Tariffs and non-tariff frictions in the world wine trade

1. Introduction

A decade has passed since Kym Anderson studied the effects of globalisation on the world wine market: ‘Globalisation is not new to the world's wine markets, but its influence over the past decade or so has increased significantly’ (Anderson and Golin, 2004: 14). Indeed, the wine sector is increasingly becoming an export-oriented industry. Nowadays, almost half of world production is concentrated in three countries (France, Italy and Spain), which account for less than a third of global consumption, while the declining consumption in their domestic markets is pushing the industry to export to distant markets. The rapid and dynamic growth of the world wine trade and the rise of new exporters and importers have largely contributed to rendering wine one of the most globally consumed drinks (Anderson and Nelgen, 2011).

In recent decades, growth in the world wine trade has been driven by a number of complementary factors,¹ including technological improvements and policy interventions. The former have reduced transport costs, bringing distant countries ever closer, while the latter have aimed to reduce price mark-ups on imported goods. At the same time, new regulations have impacted trade flows. In general terms, sanitary and phytosanitary (SPS) measures and technical barriers to trade (TBTs) have been introduced to overcome technical regulations and correct market failures for facilitating trade and to guarantee high safety and technical standards (Mahé, 1997). However, a vast majority of scholars have argued that non-tariff barriers have simply been put in place to protect domestic industries from import competition (Yue and Beghin, 2009), raising international political concern (Disdier, Fontagné and Mimouni, 2008). Wine is seldom perceived by consumers as an essential good, and it is regulated by governments accordingly, as a source of additional revenue. Although the objective of reducing negative externalities from consuming alcohol is undoubtedly noble, the real, less noble, motivation for taxing wine consumption is simply revenue seeking (Fogarty, 2010). As a result, the rate of taxation on wine is high (Freebairn, 2010). Indeed, as repeatedly argued (Foster and Spencer, 2002: 14–19), the world wine market has been overregulated, and the effective level of protectionism has probably not changed at all. Addressing this specific question is an open empirical issue which is particularly relevant to policy-makers and entrepreneurs: identification of the protectionist nature of specific TBTs set by governments may in fact facilitate changes in existing TBTs in some countries or foster the implementation of harmonised standards internationally in order to improve wine trading and marketing.

An extensive literature on the effects of trade regulations analysed the role of tariffs and non-tariff barriers and their impact in economic terms, reporting contrasting results on the role of non-tariff barriers (TBTs and SPS standards). Leamer (1990) showed how trade barriers greatly reduce trade. Heien and Sims (2000) found that the removal of tariffs and non-tariff barriers due to the establishment of the Canada–United States Free Trade Agreement (FTA) increased trade flows from Canada to the USA by 10 per cent for tariffs, and by 17 per cent for non-tariff barriers. Similarly, Otsuki, Wilson, and Sewadeh (2001) demonstrated that the European standards on aflatoxins are the main barrier to imports of African groundnuts. Similar results have been reported by Henson and Loader (2001) and Peterson and Orden (2005, 2008). Jayasinghe, Beghin and Moschini (2010) analysed trade frictions in world demand for US corn seeds and concluded that ‘tariffs matter most, followed by distance and SPS measures' while Henry de Fraham and Vancauteren (2006) stated that the removal of TBTs and the harmonisation of regulations in the food industry have greatly increased intra-EU trade. Finally, Henson, Brouder and Mitullah (2000) stressed that also measures such as food safety requirements can equally act as barriers to trade.

