1. There is growing interest in the potential of payments for ecosystem services (PES) to encourage land managers to protect and enhance the environment. However, questions remain about how PES agreements should be designed. There is a division between schemes that structure payments by action or by results, with most biodiversity PES schemes, including European agri-environment schemes, paying by action; for example incentivising land managers to carry out actions believed to increase biodiversity. Payment by results is a common incentive structure in the private sector (e.g. labourers doing piece work or no-win no-fee lawyers) but rarer in PES.
2. Using a theoretical modelling approach, we investigate the conditions under which each way of structuring payments may be more cost-effective in a biodiversity PES.
3. Payment by action is favoured where there is a clear action that can be specified at an appropriate level and to which biodiversity is sensitive. We found that payment by results is favoured in degraded landscapes as incentives are created for managers to use their private knowledge and join the scheme only if they can produce the biodiversity services targeted by the scheme. Payment by results is also favoured where biodiversity is less sensitive to conservation action and when it is difficult for a central agency to determine an appropriate level of conservation action. This is because payment by results allows individual managers to optimise their level of action.
4. The relative cost of monitoring action (compliance with an agreement to manage in a certain way) versus results (the presence of biodiversity) has a substantial effect on which payment structure is more efficient only when the central agency can accurately set an appropriate level of action. We illustrate these principles with examples based on agri-environment schemes.
5.Synthesis and applications. Payment by results deserves more attention from those designing biodiversity PES (be they agri-environment schemes in agricultural landscapes or direct payment schemes in more intact ecosystems). This paper provides a formal framework to help policy makers identify the conditions under which payment by results or payment by action is most likely to yield cost-effective biodiversity conservation.
To maximise efficiency, buyers of an ecosystem service should seek suppliers who can supply the good at the lowest cost. A challenge in the design of PES schemes is that those supplying services have more information about the costs and feasibility of supply than potential buyers. Such hidden information about costs and demand makes it difficult to design a cost-effective scheme (Moxey, White & Ozanne 1999; Ferraro 2008; Jack, Leimona & Ferraro 2008). With a flat-rate payment scheme, the highest uptake will be from suppliers with the lowest costs of fulfilling the criteria, who are consequently likely to be over-rewarded (Emerson & Gillmor 1999; Sanchez-Azofeifa et al. 2007), and there may also be uptake from those unlikely to produce the benefit. Making payments based on individual costs or selecting only regions where the benefit is likely to be produced requires substantial information, could reduce competition between suppliers and may be politically undesirable (Vickery et al. 2004). Selection contracts (Moxey, White & Ozanne 1999) and auctions (Stoneham et al. 2003; Ferraro 2008) are potential solutions for screening suppliers or creating incentives for suppliers to reveal their true costs. However, contract screening to identify suppliers with low production costs, although theoretically attractive, is difficult to implement (Ferraro 2008).
A solution that avoids the need to directly select suppliers by cost is to base payment on service production (payment by results), which creates incentives for only those suppliers able to produce the service at low cost to join the scheme and allows suppliers to use their private knowledge to maximise service production. Under any contract (whether paying by action or by results), some proxy of compliance must be measured, making monitoring central to the design of any contract (Bolton & Dewatripont 2004). Any such proxy measures used by the buyer should be ‘costly-to-fake signals’ (Ferraro 2008) or there is a risk of signals being produced but no service delivered. Payment by results requires that service delivery is measured, potentially increasing transaction costs over payment by action (where only compliance with management agreements needs to be monitored). When paying by results, suppliers risk not producing the service despite investing effort, or producing the service but the agency not detecting it. Risk-averse suppliers are therefore likely to require risk premiums when paid by results (Bolton & Dewatripont 2004).
Payment by results is common in many markets (for example piece work or no-win no-fee court cases). Although the number of PES schemes paying by results (also called conservation performance payments) rather than action is increasing (Ferraro & Gjertsen 2009; Zabel & Roe 2009), many are small initiatives. Most PES schemes pay by action. For example, large biodiversity-motivated programmes such as European agri-environment schemes typically offer payment for compliance with a set of management requirements (Vickery et al. 2004 but see Musters et al. 2001; and Wittig, Kemmermann & Zacharias 2006; for examples of payment by results) and most biodiversity PES schemes in low-income countries base payments on performance relative to indicators of action (e.g. presence or absence of wildlife snares) rather than directly measuring service provision (Milne & Niesten 2009).
To explore these issues, we examine payments by action or results for the special case of biodiversity PES. Several studies have considered the issues involved in paying for biodiversity conservation by actions or results (see for example Musters et al. 2001; Verhulst, Kleijn & Berendse 2007; Ferraro & Gjertsen 2009; Zabel & Roe 2009). However, a thorough comparative analysis within a consistent theoretical framework is lacking (Watzold & Schwerdtner 2005). Here, we develop such a framework and use a simulation model to consider the influence of properties of the targeted biodiversity service (sensitivity to management action and detectability), the management action (detectability) and existing distribution of the service in the landscape, on the provision of biodiversity services. We explore the conditions that favour payment by action or results in terms of the biodiversity gains produced for a limited budget. We then illustrate the model using examples based on agri-environment schemes.
Materials and methods
A conceptual framework for scheme design
We start by considering a simple system with one agency, one land manager and one patch. The manager is risk neutral and seeks to maximise net income from the patch. The agency has a limited budget and seeks to increase biodiversity service provision from the patch by providing a payment as an incentive for the manager to increase biodiversity. This payment may be to carry out an action known to increase biodiversity (payment by action) or may be based on the level of biodiversity service provision (payment by results). We conceptualise the biodiversity service as the presence of a species or habitat of conservation concern in the patch. The agency seeks to structure the payment to maximise the biodiversity gain for the available budget. To determine payment level, the agency must monitor the level of action when paying by action or the level of biodiversity service provision when paying by results.
A simple scheme is where there is a single action that increases provision of the biodiversity service, and the manager receives the payment in direct proportion to their level of action or the level of provision. For a given monitoring cost, the payment structures will be of equal efficiency if there is a level of action that maximises both manager net income and biodiversity provision. In this case, the agency should simply choose the scheme with the lowest monitoring costs. However, there are many ways in which real systems are more complex. For example, the relationship between the level of action and level of biodiversity provision is unlikely to be linear. Monitoring is imperfect, so the probability of detecting biodiversity is a function of monitoring effort and the nature of the biodiversity. Finally, the agency is likely to offer the scheme to multiple heterogeneous managers who will produce different levels of biodiversity at a given level of payment.
We use a simulation model to explore the effect of heterogeneity in landscape patches and diminishing returns of biodiversity provision to increasing levels of action within a patch. These assumptions are applicable to real-world payment schemes that reward habitat management (such as leaving field areas fallow) with a constant per unit cost and increase the probability of occurrence on the patch of widespread species that occur at low abundance. We provide a specific example of this in the later case study.
We assume that the managers know their patch quality and the costs of actions so that they are able to optimise the level of action to maximise their net income. Although in reality land managers will not have perfect information, we feel this assumption is justified as land managers tend to know their land at a fine scale (history and current status) and their operating costs. They are therefore in good position to take such information into account when making decisions and to learn from the outcomes of past decisions.
The agency seeks to maximise biodiversity service gain across the landscape, while the managers seek to maximise individual net income. Biodiversity service gain is measured as the increase in the mean probability of the service being present in a patch across the whole landscape. Population size and geographic range are important predictors of extinction for species (Purvis et al. 2000), and extent is an important measure of threat to habitats (Rodriguez et al. 2011). Therefore, the more patches supporting the biodiversity benefit, the better its conservation status. The agency has knowledge of the overall landscape quality (distribution of the biodiversity service) but no knowledge of individual patches. For list of model parameters and descriptions, see Table 1.
