An Experimental Analysis of the Economic Impact of Women’s Smartphone Ownership in Malawi

Previous research by members of this team found that provision of smartphones to women in low-income Tanzanian households increased annual consumption by 20% over the control group without smartphones (Roessler et al. 2021). With a benefit-cost ratio of more than 2:1, this makes smartphone provision a highly cost-effective anti-poverty intervention. The strongest economic gains from smartphone uptake accrued to those households in which the female participant retained control of the handset —what we refer to as “household property rights.” At the same time, they also reported that their households used the smartphone as well. This points to the importance of strengthening women’s property rights over smartphones and cooperative intra-household mobile technology use. In this study, focusing on 1,501 low-income married women non-phone owners in Blantyre District in Malawi, we aim to systematically test the economic impact of women’s smartphone ownership and enhanced property rights over the smartphones.

Our study has two smartphone treatment arms. One—which we consider the technical efficacy arm or individuals smartphone treatment—provides women non-phone owners with SIM cards and entry-level smartphones; registers them for mobile money and WhatsApp; and delivers training on how to use the smartphone, mobile money and WhatsApp. Participants are informed that they are the owners of the smartphones; all receive a certificate verifying their individual ownership. Overall, we hypothesize that this treatment will strengthen participants’ digital inclusion, technical efficacy to enact their preferences, and, ultimately, improve their livelihoods. As we discuss below, we employ several behavioral measures to ascertain intermediate outputs of increased digital inclusion. We will compare outcomes for the technical efficacy arm group to both a pure control group and a cash placebo control group which will receive an unconditional cash grant the equivalent value of the smartphones (roughly 70 USD) upon program enrollment.

A second treatment group—which we consider the property rights arm or couples smartphone treatment—provides the exact same technical efficacy treatment (SIM cards and entry-level smartphones to the women participants along with individual certificates of ownership; registers them for mobile money; and delivers instruction on how to use the smartphone, mobile money and WhatsApp) but it also delivers a group training with participants and their spouses, who are required to attend for participants to receive the smartphone. (See Figure 1.) The couples’ training is intended to develop beliefs around women’s use of the phone, property rights over the smartphone, and men’s public recognition of those rights in front of community members and GENET trainers. Above and beyond the gains in digital inclusion and technical efficacy, we expect the property rights arm to shift husbands’ beliefs and increase women’s control and use of the smartphone, thus translating into even greater smartphone use. We also expect the property rights arm will strengthen women’s influence over how others use the phone—ensuring it is employed for productive purposes that benefit the household at large. Together, we expect the property rights training and greater smartphone use to lead to higher household economic well-being and women’s economic empowerment at endline. We will compare outcomes for the property rights arm group to the technical efficacy arm, the cash placebo and the pure control group.

Overall, our study’s theory of change can be summed up as follows: the receipt of smartphones and smartphone training will catalyze technical efficacy in terms of increased smartphone use across a range of capabilities (mobile communications, mobile money use, mobile internet, and social media) leading to improved economic livelihoods and, ultimately, downstream effects on women’s empowerment. Moreover, we expect stronger women’s property rights and cooperative use on smartphones to further enhance the technological capabilities they derive from their devices leading to even stronger effects on economic livelihoods and women’s empowerment.

Our study will enable us to estimate the impact of the following:
A. i.) providing women non-phone owners in a low-income country with smartphones + smartphone training compared to ii.) an unconditional cash grant of the same value as the smartphone and iii.) a pure control; and
B. i.) smartphones + couples training to strengthen women’s property rights over the smartphones and shift men’s beliefs about the appropriateness of women owning smartphones to the ii.) smartphone + smartphone training as well as the iii.) cash grant and iv.) pure control.

CONTROL (N=300): No treatment.
T1: CASH PLACEBO (N=400): Distribution of unconditional transfer of the retail cash value of the smartphone ($70) at the start of program plus encouragement of women's right to decide how best to use cash transfer and cooperation with spouses over its allocation.
T2: TECHNICAL EFFICACY TREATMENT (n=400): Provision of entry-level smartphones with certificate of ownership, SIM cards, mobile money accounts, WhatsApp, and training to married women who do not own phones; encouragement of women's right to own handsets and cooperative use with spouses.
T3: PROPERTY RIGHTS TREATMENT (n=400): Technical efficacy treatment plus couples’ training designed around smartphone tech—in which husbands publicly affirm women's property rights over smartphones in front of other community members, and observe other men doing so as well; cooperative smartphone use is encouraged.

2021-06-19

2021-07-31

H1A: Receipt of a no-cost smartphone will increase individual economic well-being compared to cash placebo and pure control.

H1B: Receipt of a no-cost smartphone will increase household economic well-being compared to cash placebo and pure control.

