WPS8600 Policy Research Working Paper 8600 Trade Liberalization and Integration of Domestic Output Markets in Brazil Jose Guilherme Reis Mariana Iootty Jose Signoret Tanja Goodwin Martha Licetti Alice Duhaut Somik Lall Macroeconomics, Trade and Investment Global Practice October 2018 Policy Research Working Paper 8600 Abstract This paper describes how different policy distortions have and discusses policy options to promote better allocation of been impeding better integration of Brazil’s external and resources across the economy. The main conclusion of the internal product markets and discusses how these distor- paper is that Brazil could gain significantly from opening tions have prevented domestic firms from benefiting from to foreign trade. Yet, for Brazil to take full advantage of multiple sources of efficiency gains. The paper first focuses the opportunities that external integration offers, domestic on the costs of barriers to global integration, followed by markets also need to function better, so it is key to ensure an overview of policy induced stringencies hampering that the removal of external barriers to integration is coor- domestic integration. Drawing from general and partial dinated with the removal of internal distortions to domestic equilibrium analyses, the paper also provides evidence of market integration. potential impacts of removing some of those distortions This paper is a product of the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/research. The authors may be contacted at miootty@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Trade Liberalization and Integration of Domestic Output Markets in Brazil1 Jose Guilherme Reis, Mariana Iootty, Jose Signoret, Tanja Goodwin, Martha Licetti, Alice Duhaut and Somik Lall JEL codes: F13, F15, D24 Keywords: productivity, trade liberalization, market integration  This paper was prepared as a background paper for the “Jobs and Growth: Brazil’s Productivity Agenda” report, 1 which was launched in June 2018. 1. Introduction Opening to trade can boost productivity growth through numerous channels. The literature identifies two main groups of mechanisms through which trade can affect productivity: i) those that induce changes within firms and hence impact firm level productivity; and ii) mechanisms that trigger intra-industry reallocation of resources towards more productive firms, thereby increasing aggregate productivity. Improvements in firm level productivity caused by within firm changes can be triggered by exposure to competition stemming from output tariff reduction (the so-called import competition channel) and are associated with (observable) actions, for example, as investment in new technologies, increase in R&D, and entry in export markets.2 Within firm improvements in productivity can be also triggered by reduction of input tariffs - which allows firms to access better quality, higher variety and lower priced inputs (the so-called input channel) - or by reallocation of firm product mix towards more profitable products.3 Trade liberalization can affect aggregate productivity as well, as trade shocks are expected to reshuffle market shares towards more efficient firms,4 in this case, the magnitude of this reallocation process depends on how dispersed productivity was prior to trade liberalization shocks. Empirical evidence for Brazil has shown that the unilateral trade liberalization episode of the late 1980s - early 1990s has brought positive payoffs to productivity. Using sector level data for the 1985/97 period, Rossi Jr and Cavalcanti Ferreira (1999) show that a 10 percent reduction of import tariffs is associated with an increase in labor productivity growth of 0.88 percent per year and of TFP growth by 3.3 percent per year at the sectoral level. Muendler (2004) uses firm level data for the manufacturing industry in the 1986-1998 period to analyze how the reduction of inward trade barriers between 1990 and 1993 affected productivity and finds evidence that foreign competition pressures firms to raise productivity markedly, whereas the use of foreign inputs plays a minor role for productivity change. Schor (2004) runs a firm level analysis also for manufacturing industry and for the same period and finds evidence that along with higher competition pressure exerted by output tariff reduction, new access to better inputs also contributes to enhance productivity after trade liberalization. Lisboa, Menezes Filho and Schor (2010) use 1988-1998 data and confirm that the trade liberalization episode that took place in Brazil in the late 1980s and early 1990s brought positive impacts for productivity in the manufacturing industry and that the main driver of productivity growth was the reduction of input tariffs. Productivity dividends accruing from trade liberalization can be maximized when domestic markets are integrated in a way that resources are allocated to their most efficient uses. Easing barriers to trade is expected to trigger an intensive reallocation and churning process where resources are expected to move to more productive uses, within and between firms, sectors and regions, therefore boosting productivity growth. In this context, resource reallocation can be dampened by policy interventions. For example, Restuccia and Rogerson (2017) provide a 2 See for instance Bustos (2011) and Aw, Roberts and Xu (2011) as examples of studies covering this type of effect. 3 See Fernandes (2007), Topalova and Khandelwal (2011), and Halpern, Koren and Szeidl (2015) as examples of studies assessing the relevance of the input channel, while Bernard, Redding and Schott (2010) provide empirical evidence for the within-firm reallocation process. 4 See Melitz (2003) and Melitz and Ottaviano (2008) for a theoretical and empirical investigation on this aspect, and Pavcnik (2002) for empirical evidence on the productivity dividends coming from the reallocation effects caused by trade liberalization in Chile. 2    compelling review of how policy interventions (or lack thereof) can distort resource allocation; three main policy causes can be listed: statutory provisions that vary with firm characteristics,5 discretionary provisions favoring specific firms6 and market frictions.7 Against this backdrop, this paper aims at shedding light on the potential gains from trade liberalization in Brazil while discussing the (policy induced) stringencies that have been limiting domestic integration and distorting markets. The paper is organized in five sections besides this introduction. Section 2 shows how closed to trade the Brazilian economy is and highlights the key policy interventions that lie behind these results. Section 3 presents the potential gains accruing from trade liberalization and points to opportunities to boost productivity through trade integration. Section 4 sheds light on domestic market integration and discusses the (policy created) stringencies that have been affecting resource allocation in Brazil; three main areas are highlighted: infrastructure, use of distortive and ineffective business support policies and product market regulation and competition law enforcement. Section 5 shows the potential productivity benefits accruing from stronger competition. Section 6 concludes. 2. How closed to trade is the Brazilian economy and which policies lie behind this result? Openness to trade in Brazil is limited. Trade openness (measured as trade of goods and services over GDP) in Brazil, considering the level of per capita income, is below the predicted line, with no signs of improvement (Figures 1 and 2). In 2017, the ratio of exports plus imports over GDP was 24.11 percent relative to the world average of 56.1 percent. Although larger economies do tend to be more dependent on their domestic markets and export less, Brazil is the least open country and significantly below its benchmarked openness based on different econometric specifications and even after controlling for country size and distance to main partners (Lederman et al, 2014). Figure 1: Trade to GDP: 2000-2011(%)  Figure 2: Trade to GDP: 2012-2017 (%)  400 400 300 300 Trade to GDP(%)avg2012-17 Trade to GDP(%)avg2000-11 200 200 100 100 Brazil 0 Brazil 3 3.5 4 4.5 5 Log GDP per capita (PPP adjusted -avg 2012-17) 0 3 3.5 4 4.5 5 Log GDP per capita (PPP adjusted -avg 2000-11) Source: World Bank staff elaboration using WDI data Source: World Bank staff elaboration using WDI data                                                              5 For example, provisions of the tax code that vary with firm size; employment protection measures; product market regulation limiting size or market access; and tariffs applied to specific categories of goods. 6 Discretionary provisions made by the government or other entities (such as banks) that favor or penalize specific firms; for instance, subsidies, tax breaks, low interest loans granted to specific firms. 7 For example. monopoly power, market frictions, and enforcement of property rights. 3      Consistent with a limited trade openness, Brazil’s participation in global value chains (GVCs) is low and is relatively stronger on the seller (forward) side than on the buyer (backward) side. GVCs represent a new path for trade, whereby a country does not need the capability to produce an entire export good but instead contributes a segment of its production process. The latest OECD- WTO TiVA data for 2011 suggest that Brazil’s participation in GVC is reduced when compared to international peers and is relatively stronger on the seller (forward) side than on the buyer (backward) side.8 Figure 3. Participation in GVC Brazil vs comparator countries, 2011 (value added as share of domestic and foreign gross exports) 50 45 40 35 30 25 20 15 10 5 0 Forward Backward Source: World Bank staff elaboration using OECD TiVA data The limited openness to trade reflects the use of protectionist trade policies. This includes high tariff barriers to import. Brazil’s average (trade-weighted) tariff rate was 8.3 percent in 2015, the highest rate in comparison to other emerging and advanced economies (Figure 4).9 Overall, apart from the trade liberalization episode of the early 1990s, when tariffs fell from extremely high levels of 90 percent to 20 percent in wearing apparel, and except for small occasional amendments – such as the inclusion of tariff hikes in 1995 to account for the Mercosur list of exceptions, the generalized tariff increase in 1997 and the temporary rise (for 100 products) in 2012 - the import tariff structure in Brazil has not seen major changes since 2004 (Figure 5).                                                              