Abstract
Faced with decaying networks, poor revenue collections, and substantial sunk costs and operating losses, over the last two decades, many developing countries have embarked on electricity sector reforms. This analysis examines factors driving the lack of household access to electricity in sub-Saharan Africa, including poor basic infrastructure, inadequate incentives in public service policies, geophysical barriers, and constraints in institutional environment. Based on cross-region panel datasets from Demographic and Health Surveys of 31 African countries between 2003 and 2018, a general-to-specific model selection procedure is applied to parametric regressions, with special attention to border discontinuities between power trading agreements and related border region effects. The chosen specifications are replicated in beta-function generalised linear models and kernel regressions, which specifically account for upper and lower bounds in the dependent variable. The econometric results turn out to be fairly robust to different estimation methods and data panels and suggest that sector restructuring and regional power integration initiatives have contributed to reducing the percentage shares of households without electricity access. However, remoteness from agglomeration economies of major urban centres and lack of substantive improvements in the grid and off-grid networks between neighbouring power trading pools have left many regions lagging behind, particularly in Central Africa. Programmes of poverty alleviation, including electricity services, should be more carefully targeted by strengthening local infrastructure development, access to modern energy, and cross-border integration within and between African regional power pools.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Cost-effectiveness of a public service institution can be measured in terms of a ratio of a weighted set of outputs to the expenditure for the provision of these outputs. Since these weights reflect policy preferences, different levels of cost-effectiveness can partly be explained by trade-offs between policy objectives (Smith and Street 2005). Cost-effective solutions imply that a specific level of reliability and quality is offered on a least-cost basis (World Bank 2011).
Brown et al., (2006: 197) distinguish between conduct and structural regulation: the former is defined as “close scrutiny of actions and interactions” of utility providers and companies, including application of formal codes of conduct, while the latter concerns management or legal separation of business activities within a sector. As for other public utilities, the electricity market tends to be mixed, with competition in some activities (power generation, retail supply) and natural monopoly in others with fundamental economies of scale or scope (wire services, transmission and distribution networks). In either case, priority is often given to new infrastructure construction rather than maintenance of existing grids, often due to political convenience and/or clearer excludability as a public good (Medinilla et al., 2019: 21).
System losses can be either technical, including copper losses and transformer failures, or operational, such as electrical metre tampering and failure. In practice, electrical utility indicators, such as gaps between electricity supplied to grid and electrical consumption billed, do not allow to clearly separate the two types of losses. Household survey data often include illegal connections in electricity access rates, thus pointing to possible problems of inefficiency and quality of supply, lack of trust in public authorities, and affordability, especially in peri-urban areas (Singh 2015; Tallapragada et al., 2009).
The effectiveness of renewable energy and energy efficiency programmes highly depends of good regulatory governance, in terms of pricing and power purchase policies established and put forward by central and local government institutions, as well as coordination with other initiatives aimed at improving individuals’ health status and supporting long-term behavioural responses in favour of maintenance of improved cooking technology (Hanna et al., 2016; Brown et al., 2006: 21). Moreover, incentives for reducing the use of traditional fuel for cooking as a consequence of additional sources of income and improved dwelling conditions, can be offset by other factors, such as expenditures for school fees and increased scope for collecting firewood by a home helper (Pundo and Fraser 2006).
The literature refers to an inverted-U electricity Kuznets curve, whereby the share of transmission and distribution losses tends to increase up to nearly one-fourth of electricity billed to consumers in countries with access rates around 40% on average (Burgess et al., 2020). The criteria of cost-efficiency and affordability of utility expansion, which underpin an eligibility access threshold (Eq. (2)), should yield optimal combinations of central grid—including ‘lock-in effects’ of new grid extensions (Dalla Longa and van der Zwaan 2021: 3)—decentralised medium- and low-voltage distribution networks (minigrid), for rural communities with relatively higher population density and/or in sparsely populated areas, e.g., at less than 50 km distance from the central grid (Lucas et al., 2017: 54), and off-grid (stand-alone) electrification, usually far from the grid.
PSM is unfeasible at this geographical scale. However, it would be useful to reassess the results based on PSM applied to household-level data for villages located near African power pool borders. Similarly, a social welfare function for electricity services as expressed in Eq. (2) cannot strictly be operationalised at a cross-region level, since it would require an analysis at village level with fully consistent longitudinal multi-wave data. In both cases, one should be aware of a number of additional DHS data constraints, pointed out in section A of the Appendix.
