This manuscript was compiled on February 19, 2020 For billions of people across the globe, mobile phones enable relatively cheap and effective communication, as well as access to information and vital services on health, education, society and the economy. Drawing on context-specific evidence on the effects of the digital revolution, this study provides empirical support for the idea that mobile phones are a vehicle for sustainable development at the global scale. It does so by assembling a wealth of publicly available macro-and individual-level data, exploring a wide range of demographic and social development outcomes, and leveraging a combination of methodological approaches. Macro-level analyses covering 200+ countries reveal that mobile-phone access is associated with lower gender inequality, higher contraceptive uptake, and lower maternal and child-mortality. Individual-level analyses of survey data from sub-Saharan Africa, linked with detailed geo-spatial information, further show that women who own a mobile phone are better informed about sexual and reproductive health services and empowered to make independent decisions. Payoffs are larger among the least developed countries and among the most disadvantaged microlevel clusters. Overall, our findings suggest that boosting mobilephone access and coverage and closing digital divides, particularly among women, can be powerful tools to attain empowerment-related sustainable development goals, in an ultimate effort to enhance population health and well-being, and reduce poverty.

Mobile phones | SDGs | Gender equality T he potential for information and communication technolo- gies (ICTs) to empower marginalized communities and promote sustainable development goals (SDGs) has been recognized among scholars and policymakers (1)(2)(3). The digital revolution brought about by the diffusion of mobile phones has allowed several countries with otherwise poor infrastructuresuch as countries in sub-Saharan Africa and South Asia -to leapfrog communication technologies such as phone landlines and fixed internet connections (e.g. broadband) (see Figure 1 and S1 in the SI), with immense social implications. Growing evidence from specific contexts has shown that mobile phones -small, relatively inexpensive, yet incredibly powerful devices -can facilitate effective communication and connectivity, as well as access to information and vital services linked to health, education, and the economy.

This study is, to the best of our knowledge, the first to grow context-specific evidence about the role of mobile phones in affecting sustainable development to a global scale. We show that access to mobile phones is positively associated with multiple indicators linked to global social development, such as lower gender inequalities, enhanced contraceptive use, and lower maternal and child mortality. We do so by assembling a wealth of publicly available macro-and individual-level data * , 23 exploring a wide range of demographic and social develop-24 ment outcomes tied to women's decision-making power, health, 25 and well-being, and leveraging a combination of datasets and 26 methodological approaches, some of which (individual-level)

Although mobile phones have diffused rapidly even in remote parts of the world with otherwise poor infrastructure, digital divides persist. This study provides large-scale evidence that the expansion of mobile phones is associated with lower gender inequalities, higher contraceptive use, and lower maternal and child mortality, with bigger payoffs among the poorest countries. Micro-level analyses further show that the ownership of mobile phones has narrowed the information gap about reproductive and sexual health, and empowered women to make independent decisions. Boosting mobile-phone access and coverage and overcoming digital divides within and among the poorest countries has immense implications for sustainable development. Findings from this study speak to scholars and policymakers interested in the effect of technology on sustainable development goals. positive pregnant women (10), and increased contraceptive use 41 (11,12) and acceptability (13). The increased affordability of 42 mobile phones has the potential to facilitate autonomy and 43 empowerment-related outcomes especially for women, through 44 channels such as enhanced financial independence and better 45 labor-market prospects (14), more decision-making power in and behaviours (26,27). Mobile phones offer enhanced capabilities, including the ability to access these benefits privately, which is essential in contexts where social norms are restrictive and might hamper women's access to information resources and ability to connect directly with healthcare providers (28). This potential is explicitly acknowledged within the SDG Goal 5 (Target 5B), which seeks to harness ICTs such as mobile phones as a pathway towards women's empowerment and wellbeing. ‡ Drawing on a range of social and demographic outcomes linked to gender inequality, reproductive and sexual health, and women's empowerment, in this study we provide both macro-and individual-level evidence that this is indeed one promising pathway.

Our macro-level analyses build on pooled data for 209 countries between 1993 and 2017, and seek to provide broad associational evidence between mobile-phone diffusion and global development indicators to further motivate the micro-level investigations that follow. We combine country-level data from the International Telecommunication Union (ITU), the World Bank (WB), and the UN. Outcome measures include the Gender Inequality Index (GII), the prevalence of modern contraceptive methods, the maternal mortality ratio, and under-five child mortality. The GII is a comprehensive macroindicator of gender inequality comprising three sub-dimensions of human development -reproductive health, empowerment, and economic status -widely used by social scientists to measure gender dynamics and female empowerment (29)(30)(31), as well as by policy-makers to track progress towards the attainment of the SDGs (32). Table S2 in the Supplementary Information provides a complete description of the outcome measures.