Table 1. Model parameters and descriptions
Probability of benefit occurring in patch with no action
Level of action carried out in patch
Probability of benefit occurring in patch with action set to L
Sensitivity of biodiversity to action
Payment to manager for setting action to L (£)
Payment to manager for producing biodiversity on patch (£)
Cost to manager of setting level of action to L (£)
Cost to manager of scheme membership (£)
Fine for non-compliance (£)
Manager net income from patch (£)
Agency cost for patch being in scheme (£)
Gain in probability of benefit from individual patch
Mean gain in probability of benefit in landscape
Total agency budget (£)
Cost of agency monitoring (£ h−1)
Time spent monitoring patch (h)
Probability of the agency detecting non-compliance in patch
Probability of agency visiting manager for non-compliance monitoring
Probability of the agency detecting the benefit in patch
Probability of non-compliance detection with 1 h of monitoring
Probability of detecting benefit with 1 h of monitoring
Multiple of CL=1 paid to the manager by agency
Managers choose whether to join the scheme and, if they join, their level of management action (L). Managers join the scheme if their expected income is positive. Under payment by action, the agency prescribes a level of action (at L ≥ 1) and rewards the manager with a payment (DL=1). Under payment by results, the agency rewards the manager with a payment (DP) based upon the observed biodiversity provision in their patch, and managers choose the level of action that maximises their expected net income. Although there are numerous social influences on individual decision-making (Edwards-Jones 2006), we treat managers as risk-neutral profit maximisers, a common assumption in models of manager behaviour (e.g. Hart & Latacz-Lohmann 2005; Ohl et al. 2008). Contract theory (Bolton & Dewatripont 2004) shows that risk aversion by the agency and manager can substantially change scheme efficiency and we return to this point in the discussion.
The modelled landscape has 1000 heterogeneous patches, each with an existing probability of the biodiversity service occurring (P0 where 0 ≤ P0≤ 1). We model a range of landscapes from degraded, where most patches have a low initial probability of producing the service (a left-skewed distribution of the probability of service occurring), to landscapes, where most patches are already producing the service (right-skewed distribution). With a given level of management action, L, the probability of the service occurring increases from P0 to PL (where P0≤ PL≤ 1). The cost of action to a manager (CL) is linearly proportional to the level of action. The action provides no direct economic benefit to the manager, so in the absence of payments, the manager chooses no action (L = 0). Managers incur a fixed cost (Cm) representing the administrative burden of scheme membership. The probability of the service occurring is a diminishing returns function of the level of effort (L):
where a is the sensitivity of the service to action. The probability of occurrence (PL) increases with higher values of a, action (L) and background probability of service occurrence (P0) (Fig. 1). Under this function, managers with high P0 incur higher costs of increasing provision than those with lower values. Note that PL is the probability of the service occurring on the patch, i.e. the service is either present or absent in a given patch. However, the mean outcome across the landscape is equivalent to PL representing a patch level of service quality scaled between zero and one.
Costs and benefits under payment by action
Under payment by action, the manager is rewarded for carrying out the action at level L ≥ 1. There is no incentive to choose L > 1, so if the manager joins and complies, they chose L = 1. If the manager joins the scheme and does not comply, they choose L = 0.
The manager’s expected income, I, given scheme membership, is therefore given by:
where Pc is the probability of non-compliance detection by the agency, F is the fine for non-compliance, Cm is the fixed cost of scheme membership (the administrative costs), CL = 1 is the cost to the manager of the action, and DL = 1 is the payment for the action. Pc is a function of the probability of the agency visiting a manager (Pv) and the time (t) the agency spends on each visit (see equation 12, for details).
The manager joins the scheme if the expected return from the scheme, I, is positive. The agency can ensure this by setting: DL = 1> CL = 1+Cm. Given scheme membership, the manager will choose to comply if PcF > CL = 1. We fix the fine F at 2DL = 1 and assume that Cm is 10% of CL = 1. These assumptions mean that the cost of non-compliance is the administrative cost plus a penalty equal to the fixed payment. Reducing the fine F increases monitoring costs as more patches need to be visited to ensure compliance. We also express DL = 1 as a multiple (x) of CL = 1, so the agency only sets x; I is then:
The expected cost to the agency, E, is then given by:
and the expected biodiversity service gain, G, by:
When paying by action, the agency does not monitor the service and has no estimate of P0, PL = 1 or G.
