Production Fragmentation and Transport Mode Choice in EU-28 Exports

ABSTRACT

This study uses detailed data on exports from the European Union (EU-28) to non-EU countries to investigate the impact of global production fragmentation on the choice of transportation modes. The article finds a positive relationship between production fragmentation and air transportation. The study also finds that air transportation is more likely to be chosen for high-value products. Additionally, the positive effect of production fragmentation on the likelihood of using air transportation is more pronounced when goods are exported to medium-range markets, upper-middle-income countries, countries with better air transportation infrastructure, and time-sensitive products.

KEYWORDS:

• EU-28
• fragmentation of production
• international trade
• mode of transportation

Disclosure statement

The data that support this study’s findings are available from the corresponding author upon reasonable request.

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/08853908.2024.2312911.

Notes

¹ Besides geographic connectivity, the World Bank (2020) highlights the importance of factor endowments, market size, and institutional quality in expanding GPNs.

² Exceptions are Hornok (2011) and Wessel (2019). However, neither specifically consider the effects of fragmentation on transportation mode choice.

³ Note that the EU database only reports trade statistics by transport mode with external partners.

⁴ An alternative and more detailed approach is to use firm-level and customs data in measuring global production fragmentation (Johnson 2018; World Bank 2020). Such efforts have been underway, and the absence of this data limits us to using product-level customs data in the analysis.

⁵ BEC is a three-digit classification grouping goods by their main end use.

⁶ The structural gravity model is used extensively in trade models. For comprehensive surveys of the gravity literature, see Baldwin and Taglioni (2006), Anderson (2011), Head and Mayer (2014), and Yotov et al. (2016).

⁷ Because trade flows may not accurately reflect the real dynamics of mode choices in international trade, we use mode shares, not export values, as the dependent variable in our regressions. Empirical estimations using export flows (not reported here but available upon request) yield similar coefficients.

⁸ It is the dollar value of exports per unit of physical quantity.

⁹ Available at https://ec.europa.eu/eurostat/estat-navtree-portlet-prod/BulkDownloadListing.

¹⁰ The EU database reports trade flows of the 28 EU members by transportation mode with external partners, except for a few small states.

¹¹ Johnson and Noguera (2012) extended the concept of vertical specialization proposed by Hummels, Ishii, and Yi (2001) by integrating global input-output tables with bilateral trade data from the Global Trade Analysis Project (GTAP). They calculated the value-added content of bilateral trade to gauge cross-border production linkages.

¹² For a more detailed explanation, please refer to Box 1.3 in the World Development Report (World Bank 2020).

¹³ Various indicators are employed to measure the degree of fragmentation, each utilizing different methodologies and data sources. Feenstra and Hanson (1996), Campa and Goldberg (1997), and Hummels, Ishii, and Yi (2001) computed fragmentation indicators using input-output tables. Görg (2000), Egger and Egger (2005), and Clark (2006) employed outward and inward processing trade statistics. Andersson and Fredriksson (2000) and Kimura and Ando (2005) used intra-firm trade statistics. Yeats (2001), Ando (2006), and Kimura, Takahashi, and Hayakawa (2007) used either the volume of trade in P&C or the vertical intra-industry trade (IIT) index in intermediate goods.

¹⁴ The remaining items are classified as capital goods.

¹⁵ Available at https://www.oecd.org/sti/ind/bilateraltradeingoodsbyindustryandend-usecategory.htm.

¹⁶ Clustering standard errors at different levels (country-pair or exporter-importer-product) yielded consistent conclusions. The results are available upon request.

¹⁷ Similarly, Harrigan (2010) and Wessel (2019) reported unexpected findings on income variables’ influence on transport mode choice.

¹⁸ The sector-level analysis, available upon request, are consistent with our expectations. This suggests that TiVA indicators, while not ideal for product level fragmentation due to their sector-level nature, might be better suited for sector-based models.

¹⁹ The closest distance category, D (sea distance ≤ 1,000), was dropped from the estimations because they were perfectly collinear.

²⁰ The lower-middle income group was dropped from the estimations due to perfect collinearity with the other two income groups.