Measuring Geopolitical Fragmentation: Implications for Trade, Financial Flows, and Economic Policy

Board of Governors of the Federal Reserve System International Finance Discussion Papers ISSN 1073-2500 (Print) ISSN 2767-4509 (Online) Number 1408 May 2025 Measuring Geopolitical Fragmentation: Implications for Trade, Financial Flows, and Economic Policy Florencia Airaudo, Francois De Soyres, Keith Richards, and Ana Maria Santacreu Please cite this paper as: Airaudo, Florencia, Francois De Soyres, Keith Richards, and Ana Maria Santacreu (2025). “Measuring Geopolitical Fragmentation: Implications for Trade, Financial Flows, and Economic Policy,” International Finance Discussion Papers 1408. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/IFDP.2025.1408. NOTE: International Finance Discussion Papers (IFDPs) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the International Finance Discussion Papers Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers. Recent IFDPs are available on the Web at www.federalreserve.gov/pubs/ifdp/. This paper can be downloaded without charge from the Social Science Research Network electronic library at www.ssrn.com.

Measuring Geopolitical Fragmentation: Implications for Trade, Financial Flows, and Economic Policy* Florencia S. Airaudo Franc¸ois de Soyres Keith Richards Ana Maria Santacreu April 2025 Abstract Recentgeopoliticaltensionshaverevivedinterestinunderstandingtheeconomic consequences of geopolitical fragmentation. Using bilateral trade flows, portfolio investment data, and detailed records of economic policy interventions, we revisit widely-used geopolitical distance metrics, specifically the Ideal Point Distance (IPD) derived from United Nations General Assembly voting. We document substantial variabilityinmeasuredfragmentation,drivensignificantlybymethodologicalchoices relatedtosampleperiodsandvotecategories,especiallyinthewakeofRussia’s2022 invasion of Ukraine. Our results show robust evidence of increasing fragmentation in both trade flows and economic policy interventions among geopolitically distant countrypairs,withparticularlystrongeffectsobservedinstrategicallyimportantsectors and policy motives. In contrast, financial portfolio allocations exhibit weaker, more heterogeneous, and context-sensitive responses. These findings highlight the criticalimportanceofmethodologicaltransparencyandcarefulspecificationwhenassessing geopolitical realignments and their implications for international economic relations. Keywords: Fragmentation;Geoeconomics;Trade;FinancialFlows JELClassification: F14;F36;F50;F60 *Florencia S. Airaudo (florencia.s.airaudo@frb.gov), Franc¸ois de Soyres (francois.m.desoyres@frb.gov) andKeithRichards(keith.p.richards@frb.gov)arewiththeBoardofGovernorsoftheFederalReserveSystem. AnaMariaSantacreu(ana.m.santacreu@stls.frb.org)iswiththeFederalReserveBankofSaintLouis. Theviewsexpressedinthispaperareourown,anddonotrepresenttheviewsoftheBoardofGovernors oftheFederalReserve,theFederalReserveBankofSaintLouis,noranyotherpersonassociatedwiththe FederalReserveSystem. 1

1 Introduction Over the past decade, the trajectory of global economic integration has come under intense scrutiny due to heightened geopolitical tensions, increasing emphasis on national security,andaproliferationofpoliciesexplicitlyaimedatreshapingglobalsupplychains. While traditional indicators, such as the ratio of global trade to GDP, have suggested resilience, closer scrutiny of bilateral trade and financial flows reveals emerging patterns of fragmentation aligned with geopolitical considerations (Aiyar et al. (2023a), Gopinath et al. (2025)). Rising geopolitical tensions—exemplified notably by Russia’s invasion of Ukraine in 2022, intensified trade disputes between the United States and China, and ongoing shifts towards protectionism—have triggered substantial reallocations in both trade and financial linkages. Concurrently, policymakers have increasingly utilized economic policy interventions, such as tariffs, subsidies, and export controls, to strategically reshape economic relationships, directly influencing fragmentation patterns. These developmentshaverenewedinterestinunderstandingtheprecisedynamicsofgeopolitical fragmentation and its broader economic consequences, particularly the role of deliberate economicpolicychoices. Arapidlyexpandingliteraturequantifiesgeopoliticalfragmentationbyidentifyingalignmentblocsbasedoncountries’votingbehaviorsininternationalinstitutions,particularly theUnitedNationsGeneralAssembly(UNGA).TheseminalcontributionbyBaileyetal. (2017) introduced a spatial voting model to estimate countries’ ideal points on a geopolitical spectrum, leading to the widely-adopted Ideal Point Distance (IPD). This metric hassincebeenintegraltoanalysesexploringtheeconomicimpactsofpoliticalalignment, documentingnegativeassociationsbetweengeopoliticaldistanceandcross-bordertrade, foreign direct investment (FDI), and financial asset flows (Aiyar et al. (2023a,b); Blanga- Gubbay and Rub´ınova´ (2023); Catalan et al. (2024)). Building on the Ideal Point Distance (IPD), we develop a new measure—seg—that captures each country’s relative geopolitical alignment between the United States and China. This normalized score provides a continuous, interpretable indicator of alignment and allows us to track recent shifts in global alliances. Recent studies also emphasize how escalating U.S.-China tensions and the geopolitical fallout from Russia’s invasion of Ukraine have exacerbated fragmentation trends (Jakubik and Ruta (2023); Campos et al. (2024); Qiu et al. (2023)). Yet, despite theseinsights,significantuncertaintyremainsregardinghowsensitiveconclusionsabout fragmentationaretomethodologicalchoicesinconstructingIPDmeasures. Recent contributions have proposed alternative measures of geopolitical fragmentation (e.g.,Ferna´ndez-Villaverdeetal.(2024))andquantifiedtheheterogeneouseffectsoffragmentation on trade using foreign policy alignment data (Hakobyan et al. (2024)), highlightingtheneedforsystematic,alignment-basedmeasureslikethosewedevelophere. Inthispaper,wesystematicallyaddressthisuncertaintybyexamininghowmethodological variations in IPD specifications affect the measurement and interpretation of geoeconomicfragmentation. Specifically,werevisitcriticalmethodologicalchoices,including theselectedhistoricalsampleperiodandtheinclusionofspecificvotecategories(allvotes vs. economic votes), building upon the spatial voting framework developed by Bailey 2

et al. (2017). These choices affect both the construction of ideal point distances and the resulting assignment of countries into geopolitical blocs, which in turn shape measured fragmentationoutcomes. Ouranalysisdemonstratesthatseeminglyminormethodological variations significantly influence the interpretation and magnitude of geopolitical realignments,withsubstantialimplicationsforpolicyinterpretationandacademicresearch. Employing bilateral trade data from UN Comtrade (2001-2023) and bilateral financial flows from the IMF’s Coordinated Portfolio Investment Survey (CPIS, 2015-2023), we document substantial heterogeneity in fragmentation outcomes contingent on IPD specification. We find robust evidence of increased trade fragmentation following Russia’s 2022 invasion of Ukraine, consistent with Gopinath et al. (2025). However, the estimated impactvariesnotablydependingontheIPDmeasure: IPDscapturingrecentgeopolitical shifts yield significantly higher fragmentation effects, highlighting how acute geopolitical events, like wars or major diplomatic disputes, substantially reshape trade patterns. Conversely,IPDsbasedexclusivelyoneconomicvotesproducemoremoderatefragmentation effects, suggesting that broader political tensions have stronger repercussions on tradethanpurelyeconomicdisagreements. In contrast, financial fragmentation is generally weaker and more heterogeneous across IPD specifications. This result suggests that global financial linkages exhibit greater resilience to geopolitical shocks or are mediated through third-party financial centers, reflecting the complexity and indirect nature of financial market responses to geopolitical uncertainty. These insights emphasize that while geopolitical tensions directly disrupt trade flows, financial markets may respond in subtler, more context-specific ways, reflectingunderlyingdifferencesintradeversusfinancialintegrationstructures. Using detailed policy intervention data from the Global Trade Alert (GTA), we demonstrate that economic policy interventions explicitly reflect strategic motives aligned with geopolitical fragmentation. Policies targeting sectors crucial for national security, strategicautonomy,andresilience—suchascriticalminerals,advancedtechnology,anddigital infrastructure—are particularly prevalent and strongly correlated with geopolitical distance. This strategic targeting of policy interventions substantially amplifies fragmentationtrends,asgovernmentsactivelyreshapeeconomiclinkagesaccordingtogeopolitical priorities, suggesting that economic policy plays a central role in driving observed fragmentationpatterns. We deepen our analysis by decomposing bilateral trade flows across technology classifications (high-tech, medium-tech, and low-tech goods) using product-level data from GaulierandZignago(2010). Ourfindingsindicatesubstantialfragmentationeffectsacross all technology classes, though disruptions are particularly pronounced in medium-tech andlow-techsectors. Medium-techgoods,whichincludepetroleumandindustrialproducts,facesignificantdisruptionduetogeopoliticaltensionsaffectingenergyandresource security. Low-tech goods, easier to substitute and relocate, experience fragmentation as countries seek alternative suppliers aligned with their geopolitical blocs. In contrast, high-techgoodsdisplayrelativelysmallerdisruptions,possiblyduetoconcentratedglobal productionnetworksandsignificantbarrierstorapidrestructuring. 3

Our findings carry significant implications for research methodology and policy formulation. Methodologically, our results underscore the critical importance of transparency andcarefulsensitivityanalyseswhenconstructinggeopoliticaldistancemeasures. Froma policyperspective,understandingthenuanceddimensionsofgeopoliticalfragmentation can enable governments and international institutions to craft more targeted, effective strategies to enhance economic resilience, manage risk, and achieve strategic autonomy. In particular, our proposed seg metric provides a concise and interpretable measure of each country’s relative alignment with the U.S. versus China, offering a valuable tool to monitor geopolitical realignments. Recent literature has highlighted the close correlation between economic interdependence and geopolitical alignment. Kleinman et al. (2024),forexample,findrobustempiricalevidencethatincreasedeconomictiescorrelate strongly with greater political alignment among countries. While their analysis focuses primarily on economic relationships driving geopolitical outcomes, our paper examines the reverse direction—investigating how geopolitical distances, measured through IPDs, influencepatternsoftradeandfinancialfragmentation. The remainder of the paper is structured as follows: Section 2 details the methodologies underlying the construction of ideal points and geopolitical distance measures. Section 3 presents alternative IPD specifications, normalization into U.S.–China alignment scores, and bloc classifications. Section 4 assesses fragmentation patterns in trade and financial flowsacrossdifferentblocstructuresandgeopoliticaldistancemeasures. Section5evaluates fragmentation in economic policy interventions and explores how strategic motives amplifygeopoliticalrealignments. Section6concludes. 2 Measuring Geopolitical Alignment: The UN Voting Approach We measure geopolitical alignment using the methodology developed by Bailey et al. (2017),basedonroll-callvotingdatafromtheUnitedNationsGeneralAssembly(UNGA). Thekeyideaistotranslatevotingbehaviorintonumericalindicatorsreflectingcountries’ underlying foreign policy positions, known as ideal points. Each country is assumed to occupyaspecificpositionalongasingleideologicaldimension. Votes at the UNGA are categorized into three possible outcomes for each participating country: approval (yes), opposition (no), or neutrality (abstain). The model identifies two latent thresholds or cut-points for each vote, which delineate the ranges of ideal points correspondingtothesedistinctvotingdecisions. Forinstance,countrieswithidealpoints closetothoseofWestern-alignednationsmayvotesimilarlytotheUnitedStatesonmany issues,whilecountrieswithdifferentideologicalpreferencesmightopposeorabstain. Additionally,eachresolutionischaracterizedbyadiscriminationparameter,reflectinghow effectively it separates countries along the geopolitical alignment dimension. Votes with higher discrimination parameters are particularly informative about ideological differences and thus receive greater weight in the ideal point estimation. Conversely, less informative votes, which fail to differentiate clearly between countries, exert minimal 4

