Defence Readiness: how to measure it across industrial ecosystems and how it has been impacted by the War in Ukraine

Defence Readiness: how to measure it across industrial ecosystems and how it has been impacted by the War in Ukraine | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Defence Readiness: how to measure it across industrial ecosystems and how it has been impacted by the War in Ukraine Víctor Llaneza, André Carrascal-Incera, Esteban Fernández Vázquez This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8785807/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract This paper studies defence‑related industrial ecosystems defining an indicator that tries to capture the “defence readiness” of the industrial specialization of economies, focusing on how the war in Ukraine has affected this indicator for some specific countries. This index builds on a Category Space framework that adapts Hidalgo et al. (2007) product space methodology. The approach proposed combines semantic similarity measures with Revealed Comparative Advantage (RCA) data to quantify the relatedness between Munitions List (ML) categories and Harmonized System (HS) product codes, allowing the assessment of how national productive structures connect to defence‑related goods. The resulting networks capture the density and composition of each country’s defence‑related industrial ecosystem, showing how ready is each economy to adapt to an increase in military spending. The analysis presented here focuses on the evolution of this indicator around the world between 2021, just before the starting of the war, and 2024, where the most recent data are available. The findings of this paper show how among the countries directly involved in the war, Ukraine presents a modest expansion of its network, with new HS codes emerging in 2024, including items associated with munitions. On the other hand, Russia and Belarus show a contraction in their networks over the period. At the same time, the analysis reveals how the United States, and particularly China maintained and even increased dense and diversified networks before and after the outbreak of the war. These results provide a representation of defence‑related industrial capacity and illustrate how the underlying productive structures evolve during the period of the war in Ukraine. This novel Category Space framework and the definition of this defence readiness indicator offer new tools for comparing how different economies are positioned to respond to shifts in defence demands. Earth and environmental sciences/Environmental social sciences Physical sciences/Mathematics and computing Defence Readiness Revealed Comparative Advantage Defence Industry Semantic Similarity 1. Introduction The war in Ukraine has renewed attention on the industrial foundations of defence capability in Europe and beyond. The outbreak of the war in 2022 exposed significant differences in the ability of countries to supply military equipment, sustain production over time, and adapt their industrial structures to an increase in defence demand. These differences have become central to debates on European security, strategic autonomy, and the resilience of defence supply chains. This paper contributes to these debates by developing a method that quantifies defence‑related industrial ecosystems and applies it to the context of the Russia–Ukraine war. One visible response to this and other recent conflicts has been the rise in defence expenditure across the European Union (EU) member states. Figure 1 shows total defence expenditure as a percentage of GDP in 2014 and 2024. The data from the European Defence Agency reveal substantial spatial heterogeneity. In 2024, countries such as Ireland, Malta, Luxembourg and Austria allocated less than 1% of GDP to defence, while Poland, Estonia, Lithuania and Latvia exceeded 3%. Despite this variation, as can be also seen in Fig. 1 , most EU members increased their defence spending notably between 2014 and 2024, reflecting a clear reassessment of security priorities in the last decade. However, defence preparedness depends not only on budgetary allocations but also on the underlying industrial capacity required to produce, maintain and supply military equipment. Figure 1 . Source: Own elaboration based on data from the European Defence Agency (2025). This response has also been accompanied by the launch of new EU policy initiatives aimed at strengthening Europe’s defence industrial capacity. The European Commission’s “White Paper for European Defence – Readiness 2030”, presented in March 2025, plans a programme of up to €800 billion in defence‑related investment over four years (European Commission, 2025). A central objective of this strategy is to reinforce the European Defence Technological and Industrial Base (EDTIB). Between the start of the war and June 2023, only 22% of EU member states’ defence acquisitions were sourced within the EU (European Commission, 2024). To address this dependence, the strategy encourages member states to increase the share of defence procurement within the EU to at least 50% by 2030 and 60% by 2035 (European Commission, 2024). Similar initiatives have been adopted in the United States, Australia and the United Kingdom, where governments have launched programmes aimed at restoring or expanding domestic defence industrial capacity (Matthews et al., 2025). Understanding this industrial dimension is essential because defence production relies on complex ecosystems that extend beyond strictly military goods. Many components, materials and technologies used in defence originate in non-military sectors. The European Commission defines an industrial ecosystem as the set of actors operating along a value chain, including firms of different sizes, research organisations, suppliers and service providers (European Commission, 2020). These interdependencies make difficult to assess the defence capability of a country demands by using direct measures alone (such as defence expenditure as a percentage of GDP). In that sense, the method proposed adopts this industrial ecosystem perspective. Research on defence spending has traditionally examined its regional and sectoral effects, with a particular focus on how changes in military budgets influence local economic outcomes. Studies have shown that increases or reductions in defence expenditure generate uneven spatial impacts, depending on the structure of regional economies and their capacity to absorb positive impacts or adapt to a new scenario (Atkinson, 1993; Bernauer et al., 2009). Other contributions have analysed the links between defence spending and high‑technology sectors, the role of defence‑related R&D, and the position of smaller EU countries within a common defence market (Boddy and Lovering, 1986; Mowery, 2012; Ploom et al., 2022; Wang et al., 2012). Input–output approaches have been used to assess the overall economic effects of reallocating military expenditure to other sectors such as education or health, finding that defence spending produces lower employment effects and may slow down the long-term development of a country (Leontief, 1986; Peltier, 2023; Stamegna et al., 2024). Recent work on the Russia–Ukraine war has examined its effects also on financial markets (Li et al., 2024), global food systems (Zhang et al., 2024), and scientific capacity in Ukraine (de Rassenfosse et al., 2023), showing the range of the war’s economic and societal consequences. Related research has also revealed that shocks associated with the Russia–Ukraine war propagate through production and trade networks in heterogeneous ways. Carrascal‑Incera et al. (2025) show that the subsequent disruptions in agriculture and industry due to the war generate uneven spillover effects across European regions, with central economies experiencing larger indirect impacts. These findings highlight the importance of understanding the productive structures that shape each country’s exposition to external shocks. Another strand of work regarding defence focuses on measuring it through the concept of Defence Industrial Base (DIB), stressing the idea that defence is not only the industries directly involved in military production but it should also account for more general items required by the defence industry (Dunne, 1995) such as vehicles, clothing or fuel, for example. Finally, the latest attempts to define what constitutes a military product can be found with the classification of the defence‑related production in the Wassenaar Arrangement (Wassenaar Arrangement, 2024) or the EU Common Military List (Council of the European Union, 2025). While this literature provides valuable insights into the economic effects of defence spending and the composition of defence industries, it does not offer a proper tool to measure the industrial ecosystems that support defence production in a way that can be comparable across countries. Existing approaches rely either on direct data on military output or on broad sectoral classifications that do not capture the specific technological relationships between non-military and military goods. As a result, there is a lack of a framework capable of quantifying the productive structures that support defence capacity and of analysing how these structures respond to major geopolitical shocks such as the war in Ukraine. The objective of this paper is to develop a method to define a defence readiness that is quantified for the defence‑related industrial ecosystems using publicly available trade data and to apply it to analyse how these ecosystems evolved during the war in Ukraine. The method adapts Hidalgo’s product space framework (Hidalgo et al., 2007) to the defence field by combining semantic similarity techniques with the notion of Revealed Comparative Advantage (RCA). This approach allows us to measure the degree of relatedness between military categories defined in the Wassenaar Munitions List (ML) and the general products classified under the Harmonized System (HS). The empirical analysis performed in this paper focuses on the period 2021–2024, which covers the years immediately before the start of the invasion of Ukraine and the most recent year with available data. By means of the category space framework proposed, we are able to construct country‑specific networks and an indicator of relatedness, comparing the results by countries and over the period analysed. In particular, we base the empirical application on two hypotheses: The first one (H1) is that countries geographically closer to the conflict may experience an expansion of their defence‑related industrial ecosystem. This follows the idea that these countries may experience higher demand for military equipment, stronger policy incentives to increase production, and adjustments in industrial activity with the objective of meeting short term needs. The second one (H2) is that countries with established industrial capabilities are more likely to act as suppliers of military equipment during a conflict. In this case, this reflects the path‑dependent and slow responding nature of industrial development. Defence production relies on a particular knowledge, on specialised inputs and on technological complementarities that cannot be built in the short term. Countries with dense and diversified industrial ecosystems are therefore better positioned to scale up output when demand increases. By testing these hypotheses, the paper tries to illustrate the usefulness of the category space method for (i) understanding how military items interact with the existing productive structures and (ii) quantifying how ready are the industrial capacity of countries to absorb a potential increase of military spending, overcoming the limitations of direct measures of military production and allowing for cross‑country and temporal comparisons. After this introduction section, the rest of the paper is organised as follows. Section 2 describes the methodology and data, explaining the construction of the category space approach, the steps to link military and non-military product classifications and the characteristics of the proposed defence readiness indicator. Section 3 presents the empirical results, examining the evolution of defence‑related industrial ecosystems between 2021 and 2024 and testing the two hypotheses established in this section. Section 4 concludes by summarising the main findings and outlining potential applications of the method for future research and policy analysis. 2. Methodology and data The limitations of the existing methods to adopt an industrial ecosystem approach when estimating defence‑related industrial capacity create a need for a new analytical tool. A tool that can also rely on publicly available data, since data on military production are often restricted, incomplete or not comparable across countries. This tool is what we call the category space method. 2.1 Product and Category Space The methodology applied in this paper builds on the idea capturing the density of the economic system around military spending. For this purpose, we develop a variant of the measure of relatedness proposed by Hidalgo (2007), where he introduces the concept of product space , a network in which the proximity between products reflects their relatedness. In that paper, the proximity between products \(\:\:i,p\:\) for a country \(\:c\) is measured using pairwise conditional probabilities, specifically assessing the likelihood that a country exports one good given that it already exports another. More formally, the indicator is defined as the probability for a country \(\:c\) of having a Revealed Comparative Advantage in product \(\:i\) ( \(\:{RCA}_{i}^{c}\) ) conditional on having Revealed Comparative Advantage in product \(\:p\) ( \(\:{RCA}_{p}^{c}\) ), or vice versa. The measure of proximity \(\:{\varphi\:}_{ip}^{c}\) proposed in Hidalgo (2007) is calculated as: $$\:{\varphi\:}_{ip}^{c}=\text{min}\left\{P\left({RCA}_{i}^{c}/{RCA}_{p}^{c}\right),P\left({RCA}_{p}^{c}/{RCA}_{i}^{c}\right)\right\}$$ 1 The definition of Revealed Comparative Advantage (RCA) for country c in product i is defined following Balassa (1965) as: $$\:{RCA}_{i}^{c}=\frac{{x}_{i}^{c}/\sum\:_{i}{x}_{i}^{c}}{\sum\:_{c}{x}_{i}^{c}/\sum\:_{c}\sum\:_{i}{x}_{i}^{c}}$$ 2 where \(\:{x}_{i}^{c}\) stands for the exports of country \(\:c\) of product \(\:i\) . The concept of RCA acts as a filter to construct the Product Space for an economy, considering only goods whose exports are significant (RCA > 1), which reflects their embedded productive capabilities. This measurement of economic proximity is essential for understanding which products an economy is likely to produce in the future and, therefore, for explaining the development of countries. Based on the same concept, although adapted to our particular problem, the study builds on the idea of category space . However, instead of relying on conditional probabilities of exporting products as the pieces for constructing this space, we apply semantic techniques that estimate the degree of relatedness between military ML categories and the description of products exported, following the HS coding system, which is the international standard product classification used to categorize traded goods. 2.2 Defining a defence readiness indicator The key concept used in this paper is the idea of semantic similarity between the descriptions of categories in the ML and HS lists. Semantic measures include semantic similarity (SS), which indicates shared features between elements, and semantic relatedness (SR), which covers all types of relationships in context (Zhu et al., 2020; Costa and Leal, 2016; Li, 2003). These concepts let researchers examine words and their underlying semantic connections, imitating human reasoning. Recent advances in semantic methods allow exploration of research areas previously inaccessible (Gentzkow et al., 2019; Ittoo and van den Bosch, 2016; Hoberg and Phillips, 2016). Hoberg and Phillips (2016) developed an industry classification using web-crawling algorithms to analyse over 50,000 firm annual reports, clustering firms by product description similarity. Vázquez (2020) evaluated the defence and health industries’ capacity for innovation, noting that analysing only 16 military product codes reveals the limitations of traditional sector analysis tools. In this paper, we address this limitation by applying semantic methods to construct a “ product space ” around ML categories. These methods allow us to broaden the consideration of military products by incorporating not only strictly military items but also those highly related to the sector. Rather than building networks by just considering products, as in Hidalgo and Hausmann (2009), we create category spaces centred on the ML categories and surrounded by related products (as described in Fig. 2 ). This approach is represented as a bipartite network with two sets of nodes, HS products and ML categories, where product nodes connect directly only to ML nodes (see Fig. 2 ). Figure 2 . Source: Own elaboration. To do so, we employ semantic techniques that estimate the degree of relatedness between ML categories and HS products at the 4-digit level. In particular, we use a sentence-embedding model. Specifically, we utilize all-MiniLM-L6-v2 , a compact pretrained sentence-embedding model designed to transform short texts into fixed-length numerical vectors that represent their semantic content. This model consists of a 6-layer MiniLM Transformer architecture, trained with contrastive learning on extensive collections of sentence pairs so that semantically similar texts produce closely aligned embeddings. In practical applications, it is accessed via the sentence-transformers library, facilitating efficient execution of tasks such as semantic similarity assessment, clustering, and information retrieval. This library is a Python module based on SBERT (Sentence Bidirectional Encoder Representations from Transformers) and is specifically designed to produce semantically meaningful sentence embeddings, which can be compared using cosine-similarity (Reimers and Gurevych, 2019). In other words, this tool enables the comparison of not only words but also sentences, by transforming text into numbers and incorporating the contextual meaning of the entire sentence. Specifically, we apply this approach to quantify the semantic relatedness between ML definitions and HS code descriptions, measuring the semantic distance between them and subsequently obtaining the category space . To quantify the relatedness between ML categories and HS product codes, we define a semantic similarity indicator S k,i based on the cosine similarity of their textual embeddings. Let D k ML denote the textual description of ML category k (where k = 1, 2, ..., 22) and let D i HS denote the textual description of HS product code i at the 4-digit level (where i = 1, 2, ..., n and n ≈ 1,200 headings in the HS 2012 nomenclature). Using the sentence-transformer model all-MiniLM-L6-v2 , each textual description is mapped into a dense vector representation in a high-dimensional embedding space ℝ d (where d = 384 for this model). We denote these embedding vectors as: $$\:{e}_{k}^{ML}\:=\:f\left({D}_{k}^{ML}\right)\:\in\:\:{\mathbb{R}}^{d}\:\text{a}\text{n}\text{d}\:{e}_{j}^{HS}\:=\:f\left({D}_{j}^{HS}\right)\:\in\:\:{\mathbb{R}}^{d}$$ 3 where f (·) represents the embedding function implemented by the sentence-transformer model, which captures the semantic meaning of the input text by encoding contextual relationships between words. The semantic similarity indicator S k,i between ML category k and HS product i is then computed as the cosine similarity between their respective embedding vectors: $$\:{S}_{k,i}=\text{c}\text{o}\text{s}\left({e}_{k}^{ML},{e}_{i}^{HS}\right)=\frac{{\sum\:}_{j=1}^{d}\left({e}_{k,j}^{ML}\right)\left({e}_{i,j}^{HS}\right)}{\sqrt{{\sum\:}_{j=1}^{d}{\left({e}_{k,j}^{ML}\right)}^{2}}\sqrt{{\sum\:}_{j=1}^{d}{\left({e}_{i,j}^{HS}\right)}^{2}}}$$ 4 In Eq. ( 4 ) \(\:{e}_{k}^{ML}\) and \(\:{e}_{i}^{HS}\) represent the \(\:j\) -th component of the embedding vectors for ML category \(\:k\) and HS product \(\:i\) , respectively. This indicator ranges between 0 and 1, with values close to 1 indicating a high similarity between \(\:k\) and \(\:i\) , while values near to 0 suggest that \(\:k\) and \(\:i\) are semantically disconnected. As a consequence, it is possible to evaluate the “density” of the category space for class \(\:k\) as the sum \(\:{\sum\:}_{i=1}^{n}{S}_{k,i}\) : large values of this sum identify categories in the ML with a dense network of products connected to them. Note, however, that the elements \(\:{S}_{k,i}\) are defined solely on the semantic relationship between the textual descriptions between the two classifications lists, therefore they are common for all the \(\:\:C\) countries under study. But the capabilities of the different economies for taking advantage of these connections between the products they produce and the categories of the ML list are expected to vary across countries. One way of capturing this heterogeneity is by defining a \(\:C\times\:n\) matrix with binary cells \(\:{b}_{i}^{c}\) defined as: $$\:{b}_{i}^{c}\:=\left\{\begin{array}{c}1\:\:if\:{\text{R}\text{C}\text{A}}_{\text{i}}^{\text{c}}>1\\\:0\:\:if\:{\text{R}\text{C}\text{A}}_{\text{i}}^{\text{c}}\le\:1\end{array}\right.$$ 5 We propose an indicator of defence readiness \(\:{\phi\:}^{c}\) , which captures the relatedness of the production capacity of country \(\:c\:\) around military products as: $$\:{\phi\:}^{c}=\frac{{\sum\:}_{k=1}^{K}{\sum\:}_{i=1}^{n}{S}_{k,i}{b}_{i}^{c}}{{\sum\:}_{k=1}^{K}{\sum\:}_{i=1}^{n}{S}_{k,i}}$$ 6 The indicator \(\:{\phi\:}^{c}\) is naturally bounded between 0 and 1, and relatively large values of \(\:{\phi\:}^{c}\) indicate that country \(\:c\) presents an economic specialization (measured through the production of goods with RCAs) that allows it to exploit the relatedness with the products in the categories of the ML list. Cases like that reveal countries with a good capacity to benefit from an expansion of military expenditure. 2.3 Data This research required the handling of public open-access data provided by several globally recognised organisations. A quick description of these sources can be found below: 1. The Wassenaar Arrangement This work uses the Wassenaar Arrangement on Export Controls for Conventional Arms and Dual-Use Goods and Technologies (Wassenaar Arrangement, 2024), in force since 1996, as the main reference to define which products should be considered defence-related. Services are not included. The Arrangement, signed by 42 participating states (Argentina, Australia, Austria, Belgium, Bulgaria, Canada, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, India, Ireland, Italy, Japan, Latvia, Lithuania, Luxembourg, Malta, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Republic of Korea, Romania, Russian Federation, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland, Türkiye, Ukraine, United Kingdom and United States), aims to contribute to regional and international security by establishing a common framework for the control of exports of items included in the List of Dual-Use Goods and Technologies and the Munitions List (Wassenaar Arrangement, 2024). Participating states commit to applying shared guidelines, elements and procedures as the basis for their national export control legislation. The Wassenaar Arrangement Control Lists consist of the following Munitions List, which includes 22 categories of items designed for military use. The codes and descriptions of each category used to feed the semantic-similarity algorithm are shown in Table 1 . Table 1 Wassenaar Arrangement 2024 - Munitions List – Items 1 to 22. Code Textual description ML1 Smooth-bore weapons with a calibre of less than 20 mm, other arms and automatic weapons with a calibre of 12,7 mm (calibre 0,50 inches) or less and accessories and specially designed components therefor. ML2 Smooth-bore weapons with a calibre of 20 mm or more, other weapons or armament with a calibre greater than 12,7 mm (calibre 0,50 inches), projectors specially designed or modified for military use and accessories and specially designed components therefor. ML3 Ammunition and fuze setting devices and specially designed components therefor. ML4 Bombs, torpedoes, rockets, missiles, other explosive devices and charges and related equipment and accessories and specially designed components therefor. ML5 Fire control, surveillance and warning equipment, and related systems, test and alignment and countermeasure equipment specially designed for military use, and specially designed components and accessories therefor. ML6 Ground vehicles and components. ML7 Chemical agents, "biological agents", "riot control agents", radioactive materials, related equipment, components and materials. ML8 "Energetic materials" and related substances. ML9 Vessels of war (surface or underwater), special naval equipment, accessories, components and other surface vessels. ML10 "Aircraft", "lighter-than-air vehicles", "Unmanned Aerial Vehicles" ("UAVs"), aero-engines, “sub-orbital craft” and "aircraft" equipment, related equipment, and components, specially designed or modified for military use. ML11 Electronic equipment, "spacecraft" and components, not specified elsewhere on the Munitions List. ML12 High velocity kinetic energy weapon systems and related equipment and specially designed components therefor. ML13 Armoured or protective equipment, constructions, components, and accessories. ML14 'Specialised equipment for military training' or for simulating military scenarios, simulators specially designed for training in the use of any firearm or weapon specified by ML1 or ML2, and specially designed components and accessories therefor. ML15 Imaging or countermeasure equipment, specially designed for military use, and specially designed components and accessories therefor. ML16 Forgings, castings and other unfinished products, specially designed for items specified by ML1 to ML4, ML6, ML9, ML10, ML12 or ML19. ML17 Miscellaneous equipment, materials and "libraries" and specially designed components therefor. ML18 'Production' equipment, environmental test facilities and components. ML19 Directed Energy Weapon (DEW) systems, related or countermeasure equipment and test models and specially designed components therefor. ML20 Cryogenic and "superconductive" equipment and specially designed components and accessories therefor. ML21 "Software". ML22 "Technology". 2. Products in the Harmonized System (HS) list The embeddings of the textual descriptions of the ML categories need to be mapped together with descriptions of products exported between countries. These descriptions have been extracted from the Harmonized System (HS), developed by the World Customs Organization (WCO) and used by over 200 countries, as the framework for product classification. Specifically, we use the HS 2012 Edition at the 4-digits level (HS heading), which covers more than 1,200 categories of goods (World Customs Organization, 2012). 1 Data on actual export flows by product following this classification have been obtained from the BACI ( Base Analytique du Commerce International ) database. This dataset “ provides data on bilateral trade flows for 200 countries at the product level (5.000 products), according to Harmonized System nomenclature ” (Gaulier and Zignago, 2010) and is collected and curated by the Centre d'Etudes Prospectives et d'Informations Internationales (CEPII). The raw exports and imports data come from the UN Comtrade database, provided by the United Nations Statistical Division. However, CEPII performs additional operations to improve data accuracy, considering FOB import values and taking into account the reliability of trade data reported by each country. Specifically, this work uses the HS12_V2026.01 version of the BACI database, updated in January 2026 (CEPII, 2026). 3. Results This section is organized as follows. First, we quantify and represent the semantic relatedness between Wassenaar Munitions List (ML) categories and Harmonized System (HS) codes by constructing a technological network that captures the relationships between categories and products. Second, based on this network and the procedure described above, we present the networks for two representative countries and a global choropleth map to illustrate how the approach reveals and quantifies the relatedness between military categories and industrial capabilities. Finally, focusing on the war in Ukraine, we present the networks of the countries involved for 2021 and 2024, highlighting the potential of the method to compare economies and analyse their evolution over time. 2.1. Technological Relatedness Network between ML categories and HS Codes Figure 3 displays the product space surrounding ML categories as a bipartite network. This graph illustrates the links between ML categories and HS Codes, considering only the top 1% of HS-ML connections with the highest relatedness scores. This filter shows the most relevant connections that structure the defence-related category space. Link thickness reflects the strength of semantic relatedness between HS products and ML categories. Figure 3 . Each ML category can be interpreted as a cluster within the technological network. Figure 4 presents this segmentation, based on the network filtered at the top 1% of relatedness values for the sake of clarity in the presentation of the graph. The figure shows substantial heterogeneity across ML categories: some of them are associated with a large number of related HS codes, such as ML8 (Explosives and Propellants), ML10 (Aircraft) and ML17 (Miscellaneous Equipment); in contrast, other categories, such as ML1 (Small Arms), ML2 (Weapons > 20mms) and ML12 (Kinetic Energy Weapons) are linked to relatively few HS codes. A third group of ML categories does not appear in the graph, because none of their associated HS codes fall within the top 1% of relatedness values. This group includes ML21 (Software) and ML22 (Technology). 2 Figure 4 . 2.2. The geographical map of defence readiness Once the technological relatedness network is obtained and following the criteria proposed by Hidalgo (2007) and data on bilateral trade flows provided by BACI (Gaulier G. a., 2010), it is possible to construct annual rankings that identify countries, regions or economic blocs, whose economies are most closely related to the defence sector. It is important to note that these rankings do not show which economies produce more military goods, but those whose industrial ecosystem is most closely related with the military industry. Figure 5 presents a global choropleth map in which countries are coloured according to their indicator of defence readiness values ( \(\:{\phi\:}^{c}\) ) in 2024. The map provides a synthetic representation of the capabilities of each country in defence-related products. Moreover, it reveals a clear spatial pattern, both for the top positions and for the bottom ones. Figure 5 . To facilitate the interpretation of the map, Table 2 reports the countries with higher \(\:{\phi\:}^{c}\) in 2024, that is, the economies that are theoretically best positioned to benefit from an expansion in military spending. Table 2 Top economies by defence readiness indicator \(\:{\varvec{\phi\:}}^{\varvec{c}}\) for the year 2024. Country \(\:{\phi\:}^{c}\) China 0.4210 Germany 0.3993 Italy 0.3617 Spain 0.3157 Austria 0.3138 Turkey 0.3083 Poland 0.3004 Czechia 0.2846 France 0.2837 United Kingdom 0.2638 United States 0.2634 Japan 0.2609 Portugal 0.2508 India 0.2475 Croatia 0.2430 To further illustrate the explanatory potential of the method, Fig. 6 presents, as networks, the defence-related category space of two countries (Germany and Ireland) for 2024. These countries are chosen as examples due to their disparity. They differ markedly in defence expenditure as a share of GDP (2,1% in Germany and 0,2% in Ireland) and in their productive structures, with Germany representing a highly industrialised economy and Ireland a more service-oriented one. This comparison highlights how the network representation helps to understand the industrial ecosystem related to defence of each economy: Figure 6 . The German network shows a large number of nodes with strong interconnections, whereas the Irish network is considerably smaller, with fewer nodes and more limited connections between them. Accordingly, the network structure clearly suggests that the German economy is more closely related to the defence sector than the Irish economy. 2.3. The War in Ukraine: analysis of the economies involved Finally, we analyse the relatedness networks of the countries directly involved in the war in Ukraine in order to characterise their defence-related industrial ecosystems and their evolution over the course of the conflict. To this end, Fig. 7 presents the networks of Ukraine, Russia and Belarus before the war (2021) and in 2024, two years after the official start of the conflict. Figure 7 . The networks observed in 2021 are relatively similar across the three countries and are characteristic of economies with limited industrial capabilities related to the defence sector. Although Belarus displays a denser network, none of the three resembles the structure observed for a highly industrialised economy such as Germany (Fig. 6 ). In 2024, the Ukrainian network remains relatively sparse, although several HS codes that were absent in 2021 now appear in the network. These changes may reflect adjustments in industrial capabilities during the conflict. One illustrative example is the emergence of HS 9306, related to bombs, grenades, and other munitions, as a key node in the 2024 network. By contrast, the networks of Russia and Belarus exhibit a marked contraction over the same period. In 2024 both countries display fewer HS codes and ML categories than in 2021, indicating a substantial reduction in the density of their defence-related networks since de beginning of the conflict. To complete the analysis, we also include the networks of EU27 + UK, China and USA, economies not directly involved in the conflict but playing a significant role. As shown in Fig. 8 , these networks are relatively similar to each other and follow the structure observed in the German network, with a large number of HS nodes connected to many ML categories. This pattern reflects highly industrialised economies whose productive structure is closely linked to the defence industry. Furthermore, no substantial changes are observed over the 2021–2024 period. Figure 8 . Table 3 reports the corresponding \(\:{\phi\:}^{c}\) values for the economies involved in the war. This indicator allows us to summarize the previous networks in a single figure and to rank these economies according to the characteristics of their defence-related industrial ecosystem: Table 3 Economies involved in the war in Ukraine by \(\:{\varvec{\phi\:}}^{\varvec{c}}\) . 2021 2024 EU27 + UK 0.4786 0.4740 China 0.3929 0.4210 USA 0.2559 0.2634 Ukraine 0.1236 0.1245 Belarus 0.1850 0.1233 Russia 0.1028 0.0778 These figures confirm that Ukraine, Belarus and Russia exhibit economies weakly related to the defence sector, whereas the USA, China and the EU27 + UK are more strongly connected, particularly China and EU27 + UK.The outbreak of the war in 2022 further exacerbated these dynamics, deepening the erosion of industrial diversity and narrowing the range of goods in which Russia and Belarus remain competitive on global markets. By contrast, Ukraine, as well as the EU27 + UK, maintains networks similar to those observed before the war. The two economies that can be classified as “winners” in this process has been USA and, specially, China, which show respectively a modest and notable expansion on their network over the same period. 3. Conclusions and discussion This paper proposes the study of the category space framework as a method to study defence‑related industrial ecosystems and to define an indicator of defence readiness. By combining semantic similarity measures between lists of military products and exported goods, and including data of exports indicating the presence or not of Revealed Comparative Advantage (RCA), the approach constructs networks that capture how each country’s productive structure relates to the categories considered in the Wassenaar Munitions List (ML). The defence readiness indicator summarises the density of these networks into a single comparable figure, measuring how well positioned an economy is to absorb from an expansion of military spending. The underlying idea is that defence preparedness cannot be adequately assessed through budgetary allocations alone, as the underlying industrial ecosystem plays a fundamental role in determining whether increased spending can be translated into effective military capability. The alternative proposal presented here offers an alternative to conventional measures of defence capacity that rely on expenditure data or narrow sectoral classifications, and provides an easily replicable, data‑driven tool for cross‑country and temporal comparisons. Besides the theoretical definition of the indicator proposed, this paper calculates it for several years basing on recent data of trade flows. The global mapping of defence readiness presented in this paper reveals a clear spatial pattern. Highly industrialised economies such as China and some countries in Western Europe occupy the top positions, reflecting productive structures densely connected to the defence domain. The empirical application to the war in Ukraine comparing the years 2021 and 2024 has yielded several findings. With respect to the first hypothesis (H1), which posited that countries geographically closer to the conflict would experience an expansion of their defence‑related industrial ecosystem, the evidence is mixed. Ukraine exhibits a modest expansion of its network. This result is consistent with the expected adjustment of productive structures under conditions of increased defence demand. However, Russia and Belarus display a marked contraction of their networks over the same period, with fewer HS codes and ML categories connected in 2024 than in 2021. The direction of involvement in the war, combined with the effects of international sanctions, appears to be a decisive moderating factor: while Ukraine’s productive structure adapted to wartime needs despite, the sanctions regime eroded the industrial diversity of Russia and Belarus in global markets, reducing the range of goods in which these countries retain a revealed comparative advantage and, consequently, lowering their defence readiness indicator. The evidence for the second hypothesis (H2), which proposed that countries with established industrial capabilities are more likely to act as suppliers of military equipment during a conflict, is considerably strong. The EU27 + UK aggregate, China and the United States all maintain dense and diversified category space networks in both 2021 and 2024, consistent with the path‑dependent nature of industrial development stressed in the economic complexity literature. The structural stability of these networks over a period of significant geopolitical disruption confirms that defence readiness cannot be built in the short term but instead reflects accumulated productive capabilities, specialised knowledge and technological complementarities developed over decades. Within this group of established economies, the analysis reveals an asymmetric evolution that is relevant for understanding the geopolitical reconfiguration of defence‑related trade. China exhibits the most notable expansion of its defence readiness indicator, suggesting a deepening of the connections between its export structure and the defence sphere. The United States registers a modest increase, while the EU27 + UK remains essentially stable with a slight decline. These dynamics point to China as the principal beneficiary of the reconfiguration of defence‑related trade networks triggered by the war. From a policy perspective, the results carry important implications for the European debate on defence readiness. The EU27 + UK aggregate records the highest defence readiness value among all the economies analysed, indicating that Europe’s productive structure is, on average, more densely connected to the activities linked to defence than any other major bloc. Yet, as noted in the introduction, only 22% of EU defence acquisitions between the start of the war and June 2023 were produced within the EU. This gap suggests that the challenge for European defence policy might lie in mobilising and coordinating the productive capabilities that already exist across member states. The methodology also opens several avenues for future research. The methodology presented here can be applied at the sub‑national level by exploiting regional trade or production data to identify which regions within a country possess the productive structures most closely related to the defence sector. It can also be extended to other conflict settings or to monitor changes in defence readiness over longer time horizons. Moreover, the bipartite network structure of the approach allows for analysis from the perspective of HS codes rather than ML categories, identifying those commercial products that serve as bridges across multiple military domains and that may therefore represent strategic bottlenecks or opportunities for industrial policy. The study also has limitations that should be acknowledged. Semantic similarity captures only potential relatedness between product descriptions rather than actual production flows, and the strength of the analysis depends on the quality and comprehensiveness of the textual descriptions available in the classifications analysed. Similarly, relying on international trade data, might understate the defence readiness of countries whose defence‑related production is primarily oriented towards domestic consumption or whose trade data are incomplete or distorted, as may be the case for sanctioned economies. Finally, the 2021–2024 window captures only the early dynamics of the war, and the structural adjustments observed may not fully reflect longer‑term transformations in defence‑related industrial ecosystems that are still underway. Despite these limitations, the methodology presented here helps to address a question that has traditionally been difficult to measure: how ready are productive structures to respond to shifts in defence demands? By providing a replicable, data‑driven and comparable method, this paper contributes with a new tool for understanding the industrial foundations of defence capability and for analysing how these foundations respond to major geopolitical shocks. Declarations Author Contribution Author ContributionsVLL: Conceptualization, methodology, data curation, formal analysis, writing – original draft.AC: Methodology, data analysis, validation, writing – review and editing.­EF: Conceptualization, supervision, interpretation of results, writing – review and editing.All authors have read and approved the final manuscript. Acknowledgement The authors acknowledge the financial support of the Cátedra Sekuens de la Innovación en Asturias (CATI-25-010). Data Availability All data used in this study are publicly available:- Trade data at the product level were obtained from the BACI database compiled by CEPII ( [link](https:/www.cepii.fr/CEPII/en/bdd_modele/bdd_modele_item.asp?id=37) ), based on UN Comtrade data. Product classifications follow the Harmonized System (HS) 2012 nomenclature published by the World Customs Organization ( [link](https:/www.wcoomd.org/en/topics/nomenclature/instrument-and-tools/hs_nomenclature_previous_editions/hs_nomenclature_table_2012.aspx) ).- The definitions of defence-related categories are based on the Wassenaar Arrangement Munitions List (2024 edition) ( [link](http:/www.wassenaar.org/app/uploads/2024/12/List-of-Dual-Use-Goods-and-Technologies-and-ML-2024.pdf) ).All data sources are cited in the manuscript. The processed datasets and code used to construct the defence readiness indicator are available from the corresponding author upon reasonable request. References Atkinson, R. D. (1993). Defense spending cuts and regional economic impact: An overview. Economic Geography , 69(2), 107–122. Balassa, B. (1965). Trade liberalisation and “revealed” comparative advantage. The Manchester School , 33(2), 99–123. Bernauer, T., Koubi, V., & Ernst, F. (2009). National and regional economic consequences of Swiss defense spending. Journal of Peace Research , 46(4), 467–484. Boddy, M., & Lovering, J. (1986). High technology industry in the Bristol sub-region: the aerospace/defence nexus. Regional studies , 20(3), 217–231. Carrascal-Incera, A., Liu, W., Orea, L., & Sickles, R. C. (2025). Beyond borders: how spillovers and commercial networks shape European productivity. Journal of Productivity Analysis , 1–49. CEPII (2026). BACI, Base pour l’Analyse du Commerce International . www.cepii.fr/CEPII/en/bdd_modele/bdd_modele_item.asp?id=37 Costa, T., & Leal, J. P. (2016, May). Semantic measures: How similar? How related?. In International Conference on Web Engineering (pp. 431–438). Cham: Springer International Publishing. Council of the European Union (2025). Common military list of the European Union adopted by the Council on 24 February 2025. Official Journal of the European Union , C(1499). De Rassenfosse, G., Murovana, T., & Uhlbach, W. H. (2023). The effects of war on Ukrainian research. Humanities and Social Sciences Communications , 10(1), 1–11. Dunne, J. P. (1995). The defense industrial base. Handbook of defense economics , 1, 399–430. European Commission (2020). A New Industrial Strategy for Europe . European Commission (2024). A New European Defence Industrial Strategy: Achieving EU Readiness Through a Responsive and Resilient European Defence Industry . European Commission (2025). White Paper for European Defence – Readiness 2030 . European Defence Agency (2025). Defence Data 2024–2025 . Gaulier, G., & Zignago, S. (2010). Baci: international trade database at the product-level (the 1994–2007 version). CEPII Working Paper. Gentzkow, M., Kelly, B., & Taddy, M. (2019). Text as data. Journal of Economic Literature , 57(3), 535–574. Hidalgo, C. A., & Hausmann, R. (2009). The building blocks of economic complexity. Proceedings of the national academy of sciences , 106(26), 10570–10575. Hidalgo, C. A., Klinger, B., Barabási, A. L., & Hausmann, R. (2007). The product space conditions the development of nations. Science , 317(5837), 482–487. Hoberg, G., & Phillips, G. (2016). Text-based network industries and endogenous product differentiation. Journal of political economy , 124(5), 1423–1465. Ittoo, A., & van den Bosch, A. (2016). Text analytics in industry: Challenges, desiderata and trends. Computers in Industry , 78, 96–107. Leontief, W. (1986). Input-output economics . Oxford University Press. Li, P., Zhang, P., Guo, Y., & Li, J. (2024). How has the relationship between major financial markets changed during the Russia–Ukraine conflict?. Humanities and Social Sciences Communications , 11(1), 1–20. Li, Y., Bandar, Z. A., & McLean, D. (2003). An approach for measuring semantic similarity between words using multiple information sources. IEEE Transactions on knowledge and data engineering , 15(4), 871–882. Matthews, R., Vivoda, V., & Leal, R. (2025). Deglobalization and the rehabilitation of Western defense industrial sovereignty. Comparative Strategy , 1–27. Mowery, D. C. (2012). Defense-related R&D as a model for “Grand Challenges” technology policies. Research policy , 41(10), 1703–1715. Peltier, H. (2023). We get what we pay for: the cycle of military spending, industry power, and economic dependence. Costs of War . Ploom, I., Kalvet, T., & Tiits, M. (2022). Defence industries in small European states: Key contemporary challenges and opportunities. Journal of International Studies , 15(4). Reimers, N., & Gurevych, I. (2019). Sentence-bert: Sentence embeddings using siamese bert-networks. arXiv preprint arXiv:1908.10084 . Stamegna, M., Bonaiuti, C., Maranzano, P., & Pianta, M. (2024). The economic impact of arms spending in Germany, Italy, and Spain. Peace Economics, Peace Science and Public Policy , 30(4), 393–422. Vázquez, D. (2020). Variety patterns in defense and health technological systems: evidence from international trade data. Journal of Evolutionary Economics , 30(4), 949–988. Wang, T. P., Shyu, S. H. P., & Chou, H. C. (2012). The impact of defense expenditure on economic productivity in OECD countries. Economic Modelling , 29(6), 2104–2114. Wassenaar Arrangement (2024). List of Dual-Use Goods and Technologies and Munitions List . World Customs Organization (2012). HS Nomenclature, 2012 Edition . https://www.wcoomd.org/en/topics/nomenclature/instrument-and-tools/hs_nomenclature_previous_editions/hs_nomenclature_table_2012.aspx Zhang, H., Jiao, L., Li, C., Deng, Z., Wang, Z., Jia, Q., ... & Hu, Y. (2024). Global environmental impacts of food system from regional shock: Russia-Ukraine war as an example. Humanities and Social Sciences Communications , 11(1), 1–13. Zhu, X., Yang, X., Huang, Y., Guo, Q., & Zhang, B. (2020). Measuring similarity and relatedness using multiple semantic relations in WordNet. Knowledge and Information Systems , 62(4), 1539–1569. Footnotes A detailed description of these products is available on the website of the statistics division of the UN at: tttps://unstats.un.org/unsd/classifications/Family/Detail/32. While the analysis above focuses on ML categories, the same framework can be applied from the perspective of HS codes. This alternative view would allow for the identification of HS products that are related to multiple ML categories and can therefore be interpreted as bridges across different categories. Figures Figures 1 to 8 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Figure1.TotalDefenceExpenditureasGDPinEUmembersbetween2014and2024.tif Figure 1. Total Defence Expenditure as % GDP in EU members between 2014 and 2024. Figure2.SchematicrepresentationofCategorySpace.tiff Figure 2. Schematic representation of Category Space. Figure3.TechnologicalRelatednessNetworkbetweenMLcategoriesandHSCodes..pdf Figure 3. Technological Relatedness Network between ML categories and HS Codes. Figure4.7z Figure 4. Clusters by ML Categories. Figure5.Mapofdefencereadinesscbycountry2024.Countrieswithnodataareshowningrey.pdf Figure 5. Map of defence readiness φ c by country ; 2024. Countries with no data are shown in grey Figure6.7z Figure 6. Relatedness Network for Germany 2024 (left) and Ireland 2024 (right). Figure7.7z Figure 7. Relatedness networks by country and year. Ukraine 2021 (upper left), Ukraine 2024 (upper right), Russia 2021 (middle left), Russia 2024 (middle right), Belarus 2021 (lower left), and Belarus 2024 (lower right). Figure8.7z Figure 8. Relatedness networks by country and year: EU27+UK 2021 (upper left), b) EU27+UK 2024 (upper right), China 2021 (middle left), China 2024 (middle right), USA 2021 (lower left), and USA 2024 (lower right). Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 17 May, 2026 Reviewers agreed at journal 12 May, 2026 Reviewers invited by journal 11 May, 2026 Editor assigned by journal 11 May, 2026 Editor invited by journal 02 Mar, 2026 Submission checks completed at journal 19 Feb, 2026 First submitted to journal 19 Feb, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8785807","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":593828668,"identity":"0f22a1bf-1556-49c4-9a0e-befdba0097d7","order_by":0,"name":"Víctor Llaneza","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6ElEQVRIiWNgGAWjYBACPnYogw1MVkAYB/BpYWOGawGxzpCihQGkhbGNCIexMTM/e/Dhj11iH/v5g48r5x3O42PvMTzAUFGHRwubueEMnuTENp5kZsOz2w4Xs/GcMTjAcOYwPoeZSfNIMBuzSTCzSTZuO5zYJpGWcICxDbd32JjZv0n/MaiHapkD0/IPn8N4zKQZEg7LQbQ0gLQkHzjA2MCMT0uZZM+B43JsPMnGhg3H0oGeOnzgQMIx3H7hZ2/fJvHjTzWPfPvBhw8baqwT57c3Nn/4UIPbYThAAqkaRsEoGAWjYBSgAAAEwUhVOqSjlAAAAABJRU5ErkJggg==","orcid":"","institution":"University of Oviedo","correspondingAuthor":true,"prefix":"","firstName":"Víctor","middleName":"","lastName":"Llaneza","suffix":""},{"id":593828669,"identity":"6551cd1c-6612-4af3-a093-815ae78026f9","order_by":1,"name":"André Carrascal-Incera","email":"","orcid":"","institution":"University of Oviedo","correspondingAuthor":false,"prefix":"","firstName":"André","middleName":"","lastName":"Carrascal-Incera","suffix":""},{"id":593828670,"identity":"ac69c08f-d3ae-4d31-a3eb-f03f58dd0377","order_by":2,"name":"Esteban Fernández Vázquez","email":"","orcid":"","institution":"University of Oviedo","correspondingAuthor":false,"prefix":"","firstName":"Esteban","middleName":"Fernández","lastName":"Vázquez","suffix":""}],"badges":[],"createdAt":"2026-02-04 11:26:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8785807/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8785807/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103050863,"identity":"d0c1fd17-b559-4522-9ebf-1b7460f32fa4","added_by":"auto","created_at":"2026-02-20 07:56:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":766442,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/ad796416-bf0d-4a42-9b5a-96536af0672a.pdf"},{"id":103048320,"identity":"412ff955-88f9-4775-8fa5-8c051715bac3","added_by":"auto","created_at":"2026-02-20 07:20:18","extension":"tif","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":10823,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 1. Total Defence Expenditure as % GDP in EU members between 2014 and 2024.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure1.TotalDefenceExpenditureasGDPinEUmembersbetween2014and2024.tif","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/315910c441a7dc0448b78d6a.tif"},{"id":103050457,"identity":"cd0f6a10-b195-4dfd-bfbe-074bdd1a766c","added_by":"auto","created_at":"2026-02-20 07:50:08","extension":"tiff","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":84247,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 2. Schematic representation of Category Space.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure2.SchematicrepresentationofCategorySpace.tiff","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/a5c8cbd8814c878c725758a3.tiff"},{"id":103048321,"identity":"4ecf985e-e108-44a5-b422-99595d08c561","added_by":"auto","created_at":"2026-02-20 07:20:18","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":179436,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 3. Technological Relatedness Network between ML categories and HS Codes.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure3.TechnologicalRelatednessNetworkbetweenMLcategoriesandHSCodes..pdf","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/c6c136f4633f298f2d4d4778.pdf"},{"id":103048322,"identity":"f0c32541-35f8-42fa-89a2-0728e6fa35b4","added_by":"auto","created_at":"2026-02-20 07:20:19","extension":"7z","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":567119,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 4. Clusters by ML Categories.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure4.7z","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/92bcf735d5ed2aaa1053794d.7z"},{"id":103048325,"identity":"b5c00b62-58f6-4384-81c2-f22fc666d91b","added_by":"auto","created_at":"2026-02-20 07:20:19","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":638352,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 5. Map of defence readiness φ\u003c/strong\u003e\u003csup\u003e\u003cstrong\u003ec\u003c/strong\u003e\u003c/sup\u003e\u003cstrong\u003e by country\u003c/strong\u003e\u003cem\u003e\u003cstrong\u003e; \u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e2024.\u003c/strong\u003e \u003cstrong\u003eCountries with no data are shown in grey\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure5.Mapofdefencereadinesscbycountry2024.Countrieswithnodataareshowningrey.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/98fe5a36ed05eb690f7aa563.pdf"},{"id":103050532,"identity":"5099fb0b-a899-4e47-987f-e1f76c3c86f7","added_by":"auto","created_at":"2026-02-20 07:50:25","extension":"7z","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":145763,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 6. Relatedness Network for Germany 2024 (left) and Ireland 2024 (right).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure6.7z","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/b1b57d8e21042329cbfe7b11.7z"},{"id":103048327,"identity":"637fb055-18e2-42af-b65a-6de65c480461","added_by":"auto","created_at":"2026-02-20 07:20:19","extension":"7z","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":258744,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 7. Relatedness networks by country and year. Ukraine 2021 (upper left), Ukraine 2024 (upper right), Russia 2021 (middle left), Russia 2024 (middle right), Belarus 2021 (lower left), and Belarus 2024 (lower right).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure7.7z","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/7eac21c06c4a81f2a47a0059.7z"},{"id":103050456,"identity":"1897ef1a-b873-46c7-9b5a-e5fade148943","added_by":"auto","created_at":"2026-02-20 07:50:07","extension":"7z","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":530607,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure 8. Relatedness networks by country and year: EU27+UK 2021 (upper left), b) EU27+UK 2024 (upper right), China 2021 (middle left), China 2024 (middle right), USA 2021 (lower left), and USA 2024 (lower right).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Figure8.7z","url":"https://assets-eu.researchsquare.com/files/rs-8785807/v1/65679682d9a1025a485f98c1.7z"}],"financialInterests":"No competing interests reported.","formattedTitle":"Defence Readiness: how to measure it across industrial ecosystems and how it has been impacted by the War in Ukraine","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eThe war in Ukraine has renewed attention on the industrial foundations of defence capability in Europe and beyond. The outbreak of the war in 2022 exposed significant differences in the ability of countries to supply military equipment, sustain production over time, and adapt their industrial structures to an increase in defence demand. These differences have become central to debates on European security, strategic autonomy, and the resilience of defence supply chains. This paper contributes to these debates by developing a method that quantifies defence‑related industrial ecosystems and applies it to the context of the Russia\u0026ndash;Ukraine war.\u003c/p\u003e \u003cp\u003eOne visible response to this and other recent conflicts has been the rise in defence expenditure across the European Union (EU) member states. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows total defence expenditure as a percentage of GDP in 2014 and 2024. The data from the European Defence Agency reveal substantial spatial heterogeneity. In 2024, countries such as Ireland, Malta, Luxembourg and Austria allocated less than 1% of GDP to defence, while Poland, Estonia, Lithuania and Latvia exceeded 3%. Despite this variation, as can be also seen in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, most EU members increased their defence spending notably between 2014 and 2024, reflecting a clear reassessment of security priorities in the last decade. However, defence preparedness depends not only on budgetary allocations but also on the underlying industrial capacity required to produce, maintain and supply military equipment.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eSource: Own elaboration based on data from the European Defence Agency (2025).\u003c/p\u003e \u003cp\u003eThis response has also been accompanied by the launch of new EU policy initiatives aimed at strengthening Europe\u0026rsquo;s defence industrial capacity. The European Commission\u0026rsquo;s \u0026ldquo;White Paper for European Defence \u0026ndash; Readiness 2030\u0026rdquo;, presented in March 2025, plans a programme of up to \u0026euro;800\u0026nbsp;billion in defence‑related investment over four years (European Commission, 2025). A central objective of this strategy is to reinforce the European Defence Technological and Industrial Base (EDTIB). Between the start of the war and June 2023, only 22% of EU member states\u0026rsquo; defence acquisitions were sourced within the EU (European Commission, 2024). To address this dependence, the strategy encourages member states to increase the share of defence procurement within the EU to at least 50% by 2030 and 60% by 2035 (European Commission, 2024). Similar initiatives have been adopted in the United States, Australia and the United Kingdom, where governments have launched programmes aimed at restoring or expanding domestic defence industrial capacity (Matthews et al., 2025).\u003c/p\u003e \u003cp\u003eUnderstanding this industrial dimension is essential because defence production relies on complex ecosystems that extend beyond strictly military goods. Many components, materials and technologies used in defence originate in non-military sectors. The European Commission defines an industrial ecosystem as the set of actors operating along a value chain, including firms of different sizes, research organisations, suppliers and service providers (European Commission, 2020). These interdependencies make difficult to assess the defence capability of a country demands by using direct measures alone (such as defence expenditure as a percentage of GDP). In that sense, the method proposed adopts this industrial ecosystem perspective.\u003c/p\u003e \u003cp\u003eResearch on defence spending has traditionally examined its regional and sectoral effects, with a particular focus on how changes in military budgets influence local economic outcomes. Studies have shown that increases or reductions in defence expenditure generate uneven spatial impacts, depending on the structure of regional economies and their capacity to absorb positive impacts or adapt to a new scenario (Atkinson, 1993; Bernauer et al., 2009). Other contributions have analysed the links between defence spending and high‑technology sectors, the role of defence‑related R\u0026amp;D, and the position of smaller EU countries within a common defence market (Boddy and Lovering, 1986; Mowery, 2012; Ploom et al., 2022; Wang et al., 2012). Input\u0026ndash;output approaches have been used to assess the overall economic effects of reallocating military expenditure to other sectors such as education or health, finding that defence spending produces lower employment effects and may slow down the long-term development of a country (Leontief, 1986; Peltier, 2023; Stamegna et al., 2024).\u003c/p\u003e \u003cp\u003eRecent work on the Russia\u0026ndash;Ukraine war has examined its effects also on financial markets (Li et al., 2024), global food systems (Zhang et al., 2024), and scientific capacity in Ukraine (de Rassenfosse et al., 2023), showing the range of the war\u0026rsquo;s economic and societal consequences. Related research has also revealed that shocks associated with the Russia\u0026ndash;Ukraine war propagate through production and trade networks in heterogeneous ways. Carrascal‑Incera et al. (2025) show that the subsequent disruptions in agriculture and industry due to the war generate uneven spillover effects across European regions, with central economies experiencing larger indirect impacts. These findings highlight the importance of understanding the productive structures that shape each country\u0026rsquo;s exposition to external shocks.\u003c/p\u003e \u003cp\u003eAnother strand of work regarding defence focuses on measuring it through the concept of Defence Industrial Base (DIB), stressing the idea that defence is not only the industries directly involved in military production but it should also account for more general items required by the defence industry (Dunne, 1995) such as vehicles, clothing or fuel, for example. Finally, the latest attempts to define what constitutes a military product can be found with the classification of the defence‑related production in the Wassenaar Arrangement (Wassenaar Arrangement, 2024) or the EU Common Military List (Council of the European Union, 2025).\u003c/p\u003e \u003cp\u003eWhile this literature provides valuable insights into the economic effects of defence spending and the composition of defence industries, it does not offer a proper tool to measure the industrial ecosystems that support defence production in a way that can be comparable across countries. Existing approaches rely either on direct data on military output or on broad sectoral classifications that do not capture the specific technological relationships between non-military and military goods. As a result, there is a lack of a framework capable of quantifying the productive structures that support defence capacity and of analysing how these structures respond to major geopolitical shocks such as the war in Ukraine.\u003c/p\u003e \u003cp\u003eThe objective of this paper is to develop a method to define a defence readiness that is quantified for the defence‑related industrial ecosystems using publicly available trade data and to apply it to analyse how these ecosystems evolved during the war in Ukraine. The method adapts Hidalgo\u0026rsquo;s product space framework (Hidalgo et al., 2007) to the defence field by combining semantic similarity techniques with the notion of Revealed Comparative Advantage (RCA). This approach allows us to measure the degree of relatedness between military categories defined in the Wassenaar Munitions List (ML) and the general products classified under the Harmonized System (HS).\u003c/p\u003e \u003cp\u003eThe empirical analysis performed in this paper focuses on the period 2021\u0026ndash;2024, which covers the years immediately before the start of the invasion of Ukraine and the most recent year with available data. By means of the \u003cem\u003ecategory space\u003c/em\u003e framework proposed, we are able to construct country‑specific networks and an indicator of relatedness, comparing the results by countries and over the period analysed.\u003c/p\u003e \u003cp\u003eIn particular, we base the empirical application on two hypotheses:\u003c/p\u003e \u003cp\u003eThe first one (H1) is that countries geographically closer to the conflict may experience an expansion of their defence‑related industrial ecosystem. This follows the idea that these countries may experience higher demand for military equipment, stronger policy incentives to increase production, and adjustments in industrial activity with the objective of meeting short term needs.\u003c/p\u003e \u003cp\u003eThe second one (H2) is that countries with established industrial capabilities are more likely to act as suppliers of military equipment during a conflict. In this case, this reflects the path‑dependent and slow responding nature of industrial development. Defence production relies on a particular knowledge, on specialised inputs and on technological complementarities that cannot be built in the short term. Countries with dense and diversified industrial ecosystems are therefore better positioned to scale up output when demand increases.\u003c/p\u003e \u003cp\u003eBy testing these hypotheses, the paper tries to illustrate the usefulness of the \u003cem\u003ecategory space\u003c/em\u003e method for (i) understanding how military items interact with the existing productive structures and (ii) quantifying how ready are the industrial capacity of countries to absorb a potential increase of military spending, overcoming the limitations of direct measures of military production and allowing for cross‑country and temporal comparisons.\u003c/p\u003e \u003cp\u003eAfter this introduction section, the rest of the paper is organised as follows. Section 2 describes the methodology and data, explaining the construction of the \u003cem\u003ecategory space\u003c/em\u003e approach, the steps to link military and non-military product classifications and the characteristics of the proposed defence readiness indicator. Section 3 presents the empirical results, examining the evolution of defence‑related industrial ecosystems between 2021 and 2024 and testing the two hypotheses established in this section. Section 4 concludes by summarising the main findings and outlining potential applications of the method for future research and policy analysis.\u003c/p\u003e"},{"header":"2. Methodology and data","content":"\u003cp\u003eThe limitations of the existing methods to adopt an industrial ecosystem approach when estimating defence‑related industrial capacity create a need for a new analytical tool. A tool that can also rely on publicly available data, since data on military production are often restricted, incomplete or not comparable across countries. This tool is what we call the \u003cem\u003ecategory space\u003c/em\u003e method.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Product and Category Space\u003c/h2\u003e \u003cp\u003eThe methodology applied in this paper builds on the idea capturing the density of the economic system around military spending. For this purpose, we develop a variant of the measure of relatedness proposed by Hidalgo (2007), where he introduces the concept of \u003cem\u003eproduct space\u003c/em\u003e, a network in which the proximity between products reflects their relatedness. In that paper, the proximity between products\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:i,p\\:\\)\u003c/span\u003e\u003c/span\u003efor a country \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:c\\)\u003c/span\u003e\u003c/span\u003e is measured using pairwise conditional probabilities, specifically assessing the likelihood that a country exports one good given that it already exports another. More formally, the indicator is defined as the probability for a country \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:c\\)\u003c/span\u003e\u003c/span\u003e of having a Revealed Comparative Advantage in product \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{RCA}_{i}^{c}\\)\u003c/span\u003e\u003c/span\u003e) conditional on having Revealed Comparative Advantage in product \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:p\\)\u003c/span\u003e\u003c/span\u003e (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{RCA}_{p}^{c}\\)\u003c/span\u003e\u003c/span\u003e), or vice versa. The measure of proximity \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varphi\\:}_{ip}^{c}\\)\u003c/span\u003e\u003c/span\u003e proposed in Hidalgo (2007) is calculated as:\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:{\\varphi\\:}_{ip}^{c}=\\text{min}\\left\\{P\\left({RCA}_{i}^{c}/{RCA}_{p}^{c}\\right),P\\left({RCA}_{p}^{c}/{RCA}_{i}^{c}\\right)\\right\\}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe definition of Revealed Comparative Advantage (RCA) for country c in product i is defined following Balassa (1965) as:\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:{RCA}_{i}^{c}=\\frac{{x}_{i}^{c}/\\sum\\:_{i}{x}_{i}^{c}}{\\sum\\:_{c}{x}_{i}^{c}/\\sum\\:_{c}\\sum\\:_{i}{x}_{i}^{c}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{x}_{i}^{c}\\)\u003c/span\u003e\u003c/span\u003e stands for the exports of country \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:c\\)\u003c/span\u003e\u003c/span\u003e of product \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e. The concept of RCA acts as a filter to construct the Product Space for an economy, considering only goods whose exports are significant (RCA\u0026thinsp;\u0026gt;\u0026thinsp;1), which reflects their embedded productive capabilities. This measurement of economic proximity is essential for understanding which products an economy is likely to produce in the future and, therefore, for explaining the development of countries. Based on the same concept, although adapted to our particular problem, the study builds on the idea of \u003cem\u003ecategory space\u003c/em\u003e. However, instead of relying on conditional probabilities of exporting products as the pieces for constructing this space, we apply semantic techniques that estimate the degree of relatedness between military ML categories and the description of products exported, following the HS coding system, which is the international standard product classification used to categorize traded goods.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Defining a defence readiness indicator\u003c/h2\u003e \u003cp\u003eThe key concept used in this paper is the idea of semantic similarity between the descriptions of categories in the ML and HS lists. Semantic measures include semantic similarity (SS), which indicates shared features between elements, and semantic relatedness (SR), which covers all types of relationships in context (Zhu et al., 2020; Costa and Leal, 2016; Li, 2003). These concepts let researchers examine words and their underlying semantic connections, imitating human reasoning. Recent advances in semantic methods allow exploration of research areas previously inaccessible (Gentzkow et al., 2019; Ittoo and van den Bosch, 2016; Hoberg and Phillips, 2016). Hoberg and Phillips (2016) developed an industry classification using web-crawling algorithms to analyse over 50,000 firm annual reports, clustering firms by product description similarity. V\u0026aacute;zquez (2020) evaluated the defence and health industries\u0026rsquo; capacity for innovation, noting that analysing only 16 military product codes reveals the limitations of traditional sector analysis tools.\u003c/p\u003e \u003cp\u003eIn this paper, we address this limitation by applying semantic methods to construct a \u0026ldquo;\u003cem\u003eproduct space\u003c/em\u003e\u0026rdquo; around ML categories. These methods allow us to broaden the consideration of military products by incorporating not only strictly military items but also those highly related to the sector.\u003c/p\u003e \u003cp\u003eRather than building networks by just considering products, as in Hidalgo and Hausmann (2009), we create \u003cem\u003ecategory spaces\u003c/em\u003e centred on the ML categories and surrounded by related products (as described in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). This approach is represented as a bipartite network with two sets of nodes, HS products and ML categories, where product nodes connect directly only to ML nodes (see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eSource: Own elaboration.\u003c/p\u003e \u003cp\u003eTo do so, we employ semantic techniques that estimate the degree of relatedness between ML categories and HS products at the 4-digit level. In particular, we use a sentence-embedding model. Specifically, we utilize \u003cem\u003eall-MiniLM-L6-v2\u003c/em\u003e, a compact pretrained sentence-embedding model designed to transform short texts into fixed-length numerical vectors that represent their semantic content. This model consists of a 6-layer MiniLM Transformer architecture, trained with contrastive learning on extensive collections of sentence pairs so that semantically similar texts produce closely aligned embeddings. In practical applications, it is accessed via the sentence-transformers library, facilitating efficient execution of tasks such as semantic similarity assessment, clustering, and information retrieval. This library is a Python module based on SBERT (Sentence Bidirectional Encoder Representations from Transformers) and is specifically designed to produce semantically meaningful sentence embeddings, which can be compared using cosine-similarity (Reimers and Gurevych, 2019). In other words, this tool enables the comparison of not only words but also sentences, by transforming text into numbers and incorporating the contextual meaning of the entire sentence.\u003c/p\u003e \u003cp\u003eSpecifically, we apply this approach to quantify the semantic relatedness between ML definitions and HS code descriptions, measuring the semantic distance between them and subsequently obtaining the \u003cem\u003ecategory space\u003c/em\u003e. To quantify the relatedness between ML categories and HS product codes, we define a semantic similarity indicator \u003cem\u003eS\u003c/em\u003e\u003csub\u003e\u003cem\u003ek,i\u003c/em\u003e\u003c/sub\u003e based on the cosine similarity of their textual embeddings.\u003c/p\u003e \u003cp\u003eLet \u003cem\u003eD\u003c/em\u003e\u003csub\u003e\u003cem\u003ek\u003c/em\u003e\u003c/sub\u003e\u003csup\u003eML\u003c/sup\u003e denote the textual description of ML category \u003cem\u003ek\u003c/em\u003e (where \u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1, 2, ..., 22) and let \u003cem\u003eD\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e\u003csup\u003eHS\u003c/sup\u003e denote the textual description of HS product code \u003cem\u003ei\u003c/em\u003e at the 4-digit level (where \u003cem\u003ei\u003c/em\u003e\u0026thinsp;=\u0026thinsp;1, 2, ..., \u003cem\u003en\u003c/em\u003e and \u003cem\u003en\u003c/em\u003e\u0026thinsp;\u0026asymp;\u0026thinsp;1,200 headings in the HS 2012 nomenclature). Using the sentence-transformer model \u003cem\u003eall-MiniLM-L6-v2\u003c/em\u003e, each textual description is mapped into a dense vector representation in a high-dimensional embedding space ℝ\u003csup\u003ed\u003c/sup\u003e (where \u003cem\u003ed\u003c/em\u003e\u0026thinsp;=\u0026thinsp;384 for this model). We denote these embedding vectors as:\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:{e}_{k}^{ML}\\:=\\:f\\left({D}_{k}^{ML}\\right)\\:\\in\\:\\:{\\mathbb{R}}^{d}\\:\\text{a}\\text{n}\\text{d}\\:{e}_{j}^{HS}\\:=\\:f\\left({D}_{j}^{HS}\\right)\\:\\in\\:\\:{\\mathbb{R}}^{d}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere \u003cem\u003ef\u003c/em\u003e(\u0026middot;) represents the embedding function implemented by the sentence-transformer model, which captures the semantic meaning of the input text by encoding contextual relationships between words.\u003c/p\u003e \u003cp\u003eThe semantic similarity indicator \u003cem\u003eS\u003c/em\u003e\u003csub\u003e\u003cem\u003ek,i\u003c/em\u003e\u003c/sub\u003e between ML category \u003cem\u003ek\u003c/em\u003e and HS product \u003cem\u003ei\u003c/em\u003e is then computed as the cosine similarity between their respective embedding vectors:\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e\n$$\\:{S}_{k,i}=\\text{c}\\text{o}\\text{s}\\left({e}_{k}^{ML},{e}_{i}^{HS}\\right)=\\frac{{\\sum\\:}_{j=1}^{d}\\left({e}_{k,j}^{ML}\\right)\\left({e}_{i,j}^{HS}\\right)}{\\sqrt{{\\sum\\:}_{j=1}^{d}{\\left({e}_{k,j}^{ML}\\right)}^{2}}\\sqrt{{\\sum\\:}_{j=1}^{d}{\\left({e}_{i,j}^{HS}\\right)}^{2}}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eIn Eq.\u0026nbsp;(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{e}_{k}^{ML}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{e}_{i}^{HS}\\)\u003c/span\u003e\u003c/span\u003erepresent the \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:j\\)\u003c/span\u003e\u003c/span\u003e-th component of the embedding vectors for ML category \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:k\\)\u003c/span\u003e\u003c/span\u003e and HS product \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e, respectively. This indicator ranges between 0 and 1, with values close to 1 indicating a high similarity between \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:k\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e, while values near to 0 suggest that \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:k\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e are semantically disconnected. As a consequence, it is possible to evaluate the \u0026ldquo;density\u0026rdquo; of the \u003cem\u003ecategory space\u003c/em\u003e for class \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:k\\)\u003c/span\u003e\u003c/span\u003e as the sum \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sum\\:}_{i=1}^{n}{S}_{k,i}\\)\u003c/span\u003e\u003c/span\u003e: large values of this sum identify categories in the ML with a dense network of products connected to them.\u003c/p\u003e \u003cp\u003eNote, however, that the elements \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{S}_{k,i}\\)\u003c/span\u003e\u003c/span\u003e are defined solely on the semantic relationship between the textual descriptions between the two classifications lists, therefore they are common for all the\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\:C\\)\u003c/span\u003e\u003c/span\u003e countries under study. But the capabilities of the different economies for taking advantage of these connections between the products they produce and the categories of the ML list are expected to vary across countries. One way of capturing this heterogeneity is by defining a \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:C\\times\\:n\\)\u003c/span\u003e\u003c/span\u003e matrix with binary cells \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{b}_{i}^{c}\\)\u003c/span\u003e\u003c/span\u003e defined as:\u003cdiv id=\"Equ5\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ5\" name=\"EquationSource\"\u003e\n$$\\:{b}_{i}^{c}\\:=\\left\\{\\begin{array}{c}1\\:\\:if\\:{\\text{R}\\text{C}\\text{A}}_{\\text{i}}^{\\text{c}}\u0026gt;1\\\\\\:0\\:\\:if\\:{\\text{R}\\text{C}\\text{A}}_{\\text{i}}^{\\text{c}}\\le\\:1\\end{array}\\right.$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e5\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWe propose an indicator of defence readiness \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\phi\\:}^{c}\\)\u003c/span\u003e\u003c/span\u003e, which captures the relatedness of the production capacity of country \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:c\\:\\)\u003c/span\u003e\u003c/span\u003earound military products as:\u003cdiv id=\"Equ6\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ6\" name=\"EquationSource\"\u003e\n$$\\:{\\phi\\:}^{c}=\\frac{{\\sum\\:}_{k=1}^{K}{\\sum\\:}_{i=1}^{n}{S}_{k,i}{b}_{i}^{c}}{{\\sum\\:}_{k=1}^{K}{\\sum\\:}_{i=1}^{n}{S}_{k,i}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e6\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThe indicator \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\phi\\:}^{c}\\)\u003c/span\u003e\u003c/span\u003e is naturally bounded between 0 and 1, and relatively large values of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\phi\\:}^{c}\\)\u003c/span\u003e\u003c/span\u003e indicate that country \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:c\\)\u003c/span\u003e\u003c/span\u003e presents an economic specialization (measured through the production of goods with RCAs) that allows it to exploit the relatedness with the products in the categories of the ML list. Cases like that reveal countries with a good capacity to benefit from an expansion of military expenditure.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Data\u003c/h2\u003e \u003cp\u003eThis research required the handling of public open-access data provided by several globally recognised organisations. A quick description of these sources can be found below:\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003e1. The Wassenaar Arrangement\u003c/h3\u003e\n\u003cp\u003eThis work uses the \u003cem\u003eWassenaar Arrangement on Export Controls for Conventional Arms and Dual-Use Goods and Technologies\u003c/em\u003e (Wassenaar Arrangement, 2024), in force since 1996, as the main reference to define which products should be considered defence-related. Services are not included.\u003c/p\u003e \u003cp\u003eThe Arrangement, signed by 42 participating states (Argentina, Australia, Austria, Belgium, Bulgaria, Canada, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, India, Ireland, Italy, Japan, Latvia, Lithuania, Luxembourg, Malta, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Republic of Korea, Romania, Russian Federation, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland, T\u0026uuml;rkiye, Ukraine, United Kingdom and United States), aims to contribute to regional and international security by establishing a common framework for the control of exports of items included in the \u003cem\u003eList of Dual-Use Goods and Technologies and the Munitions List\u003c/em\u003e (Wassenaar Arrangement, 2024). Participating states commit to applying shared guidelines, elements and procedures as the basis for their national export control legislation.\u003c/p\u003e \u003cp\u003e \u003cem\u003eThe Wassenaar Arrangement Control Lists\u003c/em\u003e consist of the following Munitions List, which includes 22 categories of items designed for military use. The codes and descriptions of each category used to feed the semantic-similarity algorithm are shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eWassenaar Arrangement 2024 - Munitions List \u0026ndash; Items 1 to 22.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCode\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTextual description\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML1\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSmooth-bore weapons with a calibre of less than 20 mm, other arms and automatic weapons with a calibre of 12,7 mm (calibre 0,50 inches) or less and accessories and specially designed components therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML2\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSmooth-bore weapons with a calibre of 20 mm or more, other weapons or armament with a calibre greater than 12,7 mm (calibre 0,50 inches), projectors specially designed or modified for military use and accessories and specially designed components therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML3\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAmmunition and fuze setting devices and specially designed components therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML4\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBombs, torpedoes, rockets, missiles, other explosive devices and charges and related equipment and accessories and specially designed components therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML5\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFire control, surveillance and warning equipment, and related systems, test and alignment and countermeasure equipment specially designed for military use, and specially designed components and accessories therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML6\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGround vehicles and components.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML7\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eChemical agents, \"biological agents\", \"riot control agents\", radioactive materials, related equipment, components and materials.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML8\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\"Energetic materials\" and related substances.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML9\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eVessels of war (surface or underwater), special naval equipment, accessories, components and other surface vessels.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML10\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\"Aircraft\", \"lighter-than-air vehicles\", \"Unmanned Aerial Vehicles\" (\"UAVs\"), aero-engines, \u0026ldquo;sub-orbital craft\u0026rdquo; and \"aircraft\" equipment, related equipment, and components, specially designed or modified for military use.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML11\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eElectronic equipment, \"spacecraft\" and components, not specified elsewhere on the Munitions List.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML12\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHigh velocity kinetic energy weapon systems and related equipment and specially designed components therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML13\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eArmoured or protective equipment, constructions, components, and accessories.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML14\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e'Specialised equipment for military training' or for simulating military scenarios, simulators specially designed for training in the use of any firearm or weapon specified by ML1 or ML2, and specially designed components and accessories therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML15\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eImaging or countermeasure equipment, specially designed for military use, and specially designed components and accessories therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML16\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eForgings, castings and other unfinished products, specially designed for items specified by ML1 to ML4, ML6, ML9, ML10, ML12 or ML19.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML17\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMiscellaneous equipment, materials and \"libraries\" and specially designed components therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML18\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e'Production' equipment, environmental test facilities and components.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML19\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDirected Energy Weapon (DEW) systems, related or countermeasure equipment and test models and specially designed components therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML20\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCryogenic and \"superconductive\" equipment and specially designed components and accessories therefor.\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML21\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\"Software\".\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eML22\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\"Technology\".\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e\n\u003ch3\u003e2. Products in the Harmonized System (HS) list\u003c/h3\u003e\n\u003cp\u003eThe embeddings of the textual descriptions of the ML categories need to be mapped together with descriptions of products exported between countries. These descriptions have been extracted from the Harmonized System (HS), developed by the World Customs Organization (WCO) and used by over 200 countries, as the framework for product classification. Specifically, we use the HS 2012 Edition at the 4-digits level (HS heading), which covers more than 1,200 categories of goods (World Customs Organization, 2012).\u003csup\u003e1\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eData on actual export flows by product following this classification have been obtained from the BACI (\u003cem\u003eBase Analytique du Commerce International\u003c/em\u003e) database. This dataset \u0026ldquo;\u003cem\u003eprovides data on bilateral trade flows for 200 countries at the product level (5.000 products), according to Harmonized System nomenclature\u003c/em\u003e\u0026rdquo; (Gaulier and Zignago, 2010) and is collected and curated by the Centre d'Etudes Prospectives et d'Informations Internationales (CEPII).\u003c/p\u003e \u003cp\u003eThe raw exports and imports data come from the UN Comtrade database, provided by the United Nations Statistical Division. However, CEPII performs additional operations to improve data accuracy, considering FOB import values and taking into account the reliability of trade data reported by each country. Specifically, this work uses the HS12_V2026.01 version of the BACI database, updated in January 2026 (CEPII, 2026).\u003c/p\u003e"},{"header":"3. Results","content":"\u003cp\u003eThis section is organized as follows. First, we quantify and represent the semantic relatedness between Wassenaar Munitions List (ML) categories and Harmonized System (HS) codes by constructing a technological network that captures the relationships between categories and products. Second, based on this network and the procedure described above, we present the networks for two representative countries and a global choropleth map to illustrate how the approach reveals and quantifies the relatedness between military categories and industrial capabilities. Finally, focusing on the war in Ukraine, we present the networks of the countries involved for 2021 and 2024, highlighting the potential of the method to compare economies and analyse their evolution over time.\u003c/p\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Technological Relatedness Network between ML categories and HS Codes\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e displays the product space surrounding ML categories as a bipartite network. This graph illustrates the links between ML categories and HS Codes, considering only the top 1% of HS-ML connections with the highest relatedness scores. This filter shows the most relevant connections that structure the defence-related category space. Link thickness reflects the strength of semantic relatedness between HS products and ML categories.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eEach ML category can be interpreted as a cluster within the technological network. Figure\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents this segmentation, based on the network filtered at the top 1% of relatedness values for the sake of clarity in the presentation of the graph. The figure shows substantial heterogeneity across ML categories: some of them are associated with a large number of related HS codes, such as ML8 (Explosives and Propellants), ML10 (Aircraft) and ML17 (Miscellaneous Equipment); in contrast, other categories, such as ML1 (Small Arms), ML2 (Weapons \u0026gt;\u0026thinsp;20mms) and ML12 (Kinetic Energy Weapons) are linked to relatively few HS codes. A third group of ML categories does not appear in the graph, because none of their associated HS codes fall within the top 1% of relatedness values. This group includes ML21 (Software) and ML22 (Technology). \u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e2.2. The geographical map of defence readiness\u003c/h2\u003e \u003cp\u003eOnce the technological relatedness network is obtained and following the criteria proposed by Hidalgo (2007) and data on bilateral trade flows provided by BACI (Gaulier G. a., 2010), it is possible to construct annual rankings that identify countries, regions or economic blocs, whose economies are most closely related to the defence sector. It is important to note that these rankings do not show which economies produce more military goods, but those whose industrial ecosystem is most closely related with the military industry.\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents a global choropleth map in which countries are coloured according to their indicator of defence readiness values (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\phi\\:}^{c}\\)\u003c/span\u003e\u003c/span\u003e) in 2024. The map provides a synthetic representation of the capabilities of each country in defence-related products. Moreover, it reveals a clear spatial pattern, both for the top positions and for the bottom ones.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eTo facilitate the interpretation of the map, Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e reports the countries with higher \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\phi\\:}^{c}\\)\u003c/span\u003e\u003c/span\u003ein 2024, that is, the economies that are theoretically best positioned to benefit from an expansion in military spending.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eTop economies by defence readiness indicator \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{\\phi\\:}}^{\\varvec{c}}\\)\u003c/span\u003e\u003c/span\u003e for the year 2024.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCountry\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\phi\\:}^{c}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChina\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.4210\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGermany\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.3993\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eItaly\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.3617\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpain\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.3157\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAustria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.3138\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTurkey\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.3083\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoland\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.3004\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCzechia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2846\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFrance\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2837\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnited Kingdom\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2638\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnited States\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2634\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eJapan\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2609\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePortugal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2508\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2475\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCroatia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2430\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTo further illustrate the explanatory potential of the method, Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e presents, as networks, the defence-related category space of two countries (Germany and Ireland) for 2024. These countries are chosen as examples due to their disparity. They differ markedly in defence expenditure as a share of GDP (2,1% in Germany and 0,2% in Ireland) and in their productive structures, with Germany representing a highly industrialised economy and Ireland a more service-oriented one.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThis comparison highlights how the network representation helps to understand the industrial ecosystem related to defence of each economy:\u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThe German network shows a large number of nodes with strong interconnections, whereas the Irish network is considerably smaller, with fewer nodes and more limited connections between them. Accordingly, the network structure clearly suggests that the German economy is more closely related to the defence sector than the Irish economy.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e2.3. The War in Ukraine: analysis of the economies involved\u003c/h2\u003e \u003cp\u003eFinally, we analyse the relatedness networks of the countries directly involved in the war in Ukraine in order to characterise their defence-related industrial ecosystems and their evolution over the course of the conflict. To this end, Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e presents the networks of Ukraine, Russia and Belarus before the war (2021) and in 2024, two years after the official start of the conflict.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eThe networks observed in 2021 are relatively similar across the three countries and are characteristic of economies with limited industrial capabilities related to the defence sector. Although Belarus displays a denser network, none of the three resembles the structure observed for a highly industrialised economy such as Germany (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn 2024, the Ukrainian network remains relatively sparse, although several HS codes that were absent in 2021 now appear in the network. These changes may reflect adjustments in industrial capabilities during the conflict. One illustrative example is the emergence of HS 9306, related to bombs, grenades, and other munitions, as a key node in the 2024 network.\u003c/p\u003e \u003cp\u003eBy contrast, the networks of Russia and Belarus exhibit a marked contraction over the same period. In 2024 both countries display fewer HS codes and ML categories than in 2021, indicating a substantial reduction in the density of their defence-related networks since de beginning of the conflict.\u003c/p\u003e \u003cp\u003eTo complete the analysis, we also include the networks of EU27\u0026thinsp;+\u0026thinsp;UK, China and USA, economies not directly involved in the conflict but playing a significant role.\u003c/p\u003e \u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e, these networks are relatively similar to each other and follow the structure observed in the German network, with a large number of HS nodes connected to many ML categories. This pattern reflects highly industrialised economies whose productive structure is closely linked to the defence industry. Furthermore, no substantial changes are observed over the 2021\u0026ndash;2024 period.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e reports the corresponding \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\phi\\:}^{c}\\)\u003c/span\u003e\u003c/span\u003e values for the economies involved in the war. This indicator allows us to summarize the previous networks in a single figure and to rank these economies according to the characteristics of their defence-related industrial ecosystem:\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eEconomies involved in the war in Ukraine by \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{\\phi\\:}}^{\\varvec{c}}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2021\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2024\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEU27\u0026thinsp;+\u0026thinsp;UK\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.4786\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.4740\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChina\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.3929\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.4210\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUSA\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.2559\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.2634\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUkraine\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.1236\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.1245\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBelarus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.1850\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.1233\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRussia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.1028\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.0778\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThese figures confirm that Ukraine, Belarus and Russia exhibit economies weakly related to the defence sector, whereas the USA, China and the EU27\u0026thinsp;+\u0026thinsp;UK are more strongly connected, particularly China and EU27\u0026thinsp;+\u0026thinsp;UK.The outbreak of the war in 2022 further exacerbated these dynamics, deepening the erosion of industrial diversity and narrowing the range of goods in which Russia and Belarus remain competitive on global markets. By contrast, Ukraine, as well as the EU27\u0026thinsp;+\u0026thinsp;UK, maintains networks similar to those observed before the war. The two economies that can be classified as \u0026ldquo;winners\u0026rdquo; in this process has been USA and, specially, China, which show respectively a modest and notable expansion on their network over the same period.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Conclusions and discussion","content":"\u003cp\u003eThis paper proposes the study of the \u003cem\u003ecategory space\u003c/em\u003e framework as a method to study defence‑related industrial ecosystems and to define an indicator of defence readiness. By combining semantic similarity measures between lists of military products and exported goods, and including data of exports indicating the presence or not of Revealed Comparative Advantage (RCA), the approach constructs networks that capture how each country\u0026rsquo;s productive structure relates to the categories considered in the Wassenaar Munitions List (ML). The defence readiness indicator summarises the density of these networks into a single comparable figure, measuring how well positioned an economy is to absorb from an expansion of military spending. The underlying idea is that defence preparedness cannot be adequately assessed through budgetary allocations alone, as the underlying industrial ecosystem plays a fundamental role in determining whether increased spending can be translated into effective military capability. The alternative proposal presented here offers an alternative to conventional measures of defence capacity that rely on expenditure data or narrow sectoral classifications, and provides an easily replicable, data‑driven tool for cross‑country and temporal comparisons.\u003c/p\u003e \u003cp\u003eBesides the theoretical definition of the indicator proposed, this paper calculates it for several years basing on recent data of trade flows. The global mapping of defence readiness presented in this paper reveals a clear spatial pattern. Highly industrialised economies such as China and some countries in Western Europe occupy the top positions, reflecting productive structures densely connected to the defence domain. The empirical application to the war in Ukraine comparing the years 2021 and 2024 has yielded several findings. With respect to the first hypothesis (H1), which posited that countries geographically closer to the conflict would experience an expansion of their defence‑related industrial ecosystem, the evidence is mixed. Ukraine exhibits a modest expansion of its network. This result is consistent with the expected adjustment of productive structures under conditions of increased defence demand. However, Russia and Belarus display a marked contraction of their networks over the same period, with fewer HS codes and ML categories connected in 2024 than in 2021. The direction of involvement in the war, combined with the effects of international sanctions, appears to be a decisive moderating factor: while Ukraine\u0026rsquo;s productive structure adapted to wartime needs despite, the sanctions regime eroded the industrial diversity of Russia and Belarus in global markets, reducing the range of goods in which these countries retain a revealed comparative advantage and, consequently, lowering their defence readiness indicator.\u003c/p\u003e \u003cp\u003eThe evidence for the second hypothesis (H2), which proposed that countries with established industrial capabilities are more likely to act as suppliers of military equipment during a conflict, is considerably strong. The EU27\u0026thinsp;+\u0026thinsp;UK aggregate, China and the United States all maintain dense and diversified category space networks in both 2021 and 2024, consistent with the path‑dependent nature of industrial development stressed in the economic complexity literature. The structural stability of these networks over a period of significant geopolitical disruption confirms that defence readiness cannot be built in the short term but instead reflects accumulated productive capabilities, specialised knowledge and technological complementarities developed over decades. Within this group of established economies, the analysis reveals an asymmetric evolution that is relevant for understanding the geopolitical reconfiguration of defence‑related trade. China exhibits the most notable expansion of its defence readiness indicator, suggesting a deepening of the connections between its export structure and the defence sphere. The United States registers a modest increase, while the EU27\u0026thinsp;+\u0026thinsp;UK remains essentially stable with a slight decline. These dynamics point to China as the principal beneficiary of the reconfiguration of defence‑related trade networks triggered by the war.\u003c/p\u003e \u003cp\u003eFrom a policy perspective, the results carry important implications for the European debate on defence readiness. The EU27\u0026thinsp;+\u0026thinsp;UK aggregate records the highest defence readiness value among all the economies analysed, indicating that Europe\u0026rsquo;s productive structure is, on average, more densely connected to the activities linked to defence than any other major bloc. Yet, as noted in the introduction, only 22% of EU defence acquisitions between the start of the war and June 2023 were produced within the EU. This gap suggests that the challenge for European defence policy might lie in mobilising and coordinating the productive capabilities that already exist across member states.\u003c/p\u003e \u003cp\u003eThe methodology also opens several avenues for future research. The methodology presented here can be applied at the sub‑national level by exploiting regional trade or production data to identify which regions within a country possess the productive structures most closely related to the defence sector. It can also be extended to other conflict settings or to monitor changes in defence readiness over longer time horizons. Moreover, the bipartite network structure of the approach allows for analysis from the perspective of HS codes rather than ML categories, identifying those commercial products that serve as bridges across multiple military domains and that may therefore represent strategic bottlenecks or opportunities for industrial policy.\u003c/p\u003e \u003cp\u003eThe study also has limitations that should be acknowledged. Semantic similarity captures only potential relatedness between product descriptions rather than actual production flows, and the strength of the analysis depends on the quality and comprehensiveness of the textual descriptions available in the classifications analysed. Similarly, relying on international trade data, might understate the defence readiness of countries whose defence‑related production is primarily oriented towards domestic consumption or whose trade data are incomplete or distorted, as may be the case for sanctioned economies. Finally, the 2021\u0026ndash;2024 window captures only the early dynamics of the war, and the structural adjustments observed may not fully reflect longer‑term transformations in defence‑related industrial ecosystems that are still underway.\u003c/p\u003e \u003cp\u003eDespite these limitations, the methodology presented here helps to address a question that has traditionally been difficult to measure: how ready are productive structures to respond to shifts in defence demands? By providing a replicable, data‑driven and comparable method, this paper contributes with a new tool for understanding the industrial foundations of defence capability and for analysing how these foundations respond to major geopolitical shocks.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eAuthor ContributionsVLL: Conceptualization, methodology, data curation, formal analysis, writing \u0026ndash; original draft.AC: Methodology, data analysis, validation, writing \u0026ndash; review and editing.\u0026shy;EF: Conceptualization, supervision, interpretation of results, writing \u0026ndash; review and editing.All authors have read and approved the final manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors acknowledge the financial support of the C\u0026aacute;tedra Sekuens de la Innovaci\u0026oacute;n en Asturias (CATI-25-010).\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eAll data used in this study are publicly available:- Trade data at the product level were obtained from the BACI database compiled by CEPII ( [link](https:/www.cepii.fr/CEPII/en/bdd_modele/bdd_modele_item.asp?id=37) ), based on UN Comtrade data. Product classifications follow the Harmonized System (HS) 2012 nomenclature published by the World Customs Organization ( [link](https:/www.wcoomd.org/en/topics/nomenclature/instrument-and-tools/hs_nomenclature_previous_editions/hs_nomenclature_table_2012.aspx) ).- The definitions of defence-related categories are based on the Wassenaar Arrangement Munitions List (2024 edition) ( [link](http:/www.wassenaar.org/app/uploads/2024/12/List-of-Dual-Use-Goods-and-Technologies-and-ML-2024.pdf) ).All data sources are cited in the manuscript. The processed datasets and code used to construct the defence readiness indicator are available from the corresponding author upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAtkinson, R. D. (1993). Defense spending cuts and regional economic impact: An overview. \u003cem\u003eEconomic Geography\u003c/em\u003e, 69(2), 107\u0026ndash;122.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBalassa, B. (1965). Trade liberalisation and \u0026ldquo;revealed\u0026rdquo; comparative advantage. \u003cem\u003eThe Manchester School\u003c/em\u003e, 33(2), 99\u0026ndash;123.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBernauer, T., Koubi, V., \u0026amp; Ernst, F. (2009). National and regional economic consequences of Swiss defense spending. \u003cem\u003eJournal of Peace Research\u003c/em\u003e, 46(4), 467\u0026ndash;484.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoddy, M., \u0026amp; Lovering, J. (1986). High technology industry in the Bristol sub-region: the aerospace/defence nexus. \u003cem\u003eRegional studies\u003c/em\u003e, 20(3), 217\u0026ndash;231.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCarrascal-Incera, A., Liu, W., Orea, L., \u0026amp; Sickles, R. C. (2025). Beyond borders: how spillovers and commercial networks shape European productivity. \u003cem\u003eJournal of Productivity Analysis\u003c/em\u003e, 1\u0026ndash;49.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCEPII (2026). \u003cem\u003eBACI, Base pour l\u0026rsquo;Analyse du Commerce International\u003c/em\u003e. www.cepii.fr/CEPII/en/bdd_modele/bdd_modele_item.asp?id=37\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCosta, T., \u0026amp; Leal, J. P. (2016, May). Semantic measures: How similar? How related?. In \u003cem\u003eInternational Conference on Web Engineering\u003c/em\u003e (pp. 431\u0026ndash;438). Cham: Springer International Publishing.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCouncil of the European Union (2025). Common military list of the European Union adopted by the Council on 24 February 2025. \u003cem\u003eOfficial Journal of the European Union\u003c/em\u003e, C(1499).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDe Rassenfosse, G., Murovana, T., \u0026amp; Uhlbach, W. H. (2023). The effects of war on Ukrainian research. \u003cem\u003eHumanities and Social Sciences Communications\u003c/em\u003e, 10(1), 1\u0026ndash;11.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDunne, J. P. (1995). The defense industrial base. \u003cem\u003eHandbook of defense economics\u003c/em\u003e, 1, 399\u0026ndash;430.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEuropean Commission (2020). \u003cem\u003eA New Industrial Strategy for Europe\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEuropean Commission (2024). \u003cem\u003eA New European Defence Industrial Strategy: Achieving EU Readiness Through a Responsive and Resilient European Defence Industry\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEuropean Commission (2025). \u003cem\u003eWhite Paper for European Defence \u0026ndash; Readiness 2030\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEuropean Defence Agency (2025). \u003cem\u003eDefence Data 2024\u0026ndash;2025\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGaulier, G., \u0026amp; Zignago, S. (2010). Baci: international trade database at the product-level (the 1994\u0026ndash;2007 version). CEPII Working Paper.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGentzkow, M., Kelly, B., \u0026amp; Taddy, M. (2019). Text as data. \u003cem\u003eJournal of Economic Literature\u003c/em\u003e, 57(3), 535\u0026ndash;574.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHidalgo, C. A., \u0026amp; Hausmann, R. (2009). The building blocks of economic complexity. \u003cem\u003eProceedings of the national academy of sciences\u003c/em\u003e, 106(26), 10570\u0026ndash;10575.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHidalgo, C. A., Klinger, B., Barab\u0026aacute;si, A. L., \u0026amp; Hausmann, R. (2007). The product space conditions the development of nations. \u003cem\u003eScience\u003c/em\u003e, 317(5837), 482\u0026ndash;487.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoberg, G., \u0026amp; Phillips, G. (2016). Text-based network industries and endogenous product differentiation. \u003cem\u003eJournal of political economy\u003c/em\u003e, 124(5), 1423\u0026ndash;1465.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIttoo, A., \u0026amp; van den Bosch, A. (2016). Text analytics in industry: Challenges, desiderata and trends. \u003cem\u003eComputers in Industry\u003c/em\u003e, 78, 96\u0026ndash;107.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeontief, W. (1986). \u003cem\u003eInput-output economics\u003c/em\u003e. Oxford University Press.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, P., Zhang, P., Guo, Y., \u0026amp; Li, J. (2024). How has the relationship between major financial markets changed during the Russia\u0026ndash;Ukraine conflict?. \u003cem\u003eHumanities and Social Sciences Communications\u003c/em\u003e, 11(1), 1\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi, Y., Bandar, Z. A., \u0026amp; McLean, D. (2003). An approach for measuring semantic similarity between words using multiple information sources. \u003cem\u003eIEEE Transactions on knowledge and data engineering\u003c/em\u003e, 15(4), 871\u0026ndash;882.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatthews, R., Vivoda, V., \u0026amp; Leal, R. (2025). Deglobalization and the rehabilitation of Western defense industrial sovereignty. \u003cem\u003eComparative Strategy\u003c/em\u003e, 1\u0026ndash;27.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMowery, D. C. (2012). Defense-related R\u0026amp;D as a model for \u0026ldquo;Grand Challenges\u0026rdquo; technology policies. \u003cem\u003eResearch policy\u003c/em\u003e, 41(10), 1703\u0026ndash;1715.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePeltier, H. (2023). We get what we pay for: the cycle of military spending, industry power, and economic dependence. \u003cem\u003eCosts of War\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePloom, I., Kalvet, T., \u0026amp; Tiits, M. (2022). Defence industries in small European states: Key contemporary challenges and opportunities. \u003cem\u003eJournal of International Studies\u003c/em\u003e, 15(4).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReimers, N., \u0026amp; Gurevych, I. (2019). Sentence-bert: Sentence embeddings using siamese bert-networks. \u003cem\u003earXiv preprint arXiv:1908.10084\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eStamegna, M., Bonaiuti, C., Maranzano, P., \u0026amp; Pianta, M. (2024). The economic impact of arms spending in Germany, Italy, and Spain. \u003cem\u003ePeace Economics, Peace Science and Public Policy\u003c/em\u003e, 30(4), 393\u0026ndash;422.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eV\u0026aacute;zquez, D. (2020). Variety patterns in defense and health technological systems: evidence from international trade data. \u003cem\u003eJournal of Evolutionary Economics\u003c/em\u003e, 30(4), 949\u0026ndash;988.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang, T. P., Shyu, S. H. P., \u0026amp; Chou, H. C. (2012). The impact of defense expenditure on economic productivity in OECD countries. \u003cem\u003eEconomic Modelling\u003c/em\u003e, 29(6), 2104\u0026ndash;2114.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWassenaar Arrangement (2024). \u003cem\u003eList of Dual-Use Goods and Technologies and Munitions List\u003c/em\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Customs Organization (2012). \u003cem\u003eHS Nomenclature, 2012 Edition\u003c/em\u003e. https://www.wcoomd.org/en/topics/nomenclature/instrument-and-tools/hs_nomenclature_previous_editions/hs_nomenclature_table_2012.aspx\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang, H., Jiao, L., Li, C., Deng, Z., Wang, Z., Jia, Q., ... \u0026amp; Hu, Y. (2024). Global environmental impacts of food system from regional shock: Russia-Ukraine war as an example. \u003cem\u003eHumanities and Social Sciences Communications\u003c/em\u003e, 11(1), 1\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu, X., Yang, X., Huang, Y., Guo, Q., \u0026amp; Zhang, B. (2020). Measuring similarity and relatedness using multiple semantic relations in WordNet. \u003cem\u003eKnowledge and Information Systems\u003c/em\u003e, 62(4), 1539\u0026ndash;1569.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e A detailed description of these products is available on the website of the statistics division of the UN at:\u003c/span\u003e\u003cdiv id=\"Par46\" class=\"Para\"\u003etttps://unstats.un.org/unsd/classifications/Family/Detail/32.\u003c/div\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e While the analysis above focuses on ML categories, the same framework can be applied from the perspective of HS codes. This alternative view would allow for the identification of HS products that are related to multiple ML categories and can therefore be interpreted as bridges across different categories.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Figures","content":"\u003cp\u003eFigures 1 to 8 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"humanities-and-social-sciences-communications","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"palcomms","sideBox":"Learn more about [Humanities \u0026 Social Sciences Communications](http://www.nature.com/palcomms/)","snPcode":"41599","submissionUrl":"https://submission.springernature.com/new-submission/41599/3","title":"Humanities and Social Sciences Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Defence Readiness, Revealed Comparative Advantage, Defence Industry, Semantic Similarity","lastPublishedDoi":"10.21203/rs.3.rs-8785807/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8785807/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis paper studies defence‑related industrial ecosystems defining an indicator that tries to capture the \u0026ldquo;defence readiness\u0026rdquo; of the industrial specialization of economies, focusing on how the war in Ukraine has affected this indicator for some specific countries. This index builds on a Category Space framework that adapts Hidalgo et al. (2007) product space methodology. The approach proposed combines semantic similarity measures with Revealed Comparative Advantage (RCA) data to quantify the relatedness between Munitions List (ML) categories and Harmonized System (HS) product codes, allowing the assessment of how national productive structures connect to defence‑related goods. The resulting networks capture the density and composition of each country\u0026rsquo;s defence‑related industrial ecosystem, showing how ready is each economy to adapt to an increase in military spending. The analysis presented here focuses on the evolution of this indicator around the world between 2021, just before the starting of the war, and 2024, where the most recent data are available. The findings of this paper show how among the countries directly involved in the war, Ukraine presents a modest expansion of its network, with new HS codes emerging in 2024, including items associated with munitions. On the other hand, Russia and Belarus show a contraction in their networks over the period. At the same time, the analysis reveals how the United States, and particularly China maintained and even increased dense and diversified networks before and after the outbreak of the war. These results provide a representation of defence‑related industrial capacity and illustrate how the underlying productive structures evolve during the period of the war in Ukraine. This novel Category Space framework and the definition of this defence readiness indicator offer new tools for comparing how different economies are positioned to respond to shifts in defence demands.\u003c/p\u003e","manuscriptTitle":"Defence Readiness: how to measure it across industrial ecosystems and how it has been impacted by the War in Ukraine","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-20 07:20:07","doi":"10.21203/rs.3.rs-8785807/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-17T15:15:01+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"151431813452283039102419513660526347167","date":"2026-05-12T07:19:56+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-05-12T02:04:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-05-12T01:52:27+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-03-02T10:46:00+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-19T09:54:35+00:00","index":"","fulltext":""},{"type":"submitted","content":"Humanities and Social Sciences Communications","date":"2026-02-19T09:49:27+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"humanities-and-social-sciences-communications","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"palcomms","sideBox":"Learn more about [Humanities \u0026 Social Sciences Communications](http://www.nature.com/palcomms/)","snPcode":"41599","submissionUrl":"https://submission.springernature.com/new-submission/41599/3","title":"Humanities and Social Sciences Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Nature AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"5eaa7d29-0748-45bb-80e1-b918bc628941","owner":[],"postedDate":"February 20th, 2026","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-17T15:15:01+00:00","index":60,"fulltext":""},{"type":"reviewerAgreed","content":"151431813452283039102419513660526347167","date":"2026-05-12T07:19:56+00:00","index":59,"fulltext":""},{"type":"reviewersInvited","content":"8","date":"2026-05-12T02:04:31+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-05-12T01:52:27+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":63191355,"name":"Earth and environmental sciences/Environmental social sciences"},{"id":63191356,"name":"Physical sciences/Mathematics and computing"}],"tags":[],"updatedAt":"2026-05-12T02:08:31+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-20 07:20:07","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8785807","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8785807","identity":"rs-8785807","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}