8  The first measure captures the foreign value-added content embodied in Brazil’s gross exports, while the second one is measured as domestic value added embodied in the third-party country’s gross exports. 9 This number considers bilateral preferences. The simple average MFM tariff rate was 13.5 for Brazil in 2016. 4      Figure 4. Average ad-valorem equivalent of Figure 5. Import tariffs in Brazil: tariffs and NTMs, 2015: Brazil vs selected maximum, average, mode and minimum peers 25.0 20.0 15.0 10.0 5.0 0.0 Effective tariffs NTMs Source: estimations using UNCTAD TRAINS and UN Source: Rios and Motta Veiga (2014) COMTRADE data In addition, use of restrictive non-tariff measures (NTMs) is widespread. Beyond tariffs, NTMs and procedural obstacles in Brazil are widespread, raising the costs of trade. The ad-valorem equivalent of NTMs – a measure of effective restrictiveness of NTMs – is almost 12 percent (Figure 4) according to estimations using UNCTAD TRAINS and UN COMTRADE data. In addition, the coverage ratio, or percentage of imports subject to at least one NTM, is high in Brazil when compared to other countries: the percentage of imports subject to sanitary and phytosanitary measures and technical barriers is 66 percent and 89 percent respectively, well above the world average of 26 percent and 61 percent respectively (Figure 6). While certain NTMs, in particular with respect to sanitary and phytosanitary measures and technical barriers, may serve legitimate purposes, others might be driven by protectionist interest.10                                                                   10 Estimates used in simulation analyses, discussed later in this paper, suggest that NTMs in Brazil further increase import cost, on average, by about 12 percent. 5      Figure 6. NTM coverage: Brazil vs other Figure 7. Services trade restrictiveness index, countries, 2015 (%) Brazil relative to LAC average 100.0 89.0 70 60 80.0 50 66.4 61.1 64.6 40 30 60.0 47.5 20 10 40.0 0 25.8 22.6 20.0 5.9 0.0 Sanitary and Technical Quantity Price phytosanitary barriers controls controls Brazil Other Brazil LAC average Source: World Bank staff elaboration using UNCTAD Source: World Bank staff elaboration using World Bank STRI TRAINS data and UN COMTRADE data dataset Note: coverage ratios capture how much trade is subject to a Note: higher values mean more restrictive regimes. NTM measure. It does not reflect the level of NTM restrictiveness, just the incidence. It doesn’t include local content requirement. Among NTMs, Brazil is imposing local content requirements (LCRs) to an increasing number of products. For instance, in 2012, the Brazilian authorities issued regulations related to the industrial and trade regime for the automotive sector. The use of local content requirements was also prevalent in the oil and gas sectors, a tendency reversed in the more recent licensing rounds. A recent analysis conducted by Stone, Messent and Flaig (2015) maps the use of LCR measures applied between 2008 and April 2014 in several countries around the world. Results show that, in this period, Brazil was second only to Indonesia in the number of LCRs used, with 17 LCRs in force.11 Trade in services, a key enabler of GVC integration and productivity growth, is hampered by policy and regulatory barriers as well as by a distortive tax structure. The internationalization of production and subsequently trading in the GVC also requires attention to the role of services. Next to the fact that many services are used as inputs into the production of GVC goods, they are also essential in the smooth operation of the GVC to connect different production sides across borders. A clear example is logistics services. The latest OECD-WTO TiVA data for 2011 show that the value added of the service sector (both from domestic and foreign sources) contained in Brazil’s gross exports (49 percent) is above the average of Latin American and the Caribbean (LAC) peers (42.3 percent) but below the OECD average (54 percent). In addition, only one-tenth of the total services content of exports originates from foreign providers. This limited contribution is somehow influenced by substantial policy barriers in the service sector. On average, Brazil has more restrictions to trade in services than the average in the LAC region, according to the World                                                              11 Overall, the use of LCR tends to undermine export competitiveness over the long run and lead to suboptimal resource allocation that further affect productivity, since the price hike associated with the change from cheaper foreign suppliers to more expensive domestic suppliers leads to substitution away from these now more expensive, though more efficient, goods and services in the rest of the economy. Estimations presented also by Stone, Messent and Flaig (2015) show that the use of LCRs has caused a decline in global imports and total exports in every region. The estimated permanent reduction in total exports from Brazil due to these measures amounts to 0.65 percent.   6      Bank Services Trade Restrictiveness Index (STRI),12 with the most restrictive scores in financial and professional services, which are critical inputs across all industries for productivity growth and competitiveness (Figure 7).13 In addition to restrictive policies and regulations, there is also a distortive tax system for trading services in Brazil: the tax burden on service imports in Brazil compares unfavorably to other countries, which contributes to hamper the absorption of imported technology (Figure 8).14 Figure 8. Tax burden on service imports: Figure 9. Logistic Performance Index (2016) Brazil vs selected peer countries (%) [very low” (1) to very high” (5)] 60.0 5.0 50.0 4.0 40.0 3.0 30.0 20.0 2.0 10.0 1.0 0.0 0.0 Saudi Arabia France EUA Brazil Japan Mexico Spain China UK South Korea Italia India Argentina South Africa Netherlands Germany Source: CNI (2017) Source: World Bank staff elaboration using World Bank LPI dataset Brazil also faces challenges in trade logistics and trade facilitation. To compete in the global economy, traders require seamless supply chains, including efficient border management and clearance processes, as well as the development of efficient logistics services and improvements in both the hard and soft international logistics infrastructure. The score of overall logistics competence (and quality) in Brazil – as captured by the Logistics Performance Index (LPI) in 2016 – is below Mexico, Turkey, India, China and South Africa (Figure 9). Brazil’s overall integration in GVCs is low compared to international peers in part because of relatively lengthy and costly procedures to import and export.15 Indeed, firms’ integration to GVCs critically depends on their                                                              12   The World Bank STRI data set focuses on policies and regulations that discriminate against foreign services or foreign service providers, as well as certain key aspects of the overall regulatory environment that have a significant impact on trade in services. 13 These results are also valid when using OECD STRI; the latest numbers from 2017 show that Brazil scores worse than Mexico, Chile and Colombia for accounting, architecture, engineering and legal services and for commercial banking and insurance. For telecoms and retail, the value of the Brazil STRI index is zero. 14 The overall tax burden on services is heavier than in other sectors: while the average tax burden on the production and consumption of goods and services is 19.4 percent, it is much higher in the services segments most critical to other sectors of the economy: the tax toll exceeds 23 percent in transport and business services, 27 percent in IT services and over 30 percent in utilities (OECD, 2016). This tends to be particularly burdensome for firms that operate in supply chain organizations as they are not allowed to claim full credit for indirect taxes paid on services inputs in lieu of the "physical credit" principle in ICMS and they cannot claim credit for inputs in ISS.  15  This forces firms to adopt costly hedging strategies and complicates their ability to engage in just-in-time production or react quickly to demand shifts. Evidence suggests that inventory-holding costs can vary from 15 percent of the cost of goods per year to as much as 50 percent (Clark et al., 2016). Similarly, each day in transit is equivalent to an ad- valorem tariff ranging between 0.6 percent and 2.3 percent; and trade in components is  extremely time-sensitive   7      capacity to provide good quality products delivered on time to buyers further up in the value chain. Yet, unpredictability in customs clearance times due to physical inspection, or delays at border posts generated by excessive cargo handling in response to controls by multiple border agencies, increase uncertainty in delivery times. Despite the recent introduction of the “Portal Único de Comercio Exterior” - an electronic data interchange system that has reduced the time for documentary compliance for both exporting and importing16 - and the recent cooperation agreements between Customs and other border control agencies (as well as with third countries’ agencies17), Brazil still fares poorly in terms of monetary costs of border and documentary compliance when compared to a range of peers (Figure 10). Figure 10. Cost to export and import: border compliance in USD (2018) 1400 1200 1000 800 600 400 200 0 Argentina Indonesia Turkey India South Africa China Brazil Cost to export: Border compliance (USD) Cost to import: Border compliance (USD) Source: data from Doing Business database (2018) The use of those protectionist trade policies, particularly in tariff and NTMS, was not conducive to improve export competitiveness. On the contrary, import protection ends up acting as a direct tax on exports, making them less, not more competitive. The so-called Lerner Symmetry theorem states that an import tariff can have the same effects as an export tax—so reducing import protection is expected also to boost exports. The idea is that an increase in import tariffs appreciates the home real exchange rate as the domestic policy rate and the international interest rate differential increases. The terms of trade appreciation, in turn, induces a positive effect on consumption but a drag on real net exports (Linde and Pescatori 2017). The connection of imports                                                              (Hummels and Schauer, 2013). Customs delays also reduce exports value and export market diversification (Volpe Martincus, Carballo and Graziano, 2015). 