influenceonthealignmentestimates. Ideal points are estimated using Bayesian Markov Chain Monte Carlo (MCMC) techniques, providing posterior distributions that capture uncertainty about each country’s alignment position. The posterior mean is employed as the definitive measure of a country’sidealpointforeachyear. Geopoliticaldistancesbetweentwocountries—IdealPoint Distances(IPDs)—arethencalculatedastheabsolutedifferencesbetweentheirrespective ideal points. For example, a substantial IPD indicates significant divergence in foreign policypreferences. OurprimarydatasourceisthecomprehensiveUNGAroll-callvotingdatasetcompiledby Voeten(2021),spanningfromthefirstsessionin1946tosession78in2023. ThisdatasetincludesallvotescastwithintheGeneralAssemblyandclassifiesresolutionsintothematic categories such as colonialism, disarmament, human rights, Middle East issues, nuclear weapons, and economic development. Each resolution is tagged with specific keywords and metadata, allowing us to isolate alignment by thematic areas. Thus, we can compute IPDs based on all available resolutions or restrict the analysis to specific thematic categories,enhancingtheflexibilityandspecificityofouralignmentmeasures. This methodology offers several advantages compared to simpler indices, such as basic agreementpercentages. First,byexplicitlyestimatingvote-specificthresholds,themodel candistinguishbetweenchangesarisingfromshiftingcountrypositionsandthoseresultingfromvariationsintheresolutionagenda. Thisallowsaclearerinterpretationofalignment shifts. Second, the inclusion of vote-level discrimination parameters ensures only significant votes substantially influence alignment measures. This prevents consensus or near-unanimous votes—which do not effectively differentiate countries—from skewing alignmentestimates. 3 Alternative Specifications and Methodology Building on the ideal point estimates described in the previous section, we develop several measures of geopolitical distance to assess trade fragmentation. Our approach proceeds in three steps. First, we estimate time-varying IPDs across countries under three different assumptions about the scope and time window of the voting data. Second, we transform these IPDs into normalized alignment scores that capture countries’ relative positioning between the United States and China. Third, we construct discrete bloc segmentationsbasedonthecross-sectionaldistributionofalignmentscoresatselectedpoints intime,obtainingfourdifferentcountryclassifications. In the first stage, we construct three alternative IPD series, each reflecting a different methodologicalchoiceregardingthescopeandtimewindowofthevotingdata: • Full Historical Sample (1946–2023, All Votes): In this specification, we maintain the original Bailey et al. (2017) framework without any modifications. This means using the complete set of UNGA votes from 1946 to 2023. No alterations are made 5

totheoriginalmethodology.1 • Economic Votes Only (1971–2023): In this alternative, the primary change to the original methodology involves restricting the input dataset exclusively to resolutions categorized under economic issues, beginning from 1971 onwards. Thus, insteadofusingthefullsetofthematiccategories,theidealpointestimatesspecifically reflectcountries’alignmentsoneconomicissuesonly.2 • Post-Cold War Period (1991–2023, All Votes): Here, the alteration from the original methodology involves changing the temporal scope. While the thematic scope (all resolution categories) remains unchanged, we limit the historical voting data to the period after the Cold War, specifically from 1990 onwards. This temporal adjustment captures alignment dynamics reflective of the post–Cold War geopolitical landscape. Figure 1 presents our three estimated IPD measures over time for a selected group of country pairs. While the relative distance between countries remains similar across all IPD measures, the choice of vote subset significantly influences the calculation of geopolitical distance. Additionally, the degree of stationarity varies across measures, affecting their sensitivity to short-term geopolitical developments, as evidenced by the economic IPDs. (a) (b) (c) Figure1: IPDmeasuresforselectedcountrypairs. Note: The figure presents our three main IPD measures over time for a selected group of countrypairs. Foreachyear,bilateraldistance(IPD)iscomputedastheabsolutevalueofthe difference of ideal points between the two countries. (a) presents the IPDs estimated using UNGAvotingdatafrom1946-2023acrossallvotecategories. (b)presentstheEconomicsIPD, whichnarrowsthefocustoeconomicvoteswhilemaintaininghistoricalcoveragefrom1971 onwards.(c)shortensthehistoricalwindowusedintheIPDestimationtobeginaftertheCold War(1990–2023)whilemaintainingallvotecategories. After estimating IPDs using the strategies described above, we refine these measures in 1ThefullhistoricalsampleIPDdataareavailableat: https://dataverse.harvard.edu/dataverse/Voeten. 2Votesinvolvingembargoesareincludedundertheeconomicvotescategory,withtwocaveats:(i)embargorelated votes before 1971 are excluded due to the sample restriction, and (ii) unanimous embargo votes, which provide no information on ideological differences, are effectively excluded because they do not contributetotheestimationofidealpoints. 6

a second stage by transforming them into normalized alignment scores. These scores reflect each country’s relative positioning between the United States and China, providing a more intuitive view of global geopolitical dynamics. By normalizing the bilateral IPDs, we can assess how closely countries align with either the U.S. or China over time, facilitatingtheanalysisofshiftsinstrategicorientation. We transform each bilateral IPD series into a normalized alignment index, denoted as seg(s),definedas: IPD(s,China)−IPD(s,U.S.) seg(s) = . IPD(s,U.S.)+IPD(s,China) The seg index ranges from −1 to +1, with values near −1 indicating stronger alignment with China, values near +1 indicating stronger alignment with the United States, and values close to zero indicating relative neutrality. This transformation produces three correspondingtimeseriesofsegmeasures,eachalignedwithadifferentIPDestimation. Figure 2 illustrates how countries’ geopolitical alignments differ across these measures. In each panel, the vertical axis shows the baseline seg based on full-vote IPDs estimated through 2021, while the horizontal axis shows seg based on one of the alternative IPD definitions: in (a), geopolitical alignment is obtained using the first alternative, the 2023 IPD values; in (b), geopolitical alignment is obtained using economic votes only; and in (c), geopolitical alignment using only post-1990 votes. Each point represents a country. Points above the 45-degree line indicate countries that are relatively more China-aligned inthealternativemeasure;pointsbelowthelinearerelativelymoreU.S.-aligned. (a) (b) (c) Figure2: ComparisonofIPDmeasures. Subfigure (a) shows that from 2021 to 2023, countries shifted closer to the U.S., albeit from a strong-China starting point. This trend is especially pronounced for Advanced Foreign Economies (AFEs) (light blue dots).3 This pattern suggests that recent geopolitical developments—such as Russia’s invasion of Ukraine and escalating U.S.—China 3AFE countries are Canada, Japan, U.K., U.S., France, Germany, Italy, Spain, Switzerland, Australia, and 7

tensions—may have played a key role in reshaping alliances. The fact that the shift is morepronouncedamongAFEs,whicharehistoricallyalignedwiththeU.S.,mayindicate atighteningofallianceswithintheWesternbloc,particularlyinresponsetoeconomicand securityconcerns. In subfigure (b) we focus on economic votes. In this case, if a country falls below the 45-degree line, it is more aligned with the U.S. than when using the full vote sample. The clustering along the 45-degree line suggests that geopolitical alignment is relatively stable across voting subsets. However, we observe a concentration around -0.5 on the x-axis, indicating that some countries appear more neutral when votes on human rights, colonialism,amongothers,areexcluded. Subfigure (c) shows that the choice of the estimation sample period (i.e., starting in 1946 or 1990) does not introduce substantial differences in geopolitical alignment, suggesting thatIPDestimatesarerelativelyrobusttotheestimationsamplewindow. Figure 3 translates the IPD shifts from Figure 2(a) into a geographic visualization, showinghowcountries’alignmentsshiftedbetween2021and2023. Themaprevealsanotable pattern of nations moving closer to the U.S. despite initially being more China-aligned. Thisgeographicperspectivehighlightsregionaltrendsnotapparentinscatterplots,such as the coherent shift among European and Oceanic countries toward the U.S., while responses vary across Africa, South America, and Southeast Asia. Though the map does not capture the full magnitude of these shifts, it effectively illustrates how recent geopoliticalevents—Russia’sinvasionofUkraine,U.S.-Chinatensions,andchangingeconomic relationships—have influenced international alignments. This visualization provides geographicalcontexttothenumericaldata,showingthespatialdistributionofrealignment patternsinresponsetoevolvinggreatpowerdynamics. Sweden. EMEcountriesareChina,India,Singapore,SouthKorea,Malaysia,Indonesia,Philippines,Thailand,Mexico,Vietam,Argentina,Brazil,Chile,Colombia,Israel,Russia,andSaudiArabia. 8

