Discrete and continuous measures of consequentiality

ABSTRACT

A respondent finds a survey consequential if they believe their answer could influence the policy being addressed in the survey and if they believe that they will have to pay for the policy if implemented. Given these criteria, the literature has followed two paths to analyse consequentiality. The first uses a discrete method that separates respondents into consequential and inconsequential groups. The second interprets beliefs about consequentiality as continuous. We compare these approaches to identify their strengths and weaknesses. Using the discrete approach, we classify respondents into groups based on whether their responses satisfy various consequentiality criteria. We find that respondents in the inconsequential group have a willingness to pay that is insignificantly different from zero. For those in the consequential group, willingness to pay is positive and depends on the scope of the project. Treating consequentiality as continuous and using the hybrid choice model, we find that individuals who believe their responses will influence policy, policy consequentiality, and those who are concerned about the amenity are more likely to be in favour of the policy. Lastly, income is positively related to payment consequentiality.

KEYWORDS:

• Consequentiality
• stormwater management
• stated preferences
• hybrid choice models
• generalized structural equation method

Acknowledgements

The authors thank an anonymous journal referee for comments that have led to improvements in this paper.

Disclosure statement

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

Notes

1 This article is based on earlier research by Carson et al. (2002).

2 This proposition is in the context of a single binary choice so strategic preferences do not come into play.

3 The full survey can be viewed at: https://bit.ly/osf-survey.

4 These ‘opt-in’ panels are becoming popular in social science research due to their relatively low cost and ability to quickly collect a large amount of data. Opt-in samples are useful for exploratory research, such as here, but use of these estimates for policy analysis should be done with caution (Baker et al. 2010; Yeager et al. 2011; Lindhjem, Henrik, and Navrud 2011).

5 Conditional on having confidence in the county government, 78% think the survey will have an effect.

6 The values in Table 7 do not include sample weights. The overall value of alpha drops to 0.67 with sample weights. Standardizing the variables to have zero mean and unit standard deviation raises alpha to 0.70.

7 The null hypothesis for Bartlett's test is that the variables are uncorrelated. Rejecting the null hypothesis is evidence of partial correlation.

8 In structural equation models ovals are used to indicate latent variables and rectangles to indicate exogenous variable.

9 The order for this analysis is: definitely no, probably no, don't know, probably yes, definitely yes.

10 Kabaya (2020) controls for both policy and payment consequentiality in a multivariate probit model and finds opposing impacts on the likelihood of voting in favor of the policy.

Additional information

Funding

The authors would like to thank Appalachian State University for funding the Water Cluster: Integrating Geology, Economics and Policy Analysis through the CONducting ComplEx Research Together (CONCERT) Grants.