16 Organized as a joint effort between more than 20 agencies and the private sector, Portal Unico promotes the simplification, streamlining and cost reduction of trade-related procedures and formalities with the support of risk management, automation and information technology tools. The initiative aims to eliminate redundant formalities and document requirements, to optimize the performance of the agencies which intervene in trade, and to reduce by 40 percent the average time to export and to import. According to the 2018 Doing Business report, the average time to comply with documentary export obligations fell from 18 to 12 hours between 2016 and 2017, a reduction of a third. The average time on the import side decreased from 120 to 48 hours, a reduction of 60 percent. Brazil improved by 10 positions in the “Trading across borders” indicator. 17 Recent examples include the cooperation between Mercosur and the Pacific Alliance countries to enable the exchange of electronic trade documents. Certificates of origin, originally in paper, are already being replaced by digital documents with Argentina, Chile and Uruguay. Brazil is also working to exchange electronic phytosanitary certificates with the United States.  8      to exports has become even clearer in the context of twenty-first century trade, as GVCs can be described as “factories that cross international borders” (Taglioni and Winkler 2016). It is evident that imports are essential for exports and that reducing the costs of imports is critical for a country to be a more dynamic exporter. In this regard, the case of Embraer is emblematic. In fact, after significant growth in trade value over much of the past decade, Brazil has been losing market share in world export markets since 2012. Brazil’s exports grew robustly in the periods before and after the global trade collapse of 2008, aided by strong commodity prices on international markets. Export growth, which averaged 22.5 percent per quarter between 2010-Q1 and 2012-Q1, was above the world average; as a result, Brazil gained world market shares, indicated by the green area in Figure 11. Since 2012-Q2, however, Brazil’s export growth has been negative, retrenching on average 4.1 percent each quarter. Although export growth has been subdued worldwide, Brazil’s lower performance has resulted in world market share losses, indicated by the red area in Figure 11. Overall, Brazilian exports were mostly benefiting from geographical and sector composition effects, primarily associated with the fast growth of China (Canuto, Cavalari and Reis 2012). Excluding these composition effects, the "pure" export performance was still positive, but of much lower intensity, and smaller than that of most of the major emerging economies, a result that could be associated to the increase of protectionism in the Brazilian economy. Figure 11. Brazil’s export growth versus world Figure 12. Industrial export quality index: export growth, 2006Q1-2016Q1 Brazil vs BRICS peers 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Brazil China India Russian Federation South Africa Source: World Bank staff elaboration using World Bank’s Source: World Bank staff elaboration using UNIDO Measuring Export Competitiveness Database. dataset In addition, Brazil’s industrial export quality has been deteriorating in the past years. Economic development entails reallocating resources towards more productive firms and improving the quality of goods produced within existing sectors. Producing higher quality (and generally higher priced) products is associated with an intense “learning by exporting” process and helps to bring productivity gains (De Loecker, 2013).18 Brazil’s export quality has been deteriorating in the past years and is currently below that of most of its BRICS peers, except the Russian Federation (Figure 12). Recent analysis conducted by Henn, Papageorgiou and Spatafora (2015) corroborates this result and shows that Brazil’s overall export quality is well below the average implied by the frontier of other major emerging economies.                                                              18 Besides quality upgrading, other mechanisms drive the “learning by exporting” process and the potential productivity gains, such as investing in marketing, innovating, or dealing with foreign buyers (De Loecker, 2013). 9      3. Potential gains accruing from trade liberalization and opportunities to boost productivity through trade integration Policy simulations, obtained from a customized CGE model, point to substantial gains from trade liberalization in Brazil. A customized CGE model for Brazil is applied to assess the impacts of trade liberalization reforms on exports, imports, GDP and output.19 Three hypothetical liberalization scenarios are modeled (Figure 13). In the first scenario, called “coordinated trade reforms within Mercosur”, each Mercosur member unilaterally reduces tariffs by 50 percent with respect to non-Mercosur countries, NTMs are streamlined among the Mercosur parties20 and export taxes are eliminated among the parties. In the second scenario, a reciprocal preferential trade agreement between Mercosur and the EU is modeled where the average tariff applied by Brazil to EU products would fall from 10.7 to 3.2 percent in a 10-year time-horizon, while the average tariff in the EU for Brazilian products would fall from about 2.5 percent to close to 1 percent, NTM tariff equivalents are reduced by 15 percent and export taxes are eliminated among Mercosur and EU countries. The third scenario comprehends a preferential trade agreement between Mercosur and the Pacific Alliance, where Mercosur countries and Pacific Alliance members gradually reduce tariffs over 10 years,21,22 NTM tariff equivalents are reduced by 15 percent and export taxes are eliminated among the parties.23                                                                19 The model used is the LINKAGE model, which is a dynamic, multi-sector, multi-region model with economy-wide coverage for each region. For each economy, the model also tracks the inter-linkages between sectors through input- output transactions, as well as various sources of final demand including private and government consumption, imports, exports, and investment. The database used is the Global Trade Analysis Project (GTAP) which was modified to update the data and identify subsectors of interest for Brazil. Starting from the GTAP database 9.2, the base year of 2011 was updated to 2015 and the input-output structures for Brazil were updated to reflect the latest official tables from IBGE. The sectoral dimension in GTAP was expanded to include several new sectors of interest in Mercosur countries. These include sectors such as beef, soybean, soybean products, wine, footwear, furniture, home appliances, and auto parts that are part of more aggregate GTAP sectors. 20  Resulting in a reduction of 15 percent in the tariff equivalents for goods and services. 21  It is worth highlighting there will be already free trade between Mercosur and the Pacific Alliance by 2019.  22 In this case, the trade-weighted average tariff applied by Brazil to products from the countries in the Pacific Alliance would fall from 1.3 to 0.3 percent. The average tariff faced by Brazil in the Pacific Alliance would fall from 2.9 percent to 0.8 percent. 23 For the Mercosur-Pacific Alliance scenario, NTM tariff equivalents are reduced by 15 percent and export taxes are eliminated among the parties. Moreover, a bilateral market access is assumed so that existing liberalization among partners remain. Existing tariff barriers with Pacific Alliance countries have been lowered by previous bilateral or Mercosur agreements under the ALADI framework, especially with respect to Chile and Peru. Most tariff liberalization in this scenario would be with respect to Mexico and to a lesser extent Colombia. Liberalization with the Pacific Alliance is assumed to take place both more quickly and more comprehensively. No products are excluded from liberalization and all tariffs are either removed or partially reduced for the most sensitive products. 10      Figure 13. CGE trade liberalization scenarios “Community reforms” at Mercosur-EU Mercosur- Pacific Alliance Mercosur Tariffs Tariffs in all Mercosur countries Bilateral tariffs: in Mercosur from 10.5 Bilateral tariffs: in Mercosur from 1.3 to reduced by 50 percent to 3.1percent; in the EU from 2.5 to 0.3 percent; in the PA from 2.6 to 0.7percent 0.6percent NTMs Reduced by 15 percent intra- Reduced by 15 percent among parties; Reduced by 15 percent among parties; Mercosur; export controls export controls eliminated export controls eliminated eliminated The largest economywide gains would come from coordinated trade reforms within Mercosur. The second largest economywide gains would come a reciprocal preferential trade agreement between Mercosur and the EU. Coordinated trade reforms within Mercosur would boost exports and imports by 7.5 and 6.6 percent, respectively, while real GDP would experience a 0.93 percent increase (above baseline projections in 2030). The effects on output would be different across sectors. Figure 15 shows the output percentage deviations from the baseline in 2030. The largest negative effects are seen for manufacturing (-0.8 percent), led by retrenchments of the textiles and apparel sector (-$2.37 billion) and vehicles (-$2.17 billion). All other broad sectors would expand relative to the baseline as resources shift to other activities. On a dollar basis, the services sector expands the most (by $16.6 billion), followed by natural resources and energy products ($10.3 billion), and agriculture ($8.1 billion). The second largest economywide gains would come from a reciprocal preferential trade agreement between Mercosur and the EU. Exports and imports would increase by 5.5 and 4.9 percent, respectively, while GDP would experience a 0.58 percent increase (above baseline projections in 2030). All big sectoral aggregates would experience positive output variations (above baseline projections in 2030): manufacturing would experience the highest expansion (2.2 percent), led mainly by expansion of the footwear sector. Figure 14. Economy wide effects - deviations Figure 15. Sectoral output deviations from the from the baseline, 2030 (%) baseline, 2030 (%) 4.0 3.4 3.5 Mercosur-Pacific Alliance 3.0 2.2 0.41 2.5 2.0 1.5 1.0 Mercosur-EU 0.5 0.58 0.0 -0.5 -1.0 ‐0.8 Coordinated trade reforms Coordinated trade Mercosur-EU Mercosur-Pacific within Mercosur 0.93 reforms within Alliance Mercosur 