2.1. The Two Categories of ROO

Each PTA contains a protocol or chapter in the agreement text detailing and identifying the criteria that test whether a product meets the nationality requirement necessary to be eligible for preferential tariff treatment. The Revised Kyoto Convention of the World Customs Organization defines two main criteria to determine origin:

1. (a)

   ‘Wholly obtained or produced’ criterion (WO) and;

2. (b)

   ‘Substantial transformation’ criterion (ST).

Figure 1 classifies these criteria along the two categories of ROO described and analysed in this paper. Figure 1a lists the RWR categories. These are applied to all products seeking eligibility for preferential status. Figure 1b lists criteria that are specific to a product, hence the name of PSR. To satisfy the nationality requirement, a product must meet both the RWR and PSR criteria. Only when both are satisfied, does the product acquire preferential status.

Figure 1

Classifying rules of origin criteria necessary to meet the nationality requirement. (a) Regime-wide rules (RWR) – see Tables 3 and 4 for further description. (b) Product-specific rules (PSR) – see Table 1 for further description. Source: Authors’ elaboration.

Regime-wide rules (RWR) are classified into origin and certification provisions. To help evaluate the intent of these rules and how they affect exporters’ eligibility, we sub-divide origin and certification further into those that ostensibly add transparency and clarity, called transparency provisions and those that presumably add flexibility, called flexibility provisions. Transparency provisions presumably decrease the information costs. Likewise, we selected a category that is intended to capture flexibility, called flexibility provisions. Flexibility provisions presumably make ROO less restrictive by reducing the production adjustment cost (direct cost of compliance) in the case of origin provisions or the conformity assessment cost (indirect cost of compliance) in the case of certification provisions. This classification is to provide rough landmarks in the description and comparisons of RW rules across PTAs. Section 3 covers RW rules. Section 4 covers PSR.

Product-specific rules (PSR) listed in Figure 1b fall under two categories: Wholly Obtained (WO), and Substantial Transformation (ST) for those products that are not required to meet the WO criterion.⁵ The ST requirement specifies that the country of origin is the country where the last transformation took place. The criterion could be: (i) a Change of Tariff Classification (CTC) (i.e. the exported good must have a different tariff classification than that of any imported inputs); (ii) a Specific Process (SP) of production (e.g. the exported good must have undergone a chemical reaction); or (iii) a Value Content (VC) requirement (i.e. a minimum threshold of local value content). The three different criteria, along with qualifiers—themselves classified into Exceptions (EXC) and Allowances (ALW)—are used to meet the ST criterion. Note that qualifiers can come with each criterion. In some cases exporters have a choice among the criteria (‘either/or’) and in others the qualifiers have to be met (‘and’). Table 1 shows the combination of PSR in the PTAs covered in this paper. Section 4 covers PSR.

2.2. Decomposing Compliance Costs

Take a shirt producer in Mozambique, i, producing under perfect competition at unit cost of selling his shirt in South Africa. Neglecting transport costs, in the absence or ROO requirements, the CIF landed price will be pi*=ci=ci0 where the suffix 0 indicates that these are undistorted unit costs. This is the price received by the exporter. A Most-favored Nation (MFN) tariff, tar_i > 0 raises the price of shirts allowing domestic producers to charge the domestic price for like products, pdi=Pi*(1+tari). All exporters that are not SADC members will receive the border price.⁶ In the absence of costs associated with proving origin, the Mozambican producer would export under SADC preferential tariff of 0% and his preferential margin, marg_i would be equal to the MFN tariff, i.e. the utilization rate for product i, pur_i, would be unity:

Many studies [e.g. Cadot et al. (2006) and UNCTAD (2019)] have shown that satisfying ROO requirements have compliance costs. Then total unit cost for the exporter will include compliance costs, cic, so total costs will be ci=ci0+cic. If total unit costs inclusive of compliance costs are less than the preferential margin (here the MFN tariff), then the Mozambican shirt producer will export to South Africa under SADC preferences (2a). Assuming the Mozambican shirt producer has full information, if compliance costs exceed the preferential margin, the Mozambican exporter will export under MFN regime Equation (2b). The decision conditions (variables defined below) are:

From Equation (2a) and Equation (2b), the compliance cost component, cic, can be broken down into three sub-components:

• •

  A distorted cost component, ciD, resulting from the beneficiary being forced to source from the partner for instance meeting a minimum RVC;

• •

  An administrative cost component, ciA, related to obtaining a valid Certificate of Origin (CoO_i) and undergoing post-importation audits by customs authorities;

• •

  A rent-sharing component, µ_i, between the exporter of intermediate goods to the FTA partner exporting the final product to the producer of intermediates. For example, a South African exporter of fabric might charge a higher price for yarn sold to the captive Mozambican producer who has to meet an RVC to export to South Africa than to other shirt producers outside SADC.⁷

This decomposition applies similarly to preferential ROO and to non-preferential ROO that determine whether a product is subject to a nation’s trade policy (e.g. anti-dumping, health standard). The ROF at present covers only preferential ROO, although both share several types of ROO. In the case of preferential access, one can say that a ROO is equivalent to a conditional tax: if the ROO is not satisfied an importer must pay the relevant MFN tariff.

