Engaging Women in Mobile Money Markets

We will implement a randomized controlled trial to measure the economic and social effects of employing women as mobile money agents. On the supply side, we will increase the number of women working as mobile money agents by randomly assigning subsidies to 500 shops for hiring either male or female employees. On the demand side, we will randomize the digitization of loan repayments at 90 BRAC branches, which increases the demand for mobile money by female customers.

Supply-side intervention:

We will recruit businesses from a sample of existing agents who bKash identifies as having the potential to increase their bKash business by adding an employee, based on the volume of previous transactions, longevity as a bKash agent, and local market characteristics. We will collect basic data about the scale, staffing, and operations of up to 1,200 businesses, from which we will recruit 500 that are willing and able to meet the requirements for participating in the hiring experiment. Surveying more businesses than will be included in the randomized controlled trial serves two purposes: first, it allows us to compare the characteristics of businesses and business owners who are willing and able to identify potential female employees to a sample of businesses that are observably similar to bKash. Second, it ensures that we will have a sufficient sample of businesses that satisfy the inclusion criteria to be eligible for the intervention.

Next, we will elicit a list of potential employees from the owners of the focal businesses. Each owner will be given a week to name four potential employees -- two men and two women -- who he is willing to employ and he believes will be good at helping customers with bKash, and who are likely to accept the job. The only restriction is that potential employees may not be members of the business owner's households. We will record which businesses are interested in providing the list of four potential employees and thus interested in participating in the study, and which ones are not. This snowball sampling strategy will create a second subject pool of potential employees who, by design, are not representative of their communities.

Finally, the businesses which provided a list of potential employees will be randomly allocated to one of two treatments or to the control: 150 of the businesses will receive a subsidy to hire a new male employee; 200 will receive a subsidy to hire a new female employee; and the remaining 150 will form a pure control group that will not receive any subsidy or further encouragement to hire. In the treatment arms that require hiring a new employee, the person to be hired will be randomly picked from the list of employees that was provided by the business owner (conditional on the chosen gender). The wage subsidy will be provided for a period of six months; will decline monthly; and will be conditional on employing the assigned individual for at least 10 hours per week. The randomization will be stratified on market type (rural vs. urban, markets where agents tend to do ``cash-in" transactions vs. markets where agents do ``cash-out" transactions) and the number of other mobile money providers in the same market or bazaar.

Business owners and new employees will receive standard training and support from bKash and their "distribution houses", which are responsible for coordinating bKash activities on the ground. Subsidies will be paid to business owners on a monthly basis. The employee will be required to work at least 10 hours/week, but apart from this, the owners and employees will be free to agree to their own contract terms, including wages. Employment will be -- and will be framed as -- a private contract between the business owner and employee. Compliance with the rule of minimum presence at the store will be tested through random checks by the research team.

Demand-side intervention:

To test the effects of payment digitization, BRAC will identify 90 branches to be included in the pilot. Clients associated with 15 branches will be required to reimburse loans using mobile money. Clients in another 45 branches will be given the option to switch to digital payment. Clients associated with the remaining 30 branches will continue to repay their loans in cash. This distribution is dictated by BRAC's operational needs, and cannot be adjusted to improve statistical power.

All clients will receive basic training about the use of bKash mobile money services, and will be given information about the location of nearby bKash agents, including agents inducted as part of the cross-cutting supply side intervention.

Cross randomization:

The sample will be selected on the basis of overlap between BRAC VO meeting locations and bKash agent operations. We will cross randomize the bKash agent intervention with the BRAC digitization intervention, by stratifying the supply side intervention at the BRAC branch level.

2020-04-01

2021-05-31

We will measure the effects of employing women as mobile money agents on the economic outcomes of the small businesses that offer mobile money services and on business owners' and customers' attitudes towards women's labor force participation. We also estimate the effects of working on the incomes, skill accumulation, and autonomy of women compared to men employees; and whether the availability of female mobile money agents increases the use of mobile money by female customers.

The primary economic outcomes for employees are income from all sources (survey data; 1 month recall, to include wages and profits from income-generating-activities that are controlled by the respondent). These are measured at the individual level, for all the 2,000 potential employees in the sample.