0.00 2.00 4.00 6.00 8.00 Agri. And food Manufacturing Imports Exports Real GDP NR and energy Services Source: World Bank staff elaboration using GTAP-Linkage Source: World Bank staff elaboration using GTAP-Linkage model model 11      While the CGE results presented here are conservative estimates, their magnitude is aligned with other recent reform simulations estimated for Brazil. While the model is dynamic in the sense that capital stock can change over time, it does not include other dynamic factors such as productivity increases from endogenous growth effects via technological spillovers, “learning by doing,” and inflows of foreign technology and efficiency-seeking FDI induced by liberalization. Nevertheless, the magnitude of the results is aligned with recent empirical evidence on the potential benefits from furthering trade liberalization on Brazil. For instance, results from IMF (2017) suggest that halving tariffs on capital goods in Brazil would increase the investment-to- GDP ratio by about two percentage points. Araujo and Flaig (2016) use a CGE model and show that if Brazil reduces its import tariffs to OECD minimum level, eliminates all local content requirements and reduces export tariffs to zero, aggregate output would grow by 1.7 percent. At the firm level, higher participation in GVCs is expected to bring productivity dividends to firms. The literature on productivity effects from GVC participation points to three main transmission channels. First, through specialization in tasks. Growth of GVCs has led to increasing specialization in specific activities within value chains. Firms then can reap productivity gains by specializing in those core tasks for which they are most efficient and offshoring less efficient parts of the production process abroad (Grossman and Rossi-Hansberg, 2008). Second, through access to a larger variety and quality of intermediate inputs; as previously highlighted, there is vast empirical evidence in the literature showing the effects of importing inputs in productivity growth (see Halpern, Koren and Szeidl (2015), for instance). Third, through knowledge spillovers from multinational enterprises. To the extent that these firms tend to demand more and/or better-quality inputs from local suppliers and may also share knowledge-technology and encourage the adoption of new practices, sales of GVC-linked local intermediates to international buyers is expected to spur productivity by increasing demand for more and/or better inputs and by providing assistance to local suppliers (Taglioni and Winkler, 2016).24 The existence of productivity spillovers from structural integration in GVCs for Brazilian manufacturing firms is confirmed by recent empirical analysis. Using a cross-section of around 12,700 domestic manufacturing firms in 22 low- and middle-income countries from the World Bank’s Enterprise Surveys, Winkler (2017) explores the association between GVC integration and labor productivity for Brazil and the full sample (see Annex 1 for a brief methodological presentation). Results suggest that Brazilian manufacturing firms operating in industries with a high structural integration into GVCs, namely which equals or exceeds the 75th percentile across all 22 countries in the sample for that industry, show significantly higher labor productivity levels than those firms that operate in industries with lower GVC integration. Their difference in productivity is stronger for GVC integration as a seller (+11.2 percent on average) compared to GVC integration as a buyer (+8.5 percent on average) in the Brazil sample, holding all other variables constant. By comparison, the findings suggest that GVC integration as a seller at the industry level is correlated with a smaller labor productivity surplus (+3.8 percent on average) in the sample of 22 countries and is non-existent for high GVC integration as a buyer.                                                              24 It is worth stressing however that the knowledge spillovers from international buyers and FDI in general tend to accrue asymmetrically, benefitting firms with sufficient absorptive capacity. In this regard, the capacity of a country or a firm to absorb, adapt, and reap the full benefits of knowledge produced at the frontier depends on strategic investments in R&D, organizational know-how, and other forms of knowledge-based capital. (OECD, 2015) provides a theoretical and empirical discussion on the role of technology diffusion as a channel to boost productivity growth. 12      Integrating to GVC represents a challenge to countries that are far from trade hubs. However, there are also some opportunities. OECD (2015) suggests that location is one of the main non- policy determinants of GVC participation: evidence suggests that because GVC activity is organized around large manufacturing hubs, the larger the distance to the main manufacturing hubs in Europe, North America and Asia, the lower is the backward (buyer) engagement. On the other hand, while there is a premium to locating close to “headquarter” economies, there are also opportunities for GVC integration for countries located far from large manufacturing hubs; a recent empirical analysis conducted by Cheng et al. (2015) on the determinants of GVC participation suggests that physical distance (weighted by economy size) is negatively associated with trade in GVCs only for low-tech manufacturing sectors. Moreover, production networks serving European or U.S. markets do not need to be the only geographical focus of countries that want to integrate into GVCs. With growing markets in the Americas, a regional network of production might be increasingly sustainable. Increasing the number and depth of preferential trade agreements (PTAs) offers an important channel to increase participation in GVCs. Given the increasing unbundling of export goods production, PTAs have become the main vehicle to bring in new disciplines (such as competition policy, intellectual property, investment, etc.) that allow factories to connect across borders in a seamless way. The frequency and depth of PTAs have been increasing substantially: data presented in Hollweg and Rocha (2017) show that the number of agreements notified to WTO have increased from about 70 in 1990 to close to 300 presently in force. The same data set25 shows that more recent PTAs cover more policy areas than earlier PTAs: agreements signed before 1991 included on average 9 provisions whereas agreements signed between 2005 and 2015 included on average 15 provisions; more than 50 percent of agreements include deeper provisions such as anti-dumping and countervailing measures, rules on competition, movement of capital and intellectual property rights. Evidence presented by Hollweg and Rocha (2017) also suggests that GVC-related trade – proxied by trade in parts and components - is higher on average for countries that have signed deeper agreements (Figure 16), where depth of an agreement is the number of legally enforceable provisions.                                                                  25 The World Bank data set on PTAs content; see Annex 3 for further details. 13      Figure 16. Deep PTAs and GVC-related trade: Figure 17. Number of active agreements by simple correlations country for selected economies (2015): Latin America Source: Hollweg and Rocha (2017) Source: Hollweg and Rocha (2017) Note: Low depth – agreements with less than or equal to 15 provisions; Medium depth – agreements with 15 or more provisions but less than or equal to 30; High depth – agreements with more than 30 provisions However, Brazil sits at the margins of regional integration trends in other parts of the world, and Mercosur, the only PTA Brazil is signatory of, has a low level of depth with only six enforceable provisions. Although countries around the world have increased their participation in PTAs, especially in the last two decades, Brazil has not followed this trend. In comparison to the Latin America region, countries such as Chile, Peru, Mexico and Colombia have signed on average 13 agreements (Figure 17). Compared with other BRICS, Brazil has the lowest number of agreements, followed by South Africa with a total of 4 active agreements. Countries such as India, China and Russia have a total of 11, 8 and 20 active PTAs respectively in 2015 (Figure 18). Brazil has only one PTA, Mercosur.26 With respect to content, only 6 out of the 17 disciplines that are covered and legally enforceable27 in the Mercosur agreement are currently in force.28 If measuring the total depth of an agreement as the total number of legally enforceable provisions it includes, Brazil –and Mercosur- represent the PTA with the lowest level of depth signed by Latin American countries with the exception of agreements signed by Bolivia, Ecuador and the República                                                              26 In addition to Mercosur, Brazil has in place a PTA with Israel since 2010; this agreement however has not been notified to the WTO. Brazil is also member of 5 partial scope agreements (PSA). A PSA which is not defined or referred to in the WTO Agreement, means that the agreement covers only certain products. Partial scope agreements are notified under paragraph 4(a) of the Enabling Clause (source: https://rtais.wto.org/UserGuide/RTAIS_USER_GUIDE_EN.html). The PSAs Brazil is signatory of are the following: Global System of Trade Preferences among Developing Countries (GSTP), Latin American Integration Association (LAIA), Protocol on Trade Negotiations (PTN) and MERCOSUR-India and Mercosur-SACU. 27 See Annex 3 for a definition of enforceable provision. 28 This figure has been confirmed by experts and governments in Mercosur. Although all the disciplines that currently fall into the mandate of the WTO (WTO+ provisions) are included in the Mercosur agreement, less than half are in force. The provisions in force include: FTA Industrial and Agriculture, Anti-dumping (AD), Technical barriers to trade (TBT), GATS, and Sanitary and phytosanitary measures (SPS). Other provisions such as Customs, State Enterprises (STE) and Public Procurement, are legally enforceable but currently not in force. Mercosur also includes disciplines that go beyond the WTO mandate (WTO-X provisions) such as Competition Policy, Intellectual Property Rights (IPR) and Movement of Capital. However, none of these are currently in force.   