Abstract
Making use of considerably improved measures of infrastructure, the study assesses the impact of infrastructure on bilateral trade for a panel of 150 developed and emerging economies during the period 1992–2011. The authors make use of a gravity approach to disentangle the impact of infrastructure on trade and trade costs. Improving infrastructure endowments and quality decreases trade costs and increases international trade flows. Countries with improved infrastructure reduce not only bilateral trade costs but also multilateral trade costs. The decomposition of effects indicates that better infrastructure encourages higher export flows relative to domestic trade flows. Main results of the study prove to be robust, also when considering distinct trade categories (consumption goods, intermediates, and capital goods) for a smaller sample.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The employed trade variable is rarely differentiated in earlier studies. This is problematic since infrastructure is unlikely to have the same impact on trade in distinct product categories. As argued by Miroudot et al. (2009), distance-related trade costs are of varying importance for specific goods and services industries. Fink et al. (2005) distinguish between trade in homogeneous and differentiated products. The authors show that communication costs have a stronger impact on trade in differentiated products.
As proposed earlier by Canning (1998), Limão and Venables (2001) and Brun et al. (2005) use the density of roads, paved roads, and railways as well as the number of telephone lines per capita; see also Vijil and Wagner (2012). Bougheas et al. (1999) consider the length of the motorway network as a proxy for infrastructure (the stock of public capital is used as an alternative proxy).
For an overview on the existing macro-level empirical literature regarding the link between infrastructure and development see the survey by Straub (2011).
Time indices are dropped here to avoid clutter.
Note that \(P_{j}\) is as well the CES price index of the demand system.
Trade costs are assumed in a way that it is always (weakly) cheaper to ship directly from country \(i\) to country \(j\) than via a third country \(k\), such that \(t_{ij} \ge t_{ik} t_{kj}\).
As shown in Eq. (2), the outward multilateral resistance term is the weighted average aggregate of all bilateral trade costs the exporter faces. Respectively, Eq. (3) shows that the inward multilateral resistance term is the weighted average of all bilateral trade costs the consumer faces. See Anderson and Yotov (2010) or Larch and Yotov (2016) for details.
Note that the standard errors are clustered at the level of country pairs in the baseline estimations.
Exporter- and importer-specific covariates are absorbed in the estimation process by exporter-time and importer-time-fixed effects that are best suited to control for multilateral resistance in the structural gravity model. In such a model, the impact of any variable affecting the exporter’s or importer’s propensity to export and the respective flows to and from all destinations is hard to identify in the presence of importer- and exporter-fixed effects (Head and Mayer 2014). Eaton and Kortum (2002) also use a two-step procedure to examine how technology and geography determine patterns of bilateral trade and specialization with cross-sectional data.
The approximation of the multilateral resistance terms with a remoteness variable instead of exporter-time- and importer-time-fixed effects is another possible way to account for the multilateral resistance terms allowing for adding country-specific characteristics separately. However, this approach is not suited to disentangle the channels through which infrastructure potentially affects trade (Anderson et al. 2015; Head and Mayer 2014).
The equation for importer-time-fixed effects is equivalent.
The estimation is based on 10,000 bootstrap replications.
Time indices dropped to avoid clutter. To obtain Eq. (7), we follow Head and Ries (2010), Jacks et al. (2011), and Novy (2013). Basically, Eq. (1) is multiplied with bilateral trade flows in the opposite direction from \(j\) to \(i\) \((X_{ji} )\) and then divided by the product of the domestic trade flows of the importing and exporting country, \(X_{ii} X_{jj} ,\) where
\(X_{ii} = \frac{{Y_{i} E_{i} }}{{Y^{W} }} \left( {\frac{{t_{ii} }}{{\varPi_{i} P_{i} }}} \right)^{1 - \sigma } \quad {\text{and}}\quad X_{jj} = \frac{{Y_{j} E_{j} }}{{Y^{W} }} \left( {\frac{{t_{jj} }}{{\varPi_{j} P_{j} }}} \right)^{1 - \sigma } .\)
This eliminates both multilateral resistance terms from the gravity equation and yields
\(\frac{{{\text{X}}_{\text{ij}} X_{ji} }}{{X_{ii} X_{jj} }} = \left( {\frac{{t_{ij} t_{ji} }}{{t_{ii} t_{jj} }}} \right)^{1 - \sigma } .\)
Rearranging this equation solves for the relative trade cost parameter, i.e. the parameter of interest. To obtain an expression for the tariff equivalent, we follow Jacks et al. (2011) and Novy (2013) by taking the geometric mean of barriers to trade and subtracting one.
