Export duration and firm markups: evidence from China

The study investigates how export duration—measured as the length of time a firm continuously participates in export markets—affects firm markups, a key indicator of pricing power and market power. Motivated by the observed "low markup trap" among Chinese exporters (exporters often having lower markups than non-exporters), the paper moves beyond the binary export/non-export comparison to examine whether accumulating export experience (longer export duration) influences markups. Within the context of China's deepening openness and the need to enhance international competitiveness and value-chain gains, the research poses two main hypotheses: (1) the net effect of longer export duration on firm markups is theoretically ambiguous because duration can raise productivity (boosting markups) but may also lower distribution costs and induce market-based pricing that reduces markups; (2) export duration operates through two channels—"production efficiency" (learning-by-exporting raises productivity and lowers marginal costs, increasing markups) and "market-based pricing" (reduced distribution costs and intensified market familiarity lower ex-factory prices, decreasing markups). The study aims to provide a comprehensive view of how export market engagement over time shapes firms’ markups using Chinese manufacturing microdata.

Two strands of literature are reviewed. (1) Export duration: Early work measured export spell length and survival (e.g., Besedeš and Prusa, 2006; Volpe Martincus and Carballo, 2009; Besedeš and Nair-Reichert, 2009; Esteve-Pérez et al., 2013; Socrates et al., 2020). More recent studies examine determinants of export duration, including information externalities (Valderrama et al., 2013), tariff changes post-WTO accession (Zhou et al., 2019), FDI linkages (Kostevc and Zajc Kejžar, 2020), innovation (Chen, 2012), servitization (Cui and Liu, 2018), and export tax rebates (Anwar et al., 2019). Few investigate how export duration affects firm performance outcomes. (2) Exports and markups: Theory links exporting, productivity, and higher markups for more productive firms (Bernard et al., 2003; Melitz and Ottaviano, 2008). Empirically, many find exporters enjoy higher markups (Görg and Warzynski, 2003; De Loecker and Warzynski, 2012; Kato, 2014; Bellone et al., 2016), though some counter-evidence exists (Hornok and Muraközy, 2019). China-focused work documents a "low markup trap" (Sheng and Wang, 2012; Zhang and Zhu, 2017) and examines mechanisms via innovation and competition (Máñez et al., 2022; Garcia-Marin and Voigtländer, 2019; Caselli and Schiavo, 2020), export patterns and destinations (Feenstra and Hanson, 2004; Manova and Zhang, 2009; Kilinç, 2019; Yang, 2021). Research specifically on export duration's effect on markups is scarce. This study contributes by centering on export duration among exporters, building a theoretical model with two channels, and providing firm-level evidence from China with heterogeneity analyses.

The study combines a theoretical model with empirical analysis.

• Theoretical framework: Two symmetric economies trade; firms produce differentiated varieties only for export. Representative consumer utility follows a CES aggregator with elasticity of substitution σ>1. Each firm produces q=φ(x)K^{1−α}L^{α}, where φ(x) is firm productivity increasing in export duration x (learning-by-exporting). Marginal cost mc(x)=α^{−α}(1−α)^{−(1−α)}R^{1−α}W^{α}/φ(x). Exporting faces iceberg trade cost γ and distribution costs η(x), which decline with export duration (∂η/∂x<0). The consumer price is P=γP0+η(x). Profit maximization yields optimal ex-factory price P0(x) and a markup defined as markup(x)=P0/mc. Markup depends on φ(x) and η(x), implying ambiguous net effects of longer duration: a positive "production efficiency" channel via φ(x) and a negative "market-based pricing" channel via η(x).
• Empirical framework: Baseline panel model at firm-year level: log(markup){it}=β1 Export_T{it}+β2 X_{it}+λ_i+ν_t+μ_{it}, where Export_T is export duration, X includes controls (Size=log sales; KI=log capital-labor ratio; Wage=log average wage; Age=log firm age; Subsidy=subsidies-to-sales). Firm and year fixed effects included.
• Markup measurement: Using De Loecker and Warzynski (2012), markup μ_it=θ_i (α_it)^{ε_it}, where α_it is expenditure share of variable input v and ε_it is output elasticity from a translog production function. Output elasticity is estimated via ACF semi-parametric method (Ackerberg, Caves, Frazer, 2015) by 2-digit industry; intermediate inputs are treated as the variable input; labor is not treated as variable; capital is dynamic. A robustness alternative "accounting" markup measure is also used.
• Production function (translog) estimated: y_it=β1 l_it+β2 k_it+β3 m_it+β4 l_it^2+β5 k_it^2+β6 m_it^2+β7 l_it k_it+β8 l_it m_it+β9 k_it m_it+ω_it+ε_it, where y=log output; l, m, k are logs of labor, intermediate inputs, and capital; ω_it is TFP (also used in mechanism analysis).
• Export duration construction: Duration is measured from a firm’s first export year to exit; only the first spell during 2000–2007 is used to mitigate "multiple spells" issues; left-truncated observations (exporting before 2000) are removed.
• Data: Chinese Annual Survey of Industrial Firms (CASIF) manufacturing sample 2000–2007; matched to China Customs transactions (2000–2007) using firm names, zip codes, and phone numbers. Data cleaned following Feenstra et al. (2014), Yu (2015), Xiang et al. (2017). Markups estimated by 2-digit industry. Descriptive statistics and industry markup distributions are reported.
• Robustness checks: (i) alternative markup measure (accounting approach); (ii) lagged export duration; (iii) addressing endogeneity via system GMM and industry-level mean Export_T as IV (à la Fisman and Svensson, 2007); (iv) exclude pre-WTO period (use 2002+ subsample); (v) balanced panel (firms present in all years).
• Mechanism tests: Two regressions assess channels: (a) TFP_it on Export_T_it (OLS and IV) to test "production efficiency"; (b) Markup_it on Export_T_it and TFP_it (OLS and IV) to net out cost-side variation and infer "market-based pricing."

