The World Economy (2009) doi: 10.1111/j.1467-9701.2008.01150.x The Impacts of Alignment with Global Product Standards on Exports of Firms in Developing Countries STANDARDS, World TWEC Oxford, Original XXX 0378-5920 1467-9701 ©GALINA 2008Economy The UK Article AN Author AND MRAS KEITH Journal ANDE.EXPORTS compilation MASKUS © Blackwell Publishers Ltd. 2008 Blackwell Publishing Ltd Galina An1 and Keith E. Maskus2 1 Kenyon College and 2University of Colorado 1. INTRODUCTION S traditional trade barriers, such as tariffs and quotas, have diminished in importance, increasing attention has been paid to the trade impacts of regulatory policies, such as technical regulations and performance requirements. There can be considerable benefits from more harmonised global standards, including economies of scale, product compatibility and consumer acceptance of imports. At the same time, it is costly to comply with international technical regulations, which therefore might be impediments to trade.1 The compliance challenges are amplified for exporters in developing countries (DCs) that try to penetrate markets in industrial nations. Compliance costs may be significantly higher for such firms due to lower initial domestic standards. Nonetheless, compliance with industrial nations’ standards may provide export gains to firms in developing countries due to larger demand for their products and greater incentives to upgrade quality and production processes (Ganslandt and Markusen, 2001; Jaffee and Henson, 2004). In this paper we offer the first firm-level study of this empirical question, using data acquired from a detailed World Bank survey of exporters in several industries located in developing countries. We examine two types of government initiatives – mutual recognition agreements (MRAs) and alignment of domestic standards with international norms. An MRA is a pact between two or more countries that commits each party to recognise the results of product testing or certifications of the other in specified sectors. Thus, it does not harmonise A The authors are grateful to two anonymous referees for their comments and insights. 1 Surveys may be found in Sykes (1995), Baldwin (2000) and Maskus and Wilson (2001). 552 © 2009 The Authors Journal compilation © 2009 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA STANDARDS, MRAs AND EXPORTS 553 regulations, but rather recognises that each member’s testing procedures are adequate to guarantee product safety. It should liberalise trade by reducing the cost of duplicative testing, inspection and certification. In contrast, alignment means harmonising national regulations with international standards, such as the ISO 9000 system. This approach may increase market access by signalling to consumers in all foreign markets that certified products achieve reliable quality and safety. However, harmonisation imposes higher compliance costs than MRAs. It becomes an empirical question as to which approach is likely to raise exports by more (or reduce them by less). Our empirical contributions are two-fold. First, we employ a firm-level dataset (from the World Bank Technical Barriers to Trade Survey, described below) of exporters in developing countries. All previous studies used aggregate trade flows by industry. Second, we distinguish between two different types of policies: negotiating MRAs and aligning home measures with international standards. More specifically, we examine how MRAs and harmonisation of domestic regulations with international standards influence firm-level decisions on how much to export to five industrial regions (the United States, the European Union, Canada, Japan and Australia). We consider exports to developed countries because the issue of quality standards will most sharply arise there. Our econometric analysis finds that where poor countries negotiated MRAs there is a significant export-promoting effect for their firms, a finding that survives extensive robustness checks. However, alignment with international standards, though positively associated with such exports, does not have a significant impact. 2. RELATION TO PRIOR LITERATURE The empirical literature on the trade impacts of standards harmonisation is scarce, mostly due to data limitations and measurement problems. An early study was by Swann et al. (1996), who used counts of voluntary national and international standards recognised by the United Kingdom and Germany from the PERINORM database at the three-digit industry level.2 They found that increases in the number of British national standards raised both manufacturing imports and exports, while the number of international standards to which Britain was a party had a significantly positive effect on exports but not imports. Moenius (2004) extended that analysis by constructing a comprehensive dataset on industry-specific standards that existed in common between bilateral trading partners. He found that standards had large trade-promoting effects among the nations sharing them, while country-specific standards promoted trade in 2 PERINORM is a proprietary database listing detailed product regulations, primarily in developed countries. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 554 GALINA AN AND KEITH E. MASKUS manufacturing sectors but hindered trade in agriculture. He attributed the latter effect to information costs. Manufactured goods must be adapted to foreign markets, and country-specific standards offer information on particular requirements. In contrast, agricultural commodities may not require such adaptation and specific standards are exclusionary. Chen and Mattoo (2008) examined the effects of regional harmonisation within the European Union. In their empirical analysis they employed a data panel covering imports from 28 OECD and 14 developing countries at the threedigit level of manufacturing industries between 1986 and 2001. Using figures on the counts of harmonisation directives and MRAs at the industry level, the authors found that harmonisation increased intra-EU trade, while raising exports of industrialised non-member countries and reducing exports of most developing countries. However, a higher number of MRAs between the EU and non-member countries promoted trade among all parties, including developing nations. A few recent studies specifically address the effects of MRAs on trade. Baller (2007) studied standards policies among OECD countries between 1986 and 2003 in two sectors, telecommunication equipment and medical devices, in which MRAs and harmonisation coverage are widespread. He found that MRAs had a strong positive influence on both export probabilities and trade volumes for partner countries. Harmonisation, however, had a negligible effect on parties to the agreement. Vancauteren and Weiserbs (2003) used panel data for 15 EU countries between 1990 and 1998 and found that sectors with MRAs exhibited the smallest home bias among all industries subject to some kind of