These regressions do not include additional fixed effects to control for multilateral resistance since the trade costs measure nets out the multilateral components.
To avoid perfect collinearity, the constant and one importer-fixed effect need to be dropped, implying that all other fixed effects are only identified relative to the dropped fixed effect, thus, relative to the U.S. The U.S. is an advanced economy well-endowed with infrastructure, not changing much over time. Thus, it can be regarded as an appropriate country for normalization.
In general, larger countries have a higher multilateral resistance compared to smaller countries.
To ensure consistency with the outlined structural gravity equation, where consumers choose among and consume domestic as well as foreign varieties, we include domestic trade in the gravity estimation. Further, the inclusion of internal trade leads to theoretically consistent identification of the effects of unilateral, non-discriminatory policies like infrastructure investments.
Our measures of international integration are taken from Mario Larch's RTA Database: http://www.ewf.uni-bayreuth.de/en/research/RTA-data/index.html (accessed: February 2018).
We thank Julian Hinz for the provision of the data.
Note that the choice of the distance measure does not influence our results (results with ‘standard’ non-time-varying distances are available upon request). Other typical non-time-varying gravity variables like contiguous borders, common language, and colonial ties are captured by the pair-fixed effects (\(\mu_{ij}\)), which control as well for exporter-importer pair heterogeneity.
A detailed description of the construction of the index is provided in “Appendix A1”.
However, as shown in Sect. 4, the correlation between the index of infrastructure and other dimensions of the economic and institutional development of our sample countries is far from perfect.
The red vertical line in Fig. 1 depicts the median of the index of infrastructure (31.9); the corresponding mean amounts to 37.6, with a standard deviation of 16.1 (not shown).
In additional calculations based on the overall country sample (not shown), we found an average autoregressive (AR(1)) coefficient of 0.56 with a standard deviation of 0.36. This finding is consistent with the country-specific evidence in Fig. 3 by suggesting a moderate degree of persistence of the index of infrastructure.
As noted by an anonymous reviewer, lagging the index of infrastructure would not offer any new variation and render lags meaningless if the time series were strongly persistent.
Time indices not shown to avoid clutter.
Note that all variables in the empirical models are in logs (except exports and our measures of international integration). In contrast, the infra variables of both trading partners are used in their original form for the calculation of GL_infra. “Appendix A2” presents summary statistics for all variables.
The only exception is the distance variable which enters with the expected negative sign but turns out to be statistically insignificant at conventional levels.
We are most grateful to the anonymous reviewers for having alerted us on issues of model identification.
See also the robustness tests below where we show that this result is driven mainly by the trade relations between pairs of advanced countries, which are both well-endowed with infrastructure. In unreported estimations, we additionally controlled for the similarity in the partner countries’ GDP per capita, by including the corresponding Grubel-Lloyd index. Importantly, the coefficient on GL_infra continued to be significantly positive, even though its size declined. In other words, the correlation between the two types of similarity does not invalidate our finding for GL_infra in column (3).
However, the coefficients on the exporter’s and the importer’s infrastructure are not significantly different from each other in column (3).
Data are taken from the World Development Indicators (WDI).
The size of the coefficients on infra even increases in column (4) compared to column (3).
See Eq. (7) for details on the construction of the calibrated trade cost measure.
