1. Introduction
Energy is returning to the centre of geography. Early in the revival of energy-and-society research, geographers were reminded that energy is an “ancient discipline, modern concern”; a few years later, energy geography was described as a “fertile academic borderland”. The spatiality of energy transitions — how low-carbon technologies take root in particular places, how fossil-fuel systems decline in particular regions, and how the costs and benefits of transition are distributed across space — has become a central concern of energy geography.

Yet a flourishing literature does not automatically yield better explanations. Methodologically, energy economic geography is markedly fragmented: social network analysis maps the topology of energy trade and infrastructure; global production network theory traces value, power and embeddedness; political economy dissects the social relations and materiality that underpin energy; and transition studies chart the co-evolution of technologies and institutions. Each of these approaches produces locally valid knowledge, but few confront head-on the methodological question of how explanations are actually constructed. Here we adopt a “methods-positive” stance: rather than arguing that one method is superior to another, we offer a constructive methodological programme for one particular approach — network methods. Our central claim is that networks can be turned from descriptive tools and loose metaphors into genuine “machines of explanation” only through three methodological operations.

2. Three uses of “network” and their conflation
In economic geography, “network” exists in at least three guises. The first is network as metaphor: since the relational turn, the network has served as a general image for expressing the embeddedness and situatedness of economic action, although critics note that it may offer a “partial understanding” rather than a new paradigm. The second is network as model: social network analysis provides a suite of computable topological metrics that have been widely transplanted into studies of energy trade flows, technological collaboration and electricity infrastructure. The third is network as method: global production network (GPN) analysis treats networks as an analytical framework for organising empirical inquiry, tracing firm–region relations through the categories of value, power and embeddedness, and, in its GPN 2.0 incarnation, has developed into a dynamic theory encompassing actors, strategies and causal mechanisms.

The problem lies not in the coexistence of these three uses, but in their frequent and undeclared conflation. When metaphor is taken for model, and model for explanation, the symptoms of “fuzzy concepts” appear: topological patterns are read off directly as causal mechanisms, and centrality metrics are interpreted directly as power. In energy research, the cost of such conflation is especially high, because energy networks possess an irreducible materiality: electricity grids obey the laws of physics rather than the logics of interpersonal trust; oil and gas pipelines are doubly locked in by geographical conditions and geopolitics; and fossil-fuel systems are deeply implicated in the production of capitalist space. A network diagram driven by metaphor alone cannot distinguish physical constraint from social construction.

3. Building a triplex network of energy, value and institutions
The first operation is a reconstruction of the object of analysis. We model the empirical object of energy economic geography explicitly as a triplex network: an energy-flow layer records the physical movements of fuels, electricity and embodied energy; a value-flow layer records the distribution of capital, profit and cost along production networks; and an institutional-flow layer records the transmission of policies, regulations, standards and discourses across scales. The three layers share some nodes (firms, cities and states), but the meaning of an “edge” is independent in each layer.

The genuine explanatory focus of the triplex network lies in inter-layer coupling. Change in the energy-economic-geographical landscape rarely occurs within a single layer; it more often appears as the rupture and reconfiguration of couplings between layers. In the global photovoltaic industry, for example, the shift of value networks towards East Asia has not been matched by a synchronous shift of energy-consumption networks, and this “decoupling” between the two layers is itself a key clue for explaining differences in regional development. Likewise, as carbon-pricing institutions diffuse through institutional networks, they reshape the cost structures of value networks via inter-layer coupling, and in turn redirect the material flows of energy networks. This yields a clear and falsifiable research framework: to explain change in energy landscapes is to explain when, where and why inter-layer couplings break down and are rebuilt.

The methodological significance of this operation is that it aligns the materiality and the sociality of energy within a single explanatory framework, rather than forcing researchers to choose between them. The concept of inter-layer coupling simultaneously disciplines two kinds of overreach: a pure physicalism that reads energy landscapes as a direct function of resource endowments, and a pure social constructivism that reads material networks as arbitrary products of discourse. The triplex network requires any explanation to pass the test of evidence from all three layers.

4. Networked process tracing
The second operation concerns the organisation of the explanatory chain. Metrics of network morphology — centrality, modularity and community structure — yield only the explanandum, not the explanans. To treat “a country’s rising centrality in the energy trade network” as the end point of explanation is to answer the map’s question with the map itself.

We therefore define and develop “networked process tracing”: significant changes in network morphology — breakpoints, community reconfiguration and abrupt shifts in centrality — are treated as empirical clues from which to trace backwards the concrete mechanisms that produced them. The procedure comprises four steps: (1) identify morphological anomalies in the triplex network; (2) formulate competing mechanistic hypotheses, such as strategic coupling and decoupling, institutional change, geopolitical manoeuvring and technological shocks; (3) test the hypotheses one by one against archival, interview and firm-level evidence, seeking the “smoking-gun” evidence that discriminates between rival explanations; and (4) write the test results back into the network model, updating the understanding of node attributes and edge meanings, and begin a new round of analysis.

It should be stressed that this is not an eclectic “mixed-methods” patchwork, but an explicit division of labour along a single explanatory chain: quantitative morphology answers where and when change occurred, while qualitative process tracing answers how and why it occurred. The two kinds of evidence perform different functions within the same explanation, and the way they are combined is itself defensible, testable and reproducible.

5. Abductive iteration and the symmetry principle of explanation
The third operation concerns the philosophy of explanation. Debates in economic geography about structure and agency often presuppose that one side enjoys natural explanatory priority. We propose a “symmetry principle of explanation”: structural patterns and actor practices serve as testing conditions for each other’s explanations, and the explanatory claims of either side must be open to challenge by the evidence of the other. Morphology without mechanistic support is empty correlation; mechanistic narratives that cannot account for observed macro-level patterns are merely ad hoc stories.

The symmetry principle operates through abductive iteration: starting from an unexpected network morphology, we infer the mechanism that would best explain it; the mechanistic hypothesis is then taken back to the data and tested against structural evidence; after revision, inference begins again. This logic resonates the value of models lies not in replicating reality but in forcing conceptual clarification, and forth movement between theory and empirics in the construction of GPN 2.0. The symmetry principle also answers the conceptual clarity: how nodes are defined, what flows constitute edges, and how inter-layer coupling is measured must all be stated explicitly in every study. Conceptual clarity is not a methodological fastidiousness but the precondition for symmetrical testing — fuzzy terms make falsification impossible.

6. Conclusion
The core argument of this paper is not intended to compete with other explanatory traditions in energy economic geography. On the contrary, the triplex network functions more like a “dialogic infrastructure”: political economy can interrogate the social relations of value appropriation at the institutional-flow layer; energy-justice research can trace the spatial distribution of transition costs along inter-layer couplings; transition geography can analyse the material rhythms of technological substitution at the energy-flow layer; and political-economic analyses of “energy spaces” can supply theoretical language for inter-layer coupling. What networks provide is not a unified theory but a shared grammar that allows the conclusions of different explanatory traditions to be checked against, and articulated with, one another.

For a heterodox community such as economic geography, the difficulty of methodological dialogue often lies not in substantive disagreement but in the absence of common forms that would make disagreement discussable. Building networks into “machines of explanation” is precisely about moving disagreement from the level of metaphor — where positions merely collide — to the level of evidence and mechanism, where disagreement can be adjudicated. This is perhaps the most positive contribution that network methods can make to the collective enterprise of constructing explanations in economic geography.