Abstract
I next investigate a possible reason that environmentalists (or at least U.S. environmentalists) currently lack interest in the increasing size of the human population and its impact on nature. Specifically, I explore whether, at the present time, Americans generally exhibit little or no sensitivity to the effect that population growth has on the ecosystem of the planet. If this is the case, environmentalists will simply be manifesting the attitudes of the public as a whole. A study that used data gathered in 1969 from a nationwide sample of U.S. adults found that attitudes toward the environment were no more than modestly related to attitudes toward population increase.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Larry D. Barnett, Concern with Environmental Deterioration and Attitudes toward Population Limitation, 20 bioscience 999, 1000 tbl. 1 (1970). Cramer’s V, the measure of association, was 0.131 among all respondents. For an explanation of Cramer’s V, see Alan C. Acock & Gordon R. Stavig, A Measure of Association for Nonparametric Statistics, 57 soc. forces 1381 (1979).
- 2.
Robert L. Fischman & Lydia Barbarsh-Riley, Empirical Environmental Scholarship, 44 ecology l.q. 767, 768, 806 (2018).
- 3.
nat’l opinion res. ctr., general social surveys, 1972-2018: cumulative codebook viii, 3171–72 (Dec. 2019) [hereinafter cumulative codebook], https://gss.norc.org/get-documentation. The GSS was first conducted in 1972. Id. at viii. Prior to 2006, GSS samples were limited to persons who spoke English; in 2006 and later years, GSS samples included persons who spoke English and persons who spoke only Spanish. Id. at 3171. An overview of the General Social Survey is in tom w. smith, the general social surveys (GSS Project Report No. 32, 2016), https://gss.norc.org/get-documentation/project-reports.
- 4.
GSS questionnaires are available at https://gss.norc.org/get-documentation/questionnaires. For the two statements in the 2000 GSS questionnaire, see Nat’l Opinion Res. Ctr., Self-Administered Questionnaire: General Social Survey ([2000]), at 4, 6 [hereinafter 2000 Questionnaire], https://gss.norc.org/get-documentation/questionnaires (under “2000 Questionnaires,” select hyperlink “2000 GSS SAQ ISSP”). For the two statements in the 2010 GSS questionnaire, see Nat’l Opinion Res. Ctr., English Questionnaires: Cross-Section 2010 Version 1, at 144, 146 [hereinafter 2010 Questionnaire], https://gss.norc.org/get-documentation/questionnaires (open dropdown menu “2010 Questionnaires;” under “English Questionnaires – Cross-Section,” select hyperlink “2010 Version 1”).
- 5.
A “dummy variable” is explained in melissa a. hardy, regression with dummy variables 7–8 (1993).
- 6.
The mnemonic label that I assigned a variable is not necessarily the mnemonic label assigned by the GSS.
- 7.
2000 Questionnaire, supra note 4, at 4; 2010 Questionnaire, supra note 4, at 144. Although for measuring attitudes single-statement instruments are generally less reliable and valid than multi-statement instruments, single-statement instruments are sufficiently reliable and valid to yield acceptable estimates. Marko Sarstedt & Petra Wilczynski, More for Less? A Comparison of Single-Item and Multi-Item Measures, 69 dbw 211, 218–19, 223–24 (2009).
For a five-statement instrument gauging attitudes toward population growth, see Larry D. Barnett, Women’s Attitudes toward Family Life and U.S. Population Growth, 12 pac. sociol. rev. 95 (1969). This five-statement instrument has been criticized, but the criticism stemmed from a subjective assessment and facial examination of the instrument. Richard R. Clayton, Guttman Scaling: An Error Paradigm, 16 pac. sociol. rev. 5, 8, 10 (1973). Importantly, the critique did not include any original quantitative data suggesting that the instrument was faulty, and it ignored the quantitative evidence that had been presented backing the conclusion that the five statements in the instrument tap the same attitude. The evidence, which came from two different samples, is in: Barnett, supra, at 96 & n.3; Larry D. Barnett, U.S. Population Growth as an Abstractly-Perceived Problem, 7 demography 53, 56 (1970). Nonetheless, revision of one or more of the five statements in the instrument may be warranted today in order to take account of the change in social context that has occurred during the last half-century. See generally earl babbie, survey research methods 167–70 (2d ed. 1990), which explains the concept (Guttman Scaling) that underlies the evidence supporting the measurement integrity of the five-statement instrument.
- 8.
On the popprob statement, no interviewees were recorded as having selected the “Can’t Choose” alternative, but some interviewees were recorded as “Don’t Know.” I assumed that interviewees who selected the “Can’t Choose” alternative were recorded as “Don’t Know.”
