ECONOMETRIC ASSESSMENT OF TRANSPORT INFRASTRUCTURE INVESTMENT AND ITS INFLUENCE ON BILATERAL TRADE FLOWS BETWEEN CHINA AND KAZAKHSTAN

ABSTRACT

Background. This article estimates the marginal contribution of transport-corridor capacity to bilateral trade between China and Kazakhstan, addressing the observed divergence between accelerating infrastructure expenditure and uneven trade growth. Methods. An augmented gravity model is estimated on an annual directed-dyad panel for 2010–2025 (N = 32) by Poisson Pseudo-Maximum Likelihood (PPML) with heteroskedasticity-robust standard errors, separate export and import specifications, lag structures, and robustness checks against mirror-statistics measurement error. Results. Rail throughput capacity is a positive and statistically significant determinant of Kazakhstan’s imports from China (elasticity 0.82, p = 0.005) but not of its exports (−0.26, p = 0.57). The export–import asymmetry is significant (difference 1.08, z = 1.97, p = 0.049), confirming that exports are price- and commodity-driven (real-exchange-rate elasticity 1.05, p = 0.010) while imports are capacity-constrained. The import result is robust to source substitution (Chinese mirror statistics) and shows no residual autocorrelation (Durbin–Watson 1.97). Conclusion. Corridor capacity raises trade through the import channel, with a contemporaneous rather than lagged effect; the asymmetry implies that infrastructure investment expands transit-borne imports more than export earnings, qualifying the value-capture case for Kazakhstan.

Keywords: gravity model of trade; transport infrastructure investment; bilateral trade flows; China–Kazakhstan trade; PPML estimation; panel data; trade elasticity; rail throughput capacity.

Introduction

Over the past decade, Kazakhstan has undertaken a substantial programme of transport infrastructure investment oriented toward its trade corridor with China, including the double-tracking of the Dostyk–Moiynty line, the construction of an Almaty rail bypass, and the development of dry ports and consolidation centres [1; 2]. Over the same period the daily throughput capacity of the corridor rose from roughly 8 to 50 train pairs per day, while total bilateral trade expanded from about USD 29 billion in 2010 to USD 112 billion in 2025. Yet the growth has been uneven and, in the first half of 2025, partially decoupled from the continued rise in throughput [3]. This raises a question that descriptive analysis alone cannot resolve: what is the measurable marginal contribution of infrastructure capacity to bilateral trade once income and price effects are accounted for?

Answering this question requires a formal econometric framework rather than a narrative one. The descriptive record conflates several distinct drivers — the income of trading partners, commodity-price cycles, exchange-rate movements, and the physical capacity of the corridor — and risks attributing to infrastructure trade growth that in fact reflects a price-driven rebound [3]. The present article specifies and estimates an augmented gravity model of trade designed to isolate the infrastructure effect, and reports the resulting elasticities, hypothesis tests, and robustness diagnostics.

The objective of the study is twofold: first, to formulate an econometrically defensible specification linking transport infrastructure investment to bilateral trade flows; and second, to estimate that specification and quantify the infrastructure elasticity, its export–import asymmetry, and its lag structure. The contribution lies in adapting the standard gravity framework to the bilateral China–Kazakhstan case through an explicit corridor-capacity regressor, and in providing the first directed-dyad empirical estimates of the capacity–trade relationship for this corridor.

Theoretical Framework

The gravity model of international trade, in its modern theory-consistent form, derives bilateral trade flows from the economic mass of trading partners and the trade frictions between them [4]. Anderson and van Wincoop showed that a correctly specified gravity equation must account for multilateral resistance — the relative, rather than absolute, trade costs faced by each country — now standard practice through exporter- and importer-time fixed effects [4]. Transport infrastructure enters this framework as a determinant of bilateral trade cost: improved corridor capacity reduces the effective distance between partners, and the World Bank’s corridor-economics research links such reductions to aggregate income gains of 2–3 per cent for recipient economies [5].

Within this framework, the working hypotheses are as follows. H1: bilateral trade flows are positively and significantly associated with rail throughput capacity, after controlling for income and price effects. H2: the elasticity of Kazakhstan’s exports to China with respect to infrastructure capacity differs from that of its imports, reflecting the commodity composition of trade. H3: the infrastructure effect operates with a lag, as capacity expansion translates into trade flows only after complementary logistics services develop.