We constrain the coefficients of the parameters used in the construction of the trade costs to their values used in the construction. Unconstrained estimations yield slightly lower coefficients for all regressors. The infrastructure coefficients for the exporter and importer range from 0.09 (model with additional bilateral variables, not controlling for tariffs and rol; see below) to 0.44 (model with pair-fixed effects; not controlling for tariffs and rol) in unconstrained estimation (not shown). These estimates are a lower bound of the impact of infrastructure on bilateral trade costs.
The impact of an infrastructure improvement by one standard deviation in the case of calibrated trade costs reduces trade costs by roughly two percent.
The data are taken from the standard gravity dataset provided by CEPII (http://www.cepii.fr/CEPII/en/bdd_modele/presentation.asp?id=8, accessed: January 2018). For a detailed discussion of the data see Head et al. (2010) and Head and Mayer (2014).
A reduction in bilateral trade costs, either via a reduction in domestic trade costs or a reduction in international trade costs, is reflected in a decrease in average trade costs and an increase in market access and supplier access of the respective country. As can be seen from Table 2, the weighted average over the characteristics of each exporter’s trade costs determinants contributes significantly towards an increase in trade volumes for the importer and exporter. This in turn is as well reflected in the reduction of multilateral trade costs (see Eqs. (9) and (10) for details on the construction of the multilateral trade costs). Hence, any change in bilateral trade costs also affects multilateral trade costs. In particular, infrastructure improvements that facilitate trade with many or major trading partners increase total trade volumes, market access, and supplier capacity of a country.
Some resource-rich countries like Azerbaijan are an exception. The global demand for their natural resources appears to have offset the negative impact of deficient infrastructure on market access and supplier capacity.
Note that the total effect of infrastructure for internal trade is slightly negative, but not significant. The impact on internal trade might be partially netted out by increases in the income level.
In contrast to Portugal-Perez and Wilson (2012), we consider this interaction for both the exporting and the importing country. Results of the estimation are shown in column (5) of Table 4. The coefficient on the interaction term is relatively small and insignificant. However, its inclusion considerably reduces the coefficients for the importer’s income and the infrastructure coefficients and renders them insignificant.
Note that the choice of the infrastructure index also affects the calculation of the GL index. Moreover, the sample size is considerably smaller in the case of the energy and finance sub-indices due to a lack of observations for some low-income countries (e.g. Afghanistan, Burkina Faso, or Lesotho).
Note, however, that though transport infrastructure does not seem to affect the trade volume it has a significant impact on trade cost reduction (results not shown).
We have no plausible instrument that fulfils the exclusion restriction. That is why we cannot solve the potential endogeneity problem with instrumental variables regressions. Hence, we pursue alternative approaches to mitigate concerns about reverse causality.
Trading partners that are prominently dropped include China, Germany, United States, France, and Great Britain, i.e. countries that also score high in terms of infrastructure development.
These parameters are estimated in order to minimize the error in the index.
The log-likelihood function which is maximized with respect to \(\alpha_{l}\), \(\beta_{l}\), and \(\sigma_{l}^{2}\) for each country c is given by \(\ln L_{cl} \left( {\alpha_{l} ,\beta_{l} ,\sigma_{l} } \right) = - \frac{1}{2}\ln \left( {2\pi } \right) - \frac{1}{2}\ln \left( {Var\left[ {y_{cl} } \right]} \right) - \frac{1}{2}\left( {\frac{{y_{cl} - \alpha_{l} }}{{Var\left[ {y_{cl} } \right]^{{\frac{1}{2}}} }}} \right)^{2}\).
References
Agnosteva, D. E., Anderson, J. E., & Yotov, Y. V. (2014). Intra-national trade costs: Measurement and aggregation. National Bureau of Economic Research Working Paper Series 19872.
Allen, T., & Arkolakis, C. (2014). Trade and the topography of the spatial economy. The Quarterly Journal of Economics, 129(3), 1085–1140.
