Climate implications of electrification projects in the developing world: a systematic review

This section specifies the research fields covered in our systematic review, and defines key terms or concepts. The review aims to capture literature that addresses the development and climate implications of electrification efforts in developing countries, and thus draws from economics, engineering, and natural science literatures. Importantly, it only includes papers that obtain empirical data on the climate impacts of electrification, (field-based) measurement of energy consumption, or proxies for it (e.g. capacity or quantified use of energy services). In choosing to compile such empirical literature on climate and development responses to electrification, we limit our investigation to real-world ex-post evidence, in contrast to more theoretical approaches that may only partially align with observable patterns.

In speaking about electrification, we first appeal to a binary construct that allows for any type of electricity use (from pico-lighting to full connections to grid power). We then extend this concept of electricity access, according to its multiple dimensions of capacity, availability, reliability, quality, affordability, legality, and health and safety, following rough application of the MTF concept [33].⁶ We additionally investigate how the energy services provided or utilized (using the typology presented in Jeuland et al [27]), vary across generation technologies (fossil fuels, hydropower, wind, solar, geothermal, other) and grid type (fully grid, mini-grid, user-level off-grid, and hybrid grid + off-grid). To characterize the climate implications in each study, we derive estimates of the carbon intensity based on the combination of information on grid type, generation technology (e.g. for off-grid diesel), and/or grid emissions factors, reported by the IEA (in 2017).

Our strategy for capturing articles on this nexus was developed based on the following considerations:

• (a)  
  The need for a tailored search string combining terms in three categories: (a) country or region-specific identifiers to capture developing regions; (b) terms for electrification or power generation technologies, including grid and off-grid specifiers; and (c) emissions-relevant terms, related to electricity consumption or proxies for it. We did not include development and energy service related terms because we determined, based on initial piloting, that these further restrictions were unnecessary or even prohibitive. The final search string, included in the appendix to this article, emerged from an iterative process that added and excluded key words in an attempt to balance the scope and relevance of identified papers.⁷
• (b)  
  Implementation of the search in publicly-accessible interdisciplinary article databases of peer-reviewed literature, namely Web of Science and Scopus. We searched based on the topic field, which considers title, abstract and author keywords only, and limited the scope to English-language articles published after 1992 (and extending through 2018).

The total number of articles captured in this way was a list of 576 peer-reviewed papers; we then applied our study inclusion criteria. Specifically, each included study had to (a) cover an electrification experience in one or more LMICs; (b) be an empirical study that included data from real-world interventions and projects; (c) allow for categorization into grid, off-grid, or hybrid electrification options; and (d) provide information on emissions outcomes, or proxies from which emissions could be inferred, such as electricity consumption levels and technology identifiers. Articles did not have to make an explicit link between grid and off-grid technology and energy consumption, but did need to report on these to allow further analysis and inference based on their reported data. Furthermore, articles did not need to directly analyze emissions and development outcomes to be included in the review, although our analyses of them consider this nexus.

Finally, we were cognizant that we would miss relevant grey literature as well as relevant peer-reviewed articles given the search string and idiosyncrasies of the article databases in which that search was implemented. Thus, we reviewed abstracts and, when necessary, full texts of all references from the retained articles (following application of inclusion criteria), as well as documents citing those retained articles (using Google scholar).

The first round of screening—based solely on titles and abstracts for the original 576 articles—with the study inclusion criteria yielded a list of 135 potentially relevant articles (figure 1). A second round of screening based on a full text review further reduced this list to 32 articles. In most cases, this additional full-text screening eliminated articles that did not rely on real empirical data or that did not allow inference about the emissions involved with the interventions/technologies. The second stage identification of articles based on the abstracts and titles of forward and backward citations to these 32 retained papers yielded an additional 68 potentially relevant articles. Only 4 of these 68 met the study inclusion criteria. Table 1 provides a descriptive classification of the key research foci covered in the final sample of 36 papers, as well as the countries represented in the papers under each theme.

Zoom In Zoom Out Reset image size

After screening, articles were coded according to a system developed and discussed by all co-authors involved in the systematic review, and submitted to a pre-review of the full research protocol (see appendix in supplementary material https://stacks.iop.org/ERL/15/103010/mmedia for the full protocol that was revised based on comments obtained by two anonymous referees). This coding scheme covers the following five aspects:

• General article information.
• Variables related to electrification, specifically: (a) grid/off-grid characterization; (b) end uses for energy provided (production, home consumption, public use, general use); (c) energy services provided, based on the characterization in Jeuland et al [27]; and (d) generation technology.
• Empirical methods, specifically: (a) observational, quasi-experimental, experimental, review, or other;⁸ (b) inclusion and type of counterfactual; (c) unit of analysis; (d) data type; and (e) sample size.
• Energy use or emissions implications, specifically: (a) empirical measures of consumption or proxies for consumption, including units; (b) modeled estimates of consumption; (c) empirical or model estimates of climate-forcing emissions, as well as details related to those estimates (input variables used, forcing agents included); and
• Development implications, specifically: (a) indicators and units; (b) whether based on statistical or econometric analyses; (c) magnitude of impact; (d) significance of impact; and (e) control variables included and assessment of robustness of estimate of impact.

With this scheme, we aimed to extract all relevant and comparable information to obtain a comprehensive overview of the scopes, methodologies, energy (or emission) implications and development impacts of included articles to adequately characterize empirical and quantitative evidence on the relationship between on- and off-grid electrification strategies and climate and development or energy services-related outcomes. The coding process was completed by two researchers. All 135 articles that were initially screened based on title and abstract were subjected to a full text review by both researchers, and the relevant subset double coded to assess consistency. In the few cases with inconsistencies across coders, the disagreements were discussed and resolved through consultation among the study authors.

Overall, we find a high diversity of approaches and inconsistencies across the included studies, as well as the relatively small number of articles in the literature that met our inclusion criteria. Indeed, the themes shown in table 1, which for the most part are based on the original authors' own descriptions of their contributions, are quite heterogeneous, with the largest concentration of papers having to do with the determinants of electrification, electricity consumption behavior, or energy poverty; and electricity system design and planning or appropriate technology. Other papers are predominantly focused on the impacts of electrification or sustainability issues, and very few (only 2) discuss emissions as a central aspect of their research. Our analysis therefore focuses on a descriptive mapping of themes and topics which is supported by descriptive and comparative quantitative analyses. More in-depth analysis techniques to answer our primary research questions, such as meta-analysis, were deemed to not be appropriate due to an insufficient number of studies with comparable research approaches.

Building on this last point, we note here the main limitations that we judge to affect this review. First, we are highly constrained by the limited available literature, and the fact that it does not perfectly address the nexus of electrification's climate emissions and development implications. The limited literature means that many LMICs and regions are scarcely represented in the review, in particular West Africa, parts of Asia outside of South Asia, and many countries in Latin America. Larger, more populous LMIC countries tend to be better represented but Nigeria, the Democratic Republic of Congo, and Indonesia are notably absent. Countries further along the energy transition, such as lower middle-income countries, are better represented than less developed countries. This geographic bias may also be due to our limiting the review to English language literature, which favors Anglophone countries. Second, due to our desire to quantify emissions, we neglect qualitative literature, which may nonetheless carry many insights particularly regarding the development and well-being implications of electrification. Relatedly, again due to our desire to quantify emissions, the search prioritized identification of articles with real energy consumption estimates, but most development and even climate-relevant articles do not include such estimates, which can be difficult to obtain owing to lack of metering or detailed survey data on energy use patterns. Given these limitations (more inclusion of middle-income and metered situations), it is possible that our review over-estimates the level of consumption in many electrification experiences.

Considering first the geographical and temporal coverage in this literature, our first observation is that more than a third of the papers (14 of 36 papers) focus on the electrification experience of India (table 2). India is certainly an important country in that it is both very populous and has made great strides in extending electricity access to unconnected communities and households over the past two decades [34, 35]. It is also a leading location for adoption and experimentation with off-grid technologies and supporting business models [36]. What is striking about this relative concentration on India, considering the relatively small set in this literature overall, however, is the dearth of papers from other regions where electrification rates have also improved at different rates [37].⁹ This is perhaps most true of Sub-Saharan Africa outside of East Africa, where we identified a sum total of four papers. Latin America and East and Southeast Asia are also poorly represented in this literature, but the experiences of those regions are less relevant to our review, since they are already nearing universal electricity access and most of their countries have only marginally relied on off-grid technologies. The depth of this literature is growing over time; two thirds of the papers are from after 2014 (figure 2).

Zoom In Zoom Out Reset image size

Turning next to the relative balance of on-grid and off-grid papers, we observe that a slightly larger number focus on off-grid electrification approaches (25 papers, vs 20 papers for on-grid; some papers include both of these approaches), which is relatively surprising considering that grid-based electrification still supplies the majority of new connections in developing regions. This may reflect the relative novelty of off-grid technology and a greater perception of the need for research on the effectiveness of efforts to extend access using it.¹⁰ What is more surprising is the predominant focus on household users (relative to more intensive energy users such as businesses and industries), and on home consumption rather than production. Only 9 (25%) papers include a unit of analysis for energy consumption and use other than a household, and only 8 (22%) address production, whether home or otherwise. Furthermore, the most commonly identified energy services are consumption services: lighting (28 papers), followed by entertainment (16) and phone charging (10). These patterns partly reflect the fact that in the past off-grid electrification (for which there are more papers) infrequently allowed more intensive energy uses due to low installed capacity (table 2), and the fact that researchers writing papers on grid electrification have infrequently applied an energy services lens—perhaps because capacity is less constraining. Consistent with this idea, few papers speak to use of services that require more power—cooking, heating, or refrigeration. The tiers (1–2 vs 3–5) inferred from these data, disaggregated by grid and off-grid approaches, are shown in figure 3.

Zoom In Zoom Out Reset image size

The generation technology covered in the papers also parallels these trends. Roughly half of the studies we reviewed were related to solar energy; these encompass most off-grid studies (a few papers consider diesel generation). Relatively few studies explicitly mentioned fossil fuel-based generation, and we therefore used a 'non-specific' generation category to account for studies that relied on the grid or were otherwise unclear about generation.

The included empirical studies nearly all use observational data and analysis (30 of 36 studies); the remaining use experimental (2 studies) or quasi-experimental (5 studies; note that one such study includes a mix of observational and quasi-experimental approaches) research designs. In accordance with these methods, relatively few studies explicitly claim to identify causal links between electrification and other outcomes; those claiming to do so primarily use experimental or quasi-experimental methods. Moreover, among the observational studies, roughly half include multivariable regression methods to analyze the primary relationships of interest, but the other half do not control for other factors or are mainly aimed at descriptions of electrification experiences supported by descriptive statistics and techno-economic modeling.

One explicit goal of our study is to investigate the nexus between climate implications and electrification. Hence, as discussed in the methods section, our systematic review only targets studies that allow estimation of emissions outcomes, or proxies from which emissions can be inferred based on electricity consumption levels and technology identifiers. Only four studies reported emissions information directly, and these reports were limited to CO₂. Since the remainder required estimation of emissions from consumption or generation capacity, we standardized this approach and applied it even for the four studies that directly reported emissions estimates.

More specifically, we calculated the emission intensity of on-grid and off grid electricity consumption as well as of the investigated energy services in the respective study year and country covered by our sample. For grid-based electrification, we obtained the emissions intensity of the public grid by using data on grid-connected power plants from the IEA (2017). This assumes that national grids are mostly interconnected such that average emissions are relevant, and that the marginal generation for additional connections entails emissions that are similar to the average for the country. We also checked that our sample of countries did not include any that depend heavily (>5%) on electricity imports, which was the case for all but Mozambique (we excluded this study as a result).

For off-grid electrification, we assumed that emissions would be equal to zero for renewables (e.g. solar, micro-hydro), and used capacity and emissions estimates for non-renewables (e.g. diesel generators) taken from the IPCC emissions (IPCC 2006). We addressed data challenges that arise due to mixed electrification and inconsistent reporting conventions in our included studies by applying appropriate assumptions or estimations. Specifically, the main challenges in implementing this approach were insufficient details on the relative balance of renewable and non-renewable sources in cases with mixed electrification options, or inconsistencies in authors' reported units or consumption proxies (e.g. reporting lighting hours per day versus kWh/day). When details reported in the studies were insufficient to obtain valid estimates of consumption (e.g. lighting hours were specified, but the most common type of light bulb in the setting was not), we filled in the median consumption value from the full set of other studies covering the same energy services as the one in question, and calculated emissions according to the logic previously described.

Based on this approach, we distinguish the tier implications according to the electrification strategy (on-grid or off-grid) (figure 3). As shown, off-grid approaches, dominated by solar technology, are more likely to provide low tier access relative to grid-based approaches. Though there are relatively few studies that cover on-grid and off-grid access together, the results from these studies suggest that hybrid models tend to more closely resemble off-grid in their implications for the level of electricity service. This is likely due to the types of communities and users (more remote and lower income) targeted by the interventions that are analyzed by the respective studies.

In addition, we map the tiers of consumption or energy services used, to emissions, as shown in figures 4 and 5. Since lighting, phone charging, and some basic entertainment services have been provided by off-grid or solar technology, the lower consumption bounds for each of these studies tend to come from such solutions, whereas mini-grids, diesel generation, or the grid are required for other higher energy consuming services. Heating, cooling and cooking services are almost never seen in combination with solar power, and are predominantly provided by grid electricity. Off-grid technology can in some instances entail very high emissions that exceed those of typical grid systems, as in the case of diesel generators that provide higher tier access (figure 4). Considering the emissions implications of different energy services, we observe that lighting, phone charging, and basic entertainment in our sample have negligible climate effects—this combination of services is also easily provided by systems that rely on off-grid solar technology alone. Emissions are higher with the addition of cooking and other appliances, as well as heating. Yet low emissions situations with even these more demanding energy services are also occasionally found—most typically with less carbon intensive electricity grids, or as minority uses from off-grid solar systems.¹¹

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Our sample thus reflects well the basic intuition about the patterns of energy services associated with off-grid and renewable technologies in low income countries, that is, that these provide energy mainly for low capacity services, such as lighting, phone charging and entertainment, whereas energy-demanding tasks, at least at a large scale, have been more heavily dependent on the provision of grid power or emissions-intensive diesel. Still, the study sample also indicates a research bias towards small-scale solar installations, as opposed to mini-grids.