The primary outcomes on attitudes towards female labor force participation for business owners and neighboring businesses will be a PCA index of answers to all of the survey questions about attitudes towards female labor force participation. These are measured at the individual level, for the 500 focal business owners and 1000 neighboring business owners.

The primary outcomes on attitudes towards female labor force participation for family members of potential employees (including those potential employees who are actually hired), and market customers will be the same as above. These are measured at the individual level, for the 2,000 potential employees, 2,000 family members, and 4,500 market customers.

The primary outcomes on access to mobile money will be the number and the total value of mobile money transactions performed by female customers. These will be obtained from bKash administrative records, and will be measured at the bKash agent level, for the 450 focal businesses.

The primary outcomes for BRAC customers will be time spent on repayment and a binary for taking a subsequent loan from BRAC. These individual outcomes will be measured for 6,750 BRAC customers.

The primary outcomes for BRAC program officers will be the share of customers retained, the number of hours spent on loan collection, and the portfolio at risk.

Additional outcomes for employees will include all household income (including wages earned by all household members and profits from all of the household's income-generating activities); net transfers (transfers received from all sources, including NGOs, minus transfers sent to others); consumption (using all items in the survey module on consumption); and a PCA index of all food security items in the survey module; plus a PCA index of all of the above household outcomes. We will also measure potential' agents skills by testing their ability to follow a simple instruction received by text message; their ability to complete a transfer to the bKash account of a surveyor; their ability to help a surveyor check a bKash balance; their ability to answer questions about bKash (share of questions answered correctly); and the share of simple math questions answered correctly; plus a PCA index of all the skill measures.

Additional outcomes about agents' attitudes toward female labor force participation will include willingness-to-pay wages for a female employee; the difference in WTP for a female and male employee; and the answers to individual survey questions about attitudes.

Additional outcomes for access to mobile money will be the fraction of all transactions performed by female customers; the fraction of all transaction value attributed to female customers; the number of unique female customers; the number of new accounts opened by female customers; and a PCA of the above.

Additional outcomes for BRAC customers will be the share of payments made personally (instead of sending a proxy); a Likert scale satisfaction rating with BRAC; the share of their outstanding loans that are from BRAC; and a PCA index of these outcomes. We will also consider BRAC customers' ability to check their own mobile money balances; their knowledge about their outstanding balance with BRAC; and whether or not they know the names of any bKash agents; and a PCA of these outcomes.

Additional outcomes for BRAC program officers will be about knowledge of specific customers (the share of customers for whom BRAC agents can correctly describe current business activities and whether or not the customer is current on loan repayments) and a Likert scale measure of job satisfaction.

We will use a randomized controlled trial designed to measure the effects of employing women as mobile money agents on the economic outcomes of the small businesses that offer mobile money services and on business owners' and customers' attitudes towards women's labor force participation; on the effects of working on the incomes, skill accumulation, and autonomy of women compared to men; and whether the availability of female mobile money agents increases the use of mobile money by female customers.

On the supply side, we partner with bKash, the largest provider of mobile money in Bangladesh. We interview owners of shops that offer mobile money and elicit a list of two men and two women that each owner would be willing to take on as an employee, in return for a temporary wage subsidy. We then randomly choose whether the shop will receive no subsidy, a subsidy for hiring a randomly-selected man from the list, or a subsidy for hiring a randomly-selected woman from the list. This generates variation at the business level in whether the business adds a new female or new male employee, and variation at the potential employee level at whether individuals receive job offers.

On the demand side, we partner with BRAC, whose microfinance program has nearly five million female borrowers. BRAC is interested in transitioning the repayment of microloans from weekly cash payments to weekly mobile money payments. However, as potential barriers the program staff have identified lack of access to mobile money agents and discomfort among female clients transacting with male agents. We will identify BRAC branches whose catchment areas include borrowing groups near bKash agents participating in the hiring experiment. We will randomly assign some of those branches to switch to mobile payments and others to continue to use cash. The intersection of the supply and demand side interventions allows us to test for complementarities: does access to female agents increase repayment and customer retention among microfinance borrowers who are assigned to use mobile money for loan repayment? Does increased demand for mobile money by female customers affect the profitability of adding a female employee, or the business owner's attitude towards female employment?