14      Bolivariana de Venezuela (Figure 19).29 In relation to the BRICS, the depth of Mercosur is comparable to the one of agreements signed by China and India. Figure 18. Number of active agreements by Figure 19. PTA depth, Latin American country for selected economies (2015): BRICS countries 35 1 1 2 1 1 1 2 22 8 10 13 30 25 20 15 10 5 0 max depth avg. Depth # of agreements Source: Hollweg and Rocha (2017) Source: Hollweg and Rocha (2017) 4. Domestic market integration and policy induced stringencies affecting resource allocation For Brazil to take full advantage of the opportunities that external integration offers, domestic markets also need to function better in a way to allow efficient allocation of resources across firms, sectors and regions. However, key (policy induced) factors have been limiting domestic integration and distorting markets. The potential productivity returns accruing from further integration with the global economy can be eventually muted if the internal market is not properly integrated. Reaping the full benefits of trade liberalization will depend on how production factors move across firms, sectors and regions. In this regard, resource misallocation is ultimately influenced by policy interventions. High costs of doing business First, high costs of doing business in Brazil – traditionally referred to as Custo Brasil - tend to impose higher burden on entry of new firms that could be potentially more efficient than incumbents. Using a worldwide database, Klapper, Laeven and Rajan (2006) found that entry regulations have significant adverse effects on entrepreneurship and tend to mute the disciplining effect of competition by indiscriminately screening out small young firms that could be more productive than incumbents.30 More recently, Fuentes and Mies (2014) show that reforms tackling entry barriers become increasingly important to close the productivity gap as countries develop. Custo Brasil is high, particularly those associated with entry procedures. According to 2018 Doing Business data, entrepreneurs face substantial difficulties with key entry costs of doing business, such as business registration, dealing with construction permits, and registering property. In all                                                              29 Depth of an agreement is the number of legally enforceable provisions. 30 Evidence presented in the analysis shows that the growth in labor productivity for firms older than two years is relatively lower in naturally high-entry industries when the industry is in a country with higher bureaucratic barriers to entry. 15      these three cases, Brazil ranks in the bottom third of the Doing Business 2018 ranking (176th, 170th and 131th, respectively). With regards to operational costs, the performance is even worse for paying taxes, where Brazil occupies the 184th position out of 190 countries (Figure 20). Figure 20. Doing Business in Brazil: rank and distance to frontier Change in DB 2017 Distance Distance to to Frontier (% Frontier 2018- Topics DB 2018 Rank points) 2017 (% points) Overall 125 56.45 0.38 Starting a Business 176 65.05 0.01 Dealing with Construction Permits 170 49.83 0.04 Getting Electricity 45 82.46 1.23 Registering Property 131 52.6 0.02 Getting Credit 105 45 .. Protecting Minority Investors 43 63.33 .. Paying Taxes 184 32.97 .. Trading across Borders 139 59.78 4.21 Enforcing Contracts 47 66 .. Resolving Insolvency 80 47.46 1.69 Source: World Bank Doing Business 2018 dataset. Note: An economy’s distance to frontier score is indicated on a scale from 0 to 100, where 0 represents the worst performance and 100 the frontier Box 1. Tax structure and resource allocation in Brazil Aggregate productivity performance at country level is particularly impaired by a complex tax structure that hampers resource allocation across sectors and states. As highlighted by Apy (2017), the Brazilian tax structure with regards to goods and services is extremely complex, with four main taxes (ICMS, IPI, ISS and PIS-Cofins) that have distinct incidence coverage (cumulative, non-cumulative and mixed). More importantly, individual products are subject to different regimes depending on the industry, the way the production process is structured (vertical integration versus fragmentation) and the locality where the production process takes place. Thus, relative prices are distorted with impacts on resource allocation and productivity. As a result, firms are often induced to integrate their activities in a vertical way, even when the production of goods and services would be cheaper if outsourced. Second, the fact that ICMS is cumulative – as it follows the origin principle and has restrictions on input tax credits – impairs exports of goods and services. Third, tax wars and regional tax breaks lead to regional capital misallocation. More specifically, Brazilian states (and also municipalities) have been engaging over the years in tax competition, known as “fiscal war.” Benefiting from full administrative autonomy in setting their ICMS rates, states have used the ICMS as an industrial policy instrument by granting tax exemptions to attract investment, not only in manufacturing but also services (distribution centers). Therefore, firms are often encouraged to locate their activities in regions/states where taxes are lower, even if the production is less efficient. 16      Inadequate state of physical infrastructure Second, the inadequate state of physical infrastructure, another key component of Custo Brasil, constrains connectivity while hindering the increase of economies of scale, which is a crucial feature needed to reap the benefits of trade integration and to boost productivity growth. In principle, increasing the efficiency in overall public infrastructure is expected to raise productivity directly through higher infrastructure capital (measured as capital intensity) which makes private capital more productive. Recent analysis conducted by IMF (2015) - for a panel of 17 advanced economies during the 1985-2013 period - provided evidence that an increase in public investment equivalent to 1 percent of GDP is associated to an increase of labor productivity of 0.5 percent over the medium term, although primary through higher physical capital intensity. The impacts are higher the more efficient is the public spending system. In Brazil, the average rates of investment in overall infrastructure have been decreasing in the past decades, falling from an average of 5.42 percent over GDP in the 1970s and 1980s to an average of 2.15 percent in 2011-16 (Frischtak and Mourao, 2017). This sinking investment trend, explained by a reduction in public investment that was not counterbalanced by private sector investment,31 brought negative impacts in terms of infrastructure adequacy: Brazil scores lower than its main export competitors on qualitative indicators of infrastructure adequacy, both in terms of overall infrastructure and transport infrastructure (Figures 21 and 22). Moreover, the strong preference for road transport over other modes combined with low percentage of paved roads32 increases logistics costs. There is also evidence that inadequacy of transport infrastructure is associated with market segmentation in Brazil: evidence presented by Garcia-Escribano, Goes and Karpowicz (2015) shows a positive correlation between slower price convergence across major metropolitan areas and longer commuting times between cities in Brazil. All these factors together contribute to hamper domestic market integration, which then prevents increasing economies of scale, a crucial feature to reap the benefits of trade openness. Figure 21. Quality of overall infrastructure, rank Figure 22. Quality of road infrastructure, (2016-17) rank (2016-17) 1 = best; 144 = worst 1 = best; 144 = worst 140 122 123 140 121 123 120 120 108 95 100 87 100 80 64 65 80 59 54 60 60 41 35 34 40 23 40 26 13 14 20 20 0 0                                                              31 In addition, public infrastructure spending efficiency in Brazil, and Latin America in general, has been hampered by several factors, as listed by Fay et al. (2017), such as: weak planning, project appraisal and preparation capacity; overly rigid or myopic budgeting; difficulties with budget execution; inefficient procurement system; unclear project sustainability (caused by an imbalance between capital and current spending); uncompetitive construction industry. 32  Almost 60 percent of cargo in Brazil flows through highways of which only 14 percent is paved.  17      Source: World Economic Forum 2016-17 Report Source: World Economic Forum 2016-17 Report Excessive (and ineffective) use of business support programs Third, several business support programs seem to be inefficient in reaching their objectives and appear to undermine creative destruction. At the micro level, evaluations of specific business support programs suggest a broad lack of impact on productivity. For instance, Lazzarini et al. (2014) analyze BNDES direct activity through loans and equity funding and find evidence that BNDES mostly financed large and profitable firms, lowering their financial expenses, but with no effect on their investments and performance in the period 2002-2009. Also, size-based policies – such as Simples, the largest tax exemption program in place in Brazil – have shown no impact on key economic performance indicators. For instance, Piza (2016) finds that Simples was not effective in increasing formalization rates of small firms. More recently, Corseuil and Moura (2017) show that the same program did not have any impact on firm performance within manufacturing industry: the impacts on wages, employment and value added were all statistically null. In addition, Lei de Informática and Lei do Bem33 disproportionately benefited a small number of large firms and are not able to cover, by design, young firms that are likely to be more productive. Moreover, a recent assessment of various Brazilian programs of firm support — including productive finance, business consulting, value chain, export promotion, and innovation support – presented by Pires and Russel (2017) suggests there were few positive results on productivity or other indicators; in most cases either no impact was found or regression results were inconclusive. Overall, the evidence suggests that through these non-competitively allocated incentives, Brazil has created an unlevel playing field that has favored the profitability of less efficient firms, both small as well as larger and older firms – thereby preventing more efficient firms from expanding, and likely deterring potentially more productive firms from entering these markets (Dutz et al., 2017). Box 2. Inovar Auto: Key features and impact Inovar-Auto was created in April 2012. Its official objective was to promote R&D, improve the quality of domestically produced cars (energy efficiency was a target within this framework) and to promote investment and domestic production. The incentives provided under the program followed a two-pronged approach. First, it increased a tax levied on industrialized products (IPI) by 30 percent for all light-duty vehicles (LDVs) and light commercial vehicles. Second, it defined a set of requirements for automakers to qualify for up to 30 percent discount in the IPU as follows:  Meet a corporate average vehicle efficiency target (precisely to improve the average efficiency for new LDVs by about 12 percent from 2012 levels by 2017)  Conduct a minimum number of manufacturing and engineering activities in Brazil for at least 80 percent of produced LDV and light commercial vehicles,34 and                                                              33 Lei do Bem is an R&D tax subsidy program instituted in 2007 with the objective of expanding incentives for investments in R&D; it authorizes companies that invest in R&D and meet certain requirements to claim tax incentives automatically for certain types of spending. 