VoxEU Column Frontiers of economic research International trade

Towards “trade policy analysis 2.0”: From national comparative advantage to firm-level trade data

The world of international trade has been in constant evolution since the rise of containerisation. This column makes the case for the need to upgrade our toolbox for trade policy analysis. An upgraded "Trade Policy Analysis 2.0" would be based on firm-level statistics and a much more refined product disaggregation, both of which are now becoming widely available. 

International trade is present in everyone's life.  Be it the fruits we have at breakfast or the electrical devices we use, our daily routine depends on complex trade flows and production processes scattered across multiple countries that hardly get noticed by the final consumer. Trade flows have evolved over time and are becoming increasingly intricate, with countless parts and components crossing multiple borders at different stages of production along global supply chains, before reaching the final consumer. 

Trade realities and theories: The role of globalisation and technological progress

International trade was revolutionised by the introduction of the standardised container, which led to a considerable reduction in shipping costs (Levinson, 2006). The revolutionary technological changes did not stop with the invention of the container. Today, a limited but growing number of these containers are equipped with sophisticated global tracking technologies (GPS, radio frequency identification, satellite communications, etc.) that can locate products and shipments in real time, optimising supply chains and inventories for the ultimate benefit of consumers. Detailed firm-level trade data on actual shipments, by exporting and importing firms, with specific product details and their port of origin and entry are publicly available. The data does not stop at the docks. Producers can track in real-time their stocks on each supermarket's shelf and plan the next shipment to make sure consumers do not face shortages, while avoiding waste and costly warehousing. Firms engaged in global supply chains and those specialised in logistics have developed detailed classifications that allow the identification of producers, the location of their production facilities and the most detailed product characteristics about brands, quantity (weight, number of units, pack sizes), quality (concentration levels of various key ingredients) as well as pricing, delivery and invoicing information.

But is this multifaceted reality fully accounted for in trade theories and well reflected in the statistical and analytical support available to trade policymakers?

Over time, trade theories have made major strides in capturing such diverse realities. For a long period trade theory was elaborated not so much on the trading firm (exporter/importer) but on much more aggregate concepts and had to assume a number of simplifying assumptions. The traditional trade theory was based on concepts like national comparative advantage or factor endowments. For decades trade analyses were run based on the Ricardian and Heckscher-Ohlin models. In those models it was not really the firms but nations that engaged in trade. The "new trade theory" developed by Krugman (1979) won him the Nobel Prize and introduced new useful concepts that brought theory closer to the realities of international trade. Consumers love product varieties and firms produce differentiated products under increasing returns to scale. For policy and empirical purposes, "new trade theory" models and analyses added useful insights but they did not really distinguish between firms within sectors in terms of their characteristics and ability to export. It was only with the emergence of "new new trade theory" articulated by Melitz (2003) and subsequently developed by other trade economists (e.g. Bernard et al. 2007) that the firms became central in explaining trade flows, just as in the trade examples offered above.

The "new new trade theory"  represented a "revolution" for trade theories and their ability to capture the detailed firm-level trade reality, by putting an emphasis on the central role of firm heterogeneity and by using newly available firm-level trade datasets. The new firm-level trade datasets that are already publicly available allow researchers to identify and analyse each and every shipment of the 30 million containers trade around the world by carrier, port of departure and destination, countries, description of products and commodities at the most detailed level (e.g. 8 or 10 digit national tariff line level), equipment type, size, weight, value, currency used, shipper and consignee's street, city, zip code, plus any other detail contained in the entire bill of lading. One can also match this information with the most important firm characteristics (e.g. from publicly available balance sheets) that have an impact on export performance. So thanks to the new firm-level trade data "revolution", available trade models come nowadays much closer to business realities.

Current trade policy analysis: Strengths and weaknesses

What difference did all these major improvements - either in trade realities or theories - make for trade policy analysis? By and large, not that much.  But such technological and analytical developments clearly offer a good basis for an upgraded "Trade Policy Analysis 2.0" platform, thus bringing trade policy closer to where the action is. The traditional analytical tools we have at our disposal (such as the standard computable general equilibrium models widely used for trade policy analysis) have great strengths, notably at estimating the macroeconomic effects of trade policy, but they remain imperfect.  Current tools tend to work well at aggregate level and whenever more detailed analyses are necessary they tend to be difficult and expensive.   Think of only a few concrete trade policy examples. Antidumping actions, WTO trade disputes, sensitive tariff lines in bilateral or plurilateral negotiations, often boil down to very detailed products for which even the most detailed cross-country trade statistics (based on the Harmonised System at 6 digits, i.e. HS6) are too aggregate. In reality, within each of these HS 6 digit codes product differentiation is considerable. The same HS6 code could cover for instance an entire shelf in supermarkets, with huge variety in product quality or functionalities.