In contrast, several scholars have reconsidered the negative effects of non-tariff barriers on trade flows. Harrigan (1993) showed that trade between organisation for economic co-operation and development members is limited by tariffs and transport costs rather than by non-tariff barriers. Similarly, Fontagné, Mimouni and Pasteels (2005) stated that SPS and TBTs in fresh and processed foods are not very restrictive, Cioffi, Santeramo and Vitale (2011) and Santeramo and Cioffi (2012) found that a price ceiling is not as effective as tariffs for imported products in the EU and Henson and Jaffee (2008) examined the concept of ‘standards as catalysts' in the context of food safety standards in international trade, highlighting the need for careful analysis when considering the trade effects of TBTs.²

What is becoming clear in the extensive debate animating the literature is that the effects of non-tariff measures on trade are likely to be sector- and country-specific depending on several aspects. For instance, Disdier, Fontagné and Mimouni (2008), in a study considering 30 disaggregated industries at the HS-2³ aggregation level, found that TBTs and SPS had positive effects for 8 industries, insignificant effects for 12 and negative effects for 10 industries. Similarly, Swinnen and Vandemoortele (2011) investigated the role of food standards on trade and showed that standards may be ‘barriers’ to trade but also ‘catalysts’ to trade. These results are also confirmed by Chevassus-Lozza et al. (2008).

All told, the net effects of TBTs and SPS standards are still unclear and merit further research. Indeed, what makes the picture even harder to fathom is the empirical difficulty in measuring and comparing the actual impact of tariffs and non-tariff barriers on trade. Yue, Beghin and Jensen (2006) propose two alternative methods, the price-wedge approach and the gravity-equation approach in order to assess the tariff equivalency of TBTs. By applying the latter approach, this study seeks to tackle the above challenge. In particular, we assess the equivalence of technical barriers and SPS measures with respect to tariffs by means of an econometric analysis. The novelty of this article arises from the detail with which the trade regulations are dissected and assessed for analysing their impact on global trade flows of wine. This type of analysis is especially useful for identifying what regulations (most) efficiently achieve protectionist goals. The study is conducted on the main exporters and importers in the world wine trade, analysing data on bilateral trade flows from 1997 to 2010.

The global nature of the wine sector, the complex set of existing tariff and non-tariff barriers and the increasing trade volumes are the key motivations for justifying a specific study of trade barriers in the world wine trade, representing an ideal framework to compare tariffs and non-tariff measures. These are the elements we focus on. The remainder of the article is organised as follows: the next section details the current situation of country-specific tariffs and non-tariff regulations; the third and fourth sections describe the methodology; and the fifth section describes the empirical results. We conclude by summing up the results and discussing empirical and policy implications.

2. Trade frictions in the wine market

In recent decades the global wine trade has experienced major changes. First, since the 1980s the traditional producers (France, Italy and Spain) have witnessed a conspicuous fall in domestic consumption that has driven exports. Second, the countries in the NewWorld have increased their production potential to satisfy the new demand in foreign markets (Cembalo, Caracciolo and Pomarici, 2014). These changes have been accompanied by a geographical redistribution of wine consumption (Aizenman and Brooks, 2008), especially an increase in wine consumption in North America and Asia.

During the last 15 years, the New World countries have quadrupled their exported quantities, now accounting for 33 per cent of the world's 11 major wine producers exports.⁴ Despite the relative increase in market shares, in absolute terms the export gap that separates the main European producers from countries in the New World has increased from 68 to 107 million hectolitres. Overall the volumes of wine exported have doubled in the last 15 years, topping ten million litres in 2012. Nowadays almost half of the global wine is consumed outside the country of production (Figure 1). However, this rarely comes without an extra cost. The sections below provide a cursory review of these regulations.

2.1. Tariffs

In this study, tariff barriers for bottled wines were calculated by taking into account most favoured nation (MFN) tariffs, chosen as reference for national tariff levels, then corrected according to FTAs, if any. If the importer applied specific tariffs, these were converted into ad valorem equivalent (AVE), utilising the average import price as reference. AVE calculations were performed to aggregate the different HS codes⁵ in the same country and to compare tariffs across different countries.⁶ For importers implementing several duty lines, the median of the duties is considered here. This approach allows a central value to be obtained at the level of HS codes with eight digits and over, minimising the influence of tariff peaks (Anderson and Neary, 2003; Bouët et al., 2008; Cipollina and Salvatici, 2008).

