VoxEU Column International trade

Rules of origin in trade arrangements: Largely unnecessary, simply protectionist

Rules of origin exist to avoid trade deflection, but they distort global value chains and are costly to abide by. This column shows empirically that in preferential trade agreements, trade deflection is unlikely to be profitable because tariffs are generally low, that countries in a common free trade agreement tend to have similar external tariff levels, and that when tariff levels differ, deflection is profitable at most for one country in the pair. Moreover, transportation costs create a natural counterforce. It appears that rules of origin are primarily used to limit trade, and hence represent an instrument for trade protection. 

When two countries eliminate bilateral tariffs in a preferential trade agreement (PTA) but continue to have different tariffs vis-à-vis third countries, they require rules of origin (RoOs) that specify under which conditions a product is allowed to be traded tariff-free within the PTA. Otherwise, products from third countries could enter the PTA through the country with the lowest external tariff, therefore undercutting the other PTA members’ higher external tariffs (‘trade deflection’). In other words, RoOs are meant to define the ‘nationality’ of imports and to make PTAs preferential. Modern PTAs devote hundreds of pages to this purpose. This is true for reciprocal PTAs, to which we refer as free trade agreements (FTAs), and for non-reciprocal GSP-arrangements (see below).

However, global value chains (GVCs) make it increasingly problematic to operate trade policy on the assumption that one can clearly identify the nationality of a product (Baldwin 2016). As a consequence, FTAs are “tying up trade policy in knots and absurdities facilitating protectionist capture” (Bhagwati 1995). This is why Bhagwati has referred to the universe of FTAs a “spaghetti bowl”. An empirical literature shows that RoOs are costly to abide by, so exporters often prefer to pay tariffs rather than bear the red tape of compliance (Anson et al. 2005, Augier et al. 2005, Cadot, et al. 2014, Keck and Lendle 2012). Moreover, if RoOs are satisfied, they distort the sourcing behaviour of exporters (Conconi et al. 2016, 2018).

One may think that this is necessary collateral damage to combat trade deflection. However, as we show in a recent paper (Felbermayr et al. 2019), trade deflection is rarely profitable for traders. So, negotiators of PTAs could do with strongly relaxed RoOs, except if their true intention lies in achieving exactly those protectionist objectives of RoOs.

A simple definition

We calculate the difference between ad valorem trade costs (tariffs and transportation costs) that arise when country i imports directly from c as compared to when it imports via its PTA partner j. The scope of trade deflection is equal to that difference when the difference is positive and zero else.

An empirical question: How realistic is trade deflection?

In our empirical analysis, we use detailed global data on tariffs and transport costs to calculate how often, in the absence of RoOs, trade deflection would be profitable. For this, we need to check whether it would make economic sense for an importer in country i (e.g. the US) , which is in an FTA with country j (e.g. Canada), to source a product from a third country c (e.g. China) by first importing the product into j and from there ship it to country i.

Assume that there are no transportation costs and that trade in the relevant product is free of tariffs between countries i and j. Then, if country j has a lower tariff on the product vis-à-vis country c than country i, trans-shipping through j would be profitable. Unlike in an FTA, in a customs union trade deflection can never be profitable due to a common external tariff policy. 

Of course, in many trade arrangements there are products for which internal tariffs are not zero, and shipping a good from c through j rather than directly to i may involve additional transportation costs. Taking these complications into account is analytically straightforward. 

However, in our empirical analysis of trade deflection, we had to deal with several challenges. First, we had to assemble a comprehensive global data base on applied most favoured nation (MFN) and preferential tariffs on a harmonised product classification and for all countries in the world. Second, transportation data are available on a bilateral and product-specific level for the US and New Zealand only. For other countries, we had to estimate those costs. We find that median ad valorem equivalents of transportation costs are a bit smaller than 6%.

Third, the analysis suffers from a curse of dimensionality: for the year 2014, we observe 5,729 country pairs ij, on average 2,640 products, and 170 third countries c so that the number of observations is equal to more than 2 billion per year. This is why, in our baseline, we look at the 20 most important third countries for every country-pair x product combination, but we propose a large array of robustness checks. 

Basic results

In our baseline analysis, for 93% of all country-pair x product x third country combinations, the scope for trade deflection is exactly zero (Figure 1). For 96% of all cases it is smaller or equal to 3%, meaning that routing imports from c to i via j yields an ad valorem trade cost advantage (tariffs and transportation costs) of 3%. For 99% of cases, that gain is smaller or equal to 10%. If we weight by trade volume, the scope for trade deflection shrinks even further (not shown in the figure). 

Figure 1 Scope for trade deflection

Note: The figure shows shares of cases, for which trade deflection is less or equal to the value given on the horozintal axis.
Source: Felbermayr et al. (2019).

The “Max. Deflection” measures shown in Figure 1 (red and green lines) provide conservative upper bounds. The red line covers all cases, regardless of whether trade takes place. The second disregards all third countries c that do not export to i. Not surprisingly, in those cases there is more scope for trade deflection; nonetheless, trade deflection is still not profitable for at least 49% of all cases. When we condition on positive trade flows, the “Max. Deflection” measure is zero for 75% of all cases. Note that in all our calculations, we account only for transportation costs. Because other trade costs are also very sizeable (Anderson and van Wincoop 2004), all our results can be considered as conservative.

