Preferences for biodiversity offset contracts on arable land: a choice experiment study with farmers

Preferences for biodiversity offset contracts on arable land: a choice experiment study with farmers Abstract Biodiversity offsetting (BO) is aimed at achieving no net loss of biodiversity in the context of economic development. Through a choice experiment in Northern France, we show that farmers have a clear preference for not signing up BO contracts. The contracts they accept may only be suitable for offsetting temporary impacts on already degraded areas of natural habitat but not for permanent impacts on high-quality habitat. We find that the introduction of a conditional monetary bonus can improve the organisational and ecological efficiency of BO because it increases the enrolled acreage in a BO contract per farmer albeit at an increased cost for developers. 1. Introduction Development projects (e.g. real estate projects, the construction of transport infrastructure) often lead to the destruction or degradation of natural and semi-natural ecosystems, with consequent impacts on biodiversity, ecological processes and associated ecosystem services. Even where developers must follow a mitigation hierarchy that includes measures to first avoid and then reduce their potential impacts on biodiversity, significant residual impacts on ecosystems and species often remain. Offsetting, the last step of the mitigation hierarchy, is supposed to compensate for residual ecological losses, in order to achieve no net loss (NNL) of biodiversity once the project and its mitigation and offsetting measures have been successfully carried out (Kiesecker et al., 2010; Bull et al., 2013; Maron et al., 2018). Current offset practices rely mainly on a biophysical1 approach (Jacob et al., 2016). Given their NNL goal, biodiversity offsets (BO)2 must provide measurable ecological gains, through ecosystem management activities, that are equivalent to the ecological losses in the impacted area. To be qualified as gains, the activities must be additional to existing biodiversity conservation activities and their duration must be proportionated to that of impacts. BO schemes can be implemented by the developers themselves or by others acting on their behalf, including farmers. In the latter case, farmers are paid for voluntarily implementing the BO on behalf of public or private developers (who, in France, remain the environmental permit holder). The mitigation hierarchy and BO have been promoted as an innovative mechanism for achieving sustainable development goals and are finding their way into environmental regulations in many countries (e.g. Calvet, Ollivier and Napoléone, 2015; Gonçalves et al., 2015; Bennett et al., 2017; Wende et al., 2018). In France, however, the implementation of BO has been highly variable and often ineffective since the concept was introduced in the 1976 Nature Protection Act (Quétier, Regnery and Levrel, 2014). The French Ministry of Environment published several guidelines in recent years to improve implementation of the mitigation hierarchy (e.g. MEDTL, 2012; MEDDE, 2013). These guidance documents were partly enacted in August 2016 when a new Law on Biodiversity was passed. This is still recent which creates a somewhat fuzzy context for BO design and implementation. Across Europe, the growing uptake of NNL policies means that BO contracts are increasingly offered to farmers (Wende et al., 2018). This is particularly relevant in France, where agriculture is the dominant land use (59.5 per cent of the country’s land area, 2012 CORINE Land Cover inventory3) and farmers are likely to be approached by developers for siting development projects or for BO on farmland. However, one can question if BO requirements for achieving NNL are compatible with the constraints and interests of farmers. First, BO requirements are much more restrictive for agricultural activity than the widespread agri-environment schemes (AES) and they include longer durations of contracts, from which it is not possible to withdraw without stiff penalties. Second, there are limits to the social acceptance of BO on agricultural land. Objections by farmers relate mainly to (i) the use of arable land for development projects, (ii) the use of arable land for offsetting and (iii) the resulting pressure on arable land prices (de Billy et al., 2015; Etrillard and Pech, 2015). Some farmers, however, see BO as an opportunity to diversify their activities by implementing biodiversity-friendly activities on their land, on behalf of developers, without selling the land itself (Etrillard and Pech, 2015). The political discourse of most farm union-run bodies in France is currently shifting from opposition to a call for better farmer engagement in BO on agricultural land. A parallel question is how to organise the ecological supply for a BO linked to a given development project. Across France, BOs are currently implemented piecemeal, through the purchase and ad hoc contracting of a multitude of small sized parcels of land, including farmland. This is costly for developers to put in place and manage, makes it difficult for regulators to control and monitor BO performance and affects the ecological efficiency4 (Quétier, Regnery and Levrel, 2014). It has been shown that, in the context of BO policies, a smaller number of agents involved in the delivery of BO activities (e.g. ecosystem restoration) reduces the transaction costs of BO policy implementation (e.g. Levrel et al., 2015; Vaissière and Levrel, 2015). Moreover, large scale5 ecological restoration actions have been proven to be more effective for a wide range of ecosystems and locations (e.g. Menz, Dixon and Hobbs, 2013;,Moreno-Mateos et al., 2012 for wetlands). Large patches and a highly connected habitat network are criteria in ‘banking’ approaches for BO implementation, where activities are anticipated and pooled (aggregated) (van Teeffelen et al., 2014). Against this background, knowing the conditions under which farmers would agree to increase the acreage of land enrolled in a BO on their farm is relevant as a way to reduce the number of stakeholders involved in the implementation of a BO. In summary, in this paper, we address the need for objective and informative research on (i) the conditions under which farmers would be prepared to enrol in BO activities (or not) and (ii) the factors that influence the enrolled share of their cultivated land. Our results have policy relevance against the on-going discussion of the various technical and organisational solutions for achieving NNL in France, and more broadly in Europe (Tucker et al., 2013; Wende et al., 2018). Our results are also informative in view of the next reform of the common agricultural policy (CAP) (Pe’er et al., 2017). We used a choice experiment (CE) to assess farmer preferences for the implementation of BO contracts on arable land in Picardy, Northern France. In our experiment, BO proposals vary with different levels of contract attributes that we selected with the help of focus groups: management plan of the land restoration, contract duration, the annual payment and an additional conditional bonus. This bonus consists of a side payment offered in addition to the baseline payment. It is only paid if a farmer enrols a minimum contiguous land area on his farm. We also calculated farmers’ willingness to accept (WTA in €/ha/year) for different BO contract features. CEs are increasingly used in various research areas (transport economics, marketing, environmental economics, etc.), including the analysis of farmers’ preferences for AES. These studies seek to understand which parameters would be considered for improving participation in AES contracts, and to understand stakeholders’ preferences for biodiversity conservation schemes (e.g. Broch and Vedel, 2012; Lienhoop and Brouwer, 2015; Greiner, 2016). In addition to contract length and the payment level, which are frequently tested, many attributes related to the flexibility of the contracts are tested. These may involve technical restrictions (Bougherara and Ducos, 2006; Espinosa-Goded, Barreiro-Hurlé and Ruto 2010; Christensen et al., 2011; Kuhfuss, Préget and Thoyer, 2014; Greiner, 2016), the parcels involved (Ruto and Garrod, 2009), bonuses for agri-environmental contracts with a collective dimension (Chen et al., 2009; Kuhfuss et al., 2016) or the possibility to opt-out of the contract (Broch et al., 2013; Greiner, Bliemer and Ballweg, 2014). The administrative burden is also often studied (Bougherara and Ducos, 2006; Ruto and Garrod, 2009; Christensen et al., 2011). Additional attributes linked to the specific case studies include the width of pesticide free buffer zones (Christensen et al., 2011), cattle and tree density in measures for the protection of a specific forest ecosystem (Santos et al., 2015), the option of localised herbicide use in vineyards (Kuhfuss et al., 2016), the opportunity to return to agriculture after the contract ends (Lienhoop and Brouwer, 2015) or afforestation purposes (Broch and Vedel, 2012). To our knowledge, there are only a few CE studies dealing with the implementation of BO on arable land that assess farmers’ preferences. Le Coënt, Préget and Thoyer (2017), for example, studied farmers’ participation in agri-environmental contracts framed as BO versus a traditional biodiversity conservation programme (not tied to an ecological loss). They show that farmers have different preferences for these two situations and that their monetary demands are higher for BO contracts. Our approach differs from previous studies by considering connection and/or agglomeration at the farm level at a bonus payment for acreage enrolled in the BO. We do not seek to investigate coordination among neighbours but we are interested in a landscape-scale approach as promoted in the context of agri-environmental schemes (e.g. Gonthier et al., 2014). To our knowledge, our study is the first to use this type of bonus in a CE. Previous studies that have tested bonuses aim at the number of signed AES (e.g. Buckley, Hynes and Mechan, 2012; Banerjee et al., 2014; Kuhfuss et al., 2016) so that the number of farmers under contract reaches a predetermined threshold within a given geographical area. Other studies involve conservation or restoration contracts that are not limited to agricultural land. A significant proportion of this literature is based on laboratory experiments, with few empirical studies (Parkhurst et al., 2002; Parkhurst and Shogren, 2007; Drechsler et al., 2010; Wätzold and Drechsler, 2014). Kuhfuss et al. (2016) studied (empirically) the effect of nudging respondents to increase land enrolment. 2. Materials and methods 2.1. Methodological options 2.1.1. Modelling farmers’ decisions to enrol in a BO contract Farmers’ decision to enrol or not in a BO contract will result from the comparison of the utility they derive from different alternatives. Following Lancaster’s theory (1966) and the random utility theory (Luce, 1959; McFadden, 1974), farmer i (i = 1, …, n) will choose alternative j (j = 1, …, m) in choice card Ct (t = 1, …, T) if the alternative is the one that provides him/her the highest level of utility among all m alternatives proposed in the choice card. The utility is defined by an observable part and a random part represented by error terms. In the conditional logit (CL) model, it is supposed that the error terms are independently and identically distributed (IID) among the alternatives and across the population and that irrelevant alternatives are independent (IIA). If Ajit is a dummy variable that takes the value of 1 if alternative j is chosen by farmer i in the choice card Ct, the probability related to this choice is (see Kuhfuss et al., 2016)   P(Ajit=1)=exp⁡(X′jitβ)∑m∈Ctexp⁡(X′mitβ) (1)where Xjit denote the attributes of alternative j faced by farmer i and β is the vector of k preference parameters, representing the average importance of each attribute of the BO contract on farmers’ preferences. The hypothesis of IIA is a strong assumption. The mixed logit (ML) model relaxes this assumption and allows assessing the βki that are specific to each interviewee, and randomly distributed across the population, with a density function f(βk). It thus captures heterogeneity of farmer’s preferences. Then, conditional on vector βi, the probability that farmer i chooses alternative j in choice card Ct is (see Kuhfuss et al., 2016)   P(Ajit=1|βi)=exp⁡(X′jitβi)∑m∈Ctexp⁡(X′mitβi). (2) Thus, the probability of observing the sequence of T choices by individual i is   P(Aji1=1,…,AjiT=1)=∫∏t=1T(exp(X′jitβ)∑m∈Ctexp(X′mitβ))f(β)dβ (3)where f(β) can be specified to be normal or lognormal: β~N(b,W) or lnβ~N(b,W). The mean b and the covariance W are to be estimated by simulation (Train, 2009). We calculate the individual-level parameters for each attribute using the method proposed by Revelt and Train (2000) for ML models. 2.1.2. Calculating the WTA for implementing BO contracts Following the analysis of the attributes that explain farmers’ choices of enrolling in BO contracts, we calculate farmers’ WTA for attribute k in BO contracts. WTA can be computed from the following equation:   WTAk=−βkβmon (4)where βk and βmon are the parameters associated with attribute k and the monetary attribute, respectively. The individual farmer’s WTA for the attribute k is calculated as follows:   WTAki=−βkiβmon (5)where βki and βmon are the parameters associated to attributes k for the individual i and the monetary attribute for the full sample, respectively. 2.1.3. Modelling farmers’ decisions on the acreage to enrol in a BO contract In addition, we analyse the farmers’ decision to enrol certain acreage in a BO contract. Our dependent variable y is the enrolled share of each farmer’s used agricultural area (UAA)6 with 0<y<1. Explanatory variables are the contract attributes and other individual characteristics. To model this proportion, we followed Kuhfuss et al. (2016) and use a beta regression model instead of a simple linear model (Paolino, 2001; Ferrari and Cribari-Neto, 2004). 2.2. The Picardy case study and survey Our study uses agricultural land located in Picardy, Northern France, as a case study. Agriculture is the dominant land use, covering 75.2 per cent of Picardy (forests and urban areas cover 17 per cent and 6.7 per cent respectively), as shown in Figure 1 (2012 CORINE Land Cover inventory). Most farms have a UAA of more than 100 hectares (mean is 200 hectares) and common crops are cereals, oilseeds and other arable crops (AGRESTE, 2010). Although modified (i.e. managed), the land provides habitat for threatened and protected birds such as the corncrake (Crex crex) or the stone-curlew (Burhinus oedicnemus), and plants such as the hurtsickle (Centaurea cyanus). These species are dependent on large expanses of low-intensity cropland and permanent grasslands, habitats that have been lost over past decades (Stoate et al., 2001, 2009). Fig. 1. View largeDownload slide Land use in Picardy, Northern France (2012 CORINE Land Cover inventory). Fig. 1. View largeDownload slide Land use in Picardy, Northern France (2012 CORINE Land Cover inventory). To inform the design of our CE, we conducted three focus groups with local farmer union-run bodies (Chambres d’Agriculture) in Picardy from July 2015 to February 2016. Focus groups included technical staff and/or farmers who helped us determine the key characteristics of BO contracts (i.e. the attributes in our CE and their levels), develop realistic CE scenarios and formulate follow-up questions for our survey. A pilot survey to test the questionnaire was conducted during February and March 2016 with 26 farmers in Picardy after which the questionnaire was adjusted following the received farmer feedback. The final questionnaire includes three parts (described in Sections 2.2.1–2.2.3) and was implemented using Lime Survey (Version 2.05+ Build 150413), an online application. The time to complete the questionnaire was estimated at 15–20 minutes (10 minutes of video explanations and 5–10 minutes for responding to questions). The link to the survey was sent by the local farmer union-run bodies to all the available e-mail addresses of farmers in Picardy (5,100 contacts) at the beginning of May 2016. The survey was available for 1 month, until early June 2016. The collected answers are anonymous. The web-based approach is well suited for our case study. First, farmers in this French region are in general well connected to the Internet. According to the French National Institute of Statistics and Economic Studies (INSEE), 70.6 per cent of French farmers had an Internet connection (this figure was 78.2 per cent for French people on average) in 2012 (Gombault, 2013). The Internet is a common means of communication between the farmers and the local farmer union-run bodies and farmers make online declarations annually for CAP subsidies. Second, this method enables us to reach out to a large sample of farmers. Finally, it is well known that web-based surveys avoid bias from the presence of an interviewer. 2.2.1. Questionnaire introduction and attributes description The questionnaire was accompanied by a video presentation of the current legal framework for BO in France, as farmers are not necessarily well informed on the topic. Next, we presented a fictive development project in their region implying destruction of meadows of ecological interest which triggers a BO requirement for the developer, who would then propose BO to the farmer. If the farmer accepts one of the two contracts offered, he/she would implement the management plan of the BO on arable land on his/her farm on behalf of the developer and would be paid for this activity. Alternatively, the farmer can decline the offer and keep his/her current agricultural practices by choosing the status quo option. The eligibility rules and terms and conditions of the contract were as follows: (i) the activity has to be additional to regulatory obligations (i.e. they cannot replace cross-compliance requirements under the CAP, current AES contracts, mandatory buffer zones along watercourses and lakes in line with the EU Nitrate Directive, or with previously planned and funded programmes); (ii) farmers would get technical support provided by relevant technical and administrative staff from the local farmer union-run bodies, or other specialised public agencies or non-governmental organisations (NGOs); (iii) farmers must agree to give access to their land for ecological monitoring and compliance control by regulators. In addition, (iv) none of the parties of a BO contract has a right of withdrawal. Instead of using long written explanations, we created short and accessible videos and made them available online.7 The use of videos to introduce information at the beginning of a CE is novel.8 To our knowledge, no study has been carried out to measure the possible influence of this method on the answers compared with more widely used textual or graphical methods. Studies using virtual reality instead of text within choice cards report improvements in the attention of the interviewees or the consistency of their choices (e.g. Bateman et al., 2009; Matthews, Scarpa and Marsh, 2017; Patterson et al., 2017). To limit bias, we prepared the videos to be neutral and objective, in collaboration with focus group participants. The BO contracts have four attributes. The first attribute is the management plan. The four possible management plans (Figure 2) share a common foundation of technical specifics. This foundation implies that arable land must be converted into meadows,9 with seed mix of legumes and grasses, and subsequently mowed (not grazed). Restrictions apply to mowing (centrifugal, forbidden at night, 15 cm minimum height, etc.). In addition, the area concerned has to be at least 10 m wide and larger than 0.5 ha. The ecological ‘gain’ increases from Levels I to IV of the management plans (Figure 2), due to decreased nitrogen (N) fertilisation, further delay of mowing and the addition of a refuge zone. The refuge zone is a zone of the meadow at least 10 m wide representing 10 per cent of the contracted area that is not mowed any given year, and that can be moved from 1 year to the next within the contracted area: it is not a fixed set-aside for the duration of the contract. These four levels are increasingly accommodating practices given the life cycle of meadow dependent plant and animal species. Such BO requirements are very close to the contracts a farmer would face in a ‘real life’ BO situation given the species and other biodiversity features that trigger offsets in Northern France. This also applies to the second attribute related to contract duration. Fig. 2. View largeDownload slide The four levels of proposed management plan. Fig. 2. View largeDownload slide The four levels of proposed management plan. Contract duration is frequently mentioned as an important factor for enrolment and is likely to be important for BO contracts as well. After the focus groups consultation, chosen durations (that cannot be terminated during the contract period) were 9, 18, 25 or 40 years. The third attribute, the conditional monetary bonus, is associated with adding (ecologically relevant) conditions to the management plan. The bonus is proposed in some scenarios and can be accepted or not by the farmers. The two levels for this attribute are thus ‘available’ or ‘not available’ in the scenarios. If the bonus is available, the farmer may decide to activate this bonus, implying a €200/ha/year payment in addition to the baseline payment. However, this bonus involves further conditions: the farmer signs the contract for the scenario for at least 5 ha and the restored parcels must be grouped in one piece or included in an ecological network on the farm. The fourth attribute is the actual payment of €800/ha/year, €1100/ha/year, €1500/ha/year or €2000/ha/year. The lower limit of these amounts was chosen based on unit AES payments, in the Picardy region that relate to the following: creation and maintenance of perennial herbaceous cover (EU COUVER06), extensive management of grassland (EU HERBE03), mowing delay (EU HERBE06) and repair and maintenance of a cover of floristic or faunistic interest (EU COUVER07). The latter is similar to the maintenance of a refuge zone as included in our scenarios. The range of values was discussed and adjusted during the focus group discussions and again after the pilot study. Following Rose and Bliemer (2013), we use a wide range of payments to enable the WTA of farmers most hesitant to implement the practices to be assessed. 