The Role of Private Standards for Manufactured Food Exports from Developing Countries

World Development 101 (2018) 16–27

Contents lists available at ScienceDirect

World Development journal homepage: www.elsevier.com/locate/worlddev

The Role of Private Standards for Manufactured Food Exports from Developing Countries Malte Ehrich a,⇑, Axel Mangelsdorf b,c a

University of Göttingen, Germany Technical University Berlin, Germany c VDI/VDE Innovation + Technik GmbH, Berlin, Germany b

a r t i c l e

i n f o

Article history: Accepted 14 August 2017

Key words: agricultural trade private food standards manufactured food gravity model

s u m m a r y The relevance of non-tariff barriers for global trade flows has increased in recent decades. However, the effect of food standards—as a particular important non-tariff measure—on agricultural trade flows remains unclear. We contribute to the debate with a unique dataset [email protected]. dehat contains the number of food processing firms of 87 countries from 2008 to 2013 that are certified with the International Featured Standard (IFS). We estimate a gravity model using the one-year lag of IFS as well as IFS certification in neighboring countries as an instrument to address potential endogeneity. We find that IFS increases c.p. bilateral exports on average of seven agricultural product categories in both specifications. However, the effect remains robust only for high- and middle-income countries and disappears for low-income countries. Hence, while IFS increases exports on average, low-income countries do not benefit in terms of higher export volumes. Moreover, once we separate the dataset by sector, the trade-enhancing effect remains for bakery, dairy, and beverage sectors only. Overall, we argue that food standards are not a suitable development tool to integrate low-income countries into high-value chains per se. Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction Significant tariff reductions during previous decades belong to the most successful tools to reduce poverty (Dollar & Kraay, 2004). Many South-East Asian countries integrated into the world trade system and achieved tremendous increases in per capita income. However, not all countries benefit from the world trade system in the same way despite tariff reductions which were achieved via multilateral as well as via regional negotiation rounds. Moreover, as the relevance of tariffs as trade barriers declines, nontariff barriers (NTBs) to trade gain in quantitative as well as in qualitative importance. For example, the total amount of sanitary and phytosanitary (SPS) notifications to the World Trade Organization (WTO) as a proxy for public food safety standards increased from less than 200 in 1995 to almost 1,000 in 2015, see Figure 1. Moreover, the number of GlobalGAP producers as an important private standard increased from below 20,000 in 2004 to more than 150,000 in 2015 (GlobalGAP, 0000; Swinnen, Deconinck, Vandemoortele, & Vandeplas, 2015). The increasing relevance of standards is important for trade policies because standards usually imply significant costs of compliance which could prevent low⇑ Corresponding author. http://dx.doi.org/10.1016/j.worlddev.2017.08.004 0305-750X/Ó 2017 Elsevier Ltd. All rights reserved.

income countries in particular to benefit from global agricultural markets. Therefore, the effect of NTBs and standards in particular on trade is of deep interest for economists and policy makers that are concerned about the integration of developing countries into the world trade system (Otsuki, Wilson, & Maskus, 1999). This study addresses these concerns by looking at the effect of the International Featured Standard (IFS) as an important private food standard on agricultural trade. We employ a unique dataset which was obtained via the IFS-auditing database. It contains more than 50,000 audits from about 12,000 companies in 87 countries for seven agricultural sectors including a time-span of six years from 2008 to 2013. Second, we apply a novel instrumental variable approach which we consider to be superior compared to the standard method of taking the one-year lag which is not appropriate if the errors are autocorrelated. Thirdly, we estimate a gravity model via poisson-pseudo-maximum likelihood (PPML) which accounts for high share of zeros and heteroskedasticity (Santos Silva & Tenreyro, 2006, 2011). We apply the Baier–Bergstrand method to address multilateral resistance (Anderson & van Wincoop, 2003; Baier & Bergstrand, 2010). This approach allows us to contribute to the debate whether standards act as barriers or catalyst to trade. We find that IFS certification as a private standard increases bilateral trade flows in general which illustrates the trade-

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27

17

Figure 1. The raising relevance of standards.

increasing potential of IFS. However, the effect remains robust for high-income countries only and disappears for low-income countries once we separate by income. This finding has important policy implications. Although IFS certification increases trade on average, only high-income countries benefit in terms of larger trade volumes. This finding reduces the potential of food standards as a development tool to integrate developing countries into the world trade system. Furthermore, we show that the effect of IFS certification differs by sector. The trade-enhancing effect remains robust only for bakery, dairy, and beverage sectors. The remainder of the paper is structured as follows: Section 2 provides an overview of the current research status within the field of standards and trade. Because IFS is analyzed rarely, we devote an entire Section 3 to provide the background of this specific food standard. Section 4 explains the PPML estimation and the instrumental variable approach in particular including the control–function approach. Section 5 shows the results, which are discussed within the research context in Section 6. 2. What do we know about the effect of standards on trade? A debate entitled as ‘‘standards-as-catalyst vs. standards-asbarriers to trade” (Jaffee & Henson, 2004) emerged which—as a result—accumulated a large set of studies. Standards can either protect consumers or domestic producers. In addition, standards can either enhance or reduce trade flows. On the one hand, standards are likely to reduce trade because of high fixed costs of compliance which affect small-scale producers in particular (Herzfeld, Drescher, & Grebitus, 2011). For example, Czubala, Shepherd, and Wilson (2009) find that average compliance costs with product standards as percentage of firm sales exceed 100%. Other nonfinancial obstacles like financial literacy are also found to constraint farmers to adopt standards (Müller & Theuvsen, 2015). On the other hand, food standards can enhance trade by reducing information asymmetries (Henson & Jaffee, 2008). The westernization of diets as well as the increasing awareness of modern consumers regarding food safety makes transparent and safe food production processes quasi-mandatory for producers. Furthermore, food standards allow producers of developing countries to enter high-value chains. Private food standards in particular allow them to signal and prove high product quality. Thus, standards potentially reduce market failures due to information asymmetries which might be more relevant for developing countries (Jaffee & Henson, 2004). If private food standards are found to increase exports of developing countries in particular, this would have important policy implications. In addition to the povertyreducing effect due to larger trade volumes, food standards would facilitate equal access to global agricultural export markets. The

latter is important from a global perspective since it improves the functioning and benefits of global markets for all participants. Moreover, there are potential additional benefits at the firm level. Since trade is not only welfare-enhancing via lower consumer prices, export sectors are on average also the most competitive sectors in a country. Thus, exporting firms earn on average higher profits, employ a larger number of workers, and pay higher wages than non-exporting firms worldwide (Mayer & Ottaviano, 2007). For example Colen, Maertens, and Swinnen (2012) provide empirical support that this pattern occurs in developing countries as well. In the context of GlobalGAP certification in Senegal, the authors show that exporting firms are important drivers for job creation and productivity spillovers which underlines the potential of private food standards as a development policy tool. Empirical research results investigating the effect of food standards on agricultural trade highly depend on the corresponding context like the set of analyzed products, the set of countries, empirical method, and the type of food standard. For example, maximum residue limits (MRLs) as an important public food standard are more often—but not exclusively—found to be trade reducing than other standards (Li & Beghin, 2012; Otsuki, Wilson, & Sewadeh, 2001). The relevance of the chosen method is also highlighted by Ferro, Wilson, and Otsuki (2015) who create a restrictiveness index of MRLs for 61 importing countries. By applying the two-step Heckman procedure as illustrated by Helpman, Melitz, and Rubinstein (2008), the authors find evidence in the first stage that more stringent MRLs reduce the probability to export due to higher fixed costs. However, once the sample selection bias and the share of exporting firms are controlled for, standards have no effect on trade flows. In addition, the first-stage effect is stronger for the BRIC countries than for non-BRIC countries. Exports from low-income countries are more negatively affected by product standards than those from higher income countries. Ceteris paribus, countries export to destination markets which have the lowest fixed costs, i.e. less restrictive MRL standards. The effect of food safety standards on China’s exports is also analyzed by Chen (2008) who finds a statistically significant negative effect. According to his estimates, the effect is even stronger than imposing tariffs. Further evidence for trade-reducing effects due to more restrictive standards is—among others—also provided by Chen, Otsuki, and Wilson (2006), Yue, Kuang, Sun, Wu, and Xu (2010), Drogué and DeMaria (2012), Melo, Engler, Nahuehual, Cofre, and Barrena (2014) who all focus on the effects of MRLs on exports. Wilson, Otsuki, and Majumdsar (2003) also find that standards affect trade flows negatively, and the authors provide further evidence that the harmonization of standards enhances trade. The article that—among others—continues this debate was written by Anders and Caswell, 2009. They argue that the introduction of

