Collective Action in Irrigation Systems: Understanding Strategic Public Goods Reliance and Cooperation

Last registered on November 17, 2025

Pre-Trial

Trial Information

General Information

Title
Collective Action in Irrigation Systems: Understanding Strategic Public Goods Reliance and Cooperation
RCT ID
AEARCTR-0017211
Initial registration date
November 11, 2025

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
November 17, 2025, 7:09 AM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Region

Primary Investigator

Affiliation
University of New South Wales

Other Primary Investigator(s)

Additional Trial Information

Status
In development
Start date
2025-11-21
End date
2026-01-31
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Sri Lanka’s village tank cascade systems are among the oldest community-managed irrigation systems in the world. They form the backbone of smallholder agriculture in the dry zone, supporting rural livelihoods through the collective management of tank-stored water and its distribution through shared canal networks (Ratnayake et al, 2021). The irrigation systems have endured for centuries, sustained not only by economic incentives but also by long-standing norms, cultural expectations, and informal institutions that govern cooperation among farmers (Ostrom, 1990; Ostrom and Gardner, 1993).

In these systems, Farmers’ Organisations (FOs) act as local social planners. Before each cultivation season, they assess the tank’s water level, determine the recommended extent of land to be cultivated, and coordinate collective canal cleaning to remove weeds, silt, and debris. These actions enable water to flow efficiently from the head end closest to the tank through the middle section to the tail end. In principle,
such coordination allows the community to achieve a social optimum: cultivating the maximum possible area under water constraints while maintaining the canals in good condition. However, individual incentives often diverge from this collective optimum. Both key decisions, which are how much land to cultivate and how much labour to contribute to canal cleaning, are vulnerable to free riding. Studies on gravitational irrigation systems have identified spatial asymmetry as a major factor contributing to this divergence (Dayton-Johnson, 2000b; Dayton-Johnson and Bardhan, 2002). Upstream, or head-end, farmers can over-cultivate and extract more water with little risk, whereas downstream farmers, located in the middle and tail ends, face shortages even when they comply with FO recommendations.

Despite these structural inequalities, empirical research shows that many irrigation communities have developed mechanisms to overcome asymmetric benefits and sustain cooperation (Weissing and Ostrom, 1991;Ostrom et al., 1994; Janssen et al., 2010). Through local rule enforcement, mutual monitoring, and shared norms, farmers have historically maintained collective canal cleaning and equitable water distribution. During years of good rainfall, this location asymmetry remains hidden by the abundance of water, but under drought conditions, unequal access becomes sharply visible. However, increasing climatic uncertainty now threatens these informal equilibria through frequent and unpredictable droughts. Climate variability amplifies the spatial dependence of irrigation outcomes, transforming irrigation management into a coupled-commons dilemma in which two interdependent public goods —canal condition and irrigation water —must be sustained through voluntary cooperation.

Recent empirical evidence from Sri Lankan tank systems illustrates this challenge. Analysis of secondary data and field surveys conducted between December 2024 and May 2025 reveals a clear behavioural pattern: in drought years, both middle- and tail-end plots were cultivated to a lesser extent than recommended by FOs. Farmers in these locations reported deliberately reducing cultivated area to manage water risk, but also decreasing canal-cleaning hours because they expected limited returns from a system in which upstream farmers overused water. These location-specific responses reveal a form of strategic adaptation: farmers in disadvantaged positions withdraw from both land use and collective labour in anticipation of others’ defection. This behaviour, where private adaptation to scarcity (reducing cultivated extent) crowds out contribution to the enabling public good (canal maintenance), is the central empirical puzzle motivating this study. The key question is whether such restraint by downstream farmers during drought also leads to reduced canal maintenance effort, thereby accelerating collective decline. Over repeated droughts, even small reductions in canal effort across many farmers could accumulate into long-term deterioration of water conveyance and institutional capacity.

To identify the mechanisms linking these behaviours, the study conceptualises farmers’ seasonal decisions as a coordination game. In good seasons, when water is abundant, the game simplifies to a single public goods problem, canal cleaning, where cooperation achieves the social optimum but remains vulnerable to free riding. In drought seasons, however, the game becomes a nested coordination problem: canal cleaning and cultivation extent interact. Water allocation depends on upstream over-cultivation, so downstream farmers’ payoffs are jointly determined by others’ cultivation decisions and total canal cleaning effort. Each farmer must decide whether to continue contributing labour to a canal that may not deliver sufficient water. This study, therefore, focuses on the behavioural responses of farmers at different spatial positions, head, middle, and tail ends, under varying environmental and institutional conditions. By experimentally replicating these irrigation environments, the research isolates the causal effects of drought and upstream defection on cooperative labour, providing new evidence on how spatial asymmetry and climate variability jointly shape collective action in self-managed irrigation systems.
External Link(s)

Registration Citation

Citation
Gnanasubramaniam, Sharunya. 2025. "Collective Action in Irrigation Systems: Understanding Strategic Public Goods Reliance and Cooperation." AEA RCT Registry. November 17. https://doi.org/10.1257/rct.17211-1.0
Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
Experimental Details

Interventions

Intervention(s)
The experimental features include a framed coordination game that simulates irrigation decisions made by farmers under varying environmental and spatial conditions. The experiment consists of multiple rounds representing cultivation seasons, where participants make private, incentivised choices about labour allocation to canal cleaning (a public good) versus off-farm work (a private activity). Also, incentivised measures include two belief elicitation questions on expectations about others’ cultivation and canal cleaning decisions.
Intervention (Hidden)
Intervention Start Date
2025-12-01
Intervention End Date
2026-01-31