Figure3: EvolutionofsegmentedIPDsfrom2021to2023 Note: Themapillustrateschangesingeopoliticalalignmentsbetween2021and2023basedon segmentedIdealPointDistances(seg). Countriesshiftingtowardstrongeralignmentwiththe United States are marked accordingly, highlighting notable realignments, especially evident amongEuropeanandOceaniccountries. In the third stage, we use the segmented alignment scores to construct discrete country blocs. Our Baseline bloc classification is based on the 2021 distribution of seg derived from the 1946-2923, all votes IPDs. We construct three alternative categorizations: (i) based on the 2023 distribution from the same IPD estimation, (ii) based on the 2021 distribution of the economic-votes IPDs, and (iii) based on the 2021 distribution of the post-1990 IPDs. Thus, while we estimate three underlying IPD (and seg) time series, we generatefouralternativeblocclassificationsdependingontheyearandvotesubsetused. The formal definition of bloc membership—classifying countries as U.S.-aligned, Chinaaligned,ornonaligned—isdetailedinthenextsection. 4 Fragmentation In this section, we follow the methodology outlined by Gopinath et al. (2025) to examine whether trade and financial flows are fragmenting along geopolitical lines and whether these findings are sensitive to the specific IPD measure used. By analyzing variations in the Ideal Point Distance (IPD) specifications, we assess the robustness of observed fragmentationpatterns,particularlyconsideringrecentgeopoliticalshifts. First, we construct geopolitical alignment blocs based on countries’ seg(s) scores derived from the estimated IPDs. Countries are classified into three groups: a U.S.-aligned bloc, comprisingthoseinthetopquartileofalignmentwiththeUnitedStates;aChina-aligned bloc,comprisingthoseinthetopquartileofalignmentwithChina;andanonalignedbloc, 9

comprising all remaining countries. We apply this classification separately under each of the four IPD specifications introduced in the previous section: the baseline measure based on 2021 IPDs, the updated measure based on 2023 IPDs, the economic vote IPDs from2021,andthepost-1990IPDsfrom2021. ComparingblocassignmentsundereachalternativeIPDspecificationtothebaselineclassification, we find that 48% of countries change blocs when using the 2023 IPD, 31% change blocs with the economic vote IPD, and only 2% change blocs when using post- 1990 IPDs. The higher reclassification rate under the 2023 IPD suggests that countries arereactingtoimmediatepoliticalandeconomicpressuresratherthanmaintaininglongstandingalignments. Incontrast,therelativestabilityofblocclassificationsundertheeconomic vote and post-1990 measures indicates that economic relationships and post–Cold Waralignmentsmaybemoreresistanttoshort-termgeopoliticaldisruptions. Second,wedefinethreedummyvariablesbasedongeopoliticalblocmembership. Specifically,BetweenBloc equals1ifcountriessanddbelongtodifferentblocs,whereasWithin sd Bloc equals1ifbothcountriesbelongtothesamebloc. Lastly,Nonaligned equals1ifat sd sd leastonecountryinthepairbelongstothenonalignedbloc. Thesedummyvariablesmay vary for the same country pair sd, depending on the IPD specification and the resulting classificationintoblocs. Weestimatethefollowinggravityequation. Y = β BetweenBloc ×Post +β Nonaligned ×Post +δ +τ +ϕ +ϵ , (1) sdt 1 sd t 2 sd t sd st dt sdt where Y is the value of total trade of goods between the country s and the country sdt d or the change in the share of portfolio assets held by the reporting country s in the counterpart country d between year t and t − 1. Post is an indicator equal to 1 after Russia’s invasion of Ukraine (years 2022-2023). δ , τ , and ϕ are country-pair, source × sd st dt timeanddestination×timefixedeffects,includedinallspecifications. For trade, we estimate the gravity model using Poisson pseudo-maximum likelihood (PPML), using annual data for the period 2001-2023, from UN Comtrade. For portfolio holdings, we estimate the gravity model with OLS using semi-annual data for the period 2015s1-2023s2. Financial data contains bilateral data on countries’ holdings of cross-borderportfolioinvestment(equityordebt)securities,excludingForeignDirectInvestment(FDI),fromtheIMF’sCoordinatedPortfolioInvestmentSurvey(CPIS). Table 1 panel (I) shows the estimation results for trade under alternative IPD specifications.4 The first column presents the results using the baseline IPD measure to construct the blocs, showing evidence of geopolitical fragmentation in trade flows. The estimated coefficientindicatesthatinthepost-invasionperiod,tradeflowsbetweencountriesindifferent geopolitical blocs are 11.8% lower, on average, compared to trade flows between countries within the same bloc.5 This result is statistically significant at the 1% level. In 4TradeistheCIFvalueoftotalgoodstradedbetweenCountryAandBinmillionsofUSD.Estimationfor portfolioholdingsequationusethelagofthecountry-pairportfolioshareasanadditionalregressor. 5CoefficientinterpretationfromPoissonregressionise−0.125−1≈−11.8%. 10

contrast, trade flows between country pairs where at least one country is nonaligned are not significantly different from trade flows within the same bloc. These findings align withthoseofGopinathetal.(2025),despitedifferencesintheunderlyingtradedata.6 Table1: RegressionResults Description BaselineIPD IPDall IPDeconomic IPDall (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (1) (2) (3) (4) (I)Trade BetweenBloc×Post -0.125*** -0.276*** -0.094** -0.128*** Std. Error (0.040) (0.058) (0.038) (0.041) Nonaligned×Post -0.044 -0.066 -0.017 -0.041 Std. Error (0.070) (0.057) (0.053) (0.068) Observations 389,747 389,761 387,589 389,747 (II)PortfolioHoldings BetweenBloc×Post -0.026* -0.014 -0.000 -0.026* Std. Error (0.016) (0.016) (0.012) (0.016) Nonaligned×Post -0.021 -0.016 -0.023 -0.021 Std. Error (0.022) (0.020) (0.022) (0.021) Observations 231,450 231,971 231,578 231,450 Note: Significance thresholds: *** p < 0.01, ** p < 0.05, * p < 0.1. For trade, we estimate thegravitymodelusingPoissonpseudo-maximumlikelihood(PPML),usingannualdatafor the period 2001-2023, from UN Comtrade. Standard errors are clustered at the country-pair level.Weincludecountry-pair,source×time,anddestination×timefixedeffects.Coefficient (cid:0) (cid:1) interpretationisthefollowing: ecoefficient−1 ×100. Forportfolioholdings,weestimatethe gravitymodelwithOLSusingsemi-annualdatafortheperiod2015s1-2023s2. Financialdata containsbilateraldataoncountries’holdingsofcross-borderportfolioinvestment(equityor debt)securities,excludingForeignDirectInvestment(FDI),fromtheIMF’sCoordinatedPortfolioInvestmentSurvey(CPIS).PostisadummythatcapturesthepostinvasionofUkraine periodandtakesthevalue1fortheyears2022and2023.Eachcolumnshowstheresultsusing a different IPD measure to construct the country blocs. (1) uses the Baseline IPD measure, which is the 2021 values of IPDs estimated using UNGA voting data from 1946-2023 across allvotecategories;(2)usesthe2023valuesofIPDsestimatedusingUNGAvotingdatafrom 1946-2023acrossallvotecategories;(3)usesthe2021valuesofEconomicsIPD,whichnarrows thefocustoeconomicvoteswhilemaintaininghistoricalcoveragefrom1971onwards;(4)uses 2021valuesofIPDestimatedusingasubsampleofvotesaftertheColdWar(1990–2023)while maintainingallvotecategories. This effect strengthens with the 2023 IPD (-24.1%), reflecting recent geopolitical shifts, butweakenswitheconomicvotes(-9%),suggestingthattradealignmentfollowsbroader political ties more than economic cooperation. The post-1990 IPD yields similar results to the baseline (-12%), indicating that Cold War-era voting does not significantly affect 6In this note, we use annual goods trade data from UN Comtrade for the period 2001–2023. In contrast, Gopinathetal.(2025)usequarterlytotalbilateraltradedatafromTradeDataMonitor,aprivateprovider, for2017:Q1–2024:Q1. 11

fragmentation estimates. In all specifications, the coefficient for the interaction term Nonaligned×Post remains small and statistically insignificant, reinforcing the idea that tradeamongnonalignedcountriesislessaffectedbygeopoliticaltensions. Overall, we find that the specification of IPDs plays a crucial role in shaping conclusions about the degree of trade fragmentation, as the magnitude of the observed effects varies depending on the choice of IPD measure. The 2023 IPD best captures recent geopolitical disruptions, making it ideal for analyzing short-term realignments. The economic votes IPDprovidesaclearerpictureoftraderelationsdrivenbyeconomicdependenciesrather than broad political alliances. The post-1990 IPD offers a historically consistent view of geopolitical fragmentation, minimizing distortions from Cold War-era dynamics. The findings confirm that trade fragmentation is increasing, but its intensity depends on the chosenIPDmeasure. Panel (II) in Table 1 provides some evidence of financial fragmentation, with the share of portfolio holdings between countries in opposing blocs declining by 0.3 percentage points post-invasion in baseline and post-1990 IPDs. However, results are weaker and less robust across specifications, particularly in 2023 and economic votes IPDs. Meanwhile, the coefficients for nonaligned countries remain small and insignificant across all specifications, reinforcing the idea that financial flows among nonaligned countries were lessaffectedbygeopoliticaltensions. Overall,whilesomespecificationssuggestportfolio fragmentation along geopolitical lines, these effects are highly sensitive to the choice of IPDmeasure,contrastingwiththemorerobustandconsistentfragmentationobservedin tradeflows. The presence of offshore financial centers poses challenges for accurately identifying underlying geopolitical exposures, as such centers may obscure true investor-recipient relationships. To address this issue, in the Appendix we re-estimate our results excluding prominent financial hubs, following the methodology proposed by Coppola et al. (2021), whospecificallyexaminetheroleofthesecentersinglobalportfolioallocations.7 4.1 Alternative specification of gravity equations In this section, we explore two alternative specifications of gravity equations to evaluate whethertradeflowsfragmentalonggeopoliticallines. While our previous analysis used normalized alignment measures (seg(s)) to classify countries into distinct blocs, here we employ bilateral IPD measures directly to assess fragmentationwithoutimposingexplicitblocboundaries. WeuselaggedIPDtomitigate simultaneity concerns, assuming geopolitical alignment evolves gradually and is predetermined relative to trade and financial flows. The interaction term with Post, captures whether trade flows are affected by changes in geopolitical distance in the post-invasion period. Fixedeffectsarethesameasin1. 7Inourportfolioholdingsanalysis,wedonotincludewell-knownfinancialcenters,suchasBermuda,the BritishVirginIslands,theCaymanIslands,andHongKongSAR,assourcecountries.IntheAppendix,we presentarobustnesscheck,wherewefurtherexcludeothercountriesfrequentlyidentifiedasinternational financialcenters,includingIreland,Luxembourg,theNetherlands,andSingapore. 12