FTAs versus GSP arrangements

FTAs are reciprocal trade agreements. The General System of Preferences (GSP), in contrast, is non-reciprocal. Figure 2 shows that in 98% of all cases, our measure of trade deflection is zero for preference schemes of rich countries with developing ones. This is not surprising – while the rich countries offer tariff-free access to GSP partners for most products, the developing countries maintain their external (MFN) tariffs, which are almost never lower than those of the rich countries. Therefore, the rich country will rarely find it optimal to import a good through the developing country. For the poor country, trans-shipping through the rich one makes even less sense as imports are still subject to tariffs.

Figure 2 FTAs versus asymmetric preference schemes (GSP), with and without transportation costs



Note: The figure shows shares of cases, for which trade deflection is less or equal to the value given on the horizontal axis.
Source: Felbermayr et al. (2019).

The role of transportation costs

Accounting for transportation costs makes a sizeable difference for symmetric preferential trade agreements (FTAs). Figure 2 shows that for 86% of all product x country-pair x third-country combinations, the scope of trade deflection is zero; in 16% of all cases, this is due to the presence of transportation costs (difference between solid and dotted red line). For GSP arrangements, transportation costs make less of a difference.

Explaining the lack of scope for trade deflection

Figure 3 shows that across all trade arrangements, trade deflection is not profitable because for 63% of all cases, the importer i has lower tariffs than the FTA partner j. In these cases, our measure of trade deflection is zero, because there is no need to trans-ship. For GSPs, that share is 82%, reflecting the asymmetric nature of the arrangement; for FTAs, it is much smaller (38%). On average, both parties in an arrangement have identical external tariffs in 18% of all cases; in FTAs that share is 29%. Clearly, this nullifies the scope of trade deflection. On average, for 10% of all cases, tariffs are such that trade deflection is an option, but transportation costs eliminate the incentive. For FTAs, that share is 16%; for GSPs it is 6%.

Figure 3 For what shares of cases is trade deflection profitable?

Source: Felbermayr et al. (2019).

Sectoral heterogeneity

In our paper, we document a substantial degree of sectoral variation. For many sectors, the scope for trade deflection is virtually zero. The only areas where trade deflection is an issue are the agricultural sectors, pulp and paper, as well as the sector of works of art.

What this all means

Empirically, trade deflection is rarely profitable in most configurations. Nonetheless, virtually all existing preferential trade arrangements feature complicated RoOs for all products to combat deflection. RoOs are costly for firms, and detrimental for welfare (Conconi et al. 2016, 2018). They reduce preference utilisation rates of PTAs, as exporters prefer to pay tariffs rather than suffer the red tape of documenting their sourcing or adjust sourcing to satisfy the rules. They are the unsavoury sauce to the spaghetti bowl of bilateral trade arrangements. So, if there is no need for them, one should fundamentally review their use.

A proposal for reform

Considering the empirical findings, we suggest that in new FTAs, negotiators should agree on a full set of simple RoOs for all products, but that the requirement to prove origin should be activated only if external tariffs of FTA members differ by some minimum amount. This threshold could be product-specific in order to reflect different transportation costs and actual tariffs should be periodically evaluated against it, since applied tariffs may change over time. In GSPs, RoOs should be activated only for those products where the beneficiary country undercuts the MFN tariffs of the preference granting country. For sure, making the proof of origin conditional on actual tariff differences would go some way toward disentangling the spaghetti bowl.


Anderson, J and E van Wincoop (2004), “Trade Costs", Journal of Economic Literature 42(3): 691–751.

Anson, J, O Cadot, A Estevadeordal, J De Melo, A Suwa-Eisenmann, and B Tumurchudur (2005), “Rules of Origin in North-South Preferential Trading Arrangements with an Application to NAFTA”, Review of International Economics 13(3): 501–517.

Augier, P, M Gasiorek, and C Lai-Tong (2005), “The Impact of Rules of Origin on Trade Flows”, Economic Policy 20(43): 586–624.

Baldwin, R (2016), The Great Convergence: Information Technology and the New Globalization, Harvard University Press.

Bhagwati, J (1995), “US Trade Policy: The Infatuation with Free Trade Agreements”, in J Bhagwati and A O Krueger (eds), The Dangerous Drift to Preferential Trade Agreements, American Enterprise Institute for Public Policy Research.

Cadot, O, A Graziano, J Harris, and C Volpe (2014), “Do Rules of Origin Constrain Export Growth?”, IDB Integration and Trade Sector No. IDB-DP-350.

Conconi, P, M García Santana, L Puccio and R Venturini (2016), “The perverse effect of preferential rules of origins”,, 16 March.  

Conconi, P, M García Santana, L Puccio and R Venturini (2016), “From final goods to inputs: the protectionist effect of rules of origin”, American Economic Review 108(8): 2335-2365.

Felbermayr, G, F Teti, and E Yalcin (2019), “Rules of origin and the profitability of trade deflection", Journal of International Economics, forthcoming.

Keck, A and A Lendle (2012), “New Evidence on Preference Utilization”, WTO Staff Working Papers No. ERSD-2012-12

3,045 Reads