In practice, it is difficult to estimate compliance costs. Firms have different production costs, and prices will differ from production costs when production does not take place under perfect competition. Also business surveys of non-tariff measures conducted by ITC in 23 developing countries [ITC (2015)] found that procedural obstacles related to issuance of certificates of origin is a major concern for exporters. On the data side, only a few countries publish trade statistics according to the trade regime requested by importers.⁸ This means that data on Preference Utilization Rates (purs) are rarely available and data at the transaction level needed to estimate fixed costs are not available.⁹ Finally, keep in mind that, at best, positive utilization rates only suggest that preferential margins outweigh compliance costs. Positive utilization rates are not an estimate of compliance costs.

2.3. Describing and Measuring ROO Complexity

Three approaches are used to ‘measure’ differences in ROO. Each comes with advantages and shortcomings. Similarity between texts is straightforward but does not inform much even when controlling for word repetition and obvious synonyms. Regulatory distance goes a step further by attributing a binary value to the presence of a similar ROO. Going further, a restrictiveness index proposes an ordinal ranking of ROO in terms of their likely compliance costs. Together, these measures help see the ‘forest from the trees’. In this exercise, the large number of distinct PSR must first be screened, then translated into categories that can then be assigned an ordinal ranking. Table 2 gives details and an example [Further explanations in Gourdon et al. (2020b)].

3.1. Classifying Regime-Wide Rules (RWR) Provisions for Process and Certification

Tables 3 and 4 describe the process and certification for the eight African PTAs according to whether the provision is applied (YES) or not (NO). If the classification were the same for each provision across all PTAs, including on Regional Value Content (RVC) calculation method, all counts would be consistently either 0 or 1. Then, there would be less need for harmonization on provisions.¹⁰ The tables also show the R-index values to restrictiveness attributed to each provision, usually a (0) [100] when the transparency provision is (included) [not included] in the set of RWR. The same scoring is applied for flexibility provisions, a (0) [100] when the flexibility provision is (included) [not included].

Three patterns emerge from Tables 3 and 4. First, all PTAs have the same set of provisions on transparency for process, but not on transparency provisions for certification. Second, for both types of provisions, there is greater uniformity on transparency than on flexibility. Third, there is less uniformity on both types of provisions for certification than for process. The differences for flexibility are greater than for transparency, probably a reflection of the greater difficulty in reaching agreement on flexibility than on transparency. Table 4 also shows that the two largest PTAs, AfCFTA and TFTA, almost always show the greatest flexibility, although there is no mention of duty drawback for both and neither explicitly allows third-party invoicing.¹¹

Producing origin documentation is part of administrative costs, c^A, identified above in Equation (2a). These include fixed and variable costs. Both components depend on shipment size about which there is little evidence, although in the case of Shoprite, a South African retailer exporting across Southern Africa SADC members, the reported documentation costs are substantial.¹²

To recapitulate, all five transparency origin provisions are present across the PTAs considered here, AfCFTA included. Although documenting these provisions is time consuming, the resulting greater transparency helps to meet the original objective of preventing trade deflection from countries outside the PTA. Their presence also reduces uncertainty. On the flexibility side, AfCFTA adopted more origin provisions adding flexibility in comparison to other pre-existing PTAs (e.g. on duty drawback and exhibition). However, the roll-up provision featured in several PTAs, is not mentioned in AfCFTA.

Table 5 shows that harmonization is also needed on the choice of cumulation rule and on the method of calculation for value-added. So long as they are not coupled with other provisions, cumulation provisions are easily ranked from most restrictive (bilateral) to least restrictive (diagonal/full cumulation). With full cumulation, all inputs from non-members are considered as originating in the calculation of value-added as long as sufficient processing has been cumulatively carried out within FTA. The attributed ordinal restrictiveness scores in Table 3 reflect this ranking. For value content, the imported content method to determine non-originating content is the easiest and administratively least burdensome to document for exporters.¹³ Interestingly, as shown by the restrictiveness scores, for both criteria, so far, TFTA and AfCFTA have settled on choices that are least (or not more) restrictive than those across current FTAs.