Table 1 presents a rough picture of trade openness of the principal actors for the global trade in bottled wine in 1999 and 2010.⁷ The second and third columns show the ranking of importers and exporters of bottled wine (in value). The fourth column reports the MFN tariffs of selected countries. The fifth column shows the number of FTAs signed by country. The sixth column (referred to as the ‘effective barrier’) describes the real duty considering the reduction in the MFN tariff due to FTAs.⁸

As regards the MFN tariffs, a clear differentiation may be noted on the basis of geographical area: the countries of North America have the lowest protection of the domestic market through tariffs, while Latin American and Asian countries set very high import duties. Thanks to the world trade organization (WTO) policies aimed at encouraging free trade, the countries in question have reduced MFN tariffs applied to imported bottled wine. Nevertheless, in the particular case of the European Union, the calculated AVE indicates the contrary; this apparent contradiction is due to the fact that the EU applies specific tariffs, such that their reduction has been followed by a more than proportional reduction in the unit value of imported wine.⁹ Consequently, the percentage of the duty calculated on the basis of the unit value of the imported product has increased.

For the selected countries the (export-weighted) average decrease in the MFN tariff from 1999 to 2010 was 22.7 per cent, driven by China and Canada which experienced a more than fourfold decrease, and by Chile that reduced its MFN tariff by 40 per cent. China applied the main reduction of the tariff in both absolute and relative terms. This reduction was undoubtedly after its accession to the WTO in 2001 falling from 65 per cent to 14 per cent in only 3 years. In contrast, Canada and Chile reduced their tariffs in order to open up to the international market: indeed, Chile and Canada have stipulated a large number of FTAs. Further considerations can be made by analysing the effective protection (last column of Table 1). While in 1999 there was almost no difference between MFN tariffs and the ‘effective barrier’, in 2010 all the nations considered¹⁰ show differences between the MFN tariff and the tariff then actually applied. An emblematic case is Chile: it has stipulated numerous FTAs such that in 2010 the average tariff applied on imported wines was 90 per cent lower than the MFN tariff.

As shown in Table 2, bilateral tariffs and MFN tariffs have been largely revised during the last 15 years. Moreover, while some countries have established high MFN tariffs (above 14 per cent), others have adopted lower tariffs (below 6 per cent). The former group of countries (namely Japan, Argentina and China) have established only one FTA in favour of Chile, while, more generally, countries with low MFN tariffs tend to set a higher number of FTAs. All these give sufficient reasons for analysing these dynamics in order to quantify to what extent tariffs may slacken trade flows.

2.2. Non-tariff barriers

As tariffs have been lowered, demands for protectionism have led to new technical barriers. Among these, SPS standards and TBTs are of increasing importance (Moenius, 2004; Disdier, Fontagné and Mimouni, 2008). Several definitions exist in the literature for non-tariff barriers: Hillman (1991: 52) described TBTs as ‘any governmental device or practice other than a tariff which directly impedes the entry of imports into a country and which discriminates against imports, but does not apply with equal force on domestic production or distribution’. More recently, Beghin (2008) and Li and Beghin (2011) defined TBTs as the wide and heterogeneous set of intervention measures, other than custom duties, that influence and distort the commerce of goods, services and production factors. Attention is currently focused on the distinction between measures that have the primary aim of protecting national products and others that do not. Following this approach, the definition of a TBT is linked to its legitimacy: the term ‘barrier’ should not be used if the measure has a collateral effect on trade, but its primary aim is to correct some market inefficiency. However, establishing impartially whether a standard has a legitimate basis is no simple task (Fisher and Serra, 2000; Maskus, Wilson and Otsuki, 2000).

