The influence of values on hard issue attitudes

ABSTRACT

In light of extensive findings that Americans lack ideological structure to their attitudes, political scientists have turned to values as a source of attitudinal constraint in the mass public. However, most of this work has exclusively examined “easy” issues – those that are highly visible and familiar to voters. In this study, I conduct an original survey to investigate the extent to which values predict attitudes toward three “hard” issues: nuclear energy expansion, needle exchange programs, and laws that allow health care providers to opt out of providing medication that could terminate a pregnancy. Overall, I find that while values structure the attitudes of most individuals, those who are poorly informed about politics fail to mimic the patterns associated with the well-informed. These findings suggest that although values can be a source of attitudinal constraint in the general public, their influence is conditional upon political sophistication.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 In addition to Alvarez & Brehm’s explanation of issue difficulty and abortion laws (1995), further empirical analysis indicates that participants indeed see opt-out differently relative to abortion restrictions. My survey also included measures of participants’ self-reported understanding and certainty of the issue (ranging from 1–7), as well as whether the respondent felt they needed more information before forming an opinion (0 or 1). Participants reported that they understood abortion opt-out laws significantly less than abortion legality (p = 0.00); that they were significantly less certain about the issue (p = 0.00), and a significantly higher proportion of respondents felt they needed more information about the issue (p = 0.00).

2 I similarly find that there is greater support for equality of opportunity and less variation, compared to equality of outcome. Both values are measured on a scale ranging from 1 to 6, with higher scores representing greater support. The mean and standard deviation for equality of opportunity was 5.29 (SD = 1.03); for equality of outcome it was 4.36 (SD = 1.53).

3 This question wording is derived from the 2006 American National Election Studies pilot, which tests two types of values questions: those from the Portrait Values Questionnaire (PVQ) designed by Schwartz (2003), and those from the World Values Survey (WVS). In-depth comparisons of the two approaches have found that the PVQ is preferable to the WVS, as the binary nature of the WVS questions “do a poor job of conceptualizing the multi-faceted, variable, and generally-positive valence of human values” (Hitlin and Kramer 2007, 5). There is additionally substantial evidence speaking to the validity and reliability of the PVQ measures (see Schwartz et al. 2001).

4 In order to assess the appropriateness of combining these measures, I conducted a Principal Component factor analysis (PCA). The sophistication factor has an eigenvalue of 1.85, indicating that it explains 62% of the total variance in the set of items. The factor loadings measure the correlation of each variable with the sophistication factor. Each of the factor loadings far exceed the standard conventional minimum correlation coefficient of 0.3. The Kaiser-Meyer Olkin (KMO) measure of sampling adequacy assesses the common variance between the variables included in the PCA. The KMO value for this set of variables is .66, which exceeds the recommended value of .60 (Thompson, 2004). All of this indicates that knowledge, interest, and attention are appropriately combined into a single dimension.

5 Opposition to abortion is measured on a 4-point scale, meaning that an ordered logit model may be used to estimate results. However, comparing the results between abortion opposition (ordered logit) and support for opt-out laws (OLS) becomes far less intuitive. Thus, the results of an OLS model predicting opposition to abortion are presented here, however they are substantively identical to estimates derived from an ordered logit model.

Additional information

Funding

This work was supported by The Ohio State University.