One of the main objectives of EU trade policy is to create new business opportunities for over 2 million importing and exporting firms in Europe, including for a large number of small and medium enterprises that are successful exporters (Cernat et al, 2014). If those firms, and the tens of millions of workers they employ, are the primary beneficiaries of EU trade policy, it would be a major improvement in the robustness of trade policy analyses if "the firm" was the underlying unit of analysis. Moreover, there is growing consensus that trade policy needs to be well embedded within this broader set of economic policies, as recently reiterated by EU political leaders. Firm-level trade statistics may be the new frontier of an enhanced data-driven trade policymaking, similar to recent analytical developments underpinning other public policies and many corporate decisions. Big data is making major inroads in economics (Einav and Levin, 2014) and produced already a shift in the way public policy decisions are made, with major improvements in the efficiency of such policies. Firm-level trade data has the potential to improve trade policy analysis just as these new datasets led to major improvements in other public policies.  Firm-level trade data can bring benefits at all levels of analytical support to trade policymaking (ex-ante and ex-post). Detailed firm-level trade data might also improve communication, leading to a more meaningful engagement with stakeholders and thus reduce public misperceptions about trade policy. A broad-based firm-level approach to trade policy might offer a more fact-based dialogue. Trade policy analyses are hardly sufficiently disaggregated for individuals and communities to relate to its predicted impacts. Why should a particular country or societal group be in favour of trade liberalisation if there is little information about what trade policy can bring to them as opposed to Europe?

Towards trade policy analysis 2.0: Taking advantage of existing firm-level trade data

Firm-level theoretical and empirical analyses have demonstrated that export performance is critically determined by firm characteristics such as the ability to innovate, productivity levels, firm size, corporate governance, skills and labour market, the overall domestic business environment, etc. But few, if any, trade policy analyses have been conducted at firm-level. Therefore, the first thing one should do is to take advantage of the available firm-level trade data in the public domain around the world. These databases are already compiled and exploited by companies promoting transparency on new market opportunities for potential exporters. 

Trade policymakers can also tap into these publicly available databases. There are also good news coming from global value chains. Firms managing such supply chains rely on various standard, universal product codes and global databases developed by the logistics industry to allow participating firms (e.g. suppliers of components and final assembly firms)  to know exactly the brand, variety, quality, dimensions, essential product characteristics, and price range of billions of traded products. Such detailed databases could considerably transform the way trade policy analysis is conducted. In "Trade Policy Analysis 2.0" the unit of analysis shifts from countries and sectors to exporting and importing firms. Once the actual exporters and importers become the unit of analysis, firm-level trade data will also provide a much more refined product disaggregation.  

In "Trade Policy Analysis 2.0" the unit of analysis for trade products will also move from HS6 product classification to real product codes, the so-called Global Trade Item numbers (GTINs) that are used routinely by companies trading along the supply chain. Such GTIN-based trade statistics do not just simply record "widgets" but would contain many product attributes and differentiate different products by key characteristics of the specific variety and its producing firm. A growing number of academic articles have already been published using such barcode retail data but, despite an important analytical potential, very few have actually addressed questions relevant for trade policy analysis.

Firm-level data could also improve our understanding in other new policy areas. The overwhelming role of international investment in the creation of global supply chains is generally absent from the standard models used for policy assessment. The major role of global supply chains in shaping trade patterns also triggered a debate about the need for a major "upgrade" in multilateral trade towards a new "WTO 2.0" set of rules dealing with the intricate interaction between investment, services and intellectual property (Baldwin, 2012). The growing importance of intermediate services exported "in a box" as part of processed goods (Cernat and Kutlina-Dimitrova, 2014) also require further reflection on how to bridge the various gaps between GATS and GATT rules.

The leap towards "Trade Policy Analysis 2.0" does not mean that current analytical tools should be discarded, quite the contrary. The current computable general equilibrium models and their standard Armington structure offer a solid analytical basis and contain the seeds for their own evolutionary course towards greater policy relevance. New theoretical discussions and the wealth of empirical firm-level data used recently in extended models make the goal of building a policy-relevant model "at firm level" look more attainable than a couple of years ago.

Containers revolutionised shipping and reduced international trade costs. Firm heterogeneity revolutionised trade theory and the new firm-level data "by container" revolutionised empirical trade analyses. Finally, the big data approach stands to revolutionise economics, as well as public policies. Will all these fundamental factors revolutionise the analytical support to trade policy making towards a more systematic use of firm-level trade data?  Based on the arguments presented above, most probably the question is not "whether" but "when".

Note: A longer version of this paper was published as part of the DG TRADE Chief Economist Notes series.  The opinions expressed herein are those of the author and do not necessarily reflect the views of the European Commission.


Baldwin, R (2012), "WTO 2.0: Global governance of supply-chain trade", CEPR Policy Insight no. 64/2012, Centre for Economic Policy Research, London.

Bernard, A B, S J Redding, and P K Schott (2007), "Comparative Advantage and Heterogeneous Firms", Review of Economic Studies 74 (1): 31-66.

Cernat, L A. Norman and A. Duch T-Figueras (2014). SMEs are more important than you think! Challenges and opportunities for EU exporting SMEs. DG TRADE Chief Economist Note no. 3/2014, Brussels.

Cernat, L and Z Kutlina-Dimitrova (2014), "Thinking in a box: A ‘mode 5’ approach to service trade", DG TRADE Chief Economist Note no. 1/2014, Brussels.

Einav, L and J Levin (2014), "Economics in the age of big data", Science 346 (6210), 7 November.

Krugman, P (1979), "Increasing returns, monopolistic competition, and international trade", Journal of International Economics 9: 469-479

Levinson, M (2006). The Box: How the Shipping Container Made the World Smaller and the World Economy Bigger, Princeton University Press.

Melitz, M (2003), "The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity", Econometrica 71: 1695–1725

1,575 Reads