In the wine sector there has been a growing use of TBTs by governments in order to protect domestic markets (Anderson and Golin, 2004). As shown in Table 3, the level of intervention is very heterogeneous across countries. Indeed, the number of TBTs set by each country at the end of 2010 varied from 169 for Australia to 811 for the USA. Three cases worth further note. First, Argentina, one of the last countries to implement TBTs, issued a total of 24 TBTs just in few years.¹¹ Second, the USA and the EU have set, respectively, 19 and 10 TBTs during the last few decades. After joining the WTO in 2002, China was forced to lower its import duties, balancing the loss in level of protectionism through TBTs. In particular, China has produced new notifications at regular intervals. The other countries have made less use of these instruments: Australia has produced six TBTs, Chile four, Japan three and Canada only two. Some nations stand out for the specificity with which they have regulated the wine sector: for example the EU have issued 8 out of the 10 TBTs for products belonging to code HS2204 (wine from fresh grapes); for Argentina 15 out of 24 TBTs are specific to wine and for USA 8 out of 19. The other importing countries have not shown such zeal in meticulously regulating the sector, and wine is generally included in notifications that regulate a wider range of food products or beverages.

Currently, the WTO comprises the TBTs pertinent to wine in six categories on the basis of what they regulate: food standards, labelling, conformity assessment, packaging, food containers and human health. According to this classification, labelling is the only category used at least once by all the countries concerned. Labelling regulations refer to the requirements imposed to attest the presence of substances that might cause allergic reactions,¹² to protect specific names,¹³ to regulate the wording regarding the country or designation of origin¹⁴ and to ensure and regulate the presence of obligatory information such as brand, alcohol content and vintage.¹⁵ Some TBTs have also been inserted to discourage the consumption of alcoholic beverages, including wine, covering the impossibility of mentioning any health benefits from wine, the obligation for warnings about possible repercussions on human health and sometimes to regulate sales to young people.¹⁶ ‘Food standards’, the second most frequent category, set technical requisites of the product, regulating oenological practices in general, and the maximum contents of particular substances.¹⁷ Some TBTs that regulate sectors are generally not strict regulations in the case of wine: this particularly holds for regulations belonging to the classes packaging,¹⁸ food containers,¹⁹ conformity assessment²⁰ and human health.²¹ Such TBTs are rarely adopted and are often very country-specific.

SPS measures generally concern the fresh product trade and there is almost no recourse to such measures for the wine sector. Notably, only Argentina, Australia, China, the USA and the EU have produced SPS notifications (Table 4). Argentina adopted SPS measures to set the maximum limit of lead, arsenic and zinc in wine.²² China, by means of SPS/N/CHN/P/133 of April 2002 which regulates all alcoholic beverages, has formalised the requisites, supervision and inspection procedures of alcoholic beverages at ports and on the domestic market. The USA adopted the sanitary barrier G/SPS/N/USA/196 in November 1999, which regulates the labelling of alcoholic beverages by prohibiting, on labels or in advertising, any claim regarding the health benefits deriving from the consumption of any alcoholic beverage, unless the statement is qualified, objective, sufficiently detailed and specific. Lastly, the EU issued G/SPS/N/EEC/247 in 2004 which sets the maximum limit for ochratoxin A in different foods, including wine. Our empirical analysis will explore the restrictiveness of these regulations, calculating the tariff equivalency as a measure of their impact. The methodology adopted to achieve these goals is described in the next section.

3. Empirical framework

In terms of data sources, it was deemed appropriate to include both the largest world wine producers and the main importing nations. The set of countries chosen covers more than two-thirds of the world imports and almost 90 per cent of the global trade in bottled wine for the period 1997–2010. More specifically, we considered trade among Argentina, Australia, Canada, Chile, China, France, Germany, Italy, Japan, Spain, the UK and the USA. Data were collected from the Global Trade Atlas (GTA) database²⁵ for export values, the CEPII database²⁶ for distance, WTO, WITS²⁷ and from national customs offices for tariffs and non-tariff barriers as described in the previous section. Supply data were collected from StatOIV Extracts. Data on tariffs were obtained from the WTO official database which reports the MFN tariffs applied by importers. See Section 4 on how we computed the AVE tariffs. For non-tariff barriers the most widely adopted trade regulations are considered (Maskus, Wilson and Otsuki, 2000). Information to model count variables on the six classes of TBTs and on SPS standards was extracted from the ‘Technical Barriers to Trade Information Management System’ database and the ‘SPS Information Management System’ database, respectively.²⁸

