Evaluating smallholder livelihoods and sustainability in Indonesian cocoa value chains

The study will evaluate the impacts of a well-established sustainability and productivity training program that is being implemented by a leading NGO in the cocoa sector in Indonesia.

The Sustainable Cocoa Production Program (SCPP), implemented by Swisscontact, was initiated in 2012 to enhance the economic, social and environmental aspects of the sustainable cocoa supply chain. SCPP aims to facilitate capacity-building for smallholder cocoa farmers in selected districts across 12 cocoa-producing provinces.

The overall objective of SCPP is to increase cocoa farmers' household income from cocoa by 75% and reduce greenhouse gas emissions from the cocoa sector by 30%. It follows a three-dimensional approach: People, Profit, and Planet, meaning the Program facilitates and implements specific activities in the social, economic and environmental dimensions.

There are five modules under SCPP, namely:
1. Good Agriculture Practices (GAP) module;
2. Good Environment Practices (GEP) module;
3. Good Financial Practices (GFP) module;
4. Good Nutrition Practices (GNP) module; and
5. Certification module.

For cocoa certification, SCPP intends to implement UTZ certification in the region under study. UTZ certified cocoa are produced according to the UTZ standard that is guided by the principles of fairness and transparency. The UTZ standard operates through two sets of guidelines: Code of Conduct (which covers the growing and harvesting process) and Chain of Custody (which covers products from the moment they leave the farm to when they arrive on the shelves).

Explanation of SCPP training modules is as follows:

1. Good Agriculture Practices (GAP): as its name suggests, the module covers all good practices of cocoa farming to achieve better result of cocoa production both in quality and quantity. The module touches on subjects such as pruning, fertilising, cocoa's pests and diseases, harvesting, cocoa beans fermentation, drying, and sortation, and many more.

2. Good Environment Practices (GEP): the module covers ideal farming practices which can minimise the rate of Greenhouse Gas (GHG) emissions in the cocoa sector. Through this module, SCPP implements a holistic approach for sustainable, environmental-friendly agricultural practices.

3. Good Financial Practices (GFP): attended by the household member who manages the household finances, the GFP module concentrates on cash flow planning and record keeping, loans and savings.

4. Good Nutrition Practices (GNP): the SCPP program has been reflecting the women’s nutrient adequacy in cocoa farming households. Through these data, SCPP then implements the GNP training to gauge improvement expected in dietary intake.

5. Certification module: this module explains the benefits of certification for smallholder cocoa farmers, which include direct benefits such as premiums paid and diffused benefits such as conservation of ecosystems, increased farm yield, employment opportunities and improved labor conditions, as well as provides assistance to selected farmers in meeting certification criteria for their cocoa farms.

Moreover for the Certification module, SCPP adopts UTZ Certification control points for the Code of Conduct and Chain of Custody. The Code of Conduct consists of list of farming practices regulated under UTZ. These farming practices cover two big categories: farm maintenance and post-harvest processing.

Meanwhile, control points for the Chain of Custody consists of list of four main themes:
1. Management, which includes topics such as general documented control system, self-assessment, and subcontractors.
2. Traceability, which includes topics such as record keeping, identification of certified input and output, UTZ premium, and annual overview of volumes and credit account.
3. Product separation and identification.
4. UTZ certified claims.

2017-05-15

2019-06-30

Improved farm performance and livelihoods of cocoa farmers in North Aceh who participated in SCPP program, with primary focus on (1) cocoa farm performance (i.e., cocoa yields), (2) cocoa revenue and profit, (3) household income.

Cocoa profit will be constructed by subtracting cocoa farm expenses (labour, inputs, etc), from cocoa revenue.

Household income will be constructed by a couple methods: (1) direct question on overall household income, (2) summation over sub-categories of HH income (farm and non-farm income, and other sources of income).

The treatment villages will be randomly selected into four cohorts for successive treatment over 2017-2019 from a candidate set of villages identified with the implementation partner.

We will take the baseline listing of villages in numerous sub-districts in North Aceh district and randomly select villages to roll out the program in. I.e., if the proposed listing of 60 villages, for example, then the Research Team would randomly select 15 for batch 1, 15 for batch 2, etc, where the 15 batch 1 villages could be anywhere in the district. The Research Team may propose a stratification design to ensure greater geographic balance.

Later, from the selected villages for batch 1, the Research Team and the Implementation Team then recruit farmers to participate in the SCPP program. From these recruited farmers, the Research Team will randomly select 60%-70% as respondents. Same thing will apply for future cohorts.