2.2.2. The choice experiment The second part of the questionnaire is the CE per se. To understand which of the four attributes is most important to farmers, we combine their different levels in scenarios that describe different types of BO. Since we have four attributes with two to four levels each and choice cards with two alternatives, the full factorial design generates 16,25610 different choice cards. We used the SAS software (SAS University Edition, Version workstation 6.5–7.x virtual machine) and its command %mktruns to decide on the number of scenarios to propose to interviewees: 16 different combinations. We gathered the scenarios in pairs leading to eight choice cards (illustrated in Figure 3). We added an opt-out answer to each choice card, stating ‘I prefer to keep my current agricultural practices’ (the status quo). The use of an opt-out answer is supposed to improve the realism of the choice cards and hence improve the estimation results (Adamowicz and Boxall, 2001; Kontoleon and Yabe, 2003). Thus, farmers choose which BO they would agree to carry out on their farm. After they selected one of the contracts, an additional question appeared: we asked them how many hectares they would be able to commit to the selected BO (acreage). After the focus groups deliberation, it was decided it would not be necessary to create subgroups of choice cards (blocks); all the farmers thus answered the eight choice cards. Fig. 3. View largeDownload slide Example of one of the choice cards of the final CE survey. Fig. 3. View largeDownload slide Example of one of the choice cards of the final CE survey. We selected the eight choice cards following a two-step procedure. First, the pilot study choice cards were chosen using an orthogonal efficient design with all the prior parameters set to 0, to determine the specific prior parameters of our sample (using the NGene software). Our D-error was 0.020317, which is acceptable. We analysed the results of the pilot study with a multinomial logit model to get a better idea of the value of the prior parameters for each attribute, or for each level of attribute for the dummy variables, and then minimise the variance. Second, the final study choice cards were chosen using a Bayesian efficient design, employing the parameters obtained from the pilot study (using the NGene software). Bayesian prior parameters11 are increasingly used as they enable random priors to be defined around the value of the prior parameters from the pilot study (for a review, see Chaloner and Verdinelli, 1995). In line with the literature (e.g. Rambonilaza and Dachary-Bernard, 2007; Kaczan, Swallow and Adamowicz, 2013; Le Coënt, Préget and Thoyer, 2017), we used effects coding instead of dummy coding for the bonus variable in order to avoid confounding the coefficient of the opt-out answer (Bech and Gyrd‐Hansen, 2005; Hasan-Basri and Karim, 2013). The presence of the bonus in a BO contract is coded as 1 while the opt-out option has no bonus by definition (coded 0). The absence of the bonus in a proposed BO contract is coded −1. As suggested by Haaijer, Kamakura and Wedel (2001), we used an additional variable for the opt-out. The opt-out takes the value of 1 when the farmer chooses to keep the current practices with the four attributes taking a value of 0. When a BO contract is selected, the opt-out takes the value of 0. The opt-out variable thus captures the preference of the farmers for their current practices (Vermeulen, Goos and Vandebroek, 2008). Table 1 summarises the attribute description and variable names (cf. Section 2.2.1), attribute levels and the corresponding coding for the analysis. Table 1. Attribute description, levels and their coding to describe BO scenarios proposed to farmers BO attributes and the opt-out  Description  Levels  Coding  Management plan  Levels of management plan required by the BO contract related to: quantity of azote fertilisation (UN), date of mowing, and presence of a refuge zone  Level I: 30 UN, 20 June, no refuge zone  1  Level II: 0 UN, 20 June, no refuge zone  2  Level III: 0 UN, 20 July, no refuge zone  3  Level IV: 0 UN, 20 July, refuge zone  4  Opt-out  0  Contract duration  Total duration of the BO contract  9 years  9  18 years  18  25 years  25  40 years  40  Opt-out  0  Conditional monetary bonus  Additional payment (€200/ha/year) for ecologically relevant measures, provided that the bonus is available in the scenario  Available bonus (€200/ha/year)  +1  No bonus in a BO contract  −1  No bonus because this is the opt-out answer  0  Payment  Payment received each year by the farmer per enrolled hectare  €800/ha/year  800  €1100/ha/year  1100  €1500/ha/year  1500  €2000/ha/year  2000  Opt-out  0  Opt-out  The farmer prefers to keep the current practices  Opt-out  1  BO contract 1 or 2  0  BO attributes and the opt-out  Description  Levels  Coding  Management plan  Levels of management plan required by the BO contract related to: quantity of azote fertilisation (UN), date of mowing, and presence of a refuge zone  Level I: 30 UN, 20 June, no refuge zone  1  Level II: 0 UN, 20 June, no refuge zone  2  Level III: 0 UN, 20 July, no refuge zone  3  Level IV: 0 UN, 20 July, refuge zone  4  Opt-out  0  Contract duration  Total duration of the BO contract  9 years  9  18 years  18  25 years  25  40 years  40  Opt-out  0  Conditional monetary bonus  Additional payment (€200/ha/year) for ecologically relevant measures, provided that the bonus is available in the scenario  Available bonus (€200/ha/year)  +1  No bonus in a BO contract  −1  No bonus because this is the opt-out answer  0  Payment  Payment received each year by the farmer per enrolled hectare  €800/ha/year  800  €1100/ha/year  1100  €1500/ha/year  1500  €2000/ha/year  2000  Opt-out  0  Opt-out  The farmer prefers to keep the current practices  Opt-out  1  BO contract 1 or 2  0  Table 1. Attribute description, levels and their coding to describe BO scenarios proposed to farmers BO attributes and the opt-out  Description  Levels  Coding  Management plan  Levels of management plan required by the BO contract related to: quantity of azote fertilisation (UN), date of mowing, and presence of a refuge zone  Level I: 30 UN, 20 June, no refuge zone  1  Level II: 0 UN, 20 June, no refuge zone  2  Level III: 0 UN, 20 July, no refuge zone  3  Level IV: 0 UN, 20 July, refuge zone  4  Opt-out  0  Contract duration  Total duration of the BO contract  9 years  9  18 years  18  25 years  25  40 years  40  Opt-out  0  Conditional monetary bonus  Additional payment (€200/ha/year) for ecologically relevant measures, provided that the bonus is available in the scenario  Available bonus (€200/ha/year)  +1  No bonus in a BO contract  −1  No bonus because this is the opt-out answer  0  Payment  Payment received each year by the farmer per enrolled hectare  €800/ha/year  800  €1100/ha/year  1100  €1500/ha/year  1500  €2000/ha/year  2000  Opt-out  0  Opt-out  The farmer prefers to keep the current practices  Opt-out  1  BO contract 1 or 2  0  BO attributes and the opt-out  Description  Levels  Coding  Management plan  Levels of management plan required by the BO contract related to: quantity of azote fertilisation (UN), date of mowing, and presence of a refuge zone  Level I: 30 UN, 20 June, no refuge zone  1  Level II: 0 UN, 20 June, no refuge zone  2  Level III: 0 UN, 20 July, no refuge zone  3  Level IV: 0 UN, 20 July, refuge zone  4  Opt-out  0  Contract duration  Total duration of the BO contract  9 years  9  18 years  18  25 years  25  40 years  40  Opt-out  0  Conditional monetary bonus  Additional payment (€200/ha/year) for ecologically relevant measures, provided that the bonus is available in the scenario  Available bonus (€200/ha/year)  +1  No bonus in a BO contract  −1  No bonus because this is the opt-out answer  0  Payment  Payment received each year by the farmer per enrolled hectare  €800/ha/year  800  €1100/ha/year  1100  €1500/ha/year  1500  €2000/ha/year  2000  Opt-out  0  Opt-out  The farmer prefers to keep the current practices  Opt-out  1  BO contract 1 or 2  0  2.2.3. Follow-up questions The third part of the questionnaire included follow-up questions developed with the help of the focus groups. These questions were aimed at characterising the respondents and at improving our interpretation of the results. A first set of follow-up questions dealt with the farmers, their socio-economical profile and details about their farms. We used this information to assess the representativeness of our sample. While we tested all these respondent characteristics in the econometric analysis, only some of them had a significant effect in the model results. In the result section, we provide in a footnote a list of the characteristics that were insignificant in the analysis. Other questions were added to check the quality of our CE answers. First, to reveal possible attribute non-attendance (see Kehlbacher, Balcombe and Bennett, 2013 for a review and test), we asked respondents if they focused their answers on one specific attribute or systematically ignored an attribute. Second, we identified and removed protest responses from the analysed sample (i.e. the answers from respondents that systematically refuse to choose any contract without paying attention to their composition) because such answers do not add any information on preferences for attributes (Adamowicz et al., 1998; Barrio and Loureiro, 2013). To identify the protest responses among the farmers who had systematically chosen the opt-out answer we removed: (i) those who stopped filling the questionnaire just after the last choice card part of the questionnaire and (ii) those who answered that the main reason for the opt-out was that (a) non-withdrawal was unacceptable, (b) she/he was against the BO principle in general or (c) she/he was against the BO principle in the agricultural context. Other farmers who systematically selected the opt-out answer were kept in the sample because they are likely to have given serious attention to and compared the eight choices on offer. The three reasons mentioned above (a, b and c) were presented in the online questionnaire when a farmer systematically chose the opt-out answer12 or chose the opt-out answer at least once.13 Third, we used the follow-up questions to check the additionality of the proposed choices at the scale of the farm. As explained in Section 1, farmers were aware that the measures had to be additional with on-going activities and regulatory requirements beneficial to meadow wildlife. In addition, we asked farmers if they would compensate the loss of the production on the contracted land by increasing production on other parts of their farm (often called the leakage effect). If this involves, e.g. ploughing an existing meadow of conservation value, the additionality and purpose of the offset contract is weakened. 3. Results 3.1. Sample characteristics We received 162 answers, giving us a response rate of about 3.2 per cent. After dropping 18 protest responses, we obtained 144 useful responses, equivalent to 3,456 observations. The fact that we did not separate the eight choice cards in blocks leads to a high number of observations per farmer. Descriptive statistics of useful answers are presented in Table 2. The representativeness of the sample is good, in terms of spatial distribution of the sampled farmers (see Figure 4) and age (apart from the age class of over 60 years old). The mild under-representation of this age class may be linked to the web-based nature of our enquiry. We observed a slight over-representation of farmers with large farms (the mean UAA of the sample is of 199 ha). Moreover, cereal and oil seed crops and single household unincorporated farms are also slightly over-represented. The farmers who answered our enquiry seem to already have some knowledge on BO. While a majority (78 per cent) of them have never been contacted before by project developers for BO contract, 63 per cent had heard of BO contracts before receiving the questionnaire. We may have an over-representation of people already knowledgeable on BO, but no other regional data on this subject are available to test this. It is possible that people with a strong opinion against BO in general, or BO on agricultural land, or people who do not know BO ignored our enquiry. We may thus also have an under-representation of strong opponents to BOs. We do not have this information, which may be a limit of web-based surveys compared to face-to-face or phone interviews, where disagreements to participate could have been recorded. We discuss below the possible consequences of these under and over-representations. Overall, the representativeness of the sample is good, allowing us to draw conclusions for comparable rural areas. Table 2. Sample and population characteristics   Farms and farmers in the sample  Farms and farmers in Picardya  Gender (%)  Male  83  No data  Female  9  No answer  8  Age class (%)  [18,39]  20  18  [40,49]  33  29  [50–59]  35  32  >60  6  21  No answer  6  /  Used Agricultural Area (UAA)  Mean UAA (ha)  199  98  Number of exploitation (%):   <50 ha  5  38   [50–100] ha  14  22   >100 ha  72  40  No answer  9  /  Agricultural status (%)  Single household unincorporated farm  24  57  Jointly run farm (Groupement Agricole d'Exploitation en Commun (GAEC))  7  6  Private limited farming company (Entreprise Agricole à Responsabilité Limitée(EARL))  40  26  Other  20  11  No answer  9  /  Economic and Technical Orientation (ETO) (%)  Cereals and oilseeds  43  28  Other arable crops  21  32  Wine growing  2  3  Other cultures  1  2  Cattle livestock  3  10  Other livestock  3  11  Livestock and crops combination  13  14  No answer  14  /  Contracts previously signed by farmers (%)  BO contract  2  No data  AES  17  Other (Natura 2000, etc.)  3  No previous contracts  78  Lease duration  18 years  67  No data  9 years  13  Other  4  No answer  16    Farms and farmers in the sample  Farms and farmers in Picardya  Gender (%)  Male  83  No data  Female  9  No answer  8  Age class (%)  [18,39]  20  18  [40,49]  33  29  [50–59]  35  32  >60  6  21  No answer  6  /  Used Agricultural Area (UAA)  Mean UAA (ha)  199  98  Number of exploitation (%):   <50 ha  5  38   [50–100] ha  14  22   >100 ha  72  40  No answer  9  /  Agricultural status (%)  Single household unincorporated farm  24  57  Jointly run farm (Groupement Agricole d'Exploitation en Commun (GAEC))  7  6  Private limited farming company (Entreprise Agricole à Responsabilité Limitée(EARL))  40  26  Other  20  11  No answer  9  /  Economic and Technical Orientation (ETO) (%)  Cereals and oilseeds  43  28  Other arable crops  21  32  Wine growing  2  3  Other cultures  1  2  Cattle livestock  3  10  Other livestock  3  11  Livestock and crops combination  13  14  No answer  14  /  Contracts previously signed by farmers (%)  BO contract  2  No data  AES  17  Other (Natura 2000, etc.)  3  No previous contracts  78  Lease duration  18 years  67  No data  9 years  13  Other  4  No answer  16  aAgricultural census for 2010 (AGRESTE, 2010). Table 2. Sample and population characteristics   Farms and farmers in the sample  Farms and farmers in Picardya  Gender (%)  Male  83  No data  Female  9  No answer  8  Age class (%)  [18,39]  20  18  [40,49]  33  29  [50–59]  35  32  >60  6  21  No answer  6  /  Used Agricultural Area (UAA)  Mean UAA (ha)  199  98  Number of exploitation (%):   <50 ha  5  38   [50–100] ha  14  22   >100 ha  72  40  No answer  9  /  Agricultural status (%)  Single household unincorporated farm  24  57  Jointly run farm (Groupement Agricole d'Exploitation en Commun (GAEC))  7  6  Private limited farming company (Entreprise Agricole à Responsabilité Limitée(EARL))  40  26  Other  20  11  No answer  9  /  Economic and Technical Orientation (ETO) (%)  Cereals and oilseeds  43  28  Other arable crops  21  32  Wine growing  2  3  Other cultures  1  2  Cattle livestock  3  10  Other livestock  3  11  Livestock and crops combination  13  14  No answer  14  /  Contracts previously signed by farmers (%)  BO contract  2  No data  AES  17  Other (Natura 2000, etc.)  3  No previous contracts  78  Lease duration  18 years  67  No data  9 years  13  Other  4  No answer  16    Farms and farmers in the sample  Farms and farmers in Picardya  Gender (%)  Male  83  No data  Female  9  No answer  8  Age class (%)  [18,39]  20  18  [40,49]  33  29  [50–59]  35  32  >60  6  21  No answer  6  /  Used Agricultural Area (UAA)  Mean UAA (ha)  199  98  Number of exploitation (%):   <50 ha  5  38   [50–100] ha  14  22   >100 ha  72  40  No answer  9  /  Agricultural status (%)  Single household unincorporated farm  24  57  Jointly run farm (Groupement Agricole d'Exploitation en Commun (GAEC))  7  6  Private limited farming company (Entreprise Agricole à Responsabilité Limitée(EARL))  40  26  Other  20  11  No answer  9  /  Economic and Technical Orientation (ETO) (%)  Cereals and oilseeds  43  28  Other arable crops  21  32  Wine growing  2  3  Other cultures  1  2  Cattle livestock  3  10  Other livestock  3  11  Livestock and crops combination  13  14  No answer  14  /  Contracts previously signed by farmers (%)  BO contract  2  No data  AES  17  Other (Natura 2000, etc.)  3  No previous contracts  78  Lease duration  18 years  67  No data  9 years  13  Other  4  No answer  16  aAgricultural census for 2010 (AGRESTE, 2010). Fig. 4. View largeDownload slide Distribution of sampled farmers in Picardy, Northern France. Fig. 4. View largeDownload slide Distribution of sampled farmers in Picardy, Northern France. Although we are aware of the risk of weak additionality at the farm-scale, we cannot be sure it will be prevented: 22 per cent of the sampled farmers are already committed to AES schemes or other environmental contracts and may behave opportunistically despite our reminder of the impossibility to use such on-going contract as BO. Most of the farmers in our sample would not try to compensate for their loss of production: 58 per cent of the farmers said they would not compensate the loss of the production at the offset site by increasing the production on other plots on the farm, 19 per cent said they would and 23 per cent did not answer the question. Those who said they would transfer the production may increase the productivity of existing cultivated plots rather than ploughing existing meadows on their farms as there are restrictions to this, especially on designated sensitive meadows. This limits the risk of leakage. Despite all our efforts to improve the level of participation, our response rate is rather low. Examples of low-response rates in other CE surveys include Abildtrup et al. (2013) and Torres et al. (2015) with, respectively, 2.2 per cent and 1.1 per cent of the estimated sample, for studies on forest ecosystem services valuation. Czajkowski, Hanley and Nyborg (2017) had a response rate of 5.23 per cent on household waste recycling behaviour. It is unlikely that the low-response rate is due to the use of a web-based survey or of videos to introduce our CE. Other factors may be involved such as the controversial – and potentially unknown for a share of farmers – nature of the topic, the time of the year chosen for the survey or the potential fatigue of farmers faced with academic enquiries that do not provide any feedback.14 We may have expected that farmers with strong opinions about BO would have responded disproportionately, allowing for possible self-selection into the sample. However, this is not what we have observed: only 11 per cent of the initial sample can be considered as strong opponents to BOs on agricultural lands (the 18 removed protest responses) and we did not have strong advocates since none of the respondents accepted BO contracts in all choice cards. Follow-up questions for checking protest responses (see Section 2.2.3) gave us further insights into farmers’ behaviours. We found that 19 per cent of the 144 farmers systematically chose the opt-out option, without being protest responders, because (i) they prefer to keep their current practices and are unwilling to change whatever the payment or other considerations, (ii) they were not satisfied by the different BO contracts proposed and (iii) the BO contracts are technically not feasible on their lands. The remaining 81 per cent chose the opt-out answer at least once for reasons ii and iii as above but mainly because of the proposed payment level (too low). This shows that farmers in our sample had diverse and justified reasons to accept or refuse BO contracts, reducing the risk of self-selection into the sample. 3.2. Econometric analysis 3.2.1. Farmer’s decision to enrol in a BO contract We first conducted a CL model to analyse the decision of farmers to enrol in a BO contract. The Hausman test, implemented by comparing estimations of the full CL model and partial CL by dropping each time one alternative, showed that the IIA hypothesis is rejected in our sample (with Prob > Chi2 = 0.3101; 0.3725 and 0.4377) making the CL estimations invalid. We thus relaxed the IIA hypothesis and modelled the farmer’s decision to participate in a BO contract with a ML model. The participation is modelled first according to the attributes of BO contracts as sole factors of choice and later by integrating individual characteristics (MLogit 1 and MLogit 2 models in Table 3). As individual variables cannot be integrated alone in the model (they do not vary across one individual), they are captured through interaction variables. The payment and the interaction variables, in model MLogit 2, are fixed and the other attributes are random (normally distributed). Table 3. ML estimates Variables  MLogit 1  MLogit 2  Payment  0.00309***  0.00329***  (0.000226)  (0.000259)  Management plan  −0.325***  −0.722***  (0.0749)  (0.160)  Contract duration  −0.146***  −0.208***  (0.0161)  (0.0272)  Conditional monetary bonus  0.265***  0.258***  (0.0651)  (0.0704)  Opt-out  2.301***  1.941***  (0.367)  (0.395)  Management plan * UAA    0.169***  (0.0556)  Contract duration * partial landowner    0.0585**  (0.0252)  SD   Management plan  0.212**  0.374***  (0.104)  (0.0987)   Contract duration  0.102***  0.135***  (0.0112)  (0.0148)   Conditional monetary bonus  0.184  0.145  (0.127)  (0.136)   Opt-out  2.526***  2.264***  (0.305)  (0.352)  ll  −754.84448  −643.98198  Chi2  453.21231  387.28256  AIC  1527.689  1309.964  BIC  1583.0198  1376.2086  Observations  3456  3048  Number of farmers  144  127  Variables  MLogit 1  MLogit 2  Payment  0.00309***  0.00329***  (0.000226)  (0.000259)  Management plan  −0.325***  −0.722***  (0.0749)  (0.160)  Contract duration  −0.146***  −0.208***  (0.0161)  (0.0272)  Conditional monetary bonus  0.265***  0.258***  (0.0651)  (0.0704)  Opt-out  2.301***  1.941***  (0.367)  (0.395)  Management plan * UAA    0.169***  (0.0556)  Contract duration * partial landowner    0.0585**  (0.0252)  SD   Management plan  0.212**  0.374***  (0.104)  (0.0987)   Contract duration  0.102***  0.135***  (0.0112)  (0.0148)   Conditional monetary bonus  0.184  0.145  (0.127)  (0.136)   Opt-out  2.526***  2.264***  (0.305)  (0.352)  ll  −754.84448  −643.98198  Chi2  453.21231  387.28256  AIC  1527.689  1309.964  BIC  1583.0198  1376.2086  Observations  3456  3048  Number of farmers  144  127  Standard errors in parentheses. Significant levels: ***p < 0.01, **p < 0.05, *p < 0.1. Table 3. ML estimates Variables  MLogit 1  MLogit 2  Payment  0.00309***  0.00329***  (0.000226)  (0.000259)  Management plan  −0.325***  −0.722***  (0.0749)  (0.160)  Contract duration  −0.146***  −0.208***  (0.0161)  (0.0272)  Conditional monetary bonus  0.265***  0.258***  (0.0651)  (0.0704)  Opt-out  2.301***  1.941***  (0.367)  (0.395)  Management plan * UAA    0.169***  (0.0556)  Contract duration * partial landowner    0.0585**  (0.0252)  SD   Management plan  0.212**  0.374***  (0.104)  (0.0987)   Contract duration  0.102***  0.135***  (0.0112)  (0.0148)   Conditional monetary bonus  0.184  0.145  (0.127)  (0.136)   Opt-out  2.526***  2.264***  (0.305)  (0.352)  ll  −754.84448  −643.98198  Chi2  453.21231  387.28256  AIC  1527.689  1309.964  BIC  1583.0198  1376.2086  Observations  3456  3048  Number of farmers  144  127  Variables  MLogit 1  MLogit 2  Payment  0.00309***  0.00329***  (0.000226)  (0.000259)  Management plan  −0.325***  −0.722***  (0.0749)  (0.160)  Contract duration  −0.146***  −0.208***  (0.0161)  (0.0272)  Conditional monetary bonus  0.265***  0.258***  (0.0651)  (0.0704)  Opt-out  2.301***  1.941***  (0.367)  (0.395)  Management plan * UAA    0.169***  (0.0556)  Contract duration * partial landowner    0.0585**  (0.0252)  SD   Management plan  0.212**  0.374***  (0.104)  (0.0987)   Contract duration  0.102***  0.135***  (0.0112)  (0.0148)   Conditional monetary bonus  0.184  0.145  (0.127)  (0.136)   Opt-out  2.526***  2.264***  (0.305)  (0.352)  ll  −754.84448  −643.98198  Chi2  453.21231  387.28256  AIC  1527.689  1309.964  BIC  1583.0198  1376.2086  Observations  3456  3048  Number of farmers  144  127  Standard errors in parentheses. Significant levels: ***p < 0.01, **p < 0.05, *p < 0.1. For both models (with and without interactions) all the contract attributes have significant effects on farmers’ choice. No attribute non-attendances were observed. The model without interaction (MLogit 1) shows that the likelihood to participate decreases with more demanding management plans. Contract duration also has a significant negative effect on farmers’ utility. Participation increases with the conditional bonus and weakly increases with the payment. The fact that the bonus improves the likelihood of signing a contract is consistent with the information that the farmers generally decided to benefit from the bonus15 when they choose a contract including a bonus. The difference from 0 and high significance of the opt-out parameter means that the farmers have a strong preference for keeping their current practices.16 The introduction of individual characteristics (MLogit 2) leads to the removal of 408 observations (linked to 17 individuals) due to missing data but improves the model results (AIC and BIC are lower). Compared to the MLogit 1 model, results from MLogit 2 are similar. Farmers prefer to keep their practices and prefer contracts with limited constraints,17 including the individual monetary bonus, of short duration and well paid. Two interactions with socio-economic characteristics18 have quite significant effects: being a partial or full landowner increases the likelihood to adopt contracts with a longer duration; and having a farm with a larger area (hundreds of ha) increases the likelihood to adopt more constraining measures. Because we have an over-representation of large farms in our sample, it is possible that we underestimate the negative effect of the management plan on the decision to enrol in a BO contract. Also note that if strong opponents to BO were to be under-represented in the sample, this does not change these results because (i) they would have been removed from the econometric analysis for being protest responses and (ii) they would support the main result that farmers seem to have a strong preference for not signing BO contracts. The standard deviation (SD) coefficients reveal that preferences for all the attributes, except for the bonus, are heterogeneous among the farmers of our sample. This confirms it is interesting to consider the distribution of the individual parameters for each attribute that we have calculated using the log likelihood maximisation.19 Figure 5 presents Epanechnikov kernel density plots of the distribution of the individual parameter estimates (Epanechnikov, 1969; Silverman, 1992).20 Fig. 5. View largeDownload slide Distribution of the individual parameter for each attribute and the opt-out. (A) Management plan, (B) contract duration, (C) conditional monetary bonus and (D) opt-out (keep the current practices). Fig. 5. View largeDownload slide Distribution of the individual parameter for each attribute and the opt-out. (A) Management plan, (B) contract duration, (C) conditional monetary bonus and (D) opt-out (keep the current practices). The distributions of the individual parameter show that farmers’ preferences for the bonus and the management plan seem to be concentrated around a single value (Figure 5A and C) while there appears to be at least two groups of preferences for contract duration (Figure 5B). The two groups of the opt-out nicely illustrate that there are farmers that systematically chose to keep their practices (right-hand spike in Figure 5D, corresponding to the 19 per cent of farmers described in Section 3.1) and farmers who chose BO contracts (left-hand spike, corresponding to the remaining 81 per cent). 