standards harmonisation. Amurgo-Pacheco (2007) examined the effects of MRAs between the EU and the United States on third countries in a set of pharmaceutical products. He found that MRAs are trade-promoting instruments among members, but diminish exports from other nations. Finally, a report to the European Commission (2003) that focused on the MRA between the EU and Australia found a positive but insignificant effect of the MRA in creating trade in four industries. Case studies of the effects of MRAs on trade are almost non-existent, perhaps because such agreements are relatively recent, especially with developing countries. Two surveys were undertaken by the European Commission (1998, 2002), though it is likely that trade impacts were biased downward by the industry selection. The results suggested that for straightforward products that pose few safety concerns, such as bicycles, tanks and containers, the MRA generally raised trade within the EU. On the other hand, the application of mutual recognition principles to technically complex products, such as buses, trucks and road safety equipment, or products with safety concerns, such as food supplements, had little impact. There are a few case studies of alignment with international standards by developing-country governments, though all focus on agricultural commodities. Stanford (2002) described efforts by the government and avocado growers in Mexico to impose international standards on producers in that industry, which © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 STANDARDS, MRAs AND EXPORTS 555 resulted in lifting of the USDA ban on avocado imports to the United States in 1997. Henson and Mutillah (2004) discussed Kenyan exports of Nile perch. The government of Kenya has invested in administrative structures, inspection and certification procedures, upgrades of laboratory facilities and training. While these costs have been considerable, there are significant benefits in terms of more effective quality control. Today most perch exporters have diversified their export base beyond the EU to Australia, Japan, the United States and other locations. Jaffee and Henson (2004) examined three case studies of exports of Guatemalan raspberries, Kenyan fresh produce and Peruvian asparagus, all of which showed that quality upgrading to international and industrial-countries standards contributed to export growth. The theoretical literature on standards and trade is more plentiful, though most of it focuses on the potential cost-raising impacts of alignment and harmonisation (Maskus and Wilson, 2001). Borrowing from elements of new trade theory, it is not difficult to understand why MRAs may be more export-promoting than international standards alignment. As emphasised by Melitz (2003), highly productive firms are more likely to enter international markets because they can overcome the fixed costs of exporting more readily than can inefficient firms. Taking this as his point of departure, Baller (2007) sets out a model in which MRAs reduce fixed exporting costs because they do not force local firms to invest in testing and certification procedures that are equivalent to those abroad, which is required under harmonisation. As a result, his model predicts a larger export-promoting impact from MRAs than alignment, a finding that is widely anticipated in the policy literature, at least as a matter of degree (Baldwin, 2000; Maskus and Wilson, 2001; Chen and Mattoo, 2008).3 3. DATA AND ECONOMETRIC SPECIFICATION Our econometric approach is driven in part by the nature of the available data. Thus, we first describe the standards data before discussing specification issues. a. The World Bank Survey In 2001–02 researchers in the Development Economics Group at the World Bank commissioned a major survey in order to address how familiar firms in developing countries were with international product standards and how they responded to them.4 That effort generated the World Bank Technical Barriers to 3 In his review of EU experience, Pelkmans (2003) adds a further advantage of MRAs, in that they are less subject to capture by those who might engage in costly overregulation. 4 See Maskus et al. (2001) for an extensive discussion of these issues. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 556 GALINA AN AND KEITH E. MASKUS Trade Survey, described in detail by Wilson and Otsuki (2004). The survey covers 689 firms in 25 industries in 17 developing countries (DCs) and transition economies for either 2000 or 2001. The primary goal of that project was to evaluate the impact of foreign standards and technical barriers to trade on the operations of firms in DCs. Firms were asked extensive questions about their knowledge of such regulations in five main developed regions (United States, EU, Canada, Japan and Australia) and their reactions to them. Questions also were posed about domestic government policies and whether they hindered or enhanced the respondent firm’s ability to export. The data we employ form a subset of the overall survey. Appendix Table A1 presents the detailed industry and country composition of our dataset after cleaning and omitting observations with missing values for the variables we analyse. Our firms cover both manufacturing and agricultural products. As can be seen in Appendix Table A1, 40.9 per cent of the observations in our sample come from India, while the representation of other countries in the sample varies from 0.7 per cent (Senegal) to 7.6 per cent (South Africa). In terms of industry composition, textiles and apparel constitute 29 per cent of the sample, followed by processed food and tobacco (12 per cent) and raw agricultural products (10 per cent). The percentage of other industries varies from 0.5 per cent (materials) to 7.1 per cent (miscellaneous manufacturing commodities). Our sample contains 421 firms, each of which reported their exports, if any, to the five developed regions mentioned above and total world exports. Values were converted into thousands of US dollars at contemporaneous exchange rates. Our two policy variables are taken from the survey as well. The first, which we call STANDARD, is the total number of international standards with which the country’s government has aligned its domestic regulations, as reported by firm managers. Specifically, firms were asked the following question: ‘With which international standards did your government choose to align their national regulations?’ The question listed as choices the International Standards Organization (ISO), International Electrotechnical Commission (IEC), International Telecommunications Union (ITU), Codex Alimentarius (CODEX), Office International des Epizooties (OIE), and the Food and Agricultural Organization’s International Plant Protection Convention (IPPC). For each choice the response was either yes or left blank. We constructed STANDARD as the sum of all positive responses. Thus, if a firm reported that its government had aligned domestic regulations with ISO only, STANDARD would equal one; if it had also adopted CODEX, STANDARD would be equal to two, and so on. Detailed information is available in a supplementary data appendix on request. In summary, approximately 70 per cent of the firms did not report any known alignment with international standards, while 26 per cent reported alignment with only one standard and only four firms mentioned three. Some firms in all countries reported at least one such alignment. At the © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 STANDARDS, MRAs AND EXPORTS 557 industry level, there was at least one firm in all sectors except petroleum, leather products and paper products that listed positive alignment with international standards. The largest absolute numbers of responses arose in textiles and apparel, followed by processed food and tobacco, and raw agricultural products. As a proportion of total firms in each industry claiming alignments, the highest were in instruments and photographic and optical goods (66 per cent) and primary metals (60 per cent). The second policy variable is MRA, which was constructed from the question: ‘Are any of your top five revenue-generating products subject to mutual recognition agreements in any of the following countries/regions: US, EU, Canada, Japan, Australia?’ Similar to STANDARD, variable MRA is the sum of positive responses, with the highest reported value being the maximum of five. To summarise, most firms did not report any MRAs (87.9 per cent). For example, no firms from India, Senegal and Iran operated under any MRA with the five developed countries. These governments have not placed emphasis on negotiating such agreements on behalf of exporting firms. Among those firms reporting agreements, 35 mentioned MRAs with one developed country and 16 reported MRAs with multiple regions. Overall, adherence with international standards is more prevalent in the data than membership in MRAs, perhaps because the former may be less strict than what developed countries require in the latter. Further, alignment with international standards is a unilateral decision, which is easier than concluding an MRA. Summary statistics for exports and the policy variables are in Table 2. b. Econometric Specification In our dataset 18 per cent of all firms do not export to any of our five developed regions. Thus, our main approach is to use Tobit estimation of the following equation: EXPijk = β′ Xijk + λzijk + ajDj + δkDk + εijk. This equation determines the amount of firm i’s exports (in log form) in industry j from country k, EXPijk, summed across five developed markets (United States, EU, Canada, Australia and Japan) as a function of firm characteristics, Xijk, and government policies toward compliance with the international standards, zijk.5 The equation also includes fixed effects for industry and exporting country. The Xijk vectors include several firm characteristics reported in the survey. The first is years since the enterprise was founded (AGE), since it is likely that older firms have established domestic and international markets. Second is fixed assets 5 Because exports were zero for some firms, we actually take logs of one plus exports. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 558 GALINA AN AND KEITH E. MASKUS (FIX_A) to control for the firm’s size.6 A third factor is the extent of foreign ownership (F_OWN ), which might enhance the ability to penetrate foreign markets. A fourth characteristic is the average growth rate of sales in the past three years (GR_PAST ). Finally, we control for the type of the firm through the use of dummy variables that indicate whether the firm is a state enterprise, private company, subsidiary of a multinational firm, joint venture, or other type. Variables AGE and FIX_A enter the estimation in natural logs. The industry dummies control for unobserved industry-specific characteristics, such as technological intensity, trade barriers in importing nations and market structure. The exporter country dummies reflect geographical characteristics, aggregate trade policies of local governments, unobserved treatment by industrial countries of a particular developing country, such as preferential trade agreements, and any other international trade considerations that differ by exporting country. We assume that εijk is a random disturbance term symmetrically distributed about zero and uncorrelated with the explanatory variables. c. Blank Entries A primary problem with the construction of our policy variables was how to treat blank entries on firms’ responses about MRAs and international standards. Those entries could either indicate that there were no such measures or that the firm managers were unaware of them. In principle, we would expect most such answers to mean the absence of measures since the firms in our sample are exporters or indicated an intention to export, suggesting that managers would be aware of product regulations in foreign markets. Nonetheless, we developed several versions of the policy variables based on different assumptions. First, we assumed that blank entries truly signify no MRAs (MRA = 0) or no incorporation of international standards (STANDARD = 0).7 Second, we assumed that the number of MRAs and standards should be similar by country and industry, so we assigned average values of MRA or STANDARD, calculated by country and industry code, to missing answers. Third, we used the maximum value of STANDARD and MRA by industry and country instead of average value. Finally, we experimented with assigning averages and maximum values to all firms, not just to those with blank answers, in our fourth and fifth versions, respectively. 6 We also tried both sales and employment to control for size and the results were qualitatively the same. However, due to concern about contemporaneous endogeneity of those variables we chose the size of fixed assets since investment decisions generating capital stocks are made in previous periods, making fixed assets predetermined. In our sensitivity section we treat fixed assets as endogenous and use instrumental variables to purge correlation with unobserved firm effects. As shown below, the main regression results are very similar. 7 Even if managers were unaware of existing measures, a blank entry would suggest they do not take such measures into account in their decision-making. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 STANDARDS, MRAs AND EXPORTS 559 Our primary reported regressions employ assumption two, assigning average values for MRA and STANDARD to the blank entries, because it seems to be most realistic. However, we show results for all versions for comparison purposes. d. Endogeneity and Instrumental Variables The most significant problem we face is that our sample is a cross-section, rather than a panel.8 This issue makes identification of the causal impacts of our policy variables on exports difficult. The reason is that STANDARD and MRA may be endogenous, since the government could decide to negotiate more MRAs and align with international standards in sectors of particular comparative advantage or where exports are growing rapidly. Note that MRA negotiations and harmonisation with international standards take years, suggesting that such policies might be considered predetermined. However, firms and sectors with significant exports in the cross-section may have been high-growth entities in the past, generating a channel from exports to policy. Alternatively, both the decisions to export by firms and government involvement with trade issues might be influenced by an unobserved third factor. We address endogeneity concerns by incorporating, as instrumental variables, several standard political-economy measures at the country–industry level. Specifically, the decision to negotiate MRAs and align with international standards may be influenced by a variety of economic and political factors. Thus, for economic instruments in the manufacturing sectors we extracted several series from the World Bank’s ‘Trade, Production and Protection 1976–2004’ dataset.9 That database contains exports, employment, investment and other measures by country and industry at the three-digit International Standard Industrial Classification (ISIC) level. We developed a concordance between the ISIC and our industry breakdown, which is available in a supplemental appendix. For the agricultural sector we compiled similar figures from the World Development Indicators. We use data from these sources to construct the following variables, which are sensible instruments for interests that could affect MRA and standards policy, while plausibly uncorrelated with the error terms in export regressions using data for individual firms, given that our sample covers only a small number of enterprises in each sector–country pair. First, IND_EXP_GR is the growth of industry exports for each country in our sample between 1989 and 1996. Since we do not have the information on when MRAs and international standards were enacted, we chose this period to make 8 Unfortunately, the World Bank decided not to replicate the survey in later years. These data were compiled by the Trade Research Group in Development Economics and are available at the World Bank website. The most recent figures contain data for all of the countries in our database. Some data are missing for particular years and industries, which is why we use averages. 9 © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 560 GALINA AN AND KEITH E. MASKUS sure that we are looking at the industry’s export growth before those initiatives came into effect. Second, COMP_ADV is the average industry’s revealed comparative advantage between 1989 and 1996, calculated as: ⎛ ⎜ EXPjk ⎝ ⎛ ⎜ ∑ EXPjk ⎝ j ⎞ ∑ EXPjk ⎟⎠ k ⎞ EXP ∑ ∑ jk ⎟⎠ j k . This measure captures country k’s share of industry j’s exports in total world exports of industry j, divided by country k’s share of total world exports. A higher ratio ‘reveals’ higher comparative advantage, though there are problems with that interpretation.10 Calculated at the industry level it is a sensible instrument for trade-policy interests of individual firms. Third, EMPL is the average industry employment share by country between 1989 and 1996, calculated as the ratio of industry employment over the country’s labour force.11 Finally, HERF is the Herfindahl index of industry concentration in the United States for 1992, acquired from the US Census Bureau. In addition to economic factors, domestic political structure can also influence the amount of government involvement in trade policy. For this reason we add to the pool of instruments two government structure variables. These include a democratisation index, derived from Gastil’s political rights index (Freedom House, various years), averaged over 1989–96, and the relative size of the government, defined as the share of government spending as a percentage of GDP between 1989 and 1996. We also incorporate interaction terms between our economic instruments and government structure variables. In the estimation, we choose the instruments used for each specification from the pool of constructed IVs using the following criteria. First, they have significant correlation with the policy variables. Second, they satisfy the overidentification tests of instrument exogeneity (Hansen’s J-test). Third, we consider whether the exclusion restrictions are appropriate by testing whether the instruments have no independent influence on the dependent variable, beyond that exercised through the policy variables. We check that by entering each instrument in the second stage and testing its significance. Precise definitions of all variables and their units are presented in Table 1, while summary statistics are in Table 2.12 10 See Ballance et al. (1987) for an extensive discussion and Balassa and Bauwens (1988) for an application. 11 Labour force figures were taken from the World Bank, World Development Indicators. 12 A correlation matrix is available in the supplementary data appendix. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 STANDARDS, MRAs AND EXPORTS 561 TABLE 1 Variable Definitions Description LHS variables EXP EXP_all RHS variables STANDARD MRA GR_PAST AGE FIX_A F_OWN Log of Exports to developed countries – EU, USA, Canada, Japan, Australia (exports are in 2001 US$1,000) Log of total exports (exports are in 2001 US$1,000) The number of international standards with which the country’s government chose to align their national regulations (ISO, IEC, ITU, CODEX, OIE, IPPC)* The sum of positive responses to the questions whether any of the top five products are subject to MRA with US, EU, Canada, Japan or Australia Average growth rate of sales in the past three years (%) Log of firm’s age = Log(2001 – year of establishment) Log of value of plant’s fixed assets (in 2001 US$1,000) Share of foreign ownership: 1 – no foreign ownership, 2 – less (or equal) than 50%, 3 – more than 50% foreign ownership Type of ownership (dummies)** PRIVN Headquarter local of a privately held, non-listed company PRIVL Headquarter local of a publicly traded or listed company SUBD Subsidiary/division of a domestic enterprise SUBM Subsidiary/division of a multinational firm JOINTD Joint venture of a domestic enterprise JOINTM Joint venture of a multinational firm STATE Completely or partially state-owned company Instruments IND_EXP_GR COMP_ADV Average growth of industry exports in each country between 1989 and 1996 Average industry j’s revealed comparative advantage for country k between 1989 and 1996, calculated as ⎛ ⎜ EXPjk ⎝ ⎛ ⎜ ∑ EXPjk ⎝ j EMPL HERF GOVT DEM ⎞ ∑ EXPjk ⎟⎠ k ⎞ ∑ ∑ EXPjk ⎟⎠ j k 10 Average industry employment share by country between 1989 and 1996, calculated as the ratio of industry employment over country’s labour force × 1,000 (number of employees per 1,000 of labour force) Herfindahl index of industry concentration in the US for 1992 (divided by 1,000) Average government expenditures as a share of GDP (in %) between 1989 and 1996 Average democracy index (= 8 – Gastil’s political rights index) between 1989 and 1996 Notes: * ISO – International Standards Organization. IEC – International Electrotechnical Commission. ITU – International Telecommunications Union. CODEX – Codex Alimentarius Commission. OIE – Office International des Epizooties. IPPC – FAO International Plant Protection Convention. ** The benchmark type of ownership is cooperative/collective. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 562 GALINA AN AND KEITH E. MASKUS TABLE 2 Summary Statistics Variable Obs. Mean Std. Dev. Min. Max. EXP EXP_all STANDARD MRA GR_PAST AGE FIX_A F_OWN PRIVN PRIVL SUBD SUBM JOINTD JOINTM STATE IND_EXP_GR COMP_ADV EMPL HERF GOVT DEM 421 412 421 421 421 421 421 421 421 421 421 421 421 421 421 421 421 421 421 421 421 5.096 6.613 0.356 0.185 0.194 2.794 6.398 1.235 0.565 0.164 0.007 0.036 0.026 0.019 0.014 0.309 0.034 0.152 0.948 14.071 4.657 3.226 2.312 0.595 0.593 0.325 0.832 2.285 0.593 0.496 0.371 0.084 0.186 0.160 0.137 0.119 0.362 0.119 0.129 0.397 4.160 1.201 0 0 0 0 −0.311 0.693 0.722 1 0 0 0 0 0 0 0 0.002 0.000 −0.531 0.255 8.854 2.125 11.758 12.476 3 5 3.000 4.956 13.201 3 1 1 1 1 1 1 1 2.158 0.845 0.727 1.873 21.755 6.500 4. ECONOMETRIC RESULTS We turn next to the presentation and discussion of the statistical analysis, beginning with the base case and working through endogeneity analysis and robustness checks. a. Basic Specification First, we estimate our regression equation using Tobit in the cross-section of firms summarised in the data section, but without instrumental variables. Table 3 presents our initial estimation results using version two (average values of MRA and STANDARD are assigned to the blank entries) of our policy variables. Here we include firm characteristics and country and industry dummies. Table 3 contains three specifications, with the policy variables entering separately and together. The estimation results in columns (1)–(3) show that both alignment with international standards (STANDARD) and negotiating mutual recognition agreements (MRA) are positively and significantly (at 5 per cent and 1 per cent significance level, respectively) associated with exports to the five developed regions. The © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 STANDARDS, MRAs AND EXPORTS 563 TABLE 3 Tobit Estimation of Export Equations (1) MRA STANDARD GR_PAST AGE FIX_A F_OWN PRIVN PRIVL SUBD SUBM JOINTD JOINTM STATE Log-likelihood Pseudo-R2 Observations (2) (3) 0.818*** (0.304) 0.736 (0.502) 0.184 (0.218) 0.610*** (0.088) 0.572 (0.371) −0.244 (0.520) −0.197 (0.636) −8.113*** (2.428) 1.664 (1.236) −1.223 (1.350) −0.837 (1.325) −1.360 (1.503) 0.816*** (0.231) 0.715** (0.300) 0.687 (0.494) 0.208 (0.214) 0.581*** (0.087) 0.552 (0.365) −0.218 (0.511) −0.283 (0.626) −7.306*** (2.392) 1.684 (1.216) −0.857 (1.332) −0.606 (1.306) −1.243 (1.485) 0.870*** (0.232) 0.752 (0.497) 0.211 (0.216) 0.594*** (0.088) 0.531 (0.368) −0.143 (0.514) −0.202 (0.630) −6.866*** (2.383) 1.652 (1.226) −0.855 (1.341) −0.195 (1.304) −1.489 (1.486) −952.57 0.0859 421 −955.91 0.0827 421 −949.76 0.0886 421 Notes: All equations include industry and country fixed effects. Standard errors are in parentheses. *** Significant at 1%, ** significant at 5%, * significant at 10%. magnitude of the coefficients is similar for both policy variables. In terms of economic effects, the coefficients in this regression imply that, on average, having an MRA with an additional developed country (or region) in our sample, or aligning with one additional international system of standards, is associated with approximately 122 per cent more exports of individual firms to these developed markets.13 The magnitudes of these coefficients, though large, seem plausible. Most of the firms in the sample list no or one MRAs with developed countries, so negotiating the first such agreement might have a large impact on the exports to industrial nations for a number of reasons. It could raise demand for a firm’s products by signalling to importers abroad that those products may be brought in without further certification. Further, entry into an MRA could substantially decrease the cost of exporting and there may be learning effects that can encourage exporting to other rich nations. Similarly, almost 70 per cent of the firms reported no adoption of international standards. Thus, a harmonisation policy could require them to undertake substantial investments in production processes and quality upgrading. Alternatively, since the direction of causality is not clear in this basic Tobit case, the government may have adopted international standards in sectors that were already exporting and complying with those norms. 13 We calculate percentage change in EXP as 100(e0.8 − 1) = 122.5 per cent. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 564 GALINA AN AND KEITH E. MASKUS Concerning other control variables, the size of fixed assets is positively associated with exports to developed countries. It also appears that subsidiaries of domestic companies tend to export less to developed regions. b. Endogeneity and Instrumental Variables Estimation Next we address the endogeneity concerns regarding our policy variables, while below we consider endogeneity of fixed assets. As discussed earlier, we develop a series of potential IVs from independent data sources at the World Bank. The second-stage IV–Tobit regression results are shown in Table 4.14 We note from the first-stage regressions (not shown here) that MRA is a function of several political-economic factors, including the Herfindahl index, democracy and the interaction between them, and interactions between industry export growth and the size of the government, and industry export growth and democracy. Specifically, MRAs are more likely to be negotiated in more concentrated industries and less democratic societies. In addition, for an average firm in our sample industry export growth is associated with a higher number of MRAs.15 Column (1) shows the second-stage regression for the case where only MRA is potentially endogenous. Of primary interest is that the coefficient for MRA remains positive and significant at the 10 per cent level. Its magnitude rises from 0.87 without instrumentation in Table 3 to 0.92 in this case. Thus, once we purge MRA of its political and economic determinants, its effect on exports to developed economies becomes even greater. The second column shows the corresponding IV–Tobit estimation accounting for endogeneity of STANDARD. First-stage results suggested that alignment with international standards is more prevalent in less democratic societies and for an average firm it is negatively associated with industry export growth. In terms of our main question, the second-stage results find that, unlike MRA, the magnitude of the coefficient on STANDARD falls with IV estimation. Further, the coefficient is not significant, suggesting that there is little evidence of a causal relationship from standards alignment to exports to developed countries. In this context, MRAs appear to be more determinative means of expanding exports than are international standards for firms in poor countries. Finally, the third column shows the results when both policy variables enter the IV–Tobit estimation. Here the coefficient for MRA is still positive and significant at the 10 per cent level, while that for STANDARD is not. Overall, our evidence suggests that alignment with international standards may not 14 These regressions were run with the ivtobit routine in Stata. In order to conserve space we do not show first-stage results here but they are available on request. 15 By average we mean a firm that comes from a country with the average share of government expenditure in relation to GDP (14 per cent) and the average democracy index (4.65). © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 MRA STANDARD FIXED_A MRA_ALL GR_PAST AGE FIX_A F_OWN PRIVN PRIVL SUBD SUBM JOINTD JOINTM STATE Observations Wald χ2(43) Hansen J-test (p-value) MRA STANDARD (1) (2) 0.924* (0.534) 0.606 (1.380) MRA and STANDARD (3) 0.969* (0.594) −0.231 (1.513) MRA and FIXED_ A (4) MRA_ALL (5) 1.001* (0.589) 0.717* (0.403) 0.748 (0.497) 0.212 (0.216) 0.592*** (0.089) 0.529 (0.368) −0.141 (0.514) −0.209 (0.632) −6.827*** (2.406) 1.654 (1.225) −0.836 (1.351) −0.184 (1.307) −1.474 (1.490) 421 284,045.42 0.90767 0.756 (0.519) 0.184 (0.217) 0.614*** (0.092) 0.568 (0.371) −0.221 (0.537) −0.171 (0.654) −8.046*** (2.457) 1.648 (1.237) −1.235 (1.350) −0.721 (1.512) −1.443 (1.598) 421 36,567.00 0.34229 0.770 (0.518) 0.213 (0.217) 0.593*** (0.092) 0.527 (0.372) −0.116 (0.542) −0.192 (0.652) −6.915*** (2.512) 1.622 (1.234) −0.856 (1.358) −0.059 (1.582) −1.586 (1.602) 421 2,046.69 0.82703 0.709 (0.514) 0.131 (0.336) 0.450 (0.446) −0.170 (0.523) −0.489 (1.088) −7.064*** (2.525) 1.491 (1.333) −0.835 (1.355) −0.344 (1.406) −1.611 (1.557) 421 7,627.15 0.81440 0.612 (0.643) 0.598** (0.272) 0.136 (0.126) 0.636*** (0.055) 0.249 (0.198) 0.128 (0.284) 0.454 (0.344) 0.183 (1.104) 1.117* (0.660) −0.756 (0.780) 0.153 (0.710) −0.441 (0.772) 412 416.92 0.67357 STANDARDS, MRAs AND EXPORTS © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 TABLE 4 IV–Tobit Estimation of Second-Stage Export Equations Notes: Column headings indicate variables subject to IV estimation in first stage. All equations include industry and country fixed effects. Standard errors are in parentheses. *** Significant at 1%, ** significant at 5%, * significant at 10%. 565 566 GALINA AN AND KEITH E. MASKUS promote exports to developed countries as much as negotiation of MRAs. For further perspective, we present in Appendix Table A2 descriptive statistics for two groups of firms: those covered by MRAs and those not covered. As the figures reveal, firms that report the existence of MRAs are younger and larger on average and have a larger share of foreign ownership. None is state owned or a subsidiary of a domestic enterprise. However, a larger portion is either a subsidiary of a multinational firm or part of a joint venture. It seems that, at least in this sample, having foreign ownership or a JV partnership is correlated with the ability to use an MRA for export growth. This outcome may be due to local subsidiaries having easier access to higher-quality standards from their investment partners. There also may be greater trust in product quality if a firm has foreign business partners. c. Sensitivity Checks Given that our policy variables are measured somewhat indirectly, it is important to subject the basic results to sensitivity and robustness checks. As mentioned above, we also ran the same types of regressions using the other four versions of our policy variables, with zero entries treated in a different fashion. In Table 5 we summarise the coefficients only for the policy variables, since the results for other regressors were almost identical.16 Coefficient estimates of the policy variables are similar for versions (2)–(5). In version one the coefficients for MRA in the export equation are almost double those in the other versions and suggest unrealistically high trade elasticities. The main conclusions are consistent across the other treatments of zero entries in computing the policy variables. In the remaining analysis we only employ version two, which, as we argued above, seems the most natural construct and has the advantage over versions four and five of preserving variation across firms, rather than just across industries and countries. As a second sensitivity check, we introduced country–industry interaction dummies in addition to country and industry dummies, a rigorous specification of controls. Since our data have a limited set of observations, for this purpose we aggregated countries into five regions (Eastern Europe, Latin America and Caribbean, Middle East, South Asia and Sub-Saharan Africa) and industries into five groups (raw food, processed food and related goods, equipment, materials, and textiles). This resulted in 32 dummy variables (24 interaction, four industry and four country dummies) in our regression equations. We simply summarise the results here. The MRA coefficients remain positive and significant in all equations, though again only marginally so in the IV–Tobit case with both policy variables included. Thus, our earlier conclusion holds up: MRAs are effective 16 Full results are available on request. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 EXP Tobit (1) Version 1: Blank entries treated as zero MRA 1.229*** (0.290) STANDARD Pseudo-R2 Wald χ 2 0.0878 IV–Tobit (2) (3) (1) 2.550* (1.401) 0.802*** (0.290) 0.0829 1.174*** (0.289) 0.717*** (0.285) 0.0908 1,155.20 Version 2: Blank entries assigned the mean value for MRA and STANDARD by industry and country MRA 0.870*** 0.816*** 0.924* (0.232) (0.231) (0.534) STANDARD 0.818*** 0.715** (0.304) (0.300) Pseudo-R2 0.0859 0.0827 0.0886 Wald χ 2 284,045.42 Version 3: Blank entries assigned the maximum value for MRA and STANDARD by industry and country MRA 0.681*** 0.623*** 0.931* (0.200) (0.200) (0.513) STANDARD 0.784*** 0.676** (0.287) (0.285) Pseudo-R2 0.0848 0.0828 0.0875 10,571.31 Wald χ 2 (2) (3) 1.515 (2.328) 2.683* (1.594) −0.395 (2.663) 3,375.83 702.90 0.606 (1.380) 0.969* (0.594) −0.231 (1.513) 36,567.00 2,046.69 0.585 (1.277) 0.979* (0.601) −0.201 (1.460) 37,199.82 1,798.74 STANDARDS, MRAs AND EXPORTS © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 TABLE 5 Summary of Coefficients for Different Versions of STANDARD and MRA in Export Equations 567 568 TABLE 5 Continued Tobit (1) IV–Tobit (2) © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 (3) (1) Version 4: STANDARD and MRA are country–industry means MRA 0.896*** (0.234) STANDARD 0.855*** (0.317) Pseudo-R2 0.0862 0.0828 Wald χ 2 0.833*** (0.234) 