Green Giving and Demand for Environmental Quality: Evidence from the Giving and Volunteering Surveys Debra K. Israel* Indiana State University Working Paper * The author would like to thank Indiana State University for supporting this research through a University Research Grant. Thanks to participants at the Midwest Economics Association conference for helpful comments on an earlier version of this paper. The author may be contacted at disrael@isugw.indstate.edu.
Introduction Understanding the demand for environmental protection is important in developing environmental policy. In this paper, I examine the characteristics of environmental donors relative to other households. While the systematic examination of environmental contributions is interesting in and of itself, it also has the potential to contribute to our understanding of the demand for environmental quality. These private donations offer an opportunity to observe people actually choosing to spend some of their income on the environment rather than on other goods or services. These households are clearly demonstrating a positive marginal willingness to pay for environmental quality. These data have been used previously by economists to better understand overall charitable contributions behavior by individuals and households, but not specifically to examine environmental donations (Andreoni et al, 2003; Andreoni, Gale, and Scholz, 1995; Tiehen, 2001). Previous economic research on giving to environmental organizations was conducted at the level of the environmental organization, using tax returns (Richer, 1995). My research is unique in its approach of studying overall environmental contributions from nationally representative household data. Donations for environmental goods have been examined in the context of the contingent valuation method, comparing hypothetical and actual donations in response to contingent valuation surveys using voluntary donation mechanisms (Berrens et al., 2002; Champ and Bishop, 2001; Foster et al., 1997; MacMillan et al., 1999). My study differs from this past research in that the contributions are not solicited for the study itself and environmental donations behavior is examined for the United States as a whole. This will provide a useful
2 reference for future studies to compare the effects of different respondent and household characteristics on environmental giving. In a previous study using the most recent Giving and Volunteering Survey data from 2001 I compared and contrasted the characteristics of environmental donors with other donors and non-donors (Israel, 2007). I found that environmental donations are relatively insensitive to tax price, unlike findings for overall charitable donations. In this paper I further explore these issues utilizing a pooled data set from the 1999 and 1996 rounds of Giving and Volunteering Survey which were carried out using similar methodologies, thus allowing the data to be pooled. 1 Empirical Methods To examine the characteristics of those households which make environmental contributions a probit model is estimated where the dependent variable is one if households contribute to environmental organizations and zero otherwise (Model 1). The explanatory variables include household and respondent characteristics thought to affect environmental giving either through differences in budget constraints or preferences. Much of the literature on charitable donations behavior has focused on income and price elasticities, and household income and the tax price are included as explanatory variables (as natural logarithms). The tax price is one for non-itemizers and one minus the marginal tax rate for itemizers. The marginal tax rate is estimated using average itemized deduction levels for their income level and filing 1 The data are from the Independent Sector (website is www.independentsector.org). The 2001 survey cannot be pooled with these previous years because of significant changes in methodology in 2001 such as using telephone rather than face-to-face interviews and changing the age cutoff to 21 instead of 18.
3 status for itemizers (Statistics of Income, Table 1.2, 1995 and 1998). All dollar amounts from the 1995 data are converted to 1998 dollars using the CPI-U from the Bureau of Labor Statistics. Household characteristics that may affect environmental giving include household size, presence of children in the household, home ownership, and geographical location. The presence of children might increase concern about environmental problems, but persons concerned with environmental problems may also tend to have fewer children. To reduce this ambiguity of interpretation, children in the household is included as an indicator variable, rather than number of children. Different giving patterns may also be associated with respondent characteristics such as age, gender, race or ethnicity, education level, student status, marital status, and retirement status, and these variables are also included. Linking respondent characteristics to household giving is difficult since not all household members necessarily share the same individual characteristics. Therefore, I also estimate the relationship between the respondent characteristics and environmental giving for respondents who are the primary decision-makers for the household s charitable giving in Model 2. 2 Model 1 is estimated on the full sample, so those households who do not contribute to the environment include both households with non-environmental charitable contributions and households with no charitable contributions. However, donor households may be distinct from non-donor households, so to examine the extent that differences reflect differences in the characteristics of environmental donors relative to other donors, Model 3 is estimated including 2 This is the same question on decisionmakers that Andreoni et al. (2003) utilize in their examination of the impact of household decision-making on charitable giving.
