Context effects and heterogeneity in voluntary carbon offsetting – a choice experiment in Switzerland

Abstract

Voluntary carbon offsets allow individuals to neutralise the CO emissions from their consumption. In a choice experiment among more than a thousand Swiss consumers, we analyse how the propensity to offset varies with consumption contexts and offset project attributes. The adopted latent class model accounts for heterogeneity in consumers’ preferences and in their motivations to buy carbon offsets. We find that consumers are not only responsive to project type and quality aspects – such as government certification – but also to the consumption context. Our results suggest that carbon offsets are perceived as differentiated rather than homogeneous goods. Estimating a single willingness to pay measure for offsets, as previous studies have done, may thus be misleading. To complement our analysis, we provide a characterisation of offset customers as well as insights into the relationship between carbon offsetting and other voluntary mitigation behaviour.

Keywords:

• voluntary carbon offsets
• context effects
• heterogeneity in preferences and motivations
• choice experiment
• latent class model

Acknowledgements

We are grateful to Renate Schubert for her supportive supervision of this research and helpful comments. We also thank an anonymous referee as well as Markus Ohndorf, Andreas Ziegler, Helga Fehr, William Greene and Jürgen Meyerhoff for their valuable suggestions.

Notes

1. The concept was first applied in 1989 when the US electricity company AES paid Guatemalan farmers to plant 50million trees to compensate for its carbon emissions (House of Commons 2007; Bellassen and Leguet 2007). Within two decades, neutralising carbon emissions has emerged as a retail service in a fairly sizeable market, open to firms as well as individuals.

2. With unit t we designate metric tons throughout this paper.

3. In 2011, prices ranged from less than 1 USD/tCO to more than 100 USD/tCO, with an average of about 6 USD/tCO (Peters-Stanley and Hamilton 2012). Price differences are largely associated with the information about the project type and host country as well as the type of third-party certification by different project standards (Conte and Kotchen 2010).

4. At the current exchange rate, the estimated WTP is around 8 and 16 USD. In both studies, the offered emission reduction was a European Union emissions allowance (EUA).

5. The sampling was carried out by a marketing research firm (Intervista).

6. One exception is the share of respondents with an academic degree: according to the Swiss Statistics Office (BFS) only around 25% of the Swiss population received tertiary education, including degrees from universities, universities of applied sciences and advanced technical or pedagogical colleges (e.g.‘College of Professional Education and Training’ or ‘School of Engineering’).

7. We transferred the donations to a Swiss non-profit carbon offset provider after the survey.

8. Our pretests revealed that the existing certification schemes such as the Verified Carbon Standard or the CDM Gold Standard are unknown to many people. Therefore, we decided to inform respondents about the type of certification body. While UN and NGO certifications are currently available, there is no Swiss government certification to date. However, a UK government’s initiative, the Government’s Quality Assurance Scheme for Carbon Offsetting, launched in 2009 and discontinued in 2011 (Pure 2011), can serve as an example for government certification.

9. Each of the four contexts was assigned exactly two times in each block. The six blocks were randomly assigned to the subjects.

10. For different ways of dealing with respondent uncertainty see e.g. Ready, Whitehead, and Blomquist (1995), Champ et al. (1997), Akter, Bennett, and Akhter (2008) or Martínez-Espiñeira and Lyssenko (2012).

11. The latent class model does not make a specific assumption about the distribution of the parameter values across individuals but only approximates the underlying distribution by a discrete form (Hensher and Greene 2003).

12. The reliability of these scales can be expressed in values of Cronbach’s of 0.77 and 0.79, respectively.

13. For comparison, Champ et al. (1997) consider only the adoptions with the highest certainty level (10 in their case) as a positive response in order to identify a lower bound for WTP.

14. The results of the latent class analysis on the entire sample of 1010 respondents are provided in Table A1 in the appendix. Comparing those results with the results listed in Table 3 indicates that a strong majority of the excluded respondents can be associated with latent class 1, that is the non-offsetting class. The rest of the results, regarding both the distribution of classes and the patterns of preferences, are very similar regardless of the sample. This strong similarity indicates that the excluded 15% of respondents can be safely classified as a ‘non-offsetting’ group.

15. We used AIC and BIC criteria. Both criteria show a rapid decline between 3 and 4 classes. While AIC still slightly declines from 4 to 5 classes, BIC rises by adding the fifth class. The model did not converge with more than five (four) classes for the sample with 854 (1010) respondents.

16. Based on the question ‘Have you ever made a CO compensation payment for your personal consumption?’

17. This corresponds to the median donation.

18. The footprint indicator is calculated as the average value of the variables indicating frequency of air travel, yearly number of hotel stays, yearly mileage with personal car, frequency of weekly meat consumption. All four variables take values from one to six and are weighted equally. This average value is further increased in case respondents run their heatings on fossil fuels and are travelling business class, and it is decreased in case respondents claimed to participate in carpooling and to buy green electricity.

19. Note that having an academic degree and having heard about offsetting before the survey is weakly correlated , so that knowledge of offsetting might partly mediate this effect.