The Impact of Heterogeneity on Social Learning's Efficacy

How to cite

Alidaee, Hossein. 2022. "The Impact of Heterogeneity on Social Learning's Efficacy." AEA RCT Registry. AEARCTR-0009537. https://doi.org/10.1257/rct.9537-1.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Hossein Alidaee*

* Primary investigator

Registry ID
AEARCTR-0009537
DOI
https://doi.org/10.1257/rct.9537-1.1
First registered
Last updated
Status
In development
Trial dates
2022-08-08 – 2022-09-30
Countries
India (Odissa)
Registry record
http://www.socialscienceregistry.org/trials/9537

Abstract

Information acquisition is central to technology adoption decisions. Two common sources of information about returns include (i) central sources, such as government information campaigns, and (ii) social learning from peers. Central sources often have greater data on returns—yet, we lack empirical evidence that social learning is less persuasive. Understanding social learning's efficacy is particularly important for technologies where returns are highly heterogeneous and information acquisition is a major barrier to adoption. I propose one potential mechanism, which I refer to as context uncertainty. I will test this mechanism via a lab-in-the-field experiment with a sample of smallholder farmers. If valid, this mechanism provides a framework to improve informational interventions from central sources.

Intervention

Participants play a mobile game based vignette experiment. The game has multiple rounds. Within each, the participant must choose how intensively to adopt a new agricultural technology. The information provided to them is recommendations from characters in the game who previously tested the technology. Each round of the game features a distinct environment with respect to the distribution of recommendations and what is known about the characters. The participant's payoff for the experiment is based on their average yield in the game based on their adoption decisions.

Experimental design

The experiment randomizes the order of the information environments, the village names, and the technology names being used. However, all participants experience all members of each set.

Design details

Randomization method
I am doing a crossover design. Each participant is exposed to every round of the game. However, round order is randomized, with a total of 64 possible orderings. To assign orderings, I compute a covariate index for each participant based on listing data. Following the approach outlined in McKenzie (2022), I compute the Mahalanobis distance between all participants, find each participant's matched pairing, and subsequently create a matched quartet from the means of each pair. This generates my first round order decision randomization. There are 6 total order randomizations to create the 2^6 = 64 arms. For the remaining 5 order randomizations, I take the assignments from the previous randomization and create a cell from each matched grouping (i.e. matched quartet). I compute the index mean for this cell and find the closest matching cell. I then alternatively assign the round order accordingly.
Randomization unit
Randomization is done at the individual farmer level.
Clusters
No clustering is planned.
Observations
1600 smallholder farmers.
Arms
Each of the farmers will be allocated uniformly across the 64 possible round orderings.
Primary outcomes
Chosen adoption intensity (0-10) per round of the game.

Cited by

No citation data available for this registration yet.

Keywords