Returns to Price Transparency: Evidence from Branchless Banking Agents in Indonesia

The first part of the project aims at documenting the prevalence of informal fees and price discrimination in the market, as well as customer preferences for price transparency vis-à-vis other characteristics of the quality of services provided by the agent. To do this, we will conduct an on-line baseline survey with the agents’ customers.
Following the baseline survey, we will conduct an intervention in which the bank enforces price transparency by directly providing a random group of clients (stratified by agent) with the official prices for the main transaction services offered by branchless banking agents.
Approximately one month after the distribution of messages about the bank’s official prices, we will conduct an on-line endline survey to clients to learn about their perceptions and whether agents change their sales and price-setting strategy in response to the diffusion of information about the prices.

2023-07-10

2024-12-31

In our experiment, we will study the effects of distributing information about product prices on (i) take-up and usage of products, (ii) client perceptions, and (iii) agent responses.

1. Take-up and usage of products: The bank’s administrative data at the agent level allows us to measure the take-up of the products offered by the banks and different measures of usage of these products. Additionally, we will be able to measure the effects of the treatment on the balance held by the agent in her account, a proxy for the liquidity she has to provide clients with cash when needed.
2. Client perceptions: Through our surveys with clients, we will also capture clients’ perceptions about the agent (trust), and whether they have made transactions with other agents.
3. Agent responses: We will use surveys with clients to measure agents’ actions following the transparency treatment. These include: whether they publicly display the official price list, the actual prices they charge for the services and the dispersion of prices between clients (a measure of price discrimination), and other (complementary) marketing strategies they may be using.

This transparency experiment is designed to estimate the effects of price transparency on customer behaviour, as well as the market returns to price transparency. For this, we will work with a random sample of agents working with our bank partner. Randomly selected clients of these agents will receive a WhatsApp message from the bank summarizing the official prices for the most common types of branchless banking transactions. The clients of the remaining agents will receive a placebo message that makes no mention of the bank’s official prices. Making use of the bank’s administrative data, we will be able to document the causal effects of price transparency on relevant outcome variables, such as take-up and usage of the products offered by agents. We will also be able to observe outcomes at the agent level, such as the balance in her account. We will complement these administrative data with information obtained from surveys with clients in which we will collect detailed information on the agent’s behaviour (transparency, sales strategies, etc.) and clients’ trust on the agent. This experiment will establish the extent to which branchless banking clients can use the information on actual prices for transactions to hold their agents accountable and, as a consequence, are more (or less) likely to use branchless banking products.

Not available

Randomization will be done through STATA.

Branchless banking agent.

Yes

500 branchless banking agents.

10,000 clients (20 randomly selected clients per each of 500 agents), with an expected response rate of 20%.

Control group: 250 agents (X 20 clients per agent)
Transparency treatment group: 250 agents (X 20 clients per agent)

To ensure the soundness of our experimental design, we have conducted a series of power calculations to identify the Minimum Detectable Effect sizes for the experiment. The calculations are made under the assumptions that: 1) there is one treatment group, 2) there is control group, 3) there is the same number of clients in each group, 4) size is set at 5% and power is set 80%. The sample size of the experiment will depend on the number of clients responding to the survey. Although the response rate is expected to be around 20%, even in a scenario with a 17.5% response rate a small effect of 0.09 standard deviations would be detected.