18

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27

the Hazard Analysis Critical Control Points (HACCP) food safety standard increases trade exports of leading exporters of seafood and reduces exports of countries with lower exports of seafood. Moreover, developing countries are more likely to experience lower exports as a response to stricter standards than developed countries. Whereas public standards are set by public authorities and are usually mandatory and legally enforceable, private standards are set by the private sector with a wider scope than only food safety (Schuster & Maertens, 2015). As a result, the effect of private standards on trade is likely to differ from the effect of public standards. Whereas public standards are usually mandatory, private standards are de-jure voluntary but often de facto mandatory (Henson & Humphrey, 2010). However, integrating small-scale farmers of developing countries into global food export markets requires them to meet private food standards as well which are—such as the IFS—often set by large retailers in developed countries. This makes private food standards quasi-mandatory and therefore, an important subject to analyze. Because of the private nature, data for private standards are more difficult to obtain than for public standards which are often publicly available. As a result, private standards are less frequently analyzed. Using firm-level data of the Peruvian asparagus sector Schuster and Maertens (2015) cannot confirm that BRC, IFS, and other private standards act as catalyst to trade. Although private standards are on average more stringent than public standards (Fulponi, 2006), these have the potential to increase agricultural trade nevertheless. Masood and Brümmer (2014) find that GlobalGAP certification increases banana imports of the European Union. The trade-enhancing effect of GlobalGAP certification is also found by Colen et al., 2012 for mango and bean producers in Senegal which have larger export market shares and larger export volumes than non-certified firms. The differential effect of voluntary private standards compared to public standards on trade is also emphasized by Shepherd and Wilson (2013) who find that EUharmonized standards, that are equivalent to ISO norms, can even enhance trade. Eventually, Mangelsdorf, Portugal-Perez, and Wilson (2012) estimate the effect of Chinese public and private standards and also find a trade-enhancing effect which was most pronounced for internationally harmonized standards. Overall, few studies exist that analyze the effect of private food standards on agricultural trade. Studies use either cross-sectional data (Latouche & Chevassus-Lozza, 2015) or are based on data which are limited to specific cases. This certainly questions external validity as emphasized by Beghin, Maertens, and Swinnen (2015). The main challenges are first, the quantification of private standards and hence, data availability. Most studies do not allow to draw general conclusions because they are based on very few products and countries. Second, endogeneity arises as a result of reverse causality. A correct identification of the causal impact requires to distinguish whether it is certification that enhances trade or whether trade increases the likelihood of certification. And thirdly, a correct specification of the empirical framework requires to account for recent developments in the field of gravity modeling which became the workhorse model in empirical trade analysis (Head & Mayer, 2014).

3. IFS background and trends The increasing complexity of agricultural value chains due to fragmentation and specialization increases the necessity for sufficient transparency within value chains. Retailers need to guarantee quality and food safety of the products that they sell, but which they do not produce themselves. Moreover, to ensure the enforcement of legal contracts it is crucial to have transparent responsibil-

ities at every stage within a value chain. Therefore, the association of the German retail sector HDE1 found together with the French counterpart FCD2 the initially named International Food Standard in 2003. The IFS is applicable at every stage of a value chain apart from agricultural raw products. This private standard—today the International Featured Standard—avoids that each retailer is required to test whether their suppliers meet the imposed standards or not. Instead, retailers agreed on the same standards. These standards are continuously modified in collaboration with the retail sector. Hereby, most regulations go beyond usual food safety standards (International featured standard, 2016). The overall objectives are twofold: first, IFS ensures comparability, transparency, and quality for the consumer within a complete value chain. Second, it aims to reduce costs for the retail sector and their suppliers by harmonizing standards. Apart from the UK, where the BRC is the most relevant food standard certification body, all major retailers within Europe are member of the IFS3 which also certifies in other fields such as logistics for example. IFS does not certify products and food manufactures directly but rather via third-party certification which takes place on average once a year. All retailers that accept the IFS have access to these audit reports of their suppliers via an online database. In addition, all certified producers have access as well. But apart, access to the database and information concerning audit reports and other confidential data is not possible. Moreover, it is not only the availability of IFS data that makes our analysis distinct from previous studies. As a post-farm gate standard, which needs to be distinguished from pre-farm gate standards such as GlobalGAP that certify agricultural raw products, IFS certifies processed food. These manufactured food products yield a higher value-added than non-processed food products. Hence, certification with IFS is expected to generate even higher profit than other standards (Colen et al., 2012). We consider our data set as unique since it contains the number of certificates per country and product from 2008 to 2013. The amount of certification is an indicator for the relevance of IFS within a country. About 12,000 food-manufacturing companies are located in 87 countries including 52 developing countries.4 The total number of certification increased from about 4,000 in 2008 to almost 12,000 certificates in 2013. Europe is the major hub of IFS certification, see Figure 2. Numbers increased especially in Asia and Europe from 2008 to 2013 by almost 500% and about 100% respectively. The unequal distribution of IFS with Europe being the most relevant region is displayed in the world map in Figure 2 as well. Moreover, Figure 3 underlines some regional patterns regarding the correlation between exports and income. Countries with more IFS certification tend to have higher exports with central European countries leading both IFS and exports. Similarly, richer countries have more certified producers on average. Both patterns naturally reflect the dominance of central European countries.

4. Model specification (a) Data Since we are predominantly interested in the effect of IFS as an important private standard on bilateral trade flows, IFS is our main 1

Hauptverband des Deutschen Einzelhandels. Fédération des Entreprises du Commerce et de la Distribution. 3 Metro Group, Edeka, Rewe Group, Aldi, Lidl, Kaufland, Kaiser’s Tengelmann, Auchan, Carre-four Group, EMC—Groupe Casino, Leclerc, Monoprix, Picard, Surgelés, Provera (Cora and Supermachés Match), Système U, COOP, CONAD und Unes. 4 All products and countries including HS classification are listed in Tables 4 and 7 respectively. 2

America Europe (RHS)

2008

2009

2010

2011

2012

2013

2000

0

Number of certifications per 1000 100 200

3000 4000 5000 Number of certifications per 1000

Africa Asia

19

6000

300

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27

Year Source: IFS Audit Database

Figure 2. Regional aspects of IFS (1).

(a) IFS and exports

(b) IFS and per capita income Figure 3. Regional aspects of IFS (2).

variable of interest. As highlighted by Head and Mayer (2014), the gravity model became the ‘‘workhorse model” in empirical trade analysis. The required variables are explained briefly: Bilateral trade in current US Dollar from UN Comtrade is the dependent variable for seven different sectors: egg products, meat, fruits and vegetables, bakery products, dairy products, and beverages. We use all countries for which we have IFS data as importing as well as exporting countries, see Table 7. In addition to total export values per sector, Figure 4 also displays the export performance per continent. Hereby, we use an index equal to 100 for the year 2008. Exports declined for all four continents until 2009 due to the economic and financial crisis, peaked in 2011 and mostly increased again for 2013. Asia is the only continent that performed worse in 2013 compared to 2011. Exports of African countries perform relatively poorly as well. Exports of American countries increase most remarkably by almost 60%. The remaining variables are of standard gravity nature: we include the logarithm of GDP in current US Dollar from the World Bank as proxies for the economic mass of both trading partners. Proxies for trade costs like distance, language, and colony are obtained from CEPII whereas ad-valorem tariffs come from the ITC. Descriptive statistics are provided in Table 1. In total, we use 1,822,819 observations from which 83% of the deflated export observations are equal to zero. (b) The benchmark specification The estimation strategy of gravity models in international trade needs to address several empirical challenges. The model needs to

account for multilateral resistance (Anderson & van Wincoop, 2003), high share of zeros (Helpman et al., 2008), and heteroskedasticity (Santos Silva & Tenreyro, 2006, 2011) in particular. Country–year fixed effects are frequently used to account for multilateral resistance. However, this approach becomes computationally difficult the larger the data set becomes in terms of countries and years included. Alternatively, Baier and Bergstrand (2010) propose a different method which adjusts all trade cost proxies in such a way that multilateral resistance does not differ across countries. The Baier–Bergstrand technique is our preferred method due to the large number of observations which makes estimating country–year fixed effects computationally difficult. Moreover, we estimate a multiplicative gravity model with PPML (Santos Silva & Tenreyro, 2006) which does not require to take logs of the dependent variable and therefore, does not drop zeros. In addition, it is robust to heteroskedasticity which is usually present in trade data. The final model of the benchmark specification is defined as follows:

X ijpt ¼ expðb0 þ b1 ln IFSipt1 þ b2 ln GDPit þ b3 ln GDPjt þ b4 ln Distij þ b5 ln Tariff ijpt þ b6 Languageij þ b7 Colonyij þ b8 Contiguityij þ b9 RTAijt þ tp Þgijpt

ð1Þ

X ijpt denotes deflated exports from country i to country j of sector p in year t. IFS represents the number of certifications in the exporting country and is the main variable of interest. However, IFS is likely to be endogenous because of reverse causality. Certification might not only increase trade flows due to the beforehand explained reasons. Vice versa, products might be more likely to be certified in a specific

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27 160

20

Africa Asia

Egg products

Index, 2008=100 120

140

Meat

America Europe

Fruits & Vegetables Dairy products

100

Bakery products Fish

80

Beverages 0

20

40

60

2008

2009

2010

Imports in billion

2011

2012

2013

Year

Source: Comtrade

Source: Comtrade

(a) Exports in 2013 / product categories

(b) Export performance (2008=100)

Figure 4. Exports per product and continent.

