Motivating Banking Agents to Increase Transactions at Scale - Experimental Evidence from India
Last registered on April 02, 2020

Pre-Trial

Trial Information
General Information
Title
Motivating Banking Agents to Increase Transactions at Scale - Experimental Evidence from India
RCT ID
AEARCTR-0005625
Initial registration date
April 01, 2020
Last updated
April 02, 2020 12:09 PM EDT
Location(s)

This section is unavailable to the public. Use the button below to request access to this information.

Request Information
Primary Investigator
Affiliation
IDinsight
Other Primary Investigator(s)
PI Affiliation
IDinsight
PI Affiliation
IDinsight
Additional Trial Information
Status
On going
Start date
2019-10-01
End date
2020-07-30
Secondary IDs
Abstract
While India has made substantial progress towards financial inclusion, much remains to be done to ensure everyone utilizes, and benefits from, increased access to the formal financial system. In particular, despite increases in bank account ownership, account usage remains low. This problem is especially pronounced in rural and semi-urban areas, where few bank branches operate. India has responded by promoting "Cash-in, Cash-out" banking agents.

This trial tests whether a composite intervention focusing on banking agents increases account usage at scale. The agents in this study belong to a single banking agent network company, and they work across Andhra Pradesh and Telangana. The intervention consists primarily of providing savings calendars to the agents to provide to their customers. These agents are also trained to explain the calendar to customers, to encourage customer savings, and to track distribution of the calendars. Thirdly, agents receive SMS messages that remind them about the content of the training. Finally, managers who supervise these agents receive similar training and SMS reminders. We randomized a sample of 794 agents across treatment and control, with 402 agents assigned to the treatment group and 392 to the control. In the aggregate, we distributed a total of 64,000 calendars to the treatment agents, who were asked to distribute at least half the calendars to women customers. We will evaluate the impact of this intervention on deposits, net deposits (deposits minus withdrawals), and transactions at the level of the banking agents, at both 7 weeks (when India's lockdown to combat COVID-19 began) and three months after training.
External Link(s)
Registration Citation
Citation
Butcher, Luke, Karan Nagpal and Andrés Parrado. 2020. "Motivating Banking Agents to Increase Transactions at Scale - Experimental Evidence from India." AEA RCT Registry. April 02. https://doi.org/10.1257/rct.5625-1.0.
Sponsors & Partners

There are documents in this trial unavailable to the public. Use the button below to request access to this information.

Request Information
Experimental Details
Interventions
Intervention(s)
Intervention Start Date
2020-01-27
Intervention End Date
2020-04-01
Primary Outcomes
Primary Outcomes (end points)
cumulative deposits, weekly deposits
Primary Outcomes (explanation)
"Cumulative deposits" indicates all deposits at an agent's service point, from the end of training until a certain point in time after training - either 7 weeks (when national lockdown in India began in response to COVID-19) or 13 weeks (3 months). Deposits are reported at the level of the agent in administrative data.

"Weekly deposits" indicates all deposits at an agent's service point in any given week after training through 13 weeks (3 months).

For all variables, to correct for outliers (which were observed in baseline data), logarithmic transformations will be conducted in the main specifications. Specifications without logarithmic transformations will be included as a robustness check.
Secondary Outcomes
Secondary Outcomes (end points)
cumulative net deposits, weekly net deposits, cumulative number of transactions, weekly number of transactions
Secondary Outcomes (explanation)
"Cumulative net deposits" indicates all deposits less all withdrawals at an agent's service point, from the end of training until a certain point in time after training - either 7 weeks (when national lockdown in India began in response to COVID-19) or 13 weeks (3 months). Deposits and withdrawals are reported at the level of the agent in administrative data.

"Weekly net deposits" indicates all deposits less all withdrawals at an agent's service point in any given week after training through 13 weeks (3 months).

"Cumulative number of transactions" indicates the total number of transactions at an agent's service point, from the end of training until a certain point in time after training - either 7 weeks (when national lockdown in India began in response to COVID-19) or 13 weeks (3 months). Total number of transactions are reported at the level of the agent in administrative data.

"Weekly number of transactions" indicates the total number of transactions at an agent's service point in any given week after training through 13 weeks (3 months).

For all variables, to correct for outliers (which were observed in baseline data), logarithmic transformations will be conducted in the main specifications. Specifications without logarithmic transformations will be included as a robustness check.
Experimental Design
Experimental Design
The sample of 794 agents in this study, spread across Andhra Pradesh and Telangana, were randomized into two groups: treatment (402) and control (392). Treatment agents received a composite intervention consisting of the following: savings calendars for customers, training on savings mobilization, and SMS messages. The managers of agents received similar training and similar SMS messages.

More specifically, in order to distribute the calendars, we conducted a total of 10 training sessions for the banking agents. Each session was attended by anywhere between 14 and 60 agents. These sessions were accompanied by sessions of the implementing partner on operating banking software (The control agents did not receive this additional training). The savings calendar session focused on three components:
1. The team explained every component of the savings calendars to agents and their managers and highlighted the importance of the calendars as an avenue to build better customer relationships.
2. The team provided advice to agents and their managers on addressing the aforementioned behavioral barriers to saving that customers often face. Agents were asked to communicate to customers the importance of saving for achieving their personal goals, the security benefits of saving with the agent, and the fact that customers should deposit even small amounts into their accounts. These three recommendations were also printed on a savings encouragement sheet that was distributed to the agents.
3. The team also explained the process for monitoring the distribution of the calendars. We provided a distribution sheet to the banking agents to document which customers received calendars. Agents had to send a copy of this sheet twice a week to their managers, who would collate them for the research team.

Following the training, each agent received their set of calendars, the savings encouragement sheet, and the calendar distribution sheet. Agents at and above the median for average number of unique customers per month (based on five pre-intervention months) received 240 calendars, while those below the median received 80. 77% of agents completed the in-person training session. Sub-K separately trained and delivered materials to agents who did not complete the training session. Moreover, in the time between randomization and training,
6 agents assigned to treatment stopped working as agents, lowering the number to be trained from 402 to 396. As of 1 April, monitoring data from 82% of these 396 agents reached the research team. Agents reported distributing 39,180 of 64,000 calendars (61%).
Experimental Design Details
Not available
Randomization Method
We used the statistical software package Stata to conduct our randomization using the following procedure:
1. Use baseline administrative data provided by the implementing partner: Prior to the randomization, we acquired administrative data capturing transactions and basic agent demographics.
2. Stratified random assignment: To increase statistical power, we stratified by bank, district, and level of customer activity (high or low). For the last variable, we calculated for each agent the average number of unique customers per month from June to October 2019; agents with an average below the median were categorized as having a low number of customers. Agents at and above the median were categorized as having a high number of customers. We conducted the randomization after all the administrative data for the baseline had been properly
cleaned.
Randomization Unit
The unit of analysis and the unit of randomization are the agents. This allows us to estimate the causal impact of the intervention on agent-level transactions.
Was the treatment clustered?
No
Experiment Characteristics
Sample size: planned number of clusters
794 agents
Sample size: planned number of observations
794 agents
Sample size (or number of clusters) by treatment arms
402 treatment, 392 control
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
The MDES for average monthly deposits is 6%. The unit is INR, and the standard deviation is 390,000.
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
IRB Approval Date
IRB Approval Number
Analysis Plan
Analysis Plan Documents
Pre-Analysis Plan

MD5: 276522856f22abf1f22f65b756804803

SHA1: 879ea0c06bd70de41f5c99254817c1159709a7cb

Uploaded At: April 01, 2020