VoxEU Column International trade

How preferential is world trade?

How much of world trade is preferential? This column uses tariff data covering around 90% of world trade in 2008 to show that only 16.3% of world trade is eligible for preferences. It shows that preferences reduce the global trade-weighted tariff from 3% to 2%.

The number of regional trade agreements is increasing rapidly. This is of concern to many who care about the world trading system since non-discriminatory liberalisation, whether unilateral or at the WTO level, would be preferable (Bhagwati 2008). But how much of today’s trade is actually preferential?

This conceptually simple question has not been answered precisely, probably due in part to the lack of easily available data at the most detailed level and also to the fact that matching tariffs with import flows is often far from straightforward.1

New evidence on preferences: Tariff line data matched with bilateral trade

In recent research (Carpenter and Lendle 2011), we develop a database matching import flows with tariff data – both most-favoured-nation (MFN) and preferential – for the 20 largest importers in 2008, which covers almost 90% of world trade. Our data is at national tariff-line level for most importers, which means import flows can be matched exactly with the applicable tariffs. The main sources are the International Trade Centre’s TradeMap and Trains / WITS. We provide here a snapshot of the preferences applied to existing trade rather than an overall picture of the degree of protectionism since the latter would have to include protection of non-traded items. We also do not take into consideration that preferential trade flows are endogenous, i.e. depending on preferences.

We define preferential imports as imports that are eligible for a preferential tariff with a positive preferential margin, which is the difference between the MFN tariff and the preferential tariff. If the MFN tariff is zero, then the import is by definition non-preferential. We do not observe whether a preference is actually used, therefore our results show upper estimates of preferential trade flows. However, we know that utilisation rates are high (above 80%) at least in the EU and US markets, so the actual share of preferential trade is probably not much lower.

It would be possible to define a minimum preferential margin that is necessary to cover utilisation costs. We don’t do this, but instead show exactly how much of trade falls into different brackets of preferential margins. It turns out – perhaps unsurprisingly – that a large share of preferential trade has low margins. We also show MFN tariffs applicable to non-preferential trade; and the share of trade at zero MFN tariffs. Since our dataset covers imports by 20 countries from all their trading partners, we can show shares of preferential trade and other indicators by importer for 20 importers and by exporter for exports from all countries to the 20 covered importers. Our paper also shows other aggregates, e.g. by type of preference regime, product category etc., and a variety of other indicators.

What are the results?

We find that around half of world imports are from countries that are granted some preference – yet that does not mean that all of that trade is actually preferential. MFN rates may be zero or a product can be excluded from preferences. Overall – but not including intra-EU trade – just 16.3% of global trade has a positive preferential margin. Preferential margins are distributed as follows:

  • 10.5% of total trade has a margin of 5 percentage points or less, 3.9% has a margin of 5-10 percentage points, 1.3% has a margin of 10-20 percentage points and only 0.5% has a margin of 20 percentage points and above.
  • Around 30% of global trade is not eligible for preferences and subject to non-zero MFN rates, although almost 2/3 of that trade faces tariffs of 5% or less.
  • More than half (52%) of world trade is at MFN zero, so no preferences can be granted.

Figure 1 shows these summary statistics for all 20 importers and also for intra-EU trade.2

Figure 1. Preferential, non-preferential and zero MFN trade by importer3 (PM = preferential margin)

Shares of preferential imports with high margins (above 10%) are small for most countries, with the main exception being intra-EU trade and imports of Mexico. The overall shares of preferential trade are very high for intra-EU trade (64%), Switzerland (54%), Mexico (48%), Turkey (40%), and Canada (35%). Except for India, Russia, and the United Arab Emirates, all countries have fairly high shares of MFN zero imports.

We can aggregate the import data by exporting country.4 The results for the 30 largest exporters are shown in Figure 2. We can see that many large countries do not receive preferences for most of their exports, and if they do then preferential margins are usually low. Countries for which more than 30% of exports are eligible for preferences are Canada, Mexico, Switzerland, Turkey, and Nigeria.