Table 1 Descriptive statistics Variable Exports (UN Comtrade) IFS certification (IFS Audit database) GDP_Importer (WDI, current US-D, logs) GDP_Exporter (WDI, current US-D, logs) Distance (CEPII, logs) Tariffs (ITC, logs) Language (CEPII) Colony (CEPII) RTA (CEPII)

Obs

Mean

Std. Dev.

Min

Max

1,822,819 1,822,819 1,776,981 1,759,802 1,822,819 1,822,819 1,822,819 1,822,819 1,822,819

2.822.812 1.493.234 2.522.667 2.521.339 8.549.615 .0779106 .0881261 .0221552 .2735779

5.623.452 5.521.823 1.995.377 2.000.466 .9231002 .1487805 .2834783 .1471883 .445795

0 0 2.056.326 2.056.326 4.087.945 0 0 0 0

12264.59 694 3.030.507 3.030.507 9.892.039 2.462.076 1 1 1

sector p if trade flows are high. Therefore, in the benchmark specification IFS is introduced as a one-year lag to address partially endogeneity due to reverse causality. Country–year fixed effects are not required because we apply the Baier–Bergstrand method. Nevertheless, sector fixed effects tp remain in the model to control for sectorspecific effects. Since the Baier–Bergstrand method is usually applied at the sectoral level, we estimate Eqn. (1) sector-wise as well. (c) An instrumental variable approach Because of the above-mentioned reverse causality, we expect IFS to be endogenous. The lag of IFS as an instrument for IFS does not solve the endogeneity problem if the errors gijpt are autocorrelated.

  E IFSipt gijpt – 0 If the one-year lag of IFS was exogenous, it should not be correlated with the error term:

  E IFSipt1 gijpt ¼ 0 This argument is based on the assumption that IFS itself is correlated over time but the errors are not. However, in the presence of autocorrelation the one-year lag is not a valid IV:

gijpt ¼ q1 gijpt1 þ iijpt ^ 1 is significantly different from zero, the error If the coefficient q terms are autocorrelated and the exogeneity assumption of IFSijpt1 does not hold. If IFS is correlated with the current error term, it is also correlated with its lag if the errors are autocorrelated. The

Wooldridge test for autocorrelation rejects the null hypothesis of no autocorrelation for all usual significance levels. Therefore, an additional identification strategy is required.

(d) IFS certification in neighboring countries as an instrument Applying an instrumental variable approach allows to address the endogeneity of IFS if a valid instrument is available. Relevance and excludability are the two key requirements of a valid instrument. An instrument is relevant if it explains sufficient variation of the endogenous variable. This relevance condition can be tested empirically for example via partial R-squared of the first-stage estimation. In contrast, excludability (or strict exogeneity) as the second condition is not testable and requires arguments based on economic theory. An instrument is excludable if it affects the outcome variable only through the endogenous variable. Both requirements make it difficult and sometimes impossible to find a valid instrument. As we will argue in the following, the total number of IFS-certified producers in all neighboring countries of a particular exporting country i meets both requirements.5 This instrument has been used previously in the context of genetically modified organisms (GMO) regulation and trade (Vigani, Raimondi, & Olper, 2012). The authors use the weighted average of GMO indices of the five closest neighbors to avoid biased estimates due to endogeneity. The estimated negative effect on trade due to GMO regulation even increases in magnitude

5 We used modern grocery distribution as an instrument in an earlier version of the paper. However, the prevalence of modern grocery stores is correlated with income which is—by definition of gravity—a determinant of trade. As a result, we make use of a different instrument which is explained below.

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27

compared to the non-IV specification. In a similar approach, Djankov, Freund, and Pham (2010) use export delay of neighboring countries as an instrument of domestic trade time. Using specific characteristics of neighbors as instruments is well established in particular at the micro level. For example Chege, Andersson, and Qaim (2015, p. 398), use the ‘‘number of supermarket farmers among the five nearest neighbors” as an instrument for supermarket participation of a particular farmer. Based on Maertens and Barrett (2013), the authors argue that social interactions within a neighborhood of farmers determine adoption behavior of modern agricultural technologies. These effects and mechanisms are also present and verifiable at the macro level. There is empirical evidence for the relevance of neighbors for exports of a particular country. Kamal and Sundaram (2016) show that firms of an exporting country i are more likely to establish trade relationships with firms in an importing country j if firms of a neighboring country of i have already well-established relationships with firms in j. The authors discuss various channels like knowledge transfer in terms of destination-specific cultural and business norms as well as legal requirements. A second channel refers to cost sharing like reducing search and matching costs between sellers and buyers. Both channels are particularly relevant in the presence of imperfect information. These spillover effects of neighboring exporters are especially important for the exporting decision (extensive margin), but less relevant for export volume per firm (intensive margin) (Koenig, Mayneris, & Poncet, 2010). The authors refer this difference to the varying relevance of fixed and variable trade costs. Thus, spillover effects due to exporting neighbors reduce barriers to trade which are a result of fixed costs. This makes these findings interesting for the debate of standards and trade. Moreover, the effect is stronger at disaggregated levels such as sector and destination dimensions than at the aggregate level. Summing up, there is empirical support in related literature that export decisions of firms within a country are influenced by firms in neighboring countries. Networks between these firms facilitate knowledge transfer which we could directly link to compliance with standards like IFS. As shown by Ehrich and Hess (2015), not only monetary costs but also non-monetary costs determine compliance which emphasizes the important role of knowledge transfers. Thus, we argue that the extent of compliance with a particular standard in a country affects the likelihood of compliance in a neighboring country. Figure 5 supports this argument. It shows the positive correlation between the endogenous variables’ IFS certification and the instrument which is defined as the sum of IFS certification in all neighboring countries. The correlation is most pronounced in Western European countries like Germany, France (the initiators of IFS) but lower in Easter European countries like Ukraine, Belarus, Albania, and Armenia. However, note that some countries do have neighbors for which we do not have any IFS observation. Nevertheless, the correlation between both measures is equal to 0.51. To conclude, we are confident that the instrument is relevant and therefore, meets the first requirement. The exclusion restriction requires the instrument to affect exports only via the endogenous variable. In contrast to the benchmark specification, reverse causality is not a problem in this IV specification. There is no plausible argument why exports of a country should influence compliance with IFS in neighboring countries. Furthermore, there is also no channel through which IFS in neighboring countries could affect exports apart from IFS of the country itself. The extent of IFS certification in a country does affect for example neither GDP of its neighbor nor tariffs or other determinants of exports of their neighbor. Instead, domestic borders are exogenous such that countries cannot self-select themselves toward specific neighbors.

