The effect of fluctuations in bilateral relations on trade: evidence from China and ASEAN countries

The paper examines how fluctuations in bilateral political relations between China and ASEAN countries affect bilateral trade over 2001–2020. Against the backdrop of China’s rapid rise, deepening economic ties with ASEAN, and periods of global uncertainty (e.g., the financial crisis and COVID-19), the study addresses whether warmer political relations promote trade and how this effect differs between imports and exports. Motivated by theories linking international politics and commerce and the growing relevance of China–ASEAN relations, the authors construct high-frequency monthly indicators of political relations using GDELT event data and test three hypotheses: (H1) improved bilateral relations promote trade; (H2) improved relations promote China’s exports and imports; and (H3) directional, one-way behavioural attitudes (China toward ASEAN vs. ASEAN toward China) may differentially impact trade flows.

The review traces realist and liberal perspectives on the politics–trade nexus. Realist accounts argue political alignment and alliances increase trade, while conflict, terrorism, and wars suppress it (Pollins, 1989a; Gowa & Mansfield, 1993; Nitsch & Schumacher, 2004; Glick & Taylor, 2010; Che et al., 2015). Liberal perspectives posit that trade interdependence reduces political conflict (Polachek, 1978, 1980; Gartzke, 2007) but can also generate tensions when seen as resource competition (Reuveny & Kang, 1996). The literature addresses colonial ties, wars, and contemporary political events, with mixed findings on whether political relations predict trade (e.g., Pollins, 1989b; Fuchs & Klann, 2013; Keshk et al., 2004). Measurement approaches include UNGA voting affinity, counts of diplomatic engagements (e.g., high-level visits), and event data–based composite indices (e.g., Tsinghua’s Sino-Foreign Relations Database). Event data (GDELT) capture high-frequency political dynamics and have been used to link political relations to trade, particularly imports (Davis et al., 2019; Li et al., 2021). Gaps include limited focus on emerging countries and directional asymmetries in political behavior. The study proposes three hypotheses: H1: improved relations promote trade; H2: improved relations promote both China’s exports and imports; H3: one-way behavioural fluctuations (China→ASEAN vs. ASEAN→China) may have different effects on trade.

Data and measures: The core explanatory variable is the fluctuation of bilateral relations derived from GDELT event data (1979–present; v2.0). Events are coded using CAMEO with Goldstein Scale (−10 conflict to +10 cooperation). For each China–ASEAN dyad and month (Jan 2001–Dec 2020), the study computes a weighted average Goldstein score, weighting each event by NumArticles (the number of media mentions). Daily weighted averages are aggregated to monthly averages. Directional series capture China→ASEAN (cpir) and ASEAN→China (apir) behaviours, enabling analysis of one-way attitudes. In total, 1,204,126 interaction event records are analysed. Variables: The dependent variables are monthly China–ASEAN total trade (Intrade), China’s imports (Inimport), and China’s exports (Inexport), all log-transformed. The key explanatory variables are pir (bilateral average Goldstein score), and in alternative specifications cpir and apir. Controls include log average market size (Inmkt; average GDP of China and partner j; monthly values derived from lower-frequency data using EViews), log geographic distance (Indist; CEPII), log combined population (Inpop), and dummies for shared border (bou), same legal system (law), presence in same FTA (fta), and APEC membership (apec). Time fixed effects (γ_t) are included. Data sources: GDELT, IMF, World Bank, CEIC, UN Comtrade, CEPII. Model: Gravity-model-based panel regressions with fixed effects are estimated: - lntrade_jt = β0 + β1 pir_jt + β2 lnmkt_jt + β3 lndist_j + β4 lnpop_jt + β5 bou_j + β6 law_j + β7 fta_j + β8 apec_j + γ_t + ε_jt - lnimport_jt and lnexport_jt are estimated analogously. Alternative models replace pir with cpir and apir. Estimation and diagnostics: Panel unit root tests (IPS, LLC) indicate stationarity. Poolability and Hausman tests favour fixed effects. Cross-sectional dependence (Pesaran test) is present; hence Feasible GLS (FGLS) is employed for robustness. Additional robustness: address potential endogeneity using 2SLS and GMM with lagged pir as instrument; replace the weighted Goldstein measure with unweighted averages; control for multilateral resistance (MRES) following Head & Mayer (2002) using weighted bilateral trade freedom; and estimate Poisson Pseudo-Maximum Likelihood (PPML) to handle zero trade flows. Due to missing Myanmar export data (2001–2010), multilateral resistance analysis excludes Myanmar. Descriptive statistics (N=2,400 panel observations) summarize variables (e.g., mean pir=1.636, cpir=1.853, apir=1.810). The study also computes Trade Intensity (TI) and Hubness Measurement (HM) indices (quarterly, 2016–2020) to contextualize bilateral dependence.