• Baseline effect: Longer export duration significantly increases firm markups. In the preferred specification with firm and year fixed effects and controls, Export_T coefficient ≈ 0.0036 (SE 0.0012), significant at 1%, implying roughly a 0.36% increase in markups per additional year of export duration (Table 3, col. 4). Size and Wage raise markups; KI lowers markups; Age and Subsidy are not significant.
• Robustness:
  • Alternative markup measure (accounting): Export_T = 0.0021** (Table 4, col. 1).
  • Lagged duration: L.Export_T = 0.0051* (Table 4, col. 2).
  • System GMM: Export_T = 0.3089** with AR(1) p=0.000, AR(2) p=0.307, Hansen p=0.268 (Table 4, col. 3).
  • IV (industry mean Export_T as IV): Export_T = 0.0288***; underidentification and weak IV tests passed (Table 4, col. 4).
  • Post-WTO subsample (2002+): Export_T = 0.0039*** (Table 4, col. 5).
  • Balanced panel: Export_T = 0.0035* (Table 4, col. 6).
• Mechanisms (Table 5):
  • Production efficiency channel: Export_T → higher TFP (OLS: 0.0071***; IV: 0.0455***), consistent with learning-by-exporting lowering marginal costs and raising markups.
  • Market-based pricing channel: Controlling for TFP, Export_T negatively relates to markup (OLS: −0.0037***; IV: −0.0180***), indicating reduced product prices/market pricing power with longer duration.
  • TFP strongly and positively loads on markups (~1.026***).
• Heterogeneity:
  • By industry type (Table 6): Capital-technology-intensive firms benefit (0.0028**); labor-intensive firms show no significant effect.
  • By firm size (Table 6): Positive for both, larger for large-scale firms (small: 0.0038**; large: 0.0047***).
  • By ownership (Table 7): SOEs negative (−0.0102*), private not significant, foreign-funded positive (0.0052**).
  • By region (Table 7): Coastal firms positive (0.0035***); non-coastal not significant.
  • By export intensity (Table 8): High-intensity exporters positive (0.0085***); low-intensity not significant.
  • By trade mode (Table 8): General trade positive (0.0042***); processing trade not significant.
• Overall: The positive production-efficiency effect dominates the negative market-based pricing effect, yielding a net positive impact of export duration on markups.

The results directly address the research question by showing that while export duration simultaneously increases productivity (raising markups) and reduces market pricing power (lowering markups), the former quantitatively dominates in Chinese manufacturing. This helps explain pathways out of the observed "low markup trap" among Chinese exporters: sustained participation in export markets enables learning-by-exporting and efficiency gains that, on net, elevate markups. The findings also clarify why effects differ across firm and market contexts: firms with greater scope for productivity upgrading (capital-technology-intensive, large, foreign-funded, coastal, high export intensity, general trade) realize stronger positive net effects, while settings with limited pricing power or structural constraints (SOEs, processing trade, non-coastal, low intensity) see muted or negative effects. The robustness checks confirm that results are not artifacts of measurement approach, timing, endogeneity, WTO-related policy changes, or panel imbalance, reinforcing the credibility and relevance of the findings for trade and industrial policy.

The study develops a theoretical framework and provides firm-level evidence from China showing that longer export duration has an uncertain theoretical effect on markups because it raises productivity (production efficiency channel) but reduces market-based pricing power. Empirically, the production efficiency gains dominate, so longer export duration significantly increases markups of Chinese manufacturing firms. This result is robust across alternative markup measures, lag structures, endogeneity corrections (system GMM, IV), subsample excluding pre-WTO, and balanced panels. The paper documents pronounced heterogeneity: positive effects for capital-technology-intensive, large, foreign-funded, coastal, high export intensity, and general-trade firms; insignificant or negative effects for labor-intensive, small (smaller magnitude), SOEs (negative), non-coastal, low export intensity, and processing-trade firms. Policy implications emphasize encouraging exporters to deepen and extend export relationships to harness learning-by-exporting and productivity gains that can lift markups and international competitiveness, while tailoring support recognizing heterogeneous impacts across firm types and regions.

• Prices and marginal costs are not directly observed in CASIF; markups and mechanisms are inferred via production-function-based measures (ACF) and indirect identification (controlling for TFP), which may introduce measurement error.
• Export duration is limited to the 2000–2007 window; left-truncated spells (pre-2000 exporters) are removed, potentially affecting representativeness of long-duration exporters.
• The analysis focuses on Chinese manufacturing firms; external validity to services or other countries may be limited.
• Trade mode classification treats firms engaging in both processing and general trade as general trade (per note), which may blur differences.
• Potential remaining endogeneity or unobserved shocks cannot be entirely ruled out despite system GMM and IV strategies.