Abstract
Agriculture plays a key role in preserving human health, mitigating climate change, and combating biodiversity loss. To achieve these objectives, one approach is to remunerate voluntary farmers for implementing agri-environmental specifications on their farms through payments for environmental services (PES). These payments are generally offered to farmers in the form of a fixed amount, in exchange for implementing predefined specifications. However, in the presence of information asymmetry regarding the costs borne by farmers to comply with these specifications, calls for tenders or agri-environmental (AE) auctions, which allocate payments to farmers who submit the most competitive bids, are often seen as a more cost-effective alternative. In auctions, the highest-ranked bids are selected first until a constraint set by the buyer is reached, usually a maximum budget or an environmental objective. This mechanism theoretically reduces the informational rents provided to farmers. In practice, AE auctions are not widely used in Europe and farmers' participation rates in existing programs are often low. The idea that auctions must be both efficient from the perspective of the buyer of environmental services (ES) and acceptable to farmers is at the heart of the analyses developed in this thesis. The performance of AE auctions depends on many parameters that are extensively studied in the literature, but some, such as the type of constraint announced by the buyer or the scale of the auction, remain under-studied. This thesis, structured in four chapters, explores different aspects of AE auctions design to improve both their performance and their acceptability among farmers. The first chapter proposes an experimental comparison of an auction with an objective constraint (typically studied in auction theory) and an auction with a budget constraint (generally used in practice). This chapter relies on an innovative approach based on an online decontextualized experiment combining the strategy method and the simulation of auction results corresponding to all possible cost realizations for different groups of participants. The results indicate that, for an equivalent average budget, announcing a budget constraint is more effective than announcing a target constraint. The budget constraint allows for acquiring more ES on average. This is explained by lower bids when the announced constraint is a budget, as well as by a mechanical effect : even with identical bidding strategies, the budget constraint produces better results than the objective constraint. Indeed, the budget constraint enables acquiring more ES when bids are low (lower payments) and acquiring fewer when bids are high (higher payments). The first chapter is complemented by a technical appendix that tests, through simulations, the robustness of the results based on the payment rule, the level of competition, and whether the units exchanged are divisible. The second chapter extends the previous one by introducing a new auction format : the double constraint. In this format, the highest-ranked bids are selected until one of the two constraints, budget or objective, is reached. The results show that announcing both constraints simultaneously does not allow the buyer to obtain as many ES as when only one of the two is announced, but they are obtained at a lower average unit cost. The buyer must therefore trade off between obtaining a greater quantity of ES or maximizing the cost-effectiveness of the auction. Other treatments with a relatively higher budget level reveal that simultaneously announcing a larger budget constraint with the same objective constraint deteriorates the cost-effectiveness of the double constraint auction. The third chapter proposes to evaluate whether, to purchase a given quantity of ES, a buyer would benefit from organizing a single large-scale auction or several small-scale auctions. The analysis shows that increasing the size of an auction reduces the unit cost for the buyer without significant impact on the informational rents obtained by the bidders. The fourth chapter examines the acceptability of auctions from the farmers' perspective based on a survey experiment. The results of this survey conducted with over 4,000 farmers show that, on average, farmers are in favor of an allocation of PES through auctions compared to a fixed-price allocation on a "first come, first served" basis. However, farmers' preferences are subject to a large v vi Abstract heterogeneity and depend on their characteristics. Furthermore, respondents show a preference for bidding on the terms of the specifications rather than on the payment amount or on both simultaneously. The least popular option is the auction based solely on the payment, even compared to a fixed-price allocation. This thesis thus makes significant contributions to the study of reverse auctions and clarifies the role of key parameters in the performance and acceptability of AE auctions. The combined use of the strategy method and simulations ensures that the experimental results do not depend on random cost draws. Moreover, the results highlight the advantages and disadvantages of each type of constraint, as well as the interest in implementing AE auctions on a large scale. Finally, the survey conducted with a large sample of farmers shows that most of them are not opposed to this mode of PES allocation and the analysis of their responses clarifies the determinants of this acceptability of AE auctions.