21

5. Results We estimate the effect of IFS on trade using three different methods: first, ppml without IV and two models with IV which are estimated via the control function approach (MartínezZarzoso, 2015; Wooldridge, 2010) and general methods of moments (GMM). Furthermore, we estimate these models at the aggregate level (Table 2), by income group of the exporting country (Table 3), and at the sectoral level (see Table 6 in the Appendix A). Column one of Table 2 shows coefficients of the benchmark specification which is estimated via ppml and the one-year lag of IFS certification. The coefficient equaling to 0.27 is interpreted as elasticity and it is statistical significant. Thus, a 1% increase in IFS certification increases country i’s exports c.p. by 0.27% on average. Other elasticities match general findings obtained via ppml within the gravity framework. Coefficients of GDP variables are statistically significant and below unity. In contrast, distance and tariffs reduce trade while as other trade costs’ proxies language, colonial relationship and contiguity are trade enhancing. The coefficient of regional trade agreement is equal to 0.151 which implies a reasonable economic magnitude of 16%.6 Column two contains estimates of the IV regression of the control function approach. Most are remarkably similar compared with the non-IV method. The IFS coefficient is of smaller magnitude equaling 0.147. The key difference is that trade cost proxies become larger in magnitude, in particular tariffs. Columns three and four belong to the control function approach. Column three shows estimates of the first stage which indicates the indeed high relevance of IFS neighboring countries for IFS certification itself. Finally, GMM results are shown in column four. Most coefficients remain similar to the ppml model. Summing up, IFS is statistically and economically significant in all three specifications. IFS certification increases exports c.p. on average by about 0.147 and 0.270% if number of certified firms increases by 1%. However, results differ by income of the exporting country. Table 3 shows coefficients of the benchmark specification (columns one, five, and seven) as well as of the discussed IV specifications. We distinguish between high-income, upper-middleincome, and low-income groups (see Table 7 for exact grouping of countries). The effect of IFS on trade is positive and statistically significant in the benchmark specification for all income groups. The effect is largest for high-income countries and lowest for upper-middle-income countries. However, there are considerable differences in the IV specifications. The IFS coefficient of the CFA model increases sharply to 2.088 and to 0.718 in the GMM model. Although the high magnitude in the CFA models is questionable, these coefficients are higher than in those in the other two income groups. The IFS coefficient of the GMM model reduces 0.253 and turns even negative to 0.114 for low-income countries. The latter lacks statistical significance. The CFA model does not convergence for the middle-income groups and becomes smaller and statistically insignificant in the low-income group. Apart of the GDP coefficient of the exporter in the high-income group in the GMM model, all other coefficient are mostly as predicted by theory in terms of sign, magnitude, and statistical significance. Finally, Table 6 in the Appendix A shows results at the sectoral level. While the effect of IFS on exports remains positive in the benchmark specification using ppml (except eggs), sign and magnitude vary in the IV models. Estimates are only strictly positive and statistically significant for bakery and dairy sector. Results are mixed for meat (negative with CFA, not different from zero with GMM). IFS reduces exports of fish which remains to be discussed (see Section 6 below). Exports of fruits and vegetables are not

6


Dummies are interpreted in ppml regressions as ebD  1  100.

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27

15

22

DEU FRA NLD POLAUT BEL CZE CHE DNK

Total IFS (log) 10 5

ITA ESP HUN CHN GRC TUR BGR VNM ROM

THA GBR USACHL ZAF EGY IDN CRI GHA KEN

GTM

IND

HRV LTU RUS

SWEECU BRA CAN PER LVAIRL FIN BIH COL NOR CIVPNG EST ARG ISR GUY ARMALB NAM SUR NIC MEX IRN KAZ

SVK PRT SVN MAR

UKR

LUX

BLR

AZEPAK

0

SMR

4

6

8 10 12 Total IFS in all neighboring countries (log) lnTotal_IFS

14

Fitted values

Source: IFS, own graph Figure 5. Scatter plot of number of IFS certification in exporting country i and the sum of IFS certified producers in all neighboring countries.

Table 2 Estimation results obtained via ppml and two IV methods Variables

(1) ppml—no IV

lnGDP_Exporter lnDist lnTariff Comlang_ethno Colony Contig RTA D_Bakery D_Beverages D_Dairy D_Egg D_FV D_Fish Constant Observations

(4) IV—GMM 0.253⁄⁄⁄ (0.028)

0.270⁄⁄⁄ (0.011)

0.723⁄⁄⁄ (0.011) 0.387⁄⁄⁄ (0.011) 0.698⁄⁄⁄ (0.032) 1.295⁄⁄⁄ (0.111) 0.211⁄⁄⁄ (0.056) 0.263⁄⁄⁄ (0.087) 0.481⁄⁄⁄ (0.062) 0.151⁄⁄⁄ (0.053) 0.073 (0.073) 0.345⁄⁄⁄ (0.080) 0.382⁄⁄⁄ (0.079) 0.749⁄⁄⁄ (0.151) 0.728⁄⁄⁄ (0.075) 0.375⁄⁄⁄ (0.077) 29.006⁄⁄⁄ (0.466)

0.768⁄⁄⁄ (0.016) 0.609⁄⁄⁄ (0.027) 0.901⁄⁄⁄ (0.042) 2.219⁄⁄⁄ (0.196) 0.665⁄⁄⁄ (0.113) 0.694⁄⁄⁄ (0.131) 0.749⁄⁄⁄ (0.093) 0.299⁄⁄⁄ (0.082) 0.008 (0.183) 0.191 (0.189) 0.153 (0.185) 1.441⁄⁄⁄ (0.227) 0.960⁄⁄⁄ (0.183) 0.888⁄⁄⁄ (0.195) 35.872⁄⁄⁄ (0.725)

0.003⁄⁄⁄ (0.000) 0.000 (0.002) 0.356⁄⁄⁄ (0.002) 0.003 (0.006) 0.036⁄ (0.020) 0.026⁄⁄ (0.012) 0.020 (0.022) 0.010 (0.020) 0.019⁄⁄ (0.009) 0.089⁄⁄⁄ (0.018) 0.257⁄⁄⁄ (0.018) 0.306⁄⁄⁄ (0.019) 0.078⁄⁄⁄ (0.023) 0.145⁄⁄⁄ (0.018) 0.444⁄⁄⁄ (0.017) 8.030⁄⁄⁄ (0.062)

691,631

843,538

843,538

IFS_Neighbor lnGDP_Importer

(3) IV—CFA 1st

0.147⁄⁄⁄ (0.029)

IFS IFSt1

(2) IV—CFA

Robust standard errors in parentheses. ⁄⁄⁄ p < 0:01, ⁄⁄ p < 0:05, ⁄ p < 0:1. Baier–Bergstrand’s method used to account for multilateral resistance. FV stands for Fruits and Vegetables. CFA stands for control function approach and GMM for general methods of moments.

0.721⁄⁄⁄ (0.017) 0.390⁄⁄⁄ (0.018) 0.700⁄⁄⁄ (0.057) 1.369⁄⁄⁄ (0.177) 0.221⁄⁄⁄ (0.086) 0.278⁄⁄ (0.129) 0.495⁄⁄⁄ (0.114) 0.094 (0.079) 0.048 (0.126) 0.307⁄⁄ (0.145) 0.332⁄⁄ (0.131) 0.790⁄⁄⁄ (0.250) 0.816⁄⁄⁄ (0.129) 0.322⁄⁄ (0.136) 28.932⁄⁄⁄ (0.654) 843,538

23

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27 Table 3 IFS estimation results by income group using PPML without and IFS-certification in neighboring countries as IV Variables

(1)

(2) (3) High income

ppml—no IV

IFS_lag

IV—CFA

lnGDP_Exporter lnDist lnTariff Comlang_ethno Colony Contig RTA D_Bakery D_Beverages D_Dairy

D_Fish D_Meat Constant Observations

ppml—no IV

0.718⁄⁄⁄ (0.198)

0.616⁄⁄⁄ (0.124) 0.669⁄⁄⁄ (0.141) 0.829⁄⁄⁄ (0.115) 0.739⁄⁄ (0.369) 1.695⁄⁄⁄ (0.337) 0.318 (0.536) 2.668⁄⁄⁄ (0.261) 0.213 (0.422) 0.147 (0.313) 0.478 (0.361) 0.645⁄⁄ (0.287)

1.704⁄⁄⁄ (0.345) 0.641⁄⁄ (0.271) 1.205⁄⁄⁄ (0.332) 2.399 (3.798) 30,031

IV—GMM

(7)

(8) (9) Low income

ppml—no IV

0.253⁄⁄⁄ (0.028) ⁄⁄⁄

0.392 (0.133)

D_Egg D_FV

IV—GMM

(5) (6) Upper middle income

⁄⁄⁄

0.865⁄⁄⁄ (0.051) 1.163⁄⁄⁄ (0.240) 2.644⁄⁄⁄ (0.249) 0.311 (0.732) 0.678 (0.546) 0.307 (0.765) 0.444 (0.477) 0.557 (0.390) 0.975 (0.647) 0.132 (0.633) 0.271 (0.666) 0.775 (0.780) 3.856⁄⁄⁄ (0.678) 0.118 (0.646)