Y = β IPD ×Post +δ +τ +ϕ +ϵ , (2) sdt sdt−1 t sd st dt sdt Table 2 presents the estimation results of equation (2), using the IPDs estimated using all votes from 1946 to 2023, and those restricted to economic votes. We omit the results for IPD estimated vith all votes since the 1990s, as they are nearly identical to the baseline specification. Columns 1 and 2 report the coefficients for the interaction term IPD × Post,whilecolumns3and4includethedirecteffectofgeopoliticaldistanceIPDwithout interactionwiththepost-invasionperiod.8 Panel (I) of Table 2 presents the estimation results for total bilateral trade. In columns 1 and2,wefindstrongevidenceoftradefragmentationalonggeopoliticallinesinthepostinvasionperiod. Aone-unitincreaseingeopoliticaldistanceisassociatedwithanapproximately 8% decline in trade flows, on average. These results are statistically significant at the1%levelandremainconsistentacrossthebaselineandeconomicIPDspecifications. Table2: RegressionResultswithIPD Description IPD all votes Economic IPD IPD all votes Economic IPD IPD × Post IPD (1) (2) (3) (4) (I) Trade β Coefficient -0.087*** -0.080*** -0.068** 0.021* Std. Error (0.023) (0.024) (0.027) (0.012) Observations 374,365 371,609 374,365 371,609 (II) Portfolio holdings β Coefficient -0.075** -0.058* 0.051 0.068 Std. Error (0.038) (0.037) (0.076) (0.062) Observations 115,142 114,129 115,142 114,129 Note: Significancethresholds: ***p < 0.01, **p < 0.05, *p < 0.1. Fortrade, weestimatethe gravitymodelusingPoissonpseudo-maximumlikelihood(PPML),usingannualdataforthe period2001-2023,fromUNComtrade. Weincludecountry-pair,source×time,anddestination × time fixed effects. Standard errors are clustered at the country-pair level. Coefficient (cid:0) (cid:1) interpretationisthefollowing: ecoefficient−1 ×100. Forportfolioholdings,weestimatethe gravitymodelwithOLSusingsemi-annualdatafortheperiod2015s1-2023s2. Financialdata containsbilateraldataoncountries’holdingsofcross-borderportfolioinvestment(equityor debt)securities,excludingForeignDirectInvestment(FDI),fromtheIMF’sCoordinatedPortfolioInvestmentSurvey(CPIS).Columns(1)and(3)usetheIPDestimatedwithallvotesfrom 1946 to 2023 as explanatory variable. Columns (2) and (4) use the Economic IPD, estimated usingonlyeconomicvotesfrom1971to2023. 8In this note, we present the results using the IPD measures lagged one period to mitigate potential endogeneity concerns, as in Catalan et al. (2024). Lagging IPDs helps reduce potential reverse causality, as geopoliticalalignmentmaybothinfluenceandbeinfluencedbytradeandfinancialflows. 13

Columns 3 and 4 assess the direct effect of geopolitical distance on trade flows, independent of the post-invasion period. The estimated coefficient remains statistically significant under the baseline IPD measure, with a 7% decline in trade flows. However, the economic IPD (column 4) yields a smaller but positive coefficient (2%), suggesting that trade relationships based on economic alignment may be more resilient to geopolitical fragmentation,particularlyinperiodspriorto2022. Thecontrastingcoefficientsbetween IPD × Post (column 2) and direct IPD effects (column 4) reveal an important temporal pattern in how economic voting alignment relates to trade flows. While countries with different economic voting patterns show reduced trade after 2022 (negative IPD ×Post coefficient), they actually maintained stronger trade relationships in the pre-invasion period (positive IPD coefficient). This suggests that economic voting differences did not disrupt trade until recent geopolitical tensions transformed how such alignment matters foreconomicrelationships. Part (II) of Table 2 provides evidence of financial fragmentation, though the effects vary across specifications. In columns 1 and 2, the interaction term IPD × Post is negative and statistically significant (-0.075 and -0.058, respectively), indicating that in the postinvasion period, greater geopolitical distance is associated with a reduction in portfolio holdings between country pairs. However, in columns 3 and 4, where we estimate the direct effect of IPD independently of the post-invasion period, the coefficients are statistically insignificant. This suggests that while geopolitical distance has played a greater roleinshapingportfolioallocationsinrecentyears,itsinfluencewasmorelimitedbefore 2022. Unlike trade, where fragmentation effects appear more persistent, portfolio holdings seem more reactive to recent geopolitical shocks rather than long-standing geopoliticalalignments. 4.1.1 Distant,alignedandnonalignedcountries In this section, we examine a new definition of country blocs based on the distribution of the geopolitical distance of the country pair IPD(s,d), instead of working on the segment space. Rather than assigning a central role to the U.S. and China in defining global geopolitical alignments, this approach classifies country pairs solely based on their relative geopolitical proximity. We define a bloc of aligned country pairs, which includes countrypairsinthelowerquartileoftheIPDdistributioninagivenyear,ablocofdistant country pairs, which includes country pairs in the top quartile of the IPD distribution in a given year, and a set of nonaligned country pairs, comprising the remaining economies. When comparing bloc classifications under different IPD specifications to the baseline IPD,wefindthat25%ofcountry-pairschangeblocswhenusingthe2023IPD,30%change blocswiththeeconomicvotesIPD,andonly1.6%changeblocswhenusingdatasincethe 1990s. Thisalternativeblocdefinitionprovidesamoregeneralperspectiveongeopoliticalalignment,removingtheemphasisonspecificanchorcountriesliketheU.S.andChina. Using theseblocs,weestimatethefollowinggravityequationfordifferentIPDspecifications: 14

Y = β Distant ×Post +β Nonaligned ×Post +δ +τ +ϕ +ϵ , (3) sdt 1 sd t 2 sd t sd st dt sdt Table3: RegressionResults Description Baseline IPD IPD all IPD economic (complete, 2021) (complete, 2023) (complete, 2021) (1) (2) (3) (I) Trade Distant x Post -0.048 -0.090** -0.113** Std. Error (0.037) (0.036) (0.055) Nonaligned x Post -0.010 0.037 0.009 Std. Error (0.032) (0.032) (0.025) Observations 387,698 387,559 383,468 (II) Portfolio holdings Distant x Post -0.010 -0.010 -0.019 Std. Error (0.014) (0.014) (0.014) Nonaligned x Post -0.010 -0.001 -0.023** Std. Error (0.010) (0.010) (0.010) Observations 230,675 227,154 230,675 Note: Significancethresholds: ***p < 0.01, **p < 0.05, *p < 0.1. Fortrade, weestimatethe gravitymodelusingPoissonpseudo-maximumlikelihood(PPML),usingannualdataforthe period2001-2023,fromUNComtrade. Weincludecountry-pair,source×time,anddestination × time fixed effects. Standard errors are clustered at the country-pair level. Coefficient (cid:0) (cid:1) interpretationisthefollowing: ecoefficient−1 ×100. Forportfolioholdings,weestimatethe gravitymodelwithOLSusingsemi-annualdatafortheperiod2015s1-2023s2. Financialdata containsbilateraldataoncountries’holdingsofcross-borderportfolioinvestment(equityor debt)securities,excludingForeignDirectInvestment(FDI),fromtheIMF’sCoordinatedPortfolioInvestmentSurvey(CPIS).PostisadummythatcapturesthepostinvasionofUkraine periodandtakesthevalue1fortheyears2022and2023.Eachcolumnshowstheresultsusing a different IPD measure to construct the country blocs. (1) uses the Baseline IPD measure, which is the 2021 values of IPDs estimated using UNGA voting data from 1946-2023 across allvotecategories;(2)usesthe2023valuesofIPDsestimatedusingUNGAvotingdatafrom 1946-2023acrossallvotecategories;(3)usesthe2021valuesofEconomicsIPD,whichnarrows thefocustoeconomicvoteswhilemaintaininghistoricalcoveragefrom1971onwards. where Distant isadummyvariablethattakesthevalue1ifthecountrypairsdbelongs sd to the bloc distant and Nonaligned is a dummy variable that takes the value 1 if the sd country pair sd belongs to the bloc nonaligned. Table 3 shows the estimation results of equation(3)fortradeandchangesintheportfolioholding’sshare. 15

Thefirstpanelshowstheresultsfortrade. Thecoefficientfortheinteractionterm Distant sd ×Post is negative and statistically significant in columns 2 and 3, suggesting that trade t flowsbetweendistantcountrypairsdeclinesignificantlyinthepost-invasionperiodwhen constructing the blocs with the latest data or when restricting to economic votes. Using the economic IPD measure (column 3), trade flows between distant pairs are estimated tobearound11%lower,onaverage,comparedtoalignedpairs,underscoringtheheightened impact of geopolitical distance on trade fragmentation. In contrast, the baseline specification (column 1) shows a smaller and statistically insignificant coefficient, highlightingthesensitivityofresultstothechoiceofIPDmeasureandblocsdefinition. These findings reinforce the idea that geopolitical tensions disproportionately disrupt trade relationshipsbetweendistantcountries. Thecoefficientsfortheinteractionterm Nonaligned ×Post aresmallandstatisticallyinsd t significant across all specifications, indicating that trade flows between nonaligned pairs remain stable in the post-invasion period. This result aligns with earlier findings using U.S.- and China-centric bloc definitions, where trade relationships involving nonaligned countrieswerelessaffectedbygeopoliticaltensions. Panel (II) of Table 3 shows the results for the change in the share of portfolio holdings. Unlike trade, the effects of geopolitical fragmentation on financial flows appear more muted. The coefficients for the interaction term Distant × Post are small and statistisd t cally insignificant across all specifications, suggesting that geopolitical distance has not led to a significant reallocation of portfolio holdings in the post-invasion period when weconsidergeopoliticaldistanceinaglobalperspectiveinsteadoffocusingonUS-China blocs. Similarly, the coefficients for the interaction term Nonalignedsd × Post are also t smallandmostlyinsignificant,withtheexceptionoftheeconomicIPDspecification(column 3), where the coefficient is negative and indicates that, when considering economic alignment, the share of portfolio holdings between nonaligned country pairs decline by approximately 2.3% in the post-invasion period. Taken together, these results indicate that financial flows, while responsive to geopolitical tensions, remain substantially less sensitive compared to trade flows, reflecting the broader economic and financial considerationsthatdriveportfoliodecisions. However, the weaker evidence for financial fragmentation may also partially reflect the unique and dominant role of the United States in global financial markets. As illustrated in Figure 2, the IPD measures consistently place the U.S. in the far-right tail of the distribution, signaling its great geopolitical distance from many international counterparts. WhenweexcludetheU.S.fromtheanalysisinequation(3),theeffectsofgeopoliticaldistance on portfolio holdings become clearer and more significant. Specifically, under the economic IPD specification, portfolio investments between distant country pairs decline significantly by about 2.5% following recent geopolitical shocks.9 This finding suggests that the United States’ central role in global finance may mask underlying fragmentation trends, offsetting or moderating the impact of geopolitical tensions on international portfolioallocations. 9SeeAppendixfordetailedresults. 16