The project uses a snowball sampling scheme that begins with the identification of focal businesses in 500 markets. Owners of these businesses will identify the second primary sample, potential employees. Other samples of interest (neighboring businesses, customers, family members, and neighboring households) are defined relative to these subjects. Our study is designed to contribute to the literatures on the economic impacts of women's force participation, and on the ways in which social norms and attitudes may constrain women's employment activities.

We will use survey and administrative data collected from multiple constituents. Administrative data from bKash and BRAC and survey data collected from businesses and households will be used to measure impacts on economic outcomes for business owners and potential employees; attitudes towards labor force participation of business owners, potential employees and their families, and market customers; time use, borrowing and customer satisfaction and retention among BRAC microfinance customers; and time use of BRAC program officers. Additional baseline and outcome variables will be obtained from administrative data about mobile money transactions. For BRAC customers, we will use survey data about perceptions of mobile money and microfinance access, and BRAC administrative data to track the repayment (default) rates of customers, the number of missed payments, and the share of customers who take out subsequent loans. BRAC program officers will complete weekly logs documenting time spent on loan collection and other activities.

randomization done in office by a computer

For supply side: agent level, stratified by BRAC branch, market type (rural vs. urban, markets where agents tend to do “cash-in” transactions vs. markets where agents do “cash-out” transactions) and the number of other mobile money providers in the same market or bazaar

For demand side: randomized at the BRAC branch level, stratified by urban/rural

Yes

500 bKash agents (markets)
90 BRAC branches (sample to include one VO group per Program Officer)

Owners of potential businesses -- interview up to 1200, from which 500 that provide complete lists of potential employees will be selected for the hiring experiment Owners of focal businesses -- 500, from those who provide complete lists in the initial intake exercise Owners of neighboring businesses -- 1000 Customers of focal and neighboring businesses -- 4,500 Potential/actual employees -- 2,000 Family members of potential/actual employees -- 2,000 BRAC customers -- 6,750 BRAC program officers -- 500

Subsidy to hire a woman: 200 bKash agents (800 potential employees)
Subsidy to hire a man: 150 bKash agents (600 potential employees)
No subsidy: 150 bKash agents (600 potential employees)

Mandatory digitization: 15 branches (1,500 customers)
Voluntary digitization: 45 branches (3,000 customers)
Control: 30 branches (2,250 customers)

Women's labor force participation: unit is the (potential) employee. Mean LFP in the control group is 0.26. . power twomeans .26, sd(.44) n1(800) n2(200) power(.8) Performing iteration ... Estimated experimental-group mean for a two-sample means test t test assuming sd1 = sd2 = sd Ho: m2 = m1 versus Ha: m2 != m1; m2 > m1 Study parameters: alpha = 0.0500 power = 0.8000 N = 1,000 N1 = 800 N2 = 200 N2/N1 = 0.2500 m1 = 0.2600 sd = 0.4400 Estimated effect size and experimental-group mean: delta = 0.0975 m2 = 0.3575 Attitudes towards women's labor force participation (PCA index of men are better agents, women's duties at home make them unavailable, women are inexperienced, prefer men, pay men more -- from pilot baseline of 52 business owners) . power twomeans 0, sd(1.4031) n1(150) n2(200) power(.8) Performing iteration ... Estimated experimental-group mean for a two-sample means test t test assuming sd1 = sd2 = sd Ho: m2 = m1 versus Ha: m2 != m1; m2 > m1 Study parameters: alpha = 0.0500 power = 0.8000 N = 350 N1 = 150 N2 = 200 N2/N1 = 1.3333 m1 = 0.0000 sd = 1.4031 Estimated effect size and experimental-group mean: delta = 0.4258 m2 = 0.4258 Income of employees (sample of 2,000 potential employees, 500 of whom get hired. Control group mean from baseline pilot survey of 111 potential employees.) . power twomeans 13896.35, sd(15474.02) n1(1500) n2(500) power(.8) Performing iteration ... Estimated experimental-group mean for a two-sample means test t test assuming sd1 = sd2 = sd Ho: m2 = m1 versus Ha: m2 != m1; m2 > m1 Study parameters: alpha = 0.0500 power = 0.8000 N = 2,000 N1 = 1,500 N2 = 500 N2/N1 = 0.3333 m1 = 1.39e+04 sd = 1.55e+04 Estimated effect size and experimental-group mean: delta = 2239.7506 m2 = 1.61e+04