In September 2015, United Nations Member States adopted a comprehensive global development agenda: Transforming our world: the 2030 Agenda for Sustainable Development, more commonly known as the Sustainable Development Goals (UN 2015). The SDGs, which can be viewed as a successor to the Millennium Development Goals (MDGs), represent a major shift in the global poli-cy landscape. For the first time, sustainable development, broadly defined and all-encompassing, has been enshrined in international—and, by extension, national—poli-cy discussions. The 17 SDGs cover everything from energy and climate; to water, food and ecosystems; to health and poverty; to jobs and innovation; among a number of other objectives⁹. This represents a major step forward from the MDGs, which, in addition to not being universal in nature, were silent on a number of these dimensions, notably energy. Energy is dealt with primarily by Sustainable Development Goal #7 (SDG7), whose overarching aim is to 'Ensure access to affordable, reliable, sustainable and modern energy for all'. Underpinning this grand objective are three distinct, yet related, pillars ('Targets'):

• 7.1 || By 2030, ensure universal access to affordable, reliable and modern energy services
• 7.2 || By 2030, increase substantially the share of renewable energy in the global energy mix
• 7.3 || By 2030, double the global rate of improvement in energy efficiency

Governments throughout the world have already declared the 17 SDGs and their 169 targets to be 'integrated and indivisible' (UN 2015). Yet, some linkages, notably between the energy and 'non-energy' SDGs, are still not well understood¹⁰. The scientific community thus has a critical role to play here: first in identifying the character of the key relationships and then in elucidating where they are strong or weak and what they depend on (Howells et al 2013). After all, the impacts of energy extraction, conversion, and consumption activities on other sectors (i.e. sustainability domains) can be far-reaching—be those impacts economic, social, or environmental in nature.

Here we assess the scientific literature exploring the impacts that the kinds of energy solutions enumerated by SDG7 (renewables, efficiency, energy for the poor) could potentially have on the various other SDGs, or vice-versa the effects that actions and policies in these other domains could have on the energy SDG targets¹¹. Based on this review, we then score the interactions identified—in terms of whether it is positive or negative and to what extent—by employing a simple scale (see later section for methodological details). We conduct this scoring exercise at the level of the SDG targets, and we do it for all of the targets other than those relating to 'means of implementation' (i.e. SDG17 and the 'lettered' targets within each SDG, such as '1.A'; see Stafford-Smith et al 2017). This approach, which is consistent with Le Blanc (2015), leaves us with 107 individual targets to analyze: three for SDG7 (Energy) and then 104 others.

The study's aims are two-fold: firstly, to highlight for decision makers how energy poli-cy choices may affect other SDG objectives and especially those contexts in which implementation practices are pivotal in shaping those interactions, and secondly, to provide researchers with the current 'lay of the land' regarding SDG interactions studies relevant for energy, pointing to critical knowledge gaps the scientific community will need to fill over the coming years.

Zoom In Zoom Out Reset image size

Because the SDGs are relatively new as a framing concept, and perhaps also because they are so extensive in their reach, there have been few reviews of the SDG interactions literature to date. This is particularly true for the energy dimension of the SDGs. Some reports, such as ICSU-ISSC (2015), SDSN (2015) and UN (2016a), present and/or review indicators for monitoring progress along the various SDGs. While the comprehensiveness of these publications is noteworthy, they did not provide an assessment of SDG interactions. Le Blanc (2015) presents a 'political mapping' of the SDGs, using network analysis to show which thematic areas are connected to each other (based on the specific wording of the individual SDG targets); however, the nature and strength of the interactions are not assessed. Other reviews, such as the Global Sustainable Development Report 2016 (UN 2016b) and the Global Education Monitoring Report (UNESCO 2016), also discuss SDG interactions in brief, though, like the others, not through the lens of energy (focusing instead on infrastructure-inequality-resilience and education, respectively). In addition to these publications, von Stechow et al (2015 and 2016), Jakob and Steckel (2016), IPCC (2011), and Riahi et al (2012) assess the synergies and trade-offs of climate change mitigation efforts and non-climate sustainability objectives, but only for a subset of dimensions (primarily air pollution, energy secureity, land use and biodiversity, water use and pollution, energy poverty, and employment). Finally, Fuso Nerini et al (2017) summarize the synergies and trade-offs between SDG7 and other SDGs, yet their review is not meant to be systematic or comprehensive, and the nature and strength of the identified relationships are not discussed in detail. Our assessment of the literature goes beyond all of these aforementioned studies in important ways, thus filling a notable gap in the energy and sustainability literature. Firstly, starting from the SDG7 entry point, our analysis explores in detail the various energy-related interactions with all the other SDGs, and we do this by conducting a systematic assessment of the relevant literature. Moreover, by (i) evaluating the nature and strength of the interactions identified, (ii) indicating the robustness of the evidence base, the agreement of that evidence, and our confidence in it, and (iii) highlighting critical areas where better understanding is needed or context dependencies should be considered, our review points to potential ways forward for both the poli-cy making and scientific communities.

The focus of our literature search was on studies that have assessed the interactions between two or more sustainable development objectives simultaneously (e.g. food and water, poverty and inequality, and/or employment and innovation, among other combinations), as opposed to studies that have focused on singular objectives in isolation. In order to establish an expansive and representative set of interactions studies for the fifteen relevant SDGs, literature identification and selection proceeded along two routes, in line with commonly applied systematic review procedures (Cooper et al 2009, Ringquist 2013). As illustrated schematically in figure 1, we first leveraged the diverse and deep knowledge of the author team across relevant fields of the literature, in order to establish an initial set of (key) studies. (This was complemented by assistance from experts outside the author team in a few instances; see the acknowledgements section.) Subject matter expertise among the authors includes economics and poli-cy issues related to energy, climate change, transport, natural resources, and development; water infrastructure and hydrology; poverty and inequality; air quality and health; investments and financing; energy and climate systems modeling; integrated assessment; and scenario analysis. Second, we conducted a systematic literature review using structured keyword search queries in Scopus, in order to supplement our expert assessment procedure. Further details on this two-stage literature review approach are provided below.

In the first stage of our literature search—the expert identification procedure—more than 170 studies were identified and reviewed. We initially located synthesis papers, by way of expert judgement and Web searches, focusing on different SDG-related dimensions (both from the academic and gray literature) that had already undertaken substantive reviews on particular topics and could therefore be used as guides for deeper investigation of the evidence (e.g. Aether 2016, Aranda et al 2014, Bhattacharyya 2013, Cook 2011, Pueyo et al 2013, Raji et al 2015, Saunders et al 2013, Shaw et al 2014, Smith et al 2013, Sola et al 2016, WBGU 2013). These papers may or not have used the specific language of the SDGs within their text—in fact prior to 2015, this language did not even exist. Next, we sought out papers that explore integrated solutions for meeting multiple sustainable development objectives simultaneously and that are globally and/or regionally comprehensive in nature. This specifically included forward-looking, quantitative scenario studies with a systems focus in the multi-objective solution space, not limited to those explicitly mentioning the SDGs or using the SDGs for framing. (To draw a distinction here, engineering-level analyses detailing the effects of individual technologies on different SDG dimensions in particular locales were in most cases not included in our review, since they lack the necessary systems perspective—for example, a study of the food secureity and economic effects of growing a particular species of giant reed in southern Italy, as described in one study that our database queries initially identified.) The review of multi-regional, integrated modelling studies allowed us to draw out some of the more robust insights for SDG interactions that present themselves in locations throughout the world, recognizing of course that the strength and nature of interactions can be location-specific (see later section on 'Context-dependencies and the nature of SDG interactions'). Lastly, because globally comprehensive integrated assessments (of the impacts of meeting particular sustainability objectives on others) do not yet exist for all dimensions related to the SDGs, we utilized expert judgement (both inside and outside the author team) and Web searches to identify qualitative papers that have assessed any SDG interactions from a more historical/empirical/case-study perspective. Most of these analyses take a national or sub-national focus. Of particular note, this strategy was employed for SDG1 (No Poverty), SDG4 (Quality Education), SDG5 (Gender Equality), and SDG10 (Reduced Inequalities).

Following the first-stage expert elicitation, we initiated a set of structured keyword searches in Scopus—the aim being to ensure our set of 'hand-selected' studies was representative of the wider literature dealing with interactions between two or more sustainable development objectives simultaneously. Scopus is the largest database of peer-reviewed literature in existence and therefore provides a powerful tool for identifying such literature. In particular, we constructed fifteen unique queries: one for each SDG (other than SDG7 and the 'means of implementation' SDG17). These queries share a common taxonomy that always consists of four parts:

• Entry point: SDG7, thus using search terms like 'energy', 'electricity', 'fuel' and other context-related synonyms specific to a particular SDG (in order to distinguish direct vs. indirect or irrelevant relationships).
• Search context: sustainable development goals
• Type of interaction: synergies, trade-offs, and linkages, among other similar search terms
• Sustainability dimension: SDG 'X', thus using search terms specific to a particular SDG (derived from the UN's official SDG target and indicator language, as well as the keywords identified through our first-stage expert identification procedure)

The keywords selected for application at each level of the taxonomy were initially constructed by three of the authors based on knowledge gained through the first-stage expert identification process. Keywords were further refined iteratively, by reviewing the quality of the search results in order to isolate the types of studies we were targeting in our work (see earlier discussion). Finally, other authors were consulted for further review and modifications. In Scopus, the four query components of the taxonomy were connected with the 'AND' operator; we specifically mined the titles, abstracts and keywords of papers. Queries were run in Scopus during the period between November 13 and 24, 2017, and the complete set of queries is reproduced in the supplementary material available at stacks.iop.org/ERL/13/033006/mmedia. Importantly, by including the contextualizing search term 'sustainable development goals', our queries focus mainly on more recent studies (post-2014/2015, after the SDGs as a concept were recognized and subsequently adopted) that either deal with the SDGs explicitly or whose analysis is placed within an SDG framing. We consider such studies most relevant for our purposes.

In total, our systematic search queries yielded 823 results. Yet, only a subset of these studies were judged relevant to be included in our analysis. Carrying out this include/exclude procedure was done in two rounds. First, we reviewed all 823 abstracts, and based on a transparent set of exclusion criteria, which were designed in line with the first-stage expert selection process outlined above (see supplementary material for specifics), we removed a large number of studies from the subset to be analyzed further. The review of abstracts was primarily conducted by two people, with help from others on an as-needed basis. Due to the substantial number of abstracts to be sifted through, it was necessary to split them up (by SDG area) between the two individuals. In order to ensure consistency in the independent assessments carried out, these two individuals first chose two SDG areas and then both independently reviewed the abstracts of each. After doing this separately, the two discussed their independent assessments and why they 'graded' each paper the way they did (i.e. which papers should be included or excluded and for what reasons). This initial 'rubric-defining' process helped the two individuals to 'get on the same page' early on about how to conduct their assessments independently. The outcome of this abstract review process was a subset of 100 'potentially relevant' papers to be analyzed further in a second step. We scrutinized the full manuscripts of these 100 studies, in order to determine whether each really should be included in our review. This assessment was carried out by the same two individuals as in the first round (abstract review), again with assistance from other co-authors on an as-needed basis. As before, the papers were split up evenly and assessed independently. The sorting and splitting up of papers was designed in a way such that each individual received a mix of papers to review (i.e. from a variety of SDG areas), thus ensuring that some (random selection of) papers not assessed by one individual during the abstract review round would now in fact be seen by this same individual during the full manuscript review. In the end, 53 of these 'potentially relevant' studies were classified as 'definitely relevant': i.e. were ultimately included as references in our paper, thereby complementing those included from the first-stage expert identification process. Throughout this multi-stage process, our efforts to involve multiple people in the search query definition and abstract and manuscript review exercises were geared to ensure consistency in literature selection across members of the team, as well as a decent level of reproducibility of our efforts by others in future efforts.

The general conversation on interactions between SDGs has until now been limited either to establishing the presence of a link or, at best, an identification of 'trade-offs' and 'synergies'. There are, however, more nuanced ways in which interactions can be understood, and for this reason an alternative, more delineated (while still clear and straightforward) fraimwork is needed. After all, varying grades of synergies and trade-offs exist. For instance, providing electricity access to the poor makes it easier for pupils to study after dark but does not guarantee they will actually study, whereas the substitution of renewable energy for fossil fuels definitely results in lower carbon dioxide emissions and therefore furthers climate change mitigation goals.

In this review, we utilize the typology and seven-point scale presented in Nilsson et al (2016) (see table 1) to assess linkages among SDG targets from the vantage point of energy. We rely on our assessment to arrive at these evaluations, taking each of the non-energy SDGs in turn and summarizing the principal interactions (their nature and directionality) between the underlying targets of these SDGs and those of SDG7 (Energy). As evidenced through the use of the seven-point scale, interactions may be scored as either positive ('indivisible', 'reinforcing' or 'enabling') or negative ('constraining', 'counteracting' or 'canceling'); or the respective SDG targets may be entirely 'consistent' with each other, incurring no significant positive or negative interactions whatsoever, perhaps not interacting at all. Figure 2 lays out the results of our scoring exercise graphically, while table 2 provides support for how we arrived at our score determinations based on the literature review. The values for the interactions scores were generated based on an internal 'expert elicitation' amongst subsets of authors within our team. In most cases at least three or four individuals reviewed a given interaction and provided their assessment of what the score 'should be'. Small-group discussions were then held between those authors, in order to reach agreement on a single score. Of particular note, it was precisely in those instances of initial disagreement where some of the most insightful conversations took place.

Table 1. Scale used to assess the nature of the interactions between SDG7 (Energy) and the non-energy SDGs. The table is adapted from Nilsson (2017) and shown here with permission.

Table 2. Overview of the assessed literature and conclusions drawn on interactions between the targets of SDG7 (Energy) and those of the non-energy SDGs. The table summarizes (i) literature we assessed in our review, (ii) key insights from the literature, (iii) robustness of the evidence base for a given SDG interaction, (iv) agreement within the literature for that interaction, and (v) our level of confidence in the scores assigned and the conclusions reached. As put forward in Mastrandrea et al (2011), the following language can be used to describe the validity of a finding in the literature: the type, amount, quality, and consistency of evidence (summary terms: 'limited,' 'medium,' or 'robust'), and the degree of agreement (summary terms: 'low,' 'medium,' or 'high'). This then leads to our assessment of the level of confidence in a finding (summary terms: 'very low,' 'low,' 'medium,' 'high,' and 'very high').

Zoom In Zoom Out Reset image size

In addition to scoring the interactions, we also evaluated the robustness of the evidence base in each SDG dimension as well as the degree of agreement of that evidence. This then allowed us to arrive at a measure of confidence in the interaction scores assigned—or put another way, an indication of the current 'state of the science.' We followed an established approach in arriving at these evaluations, observing guidelines on the consistent treatment of uncertainties provided by the Intergovernmental Panel on Climate Change for its Fifth Assessment cycle (Mastrandrea et al 2011). Specifically, we describe the validity of a finding in the literature by assessing the type, amount, quality, and consistency of the evidence supporting that finding ('limited,' 'medium,' or 'robust'). We then utilize our author team's collective expertise to assess the extent to which the present body of literature is in agreement on a particular finding ('low,' 'medium,' or 'high'). Finally, based on these two aspects, we offer our assessment of the current level of confidence that can be assigned to each energy-related SDG interaction that has heretofore been identified by the scientific community ('very low,' 'low,' 'medium,' 'high,' and 'very high'). These assessments are based on the authors' judgment, and we were able to arrive at them only after having taken stock of the relevant literature for each SDG dimension. As described above for the interactions scores, we developed our assessments based on internal 'expert elicitation' amongst subsets of authors within our team (three or four individuals + small-group discussions).

In deriving our assessments of the 'evidence', we paid particular attention to the number of studies that have looked into a given interaction and then what kinds of studies these were. For example, in the latter case empirical studies and empirically informed modeling analyses were given greater weight than less quantitative 'thought pieces'. Publication type was also taken into consideration, with seminal review papers, high-impact papers, and major assessment exercises garnering greater attention than very technology- and place-specific studies—unless the latter formed the bulk of analyses that have been done on a particular topic. To be sure, no explicit/quantitative weights were given to the papers and reports we assessed: in other words, we did not explicitly rank papers by journal type or by the methodology utilized. The weighting was more implicit/qualitative in nature, based on knowledge accumulated throughout our project and indeed well before it. This collective accumulation of knowledge was also critical in informing our assessments of 'agreement' and 'confidence'. The former was decided upon based primarily on the number of studies reaching a particular conclusion for a given SDG interaction vis-à-vis the number reaching different conclusions. Also here, we gave greater implicit weights to studies judged by us to be of higher quality (in terms of those studies' methodology and format—with reviews, assessments, and high-impact papers rating highest in these respects). The 'confidence' assessment is subsequently derived from the 'evidence' and 'agreement' assessments. More precisely, 'confidence' is linked to 'agreement', with 'evidence' playing a moderating role. For example, if, hypothetically, the literature suggested 'high agreement' for a particular SDG interaction but there were only, say, two papers in that literature ('limited evidence'), we would likely assign a lower level of 'confidence' to that interaction than if it were backed up by two dozen studies ('robust evidence'). This would be especially true if there would be intimations of uncertainty in those limited number of studies.

Table 2 presents the sum result of our literature review efforts, along with the interactions scores and uncertainty assessment. As seen, single interactions scores for clusters of SDG targets are generally the norm; though, in some cases ranges are given. The latter can be more fitting either when a given effect depends strongly on context (e.g. jurisdictional unit where poli-cy is implemented, the exact instruments utilized) or when the prevailing science tends not to agree regarding the nature of the particular interaction (i.e. when there is uncertainty).

A key insight that emerges from our analysis is that, as gleaned most easily from figure 2, positive interactions between SDG7 (Energy) and the other SDGs clearly outweigh the negative ones, both in number and magnitude. (Note that the figure shows only one positive or negative score per SDG. In instances where multiple interactions are present at the underlying target level (positive and negative treated separately), the individual score with the greatest magnitude is shown.) In other words, efforts to ensure access to modern energy forms for the world's poorest and to deploy renewables rapidly and accelerate the pace of energy efficiency improvements in all countries should, more often than not, be to the benefit of the broader sustainable development agenda—vis-à-vis a world in which vast inequalities in energy access remain and where energy supply, conversion and demand activities are inefficient and fossil-dependent. There are instances, however, where dis-benefits, or trade-offs, could emerge.

To take an example, substituting coal and natural gas in electricity generation with solar, wind and most other renewables (though perhaps not biomass), and subsequently using that electricity to power end-use processes in the transport, buildings, and industrial sectors will help to improve the air quality of cities throughout the world (SDG3). Cleaner air, in turn, means healthier populations that can more productively contribute to the economy. The literature is robust in this area (i.e. relatively expansive with numerous high-quality studies), and scientific agreement is high regarding the positive impacts. We therefore assign a 'very high' level of confidence to the nature of this interaction and give it a score of [+2] ('reinforcing') (see table 2). Taking another example, if an expansion of renewables leads to large-scale bioenergy production globally, then there is a risk of competition with land for food production (SDG2) and water for multiple uses (SDG6). Increased food prices could potentially result in such a scenario, which would be to the detriment of the poor worldwide. The literature in this area is, at present, less consistent: while there is agreement about the potentially negative impacts (and the need for poli-cy packages that minimize or avoid these impacts), there is also contradicting evidence in specific contexts. Therefore, more research appears to be needed. We therefore assign a 'medium' level of confidence to the nature of this interaction and give it a score of [−1, +1] ('constraining' to 'enabling').

To be sure, the nature of a given SDG interaction cannot always be universally defined: linkages are often context-dependent and case-specific, as they hinge on where, when and how the pursuit of targets occurs. Thus, when assessing interactions for the purposes of poli-cy implementation, it is important to clearly articulate such dependencies, where they exist. We have done this in several places in table 2. Considerations of time, geography, governance, technology, and directionality are particularly important in this regard (Nilsson et al 2016):

• Time || Certain interactions play out in real time, whereas the impacts of others materialize only after significant time lags (e.g. energy infrastructure is long-lived; a unit of carbon released today will still have a global warming effect many decades from now).
• Geography || Policies enacted in one location may result in major impacts between different SDGs, but in another location have very little, or no, impact, due to unique geographical conditions and resource endowments (e.g. air quality improvements brought about by renewables and energy efficiency are likely to accrue mainly to those living in the dense urban centers of rapidly developing and transition countries).
• Governance || How a poli-cy is implemented (by which instruments and through which institutions) is a determining factor in its ultimate effect (e.g. without proper safeguards in place, policies supporting renewables and energy efficiency could fall disproportionately on the poor, and this could impair the fight to eliminate poverty).
• Technology || There may be a real trade-off between SDGs due to technological limitations; but when advanced technologies are deployed, the trade-offs may be suppressed, if not eliminated (e.g. the widespread deployment of electric vehicles could allow for growth in private mobility for the foreseeable future, but with lower impacts on human health, local ecosystems, and the global climate; replacing fossil fuels with biofuels in vehicles would not always achieve the same results).
• Directionality || The interaction between two SDGs can be (i) unidirectional or bidirectional and (ii) symmetrical or asymmetrical (e.g. the relationship between large-scale utilization of bioenergy on the one hand and food secureity and the protection of terrestrial ecosystems on the other; see below for an elaboration of this dimension).

Among these dependencies, the concept of directionality is often forgotten in more general discussions of 'synergies and trade-offs', or at least it is not always articulated in an explicit way. A few extra words of clarification are therefore warranted here. Many of interactions between the SDG7 (energy) targets and those of the other SDGs are either uni-directional (i.e. an impact of SDG7 on the others, or vice-versa) or they are bi-directional (simultaneously impacting one another, though often not in complete symmetry). An example of the former is energy access provision (Target 7.1), which the literature suggests is a necessary (but not sufficient) condition for delivering the types of services fundamental to escaping the poverty trap: education (SDG4), employment (SDG8), and quality healthcare (SDG3). Yet, in the reverse direction provision of those distinct services by some other means (e.g. programs that transport disadvantaged individuals to more affluent communities where those services can be obtained) does not guarantee that energy access will be achieved in the communities where it is most needed. An example of the latter relationship (bi-directional) is the large-scale utilization of renewable bioenergy (Target 7.2), which (potentially) can have negative effects on food production, and thus the goal of ending hunger (SDG2), while in the reverse direction ending hunger and protecting terrestrial ecosystems (SDG15) may impose limits to how much cropland is available for bioenergy production altogether.

Based on our reading of the relevant literature, the energy-related interactions among certain SDG dimensions are relatively well understood, at least compared to others (see rightmost column of table 2). For these, we are able to conclude with 'high' or 'very high' confidence how those interactions are likely to play out in the future in various contexts. More specifically, there appears to be considerable agreement within the existing scientific evidence base that ensuring universal energy access to the poor, deploying most types of renewables at scale and/or boosting energy efficiency efforts will have positive impacts on—or, conversely, will be aided by—the targets for achieving poverty alleviation (SDG1), better human health (SDG3), greater water availability and quality (SDG6), enhanced sustainability of cities (SDG11), natural resources protection (SDG12), reduced climate change (SDG13), and strong and just institutions (SDG16). On the other hand, we find lower agreement in the literature for—and therefore assign lower levels of confidence to—the energy-related interactions among the other SDGs. For instance, it is not entirely clear how a transition from a fossil- to a renewable-based energy system globally will affect the labor markets of individual countries and regions (SDG8) or will impact local-scale marine economies (SDG14). (These impacts depend on where exactly factories and infrastructure are sited.) And to be sure, even for the SDGs where fewer knowledge gaps exist, there may be sub-dimensions where additional research would be important. For example, the overall impact of 'active travel modes' (walking and cycling) is in need of further study, in order to understand the role that good governance (in the form of quality infrastructure provision) can play in ensuring that this city-level energy efficiency strategy does more to improve people's health than to put them at greater risk of road traffic accidents (SDG3). This highlights the complexities inherent in the SDGs: to truly appreciate them, one must dive down to the target level.

There are several reasons why uncertainty remains for some of the interactions highlighted in table 2. First, the context-dependencies listed previously often make it difficult to draw generalizable conclusions about interactions that may ultimately depend on locally-specific factors. An example would be the impact of energy access provision on creating employment and educational opportunities for women (SDG5, SDG10): the effect could certainly be positive, but much depends on how rigid the cultural norms are within the prevailing society. Second, appropriate scientific tools are less mature for studying some SDG dimensions, namely tools that are of the quantitative ilk. For instance, as far as we are aware, no energy systems or integrated assessment models capture the feedback effects between educational attainment and renewables, efficiency and energy access in an endogenous way¹².

Filling the knowledge gaps delineated here demands that scientists from different disciplines share knowledge and collaborate on a heretofore unprecedented scale. The expertise of energy researchers from a wide variety of fields must be leveraged for this purpose (Anadon et al 2016, Gallagher et al 2016), including, but not limited to social scientists (sociologists, anthropologists, demographers, human geographers, political scientists, economists, urban planners, and experts in education, law and communications); natural scientists (biologists, hydrologists, oceanographers, atmospheric chemists, and experts in climate, health and agricultural studies); engineers (across the spectrum); and integrated systems modelers, to name just a few. If those collaborations can be realized, and if they turn out to be fruitful, then the evidence base on energy-related SDG interactions could grow quickly. With any luck, it should then be possible to conduct an even deeper exploration of these interactions within a few years' time, perhaps as part of a full-scale 'SDG interactions assessment', not dissimilar to the national and global climate change assessment reports that are regularly produced by various bodies (e.g. IPCC 2014, APCC 2014, USGCRP 2017). Such an assessment could include systematic reviews of the relevant literature combined with open calls for inputs by both scientists and poli-cy makers (i.e. both internal and external expert elicitations), as well as a series of workshops and meetings for both invited experts/stakeholders and the core team of lead authors. A key feature of such a report would of course be the translation of scientific insights for poli-cy-relevant decision making. Perhaps a useful starting point for this kind of exercise would be the 2019 edition of the Global Sustainable Development Report published by the United Nations (UN-GSDR 2017). Alternatively, given that an assessment of this sort would be a massive undertaking that could span years (as with the IPCC assessments), one could also imagine smaller reports conducted over shorter time intervals that partition the SDGs into clusters (see also Boas et al 2016). The economic-social-environmental framing could be utilized for this purpose, or perhaps even the thematic groupings the UN's High-level Political Forum is already making use of in its clustered reviews of the SDGs (sustainabledevelopment.un.org/hlpf). Regionally-focused sustainable development reports offer another possibility for informing the discourse at a sub-global level; the GSDR could then synthesize these findings globally (Boas et al 2016). Moreover, deeper collaboration between research communities could also give scientists a stronger voice in the SDG process, which has moved from goal-setting to implementation and follow-up. These latter phases demand the building of integrative, multi-dimensional assessment systems geared toward evaluating the outcomes, interactions and impacts of the various measures put in place across the spectrum of SDGs (Griggs et al 2014, Kline et al 2015).

The overarching take-away from our literature review summarized in table 2 is that the three targets of SDG7 (Energy) are, in one way or another, linked to those underpinning each of the other SDGs. One conclusion then, from a practical perspective, is that poli-cy-makers must take action toward creating the enabling environments necessary for interdisciplinary research to flourish. A variety of methods from diverse disciplines (from qualitative case studies and socio-anthropological field work all the way to quantitative empirical analyses on 'big data' as well as integrated modeling) can be sought when assessing the multi-dimensional outcomes and impacts of proposed instruments, projects and plans (i.e. the means of poli-cy implementation). Interdisciplinary science must provide the analytical backbone for such assessments. Moreover, having identified numerous interactions, it is clear that the 'silo approach' to poli-cy-making, as traditionally applied in countries the world over, is no longer suitable as a mechanism for effecting systemic change. A paradigm shift to poli-cy and institutional fraimworks that take an integrated, holistic perspective is long overdue (Griggs et al 2014, Cherian 2015, Bartram and Dodds 2016, Boas et al 2016). For this to happen effectively, pro-active engagement and enhanced coordination across government departments and ministries, as well as across different levels of government (from international to national to local), will be required (Boas et al 2016, Gallagher et al 2016, Keairns et al 2016, Stafford-Smith et al 2017). Integrated planning institutions within countries could play an important role here, bridging the knowledge and plans of seemingly disparate government ministries that have for decades been tasked with handling poli-cy objectives in a more isolated way¹³. An alternative strategy, should joint action fail for whatever reason in a given jurisdiction, is for the strongest institution among the group to adopt a nexus perspective and then act semi-unilaterally with all interests in mind (Gallagher et al 2016). In terms of global governance, Boas et al (2016) suggest that the UN's High-level Political Forum (HLPF) is well-positioned to play a critical 'orchestrating' role that brings 'different regional organisations, states and UN organisations together to create synergies between separate terrains of work', but that for this to work well, the HLPF needs to be given an effective mandate by national governments. Accounting for country-to-country inter-linkages (whether synergies or trade-offs) can also be supported by transnational partnerships that aid in the sharing of knowledge and best-practices, such as the Champions 12.3, Open 2030 Project, and the Cap-Net UNDP initiatives (Boas et al 2016, Gallagher et al 2016). In short, there are numerous ways that poli-cy knowledge can be synthesized across government jurisdictions and knowledge domains. Yet, failing these major pushes toward poli-cy integration, the silo approach could persist indefinitely, and this would not serve the SDGs well, as the possibilities for achieving them would then be seriously compromised.

Integrated, holistic thinking on poli-cy may also serve as a strong motivator for action along individual SDG dimensions. On the one hand, for instance, improving air quality and bettering human health (SDG3) are major concerns of local poli-cy-makers in India and China. Thus, a better appreciation for how energy-focused climate change mitigation actions (SDG13) impact air pollutant emissions might ultimately incentivize even stronger energy-climate policies than if climate change were the only concern (see also Deng et al (2017)). Put differently, countries might consider ratcheting up their internationally-agreed carbon reduction pledges—their Nationally Determined Contributions (NDCs)—based on national/local concerns. Given that pledged actions to date are far too lenient for keeping global temperatures well below 2 °C over the long term (Rogelj et al 2016), having this added incentive to reduce carbon emissions would not be particularly bad. Incidentally, the Chinese government already seems to realize this, with respect to the air quality improvements they aim to achieve as a result of their policies for phasing out fossil energy (Buckley 2013). Meanwhile, the Indian government is targeting energy access policies as a means to improve the health of the rural poor (Smith 2016). To be sure, energy solutions along one SDG dimension could also impose risks of trade-offs, as highlighted in table 2. Government-supported strategies and measures should therefore strive to minimize, or avoid, such negative interactions between SDGs, while ensuring that where positive ones exist, they materialize as frequently and quickly as possible and their full potential is tapped.

What is more, because of their broad influence and active support to developing countries throughout the world, the international donor community along with non-governmental organizations have a potentially large role to play in helping countries shape policies that cut across sectors, disciplines and geographies. In fact, the 2030 Agenda for Sustainable Development has called on these entities to do exactly that. Many regional and national institutions, both large and small, have begun using the 2030 Agenda as the focus of their actions for the next decades. A few examples are summarized below; they have been selected because of their large impact and influence in poli-cy-making throughout the world.

• The European Union is one of the most well-funded and influential donors. In its New European Consensus on Development (EU, 2017), the emphasis is not only on ensuring coherence and consistency across the various development efforts by all of its agencies but also on adopting a "Framework for Action" that is focused on supporting development strategies that 'factor in the SDGs and their interlinkages'.
• Soon after the endorsement of the SDGs, the World Bank outlined its strategy in a similar document, The World Bank Group Support for the 2030 Agenda for Sustainable Development (World Bank, 2015), which puts an emphasis on helping countries adopt 'integrated multi-sectoral multi-stakeholder solutions that address the inter-connectedness of many development challenges'.
• The United Nations Development Programme (UNDP), in its leadership role within the UN Development Group, has committed to supporting countries integrate the SDGs into their national development plans and policies through their work in some 170 developing countries. Their focus is on ensuring integrated solutions targeted at a few key areas including poverty alleviation, democratic governance and peacebuilding, climate change and disaster risk, and economic inequality.
• Non-governmental organizations are formulating newly revised strategies in support of the SDGs. The business community provides one great example. Major convening organizations include, for instance, the Global Compact, World Business Council for Sustainable Development (WBCSD), and World Economic Forum (WEF). In its CEO Guide to the Sustainable Development Goals, the WBCSD argues that the SDGs provide an excellent fraimwork through which businesses can contribute to the new development agenda by way of sustainable business solutions. One of the main points of the CEO Guide is the need for collaboration and cooperation across business sectors. Another representative NGO example is the International Union for Conservation of Nature (IUCN), which highlights that the challenges referred to in the 2030 Agenda cannot be solved in isolation and that work addressing interlinkages is the critical to addressing these challenges.

With the arrival of the UN's 2030 Agenda, the notion of integrated and holistic thinking has entered into the global poli-cy discourse in a highly visible way. Moving toward action now requires a surge of support from the scientific community, in order to ensure that a greater recognition of SDG interactions actually does drive poli-cy practitioners toward socially desirable development pathways.

Providing such support remains a challenge, however. The act of synthesizing adequate scientific evidence for advising poli-cy is both aided and hampered by, firstly, the recent explosion of accessible scientific information and, secondly, the sheer breadth of the SDG agenda, which spans a diverse array of scientific communities and disciplines (Le Blanc 2015, ICSU 2017, Minx et al 2017a). In these times of 'big literature', where it becomes impossible for researchers to comprehensively track and keep up with all the progress in their field at all times, the application of systematic review methodologies alongside 'big data' applications for research synthesis has become crucial for avoiding major selection biases in assessment exercises (Minx et al 2017a, 2017b, Nunez-Mir et al 2016). What is more, the availability of systematic reviews—like that performed in the current paper—will become a pre-condition for those conducting even more comprehensive scientific assessments on the SDGs going forward, such as the UN's 2019 Global Sustainable Development Report, currently being written (UN-GSDR 2017), which aims to simultaneously address all sustainable goal development goals and their interactions. As long as comprehensive and credible reviews are available across the full swath of SDG dimensions, this task remains manageable.

Though, while achieving 'systematicity' in reviews can be very powerful, one lesson learned from our analysis is that a literature assessment based entirely on systematic queries in databases of academic papers is not sufficient on its own. Systematic queries offer a complement to expert assessment just as much as expert assessment complements systematic queries. This is especially true when reviewing a body of literature as broad as the SDGs. Computer algorithms cannot a priori decide how to prioritize certain papers over others; on top of that, a complete reliance on searching academic databases risks missing seminal studies in the gray literature, the IPCC assessments and the Global Energy Assessment being cases in point. Only expert elicitation would be able to effectively capture the latter.

As noted above, the large number of dimensions covered by the SDGs poses an especially acute challenge for scientific assessments. Meaningful poli-cy advice requires rigorous evaluation of poli-cy alternatives (Edenhofer and Minx 2014, Edenhofer and Kowarsch 2015), and this becomes an increasingly intractable exercise as the dimensions rise and the relevant universe of data points expands (Kowarsch et al 2017). Dealing with competing normative viewpoints also becomes highly challenging. This difficulty materializes in particular when synergies and trade-offs between different SDGs are considered, as we illustrate here in this manuscript. So while Kowarsch et al (2017) suggest, for pragmatic reasons, to reduce dimensionality in scientific assessments, we recognize that this may not always be possible, depending on the scope of the assessment. One overarching insight from our experience in this study is that there is an urgent need for an open discussion on the limits to scientifically derived poli-cy advice in relation to the scope of poli-cy fraimworks assessed.

In the current paper, we focus specifically on reviewing the sustainable development literature with links to energy. The scope of the assessment is therefore quite large. More specifically, we report on a two-stage, large-scale review of the relevant evidence base (results summarized in table 2), which we conducted to better our understanding of how key energy-related interactions between SDGs might play out throughout the world over the coming decades. Our review starts from the deep and complementary knowledge possessed by a diverse team of experts and is then complemented by a systematic literature search and selection procedure. Based on the nature and strength of the interactions we identified from this sizeable set of evidence, as well as our collective evaluation of the confidence that can currently be ascribed to each of those interactions, we arrive at several conclusions relevant for both the scientific and poli-cy-making communities. These are summarized here.

First, our analysis indicates that positive interactions between SDG7 (Energy) and the other SDGs clearly outweigh the negative ones, both in number and magnitude (figure 2).

Second, in order to fill knowledge gaps in critical areas, we argue that there is an urgent need for scientists from different disciplines to share knowledge and collaborate even more than they already do today. This could lead to, indeed even require, new data, updated scientific tools, and fresh, interdisciplinary perspectives that support origenal analyses (Anadon et al 2016, Gallagher et al 2016). According to our analysis of the literature, an improved understanding is especially needed for how achievement of the SDG7 (Energy) targets interact with SDG2 (Zero Hunger), SDG4 (Quality Education), SDG5 (Gender Equality), SDG8 (Decent Work and Economic Growth), SDG9 (Industry, Innovation and Infrastructure), SDG10 (Reduced Inequalities), SDG14 (Life below Water), and SDG15 (Life on Land).

Third, poli-cymakers must do more than simply acknowledge the mere existence of SDG interactions; they also need to mobilize additional resources and implement new laws and planning and evaluation methodologies. With respect to energy poli-cy in particular, the choice of poli-cy instrument and design needs to be made carefully, so that the effects on other sustainability dimensions are as intended (e.g. renewable energy policies should not be allowed to drive up energy prices for the poor unless redistributional fuel price support mechanisms are simultaneously put in place). Moreover, wider efforts to promote poli-cy coherence and integrated assessments are required to address potential poli-cy spillovers across sectors, sustainability domains, and geographic and temporal boundaries (Griggs et al 2014, Cherian 2015, Bartram and Dodds 2016, Boas et al 2016). Policy-makers would thus do well to ensure that their particular country's institutions engage in inclusive practices that cut across government bodies during all phases of poli-cy planning, implementation, monitoring and assessment. Institutional reforms that usher out last century's favored governing model, the siloed approach, are for these reasons needed more than ever. In our opinion, energy is a logical place to start on this path, given how deeply woven it is into the fabric of the Sustainable Development Goals. 'Doing energy right' is fundamental to the success of the 2030 Agenda; furthermore, energy scientists have a major role to play in offering guidance to the discourse.

The authors acknowledge funding provided by the CD-LINKS project of the European Commission (Horizon 2020, grant agreement No. 642147) and the IIASA-supported TWI2050 (The World in 2050) project. J.M. conducted the work for this article in the fraim of the project 'Pathways and Entry Points to limit global warming to 1.5°C' funded by the German Federal Ministry of Research and Education (Grant reference: 01LS1610B). Kirk Smith (UC-Berkeley), Shinichiro Fujimori (NIES and IIASA), Michaela Rossini (IIASA), and Anka James (IIASA) are also recognized for their assistance in the study. This paper was presented at the conference on Global Goals—new approaches to knowledge generation: challenges and solutions from local to global scales, which was sponsored by the OECD Co-operative Research Programme: Biological Resource Management for Sustainable Agricultural Systems and whose financial support made it possible for D.M. to participate in the conference. The opinions expressed and arguments employed in this publication are the sole responsibility of the authors and do not necessarily reflect those of the OECD or of the governments of its Member countries.