4.2 Trade Fragmentation By Technology Class Thus far, our analysis of trade has focused solely on aggregate bilateral trade flows. While the findings above suggest increasing fragmentation in international trade, recent geopoliticalevents—particularlyRussia’sinvasionofUkraineandheightenedU.S.-China tensions—highlightthatthereallocationoftradeflowsmayvaryconsiderablyacrosssectors. Increasingly, governments are adopting strategic trade policies aimed explicitly at reducing dependencies, enhancing supply chain resilience, and reinforcing alliances through “decoupling,” “de-risking,” and “friendshoring.” To investigate how these contemporarygeopoliticaldynamicsaffectdifferentsegmentsofglobaltrade,wedecompose bilateral trade flows into high-tech, medium-tech, and low-tech manufacturing classes, examiningeachsector’sevolvingpatternsandresponsestogeopoliticalpressures. For this analysis, we utilize BACI, a rich dataset from the Centre d’Etudes Prospectives et d’Informations Internationales (CEPII) that reconciles discrepancies in national trade statistics to construct bilateral trade flows at the country pair-year-product level. BACI spans from 1996-2023 and has coverage of more than 200 countries and 5,000 product codes. To ensure symmetry with our previous analysis using UN Comtrade data, we restrict the data to the period 2001-2023. To construct technology classes, we first reclassifytheproductsinBACIfromtheHS-6leveltoISICRevision3. Fromhere,wemake judgment-based designations of goods as high-tech, medium-tech, low-tech, or other. OurcategorizationsforeachtechnologyclasscanbefoundintheAppendix,section7.2. We begin by re-estimating equation (1) using the segmented IPDs interacted with our Post indicator. We run separate gravity equations for each technology class. Table 4 presents the estimation results of geopolitical distance on tech-based trade flows using oursegmenteddistribution. Asinourpreviousestimationsfortrade,wererunourgravity equations with several different IPD measures. Table 4 shows the estimation results foreachtechnologyclass. UsingthebaselineIPDmeasuretoconstructtheblocs,weseea negative coefficient for each tech-class, suggesting that in the post-invasion period, trade flows between countries of different geopolitical blocs was lower on average compared to trade flows between countries within the same bloc for each respective tech-class. The effects vary in magnitude, with the decrease in trade flows being the largest for low-tech goods (15.2%) and the smallest for high-tech (9.3%). Our results are significant at the 1% level for high and low-tech goods and the the 5% level for medium-tech goods. Across nearly all specifications and technology classes, the coefficients for Nonaligned × Post sd t are statistically insignificant with the exception of specification (2) for low-tech goods. A possible explanation for this, is that nonaligned countries may have experienced shifts in trade patterns for low-tech goods due to increased uncertainty, disruption of supply chains involving sanctioned or geopolitically distant economies, or opportunistic redirection of exports toward politically neutral markets. As such, we would expect these shifts in trade flows to be more pronounced for low-tech goods as they typically have fewer barriers to shifting suppliers or markets compared to medium- or high-tech goods evenamongnonalignedcountries. Moreover,low-techgoodstrademaybemoreaffected by geopolitical distance than high-tech trade primarily due to their higher elasticity of substitution. Unlike high-tech products, which typically require specialized inputs, ad- 17

vanced infrastructure, and stable long-term supplier relationships, low-tech goods are relatively standardized and production processes less capital-intensive. Consequently, when geopolitical tensions rise, it is substantially easier—and less costly—for countries to rapidly shift sourcing or redirect exports of low-tech goods away from geopolitically distantmarkets,amplifyingtheirsensitivitycomparedtohigh-techgoods. Consistent with our results from Table 1, the effect strengthens with the 2023 IPDs for all techclasses,particularlyformedium-techgoods. Thereisanotableincreaseinthemagnitude for trade of medium-tech goods when we consider 2023 IPDs. The particularly pronounced fragmentation observed in medium-tech goods when employing the updated 2023 IPD measure may reflect significant recent disruptions to trade in petroleum products and related commodities due to sanctions on Russia. Additionally, weaker and less significant fragmentation effects identified with economic-vote-based IPDs reinforce the interpretation that geopolitical rather than purely economic alignments primarily drive theseobservedtradedisruptions. 18

Table4: RegressionResults Description BaselineIPD IPDall IPDeconomic IPDall (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (1) (2) (3) (4) (I)High-techTrade BetweenBloc×Post -0.098*** -0.209*** -0.075** -0.098*** Std. Error (0.037) (0.056) (0.035) (0.038) Nonaligned×Post -0.133* -0.076* 0.041 -0.141 Std. Error (0.078) (0.045) (0.058) (0.074) Observations 291,816 291,702 290,284 291,831 (II)Medium-techTrade BetweenBloc×Post -0.135** -0.330*** -0.074 -0.134** Std. Error (0.055) (0.069) (0.052) (0.055) Nonaligned×Post -0.060 -0.116 -0.026 -0.045 Std. Error (0.100) (0.101) (0.085) (0.098) Observations 255,976 255,815 254,711 255,995 (III)Low-techTrade BetweenBloc×Post -0.165*** -0.213*** -0.151*** -0.176*** Std. Error (0.033) (0.039) (0.0289) (0.034) Nonaligned×Post -0.076 -0.151*** -0.066 -0.072 Std. Error (0.051) (0.041) (0.054) (0.050) Observations 288,982 288,913 287,500 289,006 Note: Significance thresholds: *** p < 0.01, ** p < 0.05, * p < 0.1. For each tech-class, we estimatethegravitymodelusingPoissonpseudo-maximumlikelihood(PPML),usingannual datafortheperiod2001-2023,fromCEPII.Weincludecountry-pair,source×time,anddestination×timefixedeffects. Standarderrorsareclusteredatthecountry-pairlevel. Coefficient (cid:0) (cid:1) interpretationisthefollowing: ecoefficient−1 ×100. Postisadummythatcapturesthepost invasion of Ukraine period and takes the value 1 for the years 2022 and 2023. Each column shows the results using a different IPD measure to construct the country blocs. (1) uses the Baseline IPD measure, which is the 2021 values of IPDs estimated using UNGA voting data from 1946-2023 across all vote categories; (2) uses the 2023 values of IPDs estimated using UNGAvotingdatafrom1946-2023acrossallvotecategories;(3)usesthe2021valuesofEconomicsIPD,whichnarrowsthefocustoeconomicvoteswhilemaintaininghistoricalcoverage from1971onwards;(4)uses2021valuesofIPDestimatedusingasubsampleofvotesafterthe ColdWar(1990–2023)whilemaintainingallvotecategories. IntheAppendix,were-estimateequations(2)and(3)separatelyforeachtechnologyclass, and find results consistent with those presented in Tables 2 and 3. These additional estimations further highlight the heterogeneity across technology classes. Although trade flows decrease among geopolitically distant country pairs for all technology categories, the magnitude of this decline differs significantly. This variation aligns intuitively with the distribution and substitutability of trade across sectors. For instance, high-tech trade, dominated by complex products such as computers, electrical equipment, machinery, and chemicals, is highly concentrated within a few key countries like the United States, 19

China, Taiwan, and major Euro-area economies.10 Consequently, rebalancing trade flows for these goods in response to geopolitical shocks, such as Russia’s invasion of Ukraine, maybeeconomicallychallengingorpracticallyinfeasible. Conversely,productionoflowtech goods, like textiles, can be reshored or decoupled more easily, leading to more pronouncedtradereallocationinthiscategory. 5 Economic policy fragmentation Giventheevidencefromprevioussectionsthattradeandfinancialflowshavefragmented along geopolitical lines, it is important to examine whether deliberate economic policies mayfurtherincentivizeorexacerbatethisfragmentation. Indeed,geopoliticalfragmentationincreasinglymanifestsnotonlyasshiftsineconomicflowsbutalsothroughtargeted policy interventions explicitly aimed at reshaping cross-border economic interactions. Theseincludetariffs,subsidies,exportcontrols,andregulatorymeasuresintendedeither toprotectdomesticinterestsorstrategicallylimitforeigncommercialactivities,reflecting underlying geopolitical motives or strategic considerations. Such policies could directly reinforce fragmentation patterns observed in global trade and financial markets. To explorethispolicy-drivendimensionofgeopoliticalfragmentation,weempiricallyanalyze detailedrecordsfromtheGlobalTradeAlert(GTA)—specificallytheNewlyImplemented PolicyOutcomes(NIPO)database—whichsystematicallydocumentseconomicpolicyinterventions adopted globally.11 This section evaluates the extent to which economic policyinterventionsalignwithgeopoliticalblocsdefinedbyvariousIPDspecifications,thus complementingandextendingourearlieranalysisoftradeandfinancialfragmentation. 5.1 Data on policy interventions In this section, we utilize the GTA NIPO database, which provides detailed records of economic policy interventions (“acts”) implemented globally since 2017. The GTA NIPO systematically classifies interventions as either distortive or liberalizing. Distortive interventions explicitly discriminate against foreign commercial interests, either by restricting market access or by providing preferential subsidies to domestic firms. Conversely, liberalizing policies are characterized by non-discriminatory interventions that enhance marketaccess. Our analysis specifically focuses on interventions where the implementing jurisdiction and affected jurisdictions differ, restricting attention exclusively to individual countries. Consequently, we exclude interventions enacted by supranational entities such as the EuropeanUnion,leavingtheirinclusionasanavenueforfutureresearch. Giventhatmostinterventionssimultaneouslyaffectmultiplejurisdictions,wefirsttransformthedataintoadyadicformat. Specifically,ifasinglepolicyinterventionaffectsmultiplecountries,werecorditseparatelyforeachaffectedcountry,wherebyeachrestriction 10SinceTaiwanisnotaUNvotingmember,IPDsforTaiwancannotbecalculateddespitetheavailabilityof tradedata. 11FordetailsontheGTANIPOdatabase,seeEvenettetal.(2024). 20

is counted once per impacted jurisdiction. In doing so, we assign each intervention to geopolitical blocs based on the geopolitical distance of the involved country pair, using both the baseline and alternative IPD measures. This dyadic structure enables a more detailedexaminationofpolicydynamicsacrossgeopoliticallines. For the remainder of the paper, we focus on dyadic relationships defined by the baseline IPD and the bloc classification into aligned, nonaligned, or distant, as outlined in Section 4.1.1. Results obtained using alternative IPD measures are presented in the appendix, while results for bloc classification within the US-China spectrum are available upon request. The GTA database also classifies a subset of interventions as “NIPO interventions,” explicitly reflecting strategic economic or geopolitical motivations. Specifically, a NIPO intervention is associated with at least one of six predefined strategic motives: (i) National Security, covering policies aimed at safeguarding national security interests, such as export controls on sensitive technologies; (ii) Resilience and Security of Supply, referring to measuresensuringstabledomesticaccesstoessentialnon-foodproductsandrawmaterials,suchascriticalminerals;(iii)StrategicCompetitiveness,involvingactionsthatpromote domesticinnovationandcompetitivenessinstrategicallyvitalsectors;(iv)ClimateChange Mitigation, capturing interventions explicitly targeting reductions in carbon emissions and facilitating transitions toward renewable energy; (v) Geopolitical Concerns, which includes measures directly addressing threats posed by particular countries or geopolitical blocs, notably economic sanctions (e.g., sanctions imposed after Russia’s invasion of Ukraine); and (vi) Digital Transformation, encompassing policies designed to support the adoption and expansion of digital technologies and infrastructure. By analyzing NIPO interventions separately from general distortions, we obtain deeper insights into the explicitlystrategicorgeopoliticalintentionsunderlyingeconomicpolicyactions. 5.2 Empirical evidence on fragmentation in economic policy Wefirstexaminetotaldistortiveannouncementsbygeopoliticalblocs,usingourbaseline IPD measure. We observe a substantial increase in distortive announcements after Russia’s invasion of Ukraine in 2022, particularly evident in interactions between countries belonging to distant and nonaligned geopolitical blocs. Notably, the sharp rise in discriminatory interventions post-2022 suggests that geopolitical fragmentation is increasingly reflectedthroughtargetedeconomicpolicies. Analyzingnetdistortiveannouncements(distortiveminusliberalizinginterventions)providesadditionalclarity. Thenetmeasureexhibitsapronouncedriseafter2022,emphasizingintensifiedfragmentationprimarilydrivenbydistortivemeasuresoutweighingliberalizinginitiatives. Thedivergencebetweenalignedandnonaligned/distantnetdiscriminatoryannouncementssuggestssignificantgeopoliticalrealignmentsineconomicpolicy. The suggestive evidence on geoeconomic fragmentation in economic policies holds regardless of the IPD measure used to classify countries into blocs. However, the relative importance of trade restrictions between distant countries increases significantly, especially in recent years, when we allocate countries into blocs using the 2023 IPD measure, 21