Figure 2. Preferential, non-preferential, and zero MFN exports for 30 largest exporters

Another way to measure the degree of preferences is the trade-weighted preferential margin, which is equivalent to the “saved duties” (due to preferences) as a share of total trade. On aggregate, preferences reduce the global trade-weighted average tariff from 3.2% to 2.1%, hence the global trade-weighted preferential margin is 1.1%. Out of the 20 importing countries that we cover in our dataset, Mexico grants by far the largest preferences for its imports with a trade-weighted margin of 9.3%. Intra-EU trade also has very high margins when compared with the MFN rates applicable to third-country imports (4.9%), whereas the average preference granted to EU imports from third countries is only 0.9%.

Figure 3. Trade-weighted preferential margins by importer

Looking at the 25 largest exporters, the countries receiving the highest margins for their exports are Turkey (5.0%), Mexico (3.0%) and the US (2.8%). This is mainly due to high preferences in the EU market (for Turkey) and high preferences granted within NAFTA. All other large countries receive much lower preferences, and some of them – for example China, Japan and Australia, receive practically no preferences.

Figure 4. Trade-weighted preferential margins by exporter – 25 largest exporters

Does that mean that preferences do not matter? Average preferential margins at the global level are small, which suggests that they do not matter much “on aggregate” for many of the largest traders. However, some countries do receive very substantial preferences for their exports, and obviously some sectors even in other countries also do so. Figure 5 shows the countries that receive the highest preferential margins on their exports. Most of them are African, Caribbean and Pacific countries who export items with high MFN tariffs duty-free to the EU (and to a lesser extent to the US), e.g. sugar, bananas, fish, and garments. Mauritius is leading, with a trade-weighted margin of 24%.5

Figure 5. Trade-weighted preferential margins by exporter – 25 exporters with the largest trade-weighted preferential margin for their exports

Recent trends and conclusion

Our dataset covers only 2008 so we cannot observe any trend in the share of preferential trade and its margins. However, the share of preferential trade will probably increase with the ongoing proliferation of regional trade agreements. But there is an opposing trend of MFN tariff liberalisation, which reduces preferential margins and could also decrease the share of preferential. A back-of-the-envelope calculation of how the recently signed trade agreements that are not covered by the 2008 data would influence the share of preferential trade once fully implemented suggests that it will probably go up from 16% to 18% (assuming static trade flows and no further MFN liberalisation).

What can we conclude so far? We have not yet used the data for any empirical tests – which is the obvious way forward, but the stylised facts presented so far suggest that preferential trade is not as important on a global scale as one might expect from looking at the number of regional trade agreements alone. By contrast, around 80% of today’s trade occurs under MFN conditions. Obviously, regional trade agreements are not only about tariff reductions on goods, but are increasingly important for services, non-tariff measures etc. Whether a share of preferential trade of 16.3% is “large” or “small” is open for debate, but it is clear that there is not much trade with high preferential margins. Preferences matter a lot for some sectors in some countries, but their impact on the bulk of global trade in goods could well be overstated.


Bhagwati, Jagdish (2008), Termites in the trading system: How preferential agreements undermine free trade, Oxford University Press.

Carpenter, T and A Lendle (2011), “How preferential is world trade?”, CTEI Working Paper, December.

Fugazza, M and A Nicita (2010), “The Value of Preferential Market Access”, Working paper.

Grether, J-M. and M Olarreaga (1998), “Preferential and Non-Preferential Trade Flows in World Trade”, WTO Staff Working Paper 10/1998.


1 The only paper we are aware of that tried to explicitly answer the same question is Grether and Olarreaga (1998), but with a slightly different definition of preferential trade – the results are therefore difficult to compare. Fugazza and Nicita (2010) use a similar dataset to create a bilateral index of protection to then test the impact of aggregated preferences on trade flows. Some of their summary statistics are hinting towards shares of preferential trade, with similar results.

2 Intra-EU trade is defined as “preferential” when the MFN rate for imports from third countries is positive.

3 The numbers do not add up to 100%, mainly due to unknown tariffs and / or unknown ad-valorem equivalents.

4 Note that not all exports are covered, but only those to the 20 markets covered in the data. This represents at least 80% of exports for most countries. Also note that results by importer and by exporter are based on the same data, so the aggregate results are identical.

5 Note that these numbers are highly dependent on the chosen ad-valorem equivalents for items such as sugar and bananas in the EU.

1,365 Reads