Primary Outcomes

Primary Outcomes (end points)
Irrigation canal cleaning hours in each round
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Relate to belief formation and coordination of expectations. These are elicited through two incentivised and structured “guess” questions in each round:
• Guess on Land: participant’s belief about the average number of extra acres cultivated by upstream farmers beyond Farmers' organization's recommendations
• Guess on Canal Cleaning: participant’s belief about the average number of canal cleaning hours contributed by others
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Each session consists of 18 farmers, divided into three groups of six who share a simulated canal. Within each group, farmers are randomly assigned spatial positions—two head-end, two middle-end, and two tailend plots. Each participant controls a 5-acre plot and 10 labour hours per season, which they can divide between:
• Canal cleaning (public activity): each hour contributes Rs. 4 per acre increase in their revenue from paddy harvest.
• Selling vegetables (private activity): each hour yields Rs. 600.

Earnings from paddy cultivation depend on the water received and the total canal cleaning hours by the group. Water availability varies by season and location.

Each session lasts ten rounds, representing consecutive cultivation seasons:
• Rounds 1–3: Good seasons (pre-shock)
• Rounds 4–6: Drought seasons (shock)
• Rounds 7–10: Good seasons (post-shock recovery)

Participants are not told how many periods over which they will make decisions, and so may think of the task as an infinite, discrete choice game. Participants earn real cash based on their decisions, with total payments capped at Rs. 1,000 per person, including a Rs. 200 show-up fee.

Participants will be randomised at the individual level, in equal proportions, into 1 of 3 groups. Within the group, they will be randomly assigned to one of the locations (head, middle and tail):
1. Group 1 – During drought rounds, upstream over-cultivates compared to what is recommended by FO. During drought rounds, upstream (head-end) farmers are told they have water to cultivate all 5 acres. Middle-end farmers are told they have water to cultivate 2 acres, and tail-end farmers only 1 acre, due to upstream over-cultivation. During good rounds, everyone has been told they have enough water to cultivate all five acres.

2. Group 2 – During drought rounds, upstream compliance to what is recommended by FO: Head-,middle-, and tail-end farmers all have water to cultivate 3 acres, as everyone follows the FO recommendation. During drought rounds, upstream (head-end) farmers are told they have water to cultivate all 5 acres. Middle and tail-end are told that their plots have enough water to cultivate three acres, as the farmers above their plots followed the FO’s recommendation and cultivated only three acres, so they have just enough water for three acres. During good rounds, everyone has been told they have enough water to cultivate all five acres.

3. Group 3 (Control) – Good seasons only: All farmers have enough water to cultivate 5 acres in every round. No water scarcity.

After receiving information about the season and water availability, each farmer privately completes a decision sheet where they record:
• Incentivised guess on Land for middle and tail end farmers: expected number of extra acres cultivated by upstream farmers.
• Incentivised guess on Canal Cleaning: expected average canal cleaning hours by others.
• Labour allocation: how many of their 10 hours to allocate to canal cleaning and how many to selling vegetables.

No participant is informed of others’ decisions. Facilitators collect all decision sheets each round and calculate payoffs based on the group’s total canal cleaning effort. At the end of each round, participants are informed privately of:
• Their total earnings from vegetable selling, own canal cleaning, and others’ canal cleaning.
• The total hours of canal cleaning in their group.
• The water available to their plot and their paddy income.

After all ten rounds, facilitators conduct a short exit survey collecting non-incentivised measures: risk preferences, reciprocity, altruism, and trust. Also, demographic and socio-economic data. Each participant’s final payment is determined by selecting one random round, adding any bonus earnings for accurate guesses, and including the Rs. 200 participation fee.
Experimental Design Details
Randomization Method
The farmers' full population list under the selected irrigation systmes will be used as the sampling frame and randomization done in office by a computer
Randomization Unit
Individual
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
N/A
Sample size: planned number of observations
279 with 10% non respondent rate
Sample size (or number of clusters) by treatment arms
93 individuals for control
93 individuals for upstream farmer's compliance
93 individuals for upstream farmer's non-compliance
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Minimum detectable size is 0.10 SD between compliance and non-compliance groups. The MDE for location heterogeneity is 0.32 for middle-end and 0.34 SD for tail-end, comparing with head-end
Supporting Documents and Materials

Documents

Document Name
Experiment script
Document Type
other
Document Description
This document contail entire script used in the experiment
File
Experiment script

MD5: 5bde8f24e69d03a422ffd372d59de7fd

SHA1: f3788eedd319377a74875919e717bd1995706455

Uploaded At: November 09, 2025

Document Name
Exit survey instrument
Document Type
survey_instrument
Document Description
This survey instrument is used after the experiment
File
Exit survey instrument

MD5: 177bb1f54de20125b1c6d9956a4071e1

SHA1: fa718eb57636bd8b38ca6e31ebbb43e2d599d6d4

Uploaded At: November 09, 2025

IRB

Institutional Review Boards (IRBs)

IRB Name
Human Research Ethics and Compliance Committee, University of New South Wales
IRB Approval Date
2025-05-20
IRB Approval Number
iRECS7927
Analysis Plan

Analysis Plan Documents

Pre- Analysis Plan

MD5: 3f56700ccd82837b0b00426f3f31bfcf

SHA1: 9dfb0eb70ff3150b93b7fb1c741cef44563ce41f

Uploaded At: November 11, 2025

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials