\documentclass[12pt,letterpaper]{article} \pdfminorversion=6 \usepackage[top=1.0in, bottom=1in, left=1in, right=1in]{geometry} \usepackage{amssymb} \usepackage{amsmath} \usepackage{graphicx} \usepackage{tabularx} \usepackage{subfloat} \usepackage{setspace} \usepackage{rotating} \usepackage{booktabs} \usepackage{longtable} \usepackage[colorlinks,citecolor=black,linkcolor=black,urlcolor=blue]{hyperref} \usepackage{xcolor} \usepackage[final]{pdfpages} \usepackage[normalem]{ulem} \usepackage{threeparttable} \usepackage[T1]{fontenc} \usepackage[utf8]{inputenc} \usepackage[active]{srcltx} \usepackage{amstext} \usepackage{natbib} \usepackage{lscape} \usepackage[titletoc]{appendix} \usepackage[justification=centering]{caption} \newcommand{\comment}[1]{} \interfootnotelinepenalty=10000 \tolerance=20000 \newcommand{\PaperTitle}{Determinants of Intermediate Goods Trade} \begin{document} \thispagestyle{empty} { % set font to helvetica (arial) to make it 508-compliant \fontfamily{phv}\selectfont \begin{center} {\LARGE \textbf{\PaperTitle}} \\ \vspace{0.25in} \vspace{1in} {\Large Lin Jones} \\\vspace{.25in} {\Large Serge Shikher} \\\vspace{.25in} {\Large Yoto V. Yotov} \vspace{1in} {\large ECONOMICS WORKING PAPER SERIES}\\ Working Paper 2025--12--A \\ \vspace{0.5in} U.S. INTERNATIONAL TRADE COMMISSION \\ 500 E Street SW \\ Washington, DC 20436 \\ \vspace{0.5in} December 2025 \\ \end{center} \vfill \vspace{1em} \noindent{\small The views expressed solely represent the opinions and professional research of the authors. The content of the working paper is not meant to represent the views of the U.S. International Trade Commission, any of its individual Commissioners, or the United States government. } \newpage \thispagestyle{empty} % remove headers, footers, and page numbers from cover page \begin{flushleft} \PaperTitle \\ Lin Jones, Serge Shikher, and Yoto V. Yotov \\ Economics Working Paper 2025--12--A\\ December 2025 \\~\\ \end{flushleft} \vfill \begin{abstract} \noindent This paper examines the impact of various determinants of trade, including geopolitical risks, on global supply chains using a structural gravity model and the newly constructed \emph{International Trade and Production Database by End Use} (ITPD-U), which disaggregates trade into intermediate, capital, and consumer goods. Results show that intermediate goods trade responds differently to standard trade determinants compared to capital and consumer goods: distance is the largest barrier, while regional agreements and currency unions have weaker effects. Geopolitical risks reduce trade overall, though its impact is notably smaller for intermediate goods, reflecting the ``stickiness'' of global supply chains. These results highlight the relatively high trade costs and structural inertia characterizing intermediate goods trade amid geopolitical uncertainty. \end{abstract} \vfill \begin{flushleft} Lin Jones\\ Country and Regional Analysis Division\\ Office of Economics\\ U.S. International Trade Commission\\ \vspace{2em} Serge Shikher\\ Research Division\\ Office of Economics\\ U.S. International Trade Commission\\ gravity@usitc.gov \\ \vspace{2em} Yoto V. Yotov\\ Drexel University, CGPA, ifo, CESifo\\ \vspace{2em} Suggested Citation:\\ \vspace{0.2cm} Jones, L., S. Shikher, and Y. V. Yotov (2025). "\PaperTitle" U.S. International Trade Commission. Office of Economics Working Paper 2025-12-A. \end{flushleft} } % end of helvetica (arial) font \comment{ \title{Geopolitical Risk and Global Supply Chains: \\ Evidence from a New Database} \author{\quad\\ Lin Jones\thanks{United States International Trade Commission (USITC).} \and \quad\\ Serge Shikher\footnotemark[1] \and \quad\\Yoto V.\ Yotov\thanks{School of Economics and Center for Global Policy Analysis (CGPA) Drexel University, ifo institute.}} \date{\quad\\\quad\\\today} \maketitle \vspace{-.1cm} \noindent \textbf{Abstract.} This paper examines the impact of geopolitical risks on global supply chains using a structural gravity model and the newly constructed \emph{International Trade and Production Database by End Use} (ITPD-U), which disaggregates trade into intermediate, capital, and consumer goods. Results show that intermediate goods trade responds differently to standard trade determinants compared to those observed in capital and consumer goods: distance is the largest barrier, while regional agreements and currency unions have weaker effects. Geopolitical risks reduce trade overall, though its impact is notably smaller for intermediate goods, reflecting the ``stickiness'' of global supply chains. These results highlight the relatively high trade costs and structural inertia characterizing intermediate goods trade amid geopolitical uncertainty. } % End of comment \clearpage \doublespacing \section{Introduction} Global supply chains (GSCs) have transformed international trade, enabling firms to optimize production across borders and integrate efficiently into global markets. However, geopolitical risks---such as potential economic sanctions, trade restrictions, blockades, and regional conflicts---pose significant threats to the stability of these interconnected networks. The prospect of such disruptions creates uncertainties that affect investment decisions and trade flows. Therefore, understanding the impact of geopolitical risks on GSCs is critical for policymakers and businesses seeking to mitigate vulnerabilities, safeguard economic stability, and reinforce national security. One of the primary transmission channels through which geopolitical risks affect GSCs is trade in intermediate products. These goods comprise about 60 percent of global merchandise trade today, and serve as essential inputs in international production networks \citep{WTO2023}. With high frequency of cross-border movements, intermediate good trade is expected to be particularly susceptible to disruptions triggered by geopolitical shocks. Any disruption in intermediate goods trade flows can generate widespread ripple effects, propagating across global supply chains, and impacting countries, industries, businesses, and consumers worldwide. To quantitatively assess the impact of geopolitical risks on trade flows and global supply chains, we use the workhorse empirical trade model---the gravity equation---and we apply the latest recommendations for gravity estimations from \cite{Larchetal2025}, as summarized in Section \ref{methods}. One of the most important contributions of this paper, which enables us to focus on GSCs and intermediate goods, is the construction of a new database---the \emph{International Trade and Production Database by End Use} (ITPD-U).\footnote{The database and its technical documentation \citep{JonesShikherYotov2025b} will be available online, on the USITC's Gravity Portal at gravity.usitc.gov.} The ITPD-U distinguishes international and domestic trade across three categories of end use—intermediate, capital, and consumer goods—as follows: \begin{itemize} \item Intermediate goods are products used as inputs in the production of other goods and services. They include raw materials, components, and semi-finished products that require further processing before reaching the final consumer. Examples include steel used in car manufacturing, semiconductor chips for smartphones, and textiles for apparel production. \item Capital goods consist of physical assets used by businesses to produce goods and services. These goods are not directly consumed but serve as tools or infrastructure in production. Examples include machinery and equipment. \item Consumer goods are finished products intended for direct consumption by individuals or households. Examples include household appliances, food, and clothing. These goods require no further processing before reaching end users. \end{itemize} The newly constructed ITPD-U offers a unique opportunity to examine the potentially heterogeneous effects of standard trade determinants across end-use categories, and more specifically, to assess the impact of geopolitical risks on GSCs and intermediate goods trade. Additionally, it enables us to use a gravity model approach to study how the standard determinants of trade flows differently affect consumer, intermediate, and capital goods. Our key findings can be summarized as follows: First, we find that intermediate goods trade responds differently to standard trade determinants and policy variables, compared to those observed in capital and consumer goods: Distance has the strongest effect on intermediate goods trade. As distance is a primary proxy and determinant for bilateral trade costs, this result suggests that bilateral trade costs for intermediate goods for a given pair of countries are significantly higher than those for the other categories. Meanwhile, contiguity---another geographic variable indicating shared borders---shows the smallest positive effect on intermediate goods trade. This may reflect the fact that intermediate goods often move within complex supply chains, where proximity alone does not guarantee trade integration. It also suggests