Does the solemn oath lower WTP responses in a discrete choice experiment application to solar energy?

ABSTRACT

A recently offered alternative to eliminating or mitigating hypothetical bias associated with stated preference surveys is the solemn oath script. While the efficacy of solemn oath script is still debatable, the objective of this analysis is to provide an initial field setting test of the solemn oath script to a particular discrete choice experiment survey application to solar energy. We conducted a discrete choice experiment survey with two treatment groups: with and without having respondents sign the solemn oath prior to taking the survey. Utilizing random parameter logit models in both preference-space and willingness to pay (WTP)-space, results provide no evidence that the solemn oath script lowers respondents’ WTP for the good in question. Either there is no hypothetical bias in this solar energy case study, which we are unable to test as there is no real expenditure at issue, or the solemn oath script may have limited application outside of the experimental lab and is not effective under every condition. Lastly, this calls for more research on the efficacy of a solemn oath script.

KEYWORDS:

• Solemn oath script
• hypothetical bias
• choice experiment survey
• solar energy
• marginal willingness-to-Pay

JEL CLASSIFICATION:

• C81
• Q42
• D12

Acknowledgments

We greatly appreciate valuable comments received from Dr. Ronald Cummings. This paper would not exist without his comments. Jamal Mamkhezri acknowledges funding received from the University of New Mexico Center for Regional Studies.

Disclosure statement

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

Notes

1 There are five different formats that researchers ask CVM questions: (1) open-ended, (2) dichotomous choice, (3) referendum decision, (4) payment cards, and (5) iterative biddings. For a brief definition on each format, for example see (Thacher et al. 2011, 21).

2 For a recent meta-analysis on various approaches utilized to address the hypothetical bias issue, see Penn and Hu (2018).

3 The valuation method, the characteristics of the good in question, the length of the cheap talk script, or a combination of these can affect the efficacy of the cheap talk script.

4 For more details on these two types of cheap talk script, see, e.g., Carson and Groves (2011).

5 For example, de-Magistris, Gracia, and Nayga (2013) prime respondents honesty by providing cues and words related to honesty prior to participating in the DCE question in hope that ‘priming can unconsciously influence peoples’ perception, evaluations, behaviour and choice’. de-Magistris et al. demonstrate that this honesty priming task, which was completed 24 times by the subjects, mitigate their respondents WTP. In contrast, Howard et al. (2017) utilizes a similar honesty priming approach coupled with the cheap talk script and finds that the honesty prime approach does not reduce hypothetical bias while the cheap talk script does.

6 Carlsson et al. use the word ‘promise’ instead of ‘swear’ as swearing upon one’s honour is not customary in neither Swedish nor Chinese courtrooms. Further, they also used a cheap talk script throughout the survey, regardless of treatment groups.

7 The authors examined the effect of their weak oath on cultural contexts by interacting oath variable with respondents’ socioeconomic characteristics.

8 Note that in their poster that was presented in 2012 (de-Magistris and Pascucci 2012), the same authors state that the solemn oath script had no effect on hypothetical bias.

9 Immediately prior to the DCE questions, the authors asked, ‘Do you feel you can promise us you will answer the questions that will follow truthfully?’ Participants were then given a dichotomous option to choose from: ‘Yes, I promise to answer the questions in the survey truthfully’ and ‘No, I cannot promise’.

10 RPSs are state-mandated policies that mandate electric utility companies to generate a portion of their electricity from renewables by a certain timeframe.

11 Based on New Mexico’s RPS, rural electric cooperatives are required to source half of what major utility companies are required.

12 Further details on Senate Bill 489 can be found at: https://www.nmlegis.gov/Sessions/19%20Regular/bills/senate/SB0489.pdf (accessed 04.10.2019).

13 It is not mentioned as to whether these carve-out percentages will uphold under the new RPS (Senate Bill 489).

14 Source: Albuquerque Economic Development. https://www.abq.org/low-risk-location.aspx (accessed 5.31.18).

15 Source: Nebraska Department of Environment and Energy, 2010. ‘Comparison of Solar Power Potential’ by http://www.neo.ne.gov/statshtml/201.htm (accessed 5.26.18).

