Abstract
This article examines heterogeneity in the preferences of households regarding participation in a Payments for Ecosystem Services (PES) programme. We argue that such heterogeneity is particularly likely for schemes that are implemented in a developing-country setting, where households differ in their degree of integration into markets. We use the case study of the Sloping Land Conversion Programme (SLCP) in China, one of the largest PES schemes in the world. As the SLCP is not voluntary in all cases, we compare the determinants of observed participation with households’ stated preferences about future participation. Our analysis uses a novel latent-class approach to model the household decision over whether to sign up to a PES programme. Allowing for variation in the parameters of the decision process, we find significant differences between households with good access to markets and those facing market imperfections.
Notes
1. $1 = approximately ¥6.
2. We focus here on quantitative analysis of participation vs. non-participation. Some qualitative evaluations of PES programmes have included questions for participants about their reasons for participating (e.g. Kosoy et al. 2007).
3. There may be indirect benefits for those using ecosystem services or providing employment to participants.
4. Six to seven households were selected from each village. On average, villages contained approximately 400 households.
5. Survey enumerators were hired graduate students from Beijing University, who were trained by the authors of the study.
6. The attributes of the future SLCP in the choice experiment were: the degree of assurance of the subsidies; the percentage of ecological forest required (as opposed to commercial forest, which can provide saleable non-timber products); whether land rental requires village leader authorisation; whether land redistribution is permitted; and the subsidy amount.
7. This analysis, along with further details on the choice experiment, can be found in Grosjean and Kontoleon (2009).
8. As different crops are grown in the sample regions, the yield variable is an index, calculated relative to the mean productivity of the primary crop planted by each household.
9. We do not include the length of time in the scheme because all households in our sample entered at approximately the same time.
10. Models are estimated in NLOGIT 3.0.