We propose to test the effects of monetary incentives for frontline service providers on the take up of a new technology to promote financial inclusion. The context is one of a large state bank which hires local branchless banking agents to introduce a new saving product in a rural and largely unbanked area of Indonesia. These agents are randomized into receiving a high vs. low piece rate for recruiting new customers.
The goal of our experiment is to shed light not only on the supply-side effect of these incentives, but also on their potential demand-side effects, where:
• Supply effect = effect of higher incentives on the effort exerted by the delivery agent.
• Demand effect = effect of higher incentives on community’s perceptions of the quality of the product and the intentions of the agent/organization who promotes the product.
While most of the literature has focused on the supply-side effect of incentives, our conjecture is that, in a context where there is incomplete information about a new technology or product, incentives can also affect take-up through a change in the potential customer’s perceptions. Higher incentives can, for instance adversely affect potential clients’ beliefs about the true intentions of the agent, i.e., they may think that the agent is promoting the product to earn money rather than because the product is beneficial to customers (thereby reducing their perceived quality of the product). On the other hand, higher financial incentives may convey that the agent has a high opportunity cost and is therefore a good worker, or the organization is “efficient” in that it pays its employees more money, and increase the perceived quality of the product or the willingness to adopt it. If financial incentives convey negative signals about the product to customers, then they may backfire even if agent’s effort increases (the supply and demand effect of incentives in opposite directions). If incentives instead convey positive signals about the product, the demand effect of incentives may instead reinforce the supply effect.
This study takes place in the context of the introduction of a new technology to promote financial inclusion in rural Indonesia. The Government of Indonesia has recently adopted a new financial inclusion initiative, and issued a law designed to promote branchless banking as means to increase access to financial services among its largely unbanked population. The model uses one local branchless banking agent per village to promote the take-up and the use of savings and current accounts. Cash-in’s and cash-out’s from customers’ accounts are done through the agents. Five large banks have been approved by the Government to provide branchless banking and we collaborate with one of them.
The experiment takes place in 400 villages in five kabupatens of East Java, in which the population is almost exclusively unbanked and in which our partner bank is expanding. Each village is served by a newly-hired local agent. The compensation scheme offered to agents included a fee for each new customer who signs up for an account, and a fee for each transaction made.
The experiment randomizes the 400 villages (one agent per village) into high and low incentive, stratifying the sample by whether the village has good internet access, distance from the village to the main bank branch, and the presence of other banks in the village:
1. “Low-incentive” treatment: the agent receives 2,000 IDR (~ 0.14 USD) for each new customer who signs up for an account.
2. “High-incentive” treatment: the agent receives four times as much (i.e., 10,000 IDR~ 0.69 USD) for each new customer who signs up for an account.
We define an individual as having “signed-up” if she opens an account and keeps a minimum balance of 20,000 IRP (~1.6USD) in her account for at least two weeks. The latter condition has been added to avoid collusion between the customer and the agent.
To estimate the effect of the treatment on take-up, we survey a random sample of 4,800 households (12 per village.) For each of these households, we will have information on whether they take-up the product at endline, and--if they have--we will ask the amount saved in the account, the number of transactions made, etc. This data will be corroborated by administrative information from the bank.
The comparison of the take-up under the high vs low incentive treatments will allow us to estimate the reduced form effect of monetary incentives on the take up of the new technology: a combination of the supply and demand side effects. To disentangle between these two, we cross randomize an equal share of the high and the low incentive treatment into two treatment arms:
• In the “Private information” treatment, the sample of households we survey are given information about the new product but are not informed about the compensation of the agent.
• In the “Public information” treatment, the sample of households we survey are informed about the product and are also informed about the compensation of the agent (i.e., they are told the fee the agents make for each new customer).
The experimental design is as follows:
Table 1: Experimental Design
T1: High Incentives + Private information (70 Villages)
T2: High Incentives + Public information (130 Villages)
T3: Low Incentives + Private information (70 Villages)
T4: Low Incentives + Public information (130 Villages)
The rationale for using the private vs information treatment is that we believe demand effects of incentives will be minimized in the private information treatment. The context we analyze is indeed one where very people infrequently talk about their earnings to each other, and potential clients in the high or low incentive treatment are likely to have the same perception of agent earning when the information is private, at least in the short term (we’ll test for this).
Our experimental design will compare take-up of the new technology among our sample households in the 4 treatment groups:
• Conditional on being in the private information treatment, the difference between the high vs. the low incentive treatment provides the supply side effect of incentives (increased effort) (T1-T3 in Table 1).
• The difference between the high and the low incentives in the public information treatment will instead provide us with the aggregate supply effect + the demand effect (T4-T2).
• The difference-in-difference between high vs low incentives in public vs public treatment ((T4-T2)-(T1-T3)) will isolate the demand effect.
To reinforce the evidence supporting the hypothesized mechanisms, these results will be complemented by: (1) the effects of the treatments on agent perceived own amount of effort, and (2) household survey questions on the amount of effort provided by the agent, and perceptions about: (a) the quality of the product, (b) their level of trust in the product, the agent and the bank, (c) the perceived identity of the agent.
In addition to the village level treatments described above, the project also tests whether financial literacy can be used as a tool to boost demand for financial products. Within each village, we offer financial literacy and business training to random sample of 5 women entrepreneurs. This variation is orthogonal to the village-level variation in incentives and is part of a separate evaluation.