Anderson, J. E., Larch, M., & Yotov, Y. V. (2015). Estimating general equilibrium trade policy effects: GE PPML. CESifo Working Paper Series No. 5592.
Anderson, J. E., & Van Wincoop, E. (2003). Gravity with gravitas: A solution to the border puzzle. American Economic Review, 93(1), 170–192.
Anderson, J. E., & van Wincoop, E. (2004). Trade costs. Journal of Economic Literature, 42(3), 691–751.
Anderson, J. E., & Yotov, Y. V. (2010). The changing incidence of geography. American Economic Review, 100(5), 2157–2186.
Bernhofen, D. M., El-Sahli, Z., & Kneller, R. (2016). Estimating the effects of the container revolution on world trade. Journal of International Economics, 98(1), 36–50.
Bougheas, S., Demetriades, P. O., & Morgenroth, E. L. W. (1999). Infrastructure, transport costs and trade. Journal of International Economics, 47(1), 169–189.
Brülhart, M. (2009). An account of global intra-industry trade, 1962–2006. World Economy, 32(3), 401–459.
Brun, J.-F., Carrère, C., Guillaumont, P., & de Melo, J. (2005). Has distance died? Evidence from a panel gravity model. World Bank Economic Review, 19(1), 99–120.
Canning, D. (1998). A database of world infrastructure stocks, 1950–1995. World Bank Economic Review, 12(3), 529–547.
Cheng, I.-H., & Wall, H. J. (2005). Controlling for heterogeneity in gravity models of trade and integration. Federal Reserve Bank of St. Louis Review, 87(1), 49–63.
Cristea, A. D. (2015). The effect of communication costs on trade in headquarter services. Review of World Economics, 151(2), 255–289.
Datta, S. (2012). The impact of improved highways on Indian firms. Journal of Development Economics, 99(1), 46–57.
Donaldson, D. (2018). Railroads of the raj: Estimating the impact of transportation infrastructure. American Economic Review, 108(4–5), 899–934.
Donaldson, D., & Hornbeck, R. (2016). Railroads and American economic growth: A “market access” approach. The Quarterly Journal of Economics, 131(2), 799–858.
Donaubauer, J., Meyer, B. E., & Nunnenkamp, P. (2016). A new global index of infrastructure: Construction, rankings and applications. World Economy, 39(2), 236–259.
Dornbusch, R., Fisher, S., & Samuelson, P. A. (1977). Comparative advantage, trade and payments in a Ricardian model with a continuum of goods. American Economic Review, 67(5), 823–839.
Duranton, G., Morrow, P. M., & Turner, M. A. (2013). Roads and trade: Evidence from the US. The Review of Economic Studies, 81(2), 681–724.
Eaton, J., & Kortum, S. (2002). Technology, geography, and trade. Econometrica, 70(5), 1741–1779.
Fally, T. (2015). Structural gravity and fixed effects. Journal of International Economics, 97(1), 76–85.
Fink, C., Mattoo, A., & Neagu, I. C. (2005). Assessing the impact of communication costs on international trade. Journal of International Economics, 67(2), 428–445.
Francois, J., & Manchin, M. (2013). Institutions, infrastructure, and trade. World Development, 46, 165–175.
Ghani, E., Goswami, A. G., & Kerr, W. R. (2014). Highway to success: The impact of the golden quadrilateral project for the location and performance of Indian manufacturing. The Economic Journal, 126(591), 317–357.
Grubel, H. G., & Lloyd, P. J. (1975). Intra-industry trade: The theory and measurement of international trade with differentiated products. London: Macmillan.
Head, K., & Mayer, T. (2014). Gravity equations: Workhorse, toolkit, and cookbook. In G. Gopinath, E. Helpman & K. Rogoff (Eds.), Handbook of international economics (Vol. 4, 131–195).
Head, K., Mayer, T., & Ries, J. (2010). The erosion of colonial trade linkages after independence. Journal of International Economics, 81(1), 1–14.
Head, K., & Ries, J. (2010). Do trade missions increase trade? Canadian Journal of Economics/Revue Canadienne d`Economique, 43(3), 754–775.