0.011⁄⁄⁄ (0.000) 0.002 (0.002) 0.061⁄⁄⁄ (0.011) 0.041⁄⁄⁄ (0.008) 0.058⁄⁄ (0.026) 0.046⁄ (0.026) 0.045 (0.030) 0.051⁄⁄ (0.024) 0.149⁄⁄⁄ (0.013) 0.304⁄⁄⁄ (0.016) 0.108⁄⁄⁄ (0.018) 0.276⁄⁄⁄ (0.018) 0.180⁄⁄⁄ (0.017) 0.881⁄⁄⁄ (0.021) 0.542⁄⁄⁄ (0.017)

0.486⁄⁄⁄ (0.123) 0.480⁄⁄ (0.200) 0.707⁄⁄⁄ (0.169) 0.792⁄ (0.415) 1.259⁄⁄⁄ (0.327) 0.224 (0.588) 2.467⁄⁄⁄ (0.377) 0.290 (0.499) 1.247⁄⁄⁄ (0.387) 0.481 (0.321) 0.505 (0.348) 0.826⁄ (0.462) 2.955⁄⁄⁄ (0.368) 0.479 (0.339)

0.752⁄⁄⁄ (0.032) 0.625⁄⁄⁄ (0.043) 0.118 (0.111) 0.291 (0.283) 0.644⁄⁄⁄ (0.164) 0.798⁄⁄⁄ (0.212) 1.095⁄⁄⁄ (0.340) 0.741⁄⁄⁄ (0.090) 0.566⁄⁄⁄ (0.214) 0.297 (0.194) 1.491⁄⁄⁄ (0.228)

2.244 (5.836) 36,837

1.334⁄⁄⁄ (0.259) 36,837

2.605 (6.666) 36,837

IV—CFA

IV—CFA 1st

(10) IV—GMM 0.114 (0.452)

0.063 (0.481) ⁄⁄⁄

0.241 (0.049)

IFS_Neighbor lnGDP_Importer

IV—CFA 1st

2.088⁄⁄⁄ (0.432)

IFS

(4)

0.443 (0.034)

0.814⁄⁄⁄ (0.029) 0.305⁄⁄⁄ (0.024) 0.419⁄⁄⁄ (0.061) 0.628⁄⁄⁄ (0.227) 0.189 (0.154) 0.919⁄⁄⁄ (0.196) 0.057 (0.125) 0.141 (0.092) 0.714⁄⁄ (0.328) 1.063⁄⁄⁄ (0.363) 2.889⁄⁄⁄ (0.398)

0.792⁄⁄⁄ (0.026) 0.765⁄⁄⁄ (0.131) 0.573⁄⁄⁄ (0.098) 4.166⁄⁄⁄ (0.400) 0.961⁄⁄⁄ (0.207) 0.581⁄⁄ (0.294) 0.335 (0.265) 0.605⁄⁄⁄ (0.146) 0.762 (0.554) 1.395⁄⁄ (0.549) 1.679⁄⁄ (0.760)

0.822⁄⁄⁄ (0.201) 0.417⁄ (0.227)

0.721⁄⁄⁄ (0.017) 0.390⁄⁄⁄ (0.018) 0.700⁄⁄⁄ (0.057) 1.369⁄⁄⁄ (0.177) 0.221⁄⁄⁄ (0.086) 0.278⁄⁄ (0.129) 0.495⁄⁄⁄ (0.114) 0.094 (0.079) 0.048 (0.126) 0.307⁄⁄ (0.145) 0.332⁄⁄ (0.131) 0.790⁄⁄⁄ (0.250) 0.816⁄⁄⁄ (0.129) 0.322⁄⁄ (0.136)

0.001⁄⁄⁄ (0.000) 0.001 (0.002) 0.278⁄⁄⁄ (0.003) 0.017⁄ (0.010) 0.164⁄⁄⁄ (0.035) 0.009 (0.015) 0.012 (0.037) 0.016 (0.039) 0.048⁄⁄⁄ (0.014) 0.445⁄⁄⁄ (0.025) 0.422⁄⁄⁄ (0.026) 0.910⁄⁄⁄ (0.023)

1.035⁄⁄⁄ (0.338) 0.465 (0.330)

1.343⁄⁄ (0.553) 0.502 (0.604)

0.375⁄⁄⁄ (0.024) 0.671⁄⁄⁄ (0.025)

0.743 (0.689) 0.879 (0.687)

35.685⁄⁄⁄ (1.773) 145,219

28.932⁄⁄⁄ (0.654) 843,538

28.928⁄⁄⁄ (1.167) 147,410

39.731⁄⁄⁄ (3.049) 181,515

6.127⁄⁄⁄ (0.094) 181,515

32.908⁄⁄⁄ (4.618) 181,515

0.801⁄⁄⁄ (0.045) 0.497⁄⁄⁄ (0.168) 0.457⁄⁄⁄ (0.110) 1.210⁄⁄⁄ (0.372) 0.371 (0.265) 0.985⁄⁄⁄ (0.294) 0.067 (0.299) 0.054 (0.143) 0.936 (0.756) 1.479⁄ (0.846) 3.084⁄⁄⁄ (1.024)

Robust standard errors in parentheses. ⁄⁄⁄ p < 0:01, ⁄⁄ p < 0:05, ⁄ p < 0:1. Baier–Bergstrand’s method used to account for multilateral resistance. FV stands for Fruits and Vegetables. CFA stands for control function approach and GMM for general methods of moments.

affected by IFS certification whereas exports of beverages only increase due to IFS in the GMM model. Exports of egg sector are either not (ppml), positively (CFA), or negatively (GMM) affected by IFS. Thus, there is again no one-suits-all effect of IFS certification on agricultural exports. Instead, effects differ not only by income but also by sector. 6. Discussion and conclusion Finally, it remains essential to relate our empirical findings to the initial research question. In particular, we discuss the potential of IFS to integrate low-income countries into global value chains. Second, we discuss potential reasons of negative coefficients that appear in some sectors. (a) IFS and the integration of low-income countries into global export markets Although we do find trade-enhancing effects of IFS certification on trade at the aggregate level, results remain robust only for highand middle-income countries. Therefore, it remains to be discussed

to what extent private food standards are a useful development policy tool to integrate low-income countries into global agricultural export markets. The effect of IFS certification on exports differs by income level of the exporting country. If our hypothesis that high-income countries benefit more from IFS certification in terms of increasing exports than countries with lower income levels, the IFS coefficients should either decline in magnitude, become negative, or loses its statistical significance. We find empirical evidence—based on GMM estimates—for the latter which are first, the effects of IFS on exports depends on income and second, low-income countries do not benefit from increasing IFS certification in terms of higher export volumes. Results of the benchmark specification based on ppml estimates are less clear. However, we consider the IV estimates as superior due to the above-discussed advantages related to reverse causality. The findings indicate that IFS certification is not sufficient to integrate low-income countries into global value chains. In other words, the trade-enhancing factors such as lower information asymmetries and lower transaction costs are not sufficiently relevant to offset factors of the trade barrier’s view such as compliance costs and non-monetary barriers such as weak public institutions.

24

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27

As shown in Figs. 2 and 3, low-income countries have on average lower compliance levels than high-income countries. However, the non-significance of the IFS coefficient does allow us to conclude only that IFS certification in low-income countries does not affect exports. Hence, we do not know whether compliance costs are actually a relevant trade barrier in this context because the empirical model does not address the likelihood of compliance, but rather that compliance does not affect trade. This might be the case because the assumed reduction of information asymmetries is not sufficient to increase demand of importing countries. Consequently, even if compliance takes place, our findings do not provide empirical support for the conclusion that compliance as such is sufficient to increase country i’s exports. The latter interpretation might contradict findings of previous studies, especially about vegetables production in Senegal (Colen et al., 2012; Maertens & Swinnen, 2009; Swinnen et al., 2015). The authors emphasize—based on a rich firm- and householdlevel dataset—the potential of standards to integrate low-income countries into global value chains, increasing rural income, and agricultural exports. However, although these analyses address multiple outcomes and not only trade, these studies remain casespecific such that external validity remains unclear. Moreover, most studies related to the Senegal project are based on the private standard GlobalGAP (or formerly EuroGAP) which is a pre-farm gate and mostly addresses agricultural raw products. In contrast, IFS is a post-farm gate standard for processed food products. The effect on trade of such a post-farm gate standard might be less pronounced, because high-income countries have on average higher tariffs on processed food products. Moreover, food processing generates on average higher value-added than the production of agricultural raw products. Most high-income countries prefer to generate the value-added themselves. This phenomenon is known as ‘‘tariff escalation” and a serious threat for integration of developing countries into global value chains (Akyol et al., 2005). Thus, a possibly trade-enhancing effect of IFS certification in developing countries might be offset by these protections trade policies of high-income countries which are distinct for processed food products. In terms of policy implications, we argue that low-income food exporters require the existence of a well-functioning national quality infrastructure. A well-functioning quality infrastructure is characterized by local certification bodies with skilled and experienced auditors as well as a national accreditation body that is internationally recognized by the network of national accreditation bodies.

