Influence of Network Exposure and Spillover Effects on Adoption Rate of Plant Food Production

Farms within Canton Lucerne are randomly assigned to one of three arms — two treatment arms (advisory services (Arm 1); advisory services plus an experimental budget (Arm 2)) and an internal control (Arm 3); farms in the external control can-tons (Aargau, Thurgau, and St. Gallen) are assigned directly to the external control (Arm 4), without random assignment among themselves.
• Arm 1 — Advisory services (n≈75): Farms first receive an initial advisory session: a three-hour on-site consultation conducting a location and poten-tial analysis with the farm operator, enabling an informed choice of a subse-quent tailored advisory track: (1) specialty crops or specialty arable farming, (2) Agri-PV combined with specialty crops, (3) regenerative plant food pro-duction, (4) agroforestry, or (5) a voucher for an equivalent advisory service of the farm's own choice. Tailored advisory is delivered by regional public advisory services and specialised consultancies.
• Arm 2 — Advisory services plus experimental budget (n≈50): Farms re-ceive the same initial and tailored advisory as Arm 1, plus a one-time un-conditional cash payment of CHF 10,000, disbursed in July 2027 via the cantonal agricultural authority, conditional on completion of the initial adviso-ry session. The same fixed sum is paid to every Arm 2 farm irrespective of farm size. Prior to disbursement, Arm 2 farms submit a non-binding intention document describing their planned use of the budget; no repayment obliga-tion applies if a farm subsequently withdraws or uses the funds differently than stated.
• Arm 3 — Internal control (n≈300–600): Located within the treatment region (Canton Lucerne); receives no treatment; participates in all three survey waves and provides farm census data access.
• Arm 4 — External control (n≈300–600): Located in Cantons Aargau, Thur-gau, and St. Gallen; receives no treatment; same survey and census partici-pation as Arm 3.
Both control groups (Arms 3 and 4) are jointly eligible for a prize draw after each survey wave (20 farms, CHF 500 each, drawn from both groups combined), as compensation for survey participation rather than a treatment component.

2026-09-14

2029-02-28

• Overall plant food production change: 7-point Likert item, retrospective 5-year (baseline wave only) and prospective 12-month and 5-year (all waves)
• Area, tree count, and livestock-unit (GVE) change:
o retrospective (baseline only) — a single item (Δ ares, agricultural land for human food production)
o prospective (all waves) — a full multi-category battery in all waves, covering agricultural land for human food production, agricultural land for feed production, specialty crops excluding vineyards, vineyards, greenhouse production, and tree count (all in ares or count), plus live-stock units by category (cattle, pigs, poultry, sheep/goats, other) in GVE

• Network Exposure (pre-treatment baseline):
• Primary specification:
NE_i^primary=A_i/B_i
Where Ai = N(contacts who have expanded plant food production, excluding feed), and Bi = N(total contacts with whom the farmer exchanged about plant food production, excluding feed)
• Alternative specifications: Weighted by raw adopter count:
NE_i^alt1=A_i/B_i ×A_i=(A_i^2)/B_i


• Weighted by intensity of interaction:
NE_i^alt2=A_i/B_i ×I_i
where I_i= mean interaction intensity (7-point Likert, "not at all intense" to "very intense") across the farmer's exchange contacts.


• Spillover effects:
• Geographical Proximity (baseline): farm's own canton (Lucerne / external canton) * majority contact location (neighbourhood / within canton / external canton)
• Membership Association (baseline): none / Lucerne farmers Association / IP Suisse / Bio Suisse / Sector/professional association / other


• Mediating Variables
• Spillover effects:
Intensity of Interaction (Y1, Y2): self-reported exchange intensity (Likert scale) with contacts identified as advisory- or experimental-budget recipients under MehrWert | Pflanzen.

• Network Exposure weighted by raw adopter count: As an alternative to the primary exposure ratio, we test a specification that additionally weights exposure by the absolute number of contacts who expanded their plant food production (Bi). Under the ratio alone, a farmer with one contact who adopted (NE=1.0) scores higher than a farmer with ten contacts, four of whom adopted (NE=0.4), despite the latter having more farmers in absolute terms in their network who expanded their plant food production. The alternative specification accounts for this by scaling the ratio by the raw adopter count.
• Network Exposure (pre-treatment baseline) weighted by intensity of interaction:
As a second alternative to the primary exposure ratio, we test a speci-fication that weights exposure by the intensity of interaction with adopting contacts, on the grounds that not all exchange relationships carry equal informational or normative weight. A high exposure ratio driven by superficial contact plausibly has less influence on the farmer's own decision-making than the same ratio sustained through frequent, in-depth exchange.

The study follows all four arms across three data-collection waves over approximately three years (baseline September–October 2026, Year 1 November 2027, Year 2 November 2028), with an optional Year 3 wave in November 2029 pending a decision on study extension.

Farm-level survey data are linked to farm census data via farm identification number across all waves, allowing comparison of stated (survey) and revealed (census) preferences. We recruit from Canton Lucerne as the treatment region and from Cantons Aargau, Thurgau, and St. Gallen as an external control region with comparable farm structural characteristics. Within both regions, farms are eligible if they are located in the respective recruitment region, are eligible for Swiss direct payments, and operate animal production at the time of recruitment (standard livestock unit > 0). Contact data for outreach is provided by the Federal Office for Agriculture (BLW); farms already at the BLW's cap on survey invitations in the current calendar year cannot be proactively invited but remain eligible and may register by contacting the project team directly; eligible farms are then granted access to the survey. Eligible farms with available contact data are invited near-exhaustively within each region. The study deliberately targets farms that have not already fully committed to plant food production; subject to sufficient overall sample size, the analysis may exclude farms for which plant food production already constitutes 91–100% of agricultural income at baseline, and farms that indicate at baseline they would not accept either treatment if assigned, with such restrictions pre-registered and applied only if they do not compromise statistical power. Eligible farms receive a written invitation letter and follow-up email, with a reminder letter and email sent to non-respondents two weeks later, and two information events (in-person and hybrid) before registration closes at the end of October 2026. Farms are informed of their arm assignment by letter post and email in early December 2026; farms assigned to Arms 1 and 2 must actively confirm participation via a confirmation link. Replacement farms are recruited from the pool of registered control farms with similar characteristics on pre-specified balancing variables where withdrawals or non-confirmations occur.

Not available

Following completion of survey registration, eligible farms in Canton Lucerne will be randomly assigned to the three study arms: 75 farms to Arm 1, 50 to Arm 2, and the re-maining eligible farms to Arm 3. Farms in the external control cantons (Aargau, Thurgau, and St. Gallen) will be assigned directly to Arm 4 and will not enter this randomization procedure.
We will use rerandomization to improve covariate balance. Candidate allocations will be generated by computer and evaluated using Mahalanobis distance across a set of co-variates grouped into priority tiers, including baseline outcomes, geographic character-istics, and other farm-level characteristics. Balance thresholds for each tier will be cali-brated beforehand using simulated allocations and fixed prior to the final randomiza-tion.
For the final randomization, a new and independent set of allocations will be generated by computer, and the allocation ultimately used will be drawn uniformly at random from those satisfying all balance criteria. Randomization will take place after survey registra-tion closes but before treatment delivery and before any post-treatment outcomes are observed.

The individual farm.

No

-

The target sample across all arms is approximately 725-925 farms.

The target sample per treatment arm is the following: Arm 1 n ≈75; Arm 2 n≈50; Arm 3 n≈300–600; Arm 4 n≈300–600