Figure 2 provides global correlations between a measure of mobile-phone diffusion -defined as the ratio between the number of mobile-phone subscriptions and the total population -and sustainable development outcomes. The left panel shows, for each country-year, the relationship between mobile-phone diffusion and the four outcomes, including the least-squares lines (in yellow). At the global level, mobilephone diffusion is negatively correlated with gender inequality as measured by the GII (r=-0.51; p-value=0.000), positively correlated with contraceptive prevalence (r=0.31; p=0.000), and negatively correlated with maternal (r=-0.37; p=0.000) and child (r=-0.47; p=0.000) mortality. These associations show non-linearities across levels of economic development measured by GDP-per-capita quintiles (right panel), and are on average higher in absolute values for the lowest quintiles. These GDP-gradients take the form of J -or reversed-Jcurves, with mobile-phone diffusion being more negatively associated with the GII, maternal, and child mortality, and more positively associated with contraceptive prevalence among the least and less developed countries.

Further analyses adjusting for controls -including panel data fixed-effects models, panel data random-effects models, and instrumental variables (IV) models -corroborate evidence from the above unadjusted associations. For simplicity, we here present results from panel data fixed-effects models only (Table 1), while estimates using different techniques, model specifications, and non-interpolated data are reported in Table S7 (Panel 3) point to higher absolute returns to technology for less developed countries documented in Figure 2 (right panel). Mobile phone coefficients in Q1-Q2 countries suggests that mobile-phone technology might serve to complement the role of other development processes such as educational expansion and economic growth, rather than substitute for it.

Coefficient estimates are robust across a variety of estimation techniques and model specifications (Table S7) and also to alternative gender inequality indicators such as the UNDP's Gender Development Index, dis-aggregated components of the GII, and the GII indicator purged of the maternal mortality component -an outcome in itself in our analysis (Table S6).

Despite multiple robustness checks, we acknowledge that our ability to make causal claims in the macro-level analyses is limited. Nonetheless, we believe that these analyses offer a novel global-level overview of the links between mobile phones and indicators of sustainable development. In so doing, the macro-level component of this study paves the way for individual-level analyses focusing on a subset of countries at the low end of the J-curve in Figure 2, towards a more causally-oriented understanding of the relationship between technology and sustainable development.

In what follows, we provide consistent and more specific evidence from augmented individual-level data in a multi-level perspective. We use nationally-representative samples of women aged 15-49 from the Demographic and Health Surveys (DHS) to study the effects of mobile-phone adoption on (social) sustainable development outcomes. Unlike macro-level data on mobile-phone diffusion, information on individual-level adoption of mobile technologies that can be linked to development outcomes is less readily available (4,33). The most recent DHS waves in some countries, however, provide data on whether female respondents own a mobile phone and we can exploit this information. The DHS also contain detailed geographical information about where respondents live, which allows us to augment DHS data with geo-coded data. We are therefore able to directly study if women who own mobile phones are more empowered to make independent decisions in their house-

holds, more knowledgeable about health-seeking behaviours (e.g., where to get tested for HIV), and have improved health outcomes (e.g., contraceptive use, access to antenatal care etc.).

Individual-level analyses are limited to Sub-Saharan Africa (SSA) for two reasons. First, the demographic transition is slowly under way in SSA, with slow fertility decline, and infant and maternal mortality remaining at very high levels (34,35).

Relatedly, SSA is the world region with the highest variation in mobile-phone adoption and diffusion. § By focusing on SSA we therefore provide evidence from an "extreme-case" scenario where policy interventions related to technology-adoption are likely to be particularly effective -as also suggested by the higher marginal returns to technology for less developed countries documented above. Second, by focusing on a subset of SSA countries we seek to ease comparability and maximize the internal consistency of our findings. In this respect, we are concerned about the potential endogeneity of mobile-phone ownership, our key explanatory variable.

Endogeneity can arise as mobile-phone ownership and social development outcomes might be jointly determined by individual characteristics that are not observed. Similarly, we cannot rule out instances of reverse causality whereby more empowered women are more likely to own a mobile phone.

To address these concerns, we select from among the available pool of DHS surveys those containing: 1) information on mobile-phone ownership, and 2) GPS coordinates (latitude and longitude) of the respondent's household. Following these criteria, we select DHS surveys providing individual-level data on women between 2015 and 2017 from seven SSA countries: Angola, Burundi, Ethiopia, Malawi, Tanzania, Uganda, and Zimbabwe. The combined data set includes more than 100,000 individual observations -for more details on the sample refer to Table S8 in tion about nighttime lights, a well-established proxy for local 265 development (37,38). We then include this proxy in our 266 model specifications while keeping the level of granularity of 267 the analysis unaltered.

To summarize, our final augmented DHS dataset includes 269 information from four different datasets that are linked through 270 GPS coordinates. As all data are publicly available (see Table 271 S1), our analyses are easily replicable. ¶ The SI section provides 272 additional descriptions of the data (S8), variables (S9), and 273 methods used in this individual-level analysis.

Figure 3 plots the estimated coefficients from a series of 275 Ordinary Least Squares (OLS) or probit (left panel) and instru-276 mental variables regression models (right panel). We regress 277 four sets of outcomes on mobile-phone ownership, namely: 278 1) women's involvement in intra-household decision-making 279 regarding contraception (1a) or women's involvement in intra-280 household decision-making as measured by a linear index 281 comprising decisions on health, household large purchases, 282 and visits to family and friends (1b); 2) women's actual use 283 of modern contraceptive methods; 3) women's knowledge of 284 health-related outcomes and, specifically, of where to get tested 285 for HIV; and, eventually, 4) information on women's antenatal 286 visits' quality during the last pregnancy, proxied by either 287 whether a woman has been tested for HIV (4a), or whether 288 she has undergone at least one antenatal visit (4b). For the 289 last two outcomes we include only women who had at least 290 one birth in the last year and for whom we know that the area 291 in which they live was covered by cellular signal in the year 292 preceding the birth. This information is obtained by combin-293 ing the interview date and the geographical location of the 294 DHS birth-history data and the Afrobarometer. Alongside the 295 aforementioned proxy for local development (nighttime lights), 296 all models account for standard socio-economic controls -re-297 spondent's education, age, household's size, employment and 298 urban/rural status ‖ -together with indicators of exposure to 299 other media (radio and TV). Models also include country and 300 year fixed effects to account for unobservable country-level 301 heterogeneity.

Results indicate that women who own a mobile phone have 303 higher probability of being involved in decision-making pro-304 cesses about contraception (β=0.01; p<0.01), higher overall 305 decision-making power within the household (β=0.15; p<0.01), 306 higher likelihood of using modern contraceptive methods 307 (β=0.02; p<0.01), and a higher likelihood of knowing where 308 to get tested for HIV with respect to women who do not 309 own a phone (β=0.03; p<0.01) . Estimated coefficients are 310 sizable, as for most outcomes the effect of owning a mobile 311 phone is roughly comparable to -if not bigger than -that 312 of living in an urban area. Results are robust to the use of 313 instrumental variables, thus adding confidence in the interpre-314 tation of the coefficients as causal estimates. Mobile-phone 315 ownership is also associated with a higher chance of being 316 tested for HIV during pregnancy (β=0.02; p<0.01) and to 317 a higher chance of having undergone at least one antenatal 318 visit (β=0.01; p<0.01), yet IV results do not allow us to in-319 terpret these last two associations as causal as the estimated 320 coefficients are not statistically significant.

Lastly, to parallel the macro-level analyses of heterogeneous 322 ¶ DHS datasets can be downloaded directly from http://dhsprogram.com.

‖ As a robustness check we also account for wealth at the household level. Results are robust to the inclusion of this control. See Table S16 globe to close digital divides as a way to foster sustainable development. Although we acknowledge that our ability to make causal claims in the macro-level analyses is limited, we see this global-level overview as a novel and essential building block towards a holistic understanding of the relationship between technology adoption and sustainable development. Further research is needed to unravel and understand the mechanisms that underlie these macro-level relationships, thus providing effective guidance to practitioners.

Our individual-level analyses for Sub-Saharan Africa have shown how women who own a mobile phone are better informed about where to access sexual and reproductive health and are better able to make their own decisions in their households, including about contraception. Through these channels of improved knowledge and enhanced decision-making power, our individual-level analyses have suggested a pathway through which the macro-level results emerge. The macroand individual-level analyses provide consistent and complementary results that support each other to to understand the broader implications of the digital revolution on social development processes.

More than twenty years ago in 1995, the Beijing Declaration and Platform for Action reaffirmed that the human rights of women and girls are an inalienable, indivisible, and integral part of the universal human rights. Since then, gender equality has been acknowledged as both a standalone SDG, and one that is inextricably linked with progress on other SDGs, including key targets associated with health and wellbeing (39). Since the 2000s, considerable improvements have been observed through SDG indicators such as contraceptive use and underfive mortality (40). Our study has highlighted how these improvements have been bolstered, among other factors, by the digital revolution, and in particular by the diffusion of mobile phones. Echoing the words pronounced in 2003 by Nobel Peace Laureate Muhammad Yunus at a conference on poverty and information and communication technologies (ICTs) (41), "the quickest way to get out of poverty right now ones (33,(42)(43)(44). Given that the costs associated with mo-