dards can be included as minimum requirements of the importing country (e.g. maximum residue limits). In this case, standards are applied in terms of mandatory and legal minimum requirements of food products. Negative effects on trade are straightforward to explain in such settings. In this study however, standards are implemented as the number of certified firms in the exporting country. There is no immediate argument why more certified firms should reduce exports as it is the case for fish. However, it is possible that more IFS certification of fish farms crowds out exports of non-IFS-certified fish such that exports of IFS-certified fish increase, but the net effect on fish exports is negative. In other words, IFS certification is likely to create winners and losers in terms of exports within countries. This is the case for the fish sector. To further validate this argument we would need firm-level data of (fish) farms. The creation of winners and losers as a result of IFS might also be present in other sectors in which the overall effect of IFS on exports remains positive. Nevertheless, it is likely to be the case that a subset of firms—probably not IFS-certified— exports less as the number of IFS-certified firms increase whereas others—probably IFS-certified—export more. To conclude, this analysis contributes to the catalysts- vs. barriers-to-trade debate by providing additional evidence for heterogeneous effects of standards on exports of low-income countries. Thus, while the trade catalysts’ view predominates on average and for high- and middle-income countries, the trade barriers’ view remains relevant for low-income countries. As a result, neither the catalysts’ nor the barriers’ view remains correct in all cases. A valid answer to the debate, which we summarize briefly in Section 2, is required to take case-specific conditions into account. However, a significant number of studies have accumulated (including this one) that conclude that standards constitute a serious threat for the integration of low-income countries into global agricultural export markets. Future research is required to complement these results of the country perspective with those at the food-processing firm level in low-income countries. Eventually, to evaluate the increasing relevance of food standards comprehensively, various perspectives ranging from macro-trade, marketing toward rural development studies, and many more need to be taken into account. Moreover, future research should deal with effects on trade of B2C standards such as Fairtrade or the MSC. Acknowledgments Financial support of the German Research Foundation (DFG) is gratefully acknowledged.

(b) On the negative effect of IFS on fish exports Appendix A It is important to highlight the different nature of standards proxies that can be used in this type of analysis. For example, stan-

See Tables 4–7.

Table 4 Products Product group

HS code

HS description

Bakery products

1704 1806 1901 1902 1903 1904 1905

Sugar confection Chocolate & other food products containing cocoa Malt extract, food preparations of flour, etc. Pasta, prepared or not, couscous Tapioca and substitutes from starch in flakes, etc. Foods prep by swell cereal, cereal n.e.s.o.i. Bread, pastry cakes, etc.

Beverages

2009 2201 2202

Fruit juices (& grape must), vegtables juice, no spirit Water, natural etc., not sweetened etc., ice & snow Water, sweetened & other non-alcoholic beverages n.e.s.o.i.

25

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27

Table 4 (continued) Product group

HS code

HS description

2203 2204 2205 2206 2207 2208

Beer made from malt Wine of fresh grapes, grape must n.e.s.o.i. Vermouth & other wine of fresh grapes with specific flavor Fermented beverages n.e.s.o.i. (cider, berry, mead) ethyl alcohol, un-denatured, n/un 80% alcohol, alcohol, denatured ethyl alcohol, un-denatured, 80% alcohol, spirit beverages, etc.

Dairy products

401 402 403 404 405 406

Milk and cream, not concentrated or sweetened Milk and cream, concentrated or sweetened Buttermilk, yogurt, kephir, etc. Whey & milk products n.e.s.o.i. Butter and other fats and oils derived from milk Cheese and curd

Egg products

407 408

Birds’ eggs, in the shell, fresh preserved or cooked Birds’ eggs, not in shell &yolks, fresh dry, etc.

Fruits and vegetables products

2001 2002 2003 2004 2005 2006 2007 2008

Vegetable, fruits, nuts, etc. Tomatoes prepared or preserved n.e.s.o.i. Mushrooms & truffles prepared or preserved n.e.s.o.i. Vegetables n.e.s.o.i. prepared or preserved, frozen Vegetables n.e.s.o.i. prepared or preserved, not frozen Fruit/nuts/fruit-peel etc., preserved by sugar Jams, fruit jellies, marmalade, etc., cooked Fruit, nuts, etc., prepared or preserved n.e.s.o.i.

Fish products

303 304 305 306 307 1604 1605

Fish, frozen (no fillets) Fish fillets, other fish, fresh, chill, or frozen Fish, dried, salted, etc., smoked Crustaceans, live, fresh, cooked Mollusks, aquatic invertebrates n.e.s.o.i. Prepared or preserved fish, caviar & caviar substitutes Crustaceans and mollusks prepared or preserved

Meat products

1601 1602

Sausages, similar prepared meat Prepared or preserved meat, meat offal & blood n.e.s.o.i.

Table 5 Importing and exporting countries Europe Albania Austria Belarus Belgium Bosnia Herzegovina Bulgaria Croatia Cyprus Czech Republic Denmark Estonia Faeroe Island Finland France Germany Greece Hungary Iceland Ireland Italy

Africa 2 1 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Latvia Lithuania Luxembourg Macedonia Netherlands Norway Poland Romania San Marino Slovakia Slovenia Spain Sweden Switzerland Turkey Ukraine United Kingdom

1 2 1 2 1 1 1 2 1 1 1 1 1 1 2 3 1

Ivory Coast Egypt Ghana Kenya Madagascar Mauritius Morocco Namibia Senegal Seychelles South Africa Tunisia Uganda Peru Suriname United States Uruguay

America 3 3 3 3 3 2 3 2 3 2 2 3 3 2 2 1 2

Antigua and Barbuda Argentina Brazil Canada Chile Colombia Costa Rica Ecuador Guatemala Guyana Honduras Mexico Nicaragua

Notes: numbers after country names refer to income groups: (1) High income, (2) Upper middle income, (3) Low income.

Asia 2 2 2 1 2 2 2 3 3 3 3 2 3

Armenia Australia Azerbaijan Bangladesh China India Indonesia Iran Israel Kazakhstan South Korea Malaysia Pakistan Papua New Guinea Philippines Russia Sri Lanka Thailand United Arab Emirates Vietnam

3 1 2 3 3 3 3 2 1 2 1 2 3 3 3 2 3 3 1 3

Bakery ppml IFSt1 lnGDP_Importer lnGDP_Exporter lnDist lnTariff Comlang_ethno Colony

RTA

26.466⁄⁄⁄ (0.689)

0.615⁄⁄⁄ (0.027) 0.573⁄⁄⁄ (0.045) 1.223⁄⁄⁄ (0.088) 2.051⁄⁄⁄ (0.588) 0.614⁄⁄ (0.308) 0.811⁄⁄⁄ (0.166) 0.574⁄⁄⁄ (0.132) 0.573⁄⁄⁄ (0.206) 0.402⁄⁄⁄ (0.051) 31.780⁄⁄⁄ (1.210)

0.645⁄⁄⁄ (0.026) 0.358⁄⁄⁄ (0.032) 0.757⁄⁄⁄ (0.076) 0.085 (0.531) 0.242⁄ (0.141) 0.597⁄⁄⁄ (0.169) 0.702⁄⁄⁄ (0.172) 0.879⁄⁄⁄ (0.143) 0.368⁄⁄⁄ (0.030) 26.563⁄⁄⁄ (1.030)

115,605

139,587

ppml

IV (CFA)

IFS (IV) Constant Observations

Meat IV (GMM)

ppml 0.156⁄⁄⁄ (0.034) 0.666⁄⁄⁄ (0.028) 0.318⁄⁄⁄ (0.031) 0.696⁄⁄⁄ (0.094) 1.052⁄⁄⁄ (0.258) 0.547⁄⁄⁄ (0.129) 0.243 (0.444) 0.826⁄⁄⁄ (0.187) 0.190 (0.182)

25.206⁄⁄⁄ (0.982)

0.664⁄⁄⁄ (0.048) 0.370⁄⁄⁄ (0.042) 0.715⁄⁄⁄ (0.167) 1.089⁄⁄⁄ (0.395) 0.536⁄⁄ (0.215) 0.590 (0.801) 1.030⁄⁄⁄ (0.361) 0.467 (0.311) 0.021 (0.083) 26.042⁄⁄⁄ (1.379)

