The COVID-19 Pandemics and Import Demand Elasticities: Evidence from China's Customs Data

Understanding how international trade reacts to exchange rate movements is a central question in international economics, and the COVID-19 period offers a unique context as policymakers attempted to stabilize trade flows. Despite notable policy interventions (e.g., FX market operations, liquidity facilities), the policy effect on exchange rate elasticity of trade during COVID-19 remains under-explored. This study fills the gap by examining how pandemic-induced policy responses in Asian economies affected the exchange rate elasticity of China's imports. Using a Python crawler, the authors assemble first-hand monthly Chinese provincial customs data (HS 8-digit) covering the COVID-19 outbreak period. China is the largest Asian importer and a central node in regional GVCs, with over 50% of imports sourced from Asia. The paper focuses on Asian partners’ exports to China to identify the impact of policy responses and foreign competition on trade during the pandemic. The main findings reveal that aggressive fiscal/monetary support in exporting countries reduced or even reversed the sign of bilateral exchange rate elasticity of China’s imports; competitors’ currency depreciation deterred exports to China, amplified by lower market concentration; strong support policies mitigated exposure to competitors’ exchange rate movements; and a product quality channel helps explain reduced exchange rate sensitivity. Results are robust across trade modes, policy measures, and alternative data, with additional analyses on regional, product heterogeneity, and extensive margin. The study contributes to COVID-19 trade literature emphasizing third-country effects and to work on competitors’ exchange rate movements, providing novel evidence with highly disaggregated monthly data and highlighting foreign competition and policy mechanisms relevant for exchange-rate and balance-of-payments policy design. The remainder of the paper presents data, a theoretical model, empirical strategy, results, and concluding remarks.

Two strands of literature are most relevant. (1) COVID-19 and trade: Prior studies document severe disruptions to global supply chains and foreign demand (e.g., Che et al. 2020; Vidya and Prabheesh 2020; Hayakawa and Mukunoki 2021; Zhao et al. 2021). Liu et al. (2021) highlight third-country effects using China’s 2020 exports, arguing for product and country heterogeneity. This paper advances the literature by using newly collected monthly provincial HS8 data to capture rapid pandemic dynamics and provincial heterogeneity in China’s imports. (2) Competitors’ exchange rates and trade elasticity: Feenstra et al. (2002) show RMB devaluation’s adverse impact on Korea’s exports; Cheung et al. (2016) find ASEAN–China competition in the US market; Mattoo et al. (2017) show China competition intensifies exchange-rate effects; Pennings (2017) emphasizes third-country exchange rates for US import and producer prices. The present study differs by analyzing COVID-19-period exchange rate elasticities with first-hand disaggregated data, integrating foreign competition and pandemic-induced policy responses, and offering policy implications on price/quantity adjustments to bilateral and third-country exchange rate movements.

Data: Monthly Chinese provincial customs data at HS 2017 8-digit level, January 2019–March 2021, including import value, quantity, and CIF unit prices in RMB. The dataset enables decomposition into price vs quantity and rich heterogeneity across provinces, products, and trade modes (ordinary vs processing trade, with processing sub-modes: pure assembly (PA) and processing with imports (PI)). Policy measures: Oxford COVID-19 Government Response Tracker (OxCGRT) monthly averages—Economic Support Index (rescaled 0.01–1) for baseline, and Containment and Health Index for robustness. Pandemic severity: monthly new confirmed COVID-19 cases per million in exporting country (JHU CRC). Exchange rates: Bilateral real exchange rate RER_CNY (CNY vs partner, deflated by partner and Chinese provincial CPI; IMF monthly ER) and competitors’ trade-weighted real exchange rate RER_Comp at exporter–product–importer level (trade-weighted average of bilateral RERs of all countries exporting product k to province j at time t). Market concentration: Herfindahl-Hirschman Index HHI at province–product–month level computed from global suppliers’ import shares for that province–product. Additional sources: ADB MRIO and UIBE GVC index (Wang et al. 2017) for GVC position; UN Comtrade (2019–2022) for extended robustness; World Bank income groups for exporter heterogeneity. Theoretical model: Extends Melitz–Ottaviano (2008) and Dhingra (2013) to include product quality differentiation and global competition, with government support captured by parameter λ (lower λ indicates stronger support). Key propositions: (1) Policy Effect—stronger support reduces responsiveness of export quantity to exchange rate movements (lower elasticity). (2) Foreign Competition Effect—competitors’ currency depreciation reduces a country’s exports (via lower competitors’ prices and more competitors, N_k^c). Empirical specification: Panel regression at exporter i–province j–HS8 product k–month t: Δln IM_{ijkt} = β0 + β1 Δln RER_CNY_{t−1} + β2 Δln RER_CNY_{t−1}×Policy_{t−1} + β3 Δln RER_Comp_{ijkt−1} + β4 Δln RER_Comp_{ijkt−1}×Policy_{t−1} + β5 Δln COVID_{it−1} + controls + FE + ε. Dependent variable IM alternates among total import value (v), unit value (uv), and quantity (q). Exchange-rate variables and HHI are lagged one month to mitigate endogeneity and account for adjustment frictions; policy not lagged in baseline but checked in robustness. Fixed effects: province-time FE, province–product–country FE, and month FE to control for unobserved heterogeneity, trends, and seasonality. Standard errors clustered at provincial level. Mechanism/Extensions: Product quality inferred following Khandelwal et al. (2013) using residuals from volume–price regressions with product and destination–month FEs; triple interactions to test quality channel. Nonlinearities assessed by adding squared Δln RER terms (Priestley & Ødegaard 2007). Heterogeneity by trade mode (ordinary vs processing; PI vs PA), exporter income group, province type (coastal vs interior), product technology class (Lall 2000), and GVC position (top vs bottom quartiles). Extensive margin (product churning) analyzed via Probit/Logit on probability of adding a product–destination link.

• Baseline bilateral exchange-rate effect and policy moderation:
  • For all goods, a 10% real RMB depreciation increases average provincial imports by about 17.58% (β1≈1.758; Table 2 col.1), and quantity response is similar (β≈1.854; col.3). This counter-intuitive sign (depreciation raising imports) echoes prior China-specific findings.
  • Policy interaction is strongly negative: Δln RER_CNY × Policy coefficients ≈ −3.296 (value) and −3.399 (quantity) for all goods (Table 2 cols.1,3), and ≈ −3.878 (value) and −4.100 (quantity) for intermediate inputs (cols.4,6). With average policy (index ≈0.52), marginal effects shrink toward zero; with above-average support, the elasticity turns intuitive (RMB appreciation raises imports), as shown in Fig. 8.
  • Adjustments operate mainly through quantities, not unit prices (policy interactions generally insignificant for uv).
• Third-country (foreign competition) exchange-rate effect:
  • Competitors’ appreciation raises a country’s exports to China (positive β3). For all goods, Δln RER_Comp ≈ 0.115 (value), 0.100 (quantity) and small but positive for unit value (Table 2 cols.1–3). Similar magnitudes for inputs (cols.4–6).
  • Policy dampens competition effect: Δln RER_Comp × Policy is negative and significant (e.g., −0.025 for value; Table 2), indicating stronger support makes exporters less vulnerable to competitors’ depreciations (Fig. 9).
• Market concentration (HHI) channels:
  • Higher HHI lowers sensitivity to competitors’ exchange rates: Δln RER_Comp × HHI negative (e.g., −0.087 for value; Table 2), implying tougher competition (lower HHI) magnifies third-country effects. Authors report a one–SD increase in HHI reduces responsiveness by about 21.94% (Column 1 discussion).
  • HHI mildly raises bilateral exchange-rate effects in some specifications (positive Δln RER_CNY × HHI in Table 2 value/quantity columns), with mixed significance.
• Pandemic severity: More new cases in exporting countries reduce import value/quantity and slightly increase unit prices (e.g., value β≈−0.015; uv β≈+0.0028; Table 2).
• Trade mode heterogeneity:
  • Processing trade is more sensitive to policy-moderated bilateral ER: for processing, Δln RER_CNY ≈ 2.979–3.203 and Δln RER_CNY × Policy ≈ −4.816 to −5.016 (Table 3 cols.4–6), larger than ordinary trade (Table 3 cols.1–3). Ordinary trade also shows significant but smaller magnitudes (e.g., −2.754 for value interaction).
  • Within processing: PI mode shows strong quantity effects (Δln RER_CNY ≈ 3.210; interaction ≈ −5.035; Table 4), while PA shows significant quantity effects and nuanced HHI interactions; unit prices generally less responsive.
• Product quality channel:
  • Using quality as dependent variable, Δln RER_CNY × Policy is negative and significant across samples (e.g., −1.586 full sample; Table 5), suggesting stronger policies are associated with rising quality (and lower ER elasticity).
  • Triple interaction Δln RER_CNY × Policy × Quality is significantly negative for value and quantity (Table 6), confirming quality as a mechanism reducing bilateral ER elasticity.
• Regional heterogeneity:
  • Exporter income group: Policy effect stronger for high-income exporters (e.g., Δln RER_CNY × Policy ≈ −1.934 significant vs. insignificant for lower-income in value; Table 8). Competition effects operate mainly via quantities in both groups.
  • Importer provinces: Coastal provinces exhibit larger (in absolute value) policy effect on bilateral ER elasticity and significant quantity adjustments; coastal imports are less sensitive to competitors’ ER than interior provinces (Table 9).
• GVC position heterogeneity:
  • Policy effect larger when importing from downstream industries (Δln RER_CNY × Policy ≈ −5.290; Table 10), while foreign competition effect is larger for upstream exporters (higher Δln RER_Comp coefficients).
• Product technology heterogeneity (Lall 2000): Policy dampening of bilateral ER elasticity is stronger for medium- and high-tech, and resources; competition effects and their policy moderation are strongest for high-tech (Table 11).
• Alternative policy measures and specification:
  • Quartile policy categorization preserves negative interaction effects across quartiles (Table 12).
  • Using OxCGRT containment and health index yields similar negative policy interactions; economic support policies have larger effects than containment policies (Tables 13–14).
• Nonlinearities: Squared ER terms are positive and significant (Table 15), implying convex relationships—large depreciations magnify baseline effects; sufficiently large appreciations can flip effects to intuitive import increases.
• Robustness checks:
  • UN Comtrade (2019–2022) confirms baseline signs/magnitudes (Table 16).
  • Dominant currency (USD, JPY) effects are insignificant when included with RMB and competitors’ ERs (Table 17).
  • Lagged ER effects on unit prices remain insignificant up to three months (Table 18).
• Extensive margin:
  • RMB depreciation increases probability of adding new imported products in processing trade; stronger policies reduce sensitivity (large negative interactions). Competition ER effects matter more for ordinary trade extensive margin; processing extensive margin less affected by policy in the competition channel (Table 19).