3.2.2. WTA for implementing BO contracts Farmers’ WTA for implementing different features of BO contracts is presented in Table 4. Table 4. WTA for BO contract implementation Attributes  WTA (€/ha/year)  SD  Min  Max  Management plan  219  54  45  365  Contract duration  63  31  −6  111  Conditional monetary bonusa  −157  18  −207  −98  Opt-out  −590  525  −1936  564  Attributes  WTA (€/ha/year)  SD  Min  Max  Management plan  219  54  45  365  Contract duration  63  31  −6  111  Conditional monetary bonusa  −157  18  −207  −98  Opt-out  −590  525  −1936  564  aEffects coded variable, the WTA must be multiplied by 2 (Le Coënt, Préget and Thoyer, 2017) because we introduced a two-units variation (instead of a one-unit variation) between the two levels of this variable. Table 4. WTA for BO contract implementation Attributes  WTA (€/ha/year)  SD  Min  Max  Management plan  219  54  45  365  Contract duration  63  31  −6  111  Conditional monetary bonusa  −157  18  −207  −98  Opt-out  −590  525  −1936  564  Attributes  WTA (€/ha/year)  SD  Min  Max  Management plan  219  54  45  365  Contract duration  63  31  −6  111  Conditional monetary bonusa  −157  18  −207  −98  Opt-out  −590  525  −1936  564  aEffects coded variable, the WTA must be multiplied by 2 (Le Coënt, Préget and Thoyer, 2017) because we introduced a two-units variation (instead of a one-unit variation) between the two levels of this variable. It is difficult to interpret the value of the WTA for the management plan because we do not know the current practices of each farmer, even if the proposed contracts all focus on converting cultivated arable land. In addition, because this is a mixed attribute, even if it is ordered, the difference between the various levels is not of the same nature. Thus, the WTA for the management plan is €219/ha/year for a composite increase in the level of constraint imposed by our four management plan levels. Farmers are prepared to accept €63/ha/year for a 1-year increase in the duration of the offset contract. Farmers are willing to forgo €157/ha/year to get the bonus: the bonus is worth €200/ha/year so farmers would still benefit. This means that the cost for the developer of implementing supplementary ecological constraints in the management plan is €43/ha/year.21 Farmers are ready to forgo €590/ha/year to keep their current practices, in other words they would only accept contracts that pay more than €590/ha/year. The individual WTAs for the opt-out were calculated from the individual parameters for the opt-out attribute and the parameter for the payment attribute that is unique because it has been set as fixed in the ML. Figure 6 presents Epanechnikov kernel density plots of the distribution of the individual WTA estimates for the opt-out. Fig. 6. View largeDownload slide Distribution of the individual WTA for the opt-out. Fig. 6. View largeDownload slide Distribution of the individual WTA for the opt-out. While the MLogit 2 model shows that farmers on average are ready to forgo €590/ha/year to keep their current practices, Figure 6 shows that farmers who are very reluctant to sign BO contracts have high payment expectations of around €1300/ha/year and that a second group of farmers has expectations of around €500/ha/year, which is actually less than the payments that were proposed in the CE. 3.2.3. Analysing farmers’ decisions on the acreage to enrol in a BO contract We had to remove one farmer (24 observations) from the sample because abnormal areas of offsets were included in this response. Among the choices of the remaining 143 farmers, 530 alternatives other than keeping the current practices were chosen, 506 values of contracted area were completed (24 values were missing) and we could analyse 459 observations (47 values of total UAA were missing so we could not calculate the enrolled proportion of the UAA). These remaining observations are linked to 104 different farmers. Farmers enrolled on average 13 per cent of the UAA (varying from less than 1 per cent to 100 per cent, SD = 25).22 We have no cases where 0 per cent of the farm area is enrolled (y= 0) because we required that farmers enrol at least 0.5 ha, increased to a minimum of 5 ha if the contract includes a conditional bonus. We did not explore the factors which lead farmers to enrol all their UAA (y= 1) because only four farmers did so. In total, we analysed decisions on acreage using 432 observations linked to 100 farmers (Table 5). Table 5. Acreage model estimates Dependent variable: acreage/enrolled proportion of the UAA  Proportion (0 < y< 1)   Management plan  −0.00450  (0.0211)   Contract duration  0.00690  (0.00509)   Conditional monetary bonus  0.0520**  (0.0240)   Payment  1.03e−05  (9.52e−05)   UAA  −0.147**  (0.0611)   Constant  −2.148***  (0.339)  Ln_phi   Constant  1.953***  (0.423)   AIC  −1291.0413   BIC  −1262.5624   Observations  432   Number of farmers  100  Dependent variable: acreage/enrolled proportion of the UAA  Proportion (0 < y< 1)   Management plan  −0.00450  (0.0211)   Contract duration  0.00690  (0.00509)   Conditional monetary bonus  0.0520**  (0.0240)   Payment  1.03e−05  (9.52e−05)   UAA  −0.147**  (0.0611)   Constant  −2.148***  (0.339)  Ln_phi   Constant  1.953***  (0.423)   AIC  −1291.0413   BIC  −1262.5624   Observations  432   Number of farmers  100  Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Table 5. Acreage model estimates Dependent variable: acreage/enrolled proportion of the UAA  Proportion (0 < y< 1)   Management plan  −0.00450  (0.0211)   Contract duration  0.00690  (0.00509)   Conditional monetary bonus  0.0520**  (0.0240)   Payment  1.03e−05  (9.52e−05)   UAA  −0.147**  (0.0611)   Constant  −2.148***  (0.339)  Ln_phi   Constant  1.953***  (0.423)   AIC  −1291.0413   BIC  −1262.5624   Observations  432   Number of farmers  100  Dependent variable: acreage/enrolled proportion of the UAA  Proportion (0 < y< 1)   Management plan  −0.00450  (0.0211)   Contract duration  0.00690  (0.00509)   Conditional monetary bonus  0.0520**  (0.0240)   Payment  1.03e−05  (9.52e−05)   UAA  −0.147**  (0.0611)   Constant  −2.148***  (0.339)  Ln_phi   Constant  1.953***  (0.423)   AIC  −1291.0413   BIC  −1262.5624   Observations  432   Number of farmers  100  Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Among the attributes of the BO, we note that payment has no significant effect on the enrolled acreage even if it has a significant but weak role in the decision to enrol some land. Only the bonus has a significant effect on the likelihood to increase the enrolled percentage of the UAA. We checked if the increase of the enrolled acreage is only linked to the bonus requirement of a minimum of 5 ha enrolled or if there is a true effect of the bonus among the 432 observations. Farmers enrolled a mean of 14 ha ([min = 0.5; max = 150], SD = 25). Table 6 shows that the enrolled acreage is greater when there is a bonus in the proposed BO contract, whether considering the observed or predicted values. Table 6. Number of enrolled hectares in the BO contracts without or with bonus   Contract without conditional monetary bonus (N = 183)  Contract with conditional monetary bonus (N = 249)  Observed individual enrolled number of hectares  13 ha [min = 0.5; max = 140] SD = 23  15 ha [min = 0.5; max = 150] SD = 26  Predicted sample enrolled number of hectares  19 ha [min = 12; max = 26] SD = 3  21 ha [min = 12; max = 29] SD = 3    Contract without conditional monetary bonus (N = 183)  Contract with conditional monetary bonus (N = 249)  Observed individual enrolled number of hectares  13 ha [min = 0.5; max = 140] SD = 23  15 ha [min = 0.5; max = 150] SD = 26  Predicted sample enrolled number of hectares  19 ha [min = 12; max = 26] SD = 3  21 ha [min = 12; max = 29] SD = 3  Table 6. Number of enrolled hectares in the BO contracts without or with bonus   Contract without conditional monetary bonus (N = 183)  Contract with conditional monetary bonus (N = 249)  Observed individual enrolled number of hectares  13 ha [min = 0.5; max = 140] SD = 23  15 ha [min = 0.5; max = 150] SD = 26  Predicted sample enrolled number of hectares  19 ha [min = 12; max = 26] SD = 3  21 ha [min = 12; max = 29] SD = 3    Contract without conditional monetary bonus (N = 183)  Contract with conditional monetary bonus (N = 249)  Observed individual enrolled number of hectares  13 ha [min = 0.5; max = 140] SD = 23  15 ha [min = 0.5; max = 150] SD = 26  Predicted sample enrolled number of hectares  19 ha [min = 12; max = 26] SD = 3  21 ha [min = 12; max = 29] SD = 3  The only individual variable having a significant effect on the enrolled percentage of the UAA is the UAA itself (hundreds of ha). A 100 ha increase in the UAA leads to a 15 per cent decrease of the enrolled percentage of the UAA. We can conclude that there is a threshold effect of the size of the BO area, in absolute and not relative terms, which leads larger farms to enrol a lower share of their UAA. 4. Discussion and conclusion BOs are aimed at achieving NNL of biodiversity in the context of economic development projects, plans or programmes, in principle and under France’s recent biodiversity law. Agricultural landscapes sometimes have low levels of biodiversity and good potential for ecological gains through ecological restoration. However, implementation of BO by farmers remains a controversial topic that has received limited attention in the scientific literature. One may wonder if long-term BO contracts with restrictive conditions for reasons of ecological performance can match farmers’ preferences and constraints. Our study aimed at identifying key factors that explain decisions by farmers to sign a BO contract and that influence the acreage they enrol. We conducted a CE study in Picardy, Northern France, targeting over 5,000 farmers. We received 144 useful responses. The response rate and the size of the sample are quite low. The answers to the follow-up questions included in the survey showed that the risk of self-selection into the sample is limited. However, it is possible that our respondents were better informed of BO than most farmers. Four attributes, describing different scenarios of BO contracts, were selected: the actual ecological restoration activities (in a management plan), the contract duration, the annual payment and the option of receiving a monetary bonus for the extent and spatial distribution of enrolled land. To feedback to our interviewees, we prepared a video23 of the main results of the study that has been shared by the local farmer union-run bodies. While farmers are often targeted by developers for implementing their BO obligations, we show that farmers are quite reluctant to adopt BO contracts which require them to convert arable cropland into grasslands (in our example), and manage these for biodiversity for a defined period of time. All the attributes considered in the CE have a significant effect on the likelihood of choosing a BO contract. The likelihood of choosing a BO measure decreases with increasing levels of constraints on management practices and the duration of the contract. Farmers have, therefore, high WTA for a composite increase in the level of constraint imposed by the management plan or a 1-year increase in contract duration. Higher payment levels and the proposal of a bonus for increased extent of interconnectedness of contracted land improve the likelihood of farmers signing up to a BO contract. Our results show that farmers’ mean WTA to change their practices is even lower than the range of payment we proposed. There is, however, a large spread in payment expectations, with some farmers in the top of the proposed range. We also show that the payment attribute does not significantly lead farmers to enrol a greater share of their UAA, unlike the conditional bonus. This bonus increases the likelihood of signing up a contract and the enrolled acreage but at higher overall cost of a BO contract for the developer. We also show that other reasons limit farmers’ willingness to enrol such as the farm’s total UAA, and that larger farms are likely to accept more constraining management plans. There are policy implications to these results. First, given the reluctance of farmers to enrol their land in complex and long-term management plans, there may be limited scope for involving farmers in the implementation of developers’ BO obligations through contracts. Thus, contracting farmers to restore arable land for biodiversity may be suitable for offsetting short term impacts on already degraded areas of natural habitat (e.g. field margins and recently established permanent grasslands) or modified habitat (e.g. impacts on species that thrive in low-intensity agricultural mosaics), but may not be suitable for long term and permanent impacts on high-quality habitat (e.g. old permanent grasslands). This is because less constraining and shorter contracts will only provide moderate biodiversity gains, which are only suitable for offsetting moderate biodiversity losses (Quétier and Lavorel, 2011). Second, reducing the number of farmers involved in implementing a developer’s BO obligations could reduce their cost, for the developer and for enforcement agencies. This means increasing the enrolled acreage per farmer and our results show that a conditional monetary bonus can favour this, but with an increased per unit area cost for developers. This extra cost can be justified in a context where regulations increasingly make developers liable for the measurable ecological outcomes of their BO (e.g. in France). We also found, however, that there are some limits to this increased enrolled acreage: the mean absolute number of hectares that farmers would be prepared to enrol may not be sufficient to cover needs. In our study, at 15 ha, it is below the threshold that Moreno-Mateos et al. (2012) consider as a lower limit for successful ecological restoration of wetlands. The results of our study may be specific to Europe, where farmers receive CAP subsidies which certainly modify their behaviour. In particular, European farmers are used to short term contracts (5 years) for most agri-environmental schemes (AES) under the CAP. A structural limit to the implementation of BO on arable land is that offsets are supposed to be effective for as long as impacts occur, and this tends to be over long periods of time, and even theoretically into perpetuity (e.g. for many public infrastructure projects). In our study, contract duration is one of the attributes that makes farmers reluctant to sign contracts. We expected this result since other studies with shorter durations made the same observation (Ruto and Garrod, 2009; Christensen et al., 2011; Greiner, Bliemer and Ballweg, 2014). In addition, land tenure patterns (e.g. whether farmers own the land they farm and their capacity to sign contracts with developers) are likely to determine the transferability of our results. In our study, being an owner of part of the cultivated land increased the likelihood to sign a longer contract. Indeed, not being the owner of the land makes it difficult to accept a BO contract because if the lease is shorter than the duration of the BO contract, farmers must be ready to enter discussions and negotiations with the owners of the land they farm, to include contract requirements in their leases (these contracts would have to be transferred to future farmers leasing the land). In France, 60 per cent of the UAA is under a lease (from 9 to 25 years), which could be a considerable impediment to successful (long term) BO implementation on farmland. An alternative is for several different farmers to enrol their leased land into a renewable short term BO contract, with a third-party entity (intermediary) ensuring that the pool of enrolled land fulfils the required ecological gain over time. Le Coënt, Préget and Thoyer (2017) discuss the pros and cons of this approach. This solution may be more or less relevant, depending on the species or habitats targeted by the BO (van Teeffelen et al., 2014; Drechsler, Johst and Wätzold, 2017). Other studies show that contract duration is not a limitation in countries where the proportion of farmers that own the land they farm is higher. This is the case in Australia for pure conservation measures (Greiner, 2016). In the United States, many farmers use BO implementation as an alternative use for their non-profitable arable land. This includes land put into wetland mitigation banks that make offsets available to developers, but require perpetual conservation easements that limit forever most of, or even any, agricultural practices (Vaissière and Levrel, 2015). Another fuzzier reason for the observed reluctance to sign demanding and long-term contracts may be the institutional and environmental uncertainty that farmers face. In our case, we voluntarily did not give details about the possibility to return the land to farming at the end of the contract, because BO is supposed to produce long lasting outcomes. Other CE studies on agri-environmental decisions did consider this (e.g. Lienhoop and Brouwer, 2015). Recent experience may also have made some farmers fearful, specifically temporary meadows under 5-year AES contracts that were subsequently designated as permanent meadows, with strong restrictions on their cultivation under the current CAP. It is unclear how CAP subsidies could or would be articulated with BO contracts (additionality goes both ways) and if BO management plans are compatible with the minimum land management provisions specified in farmland leases. Continued expansion of NNL policies in Europe will make resolving these questions a priority. More generally, our results show that governance (tenure) and economic (cost) criteria should be considered carefully when designing contract-based BO mechanisms that involve farmers. Soule, Tegene and Wiebe (2000) highlighted that adopting conservation practices on farmland is mainly guided by the potential future agronomic value of the land, considering that land owners want to be able to put their land back into agriculture. All the factors that explain farmers’ reluctance to enrol land in BO can be tied to a risk of decrease in the land price if it cannot be cultivated or built anymore. In this uncertain context, the impossibility to stop the contract may be the most limiting factor for signing up to BO (Broch et al. 2013; Greiner, Bliemer and Ballweg, 2014). Farmers may also be justifiably fearful of developers’ capacity to live up to long-term commitments to their BO requirements. Supplementary data Supplementary data are available at European Review of Agricultural Economics online. Funding This work was supported by grants from the Agence Nationale de la Recherche (ANR) as part of the ‘Investissements d’Avenir’ programme (ANR-10-LABX-0004, Lab of Excellence CeMEB; ANR-11-LABX-0002-01, Lab of Excellence ARBRE). This study also received funds from the European Union (FP7/2007–2013) under contract 308393 ‘OPERAs’. Our thanks go to the Université de Montpellier (LAMETA, renamed as CEE-M) and Biotope for supplementary financial support for this work; this paper was part of Anne-Charlotte Vaissière’s post-doctoral fellowship. Acknowledgements We would like to thank Jens Abildtrup, Marion Beaurepaire, Miguel Da Costa Nogueira, Serge Garcia, Laure Kuhfuss, Philippe Le Coënt, Vincent Martinet, Michel Pech, Raphaële Préget, Jean-Marc Rousselle, Sophie Thoyer and Gengyang Tu for their useful comments. The local farmer union-run bodies (Chambres d’Agriculture) of Picardy participated in focus groups and relayed the enquiry to the farmers. We thank them and the farmers for their involvement. Participants at the following Conferences are thanked for their useful comments: the 3rd Annual FAERE Conference (Bordeaux), the 22nd Annual EAERE Conference (Athens) and the 66th Annual AFSE Congress (Nice). We also thank three anonymous reviewers for their valuable comments on previous versions of the manuscript. Footnotes 1 More generally in environmental impact assessment, the loss of ecosystem services is rarely assessed (Tardieu et al., 2015). 2 We use BO to refer to biodiversity offset or biodiversity offsetting, indiscriminately. 3 https://www.eea.europa.eu/data-and-maps/data/copernicus-land-monitoring-service-corine. Accessed 22 February 2018. 4 The ecological efficiency of BOs is determined by their ability to reach the no net loss of biodiversity goal, meaning an in-kind and quantitative ecological equivalence between the impacted and the restored (offset) area with regards to the type of biodiversity (species, habitat types, etc.), its location, timing and duration (accounting for the temporal scale for achieving an ecological response from a restoration action). See Quétier and Lavorel (2011) for a detailed discussion. 