To obtain the maximum level of detail, all the documents issued from 1995 (i.e. from the creation of the WTO) to the end of 2010 were examined in order to identify those regarding wine and that are likely to originate extra costs. Table 5 shows the definition and the descriptive statistics of data on which the estimates are based.

4. Methodology for calculating tariff barriers

For the analysis of tariff barriers related to the exporting of products subject to code HS 220421, reference was made to the official WTO database, considering the MFN²⁹ tariff of the various importing countries for the period 1995–2010. The accuracy of the data was supported by a crosscheck with the duties reported in the WITS database, and with those of the respective national customs authorities, where possible. This procedure compensated for the initial lack of data on Chinese tariffs for 1998, 1999, 2000, 2009 and 2010 and made a clearer distinction on the basis of the six-digit HS code for the European Union. No data interpolation was therefore necessary nor the use of ‘extreme’ values to complete the dataset.

The MFN tariff was taken as the base value indicative of the duty imposed by each importing country on all the other WTO members, but the various preferential agreements (FTAs) already in force were also considered, as well as those instituted during the period considered. When there was a tariff deriving from an FTA for a given year and towards a given exporting country, this was substituted for the MFN. For the calculation of the tariffs deriving from preferential agreements reliance was initially placed on the WITS and WTO databases. Yet, following a more thorough revision, it emerged that they were partially inaccurate due to the lack of various preferential agreements, errors in the duties indicated and sometimes in the year they came into force. For this reason, the Sistema de Información Sobre Comercio Exterior (SICE) portal³⁰ was used for the analysis of all the official documents of the FTAs signed between the various countries, thus improving the information inserted in the dataset.

The tariffs were inserted in the dataset as AVE data, representative of the average import duty imposed on products subject to the code HS 220421. This operation overcame two methodological problems: on the one hand, the tariff profiles had to be aggregated where the code HS6 included tariffs (HS8 or lower) with different duties. On the other, when the tariffs did not directly affect the value of the imported good, but were instead expressed as specific tariffs, a transformation into AVE data became necessary by dividing the duty by the value of the imported good.

The first step was therefore the transformation of all the specific tariffs into AVE data, at the level of codes HS6 or HS8 when present. The choice of the reference unit value of the imported good was tackled in different ways in the past. The value of specific imports of the State in question is usually utilised for this operation, which has been shown to be entirely valid. It allows a qualitative distinction to be made of the imported goods, but suffers from estimation errors and is often not significant where trade is limited. To overcome these problems Gibson (2001) proposed the use of the import world as reference. This provides more robust data but does not permit any qualitative distinction to be made of the imports. Bouët et al. (2008) further developed this methodology by proposing the use of the average importation value of a group of reference countries, i.e. a set of countries with similar characteristics. This approach is classified midway between the two previously described and has been shown to be robust and able to partly take into account the qualitative differences. However, given that wine is a widely differentiated product, and that to minimise the estimation errors the dataset was constructed by inserting the major exporters and importers, in this study we preferred to use the value of specific imports of the State as reference.

The specific tariffs were therefore transformed into AVE using the following methodology: In response to the second methodological problem, i.e. aggregation of the different tariffs in code HS6, the different methods indicated include the simple average, weighted average and median. The simple average has been criticised because the tariffs have an irregular distribution (Cipollina and Salvatici, 2008), and it has no theoretical basis (Anderson and Neary, 2003; Bouët et al., 2008). The weighted average is instead valid but tends to underestimate the effect of high tariffs (Bouët et al., 2008), as high duties generally lead to a reduction in the quantities imported. Because the median is generally indicated as more robust and reliable, it was used for the aggregation of the tariffs in this article.