- 9.
fred c. pampel, logistic regression (2000); Francis L. Huang & Tonya R. Moon, What Are the Odds of That? A Primer on Understanding Logistic Regression, 57 gifted child q. 197 (2013).
To analyze the data, I used Stata™ IC version 12.1. The logistic command was employed for the estimation of odds ratios, and the logit command was used for the estimation of regression coefficients. statacorp, stata base reference manual: release 12, at 932, 970 (2011).
Stata automatically tests the degree of collinearity of every independent variable in a regression model and removes any independent variable whose collinearity is excessive. statacorp, supra, at 975. When testing the models in Table 2.1, Stata did not detect inordinate collinearity and thus kept all of the independent variables. To buttress the decision by Stata not to discard an independent variable, I used the Stata command vif, uncentered to estimate the Variance Inflation Factor (VIF) score of each independent variable in every model; the highest VIF score (9.0) was below the score (10.0) at which undue collinearity may exist. Further information on collinearity and the Variance Inflation Score is in larry d. barnett, societal agents in law: quantitative research 15 (2019).
- 10.
On the envimp statement, no interviewees were recorded as having selected the “Can’t Choose” alternative, but some interviewees were recorded as “Don’t Know.” I assumed that interviewees who selected the “Can’t Choose” alternative were classified as “Don’t Know.”
- 11.
Income was omitted as a control variable. The GSS in 2000 and in 2010 included two questions on yearly income (income and rincome) that employed the same income categories, but the highest category was “$25,000 or over.” cumulative codebook, supra note 3, at 216–17. Without categories for yearly incomes above $25,000, regression models using GSS data on income were thought to have an unacceptable likelihood of yielding incorrect conclusions. Educational attainment was included as a control variable, however, and is strongly correlated with income. jennifer cheeseman day & eric c. newburger, u.s. census bureau, the big payoff: educational attainment and synthetic estimates of work-life earnings (2002).
- 12.
cumulative codebook, supra note 3, at 2976, 3244–45, 3420–21.
- 13.
larry d. barnett, societal agents in law: a macrosociological approach 91–98 (2019) [hereinafter sail vol. 1].
- 14.
Paul DiMaggio, Culture and Cognition, 23 ann. rev. sociol. 253, 282 (1997).
- 15.
sail vol. 1, supra note 13, at 98.
- 16.
irwin s. kirsch et al., nat’l ctr. for educ. stat., adult literacy in america xviii–xix, 3–6, 8–11, 30–34, 38–39, 46–47, 60–61 (3d ed. 2002), available at nces.ed.gov/pubs93/93275.pdf. See Roberto M. De Anda & Pedro M. Hernandez, Literacy Skills and Earnings: Race and Gender Differences, 34 rev. black pol. econ. 231, 235, 240 (2007) (finding that the effect of literacy-skill level on the amount of earned income differs by race and by sex).
- 17.
Emily Greenman & Yi Xie, Double Jeopardy – The Interaction of Gender and Race on Earnings in the United States, 86 soc. forces 1217, 1226, 1236 (2008); Andrea E. Willson, Race and Women’s Income Trajectories: Employment, Marriage, and Income Security Over the Life Course, 50 soc. probs. 87, 91–92, 104 (2003).
- 18.
Taciano L. Milfont et al., Values Stability and Change in Adulthood: A 3-Year Longitudinal Study of Rank-Order Stability and Mean-Level Differences, 42 personality & soc. psychol. bull. 572, 575, 586 (2016); Seth Ovadia, Race, Class, and Gender Differences in High School Seniors’ Values: Applying Intersection Theory in Empirical Analysis, 82 soc. sci. q. 340, 345, 349 tbl. 2, 350 tbl. 3, 353 tbl. 4 (2001).
- 19.
sail vol. 1, supra note 13, at 105.
- 20.
Claude S. Fischer, Toward a Subcultural Theory of Urbanism, 80 am. j. sociol. 1319, 1324–30 (1975); Claude S. Fischer, The Subcultural Theory of Urbanism: A Twentieth-Year Assessment, 101 am. j. sociol. 543, 544–46 (1995). See, in addition, the text that accompanies supra note 15.
- 21.
In the GSS, the variable region reports the geographic division in which each respondent resided. cumulative codebook, supra note 3, at 233. To create the dummy variable for culture, I combined the geographic divisions in each of the four regions. For the states that form each region, see u.s. bureau of the census, geographic areas reference manual. ch. 6: statistical groupings of states and counties, at 2 fig. 6-1, 24 tbl. 6-4 (1994), available at https://www.census.gov/geo/reference/garm.html (follow “Chapter 6” hyperlink).
- 22.
sail vol. 1, supra note 13, at 98.
- 23.
Id.
- 24.
Respondents who were 89 years of age or older were coded as 89. cumulative codebook, supra note 3, at 3241, 3284.
- 25.
See id. at 198 for the method used in GSS interviews to ascertain the race of respondents. See generally Aliya Saperstein & Andrew M. Penner, Racial Fluidity and Inequality in the United States, 118 am. j. sociol. 676, 687–88, 696–98 & tbls. 2 & 3, 707–08 (2012) (reporting race-specific rates of consistency and inconsistency in the race categorizations of individuals in a large sample of U.S. residents who were studied over a period of twenty-three years).
- 26.
other was not the reference category for two reasons—(i) being a heterogeneous group, respondents of Other races provide an ambiguous referent; and (ii) being a numerically small group, respondents of Other races may produce unreliable regression coefficients and odds ratios for the relationship of other to the dependent variable. hardy, supra note 5, at 10. As to (ii), among interviewees of Other races, a total of just 77 respondents in the 2000 GSS were coded 0 or 1 on both envimp and popprob, and hence were in the data analysis; the corresponding total in the 2010 GSS was 128 respondents. Particularly in 2000, the regression results for other could have been markedly affected by a small change in the total number of respondents of Other races and/or by a small change in the distribution of respondents of Other races on envimp or on popprob.
In the two GSS waves, a substantially larger number of Blacks than of Other races were coded 0 or 1 on both envimp and popprob. The total number of Blacks so coded was 179 in the 2000 GSS and 228 in the 2010 GSS.
- 27.
At the college level, the number of years of schooling was based on courses taken for credit at other than a “business college, technical or vocational school.” cumulative codebook, supra note 3, at 3242.
- 28.
The data are from the GSS variable size. cumulative codebook, supra note 3, at 235.
- 29.
T. P. Hutchinson, Beyond Interaction: Theories, 40 quality & quantity 869, 870 (2006); james jaccard, interaction effects in logistic regression 12–13 (2001); F. David Schoorman et al., The Role of Theory in Testing Hypothesized Interactions: An Example from the Research on Escalation of Commitment, 21 j. applied soc. psychol. 1338, 1338, 1349, 1353 (1991).
- 30.
Text accompanying supra notes 16–18.
- 31.
See Marcia L. McCormick, Stereotypes as Channels and the Social Model of Discrimination, 36 st. louis u. pub. l. rev. 19 (2017) (reviewing empirical research on the nature, operation, and consequences of stereotypes, especially sex- and race-based stereotypes).
- 32.
jaccard, supra note 29, at 14.
- 33.
Text accompanying supra note 26.
- 34.
jaccard, supra note 29, at 18–20.
- 35.
In the instant study, the standard error is the standard deviation of the “sampling distribution” for the regression coefficient, i.e., the standard deviation of the dispersion that would have been obtained for the regression coefficient across numerous probability samples of adults who resided in the United States (the universe being studied). The sampling distribution assumes that, in the universe, the regression coefficient for a given independent variable is zero and that the independent variable is thus unrelated to the dependent variable. This assumption is commonly known as the null hypothesis.
The sampling distribution supplies the basis for ascertaining the probability that the numerical value of a particular regression coefficient is due solely to chance when the null hypothesis is correct, i.e., when the regression coefficient is zero in the universe. The probability, which is the level of statistical significance for the regression coefficient, represents the likelihood of making a mistake by rejecting the null hypothesis. The highest generally accepted probability is 10% (the .10 significance level).
- 36.
GSS interviewees reside in housing units that are selected using multi-stage cluster sampling. cumulative codebook, supra note 3, at 3171–72. Statistics software, when applied to data obtained through cluster sampling, produces an artificially low estimate of the standard error for a regression coefficient. Lin Wang & Xitao Fan, The Effect of Cluster Sampling Design in Survey Research on the Standard Error Statistic (paper presented at the 1997 meeting of the American Educational Research Ass’n), available at https://eric.ed.gov/?id=ED409320. However, while the need to correct standard errors arising from cluster sampling is widely acknowledged, every procedure for making the correction (including the procedure employed here) has disadvantages. Andrew Gelman, Struggles with Survey Weighting and Regression Modeling, 22 stat. sci. 153, 163–64 (2007).
For a brief explanation of cluster sampling as a form of probability sampling, see babbie, supra note 7, at 87–91.
- 37.
Larry D. Barnett, Mutual Funds, Hedge Funds, and the Public-Private Dichotomy in a Macrosociological Framework for Law 33-34 (CIRSDIG Working Paper No. 34, 2008), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1319782.
- 38.
The interaction variable was created by multiplying, for each respondent, the numerical code assigned for envimp and the numerical code assigned for year. jaccard, supra note 29, at 14.
- 39.
Maarten L. Buis, Stata Tip 87: Interpretation of Interactions in Nonlinear Models, 10 stata j. 305 (2010). The regression coefficient for the interaction variable (envyear) had a z-score (calculated with its adjusted standard error) of 0.70; the significance level of the regression coefficient, therefore, was 0.49. Since every respondent in this model had a Cook’s Statistic below 0.11, none of the respondents appeared to be an influential outlier. david w. hosmer & stanley lemeshow, applied logistic regression 173, 180 (2d ed. 2000); J. scott long & jeremy freese, regression models for categorical dependent variables using stata 151 (2d ed. 2006).
- 40.
In Model I, Cook’s Statistic for every respondent was negligible (less than 0.10), indicating that no influential outlier was present.
- 41.
See Long & Freese, supra note 39, at 178, 180 (discussing odds ratios (signified by eˆb) and percentage change in odds (signified by %)).
- 42.
The odds that an individual has a particular attribute is the ratio of (i) the probability that the attribute is present (i.e., the proportion of individuals who possess the attribute) to (ii) the probability that the attribute is absent (the proportion of individuals who do not possess the attribute). The attribute here is the belief that population growth is a problem.
- 43.
The accuracy rate of predictions as a measure of model fit is discussed in Long & Freese, supra note 39, at 104, 110–12. The rate is obtained with the estat classification command in Stata. statacorp, supra note 9, at 957.
- 44.
The regression with the expanded set of independent variables omitted envwhite because white was the reference for race. See text accompanying supra note 26.
- 45.
The accuracy rate for the expanded group of variables was 57.9%. As seen in the bottom row of the table, the accuracy rate for Model I was 58.2%.
- 46.
Jaccard, supra note 29, at 18–23.
- 47.
Sumi Cho et al., Toward a Field of Intersectionality Studies: Theory, Applications, and Praxis, 38 signs 785, 787 (2013). In addition, see Kimberle Crenshaw, Demarginalizing the Intersection of Race and Sex: A Black Feminist Critique of Antidiscrimination Doctrine, Feminist Theory and Antiracist Politics, 1989 u. chi. legal f. 139.
- 48.
Cho et al., supra note 47, at 787, 797; Mike C. Parent et al., Approaches to Research on Intersectionality: Perspectives on Gender, LGBT, and Racial/Ethnic Identities, 68 sex roles 639, 640 (2013).
As originally formulated, intersectionality theory focused on social conflict and included a political agenda. Cho et al., supra note 47, at 797, 800. However, intersectionality theory is likely to be more productive as a heuristic tool in empirical social science research when it does not emphasize social conflict and does not have a political agenda than when it does. Notably, a longstanding theory in sociology—structural functionalism—(i) posits that social equilibrium normally takes priority over social conflict in a society and (ii) omits a political agenda. Part 3.2 in infra Chap. 3.
- 49.
Liam Kofi Bright et al., Causally Interpreting Intersectionality Theory, 83 phil. sci. 60, 62 (2016).
- 50.
Id. at 63.
- 51.
Amy C. Steinbugler et al., Gender, Race, and Affirmative Action: Operationalizing Intersectionality in Survey Research, 20 gender & soc’y 805, 807–08 (2006).
- 52.
Kerryn E. Bell, Young Adult Offending: Intersectionality of Gender and Race, 21 critical criminology 103 (2013) (studying whether members of the birth cohort born in 1958 in a major U.S. city had crime-related police contacts in that city or in the metropolitan area where that city is located; using, for each of its six dependent variables, the frequency with which members of this cohort had crime-related police contacts as juveniles and as young adults through age 26; finding that White females, Black females, and Black males generally differed from White males on the dependent variables; and concluding that sex and race interact in determining the incidence of crime-related police contacts by juveniles and young adults); Huoying Wu, Can the Human Capital Approach Explain Life-Cycle Wage Differentials Between Races and Sexes? 45 econ. inquiry 24, 34 tbl. 5, 35–36 (2007) (using data from a national probability sample of U.S. residents who were 14–21 years old when initially interviewed in 1979 and who were re-interviewed in 1994; studying the interviewees from age 20–21 to age 31–33; finding differences during the course of aging between White men, White women, Black men, and Black women on several economic measures, viz., mean number of weeks employed and unemployed, mean number of weeks out of the labor force, and mean number of hours worked weekly). Accord, Lauren B. Edelman et al., Legal Discrimination: Empirical Sociolegal and Critical Race Perspectives on Antidiscrimination Law, 12 ann. rev. l. & soc. sci. 395, 401–02 (2016). See also Glenn Firebaugh et al., Why Lifespans Are More Variable Among Blacks Than Among Whites in the United States, 51 demography 2025, 2027, 2032–33, 2037–38 & fig. 7, 2041 (2014) (using data on the resident population of the United States; finding that age at death is more variable among Blacks than among Whites who die of the same cause; and concluding that this variability is mainly due to age at death among females and secondarily due to age at death among males).
