Nudging farmers to adopt more climate-friendly practices

This study presents an experiment conducted across 11 municipalities, and aimed at promoting climate-conscious farming practices through nudging interventions. The municipalities were selected based on specific criteria, including having more than 100 farms, over 10% of cropland on organic soils, and a moderate proportion of forage grassland.
We identified pairs of similar municipalities, from which we randomly selected experimental and control municipalities. The intervention included two nudge letters. The experimental design, randomized at the municipal level, sought to minimize the cross-contamination of treatment effects, acknowledging the potential influence of local discussions among farmers.

The intervention itself consists of a set of two letters sent to the trial group and no letters sent to the control group.

1) First Nudge Letter (July 2022): This letter aimed to strengthen the positive identity factors, that were identified to underlie farmers attitudes. Specifically it thanked the farmers on their existing actions, addressed questions of shared responsibility, and aimed dismantle oppositions related to climate issues.

2) Second Nudge Letter (August 2022): This letter provided up-to-date information on climate actions and attitudes of other farmers overall, and on the situation of the recipient farmers' municipality/region especially related to cultivated organic soils. It also highlighted environmentally friendly farming, particularly focusing on three practices: winter cover crops, cultivation of perennial crops on organic soils instead of annual crops, and afforestation of retired or low-yield fields. These practices represent varying degrees of ease of implementation and environmental impact. The climate impact of the three actions were communicated in the letter, as well as additional information on these practices in a separate sheet. The aim of the letter was to, first, allow for accurate social comparison, and, next, to provide practical steps on how to implement in practice. It also utilises injunctive messaging at the level of the municipality via either a positive, neutral, or negative emoticon based on how the municipality compares to regional mean percentage of cultivated organic soils with perennial crop coverage.

2022-07-11

2022-09-01

Individual level measures. Areas and shares of environmentally positive and negative farming practices (in terms of greenhouse gas emissions from agricultural land use) in relation to the total cultivated area of the farm. Considered practices / climate actions: winter cover crops, shifting from annual to perennial crops, and afforestation of abandoned and low-yielding fields.

Areas will be measured as hectares per year.

The experimental design divides the 11 municipalities into treatment and control areas. Experimental design was formed in two phases. At first, we formed pairs from 10 rural municipalities, which were as similar as possible in terms of descriptive spatial characteristics derived from register data.

The characteristics used as criteria in pair formation were:
-Total cultivated area in the municipality (in hectares)
-Number of farms in the municipality
-Distribution of farms in the municipality by size (size defined in terms of hectares of arable land)
-Distribution of farms in the municipality by type (livestock, crop, other)
-Distribution of cultivated area in the municipality by soil type (organic soil, mineral soil)
-Distribution of cultivated area on organic soils in the municipality by crop (annual, perennial)
-Whether the municipality is part of the carbon-neutral municipalities network (HINKU 2008 -programme)

In this way, we formed five pairs of municipalities. From each pair, one municipality was randomly assigned to the control group and the other to the treatment group. Randomization was carried out by coin flip in the office under the watchful eye of four researchers.

In the second phase we randomized North Ostrobothnia's regional center munincipality Oulu's postal code areas to the treatment and control groups. Oulu was distinguished from the other municipalities firstly because it stands out from the other regions of the region as an urban economic center. In addition, the structure of Oulu's agriculture was different from the other municipalities on the basis of the criterion variables used in the comparison. In the case of Oulu, we removed from the postal code list the urban areas of the city centre where it was reasonable to assume that there were no farms. Map services, zoning data, satellite and aerial photographs were used to identify these areas (Oulu Region map services and Google Maps were used as sources). The remaining 30 postal code areas were randomized into experimental groups using a random number generator called Research randomizer: https://www.randomizer.org/.

After defining the treatment area as a set of postal codes, we requested the contact details of all the farmers in the area from the Finnish Food Authority. The size of the treatment group was eventually 927 farms. This included all farmers who had applied for crop subsidies in the treatment area in the previous season and who had not refused to give up their contact details for research purposes. The treatment letters were sent to the treatment group during summer 2022.

Not available

Coin flip in the first phase, random number generator in the second phase

Municipalities in the first phase, postal codes in the second phase

Yes

A total of 40 clusters, 10 of these are municipalities in the first phase randomization, and 30 are postal codes in the second phase of randomization. The farmers in 5 of the municipalities and in 15 of the postal codes are treated and farmers in the remaining 5 municipalities and 15 postal codes are untreated.

Pretrial estimate of 3 601 farms according to total number of farms in the experimental area's munincipalities according to statistics administered by Natural Resources Institute Finland that was a partner in the research project. After receiving the contact list from Finnish Food Authority, the approximated sample size was revised to be 1920 because part of the farms had refused to give their contact details for research purposes and we were not therefore able to sent the intermention letter to these farms. The proportion of farms that had not made this ban was about 53.3% of all farms in the treatment area. We estimated the sample size using the same proportion of all farms in the experimental area. The sample size will be refined in the analysis phase when we request comparable data from the Finnish Food Authority for the control area.

5 municipalities and 15 postal codes treated; 5 municipalities and 15 postal codes untreated.

With an estimated intra-class cluster coefficient of 0.01, a within-cluster standard deviation of 10, desired power of 0.80, and an alpha level of 0.05, the minimum detectable effect size in terms of Cohen's d is approximately 0.155, equal to an 1.55% percentage point difference.