5 There is an ecological debate around the interest of having single large versus several small areas for conservation (the SLOSS debate). For instance, Verboom et al. (2001) demonstrate the stabilising role of large patches in marshland (birds habitat) networks but other studies also underline the importance of a network of smaller patches for certain species (e.g. invertebrates in Rösch et al., 2015). In our study, we actually consider both situations, the focus being on having less actors implementing few large or few networks of small areas for conservation. 6 ‘The total area taken up by arable land, permanent grassland, permanent crops and kitchen gardens used by the holding, regardless of the type of tenure or of whether it is used as a part of common land’ (Source: Eurostat, http://ec.europa.eu/eurostat/statistics-explained/index.php/Glossary:Utilised_agricultural_area_(UAA)). Accessed 22 February 2018. 7 The videos are available online for both the offsetting explanations (https://youtu.be/7rXahUmFpM8) and for the choice modelling explanations (https://youtu.be/BRLKNW-84zo) (in French). 8 Note that Navrud (1997) looked at the possibility of using video in describing complex environmental impacts and scenarios in contingent valuation studies. He showed that video can be an effective communication device as this is preferred by the respondents compared with other communication devices (text and photos), even though no significant differences in willingness-to-pay were identified. 9 We asked the farmers to only consider arable land for implementing the BO in order to reduce the heterogeneity of land they could have considered (e.g. unused fallows, permanent meadows or shrubland, and other uncultivated land), and to have offsets that could likely fit all the situations in our study area (unlike wetland restoration, for instance). This reduces the heterogeneity in opportunity costs and ecological restoration potential. 10 The full factorial design generates (4 × 4 × 2 × 4) x ((4 × 4 × 2 × 4)−1) = 16,256 choice cards. 11 To minimise the approximation error when calculating the Bayesian efficiency, we use a Gaussian quadrature. Because we have six priors parameters, the Gaussian quadrature might need a large number of draws so we decided to include; bdraws = gauss(3) in our design. 12 The other possible answers were: ‘I prefer the current use of my land’, ‘I did not like any of the contracts’, ‘the measures are not feasible in my farm’, ‘I do not have the equipment to implement the management plan for the meadow’, ‘the payment is too low’, ‘no answer’. 13 The other possible answers were: ‘I did not like any of the contracts’, ‘the two proposed measures are not feasible in my farm’, ‘the payment is too low’, ‘no answer’. 14 This risk may remain even though we explained at both the beginning and the end of the survey that farmers would receive the results from the study within the year, which they did (see Section 4). 15 Among the 312 choices made that included the conditional monetary bonus: farmers decided to benefit from the bonus in 67% of the choices; they will not benefit from the €200/ha/year because a lower than 5 ha surface was entered in 25% of the choices (in such a case during the enquiry, an error message displayed to inform the farmer that with the proposed number of ha, he/she would not benefit from the bonus to make sure the farmer understood the principle); they refused the bonus even if a higher than 5 ha surface was entered (which means that the farmer does not accept the ecological constraints of the bonus) in only 3% of the choices; and they did not answer the question, that was not mandatory, in 5% of the choices. 16 The opt-out variable is how we interpreted the Alternative Specific Constant (ASC) variable in these ML models. 17 We previously included the management plan variable as three dummy variables, with Level I set as the reference. All the contract attributes had significant effects on farmers’ choice and the parameters for the four levels of the management plan variable were ordered. The farmers have a preference for Level I over Level II, a preference for Level II over Level III and a preference for Level III over Level IV. In other words, they have a preference for the less restrictive management plans over the more restrictive ones. It is thus appropriate to interpret the management plan variable as a quantitative variable: it is continuous and ordered but not linear since the difference between each of the four levels is not similar (stop fertilising from Level I to II, delay the mowing of one more month from Level II to III, set a refuge zone from Level III to IV). 18 The other individual characteristics that were tested but that had no significant effect are: having lands already enrolled in BO or AES, gender, age class, education, status of agricultural company, department, lease duration, Economic and Technical Orientation (ETO), agricultural types (conventional, organic, etc.), income, being part of an environmental NGO, being a hunter. 19 We used the mixlbeta command following the use of the mixlogit one (STATA 14). 20 The kdensity command has been used here (STATA 14). 21 This amount is defined as the €200/ha/year of monetary bonus less the €157/ha/year the farmers are ready to forgo. 22 We used the log likelihood maximisation through the zoib command (STATA 14). This command is linked to a full zero one inflated beta model but we only used its proportion part. 23 The video is available online (https://youtu.be/9W8TvXLA7bU) (in French). References Abildtrup, J., Garcia, S., Olsen, S. B. and Stenger, A. ( 2013). Spatial preference heterogeneity in forest recreation. Ecological Economics  92: 67– 77. Google Scholar CrossRef Search ADS   Adamowicz, W. and Boxall, P., 2001. Future directions of stated choice methods for environment valuation. Paper prepared for ‘Choice Experiments: A New Approach to Environmental Valuation’ Conference, 10 April 2001, London. Adamowicz, W., Boxall, P., Williams, M. and Louvière, J. ( 1998). Stated preference approaches for measuring passive use values: choice experiments and contingent valuation. American Journal of Agricultural Economics  80( 1): 64– 75. Google Scholar CrossRef Search ADS   AGRESTE, 2010. Mémento de la statistique agricole. Edition 2010. Picardie. Banerjee, S., deVries, F. P., Hanley, N. and van Soest, D. P. ( 2014). The impact of information provision on agglomeration bonus performance: an experimental study on local networks. American Journal of Agricultural Economics  96( 4): 1009– 1029. Google Scholar CrossRef Search ADS   Barrio, M. and Loureiro, M. ( 2013). The impacts of protests responses in choice experiments: an application to a biosphere reserve management program. Forest Systems  22( 1): 94– 105. Google Scholar CrossRef Search ADS   Bateman, I. J., Day, B. H., Jones, A. P. and Jude, S. ( 2009). Reducing gain–loss asymmetry: a virtual reality choice experiment valuing land use change. Journal of Environmental Economics and Management  58( 1): 106– 118. Google Scholar CrossRef Search ADS   Bech, M. and Gyrd‐Hansen, D. ( 2005). Effects coding in discrete choice experiments. Health Economics  14( 10): 1079– 1083. Google Scholar CrossRef Search ADS   Bennett, G., Chavarria, A., Ruef, F. and Leonardi, A. ( 2017). State of European Markets 2017 Biodiversity Offsets and Compensation . Washington: Ecosystem Marketplace, 36. Bougherara, D. and Ducos, G., 2006. Farmers’ preferences over compensation contract flexibility and duration: an estimation of the effect of transaction costs using choice experiment. First Meeting of the European School on New Institutional Economics (ESNIE), Université de Paris-Sud XI, France, 26 pp. Broch, S. W., Strange, N., Jacobsen, J. B. and Wilson, K. A. ( 2013). Farmers’ willingness to provide ecosystem services and effects of their spatial distribution. Ecological Economics  92: 78– 86. Google Scholar CrossRef Search ADS   Broch, S. W. and Vedel, S. E. ( 2012). Using choice experiments to investigate the policy relevance of heterogeneity in farmer agri-environmental contract preferences. Environmental and Resource Economics  51( 4): 561– 581. Google Scholar CrossRef Search ADS   Buckley, C., Hynes, S. and Mechan, S. ( 2012). Supply of an ecosystem service – farmers’ willingness to adopt riparian buffer zones in agricultural catchments. Environmental Science & Policy  24: 101– 109. Google Scholar CrossRef Search ADS   Bull, J. W., Suttle, K. B., Gordon, A., Singh, N. J. and Milner-Gulland, E. J. ( 2013). Biodiversity offsets in theory and practice. Oryx  47( 3): 369– 380. Google Scholar CrossRef Search ADS   Calvet, C., Ollivier, G. and Napoléone, C. ( 2015). Tracking the origins and development of biodiversity offsetting in academic research and its implications for conservation: a review. Biological Conservation  192: 492– 503. Google Scholar CrossRef Search ADS   Chaloner, K. and Verdinelli, I. ( 1995). Bayesian experimental design: a review. Statistical Science  10( 3): 273– 304. Google Scholar CrossRef Search ADS   Chen, X., Lupi, F., He, G. and Liu, J. ( 2009). Linking social norms to efficient conservation investment in payments for ecosystem services. Proceedings of the National Academy of Sciences of the United States of America (PNAS)  106( 28): 11812– 11817. Google Scholar CrossRef Search ADS   Christensen, T., Pedersen, A. B., Nielsen, H. O., Mørkbak, M. R., Hasler, B. and Denver, S. ( 2011). Determinants of farmers’ willingness to participate in subsidy schemes for pesticide-free buffer zones – a choice experiment study. Ecological Economics  70( 8): 1558– 1564. Google Scholar CrossRef Search ADS   Czajkowski, M., Hanley, N. and Nyborg, K. ( 2017). Social norms, morals and self-interest as determinants of pro-environment behaviours: the case of household recycling. Environmental and Resource Economics  66( 4): 647– 670. Google Scholar CrossRef Search ADS   de Billy, V., Tournebize, J., Barnaud, G., Benoît, M., Birgand, F., Garnier, J., Lesaffre, B., Lévêque, C., de Marsily, G., Muller, S., Musy, A. and Zimmer, D. ( 2015). Compenser la Destruction de Zones Humides. Retours d’Expériences sur les Méthodes et Réflexions Inspirées par le Projet d’Aéroport de Notre-Dame-des-Landes (France). Nature Sciences Sociétés  23( 1): 27– 41. Google Scholar CrossRef Search ADS   Drechsler, M., Johst, K. and Wätzold, F. ( 2017). The cost-effective length of contracts for payments to compensate land owners for biodiversity conservation measures. Biological Conservation  207: 72– 79. Google Scholar CrossRef Search ADS   Drechsler, M., Wätzold, F., Johst, K. and Shogren, J. F. ( 2010). An agglomeration payment for cost-effective biodiversity conservation in spatially structured landscapes. Resource and Energy Economics  32( 2): 261– 275. Google Scholar CrossRef Search ADS   Epanechnikov, V. A. ( 1969). Non-parametric estimation of a multivariate probability density. Theory of Probability and Its Applications  14( 1): 153– 158. Google Scholar CrossRef Search ADS   Espinosa-Goded, M., Barreiro-Hurlé, J. and Ruto, E. ( 2010). What do farmers want from agri-environmental scheme design? A choice experiment approach. Journal of Agricultural Economics  61( 2): 259– 273. Google Scholar CrossRef Search ADS   Etrillard, C. and Pech, M. ( 2015). Mesures de Compensation Ecologique: Risques ou Opportunités pour le Foncier Agricole en France ? VertigO  15( 2). https://journals.openedition.org/vertigo/16450 Ferrari, S. L. P. and Cribari-Neto, F. ( 2004). Beta regression for modelling rates and proportions. Journal of Applied Statistics  31( 7): 799– 815. Google Scholar CrossRef Search ADS   Gombault, V., 2013. L’internet de plus en plus prisé, l’internaute de plus en plus mobile, INSEE Première 1452. Gonthier, D. J., Ennis, K. K., Farinas, S., Hsieh, H. Y., Iverson, A. L., Batáry, P., Rudolphi, J., Tscharntke, T., Cardinale, B. J. and Perfecto, I. ( 2014). Biodiversity conservation in agriculture requires a multi-scale approach. Proceedings of the Royal Society of London B: Biological Sciences  281( 1791): 20141358. doi:10.1098/rspb.2014.1358 Google Scholar CrossRef Search ADS   Gonçalves, B., Marques, A., Velho Da Maia Soares, M. A. and Pereira, H. M. ( 2015). Biodiversity offsets: from current challenges to harmonized metrics. Current Opinion in Environmental Sustainability  14: 61– 67. Google Scholar CrossRef Search ADS   Greiner, R. ( 2016). Factors influencing farmers’ participation in contractual biodiversity conservation: a choice experiment with northern Australian pastoralists. Australian Journal of Agricultural and Resource Economics  60( 1): 1– 21. Google Scholar CrossRef Search ADS   Greiner, R., Bliemer, M. and Ballweg, J. ( 2014). Design considerations of a choice experiment to estimate likely participation by north Australian pastoralists in contractual biodiversity conservation. Journal of Choice Modelling  10: 34– 45. Google Scholar CrossRef Search ADS   Haaijer, R., Kamakura, W. and Wedel, M. ( 2001). The ‘no-choice’ alternative in conjoint choice experiments. International Journal of Market Research  43: 93– 106. Hasan-Basri, B. and Karim, M. Z. A. ( 2013). The effects of coding on the analysis of consumer choices of public parks. World Applied Sciences Journal  22( 4): 500– 505. Jacob, C., Vaissière, A.-C., Bas, A. and Calvet, C. ( 2016). Investigating the inclusion of ecosystem services in ′ting. Ecosystem Services  21: 92– 102. Google Scholar CrossRef Search ADS   Kaczan, D., Swallow, B. M. and Adamowicz, W. L. ( 2013). Designing a payment for ecosystem services (PES) program to reduce deforestation in Tanzania: an assessment of payment approaches. Ecological Economics  95: 20– 30. Google Scholar CrossRef Search ADS   Kehlbacher, A., Balcombe, K. and Bennett, R. ( 2013). Stated attribute non‐attendance in successive choice experiments. Journal of Agricultural Economics  64( 3): 693– 706. Google Scholar CrossRef Search ADS   Kiesecker, J. M., Copeland, H., Pocewicz, A. and McKenney, B. ( 2010). Development by design: blending landscape level planning with the mitigation hierarchy. Frontiers in Ecology and the Environment  8: 261– 266. Google Scholar CrossRef Search ADS   Kontoleon, A. and Yabe, M. ( 2003). Assessing the impacts of alternative ‘opt-out’ formats in choice experiment studies: consumer preferences for genetically modified content and production information in food. Journal of Agricultural Policy and Research  5: 1– 43. Kuhfuss, L., Préget, R. and Thoyer, S. ( 2014). Préférences Individuelles et Incitations Collectives: Quels Contrats Agroenvironnementaux pour la Réduction des Herbicides par les Viticulteurs? Revue d’Etudes en Agriculture et Environnement / Review of Agricultural and Environmental Studies  95( 1): 111– 143. Google Scholar CrossRef Search ADS   Kuhfuss, L., Préget, R., Thoyer, S. and Hanley, N. ( 2016). Nudging farmers to enrol land into agri-environmental schemes: the role of a collective bonus. European Review of Agricultural Economics  43( 4): 609– 636. Google Scholar CrossRef Search ADS   Lancaster, K. ( 1966). A new approach to consumer theory. Journal of Political Economy  74: 132– 157. Google Scholar CrossRef Search ADS   Le Coënt, P., Préget, R. and Thoyer, S. ( 2017). Compensating environmental losses versus creating environmental gains: implications for biodiversity offsets. Ecological Economics  142: 120– 129. Google Scholar CrossRef Search ADS   Levrel, H., Frascaria-Lacoste, N., Hay, J., Martin, G., and Pioch, S. (eds.), 2015. Restaurer la nature pour atténuer les impacts du développement – Analyse des mesures compensatoires pour la biodiversité. Editions QUAE. Lienhoop, N. and Brouwer, R. ( 2015). Agri-environmental policy valuation: farmers’ contract design preferences for afforestation schemes. Land-Use Policy  42: 568– 577. Google Scholar CrossRef Search ADS   Luce, R. D. ( 1959). Individual Choice Behavior: A Theoretical Analysis . New York: Wiley. Maron, M., Brownlie, S., Bull, J. W., Evans, M. C., von Hase, A., Quétier, F., Watson, J. E. M. and Gordon, A. ( 2018). The many meanings of no net loss in environmental policy. Nature Sustainability  1: 19– 27. Google Scholar CrossRef Search ADS   Matthews, Y., Scarpa, R. and Marsh, D. ( 2017). Using virtual environments to improve the realism of choice experiments: a case study about coastal erosion management. Journal of Environmental Economics and Management  81: 193– 208. Google Scholar CrossRef Search ADS   McFadden, D. ( 1974). Conditional logit analysis of qualitative choice behavior. In: Zarembka P. (ed.), 1974. Frontiers in Econometrics . New York: Academic Press, 105– 142. MEDDE, 2013. Lignes Directrices Nationales sur la Séquence Eviter, Réduire et Compenser les Impacts sur les Milieux Naturels. MEDTL, 2012. Doctrine Relative à la Séquence Eviter, Réduire et Compenser les Impacts sur le Milieu Naturel. Menz, M. H., Dixon, K. W. and Hobbs, R. J. ( 2013). Hurdles and opportunities for landscape-scale restoration. Science  339( 6119): 526– 527. Google Scholar CrossRef Search ADS   Moreno-Mateos, D., Power, M. E., Comín, F. A. and Yockteng, R. ( 2012). Structural and functional loss in restored wetland ecosystems. PLoS Biology  10( 1): e1001247. Google Scholar CrossRef Search ADS   Navrud, S. ( 1997). Communication devices in contingent valuation surveys—Experiments with video. In: R. J. Kopp, W. W. Pommerehne and N. Schwarz (eds), Determining the Value of Non-Marketed Goods. Studies in Risk and Uncertainty 10 . Dordrecht: Springer, 273– 298. Google Scholar CrossRef Search ADS   Paolino, P. ( 2001). Maximum likelihood estimation of models with beta-distributed dependent variables. Political Analysis  9( 4): 325– 346. Google Scholar CrossRef Search ADS   Parkhurst, G. and Shogren, J. ( 2007). Spatial incentives to coordinate contiguous habitat. Ecological Economics  64( 2): 344– 355. Google Scholar CrossRef Search ADS   Parkhurst, G., Shogren, J., Bastian, C., Kivi, P., Donner, J. and Smith, R. ( 2002). Agglomeration bonus: an incentive mechanism to reunite fragmented habitat for biodiversity conservation. Ecological Economics  41( 2): 305– 328. Google Scholar CrossRef Search ADS   Patterson, Z., Darbani, J. M., Rezaei, A., Zacharias, J. and Yazdizadeh, A. ( 2017). Comparing text-only and virtual reality discrete choice experiments of neighbourhood choice. Landscape and Urban Planning  157: 63– 74. Google Scholar CrossRef Search ADS   Pe’er, G., Zinngrebe, Y., Hauck, J., Schindler, S., Dittrich, A., Zingg, S., Tscharntke, T., Oppermann, R., Sutcliffe, L. M. E., Sirami, C., Schmidt, J., Hoyer, C., Schleyer, C. and Lakner, S. ( 2017). Adding some green to the greening: improving the EU’s ecological focus areas for biodiversity and farmers. Conservation Letters  10( 5): 517– 530. Google Scholar CrossRef Search ADS   Quétier, F. and Lavorel, S. ( 2011). Assessing ecological equivalence in biodiversity offset schemes: key issues and solutions. Biological Conservation  144: 2991– 2999. Google Scholar CrossRef Search ADS   Quétier, F., Regnery, B. and Levrel, H. ( 2014). No net loss of biodiversity or paper offsets? A critical review of the French no net loss policy. Environmental Science and Policy  38: 120– 131. Google Scholar CrossRef Search ADS   Rambonilaza, M. and Dachary-Bernard, J. ( 2007). Land-use planning and public preferences: what can we learn from choice experiment method? Landscape and Urban Planning  83( 4): 318– 326. Google Scholar CrossRef Search ADS   Revelt, D. and Train, K. ( 2000). Customer-specific taste parameters and mixed logit: households’ choice of electricity supplier. Working Paper E00-274, Department of Economics . Berkeley: University of California. Rose, J. M. and Bliemer, M. C. J. ( 2013). Sample size requirements for stated choice experiments. Transportation  40( 5): 1021– 1041. Google Scholar CrossRef Search ADS   Ruto, E. and Garrod, G. ( 2009). Investigating farmers’ preferences for the design of agri-environment schemes: a choice experiment approach. Journal of Environmental Planning and Management  52( 5): 631– 647. Google Scholar CrossRef Search ADS   Rösch, V., Tscharntke, T., Scherber, C. and Batáry, P. ( 2015). Biodiversity conservation across taxa and landscapes requires many small as well as single large habitat fragments. Oecologia  179( 1): 209– 222. Google Scholar CrossRef Search ADS   Santos, R., Clemente, P., Brouwer, R., Antunes, P. and Pinto, R. ( 2015). Landowner preferences for agri-environmental agreements to conserve the montado ecosystem in Portugal. Ecological Economics  118: 159– 167. Google Scholar CrossRef Search ADS   Silverman, B. W. ( 1992). Density Estimation for Statistics and Data Analysis . London: Chapman & Hall. Soule, M., Tegene, A. and Wiebe, K. D. ( 2000). Land tenure and the adoption of conservation practices. American Journal of Agricultural Economics  84( 2): 993– 1005. Google Scholar CrossRef Search ADS   Stoate, C., Boatman, N., Borralho, R., Carvalho, C. R., Snoo, G. R. and Eden, P. ( 2001). Ecological impacts of arable intensification in Europe. Journal of Environmental Management  63: 337– 365. Google Scholar CrossRef Search ADS   Stoate, C., Báldi, A., Beja, P., Boatman, N. D., Herzon, I., van Doorn, A., de Snoo, G. R., Rakosy, L. and Ramwell, C. ( 2009). Ecological impacts of early 21st century agricultural change in Europe – a review. Journal of Environmental Management  91: 22– 46. Google Scholar CrossRef Search ADS   Tardieu, L., Roussel, S., Thompson, J. D., Labarraque, D. and Salles, J.-M. ( 2015). Combining direct and indirect impacts to assess ecosystem service loss due to infrastructure construction. Journal of Environmental Management  152: 145– 157. Google Scholar CrossRef Search ADS   Torres, A. B., MacMillan, D. C., Skutsch, M. and Lovett, J. C. ( 2015). ‘Yes-in-my-backyard’: spatial differences in the valuation of forest services and local co-benefits for carbon markets in México. Ecological Economics  109: 130– 141. Google Scholar CrossRef Search ADS   Train, K. E. ( 2009). Discrete Choice Methods with Simulation , 2nd edn. Cambridge, UK: Cambridge University Press. Google Scholar CrossRef Search ADS   Tucker, G. M., Allen, B., Conway, M., Dickie, I., Hart, K., Rayment, M., Schulp, C. and van Teeffelen, A., 2013. Policy Options for an EU No Net Loss Initiative. Report to the European Commission. Institute for European Environmental Policy, London. Vaissière, A. C. and Levrel, H. ( 2015). Biodiversity offset markets: what are they really? An empirical approach to wetland mitigation banking. Ecological Economics  110: 81– 88. Google Scholar CrossRef Search ADS   van Teeffelen, A. J., Opdam, P., Wätzold, F., Hartig, F., Johst, K., Drechsler, M., Vos, C. C., Wissel, S. and Quétier, F. ( 2014). Ecological and economic conditions and associated institutional challenges for conservation banking in dynamic landscapes. Landscape and Urban Planning  130: 64– 72. Google Scholar CrossRef Search ADS   Verboom, J., Foppen, R., Chardon, P., Opdam, P. and Luttikhuizen, P. ( 2001). Introducing the key patch approach for habitat networks with persistent populations: an example for marshland birds. Biological Conservation  100( 1): 89– 101. Google Scholar CrossRef Search ADS   Vermeulen, B., Goos, P. and Vandebroek, M. ( 2008). Models and optimal designs for conjoint choice experiments including a no-choice option. International Journal of Research in Marketing  25: 94– 103. Google Scholar CrossRef Search ADS   Wende, W., Tucker, G., Quétier, F., Rayment, M. and Darbi, M. (eds) ( 2018). Biodiversity Offsets – European Perspectives on No Net Loss of Biodiversity and Ecosystem Services . Cham: Springer. Google Scholar CrossRef Search ADS   Wätzold, F. and Drechsler, M. ( 2014). Agglomeration payment, agglomeration bonus or homogeneous payment? Resource and Energy Economics  37: 85– 101. Google Scholar CrossRef Search ADS   Author notes Review coordinated by Ada Wossink © Oxford University Press and Foundation for the European Review of Agricultural Economics 2018; all rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png European Review of Agricultural Economics Oxford University Press

Preferences for biodiversity offset contracts on arable land: a choice experiment study with farmers

Loading next page...