If the percentage share of households with access to electricity is chosen as dependent variable (i.e., elc = 1 − nelc) for the same linear regression specifications, parameter estimates are identical with reverse signs. The only exception is the intercept term, as complementary percentage rate summing to one (100%) with the constant estimated in the respective specifications on lack of access (Table 3, 4, 5, 6, and 7). Regression diagnostics are the same, including residual normality and heteroscedasticity tests, F and t tests (except for the t-statistic associated with the intercept term: e.g., for model [1] in Table 3, constant = 0.25 (t-stat 3.02***)), log-likelihood, and adjusted R-square, among others. As implicitly suggested by a reviewer, further insights for policy would be gained by examining at village and district level the extent of utility expansion coverage in terms of numbers of newly connected households, instead of percentage rates of households with (/without) electricity access at a more aggregate regional scale.
To assess the internal validity of a discontinuity design, a useful preliminary step is graphical observation of the rating variable vis-à-vis non-outcome variables (Jacob et al., 2012: 9). For instance, other conditions unaltered, closeness to a neighbouring country’s capital may widen border discontinuities (Pinkovskiy 2017: 171). Moreover, incorrectly specified functional forms can cause bias in the treatment parameter. This supports non-parametric estimation for regression discontinuity analysis: to better account for a boundary problem at threshold point, an analysis on varying window widths helps to check the robustness of locally weighted regressions (Lee and Lemieux 2010: 316–318). In the presence of spatial (/time) trends, the inclusion of interaction terms, as in Eq. (3), makes the parametric approach substantially equivalent to non-parametric estimation around the threshold(s) (Lehner 2021).
In practice, in this analysis, multicollinearity problems prevented from using other DHS round-lagged indicators of living conditions, cooking environment and cooking fuel choices, which turned out to have mostly weaker links with household access to electricity. These indicators included the percentage share of households with a pit latrine or open pit (Table 1: pitlatrine), the mean number of household persons per sleeping room, the percentage of households with a television, the percentage of households cooking outdoors, a dummy accounting for more than half of households cooking with natural gas within a region, and the percentage share of households cooking with kerosene and/or charcoal. Similarly, to avoid multicollinearity problems, some regression models only tested for time discontinuity between power pools, without simultaneous testing for border discontinuity (Table 3: [4]/[4a]), or limiting the latter to either slope dummies of distances or intercept dummies, not both. Relative to models [9b] and [9c/d] (Table 5), this choice was also due to the final specification following the general-to-specific model selection procedure (see notes under Table 3), applied to model [9]. For model [1w] (Table 7), the time effect dummy (tdum) was kept despite being discarded by this procedure.
In the DRC, hydropower accounts for more than 90% of national electricity supply, with operational capacity estimated to barely exceed half of the installed energy capacity, which is largely concentrated in the Grand Inga project (Kongo Central province). The national electricity grid is distinguished in three main networks (Kinshasa in the West, the Lakes region in the East, and Katanga in the South; Gnassou 2019: 3).
In spite of initiatives to link the two energy power sectors, SAPP still consists in a southern and a northern sector, with the former largely dependent on coal thermal generation (Botswana, Lesotho, Swaziland, Namibia, and South Africa), and the latter mainly reliant on hydropower (Angola, DRC, Malawi, Mozambique, Tanzania, Zambia, and Zimbabwe; REN21 2018: 33). Also due to this reason, in this analysis Angola and DRC are classified as CAPP members (member states of the four regional power pools in SSA are listed in Table 1).
For instance, in Angola, the enclave province of Cabinda lies mostly below four hundred metres, with a small part of territory towards the north-eastern border with the DRC reaching 800 m. elevation, thus implying zero value in the dummy topogrhet. As an opposite case, the north-western landlocked province of Cuanza Norte presents a markedly heterogeneous topography, passing from plains in the west to high mountains towards the centre and east.
For each country and DHS sub-period where available, the 1-year lagged score (/one-year lagged binary value for dummies) was chosen for each of the eight variables, including the quality index. For instance, for Angola, with a 2015–16 DHS, its electrification planning score for 2014 was 63 (/100; rise.esmap.org), which converts to 4.18 in the respective index used here (plan) for consistency with the electricity quality index. Relative to tariffsys, for which the same conversion applies, a detailed description of the three underlying criteria for this index variable is provided in ESMAP (2018: 109). The dummy variable tariffpeak can also serve for sensitivity analysis using alternative indicators of energy efficiency, due to partial overlap with tariffsys.
The use of level 2 administrative subdivisions, besides level 1, was made if DHS information was available at a finer geographical scale (e.g., in Rwanda with 30 admin2 districts vs. 5 admin1 provinces; in Namibia, the opposite holds true, with only 4 admin2 vs. 13 admin1 areas).
An additional regional power development plan has been operating since 2005 by member states of the Nile Equatorial Lakes Subsidiary Action Programme (NELSAP), which is one of the main regional development programmes promoted by the Nile Basin Initiative. Feasibility studies for NELSAP currently focus on electricity network interconnections between DRC, Burundi, Rwanda, Uganda, and Kenya (ikp.nilebasin.org).
