = 0.084. Linear fit on median income deciles has R2 = 0.061.

4

Conclusion

This paper is not finished yet.

Appendix: Relevant literature to review • On explaining policy preferences using survey responses: – The most related is Tjernstöm and Tietenberg (2008), who look at climate change policies. – Israel and Levinson (2004) set out to show that marginal willingness to pay for environmental protection is diminishing. They also use the WVS, though the third wave, a different question, a near exclusive focus on income, and linear probability rather than probit. They find no relation among low income countries, and that MWTP does decrease with GDP for some rich countries. See also Israel (2004). – There has been a lot of work on this in the sociology literature. See, for example, Bloom (1995); Franzen and Meyer (2010); Gerhards and Lengfeld (2008); Kemmelmeier et al. (2002); Marquart-Pyatt (2008). – Hainmueller and Hiscox (2007) look at immigration policies in Europe.

15

• On the role of income in determining individual preferences for environmental protection: This is best fleshed out, as an assumption, in Finus (2003).

• On the role of democratic development in national environmental policy outcomes:

See Tjernstöm and Tietenberg (2008) for a (pretty weak) empirical investigation and Congleton (1992) for a comparison of democracy versus autocracies (includes a nice theoretical part. Jones and Manuelli (2001) provides a political economy model of pollution/growth trade-off. Magnani (2001) discusses how individual preferences are converted into public policy, with an eye towards deriving an invertedU EKC. Torras and Boyce (1998) includes “political freedom” as an explanatory variable to explain differences in cross-country differences in pollution emissions. Pargal and Wheeler (1996) looks at informal (local) regulations.

• Things to read via Tjernstöm and Tietenberg (2008) : Brechin and Kempton (1994); Fransson and Gärling (1999); Martínez-Alier (1995); and Weaver (2003).

• Environmental Kuznets Curve: – Stern (2004) and Dasgupta et al. (2002) review the literature. Stern (2004) is more technical, and is extremely critical of the EKC literature, both theoretically and empirically. Copeland and Taylor (2004) also provides a critical review, concluding (quoted by Stern (2004)) “Our review of both the theoretical and empirical work on the EKC leads us to be skeptical about the existence of a simple and predictable relationship between pollution and per capita income.” Martínez-Alier (1995) tries to say that environmental protection is a luxury good that poor countries cannot afford. This seems to be debunked by the reviews above, but it still may be interesting to read. – Note that Grossman and Krueger (1993) were the first to use the term Environmental Kuznets Curve. Andreoni and Levinson (2001) provides some “simple analytics” using scale arguments. López and Mitra (2000) investigates the role of corruption. Ansuategi and Perrings (2000) considers transboundary externalities. • Franzen and Meyer (2010) identifies three theoretical proposals for explaining environmental concern:

1. Ingelhart (1995) has been the most prominent paper to promote the “postmaterialism” explanation... citizens of wealthier nations display more pro16

environmental attitudes due to a shift from materialistic to post-materialistic values in modern society (political freedom, individual self-fulfillment, and environmental protection). This has been challenged since it seems that respondents in developing countries are just as willing to pay for environmental protection. Inglehart then formulates the “objective problems and subjective values” hypothesis, in which wealthy societies want to protect the environment due to post-materialistic values and poor societies want to protect the environment because they are faced with pressing environmental problems (polluted cites, lack of clean water, etc). 2. The “Globalization phenomenon” argument holds that environmental concern has spread to all countries, regardless of their level of wealth. See Dunlap and Mertig (1995) and Dunlap and Mertig (1997) and Gelisson (2007) for support of the globalization phenomenon explanation. 3. The “propensity” or “affluence” hypothesis, which is rooted in classical economic reasoning [Baumol and Oates (1979)] and has been confirmed by Diekmann and Franzen (1999) is the hypothesis that is supported the best by Franzen and Meyer (2010). This hypothesis is closest to what I have modeled in the theory accompaniment... that environmental quality is a public good with individual demand that is normal (increasing in individual income levels). More specifically, there may be diminishing marginal WTP for environmental protection, so that the relationship between income and environmental concern is not linear, but concave [Israel and Levinson (2004)]. • Other opinion surveys about the environment: International Social Survey Program (ISSP) used by Marquart-Pyatt (2008) and Franzen and Meyer (2010).