34 For cars and light commercial vehicles these manufacturing stages are: Stamping; welding; anticorrosive treatment and painting; plastic injection; motor manufacturing; gearbox and suspension systems assembly; steering and suspension systems assembly; electrical systems assembly; axle and brake systems assembly; monoblock manufacturing or chassis assembly; final assembly, review and testing; and own laboratory infrastructure for product   18       Choose at least 2 out of 3 pre-requisites to qualify for the program – (1) investment in R&D, (2) investment in engineering, industrial technology, and supplier capacitation, and (3) participation in the Vehicle Labeling Scheme. In other words, IPI taxes would remain unchanged for those manufacturers that meet the requirements, which in principle incentivizes investments in vehicle efficiency, national production, R&D, and automotive technology. The program was limited to vehicles manufactured between 2013 and 2017, after which IPI rates return to pre-2013 levels. While the official objective of the program was to increase the sector’s competitiveness, its real motivation seems to have been to protect domestic producers from losing market share to imports. By not including instruments to promote exports (even a t the regional level), and not boosting the sustained development of suppliers’ capabilities, the program neglected the benefits of a more outward expansion approach while inflicting a longer period for technological catching up, as there was limited space to import inputs of better quality and higher variety. In fact, according to Sturgeon et al (2017), although Inovar-Auto may have shifted demand from imports to domestic production in the short-term, it did not alter the competitiveness of the industry enough to allow Brazilian production to grow through exports or through cost and price reductions in the domestic market. Despite the absence of a counterfactual, the authors present evidence suggesting lack of impact on most economic outcomes, such as: employment, wages, production and productivity. In addition, the authors show that while Inovar-Auto improved the trade balance through a reduction of imports, it did not increase the industry participation in global value chains (via increased bi-lateral trade in intermediates and knowledge-intensive services) and has not increased scale- efficiency, since automakers overinvested in different plants. In August 2017, a dispute panel at the World Trade Organization (WTO) recommended that Brazil withdraw Inovar Auto, supporting complaints of unfair competition by the European Union and Japan. The program was then closed. In July 2018, a new incentive program, Rota 2030, was launched; this program includes among other incentives a federal government grant of up to 1.5 billion reais (US$382mn) in tax credit for every year automotive companies jointly invest at least 5 billion reais on research and development.   At the regional level, development policies aiming at redistributing economic activity in less flourishing areas have been showing mixed impacts on local productivity growth and might be dampening economic agglomeration effects. Duhaut and Lall (2018) apply a propensity score method combined with a difference-in-differences estimator to assess how credit provided under the regional Constitutional Financing Funds as well as loans provided by BNDES under the Regional Dynamization Policy (the PDR) impact local productivity growth between 2008 and 2014 (see Annex 4 for a methodological presentation). When pooling all minimally comparable areas across all population densities, results suggest that receiving credits from both the Constitutional Funds and the BNDES is linked to a 12 percent higher productivity index in 2014 compared to areas that received Constitutional Funds only. This suggests that regional development programs in Brazil may generate higher productivity dividends if better coordinated.                                                              development and testing. For trucks these manufacturing stages are: Stamping; welding; anticorrosive treatment and painting; plastic injection; motor manufacturing; gearbox and suspension systems assembly; steering and suspension systems assembly; electrical systems assembly; axle and brake systems assembly; monoblock manufacturing or chassis assembly; final assembly, review and testing; final assembly of cabins or bodies, with installation of items, including acoustic and thermal, lining and finishing; production of bodies predominantly through single pieces stamped regionally; and own laboratory infrastructure for product development and testing. For chassis with an engine the manufacturing stages are: welding; anticorrosive treatment and painting; plastic injection; motor manufacturing; gearbox and suspension systems assembly; steering and suspension systems assembly; electrical systems assembly; axle and brake systems assembly; monoblock manufacturing or chassis assembly; final assembly, review and testing; production of bodies; and own laboratory infrastructure for product development and testing.   19      Despite limited benefits, the fiscal cost associated with this large set of business support programs has been increasing and not been accompanied by an effective oversight and evaluation system. Federal spending on business support measures in Brazil - encompassing tax expenditures, subsidized credit and general expenditures35 - more than doubled in real terms in the past decade, jumping from R$ 125 billion in 2006 to R$ 267 billion in 2015; all figures in 2015 values (Figure 23). The responsibility for spending and oversight of industrial policies at the federal level in Brazil is spread across a range of ministries and public entities. There is no one entity overseeing overall implementation: such an entity could in principle take advantage of synergies and avoid cross- program overlap, as well as evaluate the effectiveness and efficiency of individual programs.36 Figure 23. Total federal fiscal expenditures on sectoral business support policies   Source: Federal Revenue Service of Brazil; BNDES; FAT; Ministry of S&T (MCTI); Ministry of Industry (MDIC) Note: Values in R$ MM 2015 base year; estimated credit before 2008 and general expenditures after 2013 Tax exemptions are by far the most important component of federal spending on sectoral business support policies. One key characteristic of these exemptions is their non-neutral implementation which tends to focus in specific sectors and types of firms. Tax exemptions (expenditures) account for almost 61 percent of total sectoral business support policy spending and 2.9 percent of GDP in 2015. They lower otherwise-required tax payments and thereby provide an incentive to firms to invest in equipment, local content sourcing, upgrading or innovation. The expenditures related to exemptions have doubled in real terms over the past decade. The major components are payroll tax exemptions, introduced in 2011, the simplified tax regime for smaller companies, and the Manaus Free Zone. Compared with other countries, Brazil is an outlier with 2.9 percent of its GDP spent on tax exemptions. This compares to OECD economies and structural peers that spend approximatively 1 percent of their GDP on tax exemptions as well as other Latin American countries such as Argentina (1.0 percent), Chile (1.5 percent) and Peru (1.5 percent).                                                              35 Tax expenditures cover exemptions that result in foregone tax revenues; subsidized credit states for credit provided at below-market rates, and general expenditures include grants and subsidies to outputs and non-credit inputs, and payments for program administration, equipment and buildings, etc. 36 See Dutz et al (2017) for a detailed discussion about the main incentives granted at the federal level and how they are spread across a range of public institutions. 20      Except for the Simples size-based tax exemptions and the R&D incentives, the remaining incentives have a sectoral focus and thereby generate distortions across sectors and activities. In addition, Simples also provides incentives to smaller firms, which can create threshold effects in preventing formal firm growth. The bulk of the implicit subsidies created by incentives go to firms in trade and services, and manufacturing industries. These differences across sectors erode a level playing field by favoring some firms and sectors over others. Additional government interventions: Product market regulations and competition law enforcement Reducing cost of doing business and improving the adequacy of infrastructure are not enough to foster competition in markets and ensure domestic market integration. Effective antitrust enforcement and pro-competition regulation are also important. Even when transport and transaction costs are lower, or when costs of paying taxes, getting electricity and dealing with construction permits are reduced, there may be still firms exercising significant market power even in geographically large and integrated markets, for example, if they agree to collude on prices, or if there is a lack of regulation that prevents the abuse of dominance by firms that supply services in segments with features of a natural monopoly. Effective antitrust enforcement and pro- competition regulation minimize these risks and are key pillars of a holistic competition policy.37 In this turn, while this paper discusses specific policies that can enhance healthy market dynamics and competition, it does not include an assessment of Brazil’s competition policy implementation itself.38 Available data on product market regulation reveal that government interventions in domestic markets have not been conducive to competition. Competition enhances productivity by improving allocative efficiency, enhancing productive efficiency, and boosting innovation. Aghion, Akcigit and Howitt (2014) also provide a compelling framework linking market competition to productivity growth through creative destruction, while Conway et al (2006) and Alesina et al. (2005) trace the positive impact of pro-competition product market regulation on productivity, and Barone and Cingano (2011) find empirical evidence that pro-competition reforms in input sectors benefit productivity of the overall economy. Furthermore, the introduction of anti-cartel policy is also related to higher labor productivity growth. Symeonidis (2008) records a 20 to 30 p.p. lower labor productivity growth in industries with cartels than in industries without cartels. Some government interventions unduly inhibit firm’s ability or incentives to compete. The OECD Product Market Regulation (PMR) indicator measures the degree to which policies or regulatory measures promote or inhibit competition in areas of the product market where competition is viable. PMR data for Brazil in 2013 suggests that policies and regulations in key sectors are less conducive to competition than in comparator countries. While other BRICS countries have promoted reforms over the last decade, the overall restrictiveness of product market regulation in Brazil remained unchanged between 2008 and 2013. The general degree of restrictiveness of product market regulations in Brazil has not changed since between 2008 and 2013 and compares less favorably to other economies which on average have moved towards a more pro-competitive regulatory framework (Figure 24). In fact, while others                                                              37 Competition policy encompasses a set of policies and laws that ensure competition in the marketplace is not restricted in such a way as to reduce economic welfare (Motta, 2004). 38 For more details on what competition policy encompasses, see Kitzmueller and Licetti (2012).  21      such as China, Russia, South Africa and India have reduced their PMR scores by 9 to 17 percent, Brazil’s overall PMR score suggests little reform progress between 2008 and 2013 (Figure 25). Figure 24. PMR indicators: Brazil vs Figure 25. Changes in the PMR indicator from comparator countries, 2013 2008 to 2013- (Index scale 0 to 6 from least to most restrictive) (Greater reduction in the PMR is associated with regulations less restrictive to competition) Source: Product Market Regulation OECD database Source: Product Market Regulation OECD database Note: OECD top 5 are Austria, Denmark, Netherlands, New Note: OECD top 5 are Austria, Denmark, Netherlands, New Zealand and the UK Zealand and the UK Barriers to entrepreneurship represent the largest source of restrictiveness to competition in Brazil and are likely to slow down the reallocation process by preventing entry of potentially more productive firms and/or reinforcing the dominance in specific markets. Three main policy areas are mapped by the PMR methodology: state control, barriers to entrepreneurship, and barriers to trade and investment, economy-wide and in key sectors, including transport, retail, professional services, energy and telecommunications.39 A decomposition of the PMR Indicator for “barriers to entrepreneurship” reveals that the complexity of regulatory procedures is particularly burdensome (Figure 24). For instance, the license and permits system has seen no improvement since 2008. In addition, there are sizeable barriers to competition in services sectors – especially those needed for construction. For example, the number of tasks for which the legal, accounting and architectural services have exclusive rights (7, 9 and 8, respectively) is higher than other OECD countries; for the three professional services, there is compulsory chamber membership and a requirement for accreditation. In Brazil, engineering services can only be provided by professionals registered with a professional association (CONFEA/CREA).40 In other countries, such membership to professional associations is usually just voluntary. These associations additionally have the power to regulate the engineering profession in Brazil, which presents high barriers to entry of foreign competitors. This is critical, since Brazil graduates on average 6 engineers per 1,000 working population compared to 25 engineers per 1,000 working population                                                              39 See Koske et al (2015) for a detailed presentation of the PMR methodology. 40 Conselho Federal de Engenharia e Agronomia - CONFEA (Federal Council of Engineering and Agronomy) is the professional association entrusted with the regulation of engineers and engineering services in Brazil. CONFEA operates nationwide through its state agencies called CREAs (Regional Council of Engineering and Agronomy).   22      in the United States and Japan.41 Permanent registration by foreign engineers with the CREA may take up to 3 years, while temporary registration is subject to the CREAs’ discretion. Even with a temporary registration, foreign engineers need to have their work supervised by a registered Brazilian professional and cannot hold a majority stake or occupy a management position in a Brazilian engineering company.42 CONFEA additionally decides on the list of tasks that can be executed only by registered engineers, which is opposed to good practice where such lists are set by independent bodies.43 CREAs publish tables with engineering service fees. Even though they are not mandatory, CONFEA’s Code of Ethics establishes that engineers should not practice prices that are either excessive or that deviate from the minimum amounts that are set in CREA’s tables. Figure 26. PMR indicator decomposition Figure 27. PMR indicator decomposition through state control: Brazil, 2013 through barriers to entrepreneurship: (Index scale 0 to 6 from least to most restrictive) Brazil, 2013 (Index scale 0 to 6 from least to most restrictive) Source: World Bank staff elaboration using OECD data Source: World Bank staff elaboration using OECD data Direct government participation in markets can be distortive to competition, and competitive neutrality could be strengthened. State owned enterprises (SOEs) can be found in at least 23 markets in Brazil. While SOE participation is common, especially in sector in network sectors, effectively implemented rules that guarantee competitive neutrality are critical to ensure that all market participants (public and potential private competitors) have the right incentives to operate efficiently and competitively. In this regard, publicly-controlled firms in Brazil may benefit from undue competitive advantages as they can receive financing which is not available to public companies. While Article 173 of the Brazilian Constitution stipulates that public enterprises may not receive any tax benefits that are not available to private enterprises, this does not extend to other forms of financial transfers (such as loans) and the article does not apply to any enterprise that provides a public service.44 State control is also high when measured in terms of involvement in business operations. For instance, unlike in several other countries, as of 2013, price controls still applied to pharmaceuticals. More recently, in July 2018, the government has imposed price control measures in the road cargo transport sector.                                                              41 Based on 2015 data. See also: http://www.confea.org.br/cgi/cgilua.exe/sys/start.htm?infoid=15360&sid=1206. 42 Brazil is also not party to any of the three international agreements on mutual recognition of engineering accredited programs: the Washington Accord for Engineers, the Sydney Accord for Engineering Technologists, and the Dublin Accord for Engineering Technicians. 43 See OECD (1999) for a discussion about regulatory issues in professional services. . 44 Article 175 of the Constitution and Surpreme Court decisions relating to “Casa da Moeda”, “ECT” and “Infraero” 23      While there are regulatory barriers that hinder the efficient allocation of resources across sectors, regions and firms, some critical markets also have inherent features that reduce contestability and facilitate anti-competitive behavior. This is the case for markets such as cement, fuel distribution and LPG. In 2014, Brazil’s Competition Authority CADE sanctioned 6 domestic cement companies that represented 75 percent of the domestic cement market for forming collusive agreements. While local cement sales had doubled in the past decade, prices had increased by around 33 percent. CADE estimates that customers overpaid around US$9 billion over the two decades the cartel existed.45 A total of 14 cartel agreements have been uncovered by CADE in fuel retail and LPG markets between 2002 and 2015. In one of these, subsequent consumer savings were estimated at over U$ 10 million a year.46 5. Potential productivity dividends accruing from stronger competition Boosting competition in product markets is expected to bring positive payoffs for productivity growth in the manufacturing industry. The acceleration of trade integration combined with the removal of policy induced stringencies that hamper resource allocation would reduce distortions to market functioning while boosting competition in product markets. An econometric analysis is conducted to measure the association between productivity growth and competition intensity for Brazil. The analysis draws on the methodology presented at Aghion, Braun and Fedderke (2008) and uses data at CNAE 3-digit level from 2007–2014 Pesquisa Industrial Annual (see Annex 3 for a methodological presentation). The results suggest that higher price cost margins (implying that firms face lower levels of competition intensity) are significantly associated with lower (labor) productivity growth in the following year (see Figure 28).47 The economic magnitude of the (average) “effect” is non-negligible: a 10 percent decrease from the manufacturing overall mean price cost margin of 0.14 is associated with an increase in productivity growth of 3.4 percent per year. With regards to this average “effect”, two aspects are worth highlighting. First, the productivity dividends reflect overall competition pressure and therefore include both domestic and foreign competition pressure exerted by trade. In fact, the data show that sectors (at 3-digit level) with higher import penetration present lower price cost margins (see Figure 29), which suggests that increasing import penetration can be one effective channel to increase competition pressure.48 In addition, a recent empirical analysis conducted by Kannebley Jr, Remedio and Oliveira (2017) shows that the recurrent use of protectionist trade interventions proxied by antidumping measures in the past three decades in Brazil has been associated with an increase in markups of firms that operate in protected sectors. Second, as expected, the magnitude of the (lagged) markup coefficient varies substantially across markets: the productivity growth dividend of a 10 percent decrease in average markup by sector varies from 1.43 percent (“Roasting and Milling of Coffee”) to 7.54 percent (“Magnetic and optical media”), as displayed in Figure A3.1 in Annex 3.                                                              45 http://www.reuters.com/article/2014/05/29/brazil-cement-antitrust-idUSL1N0OE2IF20140529 46 http://www.cade.gov.br/upload/Brazil_Leniencia_Program_Brochure.pdf 47 The relationship between price cost margin and productivity is negative on average and did not present a U-shaped format, as the coefficient of the squared PCM term was not statistically significant (see column 2). This result suggests that, on average, higher degrees of competition are continuously expected to increase labor productivity growth, and there are no industries that may present ‘too much competition’. 48 Import penetration data were taken from the Confederacao Nacional da Industria website (available at http://www6.sistemaindustria.org.br/gpc/externo/estatisticaAcessoSistemaExterno.faces). 24      Figure 28. Price cost margin and labor productivity growth in Brazil's manufacturing industry (3- digit sector level analysis, 2007-2014) (1) (2) *** PCM[t-1] -2.374 -2.833*** (0.167) (0.421) PCM[t-1] squared 1.442 (1.191) Constant 0.394*** 0.425*** (0.026) (0.036) Year fixed effects Yes Yes Sector (3-digit) fixed effects Yes Yes No. of observations 691 691 W0ithin R-squared 0.334 0.336 Note1: Clustered standard errors at year level are in parentheses. ***, **, and * indicate significance at 1 percent, 5 percent, and 10 percent Note2: Results are from a fixed effects OLS regression (excluding influential outliers). The dependent variable is the real labor productivity growth, and labor productivity equals value added per worker. Price cost margin is defined as (Value added - labor costs)/gross output. All variables (value added, number of employees, labor costs and gross output) are taken from PIA sector tabulations at 3 digit level. Nominal values of value added are deflated to 2007 values using the CPI as reported by the OECD. Figure 29. Import penetration vs price cost margin (at 3-digit level); pooled observations 2007-2014   25      6. Conclusions and overall policy recommendations As a result of protectionist trade policies, Brazil’s external integration is limited, which has been preventing local firms from benefiting from multiple sources of productivity gains, such as: reaping the benefits from economies of scale accrued from access to larger markets; relaxing technological constraints through access to better-quality and competitively priced inputs, and facing stronger competition (either from import competition or entry of foreign companies), which pushes firms to increase efficiency. Evidence compiled by this paper suggests there is large scope for further trade reforms in the country. Tariffs are high, use of NTMs and local content requirements is widespread, and the number (and depth) of FTAs Brazil is signatory of is quite restricted. Unilateral tariff reforms could prioritize reducing tariffs on capital goods and intermediates; reducing tariff escalation and establishing more homogeneous levels of protection; and simplifying the tariff structure by reducing the number of tariff levels. Opening to trade goes beyond tariff liberalization. In this regard, tariff liberalization efforts should be combined with a more aggressive strategy of trade agreements, especially with more technologically sophisticated countries with significant GVC integration and learning potential for Brazilian companies (e.g. Canada, EFTA and New Zealand as warm-up for US, EU and Australia). In addition, LCRs in input markets should be removed and the tax structure on services imports should be streamlined. From the trade facilitation front, the recent unified export system with an online platform that has started to function in mid-2017 represents a major improvement; however, there is still space to streamline other initiatives, such as the Authorized Economic Operator Program where ANVISA and the Army could provide a common assessment framework and a single authorization. Reducing barriers to trade would bring substantial payoffs for Brazil, according to a customized analysis developed for this paper. Estimations from a customized CGE model show that coordinated reforms at the Mercosur level would permanently accelerate overall trade (parts, components and final goods), increasing both exports and imports by 7.5 and 6.6 percent, respectively, by 2030 relative to the baseline. These reforms would also trigger a permanent expansion of economic activity: real GDP would be 0.93 percent higher than the baseline by 2030. In addition, increasing the level of depth of Mercosur would increase Brazil’s GVC-related trade with Mercosur partners: if Mercosur had all 17 legally enforceable provisions currently in force, Brazil exports in parts and components to other Mercosur members would increase 22 percent on average (US$2,287 million). Reaping the full benefits from further integration to the global economy will depend on how production factors move across firms, sectors and regions. Evidence compiled by this paper suggests that the resource allocation process in Brazil has been hampered by multiple stringencies created by distortive “behind the border” policies. Four key issues were highlighted. First, high costs of doing business tend to impose a higher burden on entry of new firms that could be potentially more efficient than incumbents. Second, lagging transport infrastructure and lack of technology diffusion policies constrain the increase of economies of scale. Third, federal spending on industrial policies has been growing at high speed and several incentive programs seem not only to be inefficient in reaching their objectives but they also appear to undermine creative destruction. Fourth, product market regulations have not been conducive to competition. Boosting competition would bring productivity gains. Estimations presented in the paper show that a 10 26      percent decrease in the average manufacturing price-cost margin, as would likely occur with greater competition, is associated with an increase in labor productivity growth of over 3 percent per year. Against this backdrop, key policy recommendations to boost trade integration can be outlined:  Reform of the tariff schedule: The new tariff schedule should be designed to give greater rationality to the protection structure. In addition to high import tax rates, there are a number of distortions resulting from both the Brazilian tariff structure and the way it has been managed, resulting in levels of effective protection badly distributed, ad hoc administration of import tariffs, reintroduction of temporary import quotas associated with tariff reductions, and costly proliferation of tariff exemptions. A new protection structure should be designed to provide long- term predictability for producers and investors in Brazil. The trade liberalization schedule should be announced in advance and be implemented gradually over a few years and should focus on simplification of the tariff schedule, limitation in the number of levels of tax rates, a reduction of the cost of imports of intermediate products and capital goods, and a reduction in the effective rates of protection.  Reducing or streamlining the application of non-tariff measures, involving, inter alia: a. Remove local content requirements, in particular in input markets. LCRs in input markets deter integration into GVCs, limit the choice of best-global inputs, and thereby depress productivity & competitiveness of user industries. b. Reduction of the tax burden on exports. Simplifying national tax regimes, reducing ancillary costs in meeting tax requirements, and eliminating tax residues levied on exports with the payment of tax credits due to exporters is essential to give predictability and greater competitiveness to domestic production. c. Reinforcement of institutions for technical regulation and product certification (e.g., INPI and INMETRO). Participation in global value chains requires property rights guarantees for technology transfer along the chain and ability to comply with standards and technical regulations. It is essential to reduce time limits in obtaining patents and to develop the capacity to participate in the definition of international norms and regulations that affect Brazilian exports. d. Reduction of services costs. Reforms should start by reducing the high levels of restrictiveness especially in financial and professional services (substantially above the regional average). It would be important to reassess the tax treatment for imports of technical services in Brazil, essential for technology attraction, business consulting, etc. and currently highly taxed (see CNI 2016). Finally, Brazil would benefit from removing barriers to services trade and investment in international trade agreements - it should start with opening restrictions in its main trade agreement, Mercosur – much more restrictive than most trade agreements in the region, and especially comparted to LAC commitments in FTAs with the US.  Reinforcement of the Trade Facilitation agenda, including: a. Accelerate the implementation of the Single Portal for exports and imports: 27      b. Creation of a single payment system for all government fees. c. Use of risk-based assessment at the border, including all agencies involved d. Authorized Economic Operator  Pursue an aggressive strategy for (deeper) trade agreements, starting by completing ongoing negotiations Mercosur-European Union, the main market for Brazilian exports, but also including moving forward in negotiations within the Americas, including Mercosur-Pacