Bringing ideology in: the conservative white male effect on worry about environmental problems in the USA

Abstract

Extending existing scholarship on the white male effect in risk perception, we examine whether conservative white males (CWMs) are less worried about the risks of environmental problems than are other adults in the US general public. We draw theoretical and analytical guidance from the identity-protective cognition thesis explaining the white male effect and from recent political psychology scholarship documenting the heightened system-justification tendencies of political conservatives. We utilize public opinion data from nine Gallup surveys between 2001 and 2010, focusing on both a single-item indicator and a composite measure of worry about environmental problems. We find that CWMs indeed have significantly lower worry about environmental problems than do other Americans. Furthermore, the results of our multivariate regression models reveal that this CWMs effect remains significant when controlling for the direct effects of political ideology, race, and gender and the effects of nine social, demographic, and temporal control variables – as well as the effect of individuals generalized (nonenvironmental) risk perceptions. We conclude that the white male effect is due largely to CWMs, and that the latters low level of concern with environmental risks is likely driven by their social commitment to prevent new environmental regulations and repeal existing ones.

Keywords:

• white male effect
• political ideology
• race
• gender
• environmental problems

Acknowledgements

The authors thank the Gallup Organization for making the data available for analysis and thank the editors and reviewers at Journal of Risk Research for their constructive feedback.

Notes

1. A recent study finds no such WM effect in Sweden, though the authors argue that this is likely due to the relative equality between men and women in that country (Olofsson and Rashid 2011).

2. According to the vulnerability thesis, WMs feel less vulnerable to many risks than do females and nonwhites because of their dominant position in the social structure; they are, therefore, more accepting of such risks. Marshall et al. (2006, 35) point out that early works promoting the vulnerability thesis (Finucane et al. 2000; Kalof et al. 2002) used race-gender group status as a surrogate measure of vulnerability. That is, these studies used the sociodemographic characteristic of WM status as a proxy for individuals social structural position in society. Satterfield, Mertz, and Slovic (2004) were the first to explicitly test the vulnerability thesis, employing a social vulnerability score – the mean of six discrimination items – in their analyses. They found some support for the hypothesis, as the independent effects of gender and race were partly but not fully explained by their vulnerability composite measure. Gender remains a robust predictor of risk as does – to a lesser extent – race, but the influence of these two demographic variables (particularly race) is explained in part by our measures of vulnerability and environmental injustice (Satterfield, Mertz, and Slovic 2004, 128). Marshall (2004) found a similar result when including an internal efficacy measure in his analytical models. This measure of vulnerability was the strongest predictor of environmental risk perceptions among Louisiana residents; yet, even when controlling for this variable, his white male dummy variable still had a statistically significant effect. In characterizing the generally supportive evidence for the vulnerability thesis, Marshall et al. (2006, 36) perhaps put it the best: race-gender groups appear to serve as a significant but incomplete measures of vulnerability.

3. The cultural worldviews thesis has not yet been tested directly in studies of the WM effect. Nevertheless, scholars have documented the distinct political and cultural affinities of risk-accepting WMs in ways that align with cultural theory of risk (Douglas and Wildavsky 1982; Thompson, Ellis, and Wildavsky 1990; Wildavsky and Dake 1990). For instance, risk-accepting WMs are more likely than are others to have an individualist worldview (Finucane et al. 2000; Kahan et al. 2007; Palmer 2003), be antiegalitarian (Finucane et al. 2000; Flynn, Slovic, and Mertz 1994; Palmer 2003; Satterfield, Mertz, and Slovic 2004), and favor hierarchy (Flynn, Slovic, and Mertz 1994; Kahan et al. 2007). Furthermore, Kahan et al. (2007) integrate cultural theory into their identity-protective cognition thesis, which we are employing.

4. Not all of the environmental worry items we use were asked every year. Thus, when we use pooled samples in our analyses, the Ns vary across the dependent variables. Also, as typical for most national surveys, the Gallup Organization employs weighting procedures to ensure that its samples are representative of the US adult population. We report the results of multivariate models with unweighted data here.

5. In past studies documenting the WM effect, scholars operationally define risk perception in a variety of related ways. Sometimes respondents are asked to rate various hazards in terms of the risks they pose (e.g. almost no risk, slight risk, moderate risk, or high risk) either to themselves and their families (Finucane et al. 2000) or to the general public (Flynn, Slovic, and Mertz 1994; Satterfield, Mertz, and Slovic 2004). In other studies, respondents are asked to rate how risky different hazards or activities are on a numerical scale (Palmer 2003). In still other studies, respondents are asked to evaluate whether various issues or problems are not a problem, a minor problem, a moderate problem, or a serious problem (Marshall 2004; Marshall et al. 2006). We believe that our single-item indicator of worry about the quality of the environmental, our index of worry about environmental problems, and our index of generalized worry – which assess how much respondents worry about the respective problems – are reasonable proxy measures for risk perception as it has been operationally defined in the WM effect literature. They most closely follow how Marshall (2004) and colleagues (Marshall et al. 2006) operationalize risk perception. Although our worry indicators are arguably broader in scope than the more focused items that ask specifically about risks, used in several WM effect studies, using the former enables us to demonstrate that the CW effect – and now the CWM effect, in our terms – is not limited to narrow notions of risk but extend to general worry about the state of the environment and other societal problems.

6. We also tested for a WM effect by using a WM (rather than CWM) dummy variable in our expanded model for both dependent variables. We find no statistically significant WM effect on the worry about environment problems index (p = .10) and only a marginally significant WM effect on worry about quality of the environment (p = .05). This lack of strong evidence for a WM effect when controlling for political ideology suggests that focusing more specifically on the atypical environmental worry of CWMs is more efficacious than focusing on those of WMs.

7. We also regressed our measure of generalized worry on all of the predictors in the CWM model. The results provide evidence that our measure of generalized worry has considerable construct validity. Briefly, nonwhites, women, adults with lesser education and lower income, individuals who identify strongly with the environmental movement, and Democrats report higher levels of generalized, nonenvironmental worry than do their respective counterparts. Political ideology, age, full-time employment, parenthood, and religiosity have no statistically significant effect. Generalized worry perception scores did increase over the time period. Most notably, we find no CWM effect on generalized worry. CWMs do not worry less than do other adults about a range of nonenvironmental problems. This supports our argument that the CWM effect is most applicable to those types of risks, particularly environmental problems, which require governmental regulations for their amelioration.