that trade in intermediates depends more heavily on infrastructure, logistics, and production network linkages than simple geographic closeness. EU membership has a strong positive effect on consumer goods trade, a modest effect on intermediate goods trade, and a statistically insignificant impact on capital goods trade. Participation in currency unions and regional trade agreements benefits consumer goods trade the most and intermediate goods trade the least. On the other hand, WTO accessions and sanctions have a disproportionately high effect on capital goods trade (positive and negative, respectively), while their impact on consumer and intermediate goods trade is more muted. Second, we find that geopolitical risks reduce trade flows overall, though its negative effect is smaller on intermediate goods trade. We measure geopolitical risks using several different indices. The Fraser's Economic Freedom Index (EFI) and the World Bank's Worldwide Governance Indicator (WGI) are positively associated with trade, while the Political \& Social Value Distance (PSVD) index, that we construct, has a negative effect, indicating that higher geopolitical risk is associated with lower trade volumes. Notably, the impact of these risk indicators is smaller for intermediate goods trade than for consumer goods, a finding that aligns with existing theories and empirical evidence on global supply chain ``stickiness'' \citep{Grossmanetal2024}. Finally, we test the popular Geopolitical Distance (GD) index that is based on the United Nation (UN) General Assembly voting data. However, we obtain counterintuitive results—its estimated impact on trade is positive and statistically significant across all goods categories in our sample. This result confirms concerns about the validity of using this variable to predict bilateral economic relationships and highlights the need for caution when employing the UN voting data as a proxy for geopolitical alignment \citep{Broneretal2025}. The rest of this paper is structured as follows. Section \ref{literature} provides a literature review on the economic impact of geopolitical risks and the resistance of global supply chains to changes. Section \ref{data} discusses the data used in this paper and their sources, with a particular focus on the ITPD-U and geopolitical risk measures. Section \ref{first_look} highlights several global trade patterns by end use, informed by the newly constructed ITPD-U. Section \ref{methods} outlines the econometric methods employed in the analysis. Section \ref{results} summarizes the estimation results, offering insights on the effects of standard trade determinants as well as geopolitical risks on global trade and supply chain dynamics. Finally, Section \ref{conclusion} concludes the paper by highlighting key findings and outlining directions for future research and methodological advancements. \section{Literature Review \label{literature}} While most studies of global supply chain risks focus on logistical disruptions \citep{FRBNewYork2025, Arvisetal2024}, recent geopolitical events---such as Russia's invasion of Ukraine and U.S.-China trade tension---are found to have had a growing impact on global supply chains \citep{Fajgelbaum2023, Freund2024, Luo2025}. As a result, there are growing interests in investigating the effect of geopolitical risks. Geopolitical risks include adverse geopolitical events such as political instability, conflicts, sanctions, trade disputes, and policy shifts that affect economic performance, financial markets, and supply chains \citep{SPGlobal2025}. \cite{CaldaraLacoviello2018} characterize geopolitical risks as both the threat and realization of such adverse geopolitical events. Applying the Geopolitical Risk (GPR) Index---a news-based measure that they constructed,\footnote{The Geopolitical Risk (GPR) Index, developed by \cite{CaldaraLacoviello2022}, measures geopolitical risk based on the relative frequency of keywords related to geopolitical tensions—such as “war,” “terrorist threat,” and “military conflict”—appearing in leading international newspapers. The global GPR Index is constructed using articles from major U.S. and U.K. publications including The New York Times, The Wall Street Journal, and The Financial Times, with monthly coverage beginning in 1985. Country-specific GPR indices are available for 44 countries, drawing on both international and selected local news sources. Full methodology and data are available at: https://www.matteoiacoviello.com/gpr.htm.} \cite{CaldaraLacoviello2022} find that heightened geopolitical risk is associated with reduced investment and employment, increased disaster probability, and greater downside risks to the global economy. Notably, both the threat and actual occurrence of geopolitical events contribute to economic downturns. Also using the GPR index, \cite{Buptaetal2019} demonstrate that elevated geopolitical risks lead to declining trade volumes between countries, while \cite{ThakkarAyub2022} note that the adverse effects of geopolitical uncertainty are more pronounced for foreign direct investment (FDI) than for trade. Most recently, \cite{MulabdicYotov2025} show that geopolitical risks lead to significant reduction of international trade by 30 to 40 percent, and that services and agriculture sectors are particularly vulnerable to these risks. Economic policy uncertainty, driven by the threat of adverse geopolitical events such as sanctions or trade protectionism, is a key channel through which geopolitical risks affect the global economy. Using the Economic Policy Uncertainty (EPU) Index, also a news-based measure developed by \cite{Bakeretal2016},\footnote{The Economic Policy Uncertainty (EPU) Index, developed by \cite{Bakeretal2016}, quantifies policy-related economic uncertainty based on the frequency of specific keywords—such as “economic,” “policy,” and “uncertainty”—in major newspaper articles. The original U.S. index draws from ten leading U.S. newspapers, with data available monthly from 1985 onward. The EPU framework has since been extended to over 20 countries using local-language news sources. The index captures both domestic and global policy-related uncertainty and is widely used in empirical macroeconomic research. Methodological details and data access are available at: https://www.policyuncertainty.com.} \cite{Constantinescuetal2019} find that economic policy uncertainty significantly hinders trade growth, while \cite{KimLee2024} show it discourages greenfield FDI inflows. A related body of literature examines the effects of trade policy uncertainty (TPU)---a source of economic policy uncertainty---on aggregate trade flows. Synthesizing multiple measures developed in the literature, \cite{HandleyLimao2017, HandleyLimao2022} analyze specific events that increase or reduce trade policy uncertainty, and investigate the variations in risk exposure and associated impacts across industries and countries. Their findings indicate that uncertainty alone about tariff or trade agreements even without realized policy changes reduces exports, investment, employment, and output. It amplifies trade cost and affects the global distribution of production and trade. Some industries, such as those with high fixed costs, long planning horizons, or reliance on global markets, are more sensitive to risk exposure and thus react more strongly to changes in uncertainty than others. \cite{Helbigetal2021} conduct a comprehensive systematic review of 88 supply risk assessment methods, and identify political instability and regulatory uncertainty as top risk factors affecting the supply of critical minerals—an upstream industry in global supply chains. These studies often utilize indices such as the Country Political Risk Index or the World Governance Index to measure political instability. In addition, they employ metrics such as the Policy Perception Index to gauge regulatory uncertainty arising from policy shifts and regulation changes. These risks are micro-level reflections of broader geopolitical trends affecting global trade and supply chain resilience. \cite{BosoneStamato2022} find that geopolitical distance---the degree of political alignment or divergence between trading partners, measured through UN General Assembly voting patterns---has increasingly acted as a trade friction, reducing trade volumes over time. The study provides evidence of friend-shoring, where countries prioritize trade with politically aligned partners. While a rich body of literature examine the effects of geopolitical risk on the global economy, trade, foreign investment, and industries, relatively fewer studies explicitly explore its impact on global supply chains. However, some research investigates how global supply chains respond to negative shocks such as tariff changes, trade restrictions, or geopolitical instability. \cite{Grossmanetal2024} find that global supply chains are slow to adjust to negative shocks due to the stickiness of supply chain arrangements and customer-supplier relationships. Firms frequently are reluctance