Financial incentives often fail to reconcile agricultural productivity and pro-conservation behavior

The paper addresses whether financial incentives can reconcile agricultural productivity with pro-conservation behavior in contexts where pro-environment practices can yield private benefits. Many environmental management challenges involve excludability and collective action dilemmas, where prevailing responses (e.g., pesticides, lethal control, fencing) can conflict with conservation. Direct payments are common, scalable policy tools, yet evidence on whether they improve cooperation and align conservation with development goals is mixed and context-dependent. The authors synthesize data from dynamic, spatially explicit experimental games across Europe, Africa, and Asia examining three dilemmas with potential for ecosystem services to enhance production: replacing pesticides with insect-based services (NonCropShare), sharing damages while using non-lethal wildlife control (GooseBump), and managing shared agricultural land with fallowing (SharedSpace). The core research question is: in contexts where pro-environment practices may yield private gains, do payments help achieve joint production–environment outcomes?

The study situates itself within literature on collective action in natural resource management and conservation conflicts, noting mixed evidence on the effectiveness of payments for environmental services (PES) and other incentive-based instruments in aligning conservation with livelihoods. Prior reviews and meta-analyses highlight that PES performance depends on context, design, and implementation, with potential trade-offs when environmental goals displace production. Alternatively, in cases like conservation agriculture, ecosystem service enhancements can deliver private production benefits, potentially enabling synergies. The literature also emphasizes social capital and trust as drivers of pro-environment behavior, and points to challenges of scale mismatch and spillovers (e.g., displacement or leakage of production). Gender composition in resource governance shows mixed empirical results, though some evidence suggests more balanced gender representation can improve collective outcomes. Experimental games are recognized as valuable but limited tools for exploring behavior and policy interventions in complex social–ecological systems.

Design: The authors synthesized results from seven datasets comprising three families of dynamic, spatial, NetLogo-based experimental games representing collective action dilemmas in conservation: (1) NonCropShare (insect-based ecosystem services and pesticide use), (2) GooseBump (human–wildlife coexistence with non-lethal/lethal control and habitat set-aside), and (3) SharedSpace (shared-property land use with crop vs fallow/forest decisions and soil productivity dynamics). Games were played by groups of four participants on 3×3 or 6×6 landscapes with simultaneous decisions per round. Treatments varied the presence and magnitude of payments/subsidies targeting pro-conservation actions or outcomes (e.g., habitat set-aside, non-lethal control, fallowing), and in some cases property rights structures (private vs open access) and information visibility. Games incorporated uncertainty, nonlinearity, temporal variation, and spatial externalities. Sampling and settings: More than 1000 participants across seven countries (Cambodia, Vietnam, Gabon, Tanzania, UK/Orkney, Kenya, Madagascar) made over 14,000 decisions. Recruitment approaches were context-specific (village lists, markets, community events), generally targeting agricultural decision-makers. Compensation varied by site; many sites provided payments equivalent to a day’s wage and/or bonuses, though Orkney participants were not paid. Game details: NonCropShare involved choices among heavy/light pesticide use, crop planting, and non-crop habitat; coordination could allow natural enemies to replace pesticides. GooseBump involved choices among cropping/livestock raising, scaring wildlife, lethal control, and wildlife habitat; animal movement and damage created spatial externalities. SharedSpace involved choices between cropping and conserving fallow/forest; repeated cultivation reduced plot productivity, recoverable through fallow and neighboring vegetation. Outcomes and variables: For each round, three standardized outcomes were constructed: (i) production (sum of agricultural yields across players, net of subsidies), (ii) environment (game-specific: animals persisting in landscape for GooseBump; total fallow/forest for SharedSpace; cells not heavily sprayed for NonCropShare), and (iii) joint environment–production product (environment × production). All outcomes were standardized to z-scores at the dataset level, with the joint product computed from raw environment and production before standardization. Explanatory variables included observables (age, education, gender composition, primary/secondary farming income), a payment treatment dummy, and, for a 5-country subset, beliefs (e.g., trust in community/government, perceived wellbeing impacts of conservation, social norms, penalty importance) and context measures (e.g., relationship strength among players, motivations, dependence on others’ choices, consideration of intergenerational or other players’ impacts). Likert scales were harmonized to 0–1 and variables normalized (dataset-level z-scores). The unit of analysis was the game round; standard errors were clustered at the game session level. Analyses: Pooled regressions across all seven datasets examined associations of payments and observables with environment, production, and joint outcomes (Table 1). For the 5-dataset subset, models compared explanatory power of payments alone vs adding observables, beliefs, and context variables (Table 2), and then jointly estimated associations for conservation, production, and the joint outcome (Table 3). The study emphasized qualitative inference and comparative patterns across contexts rather than precise parameter estimates. Data and analysis scripts were archived and made accessible via Zenodo and Binder.