3.2. Regime-Wide Rules: Similarity and Restrictiveness

Table 6 (and Table A1) uses the textual overlap and regulatory distance measures to see how close AfCFTAs’ proposed RW rules are with those of the other FTAs. Regulatory similarity percentages are always higher than the percentages for text overlap, a reflection of the binary measurement for regulatory proximity. As the scores are only ordinal, what counts is that the ranks are close for both, an indication that the two measurements give the same information. The low percentages for textual overlaps indicate differences in “form”, i.e. in drafting style. The highest measure of proximity is with the 26-member TFTA and Agadir, followed by GAFTA. Since both were engaged in a rapprochement with the EU this is not surprising. This could also reflect anticipation by AfCFTA negotiators of a rapprochement with the EU. More surprisingly, negotiators from the two large membership PTAs, ECOWAS and COMESA do not appear to have participated actively in the drafting, especially when captured by the formal drafting as measured by textual similarity.

Comparisons at the level of the 30 RW rules reported in Table A1 show the ‘roots’ of the AfCFTA. Inspiration from Agadir is apparent for the provision on Accessories, Spare Parts and Tools; from COMESA for packaging; and from SADC for fungible materials. Overall, however, overlap is usually small (<15%), at least partly a reflection of a difference in ‘form’. Also, negotiators may have faced few requests from domestic interest groups, or they may have wished to start from a blank slate when drafting the rules.

Applying the regulatory distance measure (see Table 2) to the process and certification provisions in Tables 3 and 4 show two patterns of AfCFTA’s provisions. First, AfCFTA’s provisions on process and certification are always closer for transparency than for flexibility mechanisms (i.e. points below the 45° line in Figure A1) a reflection that adopting transparency measures is less controversial and probably easier to implement than flexibility measures. Second, transparency for origin is similar for Tripartite, Agadir, SADC and ECOWAS. Patterns on certification are similar for EAC, COMESA and SADC. And across FTAs, similarities for origin are closer than for certification.

Compared with the complexity of PSR, devising an observation rule to construct an ‘R-index’ on RWR provisions (see Table 2) is relatively straightforward. For these cases, information is usually recorded as a YES–NO choice or a choice on the score to attribute to a RVC percentage where an across-the-board RVC of 70% is less restrictive than an RVC of 85% (see R-index values in Table 5). Because the presence of transparency and flexibility provisions are assigned low restrictiveness scores (except for high RVC percentages which are assigned a higher score) and the same weight is applied across provisions, the resulting restrictiveness patterns follow closely those for similarity. By this scoring, the less prevalent certification provisions are more restrictive than the uniformly prevalent process provisions (see Figure A2).

Figure A3 shows the RWR index for the four pillars we built: transparency types and flexibility types in origin provisions, and transparency types and flexibility types in certification provisions. This index clearly shows that AfCFTA members managed to agree on less restrictive and hence more trade-facilitating regime-wide rules compared to other PTAs in the region. This is especially the case for certification provisions of the transparency type but also for those of the flexibility type. On aggregate, AfCFTA does not perform significantly better than other PTAs on origin process provisions.

In conclusion, even for the relatively straightforward-to-categorize RWR, attributing restrictiveness is highly judgmental. To ease interpretation for the reader, we gave the same weight for each component in the index. Then, why, for example, should an across-the-board high RVC percentage of 65% receive the same weight in an overall index as a RVC percentage of say 30% or a provision on packaging? One can only take refuge in noting that across-the-board observation rules give transparency to efforts at categorization.

4.1. Classifying Product-Specific Rules (PSR)

African PTAs have less complex PSR than other PTAs in the ROF database. They differ mostly along two dimensions. First, African PTAs do not use the change in chapter (CC), the most restrictive CTC. Also, they use fewer compounded rules (CH and VC/SP or CS and VC/SP) that add restrictiveness. Second, African PTAs rely more on the WO criterion. If WO is obvious for certain products (live animals etc…), because it can be restrictive for non-primary activities like food products, one can question the restrictiveness associated with the high prevalence of WO criteria compared with those on other PTAs.¹⁵ Even if less complex, differences in PSR across African PTAs are substantial. Reasons for these differences are beyond the scope of this paper but inspection of when these PSR came into effect reported in the notes to Table 1 shows that, initially, ROOs were rather across-the-board, then became progressively more complex over time with the addition of PSR.

ECOWAS and COMESA on the left of Table 1 appear less complex as they have opted for a CTC category for all products. For ECOWAS it is CH or VC. COMESA combines either CH or VC for 76% of the products, but also a recent annex replaces the CTH rule with some PSR. COMESA also offers a fifth option of RVC 25% for selected products. Yet, the notes to Table 1 show that there is diversity in this apparent across-the-board selection of PSR.