Appendix: Tables

17

Table 1: SUMMARY STATISTICS – INCOME, WILLINGNESS TO PAY, AND ENVIRONMENTAL PERFORMANCE I NDEX Country

2002 GDP per capita

Mean WTP Income

Mean WTP Tax

EPI

Vietnam Ethiopia Moldova Burkina Faso Ghana India Georgia Ukraine Indonesia Egypt China Morocco Romania Serbia Thailand Bulgaria Poland Uruguay Turkey South Africa Brazil Mexico Malaysia Chile Trinidad Taiwan South Korea Slovenia Cyprus Spain Italy Andorra Australia Germany Canada Japan United States Finland Sweden Switzerland Norway

188.80 214.95 391.66 461.76 489.77 561.58 762.03 932.78 946.63 977.88 1017.73 1323.47 1645.61 2193.76 2303.47 2425.08 2505.23 2926.45 3048.88 3067.43 3567.90 3660.07 3966.54 4171.19 6706.43

0.963 0.793 0.648 0.805 0.829 0.680 0.781 0.472 0.722 0.487 0.824 0.446 0.382 0.545 0.865 0.573 0.528 0.442 0.836 0.537 0.530 0.840 0.619 0.569 0.747 0.835 0.772 0.703 0.734 0.490 0.614 0.628 0.590 0.354 0.703 0.665 0.514 0.572 0.705 18 0.645 0.685

0.908 0.735 0.552 0.753 0.744 0.619 0.470 0.472 0.590 0.308 0.737 0.388 0.351 0.495 0.742 0.511 0.467 0.429 0.782 0.464 0.500 0.705 0.533 0.525 0.593 0.648 0.521 0.544 0.631 0.473 0.521 0.629 0.582 0.264 0.637 0.535 0.505 0.572 0.6955 0.597 0.679

73.91 58.85 70.74 44.34 70.78 60.28 82.18 74.10 66.19 76.28 65.08 72.09 71.93

11181.23 11496.61 13437.93 17656.55 23345.75 23581.01 24782.14 25910.90 26000.70 26926.48 28052.26 34005.59 34644.59 36744.29 38525.73

79.15 78.47 80.49 82.29 75.90 68.98 82.65 79.80 83.98 83.44 70.36 80.83 79.36 86.30 79.19 83.14 84.22 79.83 86.31 86.64 84.54 81.03 91.44 93.12 95.51 93.12

Notes: GDP per capita is PPP-adjusted in 1990 US$, calculated by the United Nations Statistics Division. Willingness to Pay variables are from the 2005-2008 wave of the World Values Survey, and is the

Table 2: DEPENDENT VARIABLE: WILLINGNESS TO PAY PART MARGINAL EFFECTS, WITH COUNTRY FIXED EFFECTS

OF

INCOME. PROBIT AVERAGE

Variable

Full Sample

OECD Countries

Non-OECD Countries

globalwarm (d)

0.0867*** (0.019) 0.0146*** (0.004) 0.1505*** (0.023) 0.0245* (0.013) 0.0449** (0.023) 0.0731*** (0.011) 0.0001 (0.000) 0.0001 (0.014) -0.0298*** (0.011) 0.0002** (0.000) -0.0404 (0.025) 0.0025** (0.001)

0.1276*** (0.015) 0.0166*** (0.003) 0.1561*** (0.030) 0.0089 (0.008) 0.0679*** (0.020) 0.0545*** (0.011) 0.0011*** (0.000) 0.0002 (0.015) 0.0063*** (0.002)

0.0463* (0.025) 0.0153** (0.007) 0.1284*** (0.032) 0.0469*** (0.017) 0.0191 (0.032) 0.0834*** (0.016) -0.0001 (0.000) -0.0013 (0.018) -0.0029 (0.004)