 
/lp/ou_press/preferences-for-biodiversity-offset-contracts-on-arable-land-a-choice-f7eDA0RWPX
Publisher
Oxford University Press
Copyright
© Oxford University Press and Foundation for the European Review of Agricultural Economics 2018; all rights reserved. For permissions, please e-mail: journals.permissions@oup.com
ISSN
0165-1587
eISSN
1464-3618
D.O.I.
10.1093/erae/jby006
Publisher site
See Article on Publisher Site

Abstract

Abstract Biodiversity offsetting (BO) is aimed at achieving no net loss of biodiversity in the context of economic development. Through a choice experiment in Northern France, we show that farmers have a clear preference for not signing up BO contracts. The contracts they accept may only be suitable for offsetting temporary impacts on already degraded areas of natural habitat but not for permanent impacts on high-quality habitat. We find that the introduction of a conditional monetary bonus can improve the organisational and ecological efficiency of BO because it increases the enrolled acreage in a BO contract per farmer albeit at an increased cost for developers. 1. Introduction Development projects (e.g. real estate projects, the construction of transport infrastructure) often lead to the destruction or degradation of natural and semi-natural ecosystems, with consequent impacts on biodiversity, ecological processes and associated ecosystem services. Even where developers must follow a mitigation hierarchy that includes measures to first avoid and then reduce their potential impacts on biodiversity, significant residual impacts on ecosystems and species often remain. Offsetting, the last step of the mitigation hierarchy, is supposed to compensate for residual ecological losses, in order to achieve no net loss (NNL) of biodiversity once the project and its mitigation and offsetting measures have been successfully carried out (Kiesecker et al., 2010; Bull et al., 2013; Maron et al., 2018). Current offset practices rely mainly on a biophysical1 approach (Jacob et al., 2016). Given their NNL goal, biodiversity offsets (BO)2 must provide measurable ecological gains, through ecosystem management activities, that are equivalent to the ecological losses in the impacted area. To be qualified as gains, the activities must be additional to existing biodiversity conservation activities and their duration must be proportionated to that of impacts. BO schemes can be implemented by the developers themselves or by others acting on their behalf, including farmers. In the latter case, farmers are paid for voluntarily implementing the BO on behalf of public or private developers (who, in France, remain the environmental permit holder). The mitigation hierarchy and BO have been promoted as an innovative mechanism for achieving sustainable development goals and are finding their way into environmental regulations in many countries (e.g. Calvet, Ollivier and Napoléone, 2015; Gonçalves et al., 2015; Bennett et al., 2017; Wende et al., 2018). In France, however, the implementation of BO has been highly variable and often ineffective since the concept was introduced in the 1976 Nature Protection Act (Quétier, Regnery and Levrel, 2014). The French Ministry of Environment published several guidelines in recent years to improve implementation of the mitigation hierarchy (e.g. MEDTL, 2012; MEDDE, 2013). These guidance documents were partly enacted in August 2016 when a new Law on Biodiversity was passed. This is still recent which creates a somewhat fuzzy context for BO design and implementation. Across Europe, the growing uptake of NNL policies means that BO contracts are increasingly offered to farmers (Wende et al., 2018). This is particularly relevant in France, where agriculture is the dominant land use (59.5 per cent of the country’s land area, 2012 CORINE Land Cover inventory3) and farmers are likely to be approached by developers for siting development projects or for BO on farmland. However, one can question if BO requirements for achieving NNL are compatible with the constraints and interests of farmers. First, BO requirements are much more restrictive for agricultural activity than the widespread agri-environment schemes (AES) and they include longer durations of contracts, from which it is not possible to withdraw without stiff penalties. Second, there are limits to the social acceptance of BO on agricultural land. Objections by farmers relate mainly to (i) the use of arable land for development projects, (ii) the use of arable land for offsetting and (iii) the resulting pressure on arable land prices (de Billy et al., 2015; Etrillard and Pech, 2015). Some farmers, however, see BO as an opportunity to diversify their activities by implementing biodiversity-friendly activities on their land, on behalf of developers, without selling the land itself (Etrillard and Pech, 2015). The political discourse of most farm union-run bodies in France is currently shifting from opposition to a call for better farmer engagement in BO on agricultural land. A parallel question is how to organise the ecological supply for a BO linked to a given development project. Across France, BOs are currently implemented piecemeal, through the purchase and ad hoc contracting of a multitude of small sized parcels of land, including farmland. This is costly for developers to put in place and manage, makes it difficult for regulators to control and monitor BO performance and affects the ecological efficiency4 (Quétier, Regnery and Levrel, 2014). It has been shown that, in the context of BO policies, a smaller number of agents involved in the delivery of BO activities (e.g. ecosystem restoration) reduces the transaction costs of BO policy implementation (e.g. Levrel et al., 2015; Vaissière and Levrel, 2015). Moreover, large scale5 ecological restoration actions have been proven to be more effective for a wide range of ecosystems and locations (e.g. Menz, Dixon and Hobbs, 2013;,Moreno-Mateos et al., 2012 for wetlands). Large patches and a highly connected habitat network are criteria in ‘banking’ approaches for BO implementation, where activities are anticipated and pooled (aggregated) (van Teeffelen et al., 2014). Against this background, knowing the conditions under which farmers would agree to increase the acreage of land enrolled in a BO on their farm is relevant as a way to reduce the number of stakeholders involved in the implementation of a BO. In summary, in this paper, we address the need for objective and informative research on (i) the conditions under which farmers would be prepared to enrol in BO activities (or not) and (ii) the factors that influence the enrolled share of their cultivated land. Our results have policy relevance against the on-going discussion of the various technical and organisational solutions for achieving NNL in France, and more broadly in Europe (Tucker et al., 2013; Wende et al., 2018). Our results are also informative in view of the next reform of the common agricultural policy (CAP) (Pe’er et al., 2017). We used a choice experiment (CE) to assess farmer preferences for the implementation of BO contracts on arable land in Picardy, Northern France. In our experiment, BO proposals vary with different levels of contract attributes that we selected with the help of focus groups: management plan of the land restoration, contract duration, the annual payment and an additional conditional bonus. This bonus consists of a side payment offered in addition to the baseline payment. It is only paid if a farmer enrols a minimum contiguous land area on his farm. We also calculated farmers’ willingness to accept (WTA in €/ha/year) for different BO contract features. CEs are increasingly used in various research areas (transport economics, marketing, environmental economics, etc.), including the analysis of farmers’ preferences for AES. These studies seek to understand which parameters would be considered for improving participation in AES contracts, and to understand stakeholders’ preferences for biodiversity conservation schemes (e.g. Broch and Vedel, 2012; Lienhoop and Brouwer, 2015; Greiner, 2016). In addition to contract length and the payment level, which are frequently tested, many attributes related to the flexibility of the contracts are tested. These may involve technical restrictions (Bougherara and Ducos, 2006; Espinosa-Goded, Barreiro-Hurlé and Ruto 2010; Christensen et al., 2011; Kuhfuss, Préget and Thoyer, 2014; Greiner, 2016), the parcels involved (Ruto and Garrod, 2009), bonuses for agri-environmental contracts with a collective dimension (Chen et al., 2009; Kuhfuss et al., 2016) or the possibility to opt-out of the contract (Broch et al., 2013; Greiner, Bliemer and Ballweg, 2014). The administrative burden is also often studied (Bougherara and Ducos, 2006; Ruto and Garrod, 2009; Christensen et al., 2011). Additional attributes linked to the specific case studies include the width of pesticide free buffer zones (Christensen et al., 2011), cattle and tree density in measures for the protection of a specific forest ecosystem (Santos et al., 2015), the option of localised herbicide use in vineyards (Kuhfuss et al., 2016), the opportunity to return to agriculture after the contract ends (Lienhoop and Brouwer, 2015) or afforestation purposes (Broch and Vedel, 2012). To our knowledge, there are only a few CE studies dealing with the implementation of BO on arable land that assess farmers’ preferences. Le Coënt, Préget and Thoyer (2017), for example, studied farmers’ participation in agri-environmental contracts framed as BO versus a traditional biodiversity conservation programme (not tied to an ecological loss). They show that farmers have different preferences for these two situations and that their monetary demands are higher for BO contracts. Our approach differs from previous studies by considering connection and/or agglomeration at the farm level at a bonus payment for acreage enrolled in the BO. We do not seek to investigate coordination among neighbours but we are interested in a landscape-scale approach as promoted in the context of agri-environmental schemes (e.g. Gonthier et al., 2014). To our knowledge, our study is the first to use this type of bonus in a CE. Previous studies that have tested bonuses aim at the number of signed AES (e.g. Buckley, Hynes and Mechan, 2012; Banerjee et al., 2014; Kuhfuss et al., 2016) so that the number of farmers under contract reaches a predetermined threshold within a given geographical area. Other studies involve conservation or restoration contracts that are not limited to agricultural land. A significant proportion of this literature is based on laboratory experiments, with few empirical studies (Parkhurst et al., 2002; Parkhurst and Shogren, 2007; Drechsler et al., 2010; Wätzold and Drechsler, 2014). Kuhfuss et al. (2016) studied (empirically) the effect of nudging respondents to increase land enrolment. 2. Materials and methods 2.1. Methodological options 2.1.1. Modelling farmers’ decisions to enrol in a BO contract Farmers’ decision to enrol or not in a BO contract will result from the comparison of the utility they derive from different alternatives. Following Lancaster’s theory (1966) and the random utility theory (Luce, 1959; McFadden, 1974), farmer i (i = 1, …, n) will choose alternative j (j = 1, …, m) in choice card Ct (t = 1, …, T) if the alternative is the one that provides him/her the highest level of utility among all m alternatives proposed in the choice card. The utility is defined by an observable part and a random part represented by error terms. In the conditional logit (CL) model, it is supposed that the error terms are independently and identically distributed (IID) among the alternatives and across the population and that irrelevant alternatives are independent (IIA). If Ajit is a dummy variable that takes the value of 1 if alternative j is chosen by farmer i in the choice card Ct, the probability related to this choice is (see Kuhfuss et al., 2016)   P(Ajit=1)=exp⁡(X′jitβ)∑m∈Ctexp⁡(X′mitβ) (1)where Xjit denote the attributes of alternative j faced by farmer i and β is the vector of k preference parameters, representing the average importance of each attribute of the BO contract on farmers’ preferences. The hypothesis of IIA is a strong assumption. The mixed logit (ML) model relaxes this assumption and allows assessing the βki that are specific to each interviewee, and randomly distributed across the population, with a density function f(βk). It thus captures heterogeneity of farmer’s preferences. Then, conditional on vector βi, the probability that farmer i chooses alternative j in choice card Ct is (see Kuhfuss et al., 2016)   P(Ajit=1|βi)=exp⁡(X′jitβi)∑m∈Ctexp⁡(X′mitβi). (2) Thus, the probability of observing the sequence of T choices by individual i is   P(Aji1=1,…,AjiT=1)=∫∏t=1T(exp(X′jitβ)∑m∈Ctexp(X′mitβ))f(β)dβ (3)where f(β) can be specified to be normal or lognormal: β~N(b,W) or lnβ~N(b,W). The mean b and the covariance W are to be estimated by simulation (Train, 2009). We calculate the individual-level parameters for each attribute using the method proposed by Revelt and Train (2000) for ML models. 2.1.2. Calculating the WTA for implementing BO contracts Following the analysis of the attributes that explain farmers’ choices of enrolling in BO contracts, we calculate farmers’ WTA for attribute k in BO contracts. WTA can be computed from the following equation:   WTAk=−βkβmon (4)where βk and βmon are the parameters associated with attribute k and the monetary attribute, respectively. The individual farmer’s WTA for the attribute k is calculated as follows:   WTAki=−βkiβmon (5)where βki and βmon are the parameters associated to attributes k for the individual i and the monetary attribute for the full sample, respectively. 2.1.3. Modelling farmers’ decisions on the acreage to enrol in a BO contract In addition, we analyse the farmers’ decision to enrol certain acreage in a BO contract. Our dependent variable y is the enrolled share of each farmer’s used agricultural area (UAA)6 with 0<y<1. Explanatory variables are the contract attributes and other individual characteristics. To model this proportion, we followed Kuhfuss et al. (2016) and use a beta regression model instead of a simple linear model (Paolino, 2001; Ferrari and Cribari-Neto, 2004). 2.2. The Picardy case study and survey Our study uses agricultural land located in Picardy, Northern France, as a case study. Agriculture is the dominant land use, covering 75.2 per cent of Picardy (forests and urban areas cover 17 per cent and 6.7 per cent respectively), as shown in Figure 1 (2012 CORINE Land Cover inventory). Most farms have a UAA of more than 100 hectares (mean is 200 hectares) and common crops are cereals, oilseeds and other arable crops (AGRESTE, 2010). Although modified (i.e. managed), the land provides habitat for threatened and protected birds such as the corncrake (Crex crex) or the stone-curlew (Burhinus oedicnemus), and plants such as the hurtsickle (Centaurea cyanus). These species are dependent on large expanses of low-intensity cropland and permanent grasslands, habitats that have been lost over past decades (Stoate et al., 2001, 2009). Fig. 1. View largeDownload slide Land use in Picardy, Northern France (2012 CORINE Land Cover inventory). Fig. 1. View largeDownload slide Land use in Picardy, Northern France (2012 CORINE Land Cover inventory). To inform the design of our CE, we conducted three focus groups with local farmer union-run bodies (Chambres d’Agriculture) in Picardy from July 2015 to February 2016. Focus groups included technical staff and/or farmers who helped us determine the key characteristics of BO contracts (i.e. the attributes in our CE and their levels), develop realistic CE scenarios and formulate follow-up questions for our survey. A pilot survey to test the questionnaire was conducted during February and March 2016 with 26 farmers in Picardy after which the questionnaire was adjusted following the received farmer feedback. The final questionnaire includes three parts (described in Sections 2.2.1–2.2.3) and was implemented using Lime Survey (Version 2.05+ Build 150413), an online application. The time to complete the questionnaire was estimated at 15–20 minutes (10 minutes of video explanations and 5–10 minutes for responding to questions). The link to the survey was sent by the local farmer union-run bodies to all the available e-mail addresses of farmers in Picardy (5,100 contacts) at the beginning of May 2016. The survey was available for 1 month, until early June 2016. The collected answers are anonymous. The web-based approach is well suited for our case study. First, farmers in this French region are in general well connected to the Internet. According to the French National Institute of Statistics and Economic Studies (INSEE), 70.6 per cent of French farmers had an Internet connection (this figure was 78.2 per cent for French people on average) in 2012 (Gombault, 2013). The Internet is a common means of communication between the farmers and the local farmer union-run bodies and farmers make online declarations annually for CAP subsidies. Second, this method enables us to reach out to a large sample of farmers. Finally, it is well known that web-based surveys avoid bias from the presence of an interviewer. 2.2.1. Questionnaire introduction and attributes description The questionnaire was accompanied by a video presentation of the current legal framework for BO in France, as farmers are not necessarily well informed on the topic. Next, we presented a fictive development project in their region implying destruction of meadows of ecological interest which triggers a BO requirement for the developer, who would then propose BO to the farmer. If the farmer accepts one of the two contracts offered, he/she would implement the management plan of the BO on arable land on his/her farm on behalf of the developer and would be paid for this activity. Alternatively, the farmer can decline the offer and keep his/her current agricultural practices by choosing the status quo option. The eligibility rules and terms and conditions of the contract were as follows: (i) the activity has to be additional to regulatory obligations (i.e. they cannot replace cross-compliance requirements under the CAP, current AES contracts, mandatory buffer zones along watercourses and lakes in line with the EU Nitrate Directive, or with previously planned and funded programmes); (ii) farmers would get technical support provided by relevant technical and administrative staff from the local farmer union-run bodies, or other specialised public agencies or non-governmental organisations (NGOs); (iii) farmers must agree to give access to their land for ecological monitoring and compliance control by regulators. In addition, (iv) none of the parties of a BO contract has a right of withdrawal. Instead of using long written explanations, we created short and accessible videos and made them available online.7 The use of videos to introduce information at the beginning of a CE is novel.8 To our knowledge, no study has been carried out to measure the possible influence of this method on the answers compared with more widely used textual or graphical methods. Studies using virtual reality instead of text within choice cards report improvements in the attention of the interviewees or the consistency of their choices (e.g. Bateman et al., 2009; Matthews, Scarpa and Marsh, 2017; Patterson et al., 2017). To limit bias, we prepared the videos to be neutral and objective, in collaboration with focus group participants. The BO contracts have four attributes. The first attribute is the management plan. The four possible management plans (Figure 2) share a common foundation of technical specifics. This foundation implies that arable land must be converted into meadows,9 with seed mix of legumes and grasses, and subsequently mowed (not grazed). Restrictions apply to mowing (centrifugal, forbidden at night, 15 cm minimum height, etc.). In addition, the area concerned has to be at least 10 m wide and larger than 0.5 ha. The ecological ‘gain’ increases from Levels I to IV of the management plans (Figure 2), due to decreased nitrogen (N) fertilisation, further delay of mowing and the addition of a refuge zone. The refuge zone is a zone of the meadow at least 10 m wide representing 10 per cent of the contracted area that is not mowed any given year, and that can be moved from 1 year to the next within the contracted area: it is not a fixed set-aside for the duration of the contract. These four levels are increasingly accommodating practices given the life cycle of meadow dependent plant and animal species. Such BO requirements are very close to the contracts a farmer would face in a ‘real life’ BO situation given the species and other biodiversity features that trigger offsets in Northern France. This also applies to the second attribute related to contract duration. Fig. 2. View largeDownload slide The four levels of proposed management plan. Fig. 2. View largeDownload slide The four levels of proposed management plan. Contract duration is frequently mentioned as an important factor for enrolment and is likely to be important for BO contracts as well. After the focus groups consultation, chosen durations (that cannot be terminated during the contract period) were 9, 18, 25 or 40 years. The third attribute, the conditional monetary bonus, is associated with adding (ecologically relevant) conditions to the management plan. The bonus is proposed in some scenarios and can be accepted or not by the farmers. The two levels for this attribute are thus ‘available’ or ‘not available’ in the scenarios. If the bonus is available, the farmer may decide to activate this bonus, implying a €200/ha/year payment in addition to the baseline payment. However, this bonus involves further conditions: the farmer signs the contract for the scenario for at least 5 ha and the restored parcels must be grouped in one piece or included in an ecological network on the farm. The fourth attribute is the actual payment of €800/ha/year, €1100/ha/year, €1500/ha/year or €2000/ha/year. The lower limit of these amounts was chosen based on unit AES payments, in the Picardy region that relate to the following: creation and maintenance of perennial herbaceous cover (EU COUVER06), extensive management of grassland (EU HERBE03), mowing delay (EU HERBE06) and repair and maintenance of a cover of floristic or faunistic interest (EU COUVER07). The latter is similar to the maintenance of a refuge zone as included in our scenarios. The range of values was discussed and adjusted during the focus group discussions and again after the pilot study. Following Rose and Bliemer (2013), we use a wide range of payments to enable the WTA of farmers most hesitant to implement the practices to be assessed. 