Costs and benefits under payment by results
Under payment by results, the manager is rewarded according to the level of the biodiversity service that the agency estimates for the patch. The manager chooses whether to enter the scheme and the level of action (L) to carry out (which may be zero).
The manager’s expected income I for entering the scheme is:
where Ps is a function of PL and detectability of the benefit (equation 13). The manager chooses the level of action (Lopt) that maximises I. If Lopt exists for which I is positive (DPPs> CL+ Cm), the manager joins the scheme. Using the same assumptions as for payment by action, I becomes:
The expected agency cost (E) is given by:
and the expected biodiversity service gain (G) by:
By monitoring a patch within the scheme, the agency has an estimate of PL and consequently the biodiversity service benefit produced across all patches.
Monitoring of the biodiversity service and compliance
For both schemes, the agency must decide the proportion of patches visited and how much effort to spend monitoring each visited patch. Monitoring when paying by action determines which managers should not be paid (because of non-compliance). Monitoring when paying by results determines the payment individual managers receive. If, when paying by action, the agency does not visit all managers, then it assumes that managers not visited are complying. Lax compliance monitoring creates incentives for managers not to comply. However, as described previously (by setting PcF> CL = 1), the agency can set monitoring levels to ensure compliance. When paying by results, the agency visits all patches, as managers are rewarded in proportion to their patch status.
At the individual patch level, we assume the same simple detectability function for non-compliance as in payment by action and for biodiversity service provision when paying by results (Fig. 2):
where d is the detectability of non-compliance (dc) or the service (db) when monitoring for 1 h and 0 < d < 1. The function is asymptotic, so probability of detection is always <1. For simplicity, we assume that under both payment structures, the agency chooses the length of time to monitor each visited patch that ensures a high detectability (of 0·95):
Hence, d determines the proportion of the agency’s budget that is spent on monitoring. At the landscape level, our assumptions about the number of patches visited and the monitoring levels in each patch mean that for a given value of d, agency monitoring costs will be higher when paying by results than by action. In the special case where the optimal level of action under payment by results is L = 1, payment by action is always more cost-effective.
When paying by action, the probability of detecting non-compliance (Pc) is a function of the probability of the agency visiting the manager (Pv), monitoring time (t), and the probability of detecting non-compliance with 1 h of survey time (dc):
with the agency setting Pv>1/1·8x to ensure compliance.
When paying by results, the agency surveys the patch to estimate the probability of the service occurring (Ps), a function of the probability of the service actually occurring (PL), the detectability of the service (db) and agency monitoring time (t):
with the agency setting t, so that a survey detects the service with a probability of 0·95.
Costs for different values of detectability dc and db are illustrated in Fig. 2.
Agency budget, distribution of patch properties and landscape-level gain
We represent the distribution of the initial probability of the service occurring (P0) across the landscape using a discrete approximation to the beta distribution (see Appendix S1, Supporting Information for more details). We interpret a very left-skewed distribution as a degraded landscape with a high proportion of patches having a low probability of containing the biodiversity service and a very right-skewed distribution as a landscape where most patches have high probabilities of containing the service.
The agency has a limited budget (B) and maximises the mean expected biodiversity gain (Gm) across the 1000 patches in the landscape for this budget. When B is low enough that not all managers can join the scheme, members are assumed to be a random sample of those willing to join. The only decision the agency makes is the payment level offered, D. We found the optimal payment level for each scheme that maximises Gm, (see Appendix S2, Supporting Information for optimisation method).