H2A: Receipt of a no-cost smartphone + couples training will increase women’s retention and control over the handset compared to the individual smartphone treatment.

H2B: Receipt of a no-cost smartphone + couples training will increase couples’ cooperation and trust compared to the individual smartphone treatment.

H3C: Receipt of a no-cost smartphone + couples training use will change participants’ and their husbands perceived norms about women’s phone use and ownership in their households and also among broader community compared to the individual smartphone treatment, cash placebo and control.

H3A: Receipt of a no-cost smartphone + couples training will increase individual economic well-being compared to the individual smartphone treatment, the cash placebo and pure control.

H3B: Receipt of a no-cost smartphone + couples training will increase household economic well-being compared to the individual smartphone treatment, the cash placebo and pure control.

H4A: Receipt of a no-cost smartphone will increase women’s empowerment compared to the cash placebo and pure control.

H4B: Receipt of a no-cost smartphone + couples training will increase women’s empowerment compared to the individual smartphone treatment, the cash placebo and pure control.

Measurement is guided by our theory of change that the receipt of smartphones and smartphone training will catalyze technical efficacy in terms of increased smartphone use, leading to improved economic livelihoods and, ultimately, downstream effects on women’s empowerment. And that we expect stronger women’s property rights and cooperative use on smartphones to further enhance women’s retention and control of the smartphones and thus the technological capabilities they derive from their devices leading to even stronger effects on economic livelihoods and women’s empowerment.

Thus our measures focus on: smartphone ownership and retention; smartphone use; individual and household economic livelihoods; and women’s empowerment.

• Mobile Phone retention: Smartphone ownership will be measured at a midline and endline, roughly 6 and 12 months after phone distribution, respectively, using survey and behavioral measures. The survey measure asks about whether the participant owns a phone and what type. The behavioral measure entails the enumerators’ requesting that subjects show the enumerators any phone they have with them during the survey—and recording the type and model of the phone.

Individual economic livelihood outcomes include:
• Weekly and monthly income
• Searching for and finding working
• Access to information needed for work or productive activities
• Occupational choice: farming vs market trading
• Financial inclusion (e.g., access to credit)
• Savings
• Debt

Household economic livelihoods include:
• Household Consumption: Derived from midline and endline survey items, this index measure was adapted from Suri and Jack (2017) and covers aggregate spending across 15 different common consumption baskets, such as food, fuel, transportation, water, and electricity, as well as community functions and investment. Survey items listed fully below.
• Husband Income: Derived from midline and endline surveys, this captures husbands’ weekly and monthly income, searching for and finding work; and occupational choice. We will ask participants about their husbands at midline and endline and husbands directly at endline.
• Household savings
• Household debt

Women’s empowerment

• Social Connectedness: Derived from items tracking subjects’ ability to connect to and rely on close others, this index’s components are listed below.
• Individual Efficacy: This measure uses survey items related to subjects’ self-reported ability to achieve personal goals, overcome challenges, and change problematic situations to construct an index as described below.
• Women’s Empowerment: Derived from survey items, this index will measure household influence and intimate partner violence. Index construction described below.
• Shift in Norms about Women’s Phone Use: Index built from endline survey items measuring husband’s expectations about how women in the community use phones and beliefs about others’ expectations regarding how women use phones.
• Shift in Norms about Women’s Property Rights: Index constructed from endline survey items to measure husband’s expectations that other members of the community: a) respect their spouses’ phone ownership rights and b) expect others to do the same.

Couples’ Trust and Cooperation

• Sharing smartphone with husband: Survey items based on degree to which husbands also use the smartphone.
• Knowledge sharing: Survey items based on whether they report husband in answering the question: “if you do not know how to do something on your phone, how do you figure it out?”
• Employed at endline independently with spouse and husband, this is derived from Almås et al. (2018) and entails providing individuals with a lump sum that they can transfer all to their spouse (e.g., $3 total) or keep for themselves minus a penalty they have to pay ($2 total). Among couples with high levels of trust and cooperation, we expect women to choose the higher offer to send to their spouses.

As intermediate outcomes, we expect the smartphone treatment to increase smartphone uptake and use across a range of indicators.

• Mobile Phone Use: This index uses midline and endline survey items, including use of phone for calling, messaging, and other forms of communication, to construct a measure of subjects’ utilization of the phone. Separately, we will check weekly consumption on phone credit, data and other services.
• Mobile Money (MM) Use: Derived from endline survey items, this index includes whether the person has a MM account; frequency of MM use; use MM to save; strength of preference for MM; times sent and received MM in past month. Index construction described below.
• Behavioral Mobile Money Use: Employed at midline and endline, we will offer participants a micro-grant to be paid on the spot—either $1 in cash or $2 as a mobile money transfer. We will record which mode participants chose and, if mobile money, to whose account they are sending the MM transfer (whether their own account, their spouse’s, or someone else’s)
• Internet use: Frequency of internet use and for what purposes.
• WhatsApp use: Frequency of WhatsApp use, including an administrative measure of the last time the participant was seen active on WhatsApp. Response to a short WhatsApp survey.