Hinz, J. (2017). The view from space: Theory-based time-varying distances in the gravity model. Beiträge zur Jahrestagung des Vereins für Socialpolitik 2017: Alternative Geld- und Finanzarchitekturen—Session: International Trade and Trade Reforms IV, No. E15-V2.
Jacks, D. S., Meissner, C. M., & Novy, D. (2011). Trade booms, trade busts, and trade costs. Journal of International Economics, 83(2), 185–201.
Kaufmann, D., Kraay, A., & Mastruzzi, M. (2011). The worldwide governance indicators: Methodology and analytical issues. Hague Journal on the Rule of Law, 3(2), 220–246.
Larch, M., Wanner, J., Yotov, Y. V., & Zylkin, T. (2017). The currency union effect: A PPML re-assessment with high-dimensional fixed effects. Drexel University School of Economics Working Paper 2017-07.
Larch, M., & Yotov, Y. V. (2016). General equilibrium trade policy analysis with structural gravity. WTO Staff Working Paper, No. ERSD-2016-08.
Limão, N., & Venables, A. J. (2001). Infrastructure, geographical disadvantage, transport costs, and trade. World Bank Economic Review, 15(3), 451–479.
Mayer, T., & Zignago, S. (2011). Notes on CEPII’s distances measures: The GeoDist database. CEPII Working Paper 25.
Miroudot, S., Lanz, R., & Ragoussis, A. (2009). Trade in intermediate goods and services. OECD Trade Policy Working Paper 93. Paris.
Novy, D. (2013). Gravity redux: Measuring international trade costs with panel data. Economic Inquiry, 51(1), 101–121.
Portugal-Perez, A., & Wilson, J. S. (2012). Export performance and trade facilitation reform: Hard and soft infrastructure. World Development, 40(7), 1295–1307.
Redding, S. J., & Venables, A. J. (2004). Economic geography and international inequality. Journal of International Economics, 62(1), 53–82.
Santos Silva, J. M. C., & Tenreyro, S. (2006). The log of gravity. The Review of Economics and Statistics, 88(4), 641–658.
Straub, S. (2011). Infrastructure and development: A critical appraisal of the macro-level literature. Journal of Development Studies, 47(5), 683–708.
Timmer, M. P., Dietzenbacher, E., Los, B., Stehrer, R., & de Vries, G. J. (2015). An illustrated user guide to the world input-output database: The case of global automotive production. Review of International Economics, 23(3), 575–605.
Vijil, M., & Wagner, L. (2012). Does aid for trade enhance export performance? Investigating the infrastructure channel. World Economy, 35(7), 838–868.
Volpe Martincus, C., Carballo, J., & Cusolito, A. (2017). Roads, exports and employment: Evidence from a developing country. Journal of Development Economics, 125, 21–39.
Wilson, J. S., Mann, C. L., & Otsuki, T. (2005). Assessing the benefits of trade facilitation: A global perspective. World Economy, 28(6), 841–871.
Yotov, Y. V. (2012). A simple solution to the distance puzzle in international trade. Economics Letters, 117(3), 794–798.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix A1: Methodological notes on the index of infrastructure
The following description of the employed infrastructure index draws on Sect. 3 in Donaubauer et al. (2016). An unobserved components model (UCM; or multiple-indicator model) is employed which combines information on specific aspects of infrastructure from different sources into aggregate indices of infrastructure. The UCM treats the unobserved indices of infrastructure as random variables to be estimated on the basis of observed, though imprecise and incomplete indicators of infrastructure. This approach has two major advantages: First, the constructed indices can be expected to be more precise and informative about the quantity and quality of infrastructure than any single indicator. Second, the UCM expands the number of observations to be used for comparing the quantity and quality of infrastructure.