(a) (b) Figure4: Distortiveandnettraderestrictionannouncementsbybloc Note:Figure(a)showsthesumofdistortivetraderestrictionannouncementsbyyearandbloc, and(b)showsthesumofnettraderestrictionannouncementsbyyearandbloc,calculatedas thedifferencebetweendistortiveandliberalizing. Aligned,nonalignedanddistantblocsare constructedusingtheBaselineIPD,constructedusingallvotes,in2021. usingallvotes. Incontrast,whenwefocusontheeconomicvotes,weobserveawiderdifference between the number of trade restrictions between non-aligned and distant countries,withthenon-alignedrankingfirstforallmeasures.12 5.2.1 Motivesandsectoraldistributionofrestrictions Figure 5 highlights how geopolitical blocs differ significantly in terms of policy interventionswhenwespecificallyfocusonmeasuresexplicitlydrivenbystrategicNIPOmotives. Compared to the broader set of distortive interventions, these strategically motivated actions show even clearer distinctions across blocs, with notably more pronounced differencesinbothtotaldistortive(panela)andnetdistortiveinterventions(panelb). Thissuggeststhatgeopoliticalconsiderationsplayaparticularlyimportantroleinshapingpolicy actions when strategic economic or security-related motives—such as national security, resilience,orgeopoliticalconcerns—areexplicitlyinvolved. The detailed classification by motive thus provides deeper insight into how geopolitical factors increasingly influence national economic policies, reinforcing global fragmentationdynamics. Figure 6 illustrates the evolution of distortive interventions by motive and geopolitical blocfrom2020to2023. Thefigurehighlightsasubstantialincreaseindistortiveinterventions, primarily driven by policies motivated by national security, resilience and security ofsupply,andgeopoliticalconcerns,particularlyin2023. However,thetrendsvaryacross geopoliticalblocs. Distantblocsexhibitasharpriseininterventions,predominantlyjustified by national security and resilience motives, reflecting growing concerns over strategic autonomy and resource security. Nonaligned blocs also show significant increases, but these are primarily driven by geopolitical concerns and resilience-related policies, 12Seeappendix,section7.3foradditionalresults. 22

(a) (b) Figure5: DistortiveandnetNIPOtraderestrictionannouncementsbybloc Note: Figure (a) shows the sum of NIPO distortive trade restriction announcements by year and bloc, and (b) shows the sum of NIPO net trade restriction announcements by year and bloc, calculated as the difference between distortive and liberalizing. An act is classified as NIPOifitmentionsatleastoneofthefollowingmotives: Strategiccompetitiveness,national security,digitaltransformation,resilience/securityofsupply,geopoliticalconcerns,orclimate changemitigation. Aligned,nonalignedanddistantblocsareconstructedusingtheBaseline IPD,constructedusingallvotes,in2021. suggestingamorenuancedpositioninginthegeopoliticallandscape. Incontrast,aligned blocs display a more moderate increase in distortive interventions, likely reflecting more stable policy coordination within the bloc rather than a reactive escalation of economic measures. ThesectoralanalysispresentedinFigure7providesdeeperinsightsintothestrategicnature of economic policy interventions. Distortive interventions are heavily concentrated inkeysectors,particularlycriticalminerals,dual-useproducts,advancedtechnology,and industrial raw materials, with a marked increase following 2022. However, the extent of intervention varies across geopolitical blocs. Distant and nonaligned blocs have intensifiedinterventionsinthesesectors,likelyaspartofbroadereffortstosecuretechnological leadershipandessentialresources. Incontrast,alignedblocshavealsoincreasedinterventions, but at a relatively more moderate pace, potentially reflecting a different approach to industrial policy rather than direct strategic competition. These trends highlight the sector-specific nature of economic policy fragmentation and suggest that interventions arenotonlyaresponsetogeopoliticaltensionsbutalsopartofbroadereconomicsecurity strategies. 23

Figure6: Distortiveinterventionsbymotivesandbloc Note: Distribution of NIPO distortive inteventions by motive and bloc. Aligned, nonaligned anddistantblocsareconstructedusingtheBaselineIPD,constructedusingallvotes,in2021. Overall,theclearstrategicdifferentiationacrossgeopoliticalblocs,drivenbybothmotive and sector-specific interventions, underscores the deep strategic and geopolitical dimensions shaping economic policy fragmentation. These dynamics have critical implications forthefuturestructureofglobalmarketsandinternationaleconomiccooperation. 24

Figure7: Discriminatoryinterventionsbyblocandsector Note: Distribution of NIPO distortive inteventions by sector and bloc. Aligned, nonaligned anddistantblocsareconstructedusingtheBaselineIPD,constructedusingallvotes,in2021. 5.2.2 Typesofpolicyinterventions Finally, in figure 8 we analyze the distribution among types of distortive interventions. In the GTA NIPO database, each intervention is classified into one of the following categories. First, Horizontal implies that the policy applies broadly across all sectors within a country. Second, R&D Related refers to interventions that target research, innovation, or R&D activities. Third, Infrastructure, Transport, Cargo or Logistics refers to policies related to industrial and transport infrastructure, cargo handling, and logistics. Fourth, Support Electrical Energy includes industrial policies related to electricity generation and supply. Fifth,RecyclingServiceinvolvespoliciesrelatedtorecyclingactivities. Sixth,FDIScreening Mechanism denotes procedures assessing, investigating, authorizing, conditioning, prohibiting, or reversing inward or outward FDI. Finally, Sanctions includes trade-related sanctions imposed in security or foreign-policy contexts. We omit Recycling Service and 25

FDI Screening Mechanism from the analysis, as no distortive interventions were identifiedinthesecategoriesintheperiodunderanalysis. The analysis reveals significant heterogeneity across intervention types and geopolitical alignments. Notably, sanctions have surged dramatically after 2022, predominantly among distant and nonaligned blocs, illustrating their increased reliance on economic measures explicitly aimed at isolating geopolitical rivals. Interventions related to infrastructure,transport,cargo,orlogisticsandsupportforelectricalenergyalsoshownotable increases, again largely concentrated among distant and nonaligned blocs, highlighting strategicattemptstocontrolcriticalinfrastructureandenergyresources. Figure8: Distortiveinterventionsbytypeandbloc Note: Distribution of NIPO distortive inteventions by type and bloc. Aligned, nonaligned anddistantblocsareconstructedusingtheBaselineIPD,constructedusingallvotes,in2021. WeomitRecyclingServiceandFDIScreeningMechanismfromtheanalysis,asnodistortive interventionswereidentifiedinthesecategoriesintheperiodunderanalysis. Horizontal and R&D-related restrictions display differing patterns: horizontal restric- 26

tions (broad policy measures not sector-specific) significantly increased in aligned and nonalignedblocs,indicatingbroader-basedpolicyactionsaimedatreshoringorreinforcing intra-bloc cooperation. In contrast, R&D-related restrictions spiked sharply among distant blocs, reflecting intensified strategic competition in innovation and technological advancement. These patterns emphasize that geopolitical blocs strategically choose intervention types aligningcloselywiththeirbroadereconomicandsecurityobjectives. Thepronounceduse ofsanctionsandtargetedinterventionsrelatedtoinfrastructure,R&D,andenergyamong distant blocs indicates increasingly explicit geopolitical contention. These strategic intervention patterns further deepen global economic fragmentation, reflecting a policy environmentshapedbyintensifiedgeopoliticalrivalry. 5.3 Regression estimation results Table 5 presents the estimation results of equation (3), analyzing how geopolitical fragmentation—capturedbyalternativemeasuresofgeopoliticaldistance(IPDs)—influences the frequency of distortive economic policy interventions. We use Poisson pseudo- maximum likelihood (PPML) estimations, which are well-suited for count-data structures. Thisregressionanalysiscomplementsthedescriptiveevidencepresentedinprevioussections, providing a robust quantitative assessment of the extent of economic policy fragmentationassociatedwithgeopoliticalalignment. The baseline IPD measure (column 1) yields a positive but statistically insignificant coefficient for Between Bloc × Post (0.029), indicating limited evidence of increased distortive interventionsbetweengeopoliticallydistantblocsfollowingRussia’sinvasionofUkraine in 2022 under this specification. In contrast, alternative IPD measures that better capture recent geopolitical realignments yield stronger and statistically significant results. Specifically, when employing the 2023 IPD (column 2) or the subsample IPD (column 4), the estimated coefficients increase markedly to approximately 32.7% and 30%, respectively. These results imply that after the Russian invasion of Ukraine, geopolitically distant country pairs implemented, on average, roughly 30–33% more distortive policy interventions against each other compared to aligned country pairs, reflecting substantial policy-driven fragmentation. Conversely, the IPD based exclusively on economic votes (column3)yieldsnostatisticallysignificantresults,suggestingthatobservedpolicyfragmentation is driven more strongly by broader geopolitical tensions rather than purely economic alignment. Further robustness checks with alternative fixed-effect structures areprovidedinTable16oftheappendix.1314 13Replacingsource-yearanddestination-yearfixedeffectswithsimpletimefixedeffectsamplifiesthefragmentation effects further. Under these specifications, the largest fragmentation effect occurs with the 2023 IPD measure (approximately 39.8%), highlighting an even stronger increase in targeted distortive interventionspost-invasion. Additionally,coefficientsforNonaligned×Postbecomesignificantlypositive (16.1%),indicatingintensifiedpolicyinterventionsalsoamongnonalignedcountrypairs. 14Estimationresultsofequation(3)usingnetinterventions,andestimationresultsofequations(1)and(2) forbothdependentvariablesareavailableuponrequest. 27