2.2.2. The choice experiment The second part of the questionnaire is the CE per se. To understand which of the four attributes is most important to farmers, we combine their different levels in scenarios that describe different types of BO. Since we have four attributes with two to four levels each and choice cards with two alternatives, the full factorial design generates 16,25610 different choice cards. We used the SAS software (SAS University Edition, Version workstation 6.5–7.x virtual machine) and its command %mktruns to decide on the number of scenarios to propose to interviewees: 16 different combinations. We gathered the scenarios in pairs leading to eight choice cards (illustrated in Figure 3). We added an opt-out answer to each choice card, stating ‘I prefer to keep my current agricultural practices’ (the status quo). The use of an opt-out answer is supposed to improve the realism of the choice cards and hence improve the estimation results (Adamowicz and Boxall, 2001; Kontoleon and Yabe, 2003). Thus, farmers choose which BO they would agree to carry out on their farm. After they selected one of the contracts, an additional question appeared: we asked them how many hectares they would be able to commit to the selected BO (acreage). After the focus groups deliberation, it was decided it would not be necessary to create subgroups of choice cards (blocks); all the farmers thus answered the eight choice cards. Fig. 3. View largeDownload slide Example of one of the choice cards of the final CE survey. Fig. 3. View largeDownload slide Example of one of the choice cards of the final CE survey. We selected the eight choice cards following a two-step procedure. First, the pilot study choice cards were chosen using an orthogonal efficient design with all the prior parameters set to 0, to determine the specific prior parameters of our sample (using the NGene software). Our D-error was 0.020317, which is acceptable. We analysed the results of the pilot study with a multinomial logit model to get a better idea of the value of the prior parameters for each attribute, or for each level of attribute for the dummy variables, and then minimise the variance. Second, the final study choice cards were chosen using a Bayesian efficient design, employing the parameters obtained from the pilot study (using the NGene software). Bayesian prior parameters11 are increasingly used as they enable random priors to be defined around the value of the prior parameters from the pilot study (for a review, see Chaloner and Verdinelli, 1995). In line with the literature (e.g. Rambonilaza and Dachary-Bernard, 2007; Kaczan, Swallow and Adamowicz, 2013; Le Coënt, Préget and Thoyer, 2017), we used effects coding instead of dummy coding for the bonus variable in order to avoid confounding the coefficient of the opt-out answer (Bech and Gyrd‐Hansen, 2005; Hasan-Basri and Karim, 2013). The presence of the bonus in a BO contract is coded as 1 while the opt-out option has no bonus by definition (coded 0). The absence of the bonus in a proposed BO contract is coded −1. As suggested by Haaijer, Kamakura and Wedel (2001), we used an additional variable for the opt-out. The opt-out takes the value of 1 when the farmer chooses to keep the current practices with the four attributes taking a value of 0. When a BO contract is selected, the opt-out takes the value of 0. The opt-out variable thus captures the preference of the farmers for their current practices (Vermeulen, Goos and Vandebroek, 2008). Table 1 summarises the attribute description and variable names (cf. Section 2.2.1), attribute levels and the corresponding coding for the analysis. Table 1. Attribute description, levels and their coding to describe BO scenarios proposed to farmers BO attributes and the opt-out  Description  Levels  Coding  Management plan  Levels of management plan required by the BO contract related to: quantity of azote fertilisation (UN), date of mowing, and presence of a refuge zone  Level I: 30 UN, 20 June, no refuge zone  1  Level II: 0 UN, 20 June, no refuge zone  2  Level III: 0 UN, 20 July, no refuge zone  3  Level IV: 0 UN, 20 July, refuge zone  4  Opt-out  0  Contract duration  Total duration of the BO contract  9 years  9  18 years  18  25 years  25  40 years  40  Opt-out  0  Conditional monetary bonus  Additional payment (€200/ha/year) for ecologically relevant measures, provided that the bonus is available in the scenario  Available bonus (€200/ha/year)  +1  No bonus in a BO contract  −1  No bonus because this is the opt-out answer  0  Payment  Payment received each year by the farmer per enrolled hectare  €800/ha/year  800  €1100/ha/year  1100  €1500/ha/year  1500  €2000/ha/year  2000  Opt-out  0  Opt-out  The farmer prefers to keep the current practices  Opt-out  1  BO contract 1 or 2  0  BO attributes and the opt-out  Description  Levels  Coding  Management plan  Levels of management plan required by the BO contract related to: quantity of azote fertilisation (UN), date of mowing, and presence of a refuge zone  Level I: 30 UN, 20 June, no refuge zone  1  Level II: 0 UN, 20 June, no refuge zone  2  Level III: 0 UN, 20 July, no refuge zone  3  Level IV: 0 UN, 20 July, refuge zone  4  Opt-out  0  Contract duration  Total duration of the BO contract  9 years  9  18 years  18  25 years  25  40 years  40  Opt-out  0  Conditional monetary bonus  Additional payment (€200/ha/year) for ecologically relevant measures, provided that the bonus is available in the scenario  Available bonus (€200/ha/year)  +1  No bonus in a BO contract  −1  No bonus because this is the opt-out answer  0  Payment  Payment received each year by the farmer per enrolled hectare  €800/ha/year  800  €1100/ha/year  1100  €1500/ha/year  1500  €2000/ha/year  2000  Opt-out  0  Opt-out  The farmer prefers to keep the current practices  Opt-out  1  BO contract 1 or 2  0  Table 1. Attribute description, levels and their coding to describe BO scenarios proposed to farmers BO attributes and the opt-out  Description  Levels  Coding  Management plan  Levels of management plan required by the BO contract related to: quantity of azote fertilisation (UN), date of mowing, and presence of a refuge zone  Level I: 30 UN, 20 June, no refuge zone  1  Level II: 0 UN, 20 June, no refuge zone  2  Level III: 0 UN, 20 July, no refuge zone  3  Level IV: 0 UN, 20 July, refuge zone  4  Opt-out  0  Contract duration  Total duration of the BO contract  9 years  9  18 years  18  25 years  25  40 years  40  Opt-out  0  Conditional monetary bonus  Additional payment (€200/ha/year) for ecologically relevant measures, provided that the bonus is available in the scenario  Available bonus (€200/ha/year)  +1  No bonus in a BO contract  −1  No bonus because this is the opt-out answer  0  Payment  Payment received each year by the farmer per enrolled hectare  €800/ha/year  800  €1100/ha/year  1100  €1500/ha/year  1500  €2000/ha/year  2000  Opt-out  0  Opt-out  The farmer prefers to keep the current practices  Opt-out  1  BO contract 1 or 2  0  BO attributes and the opt-out  Description  Levels  Coding  Management plan  Levels of management plan required by the BO contract related to: quantity of azote fertilisation (UN), date of mowing, and presence of a refuge zone  Level I: 30 UN, 20 June, no refuge zone  1  Level II: 0 UN, 20 June, no refuge zone  2  Level III: 0 UN, 20 July, no refuge zone  3  Level IV: 0 UN, 20 July, refuge zone  4  Opt-out  0  Contract duration  Total duration of the BO contract  9 years  9  18 years  18  25 years  25  40 years  40  Opt-out  0  Conditional monetary bonus  Additional payment (€200/ha/year) for ecologically relevant measures, provided that the bonus is available in the scenario  Available bonus (€200/ha/year)  +1  No bonus in a BO contract  −1  No bonus because this is the opt-out answer  0  Payment  Payment received each year by the farmer per enrolled hectare  €800/ha/year  800  €1100/ha/year  1100  €1500/ha/year  1500  €2000/ha/year  2000  Opt-out  0  Opt-out  The farmer prefers to keep the current practices  Opt-out  1  BO contract 1 or 2  0  2.2.3. Follow-up questions The third part of the questionnaire included follow-up questions developed with the help of the focus groups. These questions were aimed at characterising the respondents and at improving our interpretation of the results. A first set of follow-up questions dealt with the farmers, their socio-economical profile and details about their farms. We used this information to assess the representativeness of our sample. While we tested all these respondent characteristics in the econometric analysis, only some of them had a significant effect in the model results. In the result section, we provide in a footnote a list of the characteristics that were insignificant in the analysis. Other questions were added to check the quality of our CE answers. First, to reveal possible attribute non-attendance (see Kehlbacher, Balcombe and Bennett, 2013 for a review and test), we asked respondents if they focused their answers on one specific attribute or systematically ignored an attribute. Second, we identified and removed protest responses from the analysed sample (i.e. the answers from respondents that systematically refuse to choose any contract without paying attention to their composition) because such answers do not add any information on preferences for attributes (Adamowicz et al., 1998; Barrio and Loureiro, 2013). To identify the protest responses among the farmers who had systematically chosen the opt-out answer we removed: (i) those who stopped filling the questionnaire just after the last choice card part of the questionnaire and (ii) those who answered that the main reason for the opt-out was that (a) non-withdrawal was unacceptable, (b) she/he was against the BO principle in general or (c) she/he was against the BO principle in the agricultural context. Other farmers who systematically selected the opt-out answer were kept in the sample because they are likely to have given serious attention to and compared the eight choices on offer. The three reasons mentioned above (a, b and c) were presented in the online questionnaire when a farmer systematically chose the opt-out answer12 or chose the opt-out answer at least once.13 Third, we used the follow-up questions to check the additionality of the proposed choices at the scale of the farm. As explained in Section 1, farmers were aware that the measures had to be additional with on-going activities and regulatory requirements beneficial to meadow wildlife. In addition, we asked farmers if they would compensate the loss of the production on the contracted land by increasing production on other parts of their farm (often called the leakage effect). If this involves, e.g. ploughing an existing meadow of conservation value, the additionality and purpose of the offset contract is weakened. 3. Results 3.1. Sample characteristics We received 162 answers, giving us a response rate of about 3.2 per cent. After dropping 18 protest responses, we obtained 144 useful responses, equivalent to 3,456 observations. The fact that we did not separate the eight choice cards in blocks leads to a high number of observations per farmer. Descriptive statistics of useful answers are presented in Table 2. The representativeness of the sample is good, in terms of spatial distribution of the sampled farmers (see Figure 4) and age (apart from the age class of over 60 years old). The mild under-representation of this age class may be linked to the web-based nature of our enquiry. We observed a slight over-representation of farmers with large farms (the mean UAA of the sample is of 199 ha). Moreover, cereal and oil seed crops and single household unincorporated farms are also slightly over-represented. The farmers who answered our enquiry seem to already have some knowledge on BO. While a majority (78 per cent) of them have never been contacted before by project developers for BO contract, 63 per cent had heard of BO contracts before receiving the questionnaire. We may have an over-representation of people already knowledgeable on BO, but no other regional data on this subject are available to test this. It is possible that people with a strong opinion against BO in general, or BO on agricultural land, or people who do not know BO ignored our enquiry. We may thus also have an under-representation of strong opponents to BOs. We do not have this information, which may be a limit of web-based surveys compared to face-to-face or phone interviews, where disagreements to participate could have been recorded. We discuss below the possible consequences of these under and over-representations. Overall, the representativeness of the sample is good, allowing us to draw conclusions for comparable rural areas. Table 2. Sample and population characteristics   Farms and farmers in the sample  Farms and farmers in Picardya  Gender (%)  Male  83  No data  Female  9  No answer  8  Age class (%)  [18,39]  20  18  [40,49]  33  29  [50–59]  35  32  >60  6  21  No answer  6  /  Used Agricultural Area (UAA)  Mean UAA (ha)  199  98  Number of exploitation (%):   <50 ha  5  38   [50–100] ha  14  22   >100 ha  72  40  No answer  9  /  Agricultural status (%)  Single household unincorporated farm  24  57  Jointly run farm (Groupement Agricole d'Exploitation en Commun (GAEC))  7  6  Private limited farming company (Entreprise Agricole à Responsabilité Limitée(EARL))  40  26  Other  20  11  No answer  9  /  Economic and Technical Orientation (ETO) (%)  Cereals and oilseeds  43  28  Other arable crops  21  32  Wine growing  2  3  Other cultures  1  2  Cattle livestock  3  10  Other livestock  3  11  Livestock and crops combination  13  14  No answer  14  /  Contracts previously signed by farmers (%)  BO contract  2  No data  AES  17  Other (Natura 2000, etc.)  3  No previous contracts  78  Lease duration  18 years  67  No data  9 years  13  Other  4  No answer  16    Farms and farmers in the sample  Farms and farmers in Picardya  Gender (%)  Male  83  No data  Female  9  No answer  8  Age class (%)  [18,39]  20  18  [40,49]  33  29  [50–59]  35  32  >60  6  21  No answer  6  /  Used Agricultural Area (UAA)  Mean UAA (ha)  199  98  Number of exploitation (%):   <50 ha  5  38   [50–100] ha  14  22   >100 ha  72  40  No answer  9  /  Agricultural status (%)  Single household unincorporated farm  24  57  Jointly run farm (Groupement Agricole d'Exploitation en Commun (GAEC))  7  6  Private limited farming company (Entreprise Agricole à Responsabilité Limitée(EARL))  40  26  Other  20  11  No answer  9  /  Economic and Technical Orientation (ETO) (%)  Cereals and oilseeds  43  28  Other arable crops  21  32  Wine growing  2  3  Other cultures  1  2  Cattle livestock  3  10  Other livestock  3  11  Livestock and crops combination  13  14  No answer  14  /  Contracts previously signed by farmers (%)  BO contract  2  No data  AES  17  Other (Natura 2000, etc.)  3  No previous contracts  78  Lease duration  18 years  67  No data  9 years  13  Other  4  No answer  16  aAgricultural census for 2010 (AGRESTE, 2010). Table 2. Sample and population characteristics   Farms and farmers in the sample  Farms and farmers in Picardya  Gender (%)  Male  83  No data  Female  9  No answer  8  Age class (%)  [18,39]  20  18  [40,49]  33  29  [50–59]  35  32  >60  6  21  No answer  6  /  Used Agricultural Area (UAA)  Mean UAA (ha)  199  98  Number of exploitation (%):   <50 ha  5  38   [50–100] ha  14  22   >100 ha  72  40  No answer  9  /  Agricultural status (%)  Single household unincorporated farm  24  57  Jointly run farm (Groupement Agricole d'Exploitation en Commun (GAEC))  7  6  Private limited farming company (Entreprise Agricole à Responsabilité Limitée(EARL))  40  26  Other  20  11  No answer  9  /  Economic and Technical Orientation (ETO) (%)  Cereals and oilseeds  43  28  Other arable crops  21  32  Wine growing  2  3  Other cultures  1  2  Cattle livestock  3  10  Other livestock  3  11  Livestock and crops combination  13  14  No answer  14  /  Contracts previously signed by farmers (%)  BO contract  2  No data  AES  17  Other (Natura 2000, etc.)  3  No previous contracts  78  Lease duration  18 years  67  No data  9 years  13  Other  4  No answer  16    Farms and farmers in the sample  Farms and farmers in Picardya  Gender (%)  Male  83  No data  Female  9  No answer  8  Age class (%)  [18,39]  20  18  [40,49]  33  29  [50–59]  35  32  >60  6  21  No answer  6  /  Used Agricultural Area (UAA)  Mean UAA (ha)  199  98  Number of exploitation (%):   <50 ha  5  38   [50–100] ha  14  22   >100 ha  72  40  No answer  9  /  Agricultural status (%)  Single household unincorporated farm  24  57  Jointly run farm (Groupement Agricole d'Exploitation en Commun (GAEC))  7  6  Private limited farming company (Entreprise Agricole à Responsabilité Limitée(EARL))  40  26  Other  20  11  No answer  9  /  Economic and Technical Orientation (ETO) (%)  Cereals and oilseeds  43  28  Other arable crops  21  32  Wine growing  2  3  Other cultures  1  2  Cattle livestock  3  10  Other livestock  3  11  Livestock and crops combination  13  14  No answer  14  /  Contracts previously signed by farmers (%)  BO contract  2  No data  AES  17  Other (Natura 2000, etc.)  3  No previous contracts  78  Lease duration  18 years  67  No data  9 years  13  Other  4  No answer  16  aAgricultural census for 2010 (AGRESTE, 2010). Fig. 4. View largeDownload slide Distribution of sampled farmers in Picardy, Northern France. Fig. 4. View largeDownload slide Distribution of sampled farmers in Picardy, Northern France. Although we are aware of the risk of weak additionality at the farm-scale, we cannot be sure it will be prevented: 22 per cent of the sampled farmers are already committed to AES schemes or other environmental contracts and may behave opportunistically despite our reminder of the impossibility to use such on-going contract as BO. Most of the farmers in our sample would not try to compensate for their loss of production: 58 per cent of the farmers said they would not compensate the loss of the production at the offset site by increasing the production on other plots on the farm, 19 per cent said they would and 23 per cent did not answer the question. Those who said they would transfer the production may increase the productivity of existing cultivated plots rather than ploughing existing meadows on their farms as there are restrictions to this, especially on designated sensitive meadows. This limits the risk of leakage. Despite all our efforts to improve the level of participation, our response rate is rather low. Examples of low-response rates in other CE surveys include Abildtrup et al. (2013) and Torres et al. (2015) with, respectively, 2.2 per cent and 1.1 per cent of the estimated sample, for studies on forest ecosystem services valuation. Czajkowski, Hanley and Nyborg (2017) had a response rate of 5.23 per cent on household waste recycling behaviour. It is unlikely that the low-response rate is due to the use of a web-based survey or of videos to introduce our CE. Other factors may be involved such as the controversial – and potentially unknown for a share of farmers – nature of the topic, the time of the year chosen for the survey or the potential fatigue of farmers faced with academic enquiries that do not provide any feedback.14 We may have expected that farmers with strong opinions about BO would have responded disproportionately, allowing for possible self-selection into the sample. However, this is not what we have observed: only 11 per cent of the initial sample can be considered as strong opponents to BOs on agricultural lands (the 18 removed protest responses) and we did not have strong advocates since none of the respondents accepted BO contracts in all choice cards. Follow-up questions for checking protest responses (see Section 2.2.3) gave us further insights into farmers’ behaviours. We found that 19 per cent of the 144 farmers systematically chose the opt-out option, without being protest responders, because (i) they prefer to keep their current practices and are unwilling to change whatever the payment or other considerations, (ii) they were not satisfied by the different BO contracts proposed and (iii) the BO contracts are technically not feasible on their lands. The remaining 81 per cent chose the opt-out answer at least once for reasons ii and iii as above but mainly because of the proposed payment level (too low). This shows that farmers in our sample had diverse and justified reasons to accept or refuse BO contracts, reducing the risk of self-selection into the sample. 3.2. Econometric analysis 3.2.1. Farmer’s decision to enrol in a BO contract We first conducted a CL model to analyse the decision of farmers to enrol in a BO contract. The Hausman test, implemented by comparing estimations of the full CL model and partial CL by dropping each time one alternative, showed that the IIA hypothesis is rejected in our sample (with Prob > Chi2 = 0.3101; 0.3725 and 0.4377) making the CL estimations invalid. We thus relaxed the IIA hypothesis and modelled the farmer’s decision to participate in a BO contract with a ML model. The participation is modelled first according to the attributes of BO contracts as sole factors of choice and later by integrating individual characteristics (MLogit 1 and MLogit 2 models in Table 3). As individual variables cannot be integrated alone in the model (they do not vary across one individual), they are captured through interaction variables. The payment and the interaction variables, in model MLogit 2, are fixed and the other attributes are random (normally distributed). Table 3. ML estimates Variables  MLogit 1  MLogit 2  Payment  0.00309***  0.00329***  (0.000226)  (0.000259)  Management plan  −0.325***  −0.722***  (0.0749)  (0.160)  Contract duration  −0.146***  −0.208***  (0.0161)  (0.0272)  Conditional monetary bonus  0.265***  0.258***  (0.0651)  (0.0704)  Opt-out  2.301***  1.941***  (0.367)  (0.395)  Management plan * UAA    0.169***  (0.0556)  Contract duration * partial landowner    0.0585**  (0.0252)  SD   Management plan  0.212**  0.374***  (0.104)  (0.0987)   Contract duration  0.102***  0.135***  (0.0112)  (0.0148)   Conditional monetary bonus  0.184  0.145  (0.127)  (0.136)   Opt-out  2.526***  2.264***  (0.305)  (0.352)  ll  −754.84448  −643.98198  Chi2  453.21231  387.28256  AIC  1527.689  1309.964  BIC  1583.0198  1376.2086  Observations  3456  3048  Number of farmers  144  127  Variables  MLogit 1  MLogit 2  Payment  0.00309***  0.00329***  (0.000226)  (0.000259)  Management plan  −0.325***  −0.722***  (0.0749)  (0.160)  Contract duration  −0.146***  −0.208***  (0.0161)  (0.0272)  Conditional monetary bonus  0.265***  0.258***  (0.0651)  (0.0704)  Opt-out  2.301***  1.941***  (0.367)  (0.395)  Management plan * UAA    0.169***  (0.0556)  Contract duration * partial landowner    0.0585**  (0.0252)  SD   Management plan  0.212**  0.374***  (0.104)  (0.0987)   Contract duration  0.102***  0.135***  (0.0112)  (0.0148)   Conditional monetary bonus  0.184  0.145  (0.127)  (0.136)   Opt-out  2.526***  2.264***  (0.305)  (0.352)  ll  −754.84448  −643.98198  Chi2  453.21231  387.28256  AIC  1527.689  1309.964  BIC  1583.0198  1376.2086  Observations  3456  3048  Number of farmers  144  127  Standard errors in parentheses. Significant levels: ***p < 0.01, **p < 0.05, *p < 0.1. Table 3. ML estimates Variables  MLogit 1  MLogit 2  Payment  0.00309***  0.00329***  (0.000226)  (0.000259)  Management plan  −0.325***  −0.722***  (0.0749)  (0.160)  Contract duration  −0.146***  −0.208***  (0.0161)  (0.0272)  Conditional monetary bonus  0.265***  0.258***  (0.0651)  (0.0704)  Opt-out  2.301***  1.941***  (0.367)  (0.395)  Management plan * UAA    0.169***  (0.0556)  Contract duration * partial landowner    0.0585**  (0.0252)  SD   Management plan  0.212**  0.374***  (0.104)  (0.0987)   Contract duration  0.102***  0.135***  (0.0112)  (0.0148)   Conditional monetary bonus  0.184  0.145  (0.127)  (0.136)   Opt-out  2.526***  2.264***  (0.305)  (0.352)  ll  −754.84448  −643.98198  Chi2  453.21231  387.28256  AIC  1527.689  1309.964  BIC  1583.0198  1376.2086  Observations  3456  3048  Number of farmers  144  127  Variables  MLogit 1  MLogit 2  Payment  0.00309***  0.00329***  (0.000226)  (0.000259)  Management plan  −0.325***  −0.722***  (0.0749)  (0.160)  Contract duration  −0.146***  −0.208***  (0.0161)  (0.0272)  Conditional monetary bonus  0.265***  0.258***  (0.0651)  (0.0704)  Opt-out  2.301***  1.941***  (0.367)  (0.395)  Management plan * UAA    0.169***  (0.0556)  Contract duration * partial landowner    0.0585**  (0.0252)  SD   Management plan  0.212**  0.374***  (0.104)  (0.0987)   Contract duration  0.102***  0.135***  (0.0112)  (0.0148)   Conditional monetary bonus  0.184  0.145  (0.127)  (0.136)   Opt-out  2.526***  2.264***  (0.305)  (0.352)  ll  −754.84448  −643.98198  Chi2  453.21231  387.28256  AIC  1527.689  1309.964  BIC  1583.0198  1376.2086  Observations  3456  3048  Number of farmers  144  127  Standard errors in parentheses. Significant levels: ***p < 0.01, **p < 0.05, *p < 0.1. For both models (with and without interactions) all the contract attributes have significant effects on farmers’ choice. No attribute non-attendances were observed. The model without interaction (MLogit 1) shows that the likelihood to participate decreases with more demanding management plans. Contract duration also has a significant negative effect on farmers’ utility. Participation increases with the conditional bonus and weakly increases with the payment. The fact that the bonus improves the likelihood of signing a contract is consistent with the information that the farmers generally decided to benefit from the bonus15 when they choose a contract including a bonus. The difference from 0 and high significance of the opt-out parameter means that the farmers have a strong preference for keeping their current practices.16 The introduction of individual characteristics (MLogit 2) leads to the removal of 408 observations (linked to 17 individuals) due to missing data but improves the model results (AIC and BIC are lower). Compared to the MLogit 1 model, results from MLogit 2 are similar. Farmers prefer to keep their practices and prefer contracts with limited constraints,17 including the individual monetary bonus, of short duration and well paid. Two interactions with socio-economic characteristics18 have quite significant effects: being a partial or full landowner increases the likelihood to adopt contracts with a longer duration; and having a farm with a larger area (hundreds of ha) increases the likelihood to adopt more constraining measures. Because we have an over-representation of large farms in our sample, it is possible that we underestimate the negative effect of the management plan on the decision to enrol in a BO contract. Also note that if strong opponents to BO were to be under-represented in the sample, this does not change these results because (i) they would have been removed from the econometric analysis for being protest responses and (ii) they would support the main result that farmers seem to have a strong preference for not signing BO contracts. The standard deviation (SD) coefficients reveal that preferences for all the attributes, except for the bonus, are heterogeneous among the farmers of our sample. This confirms it is interesting to consider the distribution of the individual parameters for each attribute that we have calculated using the log likelihood maximisation.19 Figure 5 presents Epanechnikov kernel density plots of the distribution of the individual parameter estimates (Epanechnikov, 1969; Silverman, 1992).20 Fig. 5. View largeDownload slide Distribution of the individual parameter for each attribute and the opt-out. (A) Management plan, (B) contract duration, (C) conditional monetary bonus and (D) opt-out (keep the current practices). Fig. 5. View largeDownload slide Distribution of the individual parameter for each attribute and the opt-out. (A) Management plan, (B) contract duration, (C) conditional monetary bonus and (D) opt-out (keep the current practices). The distributions of the individual parameter show that farmers’ preferences for the bonus and the management plan seem to be concentrated around a single value (Figure 5A and C) while there appears to be at least two groups of preferences for contract duration (Figure 5B). The two groups of the opt-out nicely illustrate that there are farmers that systematically chose to keep their practices (right-hand spike in Figure 5D, corresponding to the 19 per cent of farmers described in Section 3.1) and farmers who chose BO contracts (left-hand spike, corresponding to the remaining 81 per cent). 