Input- or output-based contracts in payments for environmental services: Interacting impacts of uncertainty, partial observability, and collective action

How to cite

Malagutti, Flavio. 2026. "Input- or output-based contracts in payments for environmental services: Interacting impacts of uncertainty, partial observability, and collective action." AEA RCT Registry. AEARCTR-0019432. https://doi.org/10.1257/rct.19432-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Flavio Malagutti*

* Primary investigator

Registry ID
AEARCTR-0019432
DOI
https://doi.org/10.1257/rct.19432-1.0
First registered
Last updated
Status
Completed
Trial dates
2026-06-05 – 2026-06-18
Countries
South Africa ()
Registry record
http://www.socialscienceregistry.org/trials/19432

Abstract

This pre-analysis plan describes a research project with smallholder cattle farmers in the Eastern Cape region of South Africa. The broad goal of this study is to understand how the design of public good contracts shape individual contributions to the public good and resulting public good provision. The study design is a framed field experiment (a lab-in-the-field). Participants face five contracts varying multiple features of the payment scheme. Each contract is played both individually and in a group, so the design identifies how each contract feature performs by itself and how moving from an individual principal-agent relationship to a principal-collective agent setting affects performance. The study speaks to the design of payments for ecosystem services on communal land and to the classic input-versus-output contracting problem under partial observability.

Experimental design

The core of the data collection instrument is a multi-round public-goods game played on individual tablets.
The framing of the games is around carbon storage: contributions to the global public good are framed as being contributions towards carbon storage in the community's rangelands.
Participants play all public-goods games both in groups of about five players (a standard public-goods design) and individually (a group of one).
Group membership is reshuffled across games and drawn from the entire session of participants, so that over the course of a session a participant interacts with essentially all session-mates.

The games were designed as follows.
In each round, participants receive an endowment of 100 tokens and decide whether to fund rotational grazing, a binary choice ($g = 20$ tokens or $g = 0$), and unobserved contributions, $e$, representing a continuous choice of an additional conservation margin that others cannot observe, from zero up to the total remaining budget.
At the end of each round, the number of participants who contributed to rotational grazing are observable to all participants within the group, while individual decisions and unobserved contributions $e$ are not observable other than through their effect on the final aggregated payment outcome, which is contract-dependent and explained below.

We designed five different payment rules for the public-goods game and refer to them as contracts. Each contract is played twice: first solo (a single-person group), which serves as the individual principal-agent benchmark, and then in a group. The combination (contract by individual-or-group mode) is referred to as a game. One game is randomly selected for payment, and the number of paid rounds per game is determined by a block-random-termination rule.

The five contracts vary three design features: (a) whether the contract pays for observable (grazing) inputs or for measured output, (b) whether the output is exposed to a production shock, and (c) whether, in addition to the observed contributions, the output is affected by an additional unobservable margin, in which the individual chooses her contribution without revealing any signal to the group.

The full carbon (public good) production function is set to $F(G,E) = (2G + 1.5E)\varepsilon$, where $G$ and $E$ are the sums grazing and unobserved (extra) contributions in the community that is playing together, respectively. The production shock $\varepsilon$ is drawn from a two-point (Bernoulli) distribution with equally likely outcomes 0.5 and 1.5. The shock is therefore mean-preserving, and it is applied to the public good production after all contributions have been summed but before payments are distributed. Therefore, the contracts that inherit risk pay participants on the final outcome after the risk shock. All five contracts pay participants for their grazing contribution. Furthermore, the expected private marginal return of this contribution is set equal across all contracts within the same game-mode (solo or collective), ensuring equivalence in grazing incentives. On the other hand, only some contracts pay for unobserved contributions. Contracts C3 and C5 offer a positive return for extra contributions, while the other contracts do not.

The order in which the contracts appear is randomized across sessions. In every session, participants always begin with playing C1 then C2. From then on, the order of C3, C4, and C5 varies by community. For each contract, participants play the solo mode first, followed by the collective mode.

All in all, experimental assignment happens in two ways. First is within individuals since we have every individual engage with different types of contracts in the lab. From this engagement, we extract within-individual changes in behavior caused by the contract features they are facing. The second experimental assignment is in the order that these contracts appear. This assignment is at the session level. The variation is in which contract features comes first, second, third, fourth and fifth. That is, the order in which different contract features appear to all participants in a specific session.

Design details

Randomization method
The randomization method here applies to the session-level order of contract features. This was pre-determined before visiting the community. We had a list of communities to visit and arbitrarily chose which contract order we would show for participants in that session. This works for causal identification in our case because we make sure that all possible orders are represented and because our primary identification is within individual.
Randomization unit
Farming communities
Clusters
8
Observations
200
Arms
8 farming communities, with 25 participants in each community for a total of 200 participants
Primary outcomes
Our primary outcomes of interest come from the public goods game. They are the individual-level decisions to fund rotational grazing and unobserved contributions.

Cited by

No citation data available for this registration yet.

Keywords