The longer-term adoption patterns of organic farming practices

After the baseline data collection, we invited the 20 interviewed farmers in each treatment village to participate in a three-day training on organic farming, with farmers receiving seven hours of training per day. Based on the insights from the theoretical framework, our training intervention addressed experience, beliefs and risks. It offered extensive information about the costs and benefits of organic farming practices and hands-on training to maximize the learning effect and minimize the risk of usage. Raising awareness, changing perceptions and implementation at the community level are intended to address social risks. The training provided farmers with an introduction to organic farming and included information on potential marketing channels. Particular emphasis was placed on practical activities such as making organic fertilisers and pesticides. The training was designed by AOI in cooperation with the research team from Germany. AOI, together with its institutional members (local NGOs), delivered the training, which was held in the villages to minimize travel time for respondents. The farmers received IDR 50,000 (around USD 3.5) for each day of the training (only if they attended) to cover any transport costs and to compensate them for potentially forgone earnings. A first round of training was rolled out at the end of March 2018 and completed in May 2018. On average, 17 farmers out of the 20 invited farmers attended each training day.

A second round of training was rolled out in July 2022 and completed in August 2022. We invited the same farmers, i.e. the treatment group, to the training as in 2018. The training was designed together with the Indonesian Soil Research Institute and trainers from two P4S centers with a focus on organic farming. P4S are self-help agricultural and rural training centers. These institutions are owned and managed by farmers. They exist in most districts in Indonesia and receive financial resources from the local government. During the training, trainers provided partly a refresher training of the first training. They further augmented the training by a discussion of soil management in organic farming and by training on soil tests. Based on the soil tests that each farmer could conduct for his or her own soil, trainers provided fertilizer recommendations according to organic principles.

2022-08-01

2022-08-31

We are interested in studying the effect of treatment on several outcomes of interest. All outcomes will be measured in the fourth wave of the survey which will take place between February and March 2023. We differentiate between three families of outcomes: (i) adoption, (ii) knowledge, and (iii) perception.


Family 1: Adoption
Domain 1

1. Organic fertilizer
o Fermented manure (binary variable) =1 if the respondent applied fermented manure. Fermented manure may be produced by farmers themselves or bought.

o Organic fertilizer (binary variable) =1 if the respondent applied organic fertilizer other than manure during the last planting season. The organic fertilizer may be produced by farmers themselves or bought.

2. Organic pesticide:
o Organic pesticide (binary variable) =1 if the respondent applied self-produced or purchased organic pesticide during the last season.

3. Plant residues:
o Returned pant residues to the soil (binary variable) =1 if the respondent returned at least part of the plant residues to the soil and did not burn the remaining part.

4. Sum of organic practices used
This outcome will be a count variable from 0 to 4 for the number of practices applied. It will be coded as 4 for farmers who applied fermented manure, organic fertilizer, organic pesticide and who returned plant residues to the soil. Both purchased and self-produced inputs will be considered for this variable.

Domain 3: Chemicals

• Expenditure on chemical fertilizers in IDR/ha (continuous variable): reported expenditure for chemical fertilizers which were applied during the last planting season. This variable will be top-coded at the 95th percentile of the overall distribution.
• Expenditure on chemical pesticide in IDR/ha (continuous variable): reported expenditure for chemical pesticides which were applied during the last planting season. This variable will be top-coded at the 95th percentile of the overall distribution.



Family 2: Knowledge
The second family of outcomes will focus on knowledge.
Knowledge Score: This outcome will be a count variable from 0 to 6 for the number of correct answers to six knowledge questions. Some of the questions are open-ended, this reduces the probability that respondents get the answer right by chance. The following knowledge questions will be considered:

1. A farmer who sells his/her products as organic is allowed to a) use some chemical inputs but less than for conventional farming b) no chemical inputs c) same amount of chemical inputs as conventional farmers d) Don’t know (correct answer b)

2. What is the optimal pH level for rice (open ended question, answers between 5.5 and 7 will be coded as correct)

3. As organic farmer, is it permitted to burn plant residues? a) yes, b) no, c) Don’t know (correct answer is b)

4. If previous question was answered correctly, why is land burning not considered an acceptable practice in organic farming? (open ended questions, coded as correct if respondent mentions at least one of the following aspects: Air pollution, kills micro-organisms, reduced nutrient content)

5. Can you use animal manure directly on an organic farming? a) yes, b) no, c) I don’t know (correct answer is a)

6. If previous question was answered correctly: How can you check whether the manure is ready for use? Open ended question, coded as correct if the respondent provides one out of the following: test for color, temperature, smell, consistency)



Family 3: Perception
1. Higher market price organic products (binary variable): Respondent thinks that organic products in Indonesia are usually sold for a higher price than non-organic products.