Materials and Methods

The empirical strategy employs an augmented gravity model estimated on a directed-dyad panel. The baseline multiplicative specification is written as follows:

X_ijt = exp[ β₀ + β₁ lnGDP_it + β₂ lnGDP_jt + β₃ lnINFRA_t + β₄ lnRER_t + β₅ lnCPI_t + δD_exp ] · η_ijt     (1)

where X_ijt is the trade flow in direction ij in year t; GDP is real gross domestic product, capturing economic mass; INFRA is rail throughput capacity in train pairs per day; RER is the real bilateral exchange-rate index; CPI is a composite oil-and-metals commodity-price index; and D_exp is an export-direction dummy.

Estimation. The model is estimated by Poisson Pseudo-Maximum Likelihood (PPML) rather than by log-linearised OLS. PPML is preferred because it is consistent under heteroskedasticity — under which the log-linear form is biased through Jensen’s inequality — and accommodates zero flows directly [6]. Standard errors are heteroskedasticity-robust (HC1). The pooled model (M1) is estimated on all 32 directed-dyad observations; separate export (M2) and import (M3) models test the asymmetry hypothesis; and a lag-augmented model (M4) adds the one-period lag of capacity.

Identification and robustness. The principal threat to identification is the endogeneity of infrastructure investment: capacity may be expanded in anticipation of trade growth, biasing β₃ upward. This is addressed through (i) lagged-capacity specifications, (ii) the Ramsey RESET test for functional-form misspecification, (iii) a Durbin–Watson check for residual autocorrelation, and (iv) re-estimation on Chinese mirror statistics to gauge sensitivity to the documented 37 per cent inter-source discrepancy [3]. The variables and expected signs are summarised in Table 1.

Table 1.

Variables, measurement, and expected signs

Data. The panel is constructed at annual frequency over 2010–2025 for two directed flows (Kazakhstan’s exports to China and its imports from China), giving 32 observations. Trade values draw on the General Administration of Customs of the PRC and the Association of Financiers of Kazakhstan; GDP and exchange-rate series on the World Bank and the National Bank of Kazakhstan; and capacity measures on reports of the national railway operator [1; 3; 5]. Descriptive statistics are reported in Table 2.

Table 2.

Descriptive statistics, 2010–2025 (N = 32 directed-dyad observations)

Results

Baseline estimates. Table 3 reports the PPML coefficients with HC1 standard errors for the pooled model (M1) and the separate export (M2) and import (M3) specifications. All models fit the data closely, with pseudo-R² of 0.94 (M1), 0.95 (M2) and 0.98 (M3); the correlation between fitted and observed flows in the pooled model is 0.98. The export-direction dummy is large and significant (−0.63, p < 0.001), confirming that import flows systematically exceed exports at given income and capacity.

The central quantity of interest is β₃, the elasticity of trade with respect to capacity. In the pooled model it is positive but imprecisely estimated (0.46, p = 0.32), a consequence of pooling two directions whose responses differ in sign and magnitude. The directional models resolve this. For imports (M3), capacity is positive and highly significant: a coefficient of 0.82 (p = 0.005) implies that a 10 per cent rise in throughput capacity is associated with an 8.2 per cent increase in imports from China, holding income and prices constant. For exports (M2), by contrast, the capacity coefficient is small, negative, and insignificant (−0.26, p = 0.57); exports instead respond to the real exchange rate (1.05, p = 0.010) and to commodity prices (0.58, p = 0.018). Hypothesis H1 is therefore supported for the import channel but not the export channel.

Table 3.

PPML gravity estimates, dependent variable bilateral trade flow (USD bn)

Note: HC1 robust standard errors in parentheses. † p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001. Estimator: Poisson Pseudo-Maximum Likelihood.

Export–import asymmetry. Hypothesis H2 is tested directly by comparing the capacity elasticities of M2 and M3. The difference is 1.08 in favour of imports (z = 1.97, p = 0.049), statistically significant at the 5 per cent level. The asymmetry is economically intuitive: Kazakhstan’s exports to China are dominated by bulk commodities whose volume is constrained by price and demand rather than by corridor throughput, whereas imports of manufactured and consumer goods are containerised and capacity-sensitive. H2 is supported.

Lag structure. Hypothesis H3 is tested in the lag-augmented model (M4), which enters both contemporaneous and once-lagged capacity. The contemporaneous term remains positive (0.43) while the lagged term is small, negative, and insignificant (−0.20, p = 0.50); the two are highly collinear (the capacity series is near-monotone), so the lag adds no identified explanatory power. The evidence indicates a contemporaneous rather than a delayed effect, and H3 is not supported in this sample. The result is consistent with capacity additions on this corridor being absorbed quickly by pre-existing demand rather than waiting on complementary logistics services.