Biodiversity gain increased continuously with budget when paying by results (with the potential to set L > 1; Fig. 3). When paying by action, biodiversity increased only up to the threshold where all managers joined the scheme (setting L = 1). Therefore, when paying by action, scheme efficiency is highly dependent on the level of action set by the agency, whereas under payment by results, managers use their private knowledge to set efficient levels of action. Under both schemes, increased monitoring costs reduced overall biodiversity gain for a given budget.
Payment by results produced a higher biodiversity gain for a given budget than payment by action when budgets were high; the distribution of the initial biodiversity service was left skewed; the service was insensitive to conservation action; and compliance was costly to monitor relative to service provision (Fig. 4). Where the service was sensitive to action (high a), payment by action was favoured across a wider range of landscapes because the prescribed action level delivers high service levels, making self-selection by managers less important. When biodiversity response is insensitive, high levels of action may be required to produce results, making high payments to few managers a better strategy for the agency, incentivising those with lower P0, and hence more potential gain, to join the scheme. Payment by results is favoured in left-skewed landscapes because of the assumption of diminishing marginal returns to action (Fig. 2). Here, the majority of managers have high potential to increase biodiversity and fewer have low potential; in this case, levels of action below the fixed level prescribed under payment by action may produce substantial biodiversity benefits at a lower cost than payment by action. The cost of monitoring only has a strong effect on the choice of scheme when budgets are low, resulting in incomplete participation, and when the cost of compliance monitoring is relatively high compared with service monitoring. Conservation budgets are usually low, so the cost of monitoring may often influence scheme choice in real situations.
Under both schemes, the proportion of managers joining the scheme increased with total budget until all managers joined (Fig. S1, Supporting Information). Under payment by results, the distribution of the initial probability of producing the service had a strong effect on the proportion of managers joining the scheme; in degraded landscapes, where most patches have a low probability of producing the service (left skewed), more managers joined for a given budget than in landscapes where the majority already produced the service. There was no such relationship under payment by action. This is because when paying by results, the agency is able to use the offered payment level to trade-off the agency monitoring costs against action per manager. In degraded landscapes, managers can produce relatively large gains for little action, so the agency can set payments relatively low and allow many managers to join the scheme. This carries higher monitoring costs, because the agency must visit all managers to determine the level of results, but keeping payments low means the overall budget is met. In more right-skewed landscapes, the cost to individual managers of increasing the service provision is larger, so the agency switches to fewer managers receiving higher payments for a higher level of action, but with lower overall monitoring costs (Fig. S2, Supporting Information).
The mean level of risk incurred by the managers was proportional to the optimal payment multiplier (x) offered by the agency (Fig. S3, Supporting Information). High values of x indicate that, for expected gains to be produced, managers must invest in substantial amounts of action, with a high risk of not receiving payment because the biodiversity service is not present. Manager risk was constant across all landscapes and budgets when paying by action, but increased with the budget and right skew when paying by results.
Illustrative case study
Agri-environment schemes provide financial incentives to farmers (the manager of the land) for adopting agricultural practices that benefit biodiversity (Kleijn et al. 2006; Whittingham 2007, 2011). We consider how the landscape context of the scheme, the responsiveness of the target to management and detectability influence whether payment by results or payment by action would produce a greater biodiversity benefit for a given level of investment. The examples, while grounded in the available literature and the types of scheme on offer, are chosen to illustrate the main points that can be drawn from our model and are not intended to be detailed analyses of whether payment by results or effort should be adopted in any particular situation.
Actions aimed at improving biodiversity will be more successful in some areas than others (Whittingham 2007) depending on factors such as land use and the biology of target species (Kleijn & Sutherland 2003). Where farmers have useful knowledge about the state of the species and the likelihood of action to increase numbers, payment by results can avoid wasted payments by the agency. For example, skylark plots are left as bare ground in arable fields to provide nest sites and access to food for skylarks Alauda arvensis, a declining farmland bird (Morris et al. 2004). If managers can use their knowledge to place plots only on land likely to attract nesting skylarks, then payment by results may be the most efficient approach. Payment by action risks farmers in unsuitable locations joining the scheme and being paid to carry out ineffective action.
How responsive the target is to management action?