3.2.2. WTA for implementing BO contracts Farmers’ WTA for implementing different features of BO contracts is presented in Table 4. Table 4. WTA for BO contract implementation Attributes  WTA (€/ha/year)  SD  Min  Max  Management plan  219  54  45  365  Contract duration  63  31  −6  111  Conditional monetary bonusa  −157  18  −207  −98  Opt-out  −590  525  −1936  564  Attributes  WTA (€/ha/year)  SD  Min  Max  Management plan  219  54  45  365  Contract duration  63  31  −6  111  Conditional monetary bonusa  −157  18  −207  −98  Opt-out  −590  525  −1936  564  aEffects coded variable, the WTA must be multiplied by 2 (Le Coënt, Préget and Thoyer, 2017) because we introduced a two-units variation (instead of a one-unit variation) between the two levels of this variable. Table 4. WTA for BO contract implementation Attributes  WTA (€/ha/year)  SD  Min  Max  Management plan  219  54  45  365  Contract duration  63  31  −6  111  Conditional monetary bonusa  −157  18  −207  −98  Opt-out  −590  525  −1936  564  Attributes  WTA (€/ha/year)  SD  Min  Max  Management plan  219  54  45  365  Contract duration  63  31  −6  111  Conditional monetary bonusa  −157  18  −207  −98  Opt-out  −590  525  −1936  564  aEffects coded variable, the WTA must be multiplied by 2 (Le Coënt, Préget and Thoyer, 2017) because we introduced a two-units variation (instead of a one-unit variation) between the two levels of this variable. It is difficult to interpret the value of the WTA for the management plan because we do not know the current practices of each farmer, even if the proposed contracts all focus on converting cultivated arable land. In addition, because this is a mixed attribute, even if it is ordered, the difference between the various levels is not of the same nature. Thus, the WTA for the management plan is €219/ha/year for a composite increase in the level of constraint imposed by our four management plan levels. Farmers are prepared to accept €63/ha/year for a 1-year increase in the duration of the offset contract. Farmers are willing to forgo €157/ha/year to get the bonus: the bonus is worth €200/ha/year so farmers would still benefit. This means that the cost for the developer of implementing supplementary ecological constraints in the management plan is €43/ha/year.21 Farmers are ready to forgo €590/ha/year to keep their current practices, in other words they would only accept contracts that pay more than €590/ha/year. The individual WTAs for the opt-out were calculated from the individual parameters for the opt-out attribute and the parameter for the payment attribute that is unique because it has been set as fixed in the ML. Figure 6 presents Epanechnikov kernel density plots of the distribution of the individual WTA estimates for the opt-out. Fig. 6. View largeDownload slide Distribution of the individual WTA for the opt-out. Fig. 6. View largeDownload slide Distribution of the individual WTA for the opt-out. While the MLogit 2 model shows that farmers on average are ready to forgo €590/ha/year to keep their current practices, Figure 6 shows that farmers who are very reluctant to sign BO contracts have high payment expectations of around €1300/ha/year and that a second group of farmers has expectations of around €500/ha/year, which is actually less than the payments that were proposed in the CE. 3.2.3. Analysing farmers’ decisions on the acreage to enrol in a BO contract We had to remove one farmer (24 observations) from the sample because abnormal areas of offsets were included in this response. Among the choices of the remaining 143 farmers, 530 alternatives other than keeping the current practices were chosen, 506 values of contracted area were completed (24 values were missing) and we could analyse 459 observations (47 values of total UAA were missing so we could not calculate the enrolled proportion of the UAA). These remaining observations are linked to 104 different farmers. Farmers enrolled on average 13 per cent of the UAA (varying from less than 1 per cent to 100 per cent, SD = 25).22 We have no cases where 0 per cent of the farm area is enrolled (y= 0) because we required that farmers enrol at least 0.5 ha, increased to a minimum of 5 ha if the contract includes a conditional bonus. We did not explore the factors which lead farmers to enrol all their UAA (y= 1) because only four farmers did so. In total, we analysed decisions on acreage using 432 observations linked to 100 farmers (Table 5). Table 5. Acreage model estimates Dependent variable: acreage/enrolled proportion of the UAA  Proportion (0 < y< 1)   Management plan  −0.00450  (0.0211)   Contract duration  0.00690  (0.00509)   Conditional monetary bonus  0.0520**  (0.0240)   Payment  1.03e−05  (9.52e−05)   UAA  −0.147**  (0.0611)   Constant  −2.148***  (0.339)  Ln_phi   Constant  1.953***  (0.423)   AIC  −1291.0413   BIC  −1262.5624   Observations  432   Number of farmers  100  Dependent variable: acreage/enrolled proportion of the UAA  Proportion (0 < y< 1)   Management plan  −0.00450  (0.0211)   Contract duration  0.00690  (0.00509)   Conditional monetary bonus  0.0520**  (0.0240)   Payment  1.03e−05  (9.52e−05)   UAA  −0.147**  (0.0611)   Constant  −2.148***  (0.339)  Ln_phi   Constant  1.953***  (0.423)   AIC  −1291.0413   BIC  −1262.5624   Observations  432   Number of farmers  100  Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Table 5. Acreage model estimates Dependent variable: acreage/enrolled proportion of the UAA  Proportion (0 < y< 1)   Management plan  −0.00450  (0.0211)   Contract duration  0.00690  (0.00509)   Conditional monetary bonus  0.0520**  (0.0240)   Payment  1.03e−05  (9.52e−05)   UAA  −0.147**  (0.0611)   Constant  −2.148***  (0.339)  Ln_phi   Constant  1.953***  (0.423)   AIC  −1291.0413   BIC  −1262.5624   Observations  432   Number of farmers  100  Dependent variable: acreage/enrolled proportion of the UAA  Proportion (0 < y< 1)   Management plan  −0.00450  (0.0211)   Contract duration  0.00690  (0.00509)   Conditional monetary bonus  0.0520**  (0.0240)   Payment  1.03e−05  (9.52e−05)   UAA  −0.147**  (0.0611)   Constant  −2.148***  (0.339)  Ln_phi   Constant  1.953***  (0.423)   AIC  −1291.0413   BIC  −1262.5624   Observations  432   Number of farmers  100  Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Among the attributes of the BO, we note that payment has no significant effect on the enrolled acreage even if it has a significant but weak role in the decision to enrol some land. Only the bonus has a significant effect on the likelihood to increase the enrolled percentage of the UAA. We checked if the increase of the enrolled acreage is only linked to the bonus requirement of a minimum of 5 ha enrolled or if there is a true effect of the bonus among the 432 observations. Farmers enrolled a mean of 14 ha ([min = 0.5; max = 150], SD = 25). Table 6 shows that the enrolled acreage is greater when there is a bonus in the proposed BO contract, whether considering the observed or predicted values. Table 6. Number of enrolled hectares in the BO contracts without or with bonus   Contract without conditional monetary bonus (N = 183)  Contract with conditional monetary bonus (N = 249)  Observed individual enrolled number of hectares  13 ha [min = 0.5; max = 140] SD = 23  15 ha [min = 0.5; max = 150] SD = 26  Predicted sample enrolled number of hectares  19 ha [min = 12; max = 26] SD = 3  21 ha [min = 12; max = 29] SD = 3    Contract without conditional monetary bonus (N = 183)  Contract with conditional monetary bonus (N = 249)  Observed individual enrolled number of hectares  13 ha [min = 0.5; max = 140] SD = 23  15 ha [min = 0.5; max = 150] SD = 26  Predicted sample enrolled number of hectares  19 ha [min = 12; max = 26] SD = 3  21 ha [min = 12; max = 29] SD = 3  Table 6. Number of enrolled hectares in the BO contracts without or with bonus   Contract without conditional monetary bonus (N = 183)  Contract with conditional monetary bonus (N = 249)  Observed individual enrolled number of hectares  13 ha [min = 0.5; max = 140] SD = 23  15 ha [min = 0.5; max = 150] SD = 26  Predicted sample enrolled number of hectares  19 ha [min = 12; max = 26] SD = 3  21 ha [min = 12; max = 29] SD = 3    Contract without conditional monetary bonus (N = 183)  Contract with conditional monetary bonus (N = 249)  Observed individual enrolled number of hectares  13 ha [min = 0.5; max = 140] SD = 23  15 ha [min = 0.5; max = 150] SD = 26  Predicted sample enrolled number of hectares  19 ha [min = 12; max = 26] SD = 3  21 ha [min = 12; max = 29] SD = 3  The only individual variable having a significant effect on the enrolled percentage of the UAA is the UAA itself (hundreds of ha). A 100 ha increase in the UAA leads to a 15 per cent decrease of the enrolled percentage of the UAA. We can conclude that there is a threshold effect of the size of the BO area, in absolute and not relative terms, which leads larger farms to enrol a lower share of their UAA. 4. Discussion and conclusion BOs are aimed at achieving NNL of biodiversity in the context of economic development projects, plans or programmes, in principle and under France’s recent biodiversity law. Agricultural landscapes sometimes have low levels of biodiversity and good potential for ecological gains through ecological restoration. However, implementation of BO by farmers remains a controversial topic that has received limited attention in the scientific literature. One may wonder if long-term BO contracts with restrictive conditions for reasons of ecological performance can match farmers’ preferences and constraints. Our study aimed at identifying key factors that explain decisions by farmers to sign a BO contract and that influence the acreage they enrol. We conducted a CE study in Picardy, Northern France, targeting over 5,000 farmers. We received 144 useful responses. The response rate and the size of the sample are quite low. The answers to the follow-up questions included in the survey showed that the risk of self-selection into the sample is limited. However, it is possible that our respondents were better informed of BO than most farmers. Four attributes, describing different scenarios of BO contracts, were selected: the actual ecological restoration activities (in a management plan), the contract duration, the annual payment and the option of receiving a monetary bonus for the extent and spatial distribution of enrolled land. To feedback to our interviewees, we prepared a video23 of the main results of the study that has been shared by the local farmer union-run bodies. While farmers are often targeted by developers for implementing their BO obligations, we show that farmers are quite reluctant to adopt BO contracts which require them to convert arable cropland into grasslands (in our example), and manage these for biodiversity for a defined period of time. All the attributes considered in the CE have a significant effect on the likelihood of choosing a BO contract. The likelihood of choosing a BO measure decreases with increasing levels of constraints on management practices and the duration of the contract. Farmers have, therefore, high WTA for a composite increase in the level of constraint imposed by the management plan or a 1-year increase in contract duration. Higher payment levels and the proposal of a bonus for increased extent of interconnectedness of contracted land improve the likelihood of farmers signing up to a BO contract. Our results show that farmers’ mean WTA to change their practices is even lower than the range of payment we proposed. There is, however, a large spread in payment expectations, with some farmers in the top of the proposed range. We also show that the payment attribute does not significantly lead farmers to enrol a greater share of their UAA, unlike the conditional bonus. This bonus increases the likelihood of signing up a contract and the enrolled acreage but at higher overall cost of a BO contract for the developer. We also show that other reasons limit farmers’ willingness to enrol such as the farm’s total UAA, and that larger farms are likely to accept more constraining management plans. There are policy implications to these results. First, given the reluctance of farmers to enrol their land in complex and long-term management plans, there may be limited scope for involving farmers in the implementation of developers’ BO obligations through contracts. Thus, contracting farmers to restore arable land for biodiversity may be suitable for offsetting short term impacts on already degraded areas of natural habitat (e.g. field margins and recently established permanent grasslands) or modified habitat (e.g. impacts on species that thrive in low-intensity agricultural mosaics), but may not be suitable for long term and permanent impacts on high-quality habitat (e.g. old permanent grasslands). This is because less constraining and shorter contracts will only provide moderate biodiversity gains, which are only suitable for offsetting moderate biodiversity losses (Quétier and Lavorel, 2011). Second, reducing the number of farmers involved in implementing a developer’s BO obligations could reduce their cost, for the developer and for enforcement agencies. This means increasing the enrolled acreage per farmer and our results show that a conditional monetary bonus can favour this, but with an increased per unit area cost for developers. This extra cost can be justified in a context where regulations increasingly make developers liable for the measurable ecological outcomes of their BO (e.g. in France). We also found, however, that there are some limits to this increased enrolled acreage: the mean absolute number of hectares that farmers would be prepared to enrol may not be sufficient to cover needs. In our study, at 15 ha, it is below the threshold that Moreno-Mateos et al. (2012) consider as a lower limit for successful ecological restoration of wetlands. The results of our study may be specific to Europe, where farmers receive CAP subsidies which certainly modify their behaviour. In particular, European farmers are used to short term contracts (5 years) for most agri-environmental schemes (AES) under the CAP. A structural limit to the implementation of BO on arable land is that offsets are supposed to be effective for as long as impacts occur, and this tends to be over long periods of time, and even theoretically into perpetuity (e.g. for many public infrastructure projects). In our study, contract duration is one of the attributes that makes farmers reluctant to sign contracts. We expected this result since other studies with shorter durations made the same observation (Ruto and Garrod, 2009; Christensen et al., 2011; Greiner, Bliemer and Ballweg, 2014). In addition, land tenure patterns (e.g. whether farmers own the land they farm and their capacity to sign contracts with developers) are likely to determine the transferability of our results. In our study, being an owner of part of the cultivated land increased the likelihood to sign a longer contract. Indeed, not being the owner of the land makes it difficult to accept a BO contract because if the lease is shorter than the duration of the BO contract, farmers must be ready to enter discussions and negotiations with the owners of the land they farm, to include contract requirements in their leases (these contracts would have to be transferred to future farmers leasing the land). In France, 60 per cent of the UAA is under a lease (from 9 to 25 years), which could be a considerable impediment to successful (long term) BO implementation on farmland. An alternative is for several different farmers to enrol their leased land into a renewable short term BO contract, with a third-party entity (intermediary) ensuring that the pool of enrolled land fulfils the required ecological gain over time. Le Coënt, Préget and Thoyer (2017) discuss the pros and cons of this approach. This solution may be more or less relevant, depending on the species or habitats targeted by the BO (van Teeffelen et al., 2014; Drechsler, Johst and Wätzold, 2017). Other studies show that contract duration is not a limitation in countries where the proportion of farmers that own the land they farm is higher. This is the case in Australia for pure conservation measures (Greiner, 2016). In the United States, many farmers use BO implementation as an alternative use for their non-profitable arable land. This includes land put into wetland mitigation banks that make offsets available to developers, but require perpetual conservation easements that limit forever most of, or even any, agricultural practices (Vaissière and Levrel, 2015). Another fuzzier reason for the observed reluctance to sign demanding and long-term contracts may be the institutional and environmental uncertainty that farmers face. In our case, we voluntarily did not give details about the possibility to return the land to farming at the end of the contract, because BO is supposed to produce long lasting outcomes. Other CE studies on agri-environmental decisions did consider this (e.g. Lienhoop and Brouwer, 2015). Recent experience may also have made some farmers fearful, specifically temporary meadows under 5-year AES contracts that were subsequently designated as permanent meadows, with strong restrictions on their cultivation under the current CAP. It is unclear how CAP subsidies could or would be articulated with BO contracts (additionality goes both ways) and if BO management plans are compatible with the minimum land management provisions specified in farmland leases. Continued expansion of NNL policies in Europe will make resolving these questions a priority. More generally, our results show that governance (tenure) and economic (cost) criteria should be considered carefully when designing contract-based BO mechanisms that involve farmers. Soule, Tegene and Wiebe (2000) highlighted that adopting conservation practices on farmland is mainly guided by the potential future agronomic value of the land, considering that land owners want to be able to put their land back into agriculture. All the factors that explain farmers’ reluctance to enrol land in BO can be tied to a risk of decrease in the land price if it cannot be cultivated or built anymore. In this uncertain context, the impossibility to stop the contract may be the most limiting factor for signing up to BO (Broch et al. 2013; Greiner, Bliemer and Ballweg, 2014). Farmers may also be justifiably fearful of developers’ capacity to live up to long-term commitments to their BO requirements. Supplementary data Supplementary data are available at European Review of Agricultural Economics online. Funding This work was supported by grants from the Agence Nationale de la Recherche (ANR) as part of the ‘Investissements d’Avenir’ programme (ANR-10-LABX-0004, Lab of Excellence CeMEB; ANR-11-LABX-0002-01, Lab of Excellence ARBRE). This study also received funds from the European Union (FP7/2007–2013) under contract 308393 ‘OPERAs’. Our thanks go to the Université de Montpellier (LAMETA, renamed as CEE-M) and Biotope for supplementary financial support for this work; this paper was part of Anne-Charlotte Vaissière’s post-doctoral fellowship. Acknowledgements We would like to thank Jens Abildtrup, Marion Beaurepaire, Miguel Da Costa Nogueira, Serge Garcia, Laure Kuhfuss, Philippe Le Coënt, Vincent Martinet, Michel Pech, Raphaële Préget, Jean-Marc Rousselle, Sophie Thoyer and Gengyang Tu for their useful comments. The local farmer union-run bodies (Chambres d’Agriculture) of Picardy participated in focus groups and relayed the enquiry to the farmers. We thank them and the farmers for their involvement. Participants at the following Conferences are thanked for their useful comments: the 3rd Annual FAERE Conference (Bordeaux), the 22nd Annual EAERE Conference (Athens) and the 66th Annual AFSE Congress (Nice). We also thank three anonymous reviewers for their valuable comments on previous versions of the manuscript. Footnotes 1 More generally in environmental impact assessment, the loss of ecosystem services is rarely assessed (Tardieu et al., 2015). 2 We use BO to refer to biodiversity offset or biodiversity offsetting, indiscriminately. 3 https://www.eea.europa.eu/data-and-maps/data/copernicus-land-monitoring-service-corine. Accessed 22 February 2018. 4 The ecological efficiency of BOs is determined by their ability to reach the no net loss of biodiversity goal, meaning an in-kind and quantitative ecological equivalence between the impacted and the restored (offset) area with regards to the type of biodiversity (species, habitat types, etc.), its location, timing and duration (accounting for the temporal scale for achieving an ecological response from a restoration action). See Quétier and Lavorel (2011) for a detailed discussion. 