16 For details on the 210-Bill, see https://www.nmlegis.gov/Sessions/19%20Regular/bills/house/HB0210.pdf (accessed 03.06.2019).

17 Although we did not utilize any type of cheap talk script in our survey design, we included the idea of soft cheap talk script in which we reminded respondents about their budget constraint in the survey. The statement we included: ‘We ask you to pick the plan that you think is best, giving serious consideration to the associated costs; in other words, assume you are paying the mentioned amount. Choosing a plan implies you are willing to bear the specified additional cost on your monthly electricity bill’.

18 The text reads ‘State policymakers want to know Public Service Company of New Mexico customer opinions. What share of electricity should come from renewable sources and what role should solar energy play?’ We then stated that ‘Decisions about Public Service Company of New Mexico’s solar energy future could affect your electricity bill’ and that we will ‘have their opinions heard by state policymakers’.

19 Throughout the survey we stayed neutral and unbiased (energy-wise). For example, we included the statement ‘Some people find rooftop solar unattractive’ when we introduced rooftop solar. Similarly, we mentioned externalities that are associated with utility-scale solar when we defined them, ‘Some people find solar farms unattractive and believe they change the landscape. Birds can crash into the panels on solar farms, thinking that they’re water bodies. Solar farms interrupt deer migratory paths’.

20 Including the business-as-usual alternative violates the necessary binary condition for incentive compatibility (Carson and Groves 2007).

21 Our data is different than the previous study (Mamkhezri, Thacher, and Chermak 2020). We collected 78 more responses (34 and 44 with and without the solemn oath) for the current study.

22 Based on the result of Model 2 of Mamkhezri, Thacher, and Chermak (2020): negative when RPS, Rooftop, and Smart Meter, and positive when No Credit Banking, Water, and ASC.

23 We also included a log-normally distributed Price in the preference-space model. Results were comparable to when Price is normally distributed, and the model with log-normal Price variable did not lead to statistically significantly better fit.

24 We used mixlogit(),wtp(), and mixlogitwtp().

25 For comparison and robustness checks, we also include results from Multinomial Logit (MNL), RPL with lognormal Price, and Generalized Multinomial Logit (GMNL) models in the Appendix. Results are comparable across various models in terms of significance level and sign.

26 MNL, RPL with lognormal Price, and GMNL models’ results are included in the appendix for comparison and robustness check. These models’ results are comparable to those of the current study’s in terms of significance level and sign.

27 SmartMeter_home in WTP-space and (RPS-20)*(Rooftop-9)*Oath in preference-space variables are marginally significant at 10% level (p-values are 0.103 and 0.10 respectively).

28 The only interacted variable with Oath that results in statistically significant MWTP in WTP-space is (RPS-20)*(Rooftop-9)*Oath. However, it has the opposite sign of what we expected: A negative sign on this variable indicates that respondents who took the solemn oath support higher levels of RPS to source from solar farm rather than rooftop solar even more than those who did not sign the oath script. This variable is statistically significantly different from zero (H0: MWTP = 0: chi2 = 3.71 and p-value = 0.054).

29 We ran models 3 and 4 on the dataset used in Mamkhezri, Thacher, and Chermak (2020) (prior to collecting 78 more responses) and arrive at similar results: beside, Rooftop*Oath, (RPS-20)*(Rooftop-9)*Oath, and NoCreditBanking*solemn (only in WTP-space) variables, none of the other interacted variables with the Oath variable were statistically significant (see Table A3 of Appendix). Further, all three variables have the opposite sign (opposite to assigned signs in Table 4) indicating, if anything the solemn oath script increases MWTP for higher levels of the Rooftop and NoCreditBanking attributes.

30 Similar results were attained when utilizing the dataset used in Mamkhezri, Thacher, and Chermak (2020). See Table A4 of Appendix.

31 Table A5 of Appendix depicts Kolmogorov-Smirnov tests results.

32 See Figure A1 of Appendix for Kernel density graphs on Mamkhezri, Thacher, and Chermak’s (2020) dataset.

33 Similar to the honesty priming script used by Stevens, Tabatabaei, and Lass (2013).

Additional information

Funding

This research was partially supported by U.S. National Science Foundation award #IIA-1301346 to NM EPSCoR.