Table5: RegressionResults: DistortiveInterventions Description BaselineIPD IPDall IPDeconomic IPDall (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (1) (2) (3) (4) TotalDistortiveInterventions DistantBloc×Post 0.029 0.283*** 0.025 0.262*** (0.030) (0.049) (0.025) (0.047) Nonaligned×Post -0.028 0.163*** -0.017 0.070 (0.023) (0.046) (0.023) (0.046) PseudoR2 0.607 0.435 0.606 0.434 Observations 15,665 15,665 15,606 15,606 Note:Significancethresholds:***p<0.01,**p<0.05,*p<0.1.Resultestimationsofequation (3)usingPoissonpseudo-maximumlikelihood(PPML),usingannualdatafortheperiod2017- 2023,fromGTANIPO.Weincludecountry-pair,source×time,anddestination×timefixed effects. Standarderrorsareclusteredatthecountry-pairlevel. Coefficientinterpretationisthe (cid:0) (cid:1) following: ecoefficient−1 ×100. PostisadummythatcapturesthepostinvasionofUkraine periodandtakesthevalue1fortheyears2022and2023.Eachcolumnshowstheresultsusing a different IPD measure to construct the country blocs. (1) uses the Baseline IPD measure, which is the 2021 values of IPDs estimated using UNGA voting data from 1946-2023 across allvotecategories;(2)usesthe2023valuesofIPDsestimatedusingUNGAvotingdatafrom 1946-2023acrossallvotecategories;(3)usesthe2021valuesofEconomicsIPD,whichnarrows thefocustoeconomicvoteswhilemaintaininghistoricalcoveragefrom1971onwards;(4)uses 2021valuesofIPDestimatedusingasubsampleofvotesaftertheColdWar(1990–2023)while maintainingallvotecategories. Overall, these findings strongly support the interpretation that deliberate policy measures, such as tariffs and sanctions, contribute directly to reinforcing geopolitical fragmentation observed in trade and financial markets. The analysis underscores the importanceofmethodologicaltransparency,asconclusionsregardingeconomicpolicyfragmentationaresubstantiallyinfluencedbyhowgeopoliticaldistancesaremeasured. 6 Conclusion Our study highlights significant methodological sensitivities in measuring geoeconomic fragmentation and underscores distinct economic impacts across trade flows, financial portfolios, and economic policy interventions. Trade relationships display robust and consistent fragmentation along geopolitical lines, particularly evident in strategic technology sectors and policy interventions driven by national security and geopolitical concerns. Medium-tech and low-tech trade sectors exhibit especially pronounced responses. In contrast, financial portfolios appear comparatively resilient, with weaker and contextsensitive fragmentation effects, suggesting financial markets may mitigate some impacts ofgeopoliticaltensionsthroughmarket-basedmechanisms. These results emphasize the critical importance of methodological transparency in constructing geopolitical distance measures, as seemingly minor methodological decisions materially influence conclusions. For immediate policy concerns, IPD measures incor- 28

porating recent geopolitical events (such as the 2023 IPD) are most informative. For structural and long-term analyses, economic-vote IPDs or post-Cold War measures provide additional stability and insight. Policymakers should therefore carefully consider methodologicalchoicestoaccuratelyassessrisksanddesignstrategicresponsesthatbalance economic integration, security, and resilience in an increasingly fragmented world economy. 29

7 Appendix 7.1 Fragmentation in financial markets: robustness analysis Inthissectionweperformseveralrobustnesscheckstoouranalysisofthedegreeoffragmentationinfinancialmarkets. 7.1.1 Internationalfinancialcenters International financial centers (IFC) can blur the geopolitical distances between recipient and ultimate investor countries, as shown by the literature.15 In our portfolio holdings analysis, we do not include well-known financial centers, such as Bermuda, the British VirginIslands,theCaymanIslands,andHongKongSAR,assourcecountries. Inthissection, we present a robustness check where we further exclude other countries frequently identified as international financial centers, including Ireland, Luxembourg, the Netherlands,andSingapore. Tables 6, 7 and 8 show the estimation results for equations (1), (2) and (3), respectively, whenweexcludeIFCfromthesample. Thetablesshowthattheoverallpatternofresults remainsconsistent,withonlyminorchangesinmagnitudeandsignificance. Thispattern suggests that financial centers play a role in channeling investments across geopolitical blocs,buttheirexclusiondoesnotfundamentallyaltertheobservedfragmentationtrends inportfolioholdings. Table6: RegressionResultsofequation1excludingIFC BaselineIPD IPDall IPDeconomic IPDall (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (1) (2) (3) (4) PortfolioHoldings BetweenBloc×Post -0.018 -0.012 0.001 -0.018 Std. Error (0.016) (0.012) (0.012) (0.016) Nonaligned×Post -0.014 -0.004 -0.017 -0.014 Std. Error (0.020) (0.019) (0.021) (0.020) Observations 214,165 214,577 214,184 214,165 Significancethresholds:***p<0.01,**p<0.05,*p<0.1 15See,forinstance,Coppolaetal.(2021). 30

Table7: RegressionResultsequation(2)excludingIFC Description IPD all votes Economic IPD IPD all votes Economic IPD IPD × Post IPD (1) (2) (3) (4) Portfolio holdings β Coefficient -0.071* -0.063* -0.057 0.037 Std. Error (0.037) (0.037) (0.063) (0.037) Observations 106,503 105,544 106,503 105,544 Significancethresholds: ***p<0.01,**p<0.05,*p<0.1 Table8: RegressionResultsequation(3)excludingIFC Description Baseline IPD IPD all IPD economic (complete, 2021) (complete, 2023) (complete, 2021) (1) (2) (3) Portfolio holdings Distant x Post -0.011 -0.011 -0.014 Std. Error (0.014) (0.014) (0.014) Nonaligned x Post -0.006 -0.001 -0.017** Std. Error (0.007) (0.008) (0.008) Observations 213,286 209,899 213,286 Significancethresholds: ***p < 0.01,**p < 0.05,*p < 0.1 7.1.2 TheroleoftheUnitedStatesinfinancialmarkets In this section we show the important role played by the United States in international financial markets whenconsidering general blocs of geopolitical distance, definedby the country-pair measures. The IPD measures position the U.S. in the right tail of the distribution, indicating that it is distant from most of its international counterparts, as discussed in our analysis of Figure 2. When we exclude the U.S. from the estimation regressions in equation (3) for portfolio holdings, the effect of geopolitical distance becomes negative and statistically significant at the 1% level under the economic IPD specification. Specifically, portfolio holdings between distant country pairs decline by 2.5% in the post-invasion period. We obtain similar results if we include country-pair and time fixed effectsinsteadofcountry-pair,sourceanddestinationbytimefixedeffects. 31

Table9: RegressionResultsequation(3)excludingtheUnitedStates Description Baseline IPD IPD all IPD economic (complete, 2021) (complete, 2023) (complete, 2021) (1) (2) (3) Portfolio holdings Distant x Post Coefficient -0.014 -0.009 -0.025*** Std. Error (0.010) (0.010) (0.010) Nonaligned x Post Coefficient -0.009 0.002 -0.021** Std. Error (0.010) (0.009) (0.010) Observations 226,246 222,751 226,246 Significancethresholds: ***p<0.01,**p<0.05,*p<0.1 7.1.3 Alternativedefinitions Inthissectionweexploretheuseoftwoalternativedefinitionsforthedependentvariable to study the degree of fragmentation in financial markets: portfolio share and financial flows(inbillionsofUSD). Our sensitivity analysis further confirms the nuanced impact of geopolitical fragmentation on financial integration. Using alternative measures such as portfolio shares and financial flows, we find that financial fragmentation effects remain sensitive to IPD specifications. Whileportfoliosharefragmentationappearsweakandmostlyinsignificant,financial flows measured in dollar terms show a clear negative and statistically significant response in most specifications, indicating a decline in financial flows between geopolitically distant country pairs following the Russian invasion of Ukraine. The magnitude and significance of these results vary considerably across IPD definitions, highlighting a stronger and more consistent effect when employing recent geopolitical distance metrics, notably the 2023 IPD. This pattern reinforces the conclusion from our main analysis: geopolitical fragmentation has more pronounced and robust effects on trade and economic policies, whereas financial integration shows weaker and more context-sensitive responses. 32

Table10: RegressionResults: PortfolioShareandFinancialFlows Description BaselineIPD IPDall IPDeconomic IPDall (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (1) (2) (3) (4) (I)PortfolioShare BetweenBloc×Post 0.001 -0.028 -0.139* 0.002 (0.075) (0.099) (0.084) (0.075) Nonaligned×Post -0.182 -0.132 -0.114 -0.187 (0.214) (0.117) (0.196) (0.213) Observations 119,859 120,071 119,934 119,859 (II)FinancialFlows BetweenBloc×Post -1.807*** -2.244*** -0.652 -1.807*** (0.632) (0.714) (0.531) (0.632) Nonaligned×Post -2.722*** -2.522*** -0.199 -2.724*** (0.723) (0.786) (0.808) (0.723) Observations 234,212 234,746 234,343 234,212 Significancethresholds:***p<0.01,**p<0.05,*p<0.1 Table11: RegressionResults: PortfolioShareandFinancialFlows Description IPDallvotes EconomicIPD IPDallvotes EconomicIPD IPD×Post IPD (1) (2) (3) (4) (I)PortfolioShare β Coefficient -0.050 -0.051 -0.111 -0.039 (0.033) (0.037) (0.093) (0.045) Observations 58,355 58,204 58,355 58,204 (II)FinancialFlows β Coefficient -1.152*** -1.049*** 0.268 1.554** (0.342) (0.318) (0.649) (0.614) Observations 115,142 114,129 115,142 114,129 Significancethresholds: ***p<0.01,**p<0.05,*p<0.1 33

Table12: RegressionResults: PortfolioShareandFinancialFlows Description BaselineandEconomicIPD BaselineIPD IPDall IPDeconomic IPDall (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (1) (2) (3) (4) (I)PortfolioShare DistantBloc×Post -0.066 -0.062 -0.187** -0.057 (0.083) (0.075) (0.081) (0.084) Nonaligned×Post 0.033 0.068 0.036 0.028 (0.064) (0.057) (0.058) (0.064) Observations 233,475 229,879 233,475 233,475 (II)FinancialFlows DistantBloc×Post -1.392* -1.853*** -1.929*** -1.306∧∧ (0.838) (0.613) (0.638) (0.840) Nonaligned×Post -0.128 0.156 0.251 -0.159 (0.258) (0.225) (0.308) (0.251) Observations 233,475 229,879 233,475 233,475 Significancethresholds:***p<0.01,**p<0.05,*p<0.1 7.2 Trade by technology class: robustness analysis In this section we conduct robustness checks for our gravity equations of geopolitical distanceontradebytechnologyclassification. SimilarlytoTable3,inTable13,wegeneralizeouranalysistothefulldistributionofIPD(s,d)insteadofonlyworkingonthesegmenteddistribution. Ourresultsindicatethatwhenusingthedistributionofgeopolitical distance,tradeflowsbetweendistantcountriesforallgoodsclassesdeclinedsignificantly in the post-invasion period when we use the latest IPD values and declined for low and high-techgoodswhenonlyeconomicvotesareused. Additionally,weobservesmalland statistically insignificant relationships for nonaligned countries in the post-invasion period. These results are consistent with Table 3 as well our larger finding that results are sensitivetothechoiceofIPDmeasureandblocdefinition. In Table 14, we replicate our estimation results from equation (2) with our trade by technology class. Across all specifications and technology classifications, we find negative and statistically significant relationships between geopolitical distance and trade flows both in the post-invasion period and over the full sample period. Consistent with Table 2, the effect is slightly weaker when using Economic IPDs to generate blocs. The diminishingeffectofgeopoliticaldistanceontradeastechnologyclassincreasesisalsoevident inourresults. 34