to switch suppliers due to high sunk costs, long-term contractual commitments, and customized production dependencies, which make rapid supplier replacement both impractical and costly. \cite{Martinetal2023} note that trade relationship stickiness varies by product type. Firms dealing with products exhibiting high stickiness---such as customized or specialized inputs---are less likely to switch suppliers due to challenges such as identifying alternative sources, renegotiating long-term contracts, managing relationship-specific investments, and adjusting production integration. As a result, industries reliant on stickier inputs are more vulnerable to tariff changes, geopolitical risks, and supply chain shocks. The existing literature on global supply chain stickiness highlights the importance of assessing the impact of geopolitical risks on global supply chains through trade in intermediate goods. However, there is a limited body of economic research that analyzes trade policy through the lens of end-use classifications---such as intermediate, capital and consumer goods. Among the few, \cite{goldberg2016} note that the positive relationship between trade liberalization and economic growth is primarily driven by tariff reductions on intermediate inputs and capital goods, rather than consumer goods. More recently, \cite{amiti2020} show that tariffs on intermediate goods can lead to significant cost pass-through to domestic producers, while consumer goods tariffs more directly affect retail prices. This work reveals that trade policy impacts could be highly heterogeneous across these end use categories, underscoring the importance of disaggregating trade data by end-use to fully understand the transmission mechanisms and welfare implications of trade policy. Building on prior studies, we hypothesize that intermediate goods—such as raw materials and components used in production—responds differently to geopolitical uncertainties compared to final goods, including capital and consumer products. However, the extent of this impact likely vary across industries, depending on the stickiness of their production and input sourcing arrangements, which affect firms' ability and willingness to adjust supply chains in response to external disruptions. \section{Data and Sources \label{data}} To perform the econometric analysis of the impact of geopolitical risks on global supply chains and intermediate goods trade, we use the following data: (i) trade data by end use; (ii) data on geopolitical risk indicators; and (iii) data on other gravity and policy variables. We describe each of the three data sets along with the corresponding sources below.\\\vspace{-.5cm} \subsection{International Trade and Production Database by End Use (ITPD-U)} \noindent Recognizing that the impact of geopolitical risk on global supply chains primarily occurs via trade in intermediate goods, we constructed the \emph{International Trade and Production Database by End Use} (ITPD-U), a novel trade dataset that disaggregates domestic and bilateral merchandise trade into three end-use categories: intermediate goods, capital goods, and consumer goods \citep{JonesShikherYotov2025b}. The ITPD-U builds upon the \emph{International Trade and Production Database for Estimation} (ITPD-E), a database originally developed by \cite{Borchertetal2021a}. The third release of the ITPD-E covers 264 countries, 36-year period from 1986 to 2022,\footnote{The year coverage of 1986-2022 for agriculture, 1988-2022 for manufacturing, mining, and energy, and 2000-2022 for services} and 170 industries in all broad sectors of the economy, including 28 industries in agriculture, 7 in mining and energy, 120 in manufacturing, and 17 in services. \citep{LarchShikherYotov2025}.\footnote{Both ITPD-U and ITPD-E are available on USITC's Gravity Portal at gravity.usitc.gov.} The ITPD-U has the same coverage as ITPD-E in terms of years and countries. It only covers the 153 goods industries out of 170 industries in ITPD-E. ITPD-U expands the goods sections of the ITPD-E by three end-use categories. To construct the ITPD-U, we first disaggregate bilateral merchandise trade data from UN Comtrade into three end-use categories using the \emph{End Use Concordance} \citep{JonesBethmann2026}, described in the paragraphs below. Then we aggregate trade data by end use and map it to align with the industry classification used in the ITPD-E. UN Comtrade classifies merchandise trade data at the Harmonized System (HS) 6-digit level. Although the HS is a standard classification system for traded goods, it does not provide information on how these goods are used---whether they serve as intermediate inputs in production processes, function as capital assets that support production without becoming part of the final output, or are intended for use by final consumer. Therefore, we rely on the \emph{End Use Concordance} to assign each of traded goods at the HS 6-digit level (HS-6) to one of the four end-use categories: capital goods (CAP), consumer goods (CONS), intermediate goods (INT), and mixed-use goods (MIXED), based on its primary economic function. Before doing so, we first conduct a comprehensive review and update on the \emph{End Use Concordance}, a dataset originally produced for the construction of the Asia-Pacific Economic Cooperation (APEC) Trade-in-value-added (TiVA) database \citep{Joneswang2019}. Consistent with the UN Broad Economic Categories (BEC) framework, the \emph{End Use Concordance} incorporates input data from various sources, including the Organisation for Economic Co-operation and Development (OECD), the United Nation Statistics Division (UNSD), the World bank, and the World Input-Output Database (WIOD). The first version of the \emph{End Use Concordance} was incorporated into the UN BEC rev. 5 (2019).\footnote{UNSD, ``BEC 5 Correlations (2019),'' https://unstats.un.org/unsd/trade/classifications/BEC5Correlations.zip} This paper uses the 2024 revision. For many traded goods, the end use is unique and can be clearly assigned to a specific category. For instance, 100 percent of traded goods under HS 220410---sparkling wine---are classified as consumer goods, as this group of goods is primarily intended for consumer consumption. In contrast, certain goods serve multiple functions with both industrial and consumer applications. Accordingly, these vehicles are classified as mixed-use goods and split across multiple end use categories according to estimated shares. These estimated shares are derived from industry sources, such as industry reports, expert assessments, and data from trade associations. For example, traded goods under HS 870380---vehicles with only electric motor for propulsion (EVs)---can be either used by firms as capital goods to support production operations, or driven by individuals or households as consumer goods. Accordingly, EVs are classified as mixed-use goods. Based on available industry data such as the Vehicle Inventory and Use Survey (VIUS), it is estimated that approximately 87.5 percent of EVs are used for consumer purposes, while 12.5 percent are used as capital goods in commercial fleets (Table \ref{concordance}).\footnote{U.S. Census Bureau, ``2021: ECNSVY Vehicle Inventory and Use Survey All Vehicles,'' https://data.census.gov/table/VIUSA2021.VIUS212A?d=ECNSVY+Vehicle+Inventory+and+Use+Survey+All+Vehicles} For more information on the methodology on estimating the splits, see \cite{JonesBethmann2026}. The \emph{End Use Concordance} (revision 2024) includes over 400 such mixed-use goods, each assigned estimated shares across these three primary end-use categories. Nearly 99 percent of these mixed-use goods are assigned to two end-use categories. Only about 1 percent are assigned to three end-use categories, including those under special chapter 99 which does not describe specific products. \begin{table}[!ht] \centering \caption{\label{concordance}Illustration of the USITC End Use Concordance} \begin{tabular}{|l|l|l|r|r|r|} \hline \textbf{HS6} & \textbf{Description} & \textbf{End Use} & \textbf{CAP} & \textbf{CONS} & \textbf{INT} \\ \hline 220410 & Sparkling wine & CONS & 0.00 & 1.00 & 0.00 \\ 281122 & Silicon dioxide & INT & 0.00 & 0.00 & 1.00 \\ 840110 & Nuclear reactors & CAP & 1.00 & 0.00 & 0.00 \\ 870380 & Vehicles with electric motor for propulsion & MIXED & 0.125 & 0.875 & 0.00 \\ \hline \end{tabular} \begin{tablenotes} \footnotesize \item Source: USITC End Use Concordance (2024 revision). \end{tablenotes} \end{table} Bilateral merchandise trade data from UN Comtrade is first disaggregated into three end-use categories using the \emph{USITC End Use Concordance}. Trade value of mixed-use goods are distributed to respective end-use categories based on their estimated splits. These data are then mapped and aggregated to align with the industry classification used in the ITPD-E. To estimate domestic trade (or production) by end use, we apply the proportionality derived from total exports---under the assumption that a country's export product composition mirrors its domestic production \citep{saltarelli2020export}. This means that the division of domestic trade in country $i$, year $t$, industry $k$ into the three end uses is done using the same proportions as country $i$, year $t$, industry $k$ total exports. The resulting dataset, ITPD-U, is a novel contribution that provides harmonized information on both international trade and domestic production by end-use category. Further details on the ITPD-U construction methodology are provided in the technical appendix available on the USITC's Gravity Portal \citep{JonesShikherYotov2025b}. The ITPD-U offers several advantages for global supply chain analysis. First, the ITPD-U retains the expansive country, industry, and year coverage of the ITPD-E. Second, the ITPD-U includes both domestic and bilateral trade flows, which is crucial for mapping and identifying supply chain dependencies. Third, like the ITPD-E, the ITPD-U is constructed from raw data without relying on statistical modeling, making it well-suited for econometric estimations and empirical analysis.