A beta kernel is given by K(z) = [(1 − z) (1 + z)]/24 if |z| ≤ 1, 0 1 else (z = β′x; notes under Table 3 and Greene 2016a: E-188). Similar to the gamma density, beta kernel functions allow correction of bias errors at data boundaries, as adaptive density estimators with no need for change in smoothing bandwidth (Chen 1999; Mackenzie and Tieu 2004).
In an OLS regression with a bounded dependent variable, as the expected value of this variable approaches either one of the boundaries, its potential variance tends to decline—unless the error distribution becomes more skewed. Extreme predicted values will tend to be biased and possibly lie outside the boundary range. However, relative to regression specifications chosen from the general-to-specific model selection procedure, statistical inference did not turn out to be significantly affected if heteroscedasticity-robust t-statistics were relied on (reported for model [9] in Table 5; analogous estimates for other models not shown—Table 3, 4, and 6). Moreover, in these regression models, the bias in parameter variance estimates turned out to be both downward and upward. This implies increased probability of type I as well as type II errors according to different regressors, with no consistently larger robust standard errors relative to their homoscedastic analogues. One should also notice that the performance of heteroscedasticity-consistent standard errors, even with ad hoc small-sample adjustments, is not found to be accurate in relatively small samples (Verbeek 2012: 102–111; Wooldridge 2002: 55–58).
Restricting the window width to 1000 km around the border line removes the statistical significance of the interaction dummy distcapp, but it retains a border discontinuity pattern with a parallel shift rightward of the regression parameter fit associated to distance for regions east of this border, given dummy parameter vis-à-vis intercept term estimates (Table 7: model [5w], capp vs. constant).
References
Afanador, A., Koper, M., Lulie, M, Tesfaye, G., & Teferra, M. (2016). Off-grid rural electrification in Ethiopia. Ecofys, Cologne. www.mitigationmomentum.org. Accessed May 2019.
Anselin, L. (2005). Exploring Spatial Data with GeoDa: A Workbook. University of Illinois, Urbana-Champaign.
Arora, P., & Chong, A. (2018). Government effectiveness in the provision of public goods: The role of institutional quality. Journal of Applied Economics, 21(1), 175–196.
Belmin, C., Hoffmann, R., Elkasabi, M., & Pichler, P. (2022). LivWell: A sub-national dataset of the living conditions of women and their well-being for 52 countries. Scientific Data, 9(719), 1–14. www.nature.com. Accessed June 2023
Bensch, G., Kluve, J., & Peters, J. (2010). Rural electrification in Rwanda: An impact assessment using matching techniques, Ruhr Economic Papers, No. 231, Ruhr Universität Bochum (RUB). dnb.d-nb.de. Accessed May 2023.
Besley, T., & Burgess, R. (2000). Land reform, poverty reduction, and growth: Evidence from India. Quarterly Journal of Economics, 115(2), 389–430.
Best, R., & Burke, P. (2018). Electricity availability: A precondition for faster economic growth? Energy Economics, 74(3), 321–329.
Bigman, D., Dercon, S., Guillaume, D., & Lambotte, M. (2000). Community targeting and poverty reduction in Burkina Faso. The World Bank Economic Review, 14(1), 167–193.
Bimenyiamana, S., Asemota, G., & Li, L. (2018). The state of the power sector in Rwanda: A progressive sector with ambitious targets. Policy and Practice Review, 6(68), 1–14.
Blimpo, M., & Cosgrove-Davies, M. (2019). Electricity Access in sub-Saharan Africa: Uptake, Reliability and Complementary Factors for Economic Impact, African Development Forum Series Report, World Bank, Washington DC. www.worldbank.org. Accessed Jun 2020.
Blimpo, M., Postepska, A., & Xu, Y. (2020). Why is household electricity access low in sub-Saharan Africa? World Development, 133(C), 1–20.
Bonardi, J.-P., Urbiztondo, S., & Quélin, B. (2009). The political economy of international regulatory convergence in public utilities. International Journal of Management and Network Economics, 1(2), 232–256.
Bondonio, D. (2002). Evaluating decentralized policies: A method to compare the performance of economic development programmes across different regions or states. Evaluation, 8(1), 101–124.
Booth, S., Li, X., Baring-Gould, I., Kollanyi, D., Bharadway, A., & Weston, P. (2018). Productive Use of African Micro-Grids: Technical and Business Considerations. USAID-NREL Partnership, Washington DC / Golden-CO. nrel.gov/docs. Accessed Apr2023.
Bosker, M., & Garretsen, H. (2009). Economic development and the geography of institutions. Journal of Economic Geography, 9(3), 295–328.
Brent, R. (1986). An axiomatic basis for the three-objective social welfare function within a poverty context. Economics Letters, 20(1), 89–94.