This experiment is multi-arm parallel-group randomized trials with participants randomly assigned to one of four study groups: (1.) pure control; (2.) cash placebo; (3.) individual smartphone group; and (4.). smartphone + couples training on women’s property rights and cooperative use.

CONTROL (N=300): No treatment.
T1: CASH PLACEBO (N=400): Distribution of unconditional transfer of the retail cash value of the smartphone ($70) at the start of program plus encouragement of women's right to decide how best to use cash transfer and cooperation with spouses over its allocation.
T2: TECHNICAL EFFICACY TREATMENT (n=400): Provision of entry-level smartphones with certificate of ownership, SIM cards, mobile money accounts, WhatsApp, and training to married women who do not own phones; encouragement of women's right to own handsets and cooperative use with spouses.
T3: PROPERTY RIGHTS TREATMENT (n=400): Technical efficacy treatment plus couples’ training designed around smartphone tech—in which husbands publicly affirm women's property rights over smartphones in front of other community members, and observe other men doing so as well; cooperative smartphone use is encouraged.

After the baseline survey, we used computerized randomization (using the randomizr package in R) to block randomly assign participants to the main treatment conditions (control, cash placebo (T2), smartphone (T3), smartphone + couples training (T4)), stratifying on educational attainment, location, and baseline mobile phone ownership within the household level.

Individuals

No

1501

1501

Our sample size is 1501 women with 400 participants targeted in the couples smartphone group, 400 participants in individual smartphone group, 400 in cash placebo and 300 in pure control.

To conduct a power analysis, we use the Declare Design framework (Blair et al. 2019) and focus on comparisons between couples smartphone treatment (T3), individuals smartphone treatment (T2) and cash (T1) (N=1200). To inform our priors about expected effect sizes and variance on of our smartphone treatments on three of our main outcomes (mobile money use, household consumption, and an index of women’s economic empowerment, WEE), we have drawn from Roessler et al. (2021). For all outcomes except WEE, we report the probability of detecting an effect (significant at the 5% level) given these parameters and a sample size of 1,200 comprised of three equally sized randomly assigned groups. As we did not observe a significant effect on WEE in Tanzania, we report what effect size our cooperative use training would have to generate to detect a statistically significant effect. For the purpose of these calculations, we first use the naive assumption of 100% take up and compliance. We then assume 10% attrition in both conditions, rerunning the simulations with a reduced sample size. The results of the power analysis for each outcome, as well as the parameters used in the analyses are provided in Figure 1. Note that for the average treatment effect parameters we assume we have standardized outcome measures with a mean of 0 and standard deviation (SD) of 1, except for the mobile money behavioral measure. The Monte Carlo approach of Declare Design runs a set of simulations based on the provided parameters, generating the probability that an experiment with the provided assumptions about the distribution of the outcome, sample size, and effect size would be sufficient to detect an effect at conventional significance levels (p-value < 0.05). Overall, we are well-powered to detect a significant effect of our smartphone treatment on MM-use (key component of individual economic livelihoods) and HH consumption (a key component of household economic livelivhoods), while the treatments would need to induce a similar substantive impact on WEE to detect a significant effect on that dimension. Whether we are also able to detect significant effects between our two treatment groups—i.e., the property rights treatment (T3) compared to the technical efficacy treatment group (T2)—is an open empirical question. Based on our Tanzania study, there are reasons to believe we might be sufficiently powered to do so. In our Tanzania study, CACE effects (as defined by women in the smartphone group still owning a smartphone) compared to control were 3 to 3.5 larger than ITT on the women’s use of MM index and MM behavioral measure. As the smartphone retainers represented only a third of the smartphone group (as noted only 35% in the smartphone group still had a smartphone at endline), if our property rights treatment is able to double women’s retention of the smartphones (i.e, increase ownership rates to closer to 70% at endline) and this increases ITT effects accordingly, we would have sufficient power to detect a treatment effect of the property rights arm compared to the technical efficacy one. Outcome (Mean / SD) | Effect Size (priors from Tanzania experiment) | Power with naïve attrition assumption | Power with 10% attrition Mobile money use (0 / 1) | 0.289 | 0.98 | 0.98 Mobile money behavioral measure (0.23* / 0.03*) 0.122 | 1 | 1 Household consumption (0 / 1) 0.204 | 0.82 | 0.79 WEE (0 / 1) 0.21 | 0. 86 | 0.79