• Payments and joint outcomes: Across contexts, direct payments for pro-conservation actions generally increased pro-conservation practice but reduced production, often failing to improve the joint environment–production outcome. Only in NonCropShare did payments sometimes improve the joint outcome. In pooled models, payments tended to relax production while raising conservation, implying trade-offs rather than win–wins.
• Observables: Higher education, greater gender diversity, and increased representation of women in groups were associated with higher production and improved joint environment–production outcomes. Age and farming as main/secondary income showed context-dependent associations.
• Beliefs and context: Stronger relationships among group members and greater trust in community coordination were associated with greater pro-conservation practice but lower production, indicating that social cohesion may shift behavior toward conservation at some productive cost. Trust in national institutions did not explain variation in outcomes.
• Quantitative highlights (5-dataset subset, Table 3): Payment was negatively associated with the joint outcome (coefficient approximately −0.527, p<0.01), positively with conservation (+0.198, p<0.01), and negatively with production (−0.581, p<0.01). Education and female representation were positively associated with production and the joint outcome, while relationship strength was negatively associated with production and positively with conservation. Model R-squared values were ~0.126 (conservation), ~0.193 (production), and ~0.179 (joint outcome).
• Heterogeneity across games: The structure of the dilemma matters. In NonCropShare, coordination can align ecosystem service gains with yields, making win–wins attainable; in GooseBump and SharedSpace, production and conservation are more often in tension, with payments tending to displace production.
• Implication: Payments alone often fail to capitalize on potential win–wins and may need redesign to target both conservation and production objectives jointly, or be complemented by other instruments.

The findings address the core question by showing that while payments can promote pro-conservation actions, they frequently do not deliver joint improvements in production and environmental outcomes, except in contexts where ecosystem services clearly translate into private gains (e.g., coordinated low-pesticide regimes in NonCropShare). The degree of inherent trade-off versus synergy, determined by ecological and institutional structure and spatial–temporal scales, constrains the scope for incentives to induce win–wins. Mismatches in scale (e.g., societal biodiversity benefits vs local production costs) and spillovers/leakage can further undermine local gains. Design implications include calibrating incentives jointly to conservation and production outcomes to stimulate innovation and risk-taking that can reconcile goals (e.g., combined bonuses for habitat and yield; spatially coordinated agglomeration payments). Complementary instruments—such as insurance for wildlife damage risks, support for alternative livelihoods, and bundling of payments with social and developmental supports—may sustain outcomes without unsustainable funding. Technical assistance (e.g., transitioning to ecologically based management) and institutional strengthening at community levels can further enable coordination. The role of group composition and social context is salient: education, gender diversity, and women’s representation correlate with better production and joint outcomes, suggesting that inclusive group design can enhance collective performance. Strong intra-group relationships and trust induce conservation-oriented behavior, albeit sometimes at the expense of yields, highlighting trade-offs that programs must anticipate. While experimental games have limitations in external validity, they provide comparative insights across diverse settings, informing program design to better align livelihoods and conservation.

This synthesis shows that financial incentives often fail to reconcile agricultural productivity with pro-conservation behavior, except in settings where ecosystem services provide clear private benefits enabling coordinated win–wins. Payments generally increase conservation actions but reduce production, yielding limited improvements in joint outcomes. Group characteristics—higher education, gender diversity, and women’s representation—are associated with better production and joint performance, while strong social ties can favor conservation with potential yield costs. Policy and research should focus on: (i) joint incentive schemes targeting both conservation and production outcomes; (ii) spatially coordinated incentives (e.g., agglomeration payments); (iii) complementary risk-transfer instruments (e.g., insurance) and support for alternative livelihoods; (iv) bundling payments with social and technical support; and (v) inclusive program design to leverage diverse groups. Future research should test such designs in field settings over longer horizons, assess cross-scale spillovers, and further explore how beliefs, social networks, and institutional contexts mediate outcomes. Advances in experimental game design to incorporate longer temporal and broader spatial perspectives and non-economic behaviors will strengthen inference for sustainability policy.

• External validity: Experimental games stylize complex realities, bound options, and cannot match real-world stakes; behaviors observed may not fully generalize to field contexts.
• Context heterogeneity: Datasets differ in sampling strategies, compensation schemes (e.g., unpaid Orkney case), and local socio-ecological conditions, potentially affecting comparability.
• Short time horizons: Games capture short-term decisions; many ecosystem service benefits (and costs) manifest over longer periods.
• Measurement and harmonization: Outcomes and belief/context variables were harmonized and standardized, potentially obscuring context-specific nuances; analyses emphasize qualitative inference over precise parameter estimates.
• Scale and spillovers unobserved: Broader market-mediated spillovers or leakage beyond the game landscape are not directly captured.