Some potential targets of agri-environment schemes may respond strongly to possible management prescriptions, whereas others may not. For example, there is evidence that redshanks Tringa totanus have shown a better response to agri-environment schemes than lapwings Vanellus vanellus (Wilson, Vickery & Pendlebury 2007). Payment by results will be favoured for species where the response to management actions is weak, because it may be that individual farmers are willing to invest in substantial amounts of action when they have private information that their patch has high potential. Where the species is sensitive to management actions, self-selection by farmers is less important because the majority of farmers will get a response to their actions and so payment by action is favoured across a wider range of conditions.
Monitoring habitat condition or species presence/absence
Agri-environment schemes aiming to increase biodiversity on farmland may have a variety of targets (ranging from a single species to habitat condition). How difficult the target is to monitor, relative to management actions, will influence whether payment by action or results is favoured. For example, some agri-environment schemes aim to increase wetland bird numbers by incentivising farmers to restrict stocking density, avoid disturbing birds and control in-field and ditch water levels. Where the target is the presence of a bird species, monitoring by results may be favoured because it is relatively straightforward to detect the presence of birds on the land relative to monitoring compliance with management prescriptions. Paying farmers based on the presence of breeding wading birds on their land has been trialled with some success in the Netherlands (Musters et al. 2001). Where the target is habitat based, however, monitoring performance against the target may be more difficult, thus favouring payment by action.
We have presented a framework for analysing the advantages and disadvantages of two alternative payment structures for PES schemes (payment by action and payment by results). Our results suggest that, with limited agency knowledge and a diminishing response of biodiversity service to action, payment by results is favoured in circumstances where managers are able to use their knowledge to maximise biodiversity benefits; these include when biodiversity is less sensitive to conservation actions and when landscapes are degraded, as managers will join the scheme only if they can produce the biodiversity service. A successful example of this is a scheme in Cambodia that facilitates payments from birdwatchers to community groups dependent upon whether the visitors see all the target bird species in the area (Clements et al. 2010). Participating villages have used their understanding of the local situation to develop and enforce rules about which species are protected and put in place agreements to protect nesting and feeding sites.
For the benefits of payment by results to be fully realised, managers need to set efficient levels of action requiring knowledge of costs and of the relationship between the level of action and the level of biodiversity service. For managers to then invest useful effort, the agency measurement of results must be a good measure of manager effort. Where there is a weak relationship between effort and results, managers are unlikely to be willing to invest in action. This is a well-understood finding from contract theory where it is recognised that payment by results is only efficient when the indicator used is a reliable signal of effort (Bolton & Dewatripont 2004).
We assumed that all managers and the agency are risk neutral but a thorough exploration of more realistic distributions of attitudes to risk would provide greater insight into optimal PES design. If managers were risk averse, a risk premium would be required when paying by results with a consequent reduction in the biodiversity gain achievable for a given budget. In a specific, very simplified case of a risk-neutral agency and a single risk-averse manager, the optimal scheme is a fixed payment even if the level of action is not observed (c.f. Chapter 4, Bolton & Dewatripont 2004), suggesting that payment by action would be favoured by risk-averse managers. However, it is likely that a conservation agency will also be risk averse and will seek to make lower payments to managers when the biodiversity outcome is not certain. Where agency and managers are risk averse, an efficient scheme would include risk sharing, such as a hybrid scheme with a fixed payment plus a variable payment for results. The relative level of the two payments would then be proportional to the relative risk to the manager and agency. Our model would allow testing of this approach by changing the scheme membership costs Cm, which we interpret as fixed costs but could be interpreted as a risk premium. Manager risk would be mitigated in a system where individual managers were in control of multiple patches, such as agri-environment schemes where a combination of scheme options is applied at the farm level. In cases where there is a time-lag between an action and biodiversity gains being produced, the discount rate (including risk discounting) of managers will influence manager’s willingness to accept payment by results. Again hybrid schemes, where part of the payment is up-front, based on action, and part is given when results are achieved, may also be useful in dealing with time-lags and are seen in some payment by results schemes in operation (e.g. Musters et al. 2001).
