Mobile phones can help individuals access information, networks, and resources, allowing them to benefit both socially and economically. Yet in many lower-income countries, women lag men in phone ownership and usage for a variety of economic and normative reasons.
To boost women’s phone ownership and usage, the Government of Chhattisgarh implemented the Sanchaar Kranti Yojana (SKY) program in 2018, using the 2011 census population data as an eligibility criterion. Gram panchayats (GPs) in which the largest village fell below the population threshold of 1,000 were ineligible to receive the program benefits. We use this population cutoff to shortlist 279 treatment gram panchayats and 408 control gram panchayats in 13 districts across the state. We collect three datasets for each gram panchayat and they are as follows.
Individual Surveys In each GP, we aimed to conduct surveys with 15 female (18-45 years of age) and 15 male (18-50 years of age) respondents. Within the 687 GPs, we were able to conduct the full 30 surveys in 682 GPs (99.3 percent) and completed some surveys in a further 2 GPs (0.3 percent). This dataset has all individual-level responses.
Household Surveys In the male survey, we collect information about the rest of the household. This information is not available for households where only a female respondent was surveyed.
Community Leader Surveys We conducted key informant interviews (KIIs) with select community leaders from the largest village in each of the sampled gram panchayats. During scoping activities, we targeted multiple community leaders including the Sarpanch (a locally elected leader), the Sachiv or vice-president (a bureaucrat and another elected leader), and the Gram Rojgar Sahayak (GRS), the main local official tasked with implementing India’s public works program, NREGA. The four leaders we ultimately surveyed had the best ability to accurately answer questions outlined in our community leader survey and the most time for our survey questions. These are the ASHA or Mitanins, Anganwadi Workers (AWW), Ward Members who are part of the GP Council, and the village criers, known as Kotwal. Often when they were unavailable, we surveyed the “representative”, someone who is deputized to carry out similar functions. In some instances when the ASHA worker was busy, we conducted these surveys with more Anganwadi workers. The Anganwadi workers were administered the exact same survey as the ASHA worker. Similarly, village criers and ward members responded to the same survey.
Speedtest Surveys Enumerators conducted speed test using
Speedtest.Net when they were in the GPs. We tried to ensure coverage across the two main network providers, Jio and Airtel, using different SIM cards. These were either conducted on their own or with the Community Leader Surveys and we have more than 1 speed test per GP.
Data Collection Data collection was carried out by IDinsight’s field team. Throughout data collection, the Inclusion Economics India Centre (IEIC) research team conducted ongoing data quality checks, including high-frequency checks, spot checks, in-person back checks, phone back checks, and audio audits. These all fed into a live dashboard created by IDinsight to track productivity and data quality throughout data collection.
Sampling procedure SKY eligibility was based on a population threshold using the 2011 Indian Census. Mobile coverage was extended to covered communities, and smartphones were distributed to one adult female per household in GPs with a population of at least 1,000 in their largest village. GPs just under this population threshold were ineligible. To maximize power, our data collection strategy focused on GPs closest to the discontinuity. We determined our sample as follows:
- First, we shortlisted the 13 districts that had the highest rates of SKY implementation, excluding the capital district of Raipur, where we had run a randomized controlled trial in many SKY-eligible GPs.
- Then, using 2011 Indian census data and local randomization approaches described in Cattaneo et al. (2023), we identified a window around the population discontinuity within which one might possibly assume that treatment is as good as randomly assigned. We find that when dropping GPs that have just one village, the local randomization approach admits GPs where the population of the largest village is 99 more and less than 1000. This yields a sample of 687 GPs.
- The realized sample is smaller than 687 since some GPs were not surveyable (e.g. due to safety issues due to left-wing extremism or challenges in obtaining permission from local leaders).
- Within these selected GPs, we used voter rolls to randomly sample 15 women aged 18-45 from the individuals associated with polling booths within the GP and attempted to survey the woman and a randomly selected male household member.
Results We investigate two approaches to closing digital gender gaps. The first is a statewide program, which distributed millions of smartphones, along with free data, to women across the state of Chhattisgarh in central India. In the second study, we layered digital literacy training on top of the smartphone distribution program. Despite initially reversing the gender gap in smartphone ownership, the smartphone distribution program had no long-term impact on digital gender gaps. In contrast, low-cost digital literacy training had lasting impacts, reducing digital gender gaps and increasing women’s smartphone use. Digital literacy training also improved women’s connection with others and mental health, highlighting important areas that phones can improve women’s well-being in settings where their mobility and networks are limited.