139,587

29,540

35,356

IV (GMM)

ppml

lnGDP_Importer lnGDP_Exporter lnDist lnTariff Comlang_ethno Colony Contig RTA

Constant Observations

⁄⁄⁄

26.803⁄⁄⁄ (1.013)

0.622 (0.041) 0.523⁄⁄⁄ (0.069) 1.246⁄⁄⁄ (0.131) 1.728⁄⁄⁄ (0.310) 0.747⁄⁄⁄ (0.272) 0.941⁄⁄⁄ (0.249) 0.808⁄⁄⁄ (0.232) 0.455⁄⁄ (0.206) 0.428⁄⁄⁄ (0.075) 30.530⁄⁄⁄ (2.057)

0.635 (0.033) 0.454⁄⁄⁄ (0.054) 0.765⁄⁄⁄ (0.090) 2.520⁄⁄⁄ (0.365) 0.441⁄⁄ (0.178) 0.139 (0.183) 0.785⁄⁄⁄ (0.186) 0.574⁄⁄ (0.224) 0.289⁄⁄⁄ (0.064) 28.376⁄⁄⁄ (1.633)

64,230

77,676

77,676

IFS (IV)

0.207⁄⁄⁄ (0.035) 0.776⁄⁄⁄ (0.026) 0.253⁄⁄⁄ (0.029) 0.383⁄⁄⁄ (0.077) 1.610⁄⁄⁄ (0.291) 0.022 (0.119) 0.289⁄⁄⁄ (0.108) 0.467⁄⁄ (0.190) 0.296⁄⁄⁄ (0.099)

IV (CFA)

0.224 (0.021) 0.767⁄⁄⁄ (0.024) 0.534⁄⁄⁄ (0.019) 0.890⁄⁄⁄ (0.080) 0.770⁄⁄⁄ (0.199) 0.109 (0.151) 0.239 (0.233) 0.017 (0.145) 0.542⁄⁄⁄ (0.128)

⁄⁄⁄

IV (CFA)

F&V IV (GMM)

ppml 0.268⁄⁄⁄ (0.017) 0.743⁄⁄⁄ (0.026) 0.444⁄⁄⁄ (0.025) 0.643⁄⁄⁄ (0.055) 1.716⁄⁄⁄ (0.232) 0.303⁄⁄⁄ (0.086) 0.189 (0.138) 0.351⁄⁄⁄ (0.104) 0.344⁄⁄⁄ (0.089)

IV (CFA)

IV (GMM)

0.697⁄⁄⁄ (0.037) 0.539⁄⁄⁄ (0.037) 0.689⁄⁄⁄ (0.086) 2.097⁄⁄⁄ (0.338) 0.278⁄⁄ (0.134) 0.330 (0.228) 0.356⁄⁄ (0.182) 0.130 (0.130) 0.039 (0.034) 32.454⁄⁄⁄ (1.653) 186,496

26.212⁄⁄⁄ (1.086)

0.754⁄⁄⁄ (0.040) 0.758⁄⁄⁄ (0.071) 0.724⁄⁄⁄ (0.101) 4.835⁄⁄⁄ (0.644) 0.636⁄ (0.336) 1.092⁄⁄ (0.515) 0.840⁄⁄ (0.336) 0.158 (0.174) 0.712⁄⁄⁄ (0.166) 37.337⁄⁄⁄ (1.800)

0.780⁄⁄⁄ (0.039) 0.640⁄⁄⁄ (0.044) 0.415⁄⁄⁄ (0.129) 1.645⁄⁄⁄ (0.379) 0.205 (0.188) 0.268 (0.225) 0.706⁄⁄ (0.344) 0.312⁄⁄ (0.156) 0.644⁄⁄⁄ (0.072) 35.043⁄⁄⁄ (1.802)

31.823⁄⁄⁄ (1.127)

0.852⁄⁄⁄ (0.024) 0.855⁄⁄⁄ (0.046) 0.839⁄⁄⁄ (0.075) 5.038⁄⁄⁄ (0.301) 1.040⁄⁄⁄ (0.164) 0.566⁄⁄⁄ (0.173) 0.662⁄⁄⁄ (0.199) 0.195 (0.146) 0.063 (0.047) 45.011⁄⁄⁄ (1.238)

35,356

155,659

191,072

191,072

151,328

186,496

IV (GMM)

ppml

IV (CFA)

IV (GMM)

0.695⁄⁄⁄ (0.147) 2.193⁄⁄ (0.927) 2.041⁄⁄⁄ (0.531) 5.683⁄⁄ (2.754) 1.487 (1.027) 0.680 (0.479) 1.004⁄⁄ (0.509) 0.494 (0.679) 1.455⁄⁄ (0.705) 76.098⁄⁄⁄ (27.609) 15,432

Egg

⁄⁄⁄

⁄⁄⁄

ppml

Beverages

⁄⁄⁄

0.405 (0.023) 0.650⁄⁄⁄ (0.023) 0.366⁄⁄⁄ (0.032) 0.774⁄⁄⁄ (0.065) 2.246⁄⁄⁄ (0.196) 0.465⁄⁄⁄ (0.127) 0.139 (0.119) 0.663⁄⁄⁄ (0.135) 0.675⁄⁄⁄ (0.150)

Fish IV (GMM)

0.741⁄⁄⁄ (0.058) 0.760⁄⁄⁄ (0.086) 1.085⁄⁄⁄ (0.205) 0.967⁄⁄ (0.419) 1.136⁄⁄⁄ (0.397) 0.806⁄⁄ (0.401) 1.621⁄⁄⁄ (0.397) 0.486 (0.364) 0.178⁄⁄⁄ (0.066) 38.551⁄⁄⁄ (2.517)

Dairy

IFSt1

IV (CFA)

⁄⁄⁄

0.173 (0.126) 0.690⁄⁄⁄ (0.082) 0.341⁄⁄ (0.136) 1.352⁄⁄⁄ (0.139) 3.023⁄⁄⁄ (1.108) 0.466⁄ (0.262) 0.359 (0.359) 0.908⁄⁄⁄ (0.221) 0.137 (0.431)

33.704⁄⁄⁄ (0.844)

0.765 (0.022) 0.882⁄⁄⁄ (0.052) 0.853⁄⁄⁄ (0.077) 1.397⁄⁄⁄ (0.322) 0.453⁄⁄⁄ (0.143) 0.310⁄ (0.170) 1.139⁄⁄⁄ (0.180) 0.424⁄⁄⁄ (0.133) 0.058 (0.044) 42.735⁄⁄⁄ (1.372)

0.779 (0.036) 0.461⁄⁄⁄ (0.041) 0.883⁄⁄⁄ (0.154) 0.731⁄⁄ (0.306) 0.187 (0.214) 0.359 (0.305) 0.090 (0.272) 0.388⁄⁄ (0.192) 0.355⁄⁄⁄ (0.064) 32.406⁄⁄⁄ (1.203)

28.144⁄⁄⁄ (2.773)

0.540⁄⁄⁄ (0.109) 2.465 (1.628) 3.465⁄⁄⁄ (0.308) 2.368⁄⁄⁄ (0.838) 0.435 (0.615) 0.670 (0.552) 0.200 (0.543) 1.702⁄⁄⁄ (0.444) 4.541⁄⁄ (2.223) 41.830 (38.788)

162,819

197,919

197,919

12,450

15,432

Because estimations were based on the sectoral level, the method of Baier–Bergtrand was used to account for multilateral resistance. The command ivpoisson was used in Stata. CFA refers to control function approach, gmm to general methods of moments. First-stage results are omitted here.

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27

Contig

0.404⁄⁄⁄ (0.018) 0.640⁄⁄⁄ (0.018) 0.354⁄⁄⁄ (0.020) 0.718⁄⁄⁄ (0.049) 0.087 (0.378) 0.242⁄⁄⁄ (0.089) 0.624⁄⁄⁄ (0.104) 0.697⁄⁄⁄ (0.100) 1.019⁄⁄⁄ (0.088)

IV (CFA)

26

Table 6 Estimation results by product using ppml and IFS-certification in neighboring countries as IV using GMM and CFA

27

M. Ehrich, A. Mangelsdorf / World Development 101 (2018) 16–27 Table 7 Importing and exporting countries by continent and income group Europe Albania Austria Belarus Belgium Bosnia Herzegovina Bulgaria Croatia Cyprus Czech Republic Denmark Estonia Faeroe Island Finland France Germany Greece Hungary Iceland Ireland Italy

2 1 2 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Africa Latvia Lithuania Luxembourg Macedonia Netherlands Norway Poland Romania San Marino Slovakia Slovenia Spain Sweden Switzerland Turkey Ukraine United Kingdom

1 2 1 2 1 1 1 2 1 1 1 1 1 1 2 3 1

Ivory Coast Egypt Ghana Kenya Madagascar Mauritius Morocco Namibia Senegal Seychelles South Africa Tunisia Uganda Uruguay

America 3 3 3 3 3 2 3 2 3 2 2 3 3 2

Antigua and Barbuda Argentina Brazil Canada Chile Colombia Costa Rica Ecuador Guatemala Guyana Honduras Mexico Nicaragua Peru Suriname United States

Asia 2 2 2 1 2 2 2 3 3 3 3 2 3 2 2 1

Armenia Australia Azerbaijan Bangladesh China India Indonesia Iran Israel Kazakhstan South Korea Malaysia Pakistan Papua New Guinea Philippines Russia Sri Lanka Thailand United Arab Emirates Vietnam

3 1 2 3 3 3 3 2 1 2 1 2 3 3 3 2 3 3 1 3

Notes: numbers after country-names refer to income groups: (1) High income, (2) Upper middle income, (3) Low income.