From a conservation perspective, an advantage of payment by results over payment by action is that the monitoring involved provides useful information about biodiversity status. There are increasing calls for evidence-based decision-making in conservation and environmental management (Pullin & Stewart 2006) but to build up an evidence base, monitoring of project outcomes is necessary. Under payment by results, monitoring of the biodiversity on participating patches is an integral part of the scheme. Under payment by action, monitoring provides information on the proportion of managers complying with management agreements, but no information about biodiversity status (although high rates of non-compliance may suggest to the agency that manager costs are higher than estimated, and this information can be used to guide scheme adjustment). In our model, the relative cost of monitoring actions or results only had a strong effect on the preferred payment structure where the agency could set an appropriate level of action and where budgets were relatively low. This is because even with relatively low detectability, the majority of the costs within each scheme are incurred by the managers. In more extreme cases, for example monitoring for highly threatened but very difficult (costly) to detect species, monitoring costs may favour payment by action unless the agency is also willing to pay for knowledge of species status.
The design of a biodiversity PES is simpler when the objectives are very specific (e.g. to increase the population of a single species or area of habitat). However, many PES have more broadly defined objectives and aim to improve conditions for a number of species and/or habitat quality concurrently. Our framework could be extended to consider biodiversity PES aimed at multiple species. In fact, payment by results may be of particular value where there is a multi-species target and species’ needs are not in agreement (Drechsler et al. 2007), making prescriptive actions difficult to define. Not all existing payment by results schemes are for single species. For example, a German scheme that uses payment by results to improve grassland biodiversity on farms has a suite of 43 indicator plant species (Wittig, Kemmermann & Zacharias 2006). Conversely, payment by action will be favoured where multiple species respond positively to a single action as it is likely that monitoring compliance will have less cost than determining presence of all species of interest.
The optimal design of PES would greatly benefit from further investigation, including both empirical analysis of the results of PES schemes and extensions to our modelling framework. For example, it is likely that scheme membership in adjacent patches will increase the probability of a biodiversity service being present compared with separated patches (Dallimer et al. 2010). Therefore, explicit consideration of spatial heterogeneity would allow the additional benefits of cooperation by neighbouring land managers to be taken into account and estimates of any increased benefit from rewarding spatial agglomeration (Parkhurst & Shogren 2007) when paying by action or results. Finally, in reality, land managers are unlikely to have perfect knowledge of their probability of producing the biodiversity service, and the effect of such uncertainty also needs exploring.
Payments for ecosystem services offer exciting opportunities for increasing the efficiency of conservation interventions but there has been relatively little formal exploration of possible ways to structure such schemes. This study provides an analytical framework for analysing the effects of one key design consideration: payment by action or results in the context of biodiversity services. Many of the parameters would be costly to accurately estimate, so we do not suggest that those designing schemes should exactly parameterise the model. However, exploring the sensitivity around rough parameter estimates would inform scheme choice. We also see the contribution of the model as providing a framework enabling explicit consideration of the factors influencing the efficiency of scheme design. We demonstrate that payment by results can create incentives for land managers to use their private knowledge and skills to produce biodiversity services. In degraded landscapes, this can reduce the costs of conservation substantially, as agencies can structure schemes to avoid paying for actions on patches with a low probability of producing a biodiversity service. The field of PES is still in its infancy and although theoretical research such as this can and should inform the design of schemes, empirical research will also be needed for conservation scientists to fully develop their understanding of the optimal design of such schemes under varying real-world conditions.
We thank Isabel Pau, Lauren Coad, Gareth Edwards-Jones, Aidan Keane, the late Simon Thirgood and Matthew Sommerville for discussion and the editor and three anonymous reviewers for comments. JPGJ, EJMG and EN were funded by the Leverhulme Trust. EJMG acknowledges a Royal Society Wolfson Research Merit award, and EN a Marie Curie Fellowship.