Incorporating expert knowledge in the estimate of farmers’ opportunity cost of supplying environmental services in rural Cameroon

From Firenze University Press Journal: Bio-based and Applied Economics (BAE)

Claudiane Yanick Moukam, Department of Public Economics, University of Douala

Calvin Atewamba, College Boreal

Nature plays a crucial role in supporting human development; howev-er, the increasing demand for the Earth’s resources is leading to accelerated extinction rates and a decline in global biodiversity and ecosystem services. According to the International Panel on Biodiversity and Ecosystem Services (IPBES, 2019), the average abundance of native species in major land-based habitats has decreased by at least 20%, primarily since 1900. Additionally, more than 40% of amphibian species, nearly 33% of reef-forming corals, and over one-third of marine mammal species are currently facing threats. Recognizing this global challenge, governments worldwide are taking action to incorporate biodiversity and ecosystem services into their development plans, policies, and strategies (IPBES, 2019). These initiatives include targets such as regenerating vegetative cover in the agricultural sector, enhancing agricultural productivity, and reduc-ing the amount of land used for agriculture through the implementation of intensive agricultural systems.Farmers, being at the forefront of environmen-tal conservation in agriculture, play a crucial role. The effectiveness and efficiency of government incentive mechanisms depend not only on the specific design of the schemes (Bareille et al., 2023) but also on the val-ues farmers associate with ecosystem services and the opportunity costs associated with adopting sustainable agricultural practices (Karsenty et al., 2010; Bessie et al., 2014; Kernecker et al., 2021). By taking into account farmer preferences and expectations in the design of government incentive schemes, we can identify the fac-tors that determine the social acceptability and eco-nomic efficiency of these schemes. Conducting research to assess farmer preferences and expectations, as well as estimating farmers’ willingness to accept compensation (WTA) for providing environmental services, is essential in this context. Farmers’ WTA to participate in envi-ronmental protection programmes reflects the opportu-nity cost of supplying environmental services. In other words, farmers express their preferences by assigning selling prices to environmental services, which can be used for their valuation (Brown and Gregory, 1999; Han-ley and Czajkowski, 2019).The economic literature on the adoption of pay-ment for ecosystem services (PES) schemes using a Stat-ed Preference (SP) approach is extensive (Carson, 2012; Villanueva et al., 2017; Johnston et al., 2017; Hanley and Czajkowski, 2019; Wang and Nuppenau, 2021; Raina et al., 2021; Viaggi et al., 2022). However, most SP stud-ies rely on respondents’ hypothetical choices as data to infer their preferences and, consequently, their WTA for changes in environmental services. As noted by Haghani et al. (2021), the hypothetical nature of SP choice set-tings introduces a hypothetical bias, leading people to systematically over or understate their WTA values in SP exercises. This bias arises because no actual pay-ment is made or received in exchange for a change in the quantity or quality of environmental services. Current research on hypothetical bias in SP approaches focuses on understanding its causes and developing methods to mitigate it. One approach to mitigatehypothetical bias is the use of “cheap talk” scripts, which aim to improve the realism of hypothetical sce-narios and reduce the influence of social desirability biases. However, the effectiveness of cheap talk as a bias mitigation tool varies depending on the context and the specific script used, as highlighted by Bosworth and Tay-lor (2012) and Doyon et al. (2015).Another approach to mitigating hypothetical bias is to use “non-hypothetical” or “real” choice experi-ments (Menapace and Raffaelli, 2020; Fang et al., 2021; Cerroni et al., 2023). These experiments involve asking participants to make actual choices rather than hypo-thetical ones, and they can be conducted in laboratory or field settings. Real-choice experiments have been found to reduce hypothetical bias in some contexts, although they can be more expensive and logistically challenging to implement compared to hypothetical choice experi-ments. In addition to these methodological approaches, researchers are exploring the use of behavioral interven-tions to reduce hypothetical bias. Vossler and Holladay (2016, 2018) suggests that framing survey questions in a way that emphasizes the importance of the decision or providing feedback on the accuracy of participants’ responses may encourage more truthful and accurate responses. However, it is important to note that survey-based welfare measures for public environmental goods are often sensitive to elicitation methods, such as wheth-er the elicitation is framed as an up-or-down vote or an open-ended willingness-to-pay question. Controlling for economic incentives, Vossler and Zawojska (2020) show that most survey response formats, including single bina-ry choice, double-bounded binary choice, payment card, and open-ended formats, elicit statistically identical WTP distributions. This finding highlights that behavioral fac-tors may not be the primary drivers of elicitation effects.Overall, research on hypothetical bias in SP approaches is an active and evolving field, with ongo-ing efforts to understand its causes and develop effec-tive mitigation strategies. Reducing hypothetical bias in choice experiments requires not only careful sur-vey design but also the integration of non-survey data information and expert knowledge. Non-data