- Core effect: Bilateral relations (pir) positively and significantly affect total China–ASEAN trade. In FGLS, pir coefficients are positive and significant for total trade and exports; the import effect is not statistically significant. - Directional effects: China’s one-way attitude (cpir) positively and significantly increases total trade and exports; ASEAN’s one-way attitude toward China (apir) is not significant for trade, imports, or exports. - Controls: Larger market size (lnmkt) strongly increases trade; geographic distance (lndist) significantly reduces trade, indicating trade costs. Population size relates positively to total trade and exports but negatively to China’s imports. Shared borders and differing legal systems associate with lower trade. APEC co-membership promotes trade. Surprisingly, same-FTA membership shows a negative association with bilateral trade in this sample. - Robustness: Results hold under 2SLS (pir≈0.384, t≈3.24), GMM (pir≈0.836, t≈6.77), using unweighted Goldstein averages, controlling multilateral resistance (pir≈0.0229 vs. 0.0174 baseline), and PPML (pir≈0.0130, p<0.01). - Descriptive dependence measures: TI indices (2016–2020) indicate generally intensive China–ASEAN trade; e.g., China–Myanmar TI peaks at 5.429 (Q4 2020). HM indices show relatively higher bilateral dependence with Vietnam, Singapore, Thailand, Malaysia, and Indonesia; Vietnam’s HM peaks at 4.39 (Q4 2020). Trade dependence with most ASEAN countries trends upward over time.

Findings validate the hypotheses. H1: Improved political relations, measured at monthly frequency via GDELT Goldstein scores, are associated with higher bilateral trade between China and ASEAN. H2: The positive effect is concentrated on China’s exports; imports are not significantly affected, suggesting asymmetric sensitivity of trade components to political relations. H3: Directionality matters—China’s cooperative stance (cpir) significantly boosts bilateral trade and exports, whereas ASEAN’s stance toward China (apir) shows no significant effect, consistent with China’s outsized role in the dyad during the sample period. These results align with realist arguments that friendly ties facilitate commerce and complement literature emphasizing institutional and political determinants of trade. Control-variable patterns highlight classic gravity mechanisms (market size and distance) and the role of multilateral organizations (APEC) in promoting trade. The negative FTA coefficient is interpreted as evidence that overlapping bilateralism may fragment regionalism, reflecting the complex institutional environment in East Asia. Contextual explanations include China’s WTO accession (2001) accelerating export growth, the 2008–09 financial crisis increasing uncertainty, and post-2013 Belt and Road cooperation shaping political-economic linkages. Policy implications: maintain stable political ties, enhance mutual understanding and policy coordination, address border and legal-system frictions, improve logistics and infrastructure, and use multilateral platforms (APEC, RCEP) to bolster resilient trade ties.

The study introduces a high-frequency, event-data-based measure of bilateral relations (GDELT Goldstein scores) into a gravity framework to assess China–ASEAN trade from 2001–2020. It finds that improved bilateral relations significantly increase bilateral trade, with pronounced effects on China’s exports; imports show no significant response. Directional analysis reveals that China’s one-way cooperative attitude toward ASEAN drives trade more than ASEAN’s attitude toward China. Supplementary TI and HM indices show that China–ASEAN trade is intensive and that China’s trade dependence on most ASEAN partners has steadily grown. Contributions include: (1) methodological integration of big event data with gravity modeling at monthly frequency; (2) directional decomposition of political relations’ effects; and (3) policy-relevant evidence on the trade benefits of stable political ties and multilateral cooperation. Future research could extend the framework to other regions, incorporate sectoral or firm-level trade responses, evaluate heterogeneous effects across ASEAN members, and integrate alternative political-risk or media-text measures to triangulate political relations.

- Measurement limitations: Political relations are proxied by GDELT Goldstein scores, which depend on media coverage and weighting by NumArticles; alternative media ecosystems may bias salience. Unweighted robustness helps, but measurement error may remain. - Data frequency and interpolation: Monthly GDP is constructed from lower-frequency data via temporal disaggregation, potentially introducing noise. - Endogeneity: Although addressed with lagged instruments (2SLS, GMM), residual reverse causality between trade and political relations may persist. - Cross-sectional dependence and omitted variables: FGLS and controls mitigate but may not fully eliminate common shocks or unobserved heterogeneity (e.g., sanctions, supply-chain shocks). - Multilateral resistance estimation excludes Myanmar due to missing export data (2001–2010), slightly narrowing coverage for that robustness check. - Generalizability: Results pertain to China–ASEAN over 2001–2020 and may not generalize to other dyads or periods with different geopolitical contexts.