Effect of Gamification on Survey Participation – Experimental Evidence from Swiss Agriculture

The intervention is the chance and information about winning a voucher if the participants’ (farmer and seeds) chain is the longest. While all farmers will receive an email asking them to forward the survey to a worker, only 50% of the farmers and the following seeds will receive the information treatment as part of the invitation email. The treatment for the farmers will be assigned randomly. The treatment for the first worker will be based on whether their employer was assigned to the treatment group or the control group.

2025-06-29

2025-10-31

Chain length (number of waves the survey has been forwarded), which can be observed in the survey tool based on the participant's ID

1) Probability of farmers, seeds and sub-seeds forwarding the link
2) Probability of seeds and sub-seeds participating in the survey
Both are measured by the share of participants participating or forwarding (0/1), which can be observed in the survey tool.

Seasonal farmworkers in Switzerland are not always registered with authorities, which is why we lack complete lists of all migrant workers to sample from. Furthermore, workers are not organised, meaning there are also no non-governmental organisations with comprehensive lists of workers. However, since the employers are known to the authorities we aim to get in contact through the farmers. We receive a list from the Federal Office of Agriculture with contact details of around 1,500 Swiss horticulture farmers who work with seasonal migrant workers. The Federal Office of Agriculture will select the farmers based on random sampling. We first contact the farmers by personal email to explain the study. If they do not respond to the email, we will follow up with one reminder.
In the email, farmers will be asked to recruit a seasonal worker to participate in the initial stage of the chain (seeds). Once a worker is selected, the farmer copies the unique link to the survey and shares it with them. The workers will then be directed to the landing page of the survey, where they will first watch a video explaining the sampling process, then they will fill out the survey and provide their contact details (which is required to compensate participants). After completing the survey, they will be asked to forward the survey link to three other workers (subseeds) together with the prepared message. These subseeds again will be asked to fill out the survey and forward the link to three workers. The link can be forwarded directly through WhatsApp, email or by copying the link into another application. Each of the three links can only be used once, and the potential participants are asked if they have participated in the survey before to prevent multiple participation by the same person. If they have participated before, they will be asked to inform the sender of the link. So that they can send it to another person. The referral from subseed to subseed will be repeated until the respondents are not willing to pass it on or when the sample size has been reached, (Tyldum & Johnston, 2014) as further explained below. The process will be facilitated with the online tool Limesurvey, which also allows us to generate unique IDs to identify the referral chains (Ferrer-Rosende et al., 2023). The referral process will work autonomously between the participants. However, we will monitor the chains and interfere if we observe irregularities.

Randomisation of farmers is done by an office computer

The treatment for the farmers will be assigned randomly based on a list with the farmers' contact details. The treatment for the first worker will be based on whether their employer was assigned to the treatment group or the control group

No

No, cluster

Around 225 farmers and 120 workers (seeds)

Treatment and control group each: 750 farmers and around 225 workers based on which farmers forwarded the survey

We have around 1,500 farmer contacts. Assuming a 15% farmer participation rate (Zachmann et al., 2023a, 2023b), a 53% worker response rate (Ferrer-Rosende et al., 2023), we estimate a sample size of 225 farmers and around 120 workers. To determine the chain length, we’ll monitor key indicators to identify equilibrium. Based on earlier studies, we expect around 4–5 waves per seed (Ferrer-Rosende et al., 2023). In addition, we will conduct a power calculation comparing 50% treated chains with 50% control chains, targeting 80% power at 5% significance. Since gamification in RDS and in the agricultural contexts lacks precedent, we cannot pre-specify effect sizes. Therefore, we will recruit farmers in two waves, conducting a power calculation after the first wave.