0.724** (0.314) 0.0888 0.911* (0.525) Version 4: STANDARD and MRA are country–industry maxima MRA 0.788*** (0.203) STANDARD 0.854*** (0.278) Pseudo-R2 0.0864 0.0838 Wald χ 2 0.694*** (0.205) 0.674** (0.279) 0.0892 2.93e+06 Notes: All equations include industry and country fixed effects. Standard errors are in parentheses. *** Significant at 1%, ** significant at 5%, * significant at 10%. (2) (3) 0.618 (1.351) 0.968* (0.595) −0.275 (1.510) 31,876.26 2,014.93 0.591 (1.253) 0.923 (0.583) −0.118 (1.438) 20,082.73 2,351.96 0.893* (0.485) 56,985.58 GALINA AN AND KEITH E. MASKUS EXP STANDARDS, MRAs AND EXPORTS 569 policy initiatives for export promotion, while alignment with international standards does not seem to have much impact on firm-level manufacturing exports to developed countries. Next, we address another endogeneity concern that relates to the size of a firm, measured here by the value of fixed assets. Though the size of fixed assets is in principle predetermined from prior investments, there might still be unobserved firm characteristics that influence both the size of the firm and the amount of exports. One example is having ties with the government, which might be helpful in both export growth and asset expansion. Furthermore, a firm might make export decisions based on the discounted present value of future assets. Thus, we treat the fixed assets variable also as endogenous. The results for the case where we treat MRA and fixed assets as endogenous are in the fourth column of Table 4. In the second stage we find that the coefficient on MRA is quite similar to that in the first column. To conserve space we do not show the results for the cases with STANDARD included, for those coefficients are consistent with earlier findings. As a fourth robustness exercise we consider potential heteroscedasticity in the second-stage IV–Tobit regression residuals. When such residuals have unequal error variances, the consistency of Tobit estimates is in question (Wooldridge, 2003). The most likely source of heteroscedasticity is variation in firm size, as captured by fixed assets. Thus, to address this problem we first regress the squared residuals from the IV–Tobit specification of the export equation with MRA as the only policy variable, taken from Table 4. This regression yields: 2 = eijk 0.0896 0.5539 + * FIX _ Aijk + uijk . (0.1430) (0.0211) This result suggests there is positive correlation between squared residuals and the size of fixed assets. Accordingly, we scale all variables by the square root of (0.5539 + 0.0896 × FIX_A) and run the new regression equation, verifying that the squared residuals in the new specification are not statistically significantly related to asset size. The coefficient estimate on MRA in the scaled regression is 1.062, similar to those in Table 4, and remains significant at the 10 per cent level. This finding confirms that our main results still hold and the size of the coefficient for MRA is close to that in the basic specification. A further question to check is whether our results are dominated by the high number of Indian firms in the data sample. As noted earlier, such firms comprise nearly 41 per cent of the firms overall, yet report no MRAs in their survey responses. Running the IV–Tobit estimation after removing Indian enterprises, we find that the responsiveness of exports to the number of MRAs is somewhat larger than that in the base case in the first column of Table 4 and remains significant. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 570 GALINA AN AND KEITH E. MASKUS As a final check, we also consider the relationship between our policy variables and total exports, rather than just exports to the five developed regions. In these specifications, MRA includes all mutual recognition agreements, not just those with the developed countries. As shown in the final column of Table 4, the IV coefficient on MRA becomes insignificant. That the responsiveness of exports to all MRAs is lower when we include all export destinations accords with intuition, because standards for achieving mutual recognition are generally less rigorous in other developing countries than in developed markets. In turn, the impact on consumer acceptance in export destinations may be less. 5. CONCLUDING REMARKS This paper offers the first firm-level econometric evidence on the effects of aligning domestic product standards with international norms and of negotiating mutual recognition agreements on firm-level manufacturing exports of developing nations. Without considering endogeneity of the policy variables, we found that both measures are associated with more exports to developed countries. However, when we apply instrumental variables to deal with the problem of endogeneity we found that the influence of MRAs remains significantly positive, but that of standards alignment becomes insignificant. Note that there is no evidence from this analysis that linking with international standards actually causes exports from developing-country firms to be reduced through higher costs. These results are consistent with the findings in Baller (2007). On this evidence, it seems that MRAs are particularly stimulating for exports, presumably because they reduce the direct costs of exporting associated with certification and testing. Alignment with international standards, such as ISO and CODEX, has lesser ability to expand exports from poor countries, perhaps because those norms may be insufficient to meet the requirements of developedcountry markets. These results suggest that if governments of developing countries are interested in increasing exports to developed countries’ markets, resources could be spent on negotiating MRAs. Of course, in making such choices governments are constrained by their available resources and often it is costly to negotiate comprehensive MRAs. Further, it is not evident from these results that this approach is the first-best means of expanding exports. Overall, however, the findings in this paper support the notion that some forms of linkage to international product norms can enhance export performance of individual firms, rather than serve as export barriers. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 TABLE A1 Industry and Country Composition Code Industry Number of Observations ARG 1 2 3 4 5 6 7 8 9 11 12 13 14 15 16 18 19 20 22 BLG CHL CZE Raw agricultural products 1 1 7 1 Meat and fish products, livestock 1 1 0 1 Electrical and electrical equipment 1 2 0 6 Fabricated metal 1 1 0 0 Industrial machinery and equipment 1 1 0 0 Industrial or agricultural chemicals 2 3 2 5 Instruments and photographic and 0 0 0 1 optical goods, watches and clocks Leather and leather products 0 1 0 0 Paper and allied products 2 0 0 0 Processed food and tobacco 6 2 2 3 Rubber and plastic products 4 0 2 0 Telecommunications and 0 0 0 3 terminal equipment Textiles and apparel 1 5 1 3 Transportation equipment and 4 1 0 5 automotive parts, and dealers Lumber, wood and furniture 0 0 1 3 Primary metal and metallic ores 0 0 0 0 Petroleum and other 0 0 0 0 non-metallic minerals Miscellaneous manufactured 0 0 0 0 commodities Material 0 2 0 0 Total 24 20 15 31 Per cent 5.70 4.75 3.56 7.36 HND IND IRN JOR KEN NGA PAK PAN POL SEN ZAF UGA Total Per Cent STANDARDS, MRAs AND EXPORTS © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 APPENDIX 1 1 0 0 0 0 0 4 0 9 6 7 1 1 0 0 1 0 0 4 0 2 0 0 1 0 1 0 12 0 0 0 0 0 0 8 0 0 0 0 0 0 0 1 0 0 0 0 1 0 3 0 0 0 1 0 2 0 5 0 0 3 0 0 0 0 0 0 2 0 0 0 1 6 1 2 0 3 2 1 0 0 0 0 42 10 26 15 10 26 3 9.98 2.38 6.18 3.56 2.38 6.18 0.71 0 0 2 0 0 15 0 11 1 1 1 0 1 0 0 0 0 7 2 0 0 0 3 0 0 0 1 3 1 0 2 0 0 1 0 0 0 2 2 0 0 0 2 0 0 0 0 0 0 0 0 0 4 3 1 1 0 3 0 0 20 3 51 16 5 4.75 0.71 12.11 3.80 1.19 1 0 87 4 2 0 3 0 0 0 1 0 6 0 1 0 6 3 0 0 5 4 0 0 122 21 28.98 4.99 1 0 0 0 0 0 1 0 2 0 1 3 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 2 2 0 1 1 6 5 8 1.43 1.19 1.90 0 25 1 0 1 0 1 0 0 1 1 0 30 7.13 0 0 0 0 0 0 0 6 172 13 20 16 14 12 1.43 40.86 3.09 4.75 3.80 3.33 2.85 0 0 9 22 2.14 5.23 0 0 0 2 0.48 3 32 12 421 100.00 0.71 7.60 2.85 100.00 571 Note: ARG: Argentina; BLG: Bulgaria; CHL: Chile; CZE: Czech Republic; HND: Honduras; IND: India; IRN: Iran; JOR: Jordan; KEN: Kenya; NGA: Nigeria; PAK: Pakistan; PAN: Panama; POL: Poland; SEN: Senegal; ZAF: South Africa; UGA: Uganda. 572 Variable © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 EXP EXP_all STANDARD GR_PAST AGE FIX_A F_OWN PRIVN PRIVL SUBD SUBM JOINTD JOINTM STATE Reported MRAs Did Not Report MRAs Obs. Mean Std. Dev. Min. Max. Obs. Mean Std. Dev. Min. Max. 51 51 51 51 51 51 51 51 51 51 51 51 51 51 6.821 7.46 0.588 0.23 2.602 7.179 1.51 0.608 0.176 0 0.078 0.059 0.039 0 2.248 2.135 0.726 0.31 0.738 2.056 0.809 0.493 0.385 0 0.272 0.238 0.196 0 0.403 1.785 0 −0.31 0.693 3.398 1 0 0 0 0 0 0 0 11.76 12.48 3 1.5 4.644 13.2 3 1 1 0 1 1 1 0 370 361 370 370 370 370 370 370 370 370 370 370 370 370 4.858 6.494 0.324 0.189 2.821 6.291 1.197 0.559 0.162 0.008 0.030 0.022 0.016 0.016 3.270 2.314 0.568 0.328 0.842 2.297 0.547 0.497 0.369 0.090 0.170 0.146 0.126 0.126 0 0 0 −0.200 0.693 0.722 1 0 0 0 0 0 0 0 11.724 11.724 3 3.000 4.956 12.235 3 1 1 1 1 1 1 1 GALINA AN AND KEITH E. MASKUS TABLE A2 Descriptive Statistics for Firms According to MRA Experience STANDARDS, MRAs AND EXPORTS 573 REFERENCES Amurgo-Pacheco, A. (2007), ‘Mutual Recognition Agreements and Trade Diversion: Consequences for Developing Nations’, HEI Working Paper No. 20-2006 (Economics Section, The Graduate Institute of International Studies). Balassa, B. and L. Bauwens (1998), Changing Trade Patterns in Manufactured Goods: An Econometric Investigation (Amsterdam: North-Holland). Baldwin, R. (2000), ‘Regulatory Protectionism, Developing Nations and a Two-Tier World System’, Centre for Economic Policy and Research Discussion Paper No. 2574 (London). Ballance, R. H., H. Forstner and T. Murray (1987), ‘Consisency Tests of Alternative Measures of Comparative Advantage’, Review of Economics and Statistics, 69, 1, 157–61. Baller, S. (2007), ‘Trade Effects of Regional Standards Liberalization. A Heterogeneous Firm’s Approach’, World Bank Policy Research Working Paper No. 4124 (World Bank). Chen, M. X. and A. Mattoo (2008), ‘Regionalism in Standards: Good or Bad for Trade?’, Canadian Journal of Economics, 41, 3, 838–63. European Commission (1998), Technical Barriers to Trade, Volume 1 of Sub-Series III: Dismantling of Barriers of the Single Market Review (Luxembourg: Office for Official Publications). European Commission (2002), Report from the Commission to the Council, the European Parliament and the Economic and Social Committee. Second Biennial Report on the Application of the Principle of Mutual Recognition in the Single Market (Brussels: European Commission). European Commission (2003), The Economic Impact of Mutual Recognition Agreements on Conformity Assessment: A Review of the Costs, Benefits, and Trade Effects Resulting from the European Community MRAs Negotiated with Australia and New Zealand, Final Report (Brussels: European Commission). Freedom House (various years), Freedom in the World: Political Rights and Civil Liberties (New York: Freedom House). Ganslandt, M. and J. R. Markusen (2001), ‘Standards and Related Regulations in International Trade: A Modeling Approach’, in K. E. Maskus and J. S. Wilson (eds.), Quantifying the Impact of Technical Barriers to Trade: Can It Be Done? (Ann Arbor: University of Michigan Press). Henson, S. and W. Mutillah (2004), ‘Kenyan Exports of Nile Perch: Impact of Food Safety Standards on an Export-Oriented Supply Chain’, World Bank Policy Research Working Paper No. 3349 (World Bank). Jaffee, S. and S. Henson (2004), ‘Standards and Agro-Food Exports from Developing Countries: Rebalancing the Debate’, The World Bank Policy Research Working Paper No. 3348 (World Bank). Maskus, K. E. and J. S. Wilson (2001), ‘A Review of Past Attempts and the New Policy Context’, in K. E. Maskus and J. S. Wilson (eds.), Quantifying the Impact of Technical Barriers to Trade: Can It Be Done? (Ann Arbor: University of Michigan Press). Maskus, K. E., T. Otsuki and J. S. Wilson (2001), ‘An Empirical Framework for Analysing Technical Regulations and Trade’, in K. E. Maskus and J. S. Wilson (eds.), Quantifying the Impact of Technical Barriers to Trade: Can It Be Done? (Ann Arbor: University of Michigan Press). Melitz, M. (2003), ‘The Impact of Trade on Intra-Industry Reallocations and Aggregate Productivity Growth’, Econometrica, 71, 6, 1695–725. Moenius, J. (2004), ‘Information versus Product Adaptation: The Role of Standards in Trade’, Kellogg School of Management Working Paper (Northwestern University). Pelkmans, J. (2003), ‘Mutual Recognition in Goods and Services: An Economic Perspective’, European Network of Economic Policy Research Institutes, Working Paper No. 16. Stanford, L. (2002), ‘Constructing “Quality”: The Political Economy of Standards in Mexico’s Avocado Industry’, Agriculture and Human Values, 19, 4, 293–310. Swann, P., P. Temple and M. Shurmer (1996), ‘Standards and Trade Performance: The UK Experience’, Economic Journal, 106, 1297–313. © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009 574 GALINA AN AND KEITH E. MASKUS Sykes, A. O. (1995), Product Standards for Internationally Integrated Goods Markets (Washington, DC: Brookings Institution). Vancauteren, M. and D. Weiserbs (2003), ‘The Impact of Removal of Technical Barriers to Trade on Border Effects and Intra-Trade in the European Union’, Working Paper (Université Catholique de Louvain). Wilson, J. S. and T. Otsuki (2004), ‘Standards and Technical Regulations and Firms in Developing Countries: New Evidence from a World Bank Technical Barriers to Trade Survey’, World Bank Working Paper (World Bank). Wooldridge, J. M. (2003), Introductory Econometrics: A Modern Approach, 2nd edn (Mason, OH: Thomson South-Western). © 2009 The Authors Journal compilation © Blackwell Publishing Ltd. 2009