Detecting and preventing commission and account opening overcharging by citizen science approach
How to cite
Fujii, Tomoki; Shonchoy, Abu; Khanna, Arpita; Chawla, Vardaan. 2023. "Detecting and preventing commission and account opening overcharging by citizen science approach." AEA RCT Registry. AEARCTR-0009104. https://doi.org/10.1257/rct.9104-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
Mobile financial services (MFS) have rapidly widespread in the developing world. They have reached many
previously unbanked people but also created opportunities for dishonest agents to overcharge vulnerable
population. This is hard to resolve both for MFS providers and regulators because it is too costly to monitor each
agent closely and because restricting agents’ behavior may drive away some unbanked clients or induce some
agents to switch to a different MFS provider. As a potentially cost-effective way to address this difficulty, we
explore a citizen science approach using high-frequency real-time crowdsourced data, gathered by youth
volunteers, to detect agents’ overcharging behaviour. An obvious concern about crowdsourced data is the data
quality because data could be added only by crowdsource app users and even faked. Hence, we plan to utilize
mystery shoppers to verify the reliability and accuracy of the crowdsourced data. We will also test the efficacy
of an incentive treatment, where treated volunteers will be provided with additional monetary incentives to recruit citizen scientists. This enables us to obtain evidence on whether providing incentives for signing up increases the total number of citizen scientists willing to contribute to the citizen science approach and if the citizen science approach can help detect overcharging and serve as a departure point for a future scale-up randomized study.
Intervention
Our proposed treatment is that volunteers in the treatment group will receive additional incentives to recruit citizen scientists to assist in data collection. The feature to recruit citizen scientists will be open for all volunteers but only the ones in the treatment group will receive additional incentives if they get someone to sign up to be a citizen scientist.
Experimental design
The main purpose of data collection is to collect data related to the charges imposed by MFS agents. We will engage youth volunteers from Youth Policy Forum (YPF) in Bangladesh and Network of Active Citizens (NAC) in Uganda to be citizen scientists and they will be collecting data first-hand. We will use a crowdsourcing app called POKET app, which allow us to collect georeferenced data from the app users.
This research consists of two stages of data collection. In Stage 1, we set up the basic infrastructure of data collection In Stage 0 (“pilot”), we test the POKET App in the field for 6-8 weeks. This stage is for us to gain experience before collecting data for the purpose of research. The pilot is to ensure that the app works as expected and calibrate the questionnaire if needed. When the volunteers register themselves as a user of the app, we will collect personal information about them, including their name, gender, phone number, and email address, among others, including personality traits. We will also request the volunteers to take the photograph of the signboard and surroundings in a way that does not violate anyone’s privacy. Once their information is registered, they will spend some time registering shops and MFS agents in the area before moving on to collecting transaction-level data, where transactions include opening an account as well as cash-in and cash-out of mobile money. These transaction level data can be the volunteers own actual transaction with MFS agent, simulated transaction with MFS agent (e.g. volunteers would ask about what is involved in opening the account such as necessary documents, fees (if any), forms to be filled, etc.), and interview with acquaintances who have completed transactions with MFS agents, or exit survey with people who have completed transactions with the agents. The data collected include the fees charged, the information about the MFS agents’ shop such the location of the shop and whether the fee structure was visibly displayed, and the volunteers’ assessments of the agents. In Stage 2 discussed below, we will collect the same set of information.
We refer to the first [second] half of Stage 2 as Stage 2-a [Stage 2-b], which lasts for about four weeks. In Stage 2, volunteers collect data through simulated and actual transactions as well as acquaintance interviews. Volunteers are incentivized to collect data. They will be paid based on the total number of valid data points they produce. Volunteers who were a part of data collection during Stage 1 now become team leaders as more volunteers are recruited for Stage 2. Team leaders become a point of contact for newly recruited volunteers should they face any issues in the field. They also become in-charge of relaying instructions from the research team and other implementation partners to the volunteers in the field. At the beginning of Stage 2-b, we release to all volunteers a link/QR code. The volunteers can use it to let anyone interested to become a new volunteer. When the new volunteers register themselves to be a volunteer, the referrer’s name will be asked. If a treatment volunteer successfully refers a new volunteer with the link/QR code and the new volunteer produces at least five valid data points, the treatment volunteer will be rewarded in addition to the incentives for data collection
In of Stage 2-b, we will collect mystery shopper data. We plan to engage 5 mystery shoppers in each location, who differ in the combination of the following characteristics (male/female, young/old, educated/uneducated, experienced MFS user/non-experienced MFS user). Mystery shoppers also use POKET App to collect the data and the variables collected by the mystery shopper data are essentially the same as those collected by the volunteers. However, mystery shoppers will follow the assigned script as closely as possible and conduct actual or simulated transactions. We will have about 600 mystery-shopping transactions in each location. The mystery shopper data collected in this way will allow us to see whether the observations on overcharging from mystery shopper data matches the data generated by citizen scientists (volunteers). If they match well, citizen scientists can be a useful source of information for scientific advancement. If they do not match well, we will further probe into the discrepancy. The results may be influenced by the fact that mystery shoppers are all first-time customers for the agents. Therefore, the overcharging behavior may be different. The combination of mystery shopper data and crowdsourced data collected by volunteers elucidate the relevance of citizen science approach.
Design details
Cited by
No citation data available for this registration yet.