4 only households who make donations. In Model 4, the dependent variable in the probit model is for whether or not a household has donations, also allowing examination of how household characteristics of donors in general compare and contrast with those of environmental donors. While the characteristics of households that give to the environment are interesting, in order to understand whether or not they relate to the characteristics of households with an underlying positive demand for the environment, more information is needed. First of all, households may differ in terms of their access to charitable organizations. While everyone could pursue organizations to contribute to, the transactions costs are lower if the organization contacts the household. The Giving and Volunteering survey has information on whether or not a household was asked to make charitable contributions. It would be ideal if this question had also been asked specifically for environmental organizations, but it was not. However, those households who were not asked at all also were not asked by an environmental organization. In Model 5, a probit model is estimated to compare the characteristics of households that were asked to contribute to those that were not asked to contribute. In general, one would expect to see self-reinforcing patterns to occur between those households that are asked to give and those that are more likely to give, since organizations will try to target their asking to those households more likely to give, including those who have given in the past. However, households that are not asked to give may indeed have lower giving simply because they have not been asked. Thus it will be difficult to understand causation because of the endogeneity involved. The focus of this paper is on understanding environmental giving. The Giving and Volunteering survey also asked respondents about their level of confidence in environmental organizations. In Model 6, a regression is estimated with confidence in environmental
5 organizations as the dependent variable, where 4 is the most confident and 1 is the least confident. Respondents who refused or answered don t know are omitted from the analysis, resulting in the slightly lower sample size. Examining the results of this regression may illuminate whether the characteristics of environmental giving are a result of reduced confidence in environmental organizations (perhaps reflecting different preferences for environmental protection or preferences for a different type of funding mechanism) or if they are confident in the organizations, stem from differences in free-riding or income constraints. The regression includes the same explanatory variables as the preceding probit models with the exclusion of the tax price which should not have a separate relevance for confidence in environmental organizations. The data utilized are from the Giving and Volunteering in the United States study from the organization the Independent Sector. These are nationally representative surveys for the United States, which include information on household charitable giving by type of giving. In this study I utilize the 1999 and 1996 Giving and Volunteering data which cover household giving in 1998 and 1995, respectively. The means for the variables used in the analysis are in Table 1. Estimation Results The results from the probit model examining the characteristics of environmental donors are in Table 2. For ease of interpretation, the predicted change in probability resulting from a marginal change in the explanatory variable (df/dx) is reported instead of the parameter estimates for the Probit models. For the indicator variables, this is the change in probability
6 from a discrete change of the variable from 0 to 1. The statistical significance level reported is for the underlying parameter estimate. Examining the characteristics of households that contribute to the environment compared to all other households with Model 1, higher income households are more likely to contribute to the environment. Households with a lower tax price are also more likely to contribute to the environment. In terms of regional differences, households in the South are the least likely to contribute to the environment, whereas those in the East or Midwest are not statistically different from households in the West. In terms of respondent characteristics, females and more educated respondents are more likely to contribute to the environment. In terms of magnitude, having a college degree, relative to less than a high school degree has a large impact on the probability of contributing to the environment. African-Americans and Asians are less likely to contribute to the environment than whites and Latinos are less likely to contribute to the environment than non-latinos. None of the other included characteristics are found to have a statistically significant impact on the probability of contributing to the environment. One difficulty with examining the relationship between giving and these individual respondent characteristics is that environmental giving is at the household level and not all members of a household necessarily have the same characteristics. To examine whether these characteristics are actually linked with the giving, in Model 2 the probit estimation only includes respondents who are the primary decisionmakers in their household with respect to charitable giving. While in general the results are very similar, confirming most of the relationships found in Model 1, the parameter estimate on female is no longer statistically significant at the 10
7 percent level. On the other hand, married respondents are less likely to contribute to the environment (parameter estimate on married is not statistically significant in Model 1). The comparison in Model 1 is between environmental donors and all other households, where the other households include those who make no charitable contributions and those with charitable contributions, but not to the environment. These two types of households may be quite different. Model 3 estimates the same probit model, but only including households that make charitable contributions, so conditional on giving, comparing the characteristics of environmental givers with other givers. Most of the results are similar to those in Model 1, except that older respondents and larger households are less likely to contribute to the environment and the parameter estimate on female is not statistically significant. To further compare the characteristics of households with charitable giving to those who contribute to the environment in Model 4 (Table 3) a probit model is estimated with the same explanatory variables with charitable giving as the dependent variable (Donate = 1 if have charitable contributions). Similar to the characteristics of environmental givers seen in Model 1, higher income households are more likely to make donations and households with a lower tax price are more likely to make donations. The magnitude of the impact of tax price on donating overall is larger than is found for environmental giving. Females and more educated respondents are more likely to be donors and African-American and Latino respondents are less likely to be donors, again as is found with environmental donors in Model 1. However, some of the distinctions between donors and non-donors contrast to those found for environmental donors versus other households. Older and married respondents are more likely to be donors, whereas these parameter estimates are not statistically significant with respect to environmental donors in
8 Model 1 and in Model 2 among deciders only, age and being married are negatively related to environmental giving. Households with kids are less likely to make donations (controlling separately for household size). This is an interesting finding. One might think that through having children a household might have more opportunities to contribute to youth organizations. This result might have something to do with the way that fundraising for schools and children s activities occurs, instead of directly soliciting contributions they sell goods (e.g. Girl Scout cookies) which would not be counted as a charitable contribution, although households might legitimately see it as a substitute for making outright contributions. In another contrast, larger households are more likely to make donations, whereas they are less likely to make environmental donations when compared to other donors in Model 3. While Latinos are less likely to give to the environment, the parameter estimate is not statistically significant for overall donations. Homeowners are more likely to be donors, while homeownership is not statistically significant in environmental giving. The regional influences also differ between overall giving and environmental giving. While households in the South are the least likely to give to the environment, for overall giving, households in the South are indistinguishable from households in the West, while households in the East are more likely to donate than households from the West (they are indistinguishable from the West in terms of environmental giving). In Model 5, the dependent variable is whether or not a household was asked to contribute. The results in Model 5 indicate that the characteristics of households that are asked to contribute are quite similar to the characteristics of households that do contribute. Higher income households, those with lower tax prices, and homeowners are more likely to be asked and, as seen in Model 4, more likely to contribute. Female, older, married, and more educated
9 respondents are also more likely to be asked and are more likely to contribute. However, the magnitude of being a university graduate on being asked is larger than is the magnitude of being a university graduate on the likelihood of giving. Since most institutions of higher education ask their alumni for charitable contributions this may not be surprising. African-American, Asian, and Latino respondents are less likely to be asked to contribute. Particularly in the case of Latinos, this may mark a missed opportunity for organizations, as they are not less likely to give than non-latinos. This difference by race and ethnicity of households being asked to give could result from the self-reinforcing patterns of giving that was discussed earlier. However, it could also be a reason for the lower probabilities of giving among African-Americans and Asians. Households with kids are more likely to be asked to contribute, although as discussed previously, are less likely to contribute. Household size does not have a significant impact on being asked to contribute, perhaps another missed opportunity for organizations since larger households are more likely to contribute. In terms of regional differences, households in the Midwest are the most likely to be asked to give. This contrasts with the giving pattern, where households in the East are the most likely to give. In terms of giving, no statistically significant differences are found between the years 1995 and 1998, however, in 1998 households were more likely to be asked to give. Model 6 analyzes respondents confidence in environmental organizations. Comparing the results of Model 6 to the environmental giving in Model 1 leads to some interesting contrasts. First of all, higher income households have less confidence in environmental organizations although they are more likely to have environmental giving. Female respondents are more confident in environmental organizations, however, the evidence on this translating to a
10 greater likelihood of giving to the environment is mixed. In terms of education, while those with a high school degree are more confident in environmental organizations than those without a high school degree, and university graduates are more confident in environmental organizations than those with less education, the coefficient on respondents with some college or a technical or associate degree is not statistically different from those without a high school degree. This contrasts with the giving results which were increasing in education. Although this might be explained by the inclusion of technical degrees perhaps being linked to jobs where respondents would have less confidence in environmental organizations, it is interesting that it would not have the same impact on the probability of environmental giving. Older respondents are less confident in environmental organizations, consistent with the findings in Models 2 and 3 where older respondents are less likely to give to the environment. In 1998 respondents had more confidence in environmental organizations than in 1995, although there is not a significant difference in giving between the years. The ethnic and racial differences in attitudes are also interesting. African-Americans express the least confidence in environmental organizations. This suggests that differences in preferences or lack of trust in these organizations is part of the explanation for their lower environmental giving. On the other hand, Asians are more confident in environmental organizations than whites, and Latinos are more confident in environmental organizations than non-latinos, whereas both groups are less likely to contribute to environmental organizations. This suggests a different explanation for their lower giving, either primarily budget related, or related to free-riding. It is possible to interpret high confidence in environmental organizations as also meaning that they are confident in the organizations ability to obtain funding (therefore not in need of contributions). Interesting regional differences are
11 also found. Respondents in the Midwest and East have more confidence in environmental organizations than do those in the West, although their environmental giving patterns are indistinguishable. Conclusions In summary, environmental giving is found to be greater for higher income households and households facing a lower tax price, as is the case for overall charitable giving, and as would be predicted by economic theory. More educated respondents are more likely to give to the environment, while households from the South are less likely to give to the environment. African-Americans and Asians are less likely to give to the environment than whites and Latinos are less likely to give to the environment than non-latinos. Some of these differences may be exacerbated by the pattern of which households were asked to give, where African-Americans, Asians, and Latinos are less likely to have been asked to make charitable contributions. Examining the confidence in environmental organizations also suggests that the lower likelihood of Latinos to be environmental givers does not stem from this, as they are more likely to show confidence in environmental organizations than are non-latinos.
12 References Andreoni, James, Eleanor Brown, and Isaac Rischall. 2003. Charitable Giving by Married Couples: Who Decides and Why Does it Matter? Journal of Human Resources 38: 111-133. Andreoni, James and John Karl Scholz. 1998. An Econometric Analysis of Charitable Giving with Interdependence Preferences. Economic Inquiry 36: 410-428. Andreoni, James, William G. Gale, and John Karl Scholz. 1995. Charitable Contributions of Time and Money. University of Wisconsin Madison, Department of Economics Working Paper. Berrens, Robert P., Hank Jenkins-Smith, Alok K. Bohara, and Carol L. Silva. 2002. Further Investigation of Voluntary Contribution Contingent Valuation: Fair Share, Time of Contribution, and Respondent Uncertainty, Journal of Environmental Economics and Management 44: 144-168. Champ, Patricia A. and Richard C. Bishop. 2001. Donation Payment Mechanisms and Contingent Valuation: An Empirical Study of Hypothetical Bias. Environmental and Resource Economics 19: 383-402. Foster, Vivien, Ian J. Bateman, and David Harley. 1997. Real and Hypothetical Willingness to Pay for Environmental Preservation: A Non-experimental Comparison. Journal of Agricultural Economics 48: 123-138. Greene, William H. 2000. Economic Analysis, 4 th ed., Prentice Hall, New Jersey. Israel, Debra. 2007. Charitable Donations: Evidence of Demand for Environmental Protection? International Advances in Economic Research 13, forthcoming. MacMillan, Douglas C., Trevor S. Smart, and Andrew P. Thorburn. 1999. A Field Experiment Involving Cash and Hypothetical Charitable Donations. Environmental and Resource Economics 14: 399-412. Peloza, John, and Piers Steel. 2005. The Price Elasticities of Charitable Contributions: A Meta- Analysis. Journal of Public Policy and Marketing 24:260-272. Richer, Jerrell. 1995. Green Giving: An Analysis of Contributions to Major U.S. Environmental Groups. Resources for the Future Discussion Paper 95-39. Washington, DC. Tiehen, Laura. 2001. Tax Policy and Charitable Contributions of Money. National Tax Journal 54: 707-723.
13 Table 1. Variable Means, Pooled Data Variable Mean Std. Err. Donate to environmental organizations (envind) 0.114 0.005 Confidence in environmental organizations 2.329 0.014 Environmental donations ($ 1998) 14.192 2.400 Tax price ($ 1998) 0.947 0.002 Household income ($ 1998) 47,250.520 604.980 Female 0.511 0.007 Age 46.099 0.247 Married 0.582 0.007 High School degree 0.320 0.007 Some college or technical or associate degree 0.253 0.006 University graduate 0.249 0.006 Kids in household 0.441 0.007 Household size 3.090 0.025 East 0.266 0.006 South 0.322 0.007 Midwest 0.224 0.006 West 0.181 0.005 Homeowner 0.653 0.007 African-American 0.167 0.005 Asian 0.014 0.002 Latino 0.162 0.005 Retired 0.198 0.006 Student 0.018 0.002 Asked to contribute 0.630 0.007 Year 1998 0.487 0.007 Total donations ($ 1998) 724.509 36.442 Donate to charitable organizations (donate = 1) 0.690 0.007 Number of types of donations (0 to 12 categories) 1.809 0.028 N = 4,978 Source: Author s calculations from pooled 1996 and 1999 Giving and Volunteering Data
14 Table 2. Probit Estimation: Characteristics of Environmental Donors Model 1. Whole sample Model 2. Deciders only Model 3. Donors only df/dx s.e. df/dx s.e. df/dx s.e. ln tax price (1998 $) -0.1010 0.0328 *** -0.1999 0.0684 *** -0.1075 0.0507 ** ln household income (1998 $) 0.0335 0.0073 *** 0.0480 0.0153 *** 0.0447 0.0117 *** Female 1 0.0125 0.0075 * 0.0200 0.0161 0.0106 0.0118 Age -0.0005 0.0003-0.0022 0.0007 *** -0.0010 0.0006 * Married 1-0.0018 0.0092-0.0310 0.0179 * -0.0195 0.0150 High School degree 1 0.0552 0.0197 *** 0.0772 0.0401 ** 0.0768 0.0308 *** Some college or technical or associate degree 1 0.0972 0.0236 *** 0.0999 0.0437 *** 0.1234 0.0341 *** University graduate 1 0.1324 0.0263 *** 0.1278 0.0444 *** 0.1647 0.0352 *** Kids in household 1-0.0135 0.0116 0.0042 0.0240-0.0184 0.0182 Household size -0.0065 0.0043-0.0135 0.0084-0.0111 0.0066 * East 1 0.0159 0.0117 0.0015 0.0231 0.0182 0.0179 South 1-0.0244 0.0105 ** -0.0455 0.0218 ** -0.0394 0.0168 ** Midwest 1 0.0180 0.0123 0.0085 0.0246 0.0242 0.0189 Homeowner 1 0.0090 0.0100 0.0129 0.0190-0.0006 0.0165 African-American 1-0.0638 0.0083 *** -0.1009 0.0174 *** -0.0921 0.0145 *** Asian 1-0.0677 0.0100 *** -0.1014 0.0244 ** Latino 1-0.0482 0.0095 *** -0.0627 0.0216 *** -0.0770 0.0152 *** Retired 1-0.0084 0.0127 0.0400 0.0314-0.0175 0.0199 Student 1-0.0282 0.0258-0.0166 0.0688-0.0503 