informa-tion refers to prior knowledge or assumptions derived from sources other than observed or survey data, such as expert opinions, previous studies, or theoretical con-siderations (Knuiman and Speed, 1988; Gelman et al., 2013; Mahmoud et al., 2020; Awwad et al., 2021; Hegazy et al., 2021). Incorporating non-data information in SP studies is particularly valuable when survey data is lim-ited, noisy, biased, or when complex problems demand additional information for accurate analysis. By account-ing for non-data information, we can improve analysis accuracy, mitigate the impact of outliers or measure-ment errors, and enhance understanding of economic agent preferences and behaviors (Kadane and Lazar, 2004; Gelman et al., 2013; Kruschke, 2013). However, it should be noted that incorporating non-data informa-tion poses challenges compared to analyzing survey data alone. Despite its potential, there have been limited studies explicitly considering expert knowledge or non-data information to address hypothetical bias in choice experiments. This is partly explained by the difficulty to capture expert knowledge in current WTA modelling frameworks, which usually rely exclusivelyon survey data to estimate the unknown parameters of agent preferences. This paper explores an approach that utilizes non-data information to constrain the range of unknown parameters of agent preferences and aims to reduce hypothetical bias in estimating WTA values.To achieve our objective, we start by conducting a field survey in Barombi Mbo, a rural area in Cameroon, to gather data on the socio-economic and environmental conditions of farmers. The survey includes information on farmers’ willingness to accept (WTA) compensation for participating in agroforestry and afforestation pro-grammes. Additionally, we employ a Multidimensional Preferences Analysis (MPA), a technique used to develop spatial representations of proximities among psychologi-cal stimuli or other entities (Carroll and Chang, 1970; Wish and Carroll, 1982; Davison, 1983), to gain insights into the contextual socio-economic and environmen-tal values of the farmers in Barombi Mbo. This analy-sis helps us understand the various factors influencing farmers’ decision-making processes. We then extend a Tobit model, originally proposed by Tobin in 1958, to estimate the WTA values. The Tobit model accounts for the presence of censoring or truncation in the WTA data. Furthermore, we incorporate stochastic constraints in the model’s parameters using prior distributions. These prior distributions capture our expert knowl-edge or expectations regarding agent preferences when engaging in environmental protection programmes. By adopting a Bayesian approach, we update our knowledge based on the data and obtain posterior estimates of the model parameters. The results of our analysis indicate that a significant majority of farmers in Barombi Mbo are willing to participate in agroforestry and afforesta-tion programmes if their financial constraints are allevi-ated. Furthermore, we find that a higher socio-economic status is likely to promote pro-environmental behaviors among farmers, while increased knowledge on environ-mental protection strategies alone does not necessarily lead to eco-friendly behaviors. Based on our Bayesian estimation, the distribution of farmers’ WTA is found to be normally distributed with a mean of 10,775CFA franc and a standard deviation of 323.59CFA franc. Moreo-ver, we estimate the opportunity cost of providing envi-ronmental services for farmers in our study area to be approximately 3,290,448CFA fanc per year.Our research findings demonstrate qualitative dif-ferences from the existing literature (Moukam, 2021; Gou et al., 2021; P ́erez-S ́anchez et al., 2021). While previous studies have acknowledged the potential of employing a Bayesian approach for modeling ecosystem services (Landuyt et al., 2013; Ban et al., 2014; Uusitalo et al., 2015; Hofer et al., 2020), a review of these stud-ies reveals that the technique is not yet fully utilized. It has been highlighted in Hofer et al. (2020); Moukam (2021); Gou et al. (2021); P ́erez-S ́anchez et al. (2021) that the standard approach for modeling ecosystem ser-vice delivery relies solely on data, without incorporat-ing expert knowledge, which can lead to controversial results regarding the drivers of economic agent behavior for environmental protection. In contrast to the afore-mentioned studies, our approach incorporates expert knowledge through the utilization of prior distributions for the model parameters. By doing so, we not only pro-vide mean-ingful insights into the determinants of econom-ic agent preferences but also significantly improve the estimation of WTA compensation for participation in environmental conservation efforts. This allows us to account for situations where the available data may not adequately capture the tangible and intangible benefits of the environment. Our results suggest that the con-ditional probability of the parameters provides the best summary of the knowledge we can gain from the data.The remaining sections of the paper are structured as follows. Section 2 provides a description of the study area, emphasizing its agroecological characteristics and the availability of agricultural extension services. In Sec-tion 3, we outline the research methodology, including details on the survey design, data collection process, and analytical methods employed. The obtained descriptive statistics, research findings, and their discussions are presented in Section 4. Finally, Section 5 serves as the conclusion of the paper, summarizing the key points and providing policy implications based on the findings.

DOI: https://doi.org/10.36253/bae-13534

Read Full Text: https://oaj.fupress.net/index.php/bae/article/view/13534