Brown, A., Stern, J., Tenenbaum, B., & Gencer, D. (2006). Handbook for Evaluating Infrastructure Regulatory Systems. World Bank, Washington DC. ppp.worldbank.org. Accessed Oct 2019.
Burgert, C., Colston, J., Roy, T., & Zachary, B. (2013). Geographic displacement procedure and georeferenced data release policy for the demographic and health surveys, DHS Spatial Analysis Reports, No. 7, Calverton, Maryland. dhsprogram.com/pubs. Accessed Jun 2023.
Burgess, R., Greenstone, M., Ryan, N., & Sudarshan, A. (2020). The consequences of treating electricity as a right. Journal of Economic Perspectives, 34(1), 145–169.
Burke, M., Heft-Neal, S., & Bendavid, E. (2016). Sources of variation in under-5 mortality across sub-Saharan Africa: A spatial analysis. The Lancet, 4, 936–945. www.thelancet.com. Accessed Nov 2018.
Byaro, M., & Mmbaga, N. F. (2022). What’s new in the drivers of electricity access in sub-Saharan Africa? Scientific African, 18(10), 1–9.
Byers, L., Friedrich, J., Hennig, R., Kressig, A., Li, X., McCormick, C., & Malaguzzi Valeri, L. (2019). A global database of power plants. WRI Technical Note. www.wri.org. Accessed Oct 2019.
Butts, K. (2021). Geographic Difference-in-Discontinuities. Applied Economics Letters, 30(5), 615–619.
Chen, S. X. (1999). Beta Kernel Estimators for Density Functions. Computational Statistics and Data Analysis, 31(2), 131–145.
Chimhowu, A., D. Hulme, L. Munro 2019. The ‘new’ national development planning and global development goals: Processes and partnerships, World Development, 120(C), 76–89.
Dalla Longa, F., & van der Zwaan, B. (2021). Heart of light: An assessment of enhanced electricity access in Africa. Renewable and Sustainable Energy Reviews, 136(2), 1–18.
Demographic and Health Surveys (DHS) 2012. Sampling and Household Listing Manual: Demographic and Health Surveys Methodology, ICF International, Calverton, Maryland.
Dercon, S. (2004). Growth and shocks: Evidence from rural Ethiopia. Journal of Development Economics, 74(2), 309–329.
Dinkelman, T. (2011). The effect of rural electrification on employment: New evidence from South Africa. American Economic Review, 101(7), 3078–3108.
Doornik, J., & Hendry, D. (2018). Empirical Econometric Modelling − PcGive 15/OxMetrics 8 (Vol. I). Timberlake Consultants Ltd.
Douma, J., & Weedon, J. (2019). Analysing continuous proportions in ecology and evolution: A practical introduction to beta and Dirichlet regression. Methods in Ecology and Evolution, 10(9), 1412–1430.
Eberhard, A., Briceño-Garmendia, V., Foster, V., & Shkaratan, M. (2010). Power: Catching up. In V. Foster & C. Briceño- Garmendia (Eds.), Africa’s infrastructure: A time for transformation (pp. 181–202). Agence Française de Développement and World Bank. www.worldbank.org. Accessed Nov 2019.
Elabbas, M., de Vries, L., & Correljé, A. (2013). African power pools and electricity market design: Taking stock of regional integration in energy sectors. Energy Research and Social Science, 105, 1–18.
Energy Sector Management Assistance Program (ESMAP). (2018). Regulatory indicators for sustainable Energy (RISE), Appendices. World Bank. www.esmap.org. Accessed May 2020.
Foster, V., & Caridad Araujo, M. (2004). Does infrastructure reform work for the poor? A Case Study from Guatemala. World Bank Policy Research Working Papers, No. 3185, World Bank. openknowledge.worldbank.org. Accessed Apr 2022.
Footman, K., Benova, L., Goodman, C., Macleod, D., Lynch, C., Penn-Kekana, L., & Campbell, O. (2015). Using multi-country household surveys to understand who provides reproductive and maternal health services in low- and middle-income countries: A critical appraisal of the demographic and health surveys. Tropical Medicine and International Health, 20(5), 589–606.
Gelman, A., & Imbens, G. (2019). Why higher-order polynomials should not be used in regression discontinuity designs? Journal of Business and Economic Statistics, 37(3), 447–456.
Gertler, P., Lee, K., & Mobarak, A. (2017). Electricity reliability and economic development in cities: A microeconomic perspective. EEG (Energy and Economic Growth) State-of-knowledge paper series (No. 3.2). Oxford Policy Management and University of California Energy Institute. escholarship.org/uc. Accessed May 2019.
Gnassou, L. (2019). Addressing renewable energy conundrum in the DR Congo: Focus on Grand Inga Hydropower Dam Project. Energy Strategy Reviews, 26, 1–6.
Green, D., Leong, T., Kern, H., Berger, A., & Larimer, C. (2009). Testing the accuracy of regression discontinuity analysis using experimental benchmarks. Political Analysis, 17(4), 400–417.
Greene, W. (2016a). LIMDEP Version 11. Econometric Software Inc.
Greene, W. (2016b). NLOGIT Version 6. Econometric Software Inc.
Grosse-Puppendahl, S., Bilal, S., & Karaki, K. (2017). EU’s financial instruments for access to energy: Support in remote and poor areas in Africa. Discussion papers, No. 218. European Centre for Development Policy and Management (ECDPM). www.ecdpm.org. Accessed May 2023.
Hanna, R., Duflo, E., & Greenstone, M. (2016). Up in smoke: The influence of household behavior on the long-run impact of improved cooking stoves. American Economic Journal: Economic Policy, 8(1), 80–114.
Hausman, C., & Rapson, D. (2018). Regression discontinuity in time: Considerations for empirical applications. Annual Review of Resource Economics, 10(1), 533–552.
Imbens, G., & Lemieux, T. (2008). Regression discontinuity designs: A guide to practice. Journal of Econometrics, 142(2), 615–635.
International Energy Agency (IEA) and World Bank. (2022). Tracking SDG 7: The Energy Progress Report. World Bank.
IEA. (2010). Energy Poverty: How to Make Energy Access Universal? OECD/IEA.
IEA. (2011). World Energy Outlook 2011. OECD/IEA. www.iea.org. Accessed May 2019.
IEA. (2014). World Energy Outlook: A Focus on Energy Prospects in sub-Saharan Africa. OECD/IEA.
IEA. (2016). World Energy Outlook 2016. OECD/IEA. www.iea.org. Accessed May 2023.
IEA and World Bank. (2017). Progress towards Sustainable Energy: Global Tracking Framework 2017. World Bank.
Jackson, K. (2013). Diversity and the distribution of public goods in sub-Saharan Africa. Journal of African Economies, 22(3), 437–462.
Jacob, R., Zhu, P., Somers, M.-A., & Bloom, H. (2012). A practical guide to regression discontinuity. Manpower Demonstration Research Corporation (MDRC). www.mdrc.org. Accessed May 2019.
Joskow, P. (2014). Incentive regulation in theory and practice: Electricity distributions and transmission networks. In: Rose, N. (ed.), Economic regulation and its reform: What have we learned? (pp. 291–344). University of Chicago Press. www.nber.org. Accessed May 2023.
Kaufmann, D., Kraay, A., & Mastruzzi, M. (2010). The worldwide governance indicators: Methodology and empirical issues. World Bank Policy Research Working Papers (No. 5430). World Bank. www.govindicators.org. Accessed Oct 2019.
Keele, L., & Titiunik, R. (2015). Geographic boundaries as regression discontinuities. Political Analysis, 23(1), 127–155.
Klege, R., Visser, M., Datta, S., & Darling, M. (2022). The power of nudging: Using feedback, competition and responsibility assignment to save electricity in a non-residential stetting. Environmental and Resource Economics, 81(3), 573–589.
Köhlin, G., Sills. E., Pattanayak, S., & Wilfong. C. (2011). Energy, gender and development. What are the linkages? Where is the evidence? Social Development Working Papers (No. 125). World Bank. www.worldbank.org. Accessed Nov 2018.
Kolak, M., & Anselin, L. (2020). A spatial perspective on the econometrics of program evaluation. International Regional Science Review, 43(1–2), 128–153.
Korkovelos, A., Mentis, D., Siyal, S. H., Arderne, C., Rogner, H., Bazilian, M., Howells, M., Beck, H., & De Roo, A. (2018). A geospatial assessment of small-scale hydropower potential in sub-Saharan Africa. Energies, 11(11), 1–21. www.mdpi.com. Accessed Oct 2019.
Lee, D., & Lemieux, T. (2010). Regression discontinuity designs in economics. Journal of Economic Literature, 48(2), 281–355.
Lehner, A. 2021. A note on spatial regression discontinuity designs, Working Paper, Becker Friedman Institute, University of Chicago. (lehner.xyz/publications)
LeSage, J., & Polasek, W. (2008). Incorporating transportation network structure in spatial econometric models of commodity flows. Spatial Economic Analysis, 3(2), 225–245.
Long, J. S. (1997). Regression Models for Categorical and Limited Dependent Variables. Sage Publications.
Lucas, P., Dagnachew, A., & Hof, A. (2017). Towards universal electricity access in sub-Saharan Africa: A quantitative analysis of technology and investment requirements. Policy Report (No. 1952). PBL Netherlands Environmental Assessment Agency. www.pbl.nl. Accessed May 2023.
Maboshe, M., & Kabinga, M. (2018). Facility level access to electricity and the efficiency of maternal and child health service provision in Zambia (Report No. S41404-ZMB-1). International Growth Centre, London School of Economics. www.theigc.org. Accessed Nov 2018.
Mackenzie, M., & Tieu, K. (2004). Asymmetric Kernel Regression. IEEE Transactions on Neural Networks, 15(2), 276–282.
Markham, D. (2019). Cost reflective pricing: Not so simple in practice. AEC Analysis. June 6, Australian Energy Council (AEC). energycouncil.com.au. Accessed Nov 2021.
Marckwardt, A., & Rutstein, S.O. (1996). Accuracy of DHS-II demographic data: Gains and losses in comparison with earlier surveys. DHS Working Papers, No. 19, Macro International, Calverton, Maryland.
Medinilla, A., Byiers, B., & Karaki, K. (2019). African power pools: Regional energy, national power. ECDPM Discussion Papers (No. 244). European Centre for Development Policy Management (ECDPM), Maastricht. www.ecdmp.org. Accessed May 2020.
Michalopoulos, S., & Papaioannou, E. (2012). National institutions and subnational development in Africa. Quarterly Journal of Economics, 129(1), 151–214.
Mveyange, A. (2015). Night lights and regional income inequality in Africa. WIDER working papers (No. 85). United Nations University-WIDER, Helsinki. www.wider.unu.edu. Accessed Nov 2018.
Nazir, C. P. (2018). Coastal power plant: A hybrid solar-hydro renewable energy technology. Clean Energy, 2(2), 102–111.
Nagelkerke, N. (1991). A note on the general definition of the coefficient of determination. Biometrika, 78(3), 691–692.
Njiru, C., & Letema, S. (2018). Energy policy and its implications on standards of living in Kirinyaga, Kenya. Journal of Energy, 9, 1–12.
Odarno, L., Agarwal, A., Devi, A., & Takahashi, H. (2017). Strategies for expanding universal access to electricity services for development. WRI working paper. World Resources Institute (WRI). www.wri.org/publication. Accessed Nov 2018.
Orr, L. (1976). Income transfers as a public good: An application to AFDC. American Economic Review, 66(3), 359–371.
Peterson, G., & Muzzini, E. (2005). Decentralizing basic infrastructure services. In: World Bank, East Asia decentralizes – Making local government work (pp. 209–236). siteresources.worldbank.org. Accessed Nov 2013.
Pinkovskiy, M. (2017). Growth discontinuities at borders. Journal of Economic Growth, 22(2), 145–192.
Pundo, M., & Fraser, G. (2006). Multinomial logit analysis of household cooking fuel choice in rural Kenya: The case of Kisumu District. Agrekon, 45(1), 24–37.
Rana, A., Ngulube, M., & Foster, V. (2022). Global electricity regulatory index 2022. ESMAP Report. World Bank. www.worldbank.org. Accessed May 2023.
Rao, N., & Pachauri, S. (2017). Energy access and living standards: Some observations on recent trends. Environmental Research Letters, 12(2), 1–9.
Raschky, P. (2013). Estimating the effects of West Sumatra public asset insurance program on short-term recovery after the September 2009 earthquake, ERIA discussion papers (No. 35). Economic Research Institute for ASEAN and East Asia (ERIA). www.eria.org. Accessed Apr 2020.
Reiche, K., Tenenbaum, B., & Torres de Mästle, C. (2006). Electrification and regulation: Principles and a model law. Energy and Mining Sector Board Discussion Papers (No. 18). World Bank. ppp.worldbank.org. Accessed Oct 2019.
REN21 (Renewable Energy Policy Network 21st Century). (2018). SADC renewable energy and energy efficiency status report. REN Secretariat. www.ren21.net. Accessed Oct 2019.
Rycroft, M. (2014). The potential of small hydropower plants in Southern Africa. Energize. www.ee.co.za. Accessed Nov 2019.
Sahlberg, A., Khavari, B., Mohamed, I., & Fuso Nerini, F. (2023). Comparison of least-cost pathways towards universal electricity access in Somalia over different timelines. Energies, 16(18), 1–20.
Singh, R., Wang, X., Mendoza, J. C., & Ackom, E. (2015). Electricity (in)accessibility to the urban poor in developing countries. WIREs (Wiley Interdisciplinary Reviews) Energy and Environment, 4(4), 339–353.
Smith, P., & Street, A. (2005). Measuring the efficiency of public services: The limits of analysis, Journal of the Royal Statistical Society. Series A, 168(2), 401–417.
Taghizadeh-Hesary, F., Rasoulinezhad, E., & Phoumin, H. (2022). Ways to achieve universal access to sustainable electricity in Southeast Asia. Economic Change and Restructuring, 55(4), 2031–2050.
Tallapragada, P., Shkaratan, M., Izaguirre, A., Helleranta, J., Rahman, S., & Bergman, S. (2009). Monitoring performance of electric utilities: Indicators and benchmarking in sub-Saharan Africa. Africa Energy Unit, World Bank. www.esmap.org. Accessed May 2015.
Tessama, Z., Davis, M., Vilchis Tella, P., & Lambe, F. (2013). Mainstreaming sustainable energy access in national development planning: The case of Ethiopia. Working paper series, No. 9. Stockholm Environment Institute. www.sei-international.org. Accessed Jan 2015.
United Nations (UN). (2022). The sustainable development goals report 2022. United Nations Publications. www.un.org. Accessed May 2023.
Verbeek, M. (2012). A Guide to Modern Econometrics. Wiley.
Wing, C., & Cook, T. (2013). Strengthening the regression discontinuity design using additional design elements: A within-study comparison. Journal for Policy Analysis and Management, 32(4), 853–877.
Wooldridge, J. (2002). Econometric Analysis of Cross Section and Panel Data. The MIT Press.
World Bank. (2022). Regulatory indicators for sustainable energy: Building resilience. World Bank. www.esmap.org. Accessed Nov 2018.
World Bank. (2020a). Integrated electrification planning accelerating progress towards SDG7: Focus on low access countries in sub-Saharan Africa. Projects & Operations Results Briefs, Nov., Washington DC. worldbank.org/en/results/2020. Accessed Feb 2022.
World Bank. (2020b). Regulatory indicators for sustainable energy: Sustaining the momentum. World Bank. www.esmap.org. Accessed Feb 2022.
World Bank. (2011). One goal, two paths: Achieving universal access to modern energy in east Asia and the Pacific. World Bank. www-wds.worldbank.org. Accessed Feb 2015.
Wuepper, D., & Finger, R. (2022). Regression discontinuity designs in agricultural and environmental economics. European Review of Agricultural Economics, 50(1), 1–28.
Acknowledgements
Insightful and constructive comments from four anonymous reviewers and Rachel Macauley are gratefully acknowledged. The usual caveats apply.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
A Cross-region regression discontinuity in time: potential sources of bias
In interpreting the econometric estimates from cross-region panel datasets, one should be aware of a number of potential sources of bias and limitations in interpretability. In first instance, the statistical information offers no scope for expanding the time domain, in terms of frequency and/or series. This has implications on three grounds: (i) possible effects of policies preceding the initial period may be fully absorbed only in the second period, with no possibility of disentangling the effects of multiple institutional changes over time (Butts, 2021), (ii) some policies may target some regions more than others, thus introducing unobserved heterogeneity on both sides of borders, and (iii) exogeneity in the service initiation date and location may be weakened by sorting, adaptation, and anticipation effects by agents—at least within countries. Particularly issues related to the third point are likely to remain unobserved, and even more so at a regional scale (Hausman & Rapson, 2018).
A second source of potential bias lies in the bias-variance trade-off between risk of omitted variable bias due to unobservable factors correlated with a treatment intervention, thus implying preference for narrow windows—and need for statistical precision—favouring larger windows − . While this problem is bound to occur in many RD designs, it becomes more complex in the presence of sparseness of data near the cutoff point (Green et al., 2009; Wuepper & Finger, 2022). However, even in the presence of large sample sizes at a given window, Gelman and Imbens (2019) warn against the use of over-parameterised regression models, with smooth functions of local low-order polynomials turning out to be more reliable RD estimation methods than global high-order polynomial approximations.
If cross-country administrative boundaries are used as thresholds, unlike a one-dimensional cutoff score, a third methodological problem is the introduction of a two-dimensional distance between location points, which implies changes over space and hence possible spatial heterogeneous impacts on the estimand (Keele & Titiunik, 2015). The latter also entails a risk of biased effects in the presence of a violation of the ‘stable unit-treatment value assumption’ (SUTVA), whereby units receiving treatments do not affect those not receiving it, with no scope for diffusion and spillover effects. One way to redress this potential bias is to exclude agents who ‘might be nearly adjacent’ and subject to different treatments (ibid. 2015: 44). This is not applicable to geographically aggregate data for administrative regions adjacent to African power pool borders.
B DHS sampling design and data limitations
The demographic and health surveys (DHS) have adopted since early implementation a two-stage stratified cluster sampling design, with sample sizes usually ranging between 5000 and 30,000 households and survey variables being mostly the same across countries and time. Beyond a number of limitations and apart from non-sampling errors in the first 1984–89 DHS phase redressed in later rounds, this standard feature enables comparisons over space and time (Belmin et al., 2022: 4; Marckwardt & Rutstein, 1996). In the first stage, stratified samples of census enumeration areas are used as survey clusters, with probability proportional to size. This is followed by a second stage based on equal-probability systematic sampling of representative numbers of households. The combined approach between proportional and equal-size allocations is chosen so as to ensure precision and representativeness at the sub-national level by avoiding excessively small-sample sizes, particularly for small regions. At the same time, urban areas are slightly oversampled so as to be better comparable with rural areas. However, a limitation of DHS is the ad hoc exclusion of areas severely hit by events such as floods and severe civil disturbances (DHS 2012: 74).
Due to confidentiality, georeferencing of DHS locations is masked through random displacements of GPS coordinates: for urban clusters, distances of up to two kilometres (0–2 km); for rural clusters, up to five kilometres (0–5 km), except up to 10 km for a further randomly selected 1% of the latter clusters. Hence, another potential limitation is the extent to which countries apply administrative unit displacement restrictions of no crossing of boundaries: this has been strictly the norm, also at the sub-regional level, in surveys conducted since 2008. Notwithstanding the query that ‘average displacement and the impact of restriction[s] is very country dependent’, GPS coordinate displacement has turned out to have a relatively small impact on linking DHS household data with facility-based surveys (Burgert et al., 2013: 9, 20–21). More substantial issues originate from a lack of consistency in terms of survey round years and from reporting and recall bias by respondents (Footman et al., 2015). These issues make the stated aim of DHS of providing quality data for policy development and programme planning in need of ad hoc modelling approaches, particularly at the cross-country level. Modalities geared to partly account for partial inconsistency in the timing of surveys are discussed below (besides the ‘Sample panels and descriptive statistics’ section).
C Dataset structure and specific country cases
Countries were chosen with at least one survey round with available information on electricity access over the period 1990–2019. This yielded a DHS-based dataset of 39 countries and nearly two thousand sample observations. In a number of cases, administrative subdivisions vary over successive survey rounds within UN-SALB level 1. For instance, Burundi moved from 5 to 18 level-1 provinces, in 2010 and 2016, respectively; Cameroon shifted from 5 in 1991/98 to 12 regions in subsequent survey rounds (2004/11/18; spatialdata.dhsprogram.com). A few countries have no available information for the latest DHS round period of 2015–19. Among others, this concerns the Republic of Congo, with surveys in 2005, 2009 and 2011–12, Cote d’Ivoire, with DHS in 2005 and 2011–12, and Gabon, with DHS limited to 2000–12. With a view to preserving geographical consistency and better cross-region comparability, the econometric analysis focused on the period 2003–18.
In regressions on balanced panels in terms of different DHS waves (e.g., survey data related to 5 regions for Burundi), a dummy variable was used to distinguish countries with medium-run survey round gaps (6–9 years) as opposed to others with wider gaps (11–15 years) between available DHS rounds (Table 6, models [11], [12] and [12b]: mrwdum). This dummy is useful to avoid discrepancies if no distinction is drawn between the two country groups, with an ‘autonomous’ rate of lack of access to electricity in the second period biased in opposite directions—assuming equal dynamics over time—for the former vs. the latter country group accordingly (Table 6: mrwdum vs. constant, tdum, t(app)dum). Relative to the cross-power pool geographical framework, which accounts for border region effects (Eq. (5) and Table 6), one should notice that the sample included a broader group of countries, such as Ethiopia and Kenya in EAPP and Mozambique and Malawi in SAPP. Hence, due caution should be exercised in comparing results based on the two modelling operational frameworks (Eq. (4) vs. Equation (5); see country list in Table 1, under each regional power pool).
As for specific regions within countries, the Malian provinces of Gao, Kidal and Tombouctou were excluded from CAPP-WAPP analysis due to data limitations (Table 3). Relative to the joint power pool sample (Table 1: C/E/S/WAPP), data for Mali covered all provinces and concerned the 2006 and 2018 DHS. Relative to Senegal, geographical consistency was achieved by using DHS information according to 2005 admin1 area subdivisions, without subsequent breakdowns of new regions from pre-existing administrative areas (Kaffrine, Sedhiou and Kedougou, from Kolda, Kaolack and Tambacounda, respectively). As a relatively more complex case of administrative subdivisions over time, Tanzania has undertaken a number of changes in sub-national borders by moving from 26 (2004) to 30 (2017) admin1 areas, grouped into 8 and 9 zones respectively. Among the latter, only two zones (Eastern Province and Zanzibar) have kept administrative borders unaltered throughout the sample period. Hence, for Tanzania, this implies that geographical indicators vary to some extent over different DHS round sub-periods (lnarea, lnaltid). To a lesser extent, Uganda has also registered changes in regional boundaries over the sample period (spatialdata.dhsprogram.com).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Mainardi, S. Electrification, regulation and electricity access backlogs: regional development and border discontinuities across African power pools. Energy Efficiency 17, 34 (2024). https://doi.org/10.1007/s12053-024-10200-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s12053-024-10200-5
Keywords
- Electricity regulation
- Access backlogs
- Power trading agreements
- Border discontinuity
- Beta model/beta kernel