Quality Upgrading and Competition - Theory and Evidence from the Coffee Sector in Uganda

The aim of this project is to identify pass-through of experimentally induced variation in quality premiums along the supply chain, from downstream exporters to intermediaries to upstream producers. This project consists of two randomized control trials. For both experiments, we will partner with a large coffee exporter in Uganda (hereafter referred to as “the company”). In the first experiment, the company will offer randomly selected traders a “bonus” for bringing in premium quality coffee. In order to examine the role of market structure, we generate exogenous variation in competition for high-quality products by randomizing the saturation of the quality premium offers to traders across geographic markets. In the second experiment, we will conduct direct buying from farmers. We will offer randomly selected farmers an additional “bonus” for premium quality coffee.

In the first experiment, the company will offer randomly selected traders a “bonus” for bringing in premium quality coffee that meet the pre-determined requirements on moisture content, foreign matter and total defects. Eligible traders with qualified coffee will receive an extra bonus of up to 500 UGX/kg. This will generate exogenous variation in quality premiums offered to traders. We will then measure how much of the quality premium gets passed up the value chain to farmers and, in turn, how variation in quality premiums affect traders and farmers' incentives to invest in quality. In order to examine the role of market structure, we generate exogenous variation in competition for high-quality products by randomizing the saturation of the quality premium offers to traders across geographic markets.

In the second experiment, we will conduct direct buying from farmers. We will offer randomly selected farmers an additional “bonus” for premium quality coffee (using the same definition as in Experiment 1). Farmers in this experiment will be divided into three equal sized groups, each offered a bonus sizes ranging from 0 to 500 UGX/kg. This will therefore span the full range of possible pass-through from 0-100% and allow for non-linear estimation of supply. To ensure that the randomization is followed exactly, we will have our research staff, who are embedded at the company, be responsible for making the offer, collecting the coffee, and testing quality at farmgate. The company will reconduct these tests upon arrival at their factory gate.

2022-02-01

2023-06-30

1. Pass-through of experimentally induced variation in quality premiums along the supply chain, from downstream exporters to intermediaries to upstream producers
2. Quality provision at each point in the value chain

1. Pass-through: we will calculate pass-through rate of the experimentally-induced increase in the quality premium along the supply chain by comparing the high-quality prices among different players in the market, especially the price received by traders and the price trader paying to their upstream trader suppliers and farmer suppliers. We will collect detailed price information by quality level through high-frequency sampling surveys throughout the season to collect real physical samples of coffee. Specifically, we ask traders and farmers to keep real samples of their main coffee transactions throughout the season. For each sample, we ask traders and farmers to record the supplier/buyer of the transaction, price paid/received, and quantities purchased/sold. We will send the physical samples to a lab to measure quality along the following dimensions: moisture level, foreign matters, total defects and outturn. This will allow us to have an accurate and objective measure of prices paid/received by quality level. To complement the high-frequency quality sampling surveysm, we will conduct trader and farmer baseline, midline and endline surveys to collect additional information on how the intervention may have affected other aspects of the farmers’ coffee production and the traders’ businesses.

2. Quality provision: As described above, we will conduct high-frequency quality sampling visits to the field throughout the season to collect real physical samples of coffee. Specifically, we ask traders and farmers to keep real samples of their main coffee transactions throughout the season. We will send the physical samples to a lab to measure quality along the following dimensions: moisture level, foreign matters, total defects and outturn. This will allow us to have an accurate and objective measure of prices paid/received and quantities purchased/sold by quality level.

1. Quality investment behaviors at each point in the value chain (both intensive and extensive margins)
2. Shifting in sourcing behavior by traders (selection effects and search)

1. To collect the quality investment behaviors of both traders and farmers, we will include detailed questions in the survey to ask what traders and farmers do in each stage of coffee processing, from gathering, drying, all the way to packaging. We will also collect information on assets owned by traders and farmers. During our high-frequency quality sampling visits to the field, we will further ask traders and farmers about the processing activities they’ve undertaken since our last visit. We will later compile these answers into a single index to represent the level of quality investment.
2. To measure shifts in sourcing behavior by traders, we will collect multiple samples of farmers (see “Experimental Design” below for more details). The first is a sample of farmers from whom the trader sample buys at baseline. The second is a sample of farmers from whom the trader sample newly started trading with in a given season. We will survey the “newly added suppliers” for a subsample of traders at the end of each season. For each trader, we will measure the impact of treatment on the following outcomes:
a. Number of new farmers and traders from whom bought
b. (Pre-determined) demographic characteristics of farmers from whom bought (e.g. size of land holdings, assets/machinery ownership, number of coffee trees (older than a year old), age of farmer, etc.)

This project is a two-pronged randomized control trial. In the first experiment, we will generate exogenous variation in quality premiums faced by intermediaries. We will then measure how much of the quality premium gets passed up the value-chain to farmers. In order to examine the role of market structure, we generate exogenous variation in competition for high-quality products by randomizing the saturation of the quality premium offers to traders across geographic markets. In the second experiment, we will directly manipulate the quality premium faced by farmers, in order to measure how farmer investment in quality upgrading responds to different price incentives. This will allow us to document how limited pass-through of quality premia by domestic supply chains affects ultimate quality provision among producers.

Experiment 1: Pass-through of quality premia along supply chains

The company will offer randomly selected traders a “bonus” for bringing in premium quality coffee that meet the pre-determined requirements on moisture content, foreign matter and defect. Eligible traders with qualified coffee will receive an extra bonus of 500 UGX/kg. This will generate exogenous variation in quality premiums offered to traders. In order to examine the role of market structure, we generate exogenous variation in competition for high-quality products by randomizing the saturation of the quality premium offers to traders across geographic market (in our case, parishes).

