|
Field
Randomization Method
|
Before
The proposal calls for a several-stage randomization process, with proposed percentage allocations determined through simulations:
1. Villages are randomly selected to be included in the program. Using 2011 census data, we include villages from seven statistical districts: Dilijan (Tavush), Ijevan (Tavush), Gugark (Lori), Tumanyan (Lori), Stepanavan (Lori), Spitak (Lori), and Sevan (Gegharkunik). We then omit all villages with fewer than 275 individuals (according to the 2011 census), greater than 1400 individuals, or villages solely comprised of an ethnic minority (e.g. “Malakan”). This leads to a pool of 71 villages.
2. From each included village, 20% of households are randomly selected to be included in the program. Conditions for inclusion are that the household includes at least one permanent member who is younger than 55 (those aged 55 and older may be too unfamiliar with ICT to participate).
3. We randomize villages into three groups:
a. Group A: Apx. 16 placed into “Pure Control” villages that receive no aspects of the program in any form.
b. Group B: Apx. 27-28 placed into “Training” villages in which a random sample of households will receive training and assistance in installing and using the mobile banking application, and
c. Group C: Apx. 27-28 placed into “Incentive and training” villages, in which a random sample of households will receive the same training as “training” villages, but where adoptees/users of the app will receive incentive payment(s).
4. We randomize households in village groups “B” and “C” into two groups:
a. Training households: 20% of the randomly selected households from step 2 are placed into group invited for training and assistance for the mobile app.
b. Spillover households: 80% of the randomly selected households from step 2 are placed into a “spillover” group.
|
After
The proposal calls for a several-stage randomization process, with proposed percentage allocations determined through simulations:
1. Villages are randomly selected to be included in the program. Using 2011 census data, we include villages from seven statistical districts: Dilijan (Tavush), Ijevan (Tavush), Tumanyan (Lori), and Spitak (Lori). We then omit all villages with fewer than 275 individuals (according to the 2011 census), greater than 1400 individuals, or villages solely comprised of an ethnic minority (e.g. “Malakan”). This leads to a pool of 41 villages.
2. From each included village, 20% of households are randomly selected to be included in the program. Conditions for inclusion are that the household includes at least one permanent member who is younger than 55 (those aged 55 and older may be too unfamiliar with ICT to participate).
3. We randomize villages into three groups:
a. Group A: Apx. 30% into “Pure Control” villages that receive no aspects of the program in any form.
b. Group B: Apx. 30% placed into “Training” villages in which a random sample of households will receive training and assistance in installing and using the mobile banking application, and
c. Group C: Apx. 40% placed into “Incentive and training” villages, in which a random sample of households will receive the same training as “training” villages, but where adoptees/users of the app will receive incentive payment(s).
4. We randomize households in village groups “B” and “C” into two groups:
a. Training households: 20% of the randomly selected households from step 2 are placed into group invited for training and assistance for the mobile app.
b. Spillover households: 80% of the randomly selected households from step 2 are placed into a “spillover” group.
|
|
Field
Sample size (or number of clusters) by treatment arms
|
Before
Approximately 16 villages in pure control, 27 to 28 villages in training group, and 27 to 28 villages in training and incentive group.
|
After
Approximately 12 villages in pure control, 12 villages in training group, and 17 villages in training and incentive group.
|