-0.0175 (0.023) -0.1087*** (0.015) 0.0700*** (0.014)

-0.0584 (0.041) 0.0026* (0.001)

41830 0.055 67.97

15470 0.079 68.30

24345 0.060 68.48

incdec worldcit (d) natcit (d) hs (d) books (d) age neverfree (d) epi epi2 lgdpc polity liberal (d)

N pseudo-R2 % Correct

Notes: *, **, and *** indicate significance at 10, 5, and 1 % levels, respectively. Standard errors (in parentheses) are heteroskedasticity-robust. (d) for discrete change of dummy variable from 0 to 1.

19

Table 3: DEPENDENT VARIABLE: WILLINGNESS TO PAY HIGHER TAXES. PROBIT AVERAGE MARGINAL EFFECTS, WITH COUNTRY FIXED EFFECTS Variable

Full Sample

OECD Countries

Non-OECD Countries

globalwarm (d)

0.0716*** (0.019) 0.0175*** (0.004) 0.1453*** (0.021) 0.0112 (0.013) 0.0131 (0.020) 0.0759*** (0.011) -0.0000 (0.000) 0.0019 (0.017) -0.0430*** (0.013) 0.0003*** (0.000) -0.0257 (0.028) 0.0025** (0.001)

0.1225*** (0.016) 0.0147*** (0.004) 0.1364*** (0.035) 0.0298** (0.012) 0.0571** (0.024) 0.0412*** (0.010) 0.0010*** (0.000) 0.0086 (0.022) 0.0088** (0.004)

0.0294 (0.023) 0.0214*** (0.006) 0.1346*** (0.026) 0.0090 (0.019) -0.0251 (0.026) 0.0973*** (0.015) -0.0005 (0.000) -0.0079 (0.023) -0.0052 (0.003)

-0.0281 (0.030) -0.0652** (0.032) 0.0919*** (0.019)

-0.0484 (0.043) 0.0025** (0.001)

41744 0.044 61.81

15482 0.057 61.94

24252 0.051 63.57

incdec worldcit (d) natcit (d) hs (d) books (d) age neverfree (d) epi epi2 lgdpc polity liberal (d)

N pseudo-R2 % Correct

Notes: *, **, and *** indicate significance at 10, 5, and 1 % levels, respectively. Standard errors (in parentheses) are heteroskedasticity-robust. (d) for discrete change of dummy variable from 0 to 1.

20

Table 4: DEPENDENT VARIABLE: WILLINGNESS TO PAY PART MODEL, WITH COUNTRY FIXED EFFECTS MODEL

OF I NCOME .

LINEAR PROBABILITY

Variable

Full Sample

OECD Countries

Non-OECD Countries

globalwarm

0.0830*** (0.019) 0.0139*** (0.004) 0.1500*** (0.022) 0.0231* (0.012) 0.0454** (0.022) 0.0694*** (0.011) 0.0001 (0.000) -0.0003 (0.014) -0.0244** (0.010) 0.0002** (0.000) -0.0380 (0.024) 0.0021* (0.001)

0.1222*** (0.015) 0.0148*** (0.003) 0.1529*** (0.029) 0.0075 (0.008) 0.0654*** (0.019) 0.0507*** (0.010) 0.0011*** (0.000) -0.0017 (0.014) 0.0063*** (0.002)

0.0421* (0.025) 0.0145** (0.006) 0.1281*** (0.030) 0.0452** (0.017) 0.0230 (0.032) 0.0786*** (0.016) -0.0001 (0.000) -0.0003 (0.017) -0.0026 (0.003)

-0.0170 (0.022) -0.0966*** (0.013) 0.0651*** (0.015)

-0.0528 (0.038) 0.0021* (0.001)

41830 0.067

15470 0.097

24345 0.072

incdec worldcit natcit hs books age neverfree epi epi2 lgdpc polity liberal

N R2

Notes: *, **, and *** indicate significance at 10, 5, and 1 % levels, respectively. Standard errors (in parentheses) are heteroskedasticity-robust. (d) for discrete change of dummy variable from 0 to 1.