0.0420 Year 1998 1 0.0002 0.0079-0.0165 0.0171 0.0015 0.0125 Sample Size N=4,978 N=1,776 N=3,433 Log likelihood -1532.715-694.225-1386.371 obs. P pred. P (at means) 0.1141 0.0790 0.1599 0.1268 0.1655 0.1354 1 Indicator variables, df/dx is for discrete change from 0 to 1 ***Significant at 1% level **Significant at 5% level *Significant at 10% level Source: Author s calculations from pooled 1996 and 1999 Giving and Volunteering Data
Table 3. Probit and Regression Results: Comparisons Dependent Variable Model 4. Model 5. Donate=1 Ask=1 (Probit) (Probit) Model 6. Confidence in Environmental Organizations (Regression) Parameter s.e. estimate Variable Names df/dx s.e. df/dx s.e. ln tax price (1998 $) -0.4288 0.0669 *** -0.3360 0.0689 *** ln household income (1998 $) 0.0531 0.0107 *** 0.0620 0.0117 *** -0.0427 0.0220 ** Female 1 0.0555 0.0136 *** 0.0440 0.0144 *** 0.0792 0.0271 *** Age 0.0016 0.0006 *** 0.0022 0.0007 *** -0.0028 0.0013 ** Married 1 0.0943 0.0159 *** 0.0469 0.0168 *** -0.0427 0.0324 High School degree 1 0.0643 0.0187 *** 0.0743 0.0205 *** 0.1059 0.0451 ** Some college or technical or associate degree 1 0.1174 0.0191 *** 0.1389 0.0212 *** 0.0510 0.0480 University graduate 1 0.1379 0.0207 *** 0.1936 0.0222 *** 0.1816 0.0517 *** Kids in household 1-0.0338 0.0185 * 0.0332 0.0198 * -0.0471 0.0365 Household size 0.0114 0.0051 ** -0.0015 0.0057-0.0034 0.0101 East 1 0.0512 0.0196 *** 0.0034 0.0216 0.0850 0.0408 ** South 1 0.0072 0.0198 0.0204 0.0211 0.0039 0.0408 Midwest 1 0.0188 0.0210 0.0710 0.0216 *** 0.0951 0.0426 ** Homeowner 1 0.0711 0.0171 *** 0.0648 0.0181 *** -0.0028 0.0347 African-American 1-0.0771 0.0205 *** -0.0513 0.0214 ** -0.1092 0.0418 *** Asian 1-0.2192 0.0613 *** -0.2422 0.0612 *** 0.2122 0.1244 * Latino 1-0.0220 0.0203-0.0783 0.0221 *** 0.0788 0.0424 * Retired 1 0.0090 0.0242 0.0074 0.0258-0.0255 0.0501 Student 1 0.0213 0.0482 0.0501 0.0514 0.1094 0.1156 Year 1998 1-0.0010 0.0143 0.0427 0.0150 *** 0.1079 0.0271 *** constant 2.7345 0.2296 *** Sample Size N=4,978 N=4,978 N=4,618 Log likelihood -2,723.660-2939.677 R 2 0.0196 s.e. of regression 0.9104 obs. P pred. P (at means) 0.6896 0.7129 0.6296 0.6449 1 Indicator variables, df/dx for probit results is for discrete change from 0 to 1 ***Significant at 1% level **Significant at 5% level *Significant at 10% level Source: Author s calculations from pooled 1996 and 1999 Giving and Volunteering Data 15