2. Organic inputs sufficient (binary variable): Respondent thinks that it is possible to manage a plot without chemicals and provide the plants with all it needs.

3. Chemicals environment (binary variable): Respondent thinks that that high and frequent use of chemical fertilizer and pesticide has a negative impact on the environment.


4. Organic farming equally or more profitable (binary variable): Respondent thinks that organic farming is more or equally profitable.

Secondary outcomes will allow us to explore the mechanisms of the training impact further, in particular, whether any observed effect is driven by self-produced or purchased organic inputs.


Family 1: Adoption
1. Manure
o Applied self-produced fermented manure (binary variable) =1 if the respondent applied fermented manure which was produced by the farmer herself/himself.
o Applied purchased manure (binary variable) = 1 if the respondent applied purchased fermented manure.

2. Other organic fertilizer
o Applied self-produced organic fertilizer (binary variable) =1 if the respondent applied self-produced organic fertilizer other than manure during the last planting season.
o Applied purchased organic fertilizer (binary variable) =1 if the respondent applied purchased organic fertilizer other than manure during the last planting season.

3. Organic pesticide:
o Applied self-produced organic pesticide (binary variable) =1 if the respondent applied self-produced organic pesticide during the last season.
o Applied purchased organic pesticide (binary variable) =1 if the respondent applied purchased organic pesticide during the last season.

4. Residues:
o Burnt plant residues (binary variable) =1 if the respondent burned all or some part of the plant residues.

Domain 2: Good agricultural practices
• Lime application (binary variable) =1 if the respondent applied lime during the last planting season.
• Leaf Color Chart (binary variable) =1 if the respondent to have monitored rice plants Nitrogen levels during the last planting season.

Domain 3: Chemicals
• Chemical fertilizer application (binary variable) =1 if the respondent applied chemical fertilizer during the last planting season.

• Chemical fertilizer application quantity of N, P and K in tons/ha (continuous variable): We will estimate this variable based on the reported quantity of different fertilizer types applied during the last planting season for rice.

• Chemical pesticide application (binary variable) =1 if the respondent applied chemical pesticide during the last planting season. This variable will be measured at the respondent and plot level.

We applied a three-stage random sampling design to select 1,200 respondents. In the first stage, we randomly selected 60 villages, 30 villages in Yogyakarta and 30 villages in Tasikmalaya. In the second stage, we randomly drew farmer group(s) in selected villages. Farmer groups in Indonesia function both as social group but also as task groups for government programmes and the allocation of subsidies.
In each village, we selected a minimum of one and a maximum of three farmer groups until the total number of registered farmer group members from the selected groups was equal to or larger than 60. We set the minimum to 60 members to ensure that a sufficient number of farmers would attend our information sessions (assuming that only a fraction of the members is interested) from which we drew our study participants. After identifying the farmer groups, the members of the selected farmer groups were invited to an information session on organic farming, which was held in their village. These information sessions served two purposes: (1) to facilitate self-selection based on initial interest in organic farming and the willingness to participate in farmer group events; and (2) to collect contact details on prospective respondents. The information sessions were run by AOI. In the third sampling stage, we randomly drew 20 farmers among the attendees of each information session. These 1,200 farmers, 20 from each of the 60 villages, constituted the respondents of our survey. If there were fewer than 20 attendees at the information session, we asked the farmer group head to nominate additional farmer group members.

The treatment was randomised at the village level and consisted of training on organic farming methods and principles. Farmers from groups in control villages did not receive any training. As baseline data was not available at the time of the randomisation, we used publicly available regional data for the stratification. Specifically, we stratified the sample according to urban and rural status and the reported size of agricultural land area per village. In Tasikmalaya, we used ‘travel distance to the district capital’ as an additional stratification criterion as this region is characterised by less developed infrastructure.

Randomization done in office by a computer,

The treatment was clustered at the village level.

Yes

60 villages.

1200 farmers (minus attrition since 2018).

60 villages, 30 villages control, 30 villages treatment

Power calculations are for ITT effects. For the power calculations are estimated based on a 95% confidence interval and a power of 80% for three primary outcomes of interest. The power calculations are based on the full sample size minus the expected attrition. Based on the previous attrition rates, we expect that we will be able to interview around 1,000 respondents in 2023. Assumptions regarding the R2 are based on the adjusted R2 derived from regressing the respective outcome in 2019 on the baseline outcome as well as on a few control covariates such as age and years of schooling using only the control group sample. For example, the baseline data show a mean of 33 percent of farmers applying organic fertilizer other than manure, with a standard deviation of 0.471 and an intra-cluster correlation of 0.13. Given our average cluster size of 16.7 individuals per cluster (adjusted by attrition) and 60 clusters in total, we are powered to detect a minimum effect of 13 percentage points or 0.289 standard deviations. This decreases further if we account for additional explanatory variables that absorb some of the variance in the data.