21

Table 5: DEPENDENT VARIABLE: WILLINGNESS MODEL, WITH COUNTRY FIXED EFFECTS MODEL

TO

PAY HIGHER TAXES. LINEAR PROBABILITY

Variable

Full Sample

OECD Countries

Non-OECD Countries

globalwarm

0.0684*** (0.019) 0.0168*** (0.004) 0.1431*** (0.020) 0.0106 (0.013) 0.0134 (0.019) 0.0729*** (0.010) -0.0000 (0.000) 0.0014 (0.016) -0.0377*** (0.012) 0.0002*** (0.000) -0.0241 (0.027) 0.0023** (0.001)

0.1185*** (0.015) 0.0137*** (0.003) 0.1323*** (0.033) 0.0283** (0.012) 0.0549** (0.023) 0.0394*** (0.010) 0.0010*** (0.000) 0.0077 (0.021) 0.0085* (0.004)

0.0260 (0.023) 0.0202*** (0.006) 0.1316*** (0.025) 0.0083 (0.018) -0.0220 (0.024) 0.0921*** (0.014) -0.0004 (0.000) -0.0077 (0.021) -0.0048 (0.003)

-0.0269 (0.029) -0.0618* (0.029) 0.0871*** (0.018)

-0.0440 (0.040) 0.0021* (0.001)

41744 0.067

15482 0.097

24252 0.072

incdec worldcit natcit hs books age neverfree epi epi2 lgdpc polity liberal

N R2

Notes: *, **, and *** indicate significance at 10, 5, and 1 % levels, respectively. Standard errors (in parentheses) are heteroskedasticity-robust. (d) for discrete change of dummy variable from 0 to 1.

22

Table 6: DEPENDENT VARIABLE: WILLINGNESS TO PAY PART MARGINAL EFFECTS, COUNTRY-LEVEL REGRESSIONS

OF

INCOME. PROBIT AVERAGE

Country

globalwarm

incdec

worldcit

pr(wtpinc=1|incdec=5)

Vietnam Ethiopia Moldova Burkina Faso Ghana India Georgia Ukraine Indonesia Egypt China Morocco Romania Serbia Thailand Bulgaria Poland Uruguay Turkey South Africa Brazil Mexico Malaysia Chile Trinidad Taiwan South Korea Slovenia Cyprus Spain Italy Andorra Australia Germany Canada Japan United States Finland Sweden Switzerland Norway

0.0276∗∗∗ 0.0422∗ 0.0667∗∗ 0.0081 0.0394∗ 0.0202 0.0507∗ 0.1137∗∗∗ 0.0575∗∗ −0.0825∗∗∗ 0.0500∗∗ 0.1030∗∗ 0.0877∗∗∗ 0.3137∗∗∗ 0.0407∗∗ 0.1395∗∗∗ 0.0202 0.0276 0.1618∗∗∗ 0.0776∗∗∗ 0.0809∗∗∗ 0.0404∗ 0.0994∗∗∗ 0.1151∗∗∗ 0.1348∗∗∗ 0.0510∗∗ 0.0913∗∗∗ 0.1176∗∗∗ 0.1466∗∗∗ 0.2106∗∗∗ 0.1032∗∗ 0.0825∗∗ 0.2139∗∗∗ 0.1227∗∗∗ 0.0756∗∗∗ 0.0676 0.2310∗∗∗ 0.1647∗∗∗ 0.1017∗∗∗ 0.1654∗∗∗ 0.0978∗∗∗

0.0027 0.0249∗∗∗ 0.0287∗∗∗ 0.0022 0.0008 0.0157∗∗ 0.0021 0.0427∗∗∗ −0.0077 0.0325∗∗∗ 0.0220∗∗∗ 0.0530∗∗∗ 0.0203∗∗∗ 0.0139 0.0008 0.0439∗∗∗ 0.0157∗∗ 0.0120 0.0121∗∗ 0.0134∗∗∗ −0.0001 0.0050 −0.0055 0.0319∗∗∗ 0.0083 0.0310∗∗∗ 0.0280∗∗∗ 0.0056 0.0440∗∗∗ 0.0342∗∗∗ 0.0137∗ 0.0103 0.0070 0.0379∗∗∗ 0.0147∗∗∗ 0.0130∗ 0.0120 0.0243∗∗∗ 0.0022 0.0369∗∗∗ 23 0.0115∗∗

0.0508∗ −0.0669∗∗ 0.0899∗∗∗ 0.1141∗∗ 0.0764∗∗ 0.1473∗∗∗ 0.1010∗∗∗ 0.0306 0.0968∗∗ −0.0422∗∗ 0.0740∗∗ 0.2830∗∗∗ 0.1104∗∗∗ 0.1368∗∗∗ 0.1288∗∗ 0.1500∗∗∗ 0.1473∗∗∗ 0.0903∗∗ 0.0505 0.1661∗∗∗ 0.1352∗∗∗ 0.0613 0.1369∗∗∗ 0.2044∗∗∗ 0.0956∗∗∗ 0.0881∗∗∗ 0.1147∗∗∗ 0.1212∗∗∗ 0.1335∗∗∗ 0.3284∗∗∗ 0.1590∗∗∗ 0.1404∗∗∗ 0.0777∗∗ 0.1612∗∗∗ 0.0647∗ 0.1230 0.0974∗∗∗ 0.0508 0.0975∗∗ 0.1594∗∗∗ 0.0349

0.9747+ 0.8435+ 0.6684+ 0.8656+ 0.8548+ 0.7973+ 0.8038+ 0.4937 0.7674+ 0.5189 0.8908+ 0.5148+ 0.4037− 0.5713+ 0.8751+ 0.6565+ 0.7973+ 0.4491− 0.8733+ 0.5703+ 0.5264 0.8486+ 0.6262+ 0.6075+ 0.7716+ 0.8794+ 0.7881+ 0.7355+ 0.7186+ 0.5309 0.6537+ 0.6283+ 0.6094+ 0.3607− 0.7077+ 0.6993+ 0.5248 0.5948+ 0.7085+ 0.6497+ 0.6851+

Positive Significantly positive

40/41 35/41

38/41 24/41

38/41 33/41

References Andreoni, J. and Levinson, A. (2001) The simple analytics of the environmental Kuznets curve, Journal of Public Economics, 80, 269–286. Ansuategi, A. and Perrings, C. (2000) Transboundary externalities in the environmental transition hypothesis, Environment and Resource Economics, 17, 353–373. Baumol, W. and Oates, W. (1979) Economics, Environmental Policy, and the Quality of Life, Pentice Hall, Engleqood Cliffs, NJ. Bloom, D. (1995) International public opinion on the environment, Science, 269, 354– 358. Brechin, S. and Kempton, W. (1994) Global environmentalism: A challenge to the postmaterialism thesis?, Social Science Quarterly, 75, 245–269. Congleton, R. (1992) Political institutions and pollution control, Review of Economics and Statistics, 74, 412–421. Copeland, B. and Taylor, M. (2004) Trade, growth and the environment, Journal of Economic Literature, 42, 7–71. Dasgupta, S., Laplante, B., Wang, H. and Wheeler, D. (2002) Confronting the environmental Kuznets curve, The Jounral of Economic Perspectives, 16, 147–168. Diekmann, A. and Franzen, A. (1999) The wealth of nations and environmental concern, Environment and Behavior, 31, 540–549. Dorsch, M. (2011) On the structure of international environmental agreements, American University of Paris Working Paper. Dunlap, R. and Mertig, A. (1995) Global concern for the environment: Is affluence a prerequisite?, Journal of Social Issues, 51, 121–137. Dunlap, R. and Mertig, A. (1997) Global environmental concern: An anomoly for postmaterialism, Social Science Quarterly, 78, 24–29. Etsy, D. et al. (2008) 2008 Environmental Performance Index, Yale Center of International Earth Science Information Network, Yale University. 24

Finus, M. (2003) Game Theory and International Environmental Cooperation, XX University Press, X. Fransson, N. and Gärling, T. (1999) Environmental concern: Conceptual definitions, measurement methods, and resaarch findings, Journal of Environmental Psychology, 19, 369–382. Franzen, A. and Meyer, R. (2010) Environmental attitudes in cross-national perspective: A multilevel analysis of the ISSP 1993 and 2000, European Sociological Review, 26, 219–234. Gelisson, J. (2007) Explaining popular support for environmental protection: A multilevel analysis of 50 nations, Environment and Behavior, 39, 392–415. Gerhards, J. and Lengfeld, H. (2008) Support for European Union environmental policy by citizens of EU-member and accession states, Comparative Sociology, 7, 1–27. Grossman, G. and Krueger, A. (1993) Environmental impacts of the North American Free Trade Agreement, in The U.S.-Mexico Free Trade Agreement (Ed.) p Garber, MIT Press, Cambridge, pp. 13–56. Hainmueller, J. and Hiscox, M. (2007) Educated preferences: Explaining attitudes toward immigration in Europe, International Organization, 61, 399–442. Ingelhart, R. (1995) Public support for environmental protection: Objective problems and subjective values in 43 societies, Political Science and Politics, 28, 57–72. Ingelhart, R. (1997) Modernization and Postmodernization: Cultural, Economic, and Political Change in 43 Societies, Princeton University Press, Princeton, NJ. Israel, D. (2004) International support for environmental protection, Environment and Development Economics, X, XXX–XXX. Israel, D. and Levinson, A. (2004) Willingness to pay for environmental quality: Testable empirical implications of the growth and environmental literature, Contributions to Economic Analysis and Policy, 3, 1–29. Jones, L. and Manuelli, R. (2001) Endogenous policy choice: The case of pollution and growth, Review of Economic Dynamics, 4, 369–405.

25

Kemmelmeier, M., Krol, G. and Kim, Y. H. (2002) Values, economics, and proenvironmental attitudes in 22 societies, Cross-Cultural Research, 36, 256–285. López, R. and Mitra, S. (2000) Corruption, pollution and the environmental Kuznets curve, Journal of Environmental Economics and Management, 40, 137–150. Magnani, E. (2001) The evironmental Kuznets curve: Development path of policy result?, Environmental Modelling and Software, 16, 157–166. Marquart-Pyatt, S. (2008) Are there similar sources of environmental concern? Comparing industrialized countries, Social Science Quarterly, 89, 1312–1335. Marshall, M. and Jaggers, K. (2002) Polity IV Project: Political Regime Characteristics and Transitions, 1800-2002, University of Maryland, Maryland. Martínez-Alier, J. (1995) The environment as a luxury good or “too poor to be green?”, Economic Applications, 13, 1–10. Pargal, S. and Wheeler, D. (1996) Informal regulation of industrial pollution in developing countries: Evidence from Indonesia, Journal of Political Economy, 104, 1313–1327. Stern, D. (2004) The rise and fall of the environmental Kuznets curve, World Development, 32, 1419–1439. Teorell, J., Samanni, M., Charron, N., Holmberg, S. and Rothstein, B. (2010) The Quality of Government Dataset, The Quality of Government Institute, University of Gothenburg. Tjernstöm, E. and Tietenberg, T. (2008) Do differences in attitudes explain differences in national climate change policies?, Ecological Economics, 65, 315–324. Torras, M. and Boyce, J. (1998) Income, inequality, and pollution: A reassessment of the environmental Kuznets curve, Ecological Economics, 25, 147–160. United Nations Statistics Division, Economics Statistics Branch (2009) National Accounts Statistics Database, United Nations, New York. Weaver, A. (2003) Determinants of environmental attitudes: A five-country comparison, International Journal of Sociology, 32, 77–108. World Values Survey Association (2009) World Values Survey, 2005-2008, ASEP/JDS, Madrid. 26