The remaining PTAs – GAFTA, AGADIR, EAC and SADC - show more disparity at two levels. First is the split between CTC and no-CTC rules. In this group, about 10% of PSR with only a WO requirement. Second, both within the CTC and no-CTC categories, the requirements are often compounded requirements rather than choice requirements. PSR with compounded requirements are automatically more restrictive than PSR with the same requirements in choice form. There is also some similarity between EAC and SADC rules while COMESA—also participating in harmonization negotiations for the TFTA—has simpler PSR. It is likely that COMESA will be under pressure to adopt the more complex and, often, more restrictive, PSR of its two comprising blocs, EAC and SADC.

As in Table 6 for regime-wide rules (RWR), regulatory similarity is greater than textual overlap for PSR in Table 7. However, the much larger number (148) of standardized codes capturing PSR requirements than those capturing RWR (30) limits the informative content of aggregate measures. Excluding ECOWAS and COMESA with PSR not shared by other PTAs, the similarity in rankings for the two measures is close, suggesting that both measures convey similar information. At the same time, the cases of COMESA and ECOWAS illustrate the limitations of these measures. For ECOWAS for which there is no overlap in PSR with other PTAs in Table 1, the regulatory similarity in Table 7 is automatically reported as zero; yet, through exceptions and allowances, there are some similarities with the other PTAs that are only partially reflected in the low textual similarity scores of Table 7. Likewise, the very low percentages of textual overlap for ECOWAS do not capture the extent of regulatory similarity that would be more apparent in a scatter plot of regulatory similarity and restrictiveness scores at a heading or subheading level.

4.2. Distribution of Restrictiveness Indexes across Product Categories and Margins

Under the assumption that traders know their alternatives, data utilization of preferences (purs) would be an approximation of the upper bound of the costs of compliance [see Equation (2a)]. In their absence, observation-based indices like the R-index that cover the entire range of PSR can serve as a proxy for the restrictiveness of PSR. These are used with other indicators in the dashboard of Section 5. As a prelude, here we explore their distribution across product categories and preferential margins.

Products traded internationally are increasingly complex. Due to trade costs having fallen drastically, the share of international trade in intermediate products has grown, often along value chains. Excluding the WO requirement for some primary products, it is difficult to assess if a traded product meets the ST requirement. This is a justification for setting PSR at the HS4 level to ascertain that enough transformation takes place at the different production stages. However, as documented by Erasmus et al. (2006) for SADC, ROO have also been used as a developmental tool to encourage producers to use regional materials.

Figure 2 summarizes the distribution of R-index values by product types (different categories of products overlap since all PSR lines are distributed across each classification) and by ranges of preferential margins. The large dispersion in values for each subcategory is evidence of the challenge facing negotiators, even though some dispersion comes from the construction of the index.¹⁶

Figure 2

Distribution of restrictiveness scores by product types and preferential margins.

Source: Author’s calculations from ROF. Notes: 1. 5387 HS6 product codes in six African PTAs (percentage distribution in parenthesis). Shaded area is inter-quartile range. The bar (I) [X] indicates the (median) [mean]. 2. Percentages in parentheses are shares of HS6 codes covered by each product category. For preferential margins (indicator 4), it is a normalized sum of shares of HS6 codes in six PTAs, as the preferential margin for the same HS6 code differs by PTA. 3. Product classification “by industry” (1) from WTO World Tariff Profiles 2019. “by processing stage” (3) from the WTO; “by technological intensity” (2) from OECD; “by end use” (1) from UN (BEC). For detailed explanations, see Klotz et al. (2016).

For all product classifications, the interquartile range in R-index values is large. When classified by end-use, average restrictiveness is higher for consumption goods, and lower for capital goods and intermediate goods. This would be expected if ROO are used as a developmental tool and production capabilities are low for these categories of goods although some high R-scores are attributed to some categories of intermediate goods. Except for the other category, dispersion by technology intensity is relatively low. By processing stage, raw materials have the highest values because of the prevalence of the WO criterion. For processed goods, the high-scores capture high RVC percentages.

The classification by range of preferential margin shows higher restrictiveness for goods with high preferential margins, although there are exceptions for goods with a preferential margin of <2%. This pattern was evidenced for the North American Free Trade Area (NAFTA) where generally, utilization rates were only marginally higher for the high preference margin sectors [Carrère and de Melo (2006)]. This suggests that tailor-made PSR are also prevalent across African PTAs. For the two lowest categories, compliance costs could exceed the preferential margin. While this margin is computed on applied rather than bound tariffs, negotiators might want to agree on doing away on origin requirements altogether for the 16% of goods with a preferential margin <2%.¹⁷