We will then measure how much of the quality premium gets passed up the value chain to farmers and, in turn, how variation in quality premiums affect traders and farmers' incentives to invest in quality (though the latter will be more precisely identified in Experiment 2). To do that, we begin by mapping out the baseline supply chain network to identify the suppliers of a given trader who sells directly to the company (referred to as the “downstream trader”. We will then follow up with the suppliers to identify the suppliers’ suppliers, until we reach the producers. Once we have the baseline network, we will select a single main supply chain for a given downstream trader, which potentially include other upstream traders and a farmer. This will form our trader sample and farmer sample (for Experiment 1). For this sample of traders and farmers, we will collect detailed information on quality, prices and sales through surveys and high-frequency field visit. In particular, we will ask traders and farmers to report their purchase and sales prices of coffee to different types of sellers and buyers by quality category. Through our high-frequency visits, we will collect transaction-level data on quality, prices and quantities at seller-buyer pair level. We will also ask what quality improvement activities they have conducted. Last but not least, to capture changes in the supply chain network, we will ask traders and farmers new trading partners formed at the end of each season, and we will randomly select a sample of the newly added suppliers and bring them to the next season’s data collection activities.

Experiment 2: Farmer quality investment in response to quality premia

Experiment 1 is designed to measure the degree of pass-through of quality incentives up the supply chain to producers. However, to fully document the role that market structure plays in ultimate quality provision, we must document not only what level of quality farmers provide to the market at the current level of pass-through (which may be minimal), but also what level of quality would farmers provide to the market counterfactually if they faced greater pass-through and therefore greater quality incentives.

Experiment 2 is designed to provide these estimates. In this experiment, we will conduct direct buying from farmers. We will offer randomly selected farmers an additional “bonus” for premium quality coffee (using the same definition as in Experiment 1). Farmers in this experiment will be divided into three equal sized groups, each offered a bonus sizes ranging from 0 to 500 UGX/kg. This will therefore span the full range of possible pass-through from 0-100% and allow for non-linear estimation of supply. For this experiment, we will select a random set of farmers from adjacent areas that are similar in terms of demographics and coffee production, but which are in distinct markets to avoid the two experiments contaminating each other.

Protocols and timeline:

We will alert treated traders and farmers about their offer after conducting the baseline survey, prior to the start of the coffee season. We will also coordinate regular calls to each trader and farmer to remind him or her about the offer during the growing season, during the harvest and post-harvest periods, when most farmers and traders make their sales.

We will run this experiment for three seasons, with a goal of making the offer fairly “long run” and known well in advance, especially in the last two seasons. That said, many of the major drivers of coffee quality (selective harvesting, proper winnowing and drying, etc.) are determined close to the time of sale and are therefore likely to be more immediately responsive to changing price incentives.

We will begin by mapping out the supply chain network for around 400 traders who sell directly to the company (the downstream traders). We conduct the network survey in two steps:

Round 1: Call all the downstream traders to administer a short survey to map out their main areas of operation, i.e., main parishes of purchasing and selling coffee.

At the end of the round 1 mapping, we will assign traders to parishes and divide the parishes into experiment 1 and experiment 2, stratify by the number of downstream traders in a parish. For traders who are assigned to parishes in experiment 1, we will conduct a second round phone call to trace out their supplier network.

Round 2 : call the downstream traders assigned to parishes in experiment 1 to collect their supplier information, and then call their suppliers to ask about the suppliers’ suppliers, till we reach the farmers.

For parishes in experiment 1, we will assign 1/3 to pure control and 2/3 to half-treated. Within the half-treated parishes, half of the downstream traders operating in those parishes, randomly selected, will be receiving the bonus.

For the parishes assigned to experiment 2, we will sample 150 farmers from those areas who share similar demographic characteristics as the farmer suppliers to the downstream traders in experiment 1 (using information collected from Round 2 network mapping). We will randomly assign farmers to three different levels of bonus offer, with 50 farmers in each.

Randomization for Experiment 1 will then occur in two-stages: first at the market level, and then second, at the individual trader level. This randomized saturation design is intended to produce random variation in demand for high-quality products, such that we can examine how the pass-through of quality premiums is affected by competition along the supply chain.

For the purpose of this project, we will take “the parish” as a plausible definition of “a market” when thinking about competition among traders and their interaction with farmers in the supply chain. Our pilot data reveal that 76% of traders operate in a single parish for the main purchase of coffee. We believe randomizing at the parish level strikes the right balance between minimizing spillovers while maintaining power.

Markets (parishes) will be randomized into two treatment saturation levels: pure control (0% of traders treated) and half-treated markets (50% treated). Then, within the latter, half of the traders will be randomized into treatment.

For Experiment 2, the farmer experiment, randomization will take place at the individual level.

Yes

We expect to have about 200 parishes (clusters) for the experiments.

For experiment 1, we expect to conduct full surveys with 250 traders and 150 farmers at baseline, which include 150 downstream traders (randomly selected from an estimated number of 300 downstream traders in the parishes assigned to experiment 1), 100 of their upstream trader suppliers and 150 of their farmer suppliers (identified through the network mapping exercise). This includes the baseline survey, the high-frequency quality sampling visits during each season and the endline survey at the end of the third season. Last but not least, at the end of every season, we will bring in 50 newly added farmer suppliers and 25 newly added trader suppliers to the data collection activities in the following seasons. For experiment 2, we will administer the baseline survey, high-frequency quality sampling visits and the endline survey on all 150 farmers recruited from the respective parishes assigned to experiment 2.

For experiment 1, 100 (out of an estimated 300) traders will be selected to receive the 500 UGX/kg bonus for high-quality coffee.

For the intervention at the farmer side, we assign 50 farmers to zero bonus, 50 with a bonus of 250 UGX, and 50 with a bonus of 500 UGX.