1. where the tariff was expressed on the basis of alcohol content (e.g. EUR 10/per cent vol/hl), an alcohol content of 12 per cent on volume was adopted, which can be considered the average for wine produced on a global basis, as suggested by the WHO;

2. where the tariff was a fixed sum on the volume imported (e.g. EUR 1/L), the average importing price from each exporter country in the respective year was used as reference for the conversion;

3. where the tariff included a fixed sum on the volume to which a variable quota was added based on alcohol content (e.g. EUR 1/L + EUR 1/per cent vol/hl), the average price of specific imports per exporting country in the respective year was used as reference for the fixed sum, to which the variable quota was added, calculated on the basis of an alcohol content of 12 per cent;

4. where the tariff included a percentage on the value plus a quota based on volume/alcohol content (e.g. 15 per cent on the value + EUR 1/L, or 15 per cent on the value + EUR 1/per cent vol/hl), this quota was transformed as indicated in points 1 and 2, and then added to the ad valorem tariff;

5. where the tariff included the lowest or highest level among the different possible options (e.g. 15 per cent on the value or 125 yen/L), the two options were calculated with the methodology of the preceding points, and then choosing the one indicated by the regulation in force;

6. where the importing country imposed a maximum and/or minimum limit to the tariff, the average tariff calculated by each exporter and each year was maintained if within the set limits. Otherwise the set minimum/maximum limit was adopted as duty.

The calculation method was used for each importing nation towards all the exporters and all years, to obtain as clear and detailed a picture as possible on the evolution of customs duty during the considered period. In order to obtain the maximum accuracy of the data the official values were used, expressed in the local currency of the various customs authorities, hence without any conversion into US dollars.

5. Model results and discussion

The results for different estimation methods are reported in Table 6. We considered four alternative estimation methods. The first column reports OLS estimates in log form. By adding a constant to the dependent variable we were also able to estimate the model for observations with zero bilateral trade. The second column presents Tobit estimates, based on the Eaton and Tamura (1995) approach. The third and fourth columns report Heckman and PPML estimates, respectively. Exporters' fixed effects are included to account for unobserved heterogeneity (Cardamome, 2011).

While the estimated coefficients from the OLS, Tobit and Heckman models are quite similar, most coefficients obtained from the PPML model differ significantly from those obtained with other models (except for the parameter estimates of GDP and Language). This suggests that heteroskedasticity (rather than truncation) is responsible for the differences between PPML and the other models (Silva and Tenreyro, 2006).

P-values from the heteroskedasticity-robust RESET test (Ramsey, 1969) on OLS, Tobit and Heckman models suggest that the null-hypothesis of misspecification should not be rejected, unlike the PPML model. We conclude that the PPML model is to be preferred.³¹ A further advantage of the PPML model is that it allows us to deal with sample selection bias that may result from excluding zero observations. Although selection bias rarely affects the sign of the variable, it often influences the magnitude, statistical significance and economic interpretation of the marginal effects (Haq, Meilke and Cranfield, 2013). In the rest of this section, unless specified otherwise, we refer to the estimates from the PPML model.

Parameter estimates of GDP and distance are statistically significant and have the expected signs: positive and negative, respectively. The estimated elasticity for GDP is close to 1.4 in all specifications. This result is supported by Silva and Tenreyro (2006): they argue that the coefficients on importer's and exporter's GDPs tend to be higher than one. The role of geographical distance is significantly larger when using OLS and Heckman estimators. The estimated elasticity is negative and close to 0.4, whereas the PPML estimate is much lower (−0.11). Compared with the literature on the gravity model (Disdier and Head, 2008), the physical distance has a limited impact on the global wine trade. This result is hardly surprising because exported wine is highly priced and has relatively long storability. In addition, as underlined by Silva and Tenreyro (2006), OLS estimation exaggerates the role of geographical proximity. We conclude that transport costs have a limited role in determining trade patterns. Our explanation for this result is that product differentiation plays an important role in the bottled wine trade because imported wines are imperfect substitutes, distant importers do not replace imports from distant markets with wines sold by closer partners.

