Access to electricity and digital inclusion: evidence from mobile call detail records

The study investigates whether and how access to electricity affects mobile connectivity and digital inclusion in Sub-Saharan Africa, focusing on rural areas and women. Despite widespread network coverage in SSA, uptake and usage remain low, particularly where electricity access is limited. The authors note potential simultaneity between electricity and mobile connectivity deployment and aim to mitigate endogeneity by leveraging variations in grid availability and site-level power sources. The purpose is to quantify the effects of electrification on access (mobile subscription, smartphone ownership) and usage (calls/SMS, direction of communications) and to explore heterogeneity by gender and by characteristics of originating (urban) municipalities. The study is important for informing policies aimed at closing rural and gender digital gaps by considering energy access as a key determinant alongside price, income, and digital literacy.

Two strands are reviewed: (1) Determinants of mobile connectivity typically emphasize long-run factors (prices, income, urbanization, education, regulation, competition) largely in advanced economies (Bohlin et al., 2010; Lee et al., 2011; Lin and Wu, 2013; Gruber, 2001). (2) Electricity and development: Armey and Hosman (2016) show at country level that electricity access raises internet penetration in low-income countries; Mothobi and Grzybowski (2017) link brighter night lights and household electricity access to higher mobile phone ownership and different mobile-money usage. Rural electrification studies (Kirubi et al., 2009; Lenz et al., 2017) find gains in ICT use but lack granularity on incoming/outgoing communications and inclusivity aspects. This paper contributes micro-level evidence from Senegal linking site-level electricity availability with mobile access/usage, including directionality of communications and gender.

Data sources and construction: (1) Mobile CDRs and site power data from Orange-Sonatel (Jan–Dec 2013) covering ~9 million users and ~2000 antenna sites, including site power source (grid, solar, generator). A 'home site' is assigned per user based on night-time activity (9 p.m.–2 a.m., holidays removed); user-level indicators (smartphone ownership via TAC, outgoing/incoming calls and SMS, days of activity, network size) are aggregated to the home-site level (mean per user). Site electrification is a dummy equal to 1 if powered by the national grid; solar/generator implies no grid access. (2) 2011 household survey (ANSD): 17,891 households (168,203 individuals), with variables on mobile subscription/expenditure, household assets, electricity (grid or solar = 1), water, dwelling characteristics, earnings, and literacy. Distinguishes informal neighbor connections vs official subscriptions. (3) 2013 census sample (10% per municipality; >1 million individuals): municipality-level covariates—household access to electricity, water, fixed telephony/internet; ownership of TV, fridge, fans; individual characteristics (gender, literacy, socio-professional categories). Aggregated over 551 municipalities. Descriptive analysis: Maps link energy transmission, towers, and population density; overlay of electricity access with smartphone penetration. Econometric strategy: (a) OLS at site level: MOB_i = α + β ELEC_i + γ X_i + ε_i, where outcomes include smartphone penetration and per-user call/SMS metrics (incoming/outgoing). Controls proxy income (socio-professional composition), literacy, and network deployment costs (infrastructure like water). (b) Propensity score matching (PSM): address potential simultaneity/endogeneity by restricting to rural municipalities (low density) and single-site municipalities to reduce supply-side confounding; estimate electrification propensity via logit using the same demand/supply covariates and match electrified sites to nearest non-electrified sites based on propensity scores. Treatment sample: 91 electrified rural sites; matched to 33 unique controls; balance tests fail to reject equality. (c) Household-level PSM: Rural—657 treated households (electrified, no water) matched to three non-electrified each (972 controls) to equalize unobserved network costs; Urban—167 Dakar households with neighbor-based electricity matched to 80 non-electrified urban households outside Dakar. (d) Heterogeneity in incoming communications by origin municipality: For 36 matched rural site pairs, compute per-user surplus in incoming communications from each urban site; estimate Δy_ij = α + β X_j + μ_i + ε_ij via OLS with pair fixed effects, where X_j captures urban literacy and employment structure. Inference: Robust or Abadie–Imbens standard errors for PSM estimates; model fit and balance assessed (e.g., Table S-3).

- Access and uptake: - Site-level OLS: Electrified sites have higher smartphone penetration by 3.20 percentage points (SE 0.37, p<0.01); outgoing and incoming usage measures are also higher after controls. - Rural site-level PSM: Smartphone ownership rises by 3.09 pp (SE 0.65, p<0.01). No significant increase in outgoing calls (duration or count). Outgoing SMS increase by 8.72 per user per month (SE 2.03, p<0.01). Days of activity increase by 0.78 per month (SE 0.24, p<0.01). Network size (unique call contacts) increases by 2.84 (SE 0.84, p<0.01). - Household-level PSM (rural): Number of mobile subscribers per household increases by 0.72 (SE 0.09, p<0.01)—approximately one additional subscriber. Monthly household mobile expenditures increase by 2,398.9 FCFA (SE 798.7, p<0.01). Urban household differences are not statistically significant. - Gender (rural, PSM): Share of female subscribers is 4.23 pp higher in electrified households (SE 1.92, p<0.05). - Usage directionality: - Rural site-level PSM: Incoming communications increase substantially—number of incoming calls +4.94 (SE 1.40, p<0.01); monthly incoming call volume +7.79 minutes (SE 1.83, p<0.01); incoming SMS +5.75 (SE 1.22, p<0.01). Outgoing calls are not significantly higher; outgoing SMS are higher, consistent with cost-minimizing behavior. - Origins: Additional incoming SMS largely originate from Dakar (+3.85 per user, SE 0.83, p<0.01). Incoming calls from Dakar show small, non-significant average surplus in the matched sample; more calls are directed to Dakar overall. - Heterogeneity by origin municipality characteristics (OLS with pair fixed effects): Surplus incoming call duration and volume are positively correlated with higher shares of formal and informal wage workers in the originating urban municipalities (e.g., duration per month coefficient 174.86, p<0.01 for formal wage workers), while surplus incoming SMS are higher from more literate urban areas (coefficient 5.52, SE 2.20). On average, per urban origin site, rural users with electricity receive about 0.5 additional minutes of calls per month (significant at 1%). - Descriptive rural gaps: Electrified rural sites show roughly 5 pp higher smartphone ownership and substantially higher incoming communication than non-electrified rural sites; these gaps shrink but remain significant after matching and controls.

The findings directly address the hypothesis that electricity access enhances digital inclusion. Electrification increases access (more smartphone owners and more subscribers per household) and stimulates usage primarily via incoming communications, indicating enhanced reachability and engagement within social and economic networks. The lack of significant increases in outgoing calls coupled with higher outgoing SMS suggests users leverage the cheapest channels (including call-back SMS) when electricity is available, possibly reflecting budget constraints. Increased incoming communications from urban, higher-living-standard areas imply potential transfers of information and resources from urban to rural users, consistent with development channels facilitated by connectivity. The stronger effect for women indicates electrification can reduce the gender digital gap by enabling female ownership and participation. Together, results highlight electricity as a critical, previously underemphasized determinant of mobile connectivity beyond prices, income, and literacy, and point to complementarities between energy and telecom infrastructure for inclusive digital development.

Access to electricity is a significant driver of digital inclusion in Senegal’s rural areas. Electrification raises smartphone ownership and the number of mobile subscribers per household, with especially pronounced gains for women. Usage increases manifest mainly through greater incoming calls and SMS—particularly from urban, better-off municipalities—while outgoing calls do not significantly rise and users favor low-cost SMS. These patterns suggest electricity access enhances rural users’ integration into broader communication networks and may facilitate financial and knowledge flows from urban to rural areas. Policy implications include recognizing electricity access as a key lever for closing rural and gender digital gaps and pursuing staged rural electrification via decentralized solutions (mini-grids, solar home systems) where grid rollout lags. Future research should rigorously evaluate the impact of decentralized electrification models on electricity demand, digital adoption, and welfare outcomes, and explore mechanisms linking increased incoming communications to concrete socio-economic benefits.

- Potential endogeneity remains despite matching: electricity and telecom infrastructure may still co-locate with unobserved advantages; OLS estimates can suffer from simultaneity bias. - CDR-based measures lack rich user demographics and direct digital literacy indicators; basic literacy proxies may understate true digital skills. - Site coverage areas do not perfectly align with municipal boundaries; analysis mitigates this by using home-site assignment and focusing on single-site rural municipalities, but residual misalignment may persist. - Results rely on one operator’s (Orange-Sonatel) 2013 data in Senegal; external validity to other operators, countries, or time periods may be limited. - Urban analysis shows strong simultaneity bias; causal effects in urban settings are less convincingly identified. - Some models (heterogeneity regressions) have low R-squared, indicating substantial unexplained variance. - Proprietary data constraints limit transparency and reproducibility of full microdata analyses.