Can private input suppliers be leveraged to stimulate farmer demand for newer maize varieties?

The study partnered with a well-known distributor who supplies maize seed to the agrodealers in the region of study. The study worked with the distributor's network of agrodealers. The agrodealers were informed about the study and invited to make their seed purchases through the distributors as usual. Agrodealers in the treatment group were offered a margin discount if they ordered two specific varieties mentioned by the study team as newer varieties that are commercially available, agronomically suited to the region of study but with a relatively low market share

2023-02-18

2023-04-15

i. An indicator for stocking any of the promoted varieties
ii. The amount of seed in kg of the varieties under promotion stocked by the agro-dealer
iii. Total sales volume of the promoted varieties (kg)
iv. Average selling price of the promoted varieties (KES/kg)

All the main outcome variables will be measured using the agro dealer follow-up surveys in May 2023, November 2023, and May 2024.

i. Information-seeking effort by the agro-dealer
ii. Credit uptake
iii. Total stock and sales volumes of all the other maize varieties (kg)
iv. Agrodealer recommends any of the promoted varieties to the secret shoppers (two variables defined
v. Dealer engagement with the secret shoppers

i. Information seeking effort by the agro-dealer: An index created by adding affirmative responses (yes=1) to the following questions: whether the agrodealer have ever heard of the promoted varieties; whether the agrodealer saw any of the promoted varieties either on a neighbor’s farm, a customer’s farm, or a demo; and whether the agrodealer asked for feedback on any of the promoted varieties from any of their customer during or after the season.
ii. Credit uptake: agrodealer took any kind of credit during the season (1=yes) and the value of the credit (KES)
iii. Total stock and sales volumes of all the other maize varieties (kg) [This might be a final one as well – as it is total sale which might have been affected?]
iv. Agrodealer recommends any of the promoted varieties to the secret shoppers (two variables defined as: i) a dummy=1 if the dealer recommends any of the promoted varieties in any of the first two scenarios ii) dummy=1 if agrodealer recommends any of the promoted variety in scenario 3.

Intermediate outcomes i to iv will be measured using the agrodealer follow-up surveys in May 2023, November 2023, and May 2024.

v. Dealer engagement with the secret shoppers; a dummy=1 if the dealer asks any question before making any recommendations
vi. Agrodealer encouragement to farmers to purchase the promoted varieties: three indicators measured using the farmer level data. i) Farmer awareness of the promoted varieties (a dummy=1 if aware) ii) dummy =1 if a farmer purchase any of the promoted varieties iii) the amount in kg of seed purchased.
Intermediate outcomes v to vi will be measured using the secret shoppers exercise in March 2023, October 2023 and March 2024 seasons.

The study comprises two experimental groups: i) A comparison group and ii) an incentive (margin discount) group.

Not available

Randomization of agrodealers into the two experimatal groups was done using Stata

Markets

Yes

100

300

50 comparison; 50 treatment

The primary outcome used for power calculations is the total amount of seed of the promoted varieties stocked by an agro-dealer. Using data from a previous survey of agro-dealers conducted by CIMMYT, we calculate the average amount of seed of a new variety sold per agro dealer during the 2019 March season. We define a new variety as any variety released within the last seven years from the time of data collection. This variable is used as a proxy for the baseline levels of the outcome variable of the current study. With a mean of 130kg (standard deviation of 211 kg ), an ICC of 0.03, 100 markets with an average of 3 shops and 0.8 power, the minimum detectable effect for our study is 70kg, a 54% increase relative to the control mean. This is considered a modest to large effect according to Cohen’s D. However, we note that there may be power losses because our clusters are not equally sized. We try to minimize this by weighting our observations with the inverse of the cluster size. In addition, we note the wide range of our outcome variable (as well as some of the other outcome variables (sales and total volumes of seed stock). An inverse hyperbolic sin transformation and winsorization of the outcome variable (s) will be considered during the estimations, and this will significantly improve our power