The results concerning wine supply also deserve particular attention. Wine production in country i (Y_1i) is statistically significant (at the 5 per cent level) and positive and this captures the stimulus of domestic supply on exports. In contrast, the production in country j (importer) is negative and significant at the 1 per cent level: we conclude that home bias induces trade resistance. The elasticity of the former variable (0.475) is more than twice that of the latter (−0.198). ‘Language’ is statistically significant: its impact is large and positive (1.204). These results are in agreement with previous studies (Disdier, Fontagné and Mimouni, 2008; Seccia, Carlucci and Santeramo, 2009; Grant and Boys, 2012; Kandilov and Grennes, 2012).

The estimated coefficients of ‘tariffs’ (t_j) are negative in all specifications. The PPML estimate indicates an elasticity of trade to tariffs of −0.472: a 1 per cent increase in tariffs would reduce trade by 0.47 per cent ceteris paribus. Frictions from restrictive technical barriers are effective: this result is in line with those of several authors (Heien and Sims, 2000; Olper and Raimondi, 2008; Liu and Yue, 2009). In particular, we found that SPS measures do not inhibit trade, while technical barriers are considerable frictions to exports. The estimated coefficients of country-specific technical barriers, if statistically significant, are negative. The estimated coefficients of ‘food containers' are statistically not significant. We argue that such a barrier is non-prohibitive. ‘Human health’ and ‘conformity assessment’ are statistically significant at 5 and 1 per cent, respectively. ‘Labelling’ is also statistically significant (at 10 per cent) and negative. Barriers due to ‘food standards’, despite being widely adopted, do not seem to be prohibitive: in all specifications the estimated coefficients are statistically not significant. Our interpretation for this result is that, while food standards on wine are motivated by food safety arguments to protect consumers, modern techniques and innovations in the wine industry allow international standards to be easily satisfied.

The results on equivalent tariffs for technical barriers (AVE-TBTs) are reported in Table 7. The estimates represent the change in tariff that would be equivalent to the imposition of TBTs. Moreover, we evaluated the actual impact of both tariffs and non-tariff barriers on the world wine trade by computing the marginal effect of trade frictions and the aggregate impact on exports.³²

‘Conformity assessment’ and ‘human health’ are equivalent to 2.84 and 2.21 tariffs, respectively. Thus we can argue that the barriers are prohibitive given that the weighted average tariffs in 2010 were close to 5.04.³³ Being country-specific, these technical barriers tend to be very prohibitive as they raise the average transaction costs incurred by exporters.

The tariff equivalent for ‘packaging’, adopted only by the USA, is assessed as being close to unity. Several considerations have to be made on this country. First, the USA system of protectionism is complex and includes a variety of technical barriers such that the marginal contribution of each measure is relatively low. Moreover, the USA being the main world market can exert market power in order to protect its growing domestic market (e.g. Californian supply). Figure 2 shows the dynamics of the total trade frictions per year, as composed by the two components, equivalent AVE for technical barriers (AVE-TBTs) and AVE of tariffs. It is worth noting that while AVE tariffs show a declining trend over the years, and AVE-TBTs an increasing trend, the total frictions pattern is quite steady and does not change significantly during the period 1997–2010 (weighting the measures according to actual trade flows). The result provides statistical and quantitative evidence of the real immutability of overall friction in the world wine trade. Moreover, Figure 3 breaks the total impact of the TBTs down into the different and prohibitive components, providing a measure of the frictions in terms of real export values. While ‘conformity assessment’ and ‘human health’ TBTs showed the greatest marginal impact on trade, they only have a potential and not a real impact on trade given their low diffusion to date. In contrast, notifications on ‘labelling’ have globally been the main non-tariff friction in world wine trade, matching in 2010 the impact of tariff barriers, although the AVE-TBT per notification (0.45) is the lowest among the restrictive TBTs.