\footnote{ITPD-E is constructed using four primary data sources, each covering distinct industry sectors: For agriculture, trade and production data come from the Food and Agriculture Organization of the United Nation (FAO) statistical database. For manufacturing and Mining and Energy, trade data are obtained from the UN Comtrade Database, while production data come from INDSTAT, maintained by the United Nations Industrial Development Organization (UNIDO). ITPD-E also includes services industries, which are not included in ITPD-U. See \cite{Borchertetal2021a} and \citep{LarchShikherYotov2025} for further details on the construction and features of the ITPD-E.} Fourth, disaggregating trade flows by end use allows us to study global supply chains more effectively, as trade in intermediate goods tends to respond differently to various trade determinants than trade in other types of goods. \\\vspace{-.5cm} \subsection{Geopolitical Risk Measures} We employ multiple indicators to capture different dimensions of geopolitical risks at both the national and bilateral levels. These include country-specific measures of economic and political risks, as well as bilateral geopolitical risks specific to trading partner pairs \citep{Jones2026}. While GPR and EPU indices are commonly used in the related literature, we opt not to use them in our analysis due to limited country coverage, high monthly volatility, and potential media bias. To measure country-specific economic risks, we rely on the Economic Freedom Index (EFI), published annually by the Fraser Institute. This composite index is built on five core areas---each representing a distinct dimension of a country’s economic conditions that are influenced by related policies \citep{Gwartneyetal2024}. We do estimations with both the overall EFI and its individual components: \begin{itemize} \item Size of government. It measures government consumption, transfers, subsidies, and the extent of state-owned enterprises. Countries with smaller government, lower marginal tax, and less state ownership of assets earn higher ratings in this component. \item Legal system and property rights. It assesses the rule of law, judicial independence, impartial courts, and protection of property rights. Strong legal institutions are essential for secure economic transactions. \item Sound money. It evaluates inflation levels, money growth, and freedom to own foreign currency. Stable monetary policy and low inflation are key to preserving purchasing power and economic predictability. \item Freedom to trade internationally. It captures tariffs, trade barriers, exchange rate controls, and capital market restrictions. Greater openness to international trade is associated with higher economic freedom. \item Regulation. It examines credit market regulations, labor market flexibility, and business regulations. Countries with fewer and less burdensome restrictions earn higher ratings in this component. \end{itemize} Each area is scored from 0 to 10, with higher scores indicating greater economic freedom. The resulting composite EFI has a similar score ranging from 1 and 10 \citep{fraserdataset2024}. A higher value indicates a more favorable economic environment, and thus, lower economic risks.\footnote{Fraser Institute, ``Economic Freedom,'' accessed February 24, 2025.} To measure country-specific political risks, we use the World Bank’s Worldwide Governance Indicators (WGI), which evaluate a country’s institutional and governance quality based on six key dimensions: voice and accountability, control of corruption, government effectiveness, political stability and absence of violence or terrorism, rule of law, and regulatory quality \citep{Kaufmannetal2010}. The WGI database covers over 200 countries from 1996 to 2022. The scores ranging from -2.5 to 2.5 \citep{wgi2024}. Higher values indicate stronger governance, and, consequently, lower political risk.\footnote{World Bank, ``Worldwide Governance Indicators,'' accessed February 24, 2025.} To quantify bilateral geopolitical risks, we construct two measures---the Political and Social Value Distance (PSVD) and Geopolitical Distance (GD)---to assess the degree of alignment or divergence between countries based on their democratic conditions or foreign policy stance. These measures serve as a quantitative proxy for geopolitical risks in bilateral trade relations. To construct GD and PSVD measures, we employ Euclidean distance between an array of country-specific measures, which calculates the degree of dissimilarity between countries based on their respective values. The first measure, the PSVD, is derived from the Freedom in the World index (FIW), published annually by the Freedom House. This dataset provides a comprehensive assessment of democratic conditions across 210 countries and territories worldwide. The FIW index is composed of two main categories: Political Rights (scored out of 40) and Civil Liberties (scored out of 60). Political Rights includes three subcategories: the electoral process, political pluralism and participation, and the functioning of government. Civil Liberties is divided into four subcategories: freedom of expression and belief, associational and organizational rights, rule of law, and personal autonomy and individual rights. The FIW scores range from 0 to 100, with higher values indicating higher political freedom \citep{freedomhouse2024fiw}.\footnote{Freedom House, ``Freedom in the World 2025: A Global Survey of Political Rights and Civil Liberties,'' accessed January 22, 2025.} The PSVD is calculated based on the Euclidean distance between countries in both Political Rights and Civil Liberties scores \citep{Jones2026}. The second measure, the GD, is derived from the United Nations General Assembly (UNGA) Voting Data, which contains ideal point estimates based on countries’ voting patterns in UNGA sessions from 1946 to 2023. These ideal points range from 3.2 to -3.2, reflecting a country's foreign policy preferences and indicating its degree of alignment with the U.S.-led liberal order \citep{baileyetal2017}.\footnote{The ideal point estimates are estimated by sessions instead of calendar year in the dataset. This paper converts them into the yearly data by taking the average of three consecutive sessions for a given year.} The GD is calculated using the Euclidean distance between countries' ideal point estimates. The UN voting-based GD measure has been increasingly used in international economic literature for estimating the impact of political relations on trade\citep{airaudoetal2025,Freund2024,BosoneStamato2022}. However, empirical findings have been mixed, as UN voting patterns reflect a range of strategic, ideological, and contextual factors that vary with the nature of individual resolutions.\\\vspace{-.5cm} \subsection{Other Gravity and Policy Variables} We employ several other gravity and policy variables from various sources. Specifically, we rely on the USITC's \emph{Dynamic Gravity Dataset} (DGD) for data on the standard gravity variables, including bilateral distance, contiguous borders, common official language, and colonial relationships \citep{GurevichHerman2018}.\footnote{The dataset is available at https://www.usitc.gov/data/gravity/dgd.htm} In addition, we use the DGD for data on membership in the European Union (EU) and the World Trade Organization (WTO). For data on regional trade agreements (RTAs), we use Mario Larch's \emph{Regional Trade Agreements Database} \citep{EggerLarch2008}.\footnote{https://www.ewf.uni-bayreuth.de/en/research/RTA-data/index.html} Finally, data on sanctions are from the latest edition of the \emph{Global Sanctions Database} (GSDB) \citep{Felbermayretal2020EER, Syropoulosetal2024}.