The study addresses how pandemic-induced policies and foreign competition shaped the exchange-rate elasticity of China’s imports during COVID-19. The central finding is that stronger fiscal/monetary support in exporting Asian economies substantially reduced exporters’ sensitivity to bilateral exchange-rate movements and to third-country depreciations, largely through quantity adjustments rather than prices. This aligns with the theoretical model in which policies accelerate recovery, stabilize supply chains, and facilitate quality upgrading, thereby mitigating exchange-rate-driven relative price effects. The prominent foreign competition (third-country) effect underscores that exporters’ outcomes depend not only on their bilateral exchange rate with China but also on competitors’ currencies. Market structure matters: less concentrated competitive environments amplify third-country exchange-rate effects, whereas higher concentration dampens them. Heterogeneity by trade mode (processing vs ordinary), product technology, GVC position, exporter income, and province type demonstrates that policy and competition channels are context-dependent, especially pronounced for intermediate inputs, processing trade (PI), high-tech goods, and downstream industries. Collectively, the results suggest that during systemic shocks, domestic support policies can serve as a buffer, enabling firms to maintain export volumes to China despite adverse currency movements by themselves or competitors, thereby stabilizing regional GVCs.

Using first-hand, highly disaggregated monthly Chinese provincial HS8 import data spanning the COVID-19 outbreak, the paper shows that: (1) bilateral and third-country exchange rates significantly affect China’s imports, predominantly via quantities; (2) pandemic-era economic support policies in exporting countries reduce the bilateral exchange-rate elasticity and dampen exposure to competitors’ depreciations, sometimes reversing counter-intuitive baseline elasticities; (3) market concentration, trade modes, product technology, GVC position, and regional characteristics systematically shape these effects; and (4) a product-quality channel helps explain attenuated sensitivity to exchange rates under stronger policies. The findings are robust to alternative policy measures, nonlinearities, extended data (UN Comtrade), and additional controls (dominant currency, lags). Policy implications include considering competitors’ currencies and recovery speeds when designing trade and macro policies, and recognizing that economic support measures are more effective than containment policies in stabilizing export capacity. Future research could extend to services trade, exploit newer customs data as released, and utilize COVID-period input–output tables to deepen analysis of GVC integration and dynamic policy effects.

• Data coverage: The provincial customs dataset spans January 2019–March 2021; extended analyses rely on UN Comtrade for 2019–2022, but provincial granularity beyond March 2021 is unavailable.
• Data access: The first-hand customs dataset was collected via a custom Python crawler and is available from authors upon reasonable request; full public replication may be constrained.
• Identification/endogeneity: Although exchange-rate and HHI variables are lagged and rich fixed effects are used, residual endogeneity may remain (e.g., exchange rates influenced by COVID-19 and policy responses). The assumption of immediate policy effects is tested but may not capture all lags.
• Price pass-through: Unit price responses are generally insignificant at monthly frequency; longer horizons or alternative pricing microdata could reveal delayed or invoicing-currency-specific pass-through.
• Extensive margin: Without firm-level entry/exit data, extensive margin results capture product–destination churning at the province level rather than firm dynamics.
• Model simplifications: The theoretical framework abstracts from some real-world complexities (e.g., contract rigidities beyond one-month lags, heterogeneous invoicing currencies).