6. Conclusions

The expansion of the wine trade over the last decade is due to many factors which include the Uruguay Round/WTO agreements that have led to a progressive lowering of the import duties on agricultural products. As reported by Anderson and Golin (2004), several countries have attempted to maintain some level of protection of their domestic market by stipulating technical requisites that the imported products must satisfy. These new instruments have spread rapidly, as testified by the WTO notifications issued, and have animated academic debates that are not just confined to the literature of agricultural economics, but have a much wider audience (see Henson and Jaffe, 2008; Kee, Nicita and Olarreaga, 2009; Jacks, Meissner and Novy, 2011, among others). Our analysis shows how the principal wine-importing countries are behaving, and in particular how tariffs and non-tariff regulations have impacted trade flows over the period 1997–2010. An extended version of the gravity model was applied to a panel dataset which includes data on 90 per cent of global trade of bottled wine, and detailed information on technical barriers and SPS measures. In order to focus on the economic aspects and avoid cumbersome notation, the econometric formulation was admittedly simple, while the estimates were carried out through the PPML estimator.

The study involved two innovations. First, this is one of the few works where the impact of non-tariff measures on a specific sector has been carefully evaluated and compared type by type to ad valorem tariffs. Second, assessing the impact of non-tariff barriers by using count data is a rather novel way to quantify the impacts of regulations. The main limitation in adopting count data relies on the excessive costs incurred in gathering country- and line-specific regulations. By focusing on a single-specific product we were able to overcome this limitation.

All in all, the econometric estimation of the gravity equation shows that tariff and non-tariff barriers do matter considerably in the world wine trade. Our findings are in line with previous studies on trade costs (Jayasinghe, Beghin and Moschini, 2010) and the wine trade (Raimondi and Olper, 2011). Jayasinghe, Beghin and Moschini (2010) found that world demand for corn seeds is mainly inhibited by tariffs and distance, and only in a limited way by SPS measures. We found similar evidence for the world wine trade. Our results are also in line with those provided by Kee, Nicita and Olarreaga (2009) who estimate, for agricultural regulations, an average AVE-TBTs of 20 per cent, which for our estimates would represent a top boundary.

However, not all non-tariff barriers are equal: our in-depth analysis shows that, in some cases, TBTs are neither binding nor catalysts for trade. In particular, some trade regulations were assessed to be as stringent as 3 per cent of ad valorem tariffs, while others including SPS measures are totally negligible. This is an important result as it may justify the removal of inefficient technical standards on imported wine, whilst suggesting a target to aim at when seeking to liberalise the international wine trade. Furthermore, for policy-makers who are often interested in the impact of standards and regulations on international trade and competitiveness, the current analysis provides tariff equivalencies that may help them gain insights and negotiate trade agreements to enhance wine trade flows. Finally, we show that the decreasing trend for tariffs has for the most part been compensated by more stringent technical barriers. The overall result, foreseen by Anderson and Golin (2004) 10 years ago, is that frictions in the world wine trade have not changed during the past 15 years.

Our analysis is not exempt from potential improvements. First, our dataset includes only trade observed among 12 countries during the period 1997–2010. Although we capture more than 90 per cent of world wine exports and two-thirds of global import flows, our results cannot explain trade patterns among small traders. Second, our comparative analysis of tariffs and non-tariff barriers, challenged by the complex regulations in force, might not be entirely satisfactory and could invite further research. Indeed, empirical work on technical regulations in world trade represents a promising area of research.

Acknowledgements

The authors thank Alessandro Olper and two anonymous referees whose comments have greatly improved this article, and Mark Walters for editing the manuscript.

References