To separately identify the effects of UTZ certification, the Research Team will compare farmers who “just” qualify for being invited into certification in a given year, and those who “just” miss out. To facilitate this, we will work with the implementing partner to generate a quantitative index on which farmers are judged in their readiness for certification, based on implementing partner’s internal assessment system.

The identification process for first batch of 500 farmers will be ongoing through May, 2017, while identification of the remaining farmers will be ongoing during the respective batches. Batch 1 will receive the program in second quarter 2017 (training of farmers would start by May, 2017). Batches 2-4 will receive the training program subsequently, with batch 4 not receiving the program before 2019.

From these selected 15 villages for batch 1, the Research Team and the Implementation Team will recruit 500 farmers who will later be randomised (60%-70% of total farmers; 20 respondents per village) to determine the sample respondents for the research.

The certification program will use UTZ certification. Farmer groups will be offered the option to join the UTZ program as they are assessed to be ready based on implementing partner’s internal assessment process. A large cocoa buyer is involved as a potential buyer of certified cocoa, the company is not directly involved in the delivery of the program. There may also be a possibility to sell fermented cocoa to a Jakarta-based processor for the domestic market.

Furthermore, for UTZ certification, the comparison would be made between farmer groups who were relatively close to the boundary of whether or not to be invited to join certification, and qualified, to those who were relatively close to the boundary, but not invited. Given that farmer groups just below and just about the boundary are not likely to be drastically different, if we observe large differences between the two then we could attribute those to the certification program.

Proposed Timeline:
1. March, 2017. Full sample (~60 villages) identified. Full sample then are randomised to generate list of villages for each batch.
2. April 25 - May 5, 2017. Training of Trainers for Key Farmer in batch 1 selected villages.
3. May 8-12, 2017. Training of enumerators.
4. May 12, 2017. Implementing partner will provide a full listing of all farmers (names, contact details), their groups, and the project villages, providing GPS coordinates where possible.
5. May 15, 2017. Training of batch 1 farmers begins. Surveys ongoing.
6. 2017-2018. Batches 2 and 3 receive the program.
7. 2019 or later. Batch 4 receives the program.

Randomization done in office by a computer.

Village-level

Yes

68 villages

1,360 cocoa farmers and 68 heads of village

First cohort 2017: 1 treatment cohort (19 villages), 3 control cohorts (49 villages)
Second cohort 2017: 2 treatment cohorts (36 villages), 2 control cohorts (32 villages)
2018: 3 treatment cohorts (51 villages), 1 control cohort (17 villages)
2019: 4 treatment cohorts (68 villages)

Based on following power calculation (using clustersampsi in Stata), we calculate a detectable effect of 122.57 kg in cocoa yield. Implementing partner has told us we can expect an effect of 254 kg (a 75% increase in yield on a baseline average of 339.0 kg of coca). mean 1: 339.00 standard deviation 1: 348.00 significance level: 0.05 power: 0.90 baseline measures adjustment (correlation): 0.00 average cluster size: 20 number of clusters per arm: 32 coefficient of variation (of cluster sizes): 0.00 intra cluster correlation (ICC): 0.14 clustersampsi estimated parameters: Firstly, under individual randomisation: detectable difference: 63.06 If, trying to detect an increasing outcome then: corresponding mean 2: 402.06 If, trying to detect a decreasing outcome then: corresponding mean 2: 275.94 Then, allowing for cluster randomisation: design effect: 3.66 detectable difference: 122.57 If, trying to detect an increasing outcome then: corresponding mean 2: 461.57 If, trying to detect a decreasing outcome then: corresponding mean 2: 216.43 We do not have good baseline data on other outcomes of interest (income, revenues, profits) to make similar calculations in levels. In terms of standard deviations, we can detect a 0.35 standard deviation effect size in the other outcomes. mean 1: 0.00 standard deviation 1: 1.00 significance level: 0.05 power: 0.90 baseline measures adjustment (correlation): 0.00 average cluster size: 20 number of clusters per arm: 32 coefficient of variation (of cluster sizes): 0.00 intra cluster correlation (ICC): 0.14 clustersampsi estimated parameters: Firstly, under individual randomisation: detectable difference: 0.18 If, trying to detect an increasing outcome then: corresponding mean 2: 0.18 If, trying to detect a decreasing outcome then: corresponding mean 2: -0.18 Then, allowing for cluster randomisation: design effect: 3.66 detectable difference: 0.35 If, trying to detect an increasing outcome then: corresponding mean 2: 0.35 If, trying to detect a decreasing outcome then: corresponding mean 2: -0.35