Table13: RegressionResults Description BaselineIPD IPDall IPDeconomic IPDall (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (1) (2) (3) (4) (I)High-techTrade Distant×PostCoefficient -0.0261 -0.095*** -0.156** -0.050 Std. Error (0.038) (0.036) (0.066) (0.042) Nonaligned×PostCoefficient 0.012 0.016 0.011 0.017 Std. Error (0.024) (0.031) (0.025) (0.024) Observations 270,157 261,452 267,739 270157 (II)Medium-techTrade Distant×PostCoefficient -0.098* -0.182*** -0.031 -0.086 Std. Error (0.055) (0.062) (0.071) (0.058) Nonaligned×PostCoefficient -0.083* -0.089* -0.007 -0.0864** Std. Error (0.040) (0.046) (0.037) (0.037) Observations 232,540 225,038 230,788 232,540 (III)Low-techTrade Distant×PostCoefficient -0.0827** -0.109*** -0.121*** -0.119*** Std. Error (0.032) (0.034) (0.039) (0.033) Nonaligned×PostCoefficient -0.010 0.020 -0.010 -0.016 Std. Error (0.021) (0.026) (0.022) (0.021) Observations 267,022 259,244 264,616 267,022 Significancethresholds:***p<0.01,**p<0.05,*p<0.1 35

Table14: RegressionResultswithIPD Description IPD all votes Economic IPD IPD all votes Economic IPD IPD × Post IPD (1) (2) (3) (4) (I) High-tech Trade β Coefficient -0.078*** -0.072*** -0.088*** -0.075*** Std. Error (0.020) (0.019) (0.027) (0.024) Observations 292,042 292,042 290,161 292,042 (II) Medium-tech Trade β Coefficient -0.106*** -0.098*** -0.123*** -0.106*** Std. Error (0.027) (0.025) (0.034) (0.031) Observations 256,506 256,506 256,506 256,506 (III) Low-tech Trade β Coefficient -0.082*** -0.076*** -0.090*** -0.075*** Std. Error (0.014) (0.013) (0.021) (0.020) Observations 288,251 288,251 288,251 288,251 Significancethresholds: ***p<0.01,**p<0.05,*p<0.1 Table15: ISICRevision–3TechnologyClassifications High-tech 30,33: Computer,electronicandopticalproducts 31-32: Electricalequipment 24: Chemicalsandpharmaceuticals 29: Machineryandequipment 34-35: Transportequipment Low-tech 15-16: Food,beverages,tobacco 17-19: Textiles,apparel,leather 20-22: Woodproducts 36: Furniture&OtherManufacturing 26: Non-metallicmineralproducts Medium-tech 27-28: Basicmetalsandmetalproducts 25: Rubberandplastics 23: Petroleumproducts 36

7.3 Economic policy fragmentation 7.3.1 Motivesdistributionofrestrictions Forcompleteness,inthissectionwepresentthedistributionofdistortiveinterventionsby motiveandbloc,underalternativeIPDs. Figure9showstheresultsthatclassifycountries into aligned, nonaligned, and distant using the 2023 IPD with all votes, while 10 uses only economic votes. The objective is to examine whether the distribution of economic policy interventionsacrossstrategicmotivesvarieswhenusingdifferentIPDdefinitions. Reclassifying country blocs with alternative IPD metrics does not fundamentally alter the conclusions from our main analysis, but highlights important nuances. Using the 2023 IPD, distortive interventions related to national security, resilience, and geopolitical concerns become more prominent among distant blocs, reflecting recent geopolitical realignments. Conversely, when employing the economic-votes IPD, interventions driven explicitlybyeconomicmotivessuchasresilienceandstrategiccompetitivenessgainrelative importance. These findings reinforce that methodological choices in defining geopolitical blocs affect not just the observed magnitude but also the strategic composition of economicpolicyfragmentation. Figure9: Countryblocsdefinedusingthe2023IPDwithallvotes. Notes: DistributionofNIPOdistortiveinteventionsbymotiveandbloc. Aligned,nonaligned anddistantblocsareconstructedusingtheIPD,constructedusingallvotes,in2023. 37

Figure10: Countryblocsdefinedusingthe2021EconomicIPD. Notes: DistributionofNIPOdistortiveinteventionsbymotiveandbloc. Aligned,nonaligned anddistantblocsareconstructedusingtheEconomicIPD,constructedusingeconomicvotes, in2021. 7.3.2 Sectoraldistributionofrestrictions Figure 11 shows the sectoral results that classify countries using the IPD 2023 with all votes, while 12 uses only economic votes. This analysis further confirms our main conclusion regarding the strategic targeting of economic policy interventions, though differences emerge when varying IPD definitions. The sectoral distribution of policy interventions remains broadly consistent across IPD specifications, with critical minerals, dual-use products, advanced technology, and industrial raw materials consistently targeted by geopolitically distant and nonaligned blocs. However, notable differences arise dependingontheIPDmeasurechosen: usingthe2023IPD,theconcentrationofinterventionsinadvancedtechnologiesandcriticalsectorsisparticularlypronouncedamongdistantblocs,highlightingrecentshiftstowardintensifiedstrategiccompetition. Incontrast, when using the economic-votes IPD, the sectoral focus appears less sharply differentiated between blocs, suggesting that economic-voting-based alignment captures broader and less polarizing sectoral interventions. Overall, the core findings regarding sectoral targeting persist, though their intensity and polarization vary with IPD methodological choices. 38

Figure11: Countryblocsdefinedusingthe2023IPDwithallvotes. Notes: DistributionofNIPOdistortiveinteventionsbysectorandbloc. Aligned,nonaligned anddistantblocsareconstructedusingtheIPDconstructedusingallvotes,in2023. 39

Figure12: Countryblocsdefinedusingthe2021EconomicIPD. Notes: DistributionofNIPOdistortiveinteventionsbymotiveandbloc. Aligned,nonaligned anddistantblocsareconstructedusingtheEconomicIPD,constructedusingeconomicvotes, in2021. 7.3.3 Additionalresults In table 16 we show the sensitivity of the results of the estimation of equation (3) for distortive interventions to the inclusion of different sets of fixed effects. Removing sourceyearanddestination-yearfixedeffectsandaddingsimpletimefixedeffects(columns5–8) further amplifies fragmentation effects. Under these specifications, the largest observed fragmentation effects occur with the 2023 IPD specification (approximately 39.8%), suggesting an even greater increase in targeted distortive interventions post-invasion. Additionally, the Nonaligned × Post coefficients also become significantly positive (16.1%), indicatingthatnonalignedcountriesalsointensifiedtheirpolicyinterventions. 40

Table16: RegressionResults: DistortiveInterventions Description BaselineIPD IPDall IPDeconomic IPDall BaselineIPD IPDall IPDeconomic IPDall (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (complete,2021) (complete,2023) (complete,2021) (subsample,2021) (I)TotalDistortiveInterventions DistantBloc×Post 0.029 0.283*** 0.025 0.262*** 0.073** 0.335*** 0.021 0.293*** (0.030) (0.049) (0.025) (0.047) (0.033) (0.052) (0.030) (0.049) Nonaligned×Post -0.028 0.163*** -0.017 0.070 0.006 0.149*** -0.038* 0.164*** (0.023) (0.046) (0.023) (0.046) (0.024) (0.048) (0.022) (0.045) PseudoR2 0.607 0.435 0.606 0.434 0.607 0.435 0.607 0.435 Observations 15,665 15,665 15,606 15,606 15,604 15,604 15,665 15,665 Country-PairFE Y Y Y Y Y Y Y Y TimeFE N N N N Y Y Y Y Source×YearFE Y Y Y Y N N N N Destination×YearFE Y Y Y Y N N N N Notes: Significance thresholds: *** p < 0.01, ** p < 0.05, * p < 0.1. Result estimations of equation (3) using Poisson pseudo-maximum likelihood (PPML), using annual data for the period 2017-2023, from GTA NIPO. Standard errors are clustered at the country-pair level. (cid:0) (cid:1) Coefficientinterpretationisthefollowing: ecoefficient−1 ×100. 41

References Aiyar,M.S.,Chen,M.J.,Ebeke,C.,Ebeke,M.C.H.,Garcia-Saltos,M.R.,Gudmundsson, T.,Ilyina,M.A.,Kangur,M.A.,Kunaratskul,T.,Rodriguez,M.S.L.,etal.(2023a). Geoeconomicfragmentationandthefutureofmultilateralism. InternationalMonetaryFund. Aiyar, S., Presbitero, A., and Ruta, M. (2023b). Geoeconomic Fragmentation: The Economic RisksfromaFracturedWorldEconomy. CEPRPress. Bailey, M. A., Strezhnev, A., and Voeten, E. (2017). Estimating dynamic state preferences fromunitednationsvotingdata. JournalofConflictResolution,61(2):430–456. Blanga-Gubbay,M.andRub´ınova´,S.(2023). Istheglobaleconomyfragmenting? TechnicalReportERSD-2023-10,WorldTradeOrganization. Campos, R. G., Freund, C., and Ruta, M. (2024). The economics of fragmentation: the geopolitical risks of a fractured world economy. Technical report, International MonetaryFund. Catalan, M. M., Fendoglu, M. S., and Tsuruga, T. (2024). A gravity model of geopolitics and financialfragmentation. InternationalMonetaryFund. Coppola, A., Maggiori, M., Neiman, B., and Schreger, J. (2021). Redrawing the map of global capital flows: The role of cross-border financing and tax havens. The Quarterly JournalofEconomics,136(3):1499–1556. Evenett, S., Jakubik, A., Mart´ın, F., and Ruta, M. (2024). The return of industrial policy in data. TheWorldEconomy,47(7):2762–2788. Ferna´ndez-Villaverde, J., Mineyama, T., and Song, D. (2024). Are we fragmented yet? measuring geopolitical fragmentation and its causal effect. NBER Working Paper No. 32038. Gaulier,G.andZignago,S.(2010). Baci: Internationaltradedatabaseattheproduct-level. the1994-2007version. WorkingPapers2010-23,CEPII. Gopinath, G., Gourinchas, P.-O., Presbitero, A. F., and Topalova, P. (2025). Changing globallinkages: Anewcoldwar? JournalofInternationalEconomics,153:104042. Hakobyan, S., Meleshchuk, S., and Zymek, R. (2024). Divided we fall: Differential exposuretogeopoliticalfragmentationintrade. IMFWorkingPaper. Jakubik, A. and Ruta, M. (2023). Trading with friends in uncertain times. Journal of Policy Modeling,45(4):768–780. Kleinman, B., Liu, E., and Redding, S. J. (2024). International friends and enemies. AmericanEconomicJournal: Macroeconomics,16(4):350–385. Qiu, H., Shin, H. S., and Zhang, L. S. (2023). Mapping the realignment of global value chains. TechnicalReport78,BankforInternationalSettlements. 42