The UCM closely resembles the procedures and steps described by Kaufmann et al. (2011) to arrive at their well-known indices on governance. Estimated summary measures of infrastructure for each country c are obtained by condensing the information on observed infrastructure indicators \(l\). The idea is that all observable infrastructure indicators include a commonly unobserved infrastructure component that reflects the underlying infrastructure endowment of a country, which by itself is only imperfectly measurable. Therefore, the expected value of the distribution of an unobserved common component of infrastructure, \(I_{c}\), is obtained conditional on the observed data of specific indicators for country c:
where \(y_{cl}\) represent the scores of country \(c\) and infrastructure indicator \(l \in \left[ {1,L} \right]\), \(w_{cl}\) represent the corresponding weights, and \(\alpha_{l}\) and \(\beta_{l}\) are parameters to be estimated (see below). Thus, the infrastructure index is estimated as the sum over all L observed infrastructure indicators weighted by the individual sources according to their precision.
The observed scores, \(y_{cl}\), are expressed in the UCM as a linear function of the unobserved common component of infrastructure \(I_{c}\) and an error term \(\varepsilon_{cl}\) capturing perception error and sampling variation in each indicator:
The parameters \(\alpha_{l}\) and \(\beta_{l}\) map the unobserved terms into the observed data space while accounting for different underlying data sources and units of measurement.Footnote 52 The error term is assumed to be independent and identically distributed with mean \(E\left[ {\varepsilon_{cl} } \right] = 0\) and variance \(Var\left[ {\varepsilon_{cl} } \right] = \sigma_{l}^{2}\). The variance differs across indicators but is the same across countries. Further assuming the errors to be independent across sources (\(E\left[ {\varepsilon_{cl} \varepsilon_{cm} } \right] = 0\) for \(l \ne m\)) allows to identify the particular unobserved information from each data source that feeds into the overall infrastructure index. Thus, the correlation between two different data sources can be exclusively attributed to the common underlying unobserved infrastructure \(I_{c}\). To obtain estimates for \(\alpha_{l}\), \(\beta_{l}\) and \(\sigma_{l}^{2}\), the log-likelihood function of the observed infrastructure data is maximized subject to \(\alpha_{l}\), \(\beta_{l}\) and \(\sigma_{l}^{2}\), assuming a normally distributed random variable \(I_{c}\) with mean zero and standard deviation one.Footnote 53 To facilitate the calculations, it is assumed that \(I_{c}\) and \(\varepsilon_{cl}\) are jointly normally distributed.
The UCM allows obtaining internal weights for indicators that feed into the aggregated index from the estimated variance. The weights \(w_{cl}\) are a decreasing function of the variance of the infrastructure indicator l and an increasing function of the variance of all infrastructure indicators:
Hence, the lower the variance of indicator l, the higher its precision and the weight assigned to the respective indicator.
This approach facilitates the estimation of the weights assigned to each indicator. From a theoretical point of view this is clearly preferable to calculating unweighted averages as is often done for cross-sectional composite infrastructure indices (e.g. Limão and Venables 2001). The indices are rescaled such that they are comparable across countries and over time.
Appendix A2
Summary statistics
Mean | Std. dev. | Min | Max | |
|---|---|---|---|---|
X (exports) | 2627.73 | 97,330.22 | 0.00 | 13,700,000.00 |
gdppc | 8.82 | 1.23 | 5.97 | 11.26 |
pop | 16.37 | 1.50 | 12.89 | 21.02 |
dist Hinz | 8.58 | 0.92 | 1.71 | 9.89 |
rta | 0.20 | 0.40 | 0.00 | 1.00 |
cu | 0.05 | 0.21 | 0.00 | 1.00 |
infra | 3.56 | 0.48 | 1.28 | 4.61 |
GL _infra | 4.27 | 0.28 | 1.96 | 4.61 |
About this article
Cite this article
Donaubauer, J., Glas, A., Meyer, B. et al. Disentangling the impact of infrastructure on trade using a new index of infrastructure. Rev World Econ 154, 745–784 (2018). https://doi.org/10.1007/s10290-018-0322-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10290-018-0322-8