Robustness and diagnostics. The import result is robust along three dimensions. First, re-estimating M3 on Chinese mirror statistics (scaling the dependent variable by the documented 37 per cent inter-source gap) leaves the capacity elasticity essentially unchanged at 0.82 (p = 0.005), so the finding does not depend on the choice of statistical source. Second, the Ramsey RESET test on M3 returns p = 0.059, providing no evidence of functional-form misspecification at the 5 per cent level. Third, the Durbin–Watson statistic on the import residuals is 1.97, indicating no residual autocorrelation. Table 4 collects the diagnostic and hypothesis-test results.

Table 4.

Hypothesis tests and diagnostics

Discussion

The estimates should be read through the distinction between statistical and economic significance. The import elasticity of 0.82 is both: statistically robust and large enough to matter for policy, implying that the roughly six-fold expansion of corridor capacity over 2010–2025 can account for a substantial share of the parallel growth in Chinese imports. Crucially, however, the same capacity expansion has no measurable effect on Kazakhstan’s exports, which remain governed by commodity prices and the real exchange rate. The corridor therefore behaves asymmetrically: it lowers the cost of bringing Chinese manufactured goods into Kazakhstan more than it lowers the cost of moving Kazakhstani commodities outward.

This asymmetry has a direct bearing on the value-capture question. A capacity-driven rise in imports expands transit and consumption but does not, by itself, generate export earnings or local value added; whether corridor investment yields net economic benefit for Kazakhstan thus depends on institutional arrangements — return-cargo logistics, processing-for-export, and customs facilitation — that the gravity model does not capture. The econometric result quantifies the trade response; the distribution of the resulting value remains an institutional matter.

Three limitations require emphasis. First, the endogeneity of infrastructure investment cannot be fully eliminated; the lag-based and mirror-statistics checks mitigate but do not resolve it, and the import elasticity should be read as an upper bound where reverse causality is plausible. Second, the annual sample of 32 directed-dyad observations is small, which limits the precision of the income coefficients in particular and prevents the inclusion of high-dimensional fixed effects; the directional split, rather than full multilateral-resistance fixed effects, is the identification device available at this frequency. Third, the documented divergence between Chinese and Kazakhstani trade statistics introduces measurement error in the dependent variable, addressed here through source substitution but not eliminated. These constraints point toward extension with monthly or commodity-disaggregated data in future work.

Conclusion

This article specified and estimated an augmented gravity model of the influence of transport-corridor capacity on bilateral trade between China and Kazakhstan over 2010–2025. Using PPML on a directed-dyad panel, it finds that rail throughput capacity is a positive and significant determinant of Kazakhstan’s imports from China, with an elasticity of 0.82, but has no significant effect on its exports, which are price- and commodity-driven. The export–import asymmetry is statistically significant (p = 0.049), the effect is contemporaneous rather than lagged, and the import result is robust to mirror-statistics measurement error, functional-form testing, and residual-autocorrelation checks.

The principal methodological contribution is the introduction of an explicit corridor-capacity regressor into a theory-consistent gravity equation estimated separately by trade direction, which isolates the marginal contribution of infrastructure from income and price effects. The principal substantive finding — that capacity expansion raises imports far more than exports — qualifies the policy case for corridor investment: converting transit capacity into durable trade gains for Kazakhstan will depend less on further capacity additions than on the institutional arrangements that determine whether the corridor carries Kazakhstani value outward as well as Chinese goods inward.

References:

1. Kazakhstan launches crucial Dostyk-Moyynty double track railway [Electronic resource] // RailFreight.com. — 2025. — URL: https://www.railfreight.com (date of access: 05.06.2026).
2. Kazakhstan to enhance connectivity via improved transport infrastructure this year [Electronic resource] // The Astana Times. — 2025. — URL: https://astanatimes.com (date of access: 05.06.2026).
3. Review of foreign trade of the Republic of Kazakhstan for the first half of 2025 [Electronic resource] // Association of Financiers of Kazakhstan. — 2025. — URL: https://afk.kz (date of access: 05.06.2026).
4. Anderson J.E., van Wincoop E. Gravity with Gravitas: A Solution to the Border Puzzle // American Economic Review. — 2003. — Vol. 93, No. 1. — P. 170–192.
5. Belt and Road Economics: Opportunities and Risks of Transport Corridors. — Washington: World Bank, 2019. — 139 p.
6. Santos Silva J.M.C., Tenreyro S. The Log of Gravity // The Review of Economics and Statistics. — 2006. — Vol. 88, No. 4. — P. 641–658.