5 There is an ecological debate around the interest of having single large versus several small areas for conservation (the SLOSS debate). For instance, Verboom et al. (2001) demonstrate the stabilising role of large patches in marshland (birds habitat) networks but other studies also underline the importance of a network of smaller patches for certain species (e.g. invertebrates in Rösch et al., 2015). In our study, we actually consider both situations, the focus being on having less actors implementing few large or few networks of small areas for conservation. 6 ‘The total area taken up by arable land, permanent grassland, permanent crops and kitchen gardens used by the holding, regardless of the type of tenure or of whether it is used as a part of common land’ (Source: Eurostat, http://ec.europa.eu/eurostat/statistics-explained/index.php/Glossary:Utilised_agricultural_area_(UAA)). Accessed 22 February 2018. 7 The videos are available online for both the offsetting explanations (https://youtu.be/7rXahUmFpM8) and for the choice modelling explanations (https://youtu.be/BRLKNW-84zo) (in French). 8 Note that Navrud (1997) looked at the possibility of using video in describing complex environmental impacts and scenarios in contingent valuation studies. He showed that video can be an effective communication device as this is preferred by the respondents compared with other communication devices (text and photos), even though no significant differences in willingness-to-pay were identified. 9 We asked the farmers to only consider arable land for implementing the BO in order to reduce the heterogeneity of land they could have considered (e.g. unused fallows, permanent meadows or shrubland, and other uncultivated land), and to have offsets that could likely fit all the situations in our study area (unlike wetland restoration, for instance). This reduces the heterogeneity in opportunity costs and ecological restoration potential. 10 The full factorial design generates (4 × 4 × 2 × 4) x ((4 × 4 × 2 × 4)−1) = 16,256 choice cards. 11 To minimise the approximation error when calculating the Bayesian efficiency, we use a Gaussian quadrature. Because we have six priors parameters, the Gaussian quadrature might need a large number of draws so we decided to include; bdraws = gauss(3) in our design. 12 The other possible answers were: ‘I prefer the current use of my land’, ‘I did not like any of the contracts’, ‘the measures are not feasible in my farm’, ‘I do not have the equipment to implement the management plan for the meadow’, ‘the payment is too low’, ‘no answer’. 13 The other possible answers were: ‘I did not like any of the contracts’, ‘the two proposed measures are not feasible in my farm’, ‘the payment is too low’, ‘no answer’. 14 This risk may remain even though we explained at both the beginning and the end of the survey that farmers would receive the results from the study within the year, which they did (see Section 4). 15 Among the 312 choices made that included the conditional monetary bonus: farmers decided to benefit from the bonus in 67% of the choices; they will not benefit from the €200/ha/year because a lower than 5 ha surface was entered in 25% of the choices (in such a case during the enquiry, an error message displayed to inform the farmer that with the proposed number of ha, he/she would not benefit from the bonus to make sure the farmer understood the principle); they refused the bonus even if a higher than 5 ha surface was entered (which means that the farmer does not accept the ecological constraints of the bonus) in only 3% of the choices; and they did not answer the question, that was not mandatory, in 5% of the choices. 16 The opt-out variable is how we interpreted the Alternative Specific Constant (ASC) variable in these ML models. 17 We previously included the management plan variable as three dummy variables, with Level I set as the reference. All the contract attributes had significant effects on farmers’ choice and the parameters for the four levels of the management plan variable were ordered. The farmers have a preference for Level I over Level II, a preference for Level II over Level III and a preference for Level III over Level IV. In other words, they have a preference for the less restrictive management plans over the more restrictive ones. It is thus appropriate to interpret the management plan variable as a quantitative variable: it is continuous and ordered but not linear since the difference between each of the four levels is not similar (stop fertilising from Level I to II, delay the mowing of one more month from Level II to III, set a refuge zone from Level III to IV). 18 The other individual characteristics that were tested but that had no significant effect are: having lands already enrolled in BO or AES, gender, age class, education, status of agricultural company, department, lease duration, Economic and Technical Orientation (ETO), agricultural types (conventional, organic, etc.), income, being part of an environmental NGO, being a hunter. 19 We used the mixlbeta command following the use of the mixlogit one (STATA 14). 20 The kdensity command has been used here (STATA 14). 21 This amount is defined as the €200/ha/year of monetary bonus less the €157/ha/year the farmers are ready to forgo. 22 We used the log likelihood maximisation through the zoib command (STATA 14). This command is linked to a full zero one inflated beta model but we only used its proportion part. 23 The video is available online (https://youtu.be/9W8TvXLA7bU) (in French). References Abildtrup, J., Garcia, S., Olsen, S. B. and Stenger, A. ( 2013). Spatial preference heterogeneity in forest recreation. Ecological Economics  92: 67– 77. Google Scholar CrossRef Search ADS   Adamowicz, W. and Boxall, P., 2001. Future directions of stated choice methods for environment valuation. Paper prepared for ‘Choice Experiments: A New Approach to Environmental Valuation’ Conference, 10 April 2001, London. Adamowicz, W., Boxall, P., Williams, M. and Louvière, J. ( 1998). Stated preference approaches for measuring passive use values: choice experiments and contingent valuation. American Journal of Agricultural Economics  80( 1): 64– 75. Google Scholar CrossRef Search ADS   AGRESTE, 2010. Mémento de la statistique agricole. Edition 2010. Picardie. Banerjee, S., deVries, F. P., Hanley, N. and van Soest, D. P. ( 2014). The impact of information provision on agglomeration bonus performance: an experimental study on local networks. American Journal of Agricultural Economics  96( 4): 1009– 1029. Google Scholar CrossRef Search ADS   Barrio, M. and Loureiro, M. ( 2013). The impacts of protests responses in choice experiments: an application to a biosphere reserve management program. Forest Systems  22( 1): 94– 105. Google Scholar CrossRef Search ADS   Bateman, I. J., Day, B. H., Jones, A. P. and Jude, S. ( 2009). Reducing gain–loss asymmetry: a virtual reality choice experiment valuing land use change. Journal of Environmental Economics and Management  58( 1): 106– 118. Google Scholar CrossRef Search ADS   Bech, M. and Gyrd‐Hansen, D. ( 2005). Effects coding in discrete choice experiments. Health Economics  14( 10): 1079– 1083. Google Scholar CrossRef Search ADS   Bennett, G., Chavarria, A., Ruef, F. and Leonardi, A. ( 2017). State of European Markets 2017 Biodiversity Offsets and Compensation . Washington: Ecosystem Marketplace, 36. Bougherara, D. and Ducos, G., 2006. Farmers’ preferences over compensation contract flexibility and duration: an estimation of the effect of transaction costs using choice experiment. First Meeting of the European School on New Institutional Economics (ESNIE), Université de Paris-Sud XI, France, 26 pp. Broch, S. W., Strange, N., Jacobsen, J. B. and Wilson, K. A. ( 2013). Farmers’ willingness to provide ecosystem services and effects of their spatial distribution. Ecological Economics  92: 78– 86. Google Scholar CrossRef Search ADS   Broch, S. W. and Vedel, S. E. ( 2012). Using choice experiments to investigate the policy relevance of heterogeneity in farmer agri-environmental contract preferences. Environmental and Resource Economics  51( 4): 561– 581. Google Scholar CrossRef Search ADS   Buckley, C., Hynes, S. and Mechan, S. ( 2012). Supply of an ecosystem service – farmers’ willingness to adopt riparian buffer zones in agricultural catchments. Environmental Science & Policy  24: 101– 109. Google Scholar CrossRef Search ADS   Bull, J. W., Suttle, K. B., Gordon, A., Singh, N. J. and Milner-Gulland, E. J. ( 2013). Biodiversity offsets in theory and practice. Oryx  47( 3): 369– 380. Google Scholar CrossRef Search ADS   Calvet, C., Ollivier, G. and Napoléone, C. ( 2015). Tracking the origins and development of biodiversity offsetting in academic research and its implications for conservation: a review. Biological Conservation  192: 492– 503. Google Scholar CrossRef Search ADS   Chaloner, K. and Verdinelli, I. ( 1995). Bayesian experimental design: a review. Statistical Science  10( 3): 273– 304. Google Scholar CrossRef Search ADS   Chen, X., Lupi, F., He, G. and Liu, J. ( 2009). Linking social norms to efficient conservation investment in payments for ecosystem services. Proceedings of the National Academy of Sciences of the United States of America (PNAS)  106( 28): 11812– 11817. Google Scholar CrossRef Search ADS   Christensen, T., Pedersen, A. B., Nielsen, H. O., Mørkbak, M. R., Hasler, B. and Denver, S. ( 2011). Determinants of farmers’ willingness to participate in subsidy schemes for pesticide-free buffer zones – a choice experiment study. Ecological Economics  70( 8): 1558– 1564. Google Scholar CrossRef Search ADS   Czajkowski, M., Hanley, N. and Nyborg, K. ( 2017). Social norms, morals and self-interest as determinants of pro-environment behaviours: the case of household recycling. Environmental and Resource Economics  66( 4): 647– 670. Google Scholar CrossRef Search ADS   de Billy, V., Tournebize, J., Barnaud, G., Benoît, M., Birgand, F., Garnier, J., Lesaffre, B., Lévêque, C., de Marsily, G., Muller, S., Musy, A. and Zimmer, D. ( 2015). Compenser la Destruction de Zones Humides. Retours d’Expériences sur les Méthodes et Réflexions Inspirées par le Projet d’Aéroport de Notre-Dame-des-Landes (France). Nature Sciences Sociétés  23( 1): 27– 41. Google Scholar CrossRef Search ADS   Drechsler, M., Johst, K. and Wätzold, F. ( 2017). The cost-effective length of contracts for payments to compensate land owners for biodiversity conservation measures. Biological Conservation  207: 72– 79. Google Scholar CrossRef Search ADS   Drechsler, M., Wätzold, F., Johst, K. and Shogren, J. F. ( 2010). An agglomeration payment for cost-effective biodiversity conservation in spatially structured landscapes. Resource and Energy Economics  32( 2): 261– 275. Google Scholar CrossRef Search ADS   Epanechnikov, V. A. ( 1969). Non-parametric estimation of a multivariate probability density. Theory of Probability and Its Applications  14( 1): 153– 158. Google Scholar CrossRef Search ADS   Espinosa-Goded, M., Barreiro-Hurlé, J. and Ruto, E. ( 2010). What do farmers want from agri-environmental scheme design? A choice experiment approach. Journal of Agricultural Economics  61( 2): 259– 273. Google Scholar CrossRef Search ADS   Etrillard, C. and Pech, M. ( 2015). Mesures de Compensation Ecologique: Risques ou Opportunités pour le Foncier Agricole en France ? VertigO  15( 2). https://journals.openedition.org/vertigo/16450 Ferrari, S. L. P. and Cribari-Neto, F. ( 2004). Beta regression for modelling rates and proportions. Journal of Applied Statistics  31( 7): 799– 815. Google Scholar CrossRef Search ADS   Gombault, V., 2013. L’internet de plus en plus prisé, l’internaute de plus en plus mobile, INSEE Première 1452. Gonthier, D. J., Ennis, K. K., Farinas, S., Hsieh, H. Y., Iverson, A. L., Batáry, P., Rudolphi, J., Tscharntke, T., Cardinale, B. J. and Perfecto, I. ( 2014). Biodiversity conservation in agriculture requires a multi-scale approach. Proceedings of the Royal Society of London B: Biological Sciences  281( 1791): 20141358. doi:10.1098/rspb.2014.1358 Google Scholar CrossRef Search ADS   Gonçalves, B., Marques, A., Velho Da Maia Soares, M. A. and Pereira, H. M. ( 2015). Biodiversity offsets: from current challenges to harmonized metrics. Current Opinion in Environmental Sustainability  14: 61– 67. Google Scholar CrossRef Search ADS   Greiner, R. ( 2016). Factors influencing farmers’ participation in contractual biodiversity conservation: a choice experiment with northern Australian pastoralists. Australian Journal of Agricultural and Resource Economics  60( 1): 1– 21. Google Scholar CrossRef Search ADS   Greiner, R., Bliemer, M. and Ballweg, J. ( 2014). Design considerations of a choice experiment to estimate likely participation by north Australian pastoralists in contractual biodiversity conservation. Journal of Choice Modelling  10: 34– 45. Google Scholar CrossRef Search ADS   Haaijer, R., Kamakura, W. and Wedel, M. ( 2001). The ‘no-choice’ alternative in conjoint choice experiments. International Journal of Market Research  43: 93– 106. Hasan-Basri, B. and Karim, M. Z. A. ( 2013). The effects of coding on the analysis of consumer choices of public parks. World Applied Sciences Journal  22( 4): 500– 505. Jacob, C., Vaissière, A.-C., Bas, A. and Calvet, C. ( 2016). Investigating the inclusion of ecosystem services in ′ting. Ecosystem Services  21: 92– 102. Google Scholar CrossRef Search ADS   Kaczan, D., Swallow, B. M. and Adamowicz, W. L. ( 2013). Designing a payment for ecosystem services (PES) program to reduce deforestation in Tanzania: an assessment of payment approaches. Ecological Economics  95: 20– 30. Google Scholar CrossRef Search ADS   Kehlbacher, A., Balcombe, K. and Bennett, R. ( 2013). Stated attribute non‐attendance in successive choice experiments. Journal of Agricultural Economics  64( 3): 693– 706. Google Scholar CrossRef Search ADS   Kiesecker, J. M., Copeland, H., Pocewicz, A. and McKenney, B. ( 2010). Development by design: blending landscape level planning with the mitigation hierarchy. Frontiers in Ecology and the Environment  8: 261– 266. Google Scholar CrossRef Search ADS   Kontoleon, A. and Yabe, M. ( 2003). Assessing the impacts of alternative ‘opt-out’ formats in choice experiment studies: consumer preferences for genetically modified content and production information in food. Journal of Agricultural Policy and Research  5: 1– 43. Kuhfuss, L., Préget, R. and Thoyer, S. ( 2014). Préférences Individuelles et Incitations Collectives: Quels Contrats Agroenvironnementaux pour la Réduction des Herbicides par les Viticulteurs? Revue d’Etudes en Agriculture et Environnement / Review of Agricultural and Environmental Studies  95( 1): 111– 143. Google Scholar CrossRef Search ADS   Kuhfuss, L., Préget, R., Thoyer, S. and Hanley, N. ( 2016). Nudging farmers to enrol land into agri-environmental schemes: the role of a collective bonus. European Review of Agricultural Economics  43( 4): 609– 636. Google Scholar CrossRef Search ADS   Lancaster, K. ( 1966). A new approach to consumer theory. Journal of Political Economy  74: 132– 157. Google Scholar CrossRef Search ADS   Le Coënt, P., Préget, R. and Thoyer, S. ( 2017). Compensating environmental losses versus creating environmental gains: implications for biodiversity offsets. Ecological Economics  142: 120– 129. Google Scholar CrossRef Search ADS   Levrel, H., Frascaria-Lacoste, N., Hay, J., Martin, G., and Pioch, S. (eds.), 2015. Restaurer la nature pour atténuer les impacts du développement – Analyse des mesures compensatoires pour la biodiversité. Editions QUAE. Lienhoop, N. and Brouwer, R. ( 2015). Agri-environmental policy valuation: farmers’ contract design preferences for afforestation schemes. Land-Use Policy  42: 568– 577. Google Scholar CrossRef Search ADS   Luce, R. D. ( 1959). Individual Choice Behavior: A Theoretical Analysis . New York: Wiley. Maron, M., Brownlie, S., Bull, J. W., Evans, M. C., von Hase, A., Quétier, F., Watson, J. E. M. and Gordon, A. ( 2018). The many meanings of no net loss in environmental policy. Nature Sustainability  1: 19– 27. Google Scholar CrossRef Search ADS   Matthews, Y., Scarpa, R. and Marsh, D. ( 2017). Using virtual environments to improve the realism of choice experiments: a case study about coastal erosion management. Journal of Environmental Economics and Management  81: 193– 208. Google Scholar CrossRef Search ADS   McFadden, D. ( 1974). Conditional logit analysis of qualitative choice behavior. In: Zarembka P. (ed.), 1974. Frontiers in Econometrics . New York: Academic Press, 105– 142. MEDDE, 2013. Lignes Directrices Nationales sur la Séquence Eviter, Réduire et Compenser les Impacts sur les Milieux Naturels. MEDTL, 2012. Doctrine Relative à la Séquence Eviter, Réduire et Compenser les Impacts sur le Milieu Naturel. Menz, M. H., Dixon, K. W. and Hobbs, R. J. ( 2013). Hurdles and opportunities for landscape-scale restoration. Science  339( 6119): 526– 527. Google Scholar CrossRef Search ADS   Moreno-Mateos, D., Power, M. E., Comín, F. A. and Yockteng, R. ( 2012). Structural and functional loss in restored wetland ecosystems. PLoS Biology  10( 1): e1001247. Google Scholar CrossRef Search ADS   Navrud, S. ( 1997). Communication devices in contingent valuation surveys—Experiments with video. In: R. J. Kopp, W. W. Pommerehne and N. Schwarz (eds), Determining the Value of Non-Marketed Goods. Studies in Risk and Uncertainty 10 . Dordrecht: Springer, 273– 298. Google Scholar CrossRef Search ADS   Paolino, P. ( 2001). Maximum likelihood estimation of models with beta-distributed dependent variables. Political Analysis  9( 4): 325– 346. Google Scholar CrossRef Search ADS   Parkhurst, G. and Shogren, J. ( 2007). Spatial incentives to coordinate contiguous habitat. Ecological Economics  64( 2): 344– 355. Google Scholar CrossRef Search ADS   Parkhurst, G., Shogren, J., Bastian, C., Kivi, P., Donner, J. and Smith, R. ( 2002). Agglomeration bonus: an incentive mechanism to reunite fragmented habitat for biodiversity conservation. Ecological Economics  41( 2): 305– 328. Google Scholar CrossRef Search ADS   Patterson, Z., Darbani, J. M., Rezaei, A., Zacharias, J. and Yazdizadeh, A. ( 2017). Comparing text-only and virtual reality discrete choice experiments of neighbourhood choice. Landscape and Urban Planning  157: 63– 74. Google Scholar CrossRef Search ADS   Pe’er, G., Zinngrebe, Y., Hauck, J., Schindler, S., Dittrich, A., Zingg, S., Tscharntke, T., Oppermann, R., Sutcliffe, L. M. E., Sirami, C., Schmidt, J., Hoyer, C., Schleyer, C. and Lakner, S. ( 2017). Adding some green to the greening: improving the EU’s ecological focus areas for biodiversity and farmers. Conservation Letters  10( 5): 517– 530. Google Scholar CrossRef Search ADS   Quétier, F. and Lavorel, S. ( 2011). Assessing ecological equivalence in biodiversity offset schemes: key issues and solutions. Biological Conservation  144: 2991– 2999. Google Scholar CrossRef Search ADS   Quétier, F., Regnery, B. and Levrel, H. ( 2014). No net loss of biodiversity or paper offsets? A critical review of the French no net loss policy. Environmental Science and Policy  38: 120– 131. Google Scholar CrossRef Search ADS   Rambonilaza, M. and Dachary-Bernard, J. ( 2007). Land-use planning and public preferences: what can we learn from choice experiment method? Landscape and Urban Planning  83( 4): 318– 326. Google Scholar CrossRef Search ADS   Revelt, D. and Train, K. ( 2000). Customer-specific taste parameters and mixed logit: households’ choice of electricity supplier. Working Paper E00-274, Department of Economics . Berkeley: University of California. Rose, J. M. and Bliemer, M. C. J. ( 2013). Sample size requirements for stated choice experiments. Transportation  40( 5): 1021– 1041. Google Scholar CrossRef Search ADS   Ruto, E. and Garrod, G. ( 2009). Investigating farmers’ preferences for the design of agri-environment schemes: a choice experiment approach. Journal of Environmental Planning and Management  52( 5): 631– 647. Google Scholar CrossRef Search ADS   Rösch, V., Tscharntke, T., Scherber, C. and Batáry, P. ( 2015). Biodiversity conservation across taxa and landscapes requires many small as well as single large habitat fragments. Oecologia  179( 1): 209– 222. Google Scholar CrossRef Search ADS   Santos, R., Clemente, P., Brouwer, R., Antunes, P. and Pinto, R. ( 2015). Landowner preferences for agri-environmental agreements to conserve the montado ecosystem in Portugal. Ecological Economics  118: 159– 167. Google Scholar CrossRef Search ADS   Silverman, B. W. ( 1992). Density Estimation for Statistics and Data Analysis . London: Chapman & Hall. Soule, M., Tegene, A. and Wiebe, K. D. ( 2000). Land tenure and the adoption of conservation practices. American Journal of Agricultural Economics  84( 2): 993– 1005. Google Scholar CrossRef Search ADS   Stoate, C., Boatman, N., Borralho, R., Carvalho, C. R., Snoo, G. R. and Eden, P. ( 2001). Ecological impacts of arable intensification in Europe. Journal of Environmental Management  63: 337– 365. Google Scholar CrossRef Search ADS   Stoate, C., Báldi, A., Beja, P., Boatman, N. D., Herzon, I., van Doorn, A., de Snoo, G. R., Rakosy, L. and Ramwell, C. ( 2009). Ecological impacts of early 21st century agricultural change in Europe – a review. Journal of Environmental Management  91: 22– 46. Google Scholar CrossRef Search ADS   Tardieu, L., Roussel, S., Thompson, J. D., Labarraque, D. and Salles, J.-M. ( 2015). Combining direct and indirect impacts to assess ecosystem service loss due to infrastructure construction. Journal of Environmental Management  152: 145– 157. Google Scholar CrossRef Search ADS   Torres, A. B., MacMillan, D. C., Skutsch, M. and Lovett, J. C. ( 2015). ‘Yes-in-my-backyard’: spatial differences in the valuation of forest services and local co-benefits for carbon markets in México. Ecological Economics  109: 130– 141. Google Scholar CrossRef Search ADS   Train, K. E. ( 2009). Discrete Choice Methods with Simulation , 2nd edn. Cambridge, UK: Cambridge University Press. Google Scholar CrossRef Search ADS   Tucker, G. M., Allen, B., Conway, M., Dickie, I., Hart, K., Rayment, M., Schulp, C. and van Teeffelen, A., 2013. Policy Options for an EU No Net Loss Initiative. Report to the European Commission. Institute for European Environmental Policy, London. Vaissière, A. C. and Levrel, H. ( 2015). Biodiversity offset markets: what are they really? An empirical approach to wetland mitigation banking. Ecological Economics  110: 81– 88. Google Scholar CrossRef Search ADS   van Teeffelen, A. J., Opdam, P., Wätzold, F., Hartig, F., Johst, K., Drechsler, M., Vos, C. C., Wissel, S. and Quétier, F. ( 2014). Ecological and economic conditions and associated institutional challenges for conservation banking in dynamic landscapes. Landscape and Urban Planning  130: 64– 72. Google Scholar CrossRef Search ADS   Verboom, J., Foppen, R., Chardon, P., Opdam, P. and Luttikhuizen, P. ( 2001). Introducing the key patch approach for habitat networks with persistent populations: an example for marshland birds. Biological Conservation  100( 1): 89– 101. Google Scholar CrossRef Search ADS   Vermeulen, B., Goos, P. and Vandebroek, M. ( 2008). Models and optimal designs for conjoint choice experiments including a no-choice option. International Journal of Research in Marketing  25: 94– 103. Google Scholar CrossRef Search ADS   Wende, W., Tucker, G., Quétier, F., Rayment, M. and Darbi, M. (eds) ( 2018). Biodiversity Offsets – European Perspectives on No Net Loss of Biodiversity and Ecosystem Services . Cham: Springer. Google Scholar CrossRef Search ADS   Wätzold, F. and Drechsler, M. ( 2014). Agglomeration payment, agglomeration bonus or homogeneous payment? Resource and Energy Economics  37: 85– 101. Google Scholar CrossRef Search ADS   Author notes Review coordinated by Ada Wossink © Oxford University Press and Foundation for the European Review of Agricultural Economics 2018; all rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

European Review of Agricultural EconomicsOxford University Press

Published: Mar 29, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off