|
Field
Randomization Method
|
Before
We have access to baseline data. Randomization will be done on a computer after bloc matching.
|
After
We have access to baseline data. Randomization will be done on a computer after bloc matching. An algorithm was developed to perform a hierarchical clustering in groups of equal cluster size using nearest neighbor matching. The algorithm also tries to maximize distance between households within each cluster (based on GPS coordinates) to minimize spillover effects. We have used the following cluster variables: household size, age of household head, average productivity of potato, gender of household head, total area of potato cultivation , consumption per capita, distance to input distributor, amount of credit received in the last year and whether the household had access to extension for potato growing. The algorithm can be found here (http://is.gd/c8m8MZ), the script that generates the randomization here (http://is.gd/F8lWJb) and the allocation of treatment here (http://is.gd/twOwFB). The last is also uploaded as supporting materials to this registration.
|