References Akyol, M., & Beghin, J. C. (Eds.). (2005). Global agricultural trade and developing countries. Trade and development series. Washington, D.C: World Bank. Anders, S. M., & Caswell, J. A. (2009). Standards as barriers versus standards as catalysts: Assessing the impact of haccp implementation on U.S. seafood imports. American Journal of Agricultural Economics, 91(2), 310–321. Anderson, J. E., & van Wincoop, E. (2003). Gravity with gravitas: A solution to the border puzzle. American Economic Review, 93(1), 170–192. Baier, S., & Bergstrand, J. H. (2010). Approximating general equilibrium impacts of trade liberalizations using the gravity equation. In P. van Bergeijk & S. Brakman (Eds.), The gravity model in international trade: Advances and applications (pp. 88–134). Cambridge University Press. Beghin, J. C., Maertens, M., & Swinnen, J. (2015). Nontariff measures and standards in trade and global value chains. Annual Review of Resource Economics, 7(1), 425–450. Chege, C. G., Andersson, C. I., & Qaim, M. (2015). Impacts of supermarkets on farm household nutrition in kenya. World Development, 72, 394–407. Chen, C. (2008). Measuring the effect of food safety standards on China’s agricultural exports. Review of World Economics, 144(1), 83–106. Chen, M., Otsuki, T., & Wilson, J. S. (2006). Do standards matter for export success? World Bank Policy Research Working Paper, (3809). Colen, L., Maertens, M., & Swinnen, J. (2012). Private standards, trade and poverty: Globalgap and horticultural employment in senegal. The World Economy, 35(8), 1073–1088. Czubala, W., Shepherd, B., & Wilson, J. S. (2009). Help or hindrance? The impact of harmonised standards on African exports. Journal of African Economies, 18(5), 711–744. Djankov, S., Freund, C., & Pham, C. (2010). Trading on time. The Review of Economics and Statistics, 92(1), 166–173. Dollar, D., & Kraay, A. (2004). Trade, growth, and poverty. The Economic Journal, 114, 22–49. Drogué, S., & DeMaria, F. (2012). Comparing apples with pears. how differences in pesticide residues regulations impact trade? Working Paper MOISA. Ehrich, M., & Hess, S. (2015). The ability of organisations to adopt foreign trade standards. GlobalFood Discussion Paper, (73). Ferro, E., Wilson, J. S., & Otsuki, T. (2015). The effect of product standards on agricultural exports. Food Policy, 50, 68–79. Fulponi, L. (2006). Private voluntary standards in the food system: The perspective of major food retailers in oecd countries. Food Policy, 31(1), 1–13. GlobalGAP. Globalgap annual reports. Head, K., & Mayer, T. (2014). Gravity equations: Workhorse, toolkit, and cookbook. In Handbook of international economics, (Vol. 4, pp. 131–195). Helpman, E., Melitz, M. J., & Rubinstein, Y. (2008). Estimating trade flows: Trading partners and trading volumes. Quarterly Journal of Economics, 73(2). Henson, S., & Humphrey, J. (2010). Understanding the complexities of private standards in global agri-food chains as they impact developing countries. The Journal of Development Studies, 46(9), 1628–1646. Henson, S., & Jaffee, S. (2008). Understanding developing country strategic responses to the enhancement of food safety standards. The World Economy, 31(4), 548–568. Herzfeld, T., Drescher, L. S., & Grebitus, C. (2011). Cross-national adoption of private food quality standards. Food Policy, 36(3), 401–411. International featured standard (2016). Jaffee, S., & Henson, S. (2004). Standards and agro-food exports from developing countries: Rebalancing the debate. World Bank Policy Research Working Paper, (3348). Kamal, F., & Sundaram, A. (2016). Buyer–seller relationships in international trade: Do your neighbors matter? Journal of International Economics, 102, 128–140.

Koenig, P., Mayneris, F., & Poncet, S. (2010). Local export spillovers in France. European Economic Review, 54(4), 622–641. Latouche, K., & Chevassus-Lozza, E. (2015). Retailer supply chain and market access: Evidence from French agri-food firms certified with private standards. The World Economy, 38(8), 1312–1334. Li, Y., & Beghin, J. C. (2012). A meta-analysis of estimates of the impact of technical barriers to trade. Journal of Policy Modeling, 34(3), 497–511. Maertens, A., & Barrett, C. B. (2013). Measuring social networks’ effects on agricultural technology adoption. American Journal of Agricultural Economics, 95(2), 353–359. Maertens, M., & Swinnen, J. (2009). Trade, standards, and poverty: Evidence from senegal. World Development, 37(1), 161–178. Mangelsdorf, A., Portugal-Perez, A., & Wilson, J. S. (2012). Food standards and exports: Evidence for China. World Trade Review, 11(03), 507–526. Martínez-Zarzoso, I. (2015). The trade effects of foreign aid: An empirical perspective. In B. Mak Arvin & B. Lew (Eds.), Handbook on the economics of foreign aid (pp. 179–196). Cheltenham: Edward Elgar Publishing. Masood, A., & Brümmer, B. (2014). Impact of globalgap certification on eu banana imports: A gravity modeling approach. GlobalFood Discussion Paper, (49). Mayer, T., & Ottaviano, G. (2007). The happy few: The internationalisation of european firms: New facts based on firm-level evidence. Bruegel Blueprint Series, (3). Melo, O., Engler, A., Nahuehual, L., Cofre, G., & Barrena, J. (2014). Do sanitary, phytosanitary, and quality-related standards affect international trade? Evidence from Chilean fruit exports. World Development, 54, 350–359. Müller, A., & Theuvsen, L. (2015). Food safety standards in the guatemalan fresh pea sector: The rolf of financial literacy in technology adoption. GlobalFood Discussion Paper, (68). Otsuki, T., Wilson, J.S., & Maskus, K. (1999). Quantifying the impact of technical barriers to trade: A framework for analysis. World Bank Policy Research Working Paper. Otsuki, T., Wilson, J. S., & Sewadeh, M. (2001). Saving two in a billion: Quantifying the trade effect of European food safety standards on African exports. Food Policy, 26, 495–514. Santos Silva, J., & Tenreyro, S. (2006). The log of gravity. The Review of Economics and Statistics, 88(4), 641–658. Santos Silva, J., & Tenreyro, S. (2011). Further simulation evidence on the performance of the Poisson pseudo-maximum likelihood estimator. Economics Letters, 112(2), 220–222. Schuster, M., & Maertens, M. (2015). The impact of private food standards on developing countries’ export performance: An analysis of asparagus firms in Peru. World Development, 66, 208–221. Shepherd, B., & Wilson, N. L. (2013). Product standards and developing country agricultural exports: The case of the European union. Food Policy, 42, 1–10. Swinnen, J., Deconinck, K., Vandemoortele, T., & Vandeplas, A. (2015). Quality standards, value chains, and international development. New York: Cambridge University Press. Vigani, M., Raimondi, V., & Olper, A. (2012). International trade and endogenous standards: The case of gmo regulations. World Trade Review, 11(03), 415–437. Wilson, J. S., Otsuki, T., & Majumdsar, B. (2003). Balancing food safety and risk: Do drug residue limits affect international trade in beef? Journal of International Trade & Economic Development, 12(4), 377–402. Wooldridge, J. (2010). Econometric analysis of cross section and panel data (2 ed.). Cambridge, MA: MIT Press. Yue, N., Kuang, H., Sun, L., Wu, L., & Xu, C. (2010). An empirical analysis of the impact of eu’s new food safety standards on China’s tea export. International Journal of Food Science & Technology, 45(4), 745–750.