\footnote{https://www.globalsanctionsdatabase.com/} \section{Global Trade by End Use\label{first_look}} The newly developed ITPD-U database provides a novel lens that allows us to analyze international trade by three end-use categories. Intermediate goods constitute the largest share of global trade. On average, they represent approximately 60 percent of total merchandise trade across all years covered in the ITPD-U. This share increased gradually from 56 percent in 1988 to 64 percent in 2008. Following the global financial crisis, the share exhibited modest fluctuations, reaching 65 percent in 2022—the most recent year in the database (see Figure \ref{fig:shares}). The share of capital goods in global trade declined steadily from 19.6 percent in 1988 to 14.5 percent in 2013. While it experienced a modest rebound between 2015 and 2020, the downward trend resumed thereafter, reaching 14 percent in 2022. In contrast, the share of consumer goods fluctuated between 21 percent and 26 percent over this period. On average, it was slightly higher during the 1990s compared to more recent years. \begin{figure} \caption{International trade by end use, shares in total} \centering \includegraphics[width=0.7\linewidth]{Figure.PNG} \label{fig:shares} \caption*{Source: International Trade and Production Database by End Use (ITPD-U)} \end{figure} The end-use composition of exports and imports varies across countries. In the case of the United States, exports are predominantly composed of intermediate goods, whose share rose from just over 50 percent to more than two-thirds in recent decades. The share of capital goods in U.S. exports declined from around 25 percent to about 16 percent, while the share of consumer goods remained relatively stable at around 16 percent. The composition of U.S. imports exhibited little change between 1988 and 2022: intermediate goods accounted for about 50 percent, consumer goods for 31 percent, and capital goods for about 19 percent. The U.S. trade deficit is heavily concentrated in consumer goods, which made up 60 percent of the total deficit in 2022. Intermediate goods contributed 17 percent to the deficit in that year—a notable shift from the 1990s, when trade in intermediates was nearly balanced before turning into a deficit in 2000. Similarly, trade in capital goods was close to balanced throughout the 1990s but has since shifted into deficit, comprising 24 percent of the overall trade deficit in 2022. Zooming in on China, we find that the share of intermediate goods in China's exports declined from 55\% in 1988 to 30\% in 1993 and then slowly increased to 48\% in 2022. The share of consumption goods was increasing between 1988 and 1993, from 43\% to 62\%. After 1993 it went on a slow decline, falling to 31\% in 2022. The share of capital goods steadily increased from 2\% in 1988 to 21-23\% in 2020-2022. The composition of China's imports has also changed over the years, but not as much as the composition of China's exports. The share of intermediate goods in imports has increased from around 70\% in the late 1980s and early 1990s to 75-80\% after 2005. The share of consumption goods was 5-10\% from 1988 until 2013. In more recent years it grew to 10-13\%. The share of capital goods was 20-30\% from 1988 to 2004 and then declined to around 10\% in 2022. The ITPD-U also provides new insights into bilateral trade patterns. An example is the trade patterns between the United States and China, where the end-use composition of imports and exports differs substantially. The United States primarily exports intermediate goods to China, while it imports mostly consumer goods. In 2022, 73 percent of U.S. exports to China consisted of intermediate goods, followed by 15 percent in capital goods, and 12 percent in consumer goods. Over the years, the share of consumer goods in U.S. exports increased modestly, while the share of capital goods in U.S. exports declined. On the other side, consumer goods led China’s exports to the United States in 2022, accounting for 48 percent of the total, followed by 30 percent in intermediate goods and 22 percent in capital goods. Notably, the composition of China’s exports to the United States has shifted in recent decades: the share of consumer goods has gradually declined, whereas intermediate and capital goods have gained more prominence. These examples demonstrate that the ITPD-U enables us to gain useful insights into trade patterns based on different end uses. In the following section, we perform a systematic analysis of the determinants of international trade for these different types of goods using the ITPD-U. \section{Econometric Methods\label{methods}} We obtain our estimates using the ITPD-U covering for 264 countries from 1986 to 2022 with the workhorse model of trade --- the gravity equation, and we apply the recent recommendations for gravity estimations of \cite{Larchetal2025}. Specifically, we estimate alternative specifications of the following econometric model:\footnote{Our econometric specification is consistent with a wide class of theoretical foundations, which, subject to parameter interpretation, lead to isomorphic gravity equations \citep{Arkolakisetal2012}.} \footnotesize \begin{eqnarray} X^k_{ij,t} &\!\! = \!\! & \exp[RISK_{ij,t}\times\beta^k+POLICY_{ij,t}\times\alpha^k+GRAV_{ij}\times\gamma^k+GLOB^k_{ij,t}+\pi^k_{i,t}+\chi^k_{j,t}]\times\epsilon^k_{ij,t}.\label{est_main} \end{eqnarray}\normalsize $X^k_{ij,t}$ denotes nominal trade flows from exporter $i$ to importer $j$ in industry $k$ at time $t$, i.e., the dependent variable includes directional trade (exports and imports) at the industry level at each point of time. Following \cite{Eggeretal2022}, we use data for consecutive years, instead of data with intervals as suggested by \cite{ChengWall2005}. $X^k_{ij,t}$ includes both cross-border/international and domestic/internal trade flows, which is consistent with theory and has important implications for gravity estimations of the effects of various policies \citep{Yotov2022}. In fact, following \cite{Beverellietal2024}, the use of domestic trade flows is exactly what would enable us to identify the impact of country-specific risk on international trade relative to domestic sales in our econometric model. Due to the separability of the theoretical gravity model \citep{AndersonvanWincoop2004,Costinotetal2012}, equation~(\ref{est_main}) can be estimated at any level of aggregation, e.g., product, industry, sector, or aggregate. This is important for the current purposes, as it enables us to estimate the model by pooling (not aggregating) the data for individual industries within each of the three end-use goods categories in our data.\footnote{This also implies that, in principle, we could estimate the model separately, and consistent with theory, for each of the 170 industries from the ITPD-E.} Motivated by \cite{SantosSilvaTenreyro2006}, we use the Poisson Pseudo Maximum Likelihood (PPML) as our preferred estimator. Due to its multiplicative form, the PPML estimator enables us to include and take advantage of the information contained in the zeros in our sample. In addition, and more importantly, the PPML estimator successfully handles heteroskedasticity in trade flows data, which, due to Jensen's inequality, renders the corresponding OLS estimates inconsistent. We estimate our model with the \textbf{ppmlhdfe} command of \cite{Correiaetal2020}, and, following the standard approach in the gravity literature \citep{EggerTarlea2015,Pfaffermayr2022}, we cluster the standard errors by country pair. Turning to the covariates in our model, the most important term on the right-hand side of equation (\ref{est_main}) is $RISK_{ij,t}$, which is a vector that includes our proxies for risk. As discussed earlier, we test with four alternative risk indices: (i) the country-specific economic risk measure---Economic Freedom Indexes ($EFI_{i,t}$) of the Fraser Institute;\footnote{In the main analysis, we will employ the average EFI index. However, in robustness analysis that are reported in the appendix, we also obtain results for each of the five separate EFI categories.} (ii) the country-specific political risk measure---the World Bank’s Worldwide Governance Indicators ($WGI_{i,t}$); (iii) the political and social value distance ($PSVD_{ij,t}$), which is derived from the Freedom in the World index by the Freedom House; and (iv) the bilateral geopolitical distance ($GD_{ij,t}$), which is derived from the United Nations General Assembly voting data.\footnote{The sources and construction of these and all other variables in our econometric model are described in the data section.} The vector $POLICY_{ij,t}$ in equation~(\ref{est_main}) includes the following time-varying bilateral policy variables: membership in Regional Trade Agreements ($RTA_{ij,t}$), membership in Customs Unions ($CU_{ij,t}$), membership in the European Union ($EU_{ij,t}$), membership in the World Trade Organization ($WTO_{ij,t}$), and complete trade sanctions ($SANCT_{ij,t}$). The vector $GRAV_{ij}$ includes the four proxies for time-invariant trade frictions that are used most often in the gravity literature, including bilateral distance ($DIST_{ij}$), and indicators for common official language ($LANG_{ij}$), the presence of colonial relationships ($CLNY_{ij}$), and contiguous borders ($CNTG_{ij}$). Some recent econometric gravity models do not include time-invariant gravity variables, and instead rely on country-pair fixed effects. The reason for our choice is that we want to explore the heterogeneous impact of these widely-used gravity covariates across the three end-use goods categories in our data. In addition, the use of country-pair fixed effects would absorb most of the variation in our bilateral risk proxies, which do vary over time but very slowly. Equation (\ref{est_main}) includes three sets of fixed effects. $GLOB^k_{ij,t}$ denotes a vector of time-varying border indicators, which take a value of one for international flows and are equal to zero for domestic sales for each year in our sample. The estimates on these dummy variables capture the impact of common (de-)globalization trends that have affected the international relative to the domestic trade, e.g., improvements in communication, transportation, global recessions, etc.\footnote{\cite{BergstrandLarchYotov2015} demonstrate that the estimates of the effects of RTAs are biased if such globalization dummies are not included in the gravity model.} Thus, our risk variables would capture effects that are in addition to any common (de-)globalization trends among the countries in our sample. Guided by theory, equation (\ref{est_main}) also includes source-industry-time and destination-industry-time fixed effects ($\pi^k_{i,t}$ and $\chi^k_{j,t}$, respectively) to account {for structural multilateral resistances terms \citep{AndersonvanWincoop2003} and for any country-time-specific determinants of trade on the source and the destination side.} When the model is estimated with pooled data, the corresponding fixed effects remain at the source-industry-time and the destination-industry-time dimensions, respectively. Thus, regardless of the level of aggregation, the exporter and importer fixed effects account for and absorb all country-industry-time-specific characteristics that influence bilateral trade flows. \section{Estimation Results and Analysis \label{results}} The objective of this section is twofold. First, we examine potential heterogeneity in how standard determinants of trade affect consumer, intermediate, and capital goods. Second, we assess and compare the impact of alternative geopolitical risk measures on trade for each end-use category of goods. Our main findings are presented in Table \ref{tab_efi} to \ref{tab_gd}, which report estimates by end use obtained from specification (\ref{est_main}). Table \ref{tab_efi} and \ref{tab_wgi} report estimates for country-specific geopolitical risks, and Table \ref{tab_psvd} and \ref{tab_gd} report estimates for bilateral geopolitical risks. Each panel has the same set of covariates, with the exception of geopolitical risk indicators (RISK), which vary across Table \ref{tab_efi} through \ref{tab_gd}, as detailed below. \begin{table}[h!] \begin{center} \caption{\label{tab_efi}Gravity Estimates Across Goods by End Use, EFI} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.745 & -0.885 & -0.669 \\ & (.032)**** & (.026)**** & (.032)**** \\ CNTG & 0.449 & 0.338 & 0.432 \\ & (.058)**** & (.051)**** & (.062)**** \\ CLNY & 0.047 & 0.103 & 0.008 \\ & (.081) & (.064) & (.076) \\ LANG & 0.229 & 0.224 & 0.172 \\ & (.044)**** & (.041)**** & (.048)*** \\ EU & 0.608 & 0.311 & -0.087 \\ & (.064)**** & (.072)**** & (.081) \\ CU & 0.930 & 0.496 & 0.616 \\ & (.112)**** & (.09)**** & (.137)**** \\ RTA & 0.269 & 0.241 & 0.260 \\ & (.044)**** & (.041)**** & (.045)**** \\ WTO & 0.311 & 0.491 & 1.074 \\ & (.109)** & (.079)**** & (.125)**** \\ SANCT & -1.452 & -1.957 & -2.167 \\ & (.389)*** & (.459)**** & (.358)**** \\ Risk-EFI & 1.503 & 0.847 & 1.104 \\ & (.063)**** & (.051)**** & (.09)**** \\ \hline\hline \multicolumn{4}{p{0.52\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}) for each of the three categories of goods based on end use in our sample using the country-specific Economic Freedom Index (EFI). The dependent variable is trade in levels and the estimator is PPML. All estimates are obtained with exporter-industry-time, importer-industry-time, and border-industry fixed effects, whose estimates are omitted for brevity. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} Table \ref{tab_efi} presents the estimates using Fraser's Economic Freedom Index (EFI)---serving as a proxy for a country's economic policy risk. As discussed in section \ref{data} on data, the EFI is a composite measure, evaluating five dimensions of economic conditions in a country that are influenced by related economic policies. Panel A also includes the estimated effects of the standard gravity covariates across the three end-use categories of goods in our sample. Several notable findings emerge from Table \ref{tab_efi}. First, the estimated coefficient on distance is negative and statistically significant across all three end-use categories, with a substantially stronger effect observed for intermediate goods. Because distance serves as a key proxy or determinant for bilateral trade costs, capturing frictions such as logistics expenses and transport infrastructure, this finding suggests that for a given pair of countries trade costs are notably higher for intermediate goods compared to consumer and capital goods. This may reflect their greater reliance on just-in-time delivery or more complex logistical coordination \citep{pisch2020}. The heightened sensitivity of intermediate goods trade to distance also implies that firms may favor geographically proximate suppliers and prefer regional supply chain structures to reduce disruption and costs. The results for the other geography-related variable, contiguity (CNTG)---whose positive effect is the smallest for intermediate goods---further supported this interpretation. The weaker positive effect of CNTG for intermediate goods suggests that being neighbors doesn't boost intermediate goods trade as much as it does for capital or consumer goods. This may reflect the fact that intermediate goods often move within complex supply chains, where proximity alone does not guarantee trade integration. It also suggests that trade in intermediates depends more heavily on infrastructure, logistics, and production network linkages than simple geographic closeness. Consistent with gravity estimates employing PPML \citep{SantosSilvaTenreyro2006}, we do not find statistically significant effect of colonial relationships (CLNY) on trade flows, regardless of the type of goods. In addition, the estimated impact of sharing a common official language is similar for intermediate and consumer goods, but notably smaller for capital goods. Table \ref{tab_efi} also reveals several interesting findings regarding the impact of policy variables. Most estimates associated with memberships of European Union (EU), customs unions (CU), and regional trade agreements (RTAs) are large, positive, and statistically significant. The EU effect for capital goods is the sole exception, which is small, negative, and statistically insignificant. Notably, across all three policy measures, the effects are consistently smallest for intermediate goods (excluding the EU-capital goods case). This suggests that, relative to other product categories, the standard forms of economic integration have been less effective at promoting trade in intermediate goods. That intermediate goods trade often stems from established, sticky production networks, and that supply chains may already be optimized within regions, are among the possible explanations for the relatively smaller marginal gains from joining EU, CU, or an RTA. The estimated effects of WTO membership are also positive, large, and statistically significant across all three categories of goods. However, the impact is stronger for capital goods and intermediate goods, suggesting that WTO accessions have been most beneficial for capital goods trade. One possible explanation for this result is that our sample excludes the early years of the General Agreement on Tariffs and Trade (GATT). Therefore, the identified WTO effects primarily capture the influence of accessions by more heterogeneous countries after 1986---particularly those with diverse levels of development and comparative advantage.\footnote{The General Agreement on Tariffs and Trade (GATT) was established in 1947 by 23 founding members---mostly high-income, industrialized economies with relatively similar legal systems, levels of development, and approaches to trade policy (e.g. the United States, the United Kingdom, Canada, Australia, and Western European countries). By contrast, accessions from the 1980s onward, especially under the WTO, brought in countries with much greater diversity in terms of development level, institutional quality, and trade orientations, including many emerging markets and former centrally planned economies.} These later entrants are more likely to be less developed and, as such, tend to experience larger gains from trade liberalization than high-income countries \citep{pengetal2015}. The stronger WTO effects for capital goods may thus be driven by the demand from these newer members' accelerated industrialization and integration into global production networks, rather than by changes among long-standing liberalized economies. Table \ref{tab_efi} presents the estimated effects of sanctions that reveal two clear patterns. First, the coefficients are consistently large, negative, and statistically significant across all product categories, indicating that sanctions have been broadly effective in curtailing trade flows. Second, the trade-suppressing effects of sanctions are strongest for capital goods, followed by intermediate goods, and weakest for consumer goods. This result aligns with both economic intuition and policy dynamics. Sanctions frequently target strategic sectors like aerospace, defense, telecommunications, and industrial machinery—most of which fall under capital goods. Since capital goods are often less substitutable in the short run, a sanctioned country may find it harder to replace them. Intermediate goods may offer slightly more flexibility via alternative sourcing, while consumer goods tend to be easiest to substitute or smuggle, especially through informal channels or third countries. In addition, sanctions often exclude some consumer goods for humanitarian reasons. Finally, Table \ref{tab_efi} shows that the coefficients for the Fraser's Economic Freedom Index (EFI)---our country-specific economic policy risk indicator---are sizable, positive, and statistically significant across the three product categories. The results are consistent with our expectation, given that higher EFI values reflect a more favorable economic policy environment with lower risk. Importantly, the estimated EFI effect for intermediate goods is significantly smaller than that for capital goods, and nearly half the magnitude of the effect observed for consumer goods. This comparatively lower sensitivity of intermediate goods trade to economic policy risk is consistent with existing theories and empirical evidence on global supply chain ``stickiness'' \citep{Grossmanetal2024}. In particular, it reflects the tendency of firms to maintain stable sourcing relationships for intermediate inputs, even in the face of shifting policy environments or economic uncertainty. Such inertia in production networks may arise from high switching costs, technical compatibility requirements, or long-term supplier contracts, which collectively reduce the responsiveness of intermediate goods trade to macro-level risk factors. The results are consistent when employing the individual components of the EFI, as described in the data section. The corresponding results are reported in Tables \ref{tab_efi1}, \ref{tab_efi2}, \ref{tab_efi3}, \ref{tab_efi4}, \ref{tab_efi5} of the Appendix. \begin{table}[h!] \begin{center} \caption{\label{tab_wgi}Gravity Estimates Across Goods by End Use, WGI} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.774 & -0.895 & -0.686 \\ & (0.033)**** & (0.027)**** & (0.034)**** \\ CNTG & 0.395 & 0.319 & 0.407 \\ & (0.062)**** & (0.052)**** & (0.066)**** \\ CLNY & 0.039 & 0.086 & -0.003 \\ & (0.081) & (0.066) & (0.079) \\ LANG & 0.296 & 0.257 & 0.233 \\ & (0.048)**** & (0.042)**** & (0.052)**** \\ EU & 0.410 & 0.186 & -0.227 \\ & (0.081)**** & (0.079)* & (0.096)* \\ CU & 0.905 & 0.498 & 0.590 \\ & (0.115)**** & (0.093)**** & (0.139)**** \\ RTA & 0.251 & 0.236 & 0.233 \\ & (0.047)**** & (0.041)**** & (0.047)**** \\ WTO & 0.660 & 0.618 & 1.212 \\ & (0.098)**** & (0.073)**** & (0.114)**** \\ SANCT & -1.405 & -1.839 & -2.089 \\ & (0.374)*** & (0.403)**** & (0.337)**** \\ Risk-WGI & 1.534 & 0.971 & 1.130 \\ & (0.079)**** & (0.070)**** & (0.103)**** \\ \hline\hline \multicolumn{4}{p{0.57\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}) for each of the three categories of goods based on end use in our sample using the country-specific Worldwide Governance Index (WGI). The dependent variable is trade in levels and the estimator is PPML. All estimates are obtained with exporter-industry-time, importer-industry-time, and border-industry fixed effects, whose estimates are omitted for brevity. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} Table \ref{tab_wgi} presents the estimation results using the World Bank's Worldwide Governance Index (WGI), serving as a proxy for a country's political risk. Our main findings and conclusions remain robust and consistent. The estimated effects are similarly strong and statistically significant, consistent with the findings in Table \ref{tab_efi} as well as those of \cite{Beverellietal2024}. Notably, intermediate goods trade also exhibits the lowest sensitivity to political risk, reinforcing the notion of global supply chain stickiness. \begin{table}[h!] \begin{center} \caption{\label{tab_psvd}Gravity Estimates Across Goods by End Use, PSVD} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.748 & -0.859 & -0.621 \\ & (.043)**** & (.03)**** & (.038)**** \\ CNTG & 0.408 & 0.316 & 0.430 \\ & (.07)**** & (.068)**** & (.082)**** \\ CLNY & 0.0830 & 0.090 & 0.051 \\ & (.091) & (.076) & (.09) \\ LANG & 0.295 & 0.294 & 0.269 \\ & (.053)**** & (.047)**** & (.059)**** \\ EU & 0.761 & 0.491 & 0.082 \\ & (.131)**** & (.097)**** & (.127) \\ CU & 0.684 & 0.437 & 0.545 \\ & (.131)**** & (.095)**** & (.15)*** \\ RTA & 0.180 & 0.165 & 0.197 \\ & (.059)** & (.048)*** & (.058)*** \\ WTO & 1.706 & 1.288 & 2.192 \\ & (.11)**** & (.08)**** & (.116)**** \\ SANCT & -2.216 & -2.631 & -2.822 \\ & (.325)**** & (.288)**** & (.327)**** \\ Risk-PSVD & -0.005 & -0.004 & -0.003 \\ & (.002)** & (.001)** & (.002)* \\ \hline\hline \multicolumn{4}{p{0.54\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}) for each of the three categories of goods based on end use in our sample using the bilateral Political \& Social Value Distance (PSVD) index. The dependent variable is trade in levels and the estimator is PPML. All estimates are obtained with exporter-industry-time, importer-industry-time, and border-industry fixed effects, whose estimates are omitted for brevity. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} Table \ref{tab_psvd} reports the estimation results based on the Political \& Social Value Distance (PSVD) index, one of two proxies for bilateral geopolitical risk employed in this paper. Several key findings from Panel A and B, such as the heterogeneous effects of geographic distance and contiguity, remain robust in Panel C. However, notable differences emerge, particularly in the magnitudes of effects for most policy variables across all product categories. This may reflect the PSVD's correlation with these policy variables. PSVD coefficients are consistently negative and statistically significant across all product categories, indicating a dampening effect of bilateral geopolitical risk on trade flows. However, this effect is smaller than that associated with country-specific risk measures such as the EFI and WGI, suggesting that bilateral geopolitical tensions exert a comparatively smaller influence on trade flows. Consistent with the estimates in Table \ref{tab_efi} and \ref{tab_wgi} and the global supply chain ``stickiness'' hypothesis, the impact on intermediate goods trade is less significant than on consumer goods, implying that entrenched production relationships are more resilient to geopolitical divergence between trading partners. \begin{table}[h!] \begin{center} \caption{\label{tab_gd}Gravity Estimates Across Goods by End Use, GD} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.795 & -0.893 & -0.634 \\ & (.044)**** & (.031)**** & (.037)**** \\ CNTG & 0.408 & 0.328 & 0.453 \\ & (.071)**** & (.067)**** & (.078)**** \\ CLNY & 0.274 & 0.270 & 0.230 \\ & (.087)** & (.077)*** & (.093)* \\ LANG & 0.277 & 0.291 & 0.246 \\ & (.054)**** & (.048)**** & (.057)**** \\ EU & 0.859 & 0.532 & 0.187 \\ & (.126)**** & (.095)**** & (.123) \\ CU & 0.654 & 0.403 & 0.515 \\ & (.127)**** & (.092)**** & (.149)*** \\ RTA & 0.350 & 0.265 & 0.343 \\ & (.059)**** & (.046)**** & (.063)**** \\ WTO & 1.693 & 1.289 & 2.059 \\ & (.106)**** & (.074)**** & (.105)**** \\ SANCT & -2.195 & -2.644 & -2.738 \\ & (.284)**** & (.249)**** & (.27)**** \\ RISK-GD & 0.187 & 0.096 & 0.185 \\ & (.029)**** & (.024)**** & (.031)**** \\ \hline\hline \multicolumn{4}{p{0.53\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}) for each of the three categories of goods based on end use in our sample using the bilateral Geopolitical Distance (GD) index. The dependent variable is trade in levels and the estimator is PPML. All estimates are obtained with exporter-industry-time, importer-industry-time, and border-industry fixed effects, whose estimates are omitted for brevity. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} Lastly, Table \ref{tab_gd} shows the estimation results based on the Geopolitical Distance (GD) index, a widely used bilateral geopolitical risk indicator based on UN General Assembly voting results. As in Table \ref{tab_efi}, \ref{tab_wgi}, and \ref{tab_psvd}, several key findings—such as the heterogeneous effects of distance and contiguity—remain robust and consistent. Meanwhile, the estimated magnitudes of effects for most policy variables differ markedly from those in Table \ref{tab_efi} and \ref{tab_wgi}, likely reflecting the similar correlation issue between the GD index and policy variables. However, the GD coefficients reveal a positive and statistically significant association with trade---a somewhat unexpected result that reinforce existing concerns about the validity of using the GD index, or the underlying UN voting results, as a reliable predictor of bilateral economic relations. \section{Conclusion\label{conclusion}} This paper provides a novel investigation into how standard trade determinants and geopolitical risk affect global supply chains, with a specific focus on intermediate goods trade. We construct the ITPD-U database—unprecedented in scope—covering domestic and international trade by end use for 153 industries across 265 countries over 35 years. By leveraging this new database and applying the latest gravity modeling techniques, our findings reveal that intermediate goods, accounting for about 60 percent of global merchandise trade, respond differently to several standard trade determinants as well as geopolitical risk, compared to capital and consumer goods. Distance---a key proxy or determinant for trade cost---has the largest dampening effect on intermediate goods trade, likely due to the logistical complexity and coordination demand of global supply chains. Contiguity offers a relatively small effect on intermediate goods trade, suggesting that sharing border alone offers limited additional benefit once supply chains are regionally optimized. Economic integration measures such as EU, CU, or RTA memberships most benefit consumer goods trade, while their influence on intermediate goods trade is comparatively smaller. By contrast, WTO accession and sanctions have disproportionate large effects on capital goods trade. Together, these patterns point to important asymmetries in how different trade policies shape global commerce across product categories. Geopolitical risk depresses trade across all product types, though its impact is significantly smaller for intermediate goods, underscoring the inertia as well as dependence of established sourcing relationships, which aligns with the concept of supply chain ``stickiness.'' Our findings highlight that building resilient supply chains requires a deeper understanding of how firms respond to geopolitical uncertainty---not only through diversification, nearshoring, or friendshoring, but also through institutional frameworks that reduce switching costs and through building durable trust with sourcing sources. This paper fills a longstanding gap in the gravity trade literature by revealing how standard trade determinants affect goods differently depending on their role in global production network. The ITPD-U database developed for this study enables these insights by offering granular trade flow data disaggregated by end use and incorporating domestic trade. \clearpage \phantomsection \bibliographystyle{econometrica} \bibliography{GVC_risk} \clearpage \section*{Appendix} \begin{table}[h!] \begin{center} \caption{\label{tab_efi1}Gravity Estimates Across Goods by End Use, EFI 1---Size of Government} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.750 & -0.859 & -0.619 \\ & (.041)**** & (.029)**** & (.034)**** \\ CNTG & 0.439 & 0.354 & 0.482 \\ & (.068)**** & (.065)**** & (.077)**** \\ CLNY & 0.117 & 0.127 & 0.105 \\ & (.084) & (.073) & (.079) \\ LANG & 0.272 & 0.259 & 0.166 \\ & (.052)**** & (.047)**** & (.057)** \\ EU & 0.859 & 0.605 & 0.255 \\ & (.102)**** & (.083)**** & (.097)** \\ CU & 0.655 & 0.444 & 0.536 \\ & (.132)**** & (.097)**** & (.175)** \\ RTA & 0.201 & 0.171 & 0.219 \\ & (.055)*** & (.045)*** & (.054)**** \\ WTO & 1.559 & 1.119 & 1.803 \\ & (.102)**** & (.075)**** & (.109)**** \\ SANCT & -1.541 & -1.953 & -2.134 \\ & (.388)**** & (.45)**** & (.339)**** \\ EFI\_1 & 0.175 & 0.217 & 0.471 \\ & (.075)* & (.049)**** & (.092)**** \\ \hline\hline \multicolumn{4}{p{0.5\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}). The results replicate those from Table \ref{tab_efi} but with the EFI 1---Size of Government. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} \clearpage \begin{table}[h!] \begin{center} \caption{\label{tab_efi2}Gravity Estimates Across Goods by End Use, EFI 2---Legal System and Property Rights} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.750 & -0.879 & -0.673 \\ & (.033)**** & (.026)**** & (.033)**** \\ CNTG & 0.406 & 0.325 & 0.400 \\ & (.062)**** & (.053)**** & (.065)**** \\ CLNY & 0.0290 & 0.0820 & -0.00700 \\ & (.081) & (.065) & (.078) \\ LANG & 0.301 & 0.266 & 0.237 \\ & (.047)**** & (.042)**** & (.05)**** \\ EU & 0.472 & 0.230 & -0.188 \\ & (.077)**** & (.076)** & (.09)* \\ CU & 0.910 & 0.515 & 0.618 \\ & (.114)**** & (.092)**** & (.138)**** \\ RTA & 0.283 & 0.256 & 0.261 \\ & (.047)**** & (.042)**** & (.046)**** \\ WTO & 0.637 & 0.602 & 1.198 \\ & (.098)**** & (.075)**** & (.114)**** \\ SANCT & -1.381 & -1.831 & -2.118 \\ & (.39)*** & (.416)**** & (.347)**** \\ EFI\_2 & 0.808 & 0.521 & 0.652 \\ & (.04)**** & (.036)**** & (.052)**** \\ \hline\hline \multicolumn{4}{p{0.5\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}). The results replicate those from Table \ref{tab_efi} but with the EFI 2---Legal System and Property Rights. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} \clearpage \begin{table}[h!] \begin{center} \caption{\label{tab_efi3}Gravity Estimates Across Goods by End Use, EFI 3---Sound Money} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.761 & -0.875 & -0.654 \\ & (.036)**** & (.027)**** & (.034)**** \\ CNTG & 0.448 & 0.360 & 0.460 \\ & (.065)**** & (.056)**** & (.068)**** \\ CLNY & 0.0390 & 0.0770 & 0.00900 \\ & (.086) & (.067) & (.081) \\ LANG & 0.285 & 0.274 & 0.235 \\ & (.049)**** & (.043)**** & (.053)**** \\ EU & 0.470 & 0.260 & -0.133 \\ & (.103)**** & (.084)** & (.107) \\ CU & 0.828 & 0.489 & 0.625 \\ & (.118)**** & (.091)**** & (.141)**** \\ RTA & 0.261 & 0.234 & 0.249 \\ & (.05)**** & (.042)**** & (.051)**** \\ WTO & 0.806 & 0.693 & 1.261 \\ & (.102)**** & (.072)**** & (.113)**** \\ SANCT & -1.515 & -1.885 & -2.130 \\ & (.393)*** & (.421)**** & (.35)**** \\ EFI\_3 & 1.006 & 0.627 & 0.810 \\ & (.059)**** & (.041)**** & (.072)**** \\ \hline \hline \multicolumn{4}{p{0.5\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}). The results replicate those from Table \ref{tab_efi} but with the EFI 3---Sound Money. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} \clearpage \begin{table}[h!] \begin{center} \caption{\label{tab_efi4}Gravity Estimates Across Goods by End Use, EFI 4---Freedom to Trade Internationally} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.807 & -0.897 & -0.696 \\ & (.037)**** & (.027)**** & (.035)**** \\ CNTG & 0.409 & 0.338 & 0.428 \\ & (.064)**** & (.054)**** & (.068)**** \\ CLNY & -0.0220 & 0.0410 & -0.0370 \\ & (.086) & (.067) & (.084) \\ LANG & 0.310 & 0.280 & 0.247 \\ & (.051)**** & (.044)**** & (.056)**** \\ EU & 0.178 & 0.0880 & -0.354 \\ & (.104) & (.084) & (.11)** \\ CU & 0.612 & 0.334 & 0.374 \\ & (.124)**** & (.097)*** & (.143)** \\ RTA & 0.230 & 0.217 & 0.233 \\ & (.049)**** & (.041)**** & (.048)**** \\ WTO & 0.275 & 0.378 & 0.989 \\ & (.117)* & (.085)**** & (.136)**** \\ SANCT & -1.493 & -1.913 & -2.128 \\ & (.399)*** & (.476)**** & (.34)**** \\ EFI\_4 & 1.311 & 0.763 & 0.917 \\ & (.072)**** & (.056)**** & (.086)**** \\ \hline\hline \multicolumn{4}{p{0.5\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}). The results replicate those from Table \ref{tab_efi} but with the EFI 4---Freedom to Trade Internationally. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} \clearpage \begin{table}[h!] \begin{center} \caption{\label{tab_efi5}Gravity Estimates Across Goods by End Use, EFI 5---Regulation} \begin{tabular}{lccc} \hline\hline & Cons. & Interm. & Cap. \\ \hline DIST & -0.724 & -0.871 & -0.641 \\ & (.031)**** & (.026)**** & (.03)**** \\ CNTG & 0.448 & 0.331 & 0.438 \\ & (.058)**** & (.052)**** & (.063)**** \\ CLNY & 0.0530 & 0.106 & 0.00600 \\ & (.078) & (.064) & (.075) \\ LANG & 0.212 & 0.217 & 0.164 \\ & (.045)**** & (.041)**** & (.048)*** \\ EU & 0.781 & 0.431 & 0.0470 \\ & (.064)**** & (.072)**** & (.079) \\ CU & 0.958 & 0.528 & 0.665 \\ & (.107)**** & (.088)**** & (.132)**** \\ RTA & 0.277 & 0.240 & 0.276 \\ & (.043)**** & (.041)**** & (.044)**** \\ WTO & 0.685 & 0.650 & 1.244 \\ & (.093)**** & (.074)**** & (.111)**** \\ SANCT & -1.493 & -1.896 & -2.208 \\ & (.376)**** & (.418)**** & (.352)**** \\ EFI\_5 & 1.020 & 0.592 & 0.781 \\ & (.042)**** & (.037)**** & (.059)**** \\ \hline\hline \multicolumn{4}{p{0.5\linewidth}}{\scriptsize\textbf{Notes}: This table reports gravity estimates that are obtained from equation (\ref{est_main}). The results replicate those from Table \ref{tab_efi} but with the EFI 5--Regulation. {**} \(p<0.10\), {***} \(p<.05\), {****} \(p<.01\). See text for further details.}\\ \end{tabular} \end{center} \end{table} \end{document}