- 53.
Regressions were not done for respondents whose race was Other. One reason for this omission is that, in Model 1, the odds ratio for other was not statistically significant. In addition, see supra note 26.
- 54.
The results for Model V (Black men) omit three Black men who were considered to be influential outliers. See infra Appendix to this chapter.
- 55.
Research on the relationship in Western Europe is needed given the ecological footprint of that region (supra Figure 1.3 and accompanying text). However, none of the eight rounds of the European Social Survey has obtained information that allows a quantitative test of whether, among residents of participating countries, concern with the environment is related to concern with population growth. The questionnaires for these rounds are available at European Social Survey, Data and Documentation by Round/Year, https://www.europeansocialsurvey.org/data/round-index.html.
- 56.
Respondents were dropped only when estimating Model V. The data for Model II, Model III, and Model IV did not appear to contain an influential outlier. The maximum Cook’s Statistic for a respondent was 0.18 in Model II, 0.18 in Model III, and 0.20 in Model IV. As to Model I, see supra note 40.
Each of the three Black male respondents omitted from the data for Model V had a Cook’s Statistic of 0.94 while all of the remaining Black male respondents in the data had a Cook’s Statistic under 0.47. The three omitted respondents, therefore, met one of the criteria recommended for deciding whether a case should be examined to ascertain if it is an influential outlier — the Cook’s Statistic for each of the three omitted respondents was much larger than the Cook’s Statistic for other respondents. hosmer & lemeshow, supra note 39, at 180; Long & Freese, supra note 39, at 151.
- 57.
The three respondents reported their age as 69 and their years of formal schooling as ten or less. Two of the three respondents reported that they had no more than five years of formal schooling.
- 58.
The three respondents were located in the South. However, among Black men in all geographic regions who were interviewed in 2010, who in that year were aged 65 or older, and who had less than twelve years of formal education, the three respondents were the only interviewees to be in the category coded 0 on envimp and in the category coded 1 on popprob.
- 59.
With the three respondents in the data, the z-score of the regression coefficient for envimp among Black men (n = 137), computed with the adjusted standard error, was 1.77 (regression coefficient = 0.807, odds ratio = 2.242). The significance level of a z-score = 1.77 is .08 and hence ≤ .10. Without the three respondents in the data, the z-score of the regression coefficient for envimp among Black men (n = 134), computed with the adjusted standard error, was 1.98 (regression coefficient = 0.922, odds ratio = 2.516); the significance level of a z-score = 1.98 is ≤ .05.
Author information
Authors and Affiliations
Corresponding author
Appendix: Model V
Appendix: Model V
Each of the three Black respondents who were deemed to be influential outliers and hence excluded from the results for Model V had a Cook’s Statistic that, although less than 1.00, was relatively large.Footnote 56 Notably, all three respondents were elderly and had little formal schooling,Footnote 57 characteristics that could have reduced their ability to understand, and hence could have affected their responses to, the envimp statement (on which they were in the category coded 0) and the popprob statement (on which they were in the category coded 1). In addition, the three respondents were in the same GSS sample (viz., 2010), and no other Black male in that sample who was of similar age and educational attainment answered envimp and popprob in the way that the three respondents did, i.e., indicated that they were unconcerned with the environment but were concerned with continued population growth.Footnote 58 Because the three interviewees were unique in the 2010 sample and had a Cook’s Statistic that was twice the size of the Cook’s Statistic for any other Black male, they were treated as influential outliers and were excluded from the data used to estimate the odds ratios in Model V.Footnote 59
Rights and permissions
Copyright information
© 2021 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this chapter
Cite this chapter
Barnett, L.D. (2021). An Empirical Study of Americans’ Attitudes. In: Demography and the Anthropocene. SpringerBriefs in Environment, Security, Development and Peace, vol 35. Springer, Cham. https://doi.org/10.1007/978-3-030-69428-9_2
Download citation
DOI: https://doi.org/10.1007/978-3-030-69428-9_2
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-69427-2
Online ISBN: 978-3-030-69428-9
eBook Packages: Law and CriminologyLaw and Criminology (R0)