Bridging nestedness and economic complexity in multilayer world trade networks

International trade flows can be represented as complex networks capturing interactions among many economic agents. Prior work modeled world trade with either monopartite product-space or bipartite country–product networks to quantify economic complexity. A robust empirical feature of these systems is hierarchical nestedness, widely studied in ecological and socioeconomic networks. However, standard bipartite representations neglect to whom products are shipped, potentially missing key structure. Multilayer network modeling—where nodes are countries and each layer corresponds to a product’s trade network—retains commodity-specific directionality and reveals within-layer nestedness. In this work, the authors focus on within-layer nestedness of the multilayer world trade network and show that the nestedness of a product’s trading layer offers a meaningful measure of product complexity. High-complexity products entail technology barriers and are exported by advanced countries to many destinations, generating nested, asymmetric trade patterns. The study uses nestedness to (i) quantify product complexity and its evolution, (ii) assess countries’ trade competitiveness across products of differing nestedness (with a comparison between China and the United States), and (iii) build an aggregated nestedness-based index to predict nations’ future economic growth.

The paper situates itself in the economic complexity literature that uses network science to analyze world trade (e.g., ubiquity/diversity, ECI, and Fitness/Complexity indices). Nestedness—originally documented in ecological mutualistic networks—has been observed in economic systems and linked to stability and structural organization. Reviews (e.g., Mariani et al., 2019) survey nestedness across disciplines. Prior trade-network studies often used single-layer (monopartite or bipartite) models; multilayer approaches (commodity-specific directed layers) capture additional information on seller–buyer relations (Barigozzi et al., 2010; Mastrandrea et al., 2014; Alves et al., 2019). Methods for multilayer networks (community detection, cascading failures, centrality, reducibility) have advanced rapidly (Boccaletti et al., 2014). The authors build on these strands, proposing within-layer nestedness as an alternative, interpretable complexity metric and comparing it with established measures (ubiquity, ECI, FCI).

Data: International trade data from UN Comtrade and the MIT Atlas resources, covering 261 countries and 786 SITC product categories, years 1976–2014. Network construction: For each product p and year, construct a directed layer where nodes are countries and edges represent exports from country c (exporter) to country d (importer). The layer adjacency/value matrix M^p has elements v_cd^p equal to the dollar volume exported from c to d in product p. To define significant trade relations, compute a z-score per product p: z = (⟨v⟩ − v)/σ^p, where ⟨v⟩ and σ^p are the mean and standard deviation of v over all c→d flows for product p. Binarize edges: if z ≥ 0 set v = 1 (significant link), else v = 0. This yields a multilayer world trade network with a binary adjacency matrix per product layer. Nestedness metric (NODF): For each layer’s adjacency matrix (rows and columns sorted by decreasing out-degree and in-degree, respectively), compute Nested Overlap and Decreasing Fill (NODF). For a pair of nodes i and j with degrees k_i and k_j (k_i ≥ k_j), define pairwise nestedness N_ij = N_ij_common / min(k_i, k_j), where N_ij_common is the number of shared neighbors. Otherwise, N_ij = 0. The matrix nestedness η is the average over all eligible node pairs. η ∈ [0,1], with higher values indicating a more perfectly nested (triangular) structure. Illustrative examples and step-by-step calculation are provided in the Supplementary Information. Descriptive assessment: Compute η for each of the 786 product layers and analyze its distribution. Example products show a range of nestedness (e.g., Bovine ≈ 0.12; Typewriters ≈ 0.37; Medical Instruments ≈ 0.60). Products are grouped by SITC sector to compare average nestedness across sectors. Auxiliary measures used in results: Technology Achievement Index (TAI) to proxy countries’ technological capability; compute mean TAI of top-10 out-degree exporters per product layer. Network asymmetry Asy = (k_in − k_out)/N (k_in, k_out are node in/out-degrees; N is number of nodes) to assess directionality imbalance; examine its relationship with η across products. Temporal analyses: Track η over time (1986–2014) at multiple SITC aggregation levels (1- to 4-digit). Define a ranking-change indicator ν for the top-20 products at time t: ν = [Σ_i rank(i, t+Δt)] / [Σ_i rank(i, t)], evaluated for Δt ∈ [−10, 10], to compare growth vs. decay dynamics of nestedness rankings. Competitiveness and prediction: Define product-level trade competitiveness κ(c,p) for country c as the number of destination countries it exports to in product p’s layer (sum over row c). Construct a global trade competitiveness index GTC(c) = Σ_p κ(c,p) · η(p), weighting products by their nestedness. Normalize η and κ within-year as needed for comparative heatmaps and time trends. Correlate GTC with future macroeconomic indicators (GDP, GDP per capita, GDP(PPP), GDP(PPP) per capita) at horizons Δt up to 20 years and assess predictive power via linear regression.

• Layer nestedness as complexity: Product-layer nestedness η varies widely across 786 products. High-η layers correspond to complex, technology-intensive products; low-η layers to primary/raw materials. Example η values: Bovine ≈ 0.12; Typewriters ≈ 0.37; Medical Instruments ≈ 0.60.
• Technological capability: Mean TAI of top-10 out-degree exporters in a product layer increases with η, indicating that high-η products are exported primarily by high-technology countries.
• Asymmetry: Network asymmetry (Asy) is positively correlated with η across products; high-η (complex) products display stronger in/out-degree imbalances consistent with technology barriers and seller–buyer asymmetry.
• Sector rankings (SITC 1-digit): Machinery and chemicals-related sectors (8 Miscellaneous manufactured articles; 7 Machinery and transport equipment; 5 Chemicals; 6 Manufactured goods chiefly by materials) have higher mean η; raw-material sectors (3 Mineral fuels; 4 Animal and vegetable oils, fats and waxes; 2 Crude materials inedible; 0 Food and live animals; 1 Beverages and tobacco) have lower η; sector 9 (Unclassified commodities/transactions) is intermediate.
• Product lists: Top-10 highest-η products are sophisticated machinery/instruments (e.g., Analog instruments for physical analysis η≈0.664, X-Ray equipment, Medical instruments). Bottom-10 lowest-η products are primary commodities (e.g., Electric current, Uranium and thorium, Tin, Cotton seed oil).
• Temporal evolution: Across aggregation levels (1–4 digit), average nestedness tends to weaken over 1986–2014; however, the most nested products change little. The ranking-change measure ν shows asymmetric dynamics: increases in nestedness rankings occur faster than subsequent declines; average ν peaks for Δt > 0 and decays slowly for Δt < 0, indicating growth of product nestedness outpaces relevance decay.
• Country competitiveness (China vs USA): In 2010, China exhibits strong κ(c,p) for mid-η products (η≈0.35–0.5; e.g., footwear, motorcycles, electronics), while the USA excels both in low-η raw materials and high-η sophisticated products (η>0.5). Over 1976–2014, the USA maintains strong competitiveness, especially at low and high η; China’s competitiveness rises notably after 1990 and strongly after 2005, particularly for high-η products.
• Predictive power: The aggregated index GTC(c) correlates with future macroeconomic indicators, with higher correlations for GDP and GDP(PPP) than per-capita metrics, typically peaking around a 10-year horizon. Linear models using GTC predict GDP per capita growth reasonably well across countries; fast-growing economies (e.g., China, India) exceed predictions, while mature economies (e.g., USA, Japan, UK) tend to be overpredicted. Overall, results indicate GTC is a useful leading indicator of economic performance.

By modeling international trade as a multilayer network and quantifying within-layer nestedness, the study links a structural network property directly to product complexity. High nestedness reflects technology-based export asymmetries: a small set of technologically advanced countries export complex products broadly, while many countries import them, producing a triangular adjacency pattern. This framework refines insights beyond bipartite country–product models by incorporating destination information. The nestedness-based product ranking aligns with intuitive sectoral complexity and correlates with exporters’ technological capability, validating the measure. Temporal analyses suggest that while overall nestedness may weaken as technologies diffuse, the most complex products retain stability, and increases in nestedness tend to occur more quickly than their subsequent declines. At the country level, counting export destinations within product layers offers an interpretable competitiveness measure that, when aggregated and weighted by product nestedness (GTC), captures nations’ structural advantages. The China–USA comparison illustrates divergent specialization patterns and the evolution of capabilities. Finally, the significant correlations between GTC and future GDP metrics emphasize the relevance of nestedness-informed structure as a forward-looking indicator, complementing established economic complexity measures.

The paper bridges nestedness and economic complexity by representing world trade as a product-resolved multilayer network and using within-layer nestedness as an alternative measure of product complexity. This approach yields product rankings consistent with common understanding, connects high nestedness to technological barriers and asymmetric export structures, and provides interpretable country-level competitiveness metrics. The aggregated nestedness-weighted competitiveness index (GTC) shows notable predictive power for future economic performance, particularly over decadal horizons. Future research directions include uncovering the mechanisms that generate triangular (nested) structures in trade layers, designing improved nestedness metrics tailored to directed, weighted, multilayer data, and enhancing predictive models by incorporating additional factors or heterogeneous development paths. The framework may aid analyses of shocks and emerging trade barriers (e.g., geopolitical events, pandemics) by revealing structural vulnerabilities and resilience in global trade.

• Model simplifications: Links are binarized using a product-specific z-score threshold, potentially discarding informative intensity variations in trade volumes.
• Metric choice: The study employs a commonly used nestedness metric (NODF); alternative or improved measures for directed, weighted, multilayer networks could refine results.
• Mechanisms: The work focuses on utilizing nestedness rather than explaining its emergence; mechanisms driving triangular structures remain open.
• Predictive scope: While GTC correlates with future GDP metrics, prediction errors vary (e.g., underestimation for fast-growing economies, overestimation for mature economies). Stronger forecasts may require incorporating additional variables and differentiating development models.
• Generalizability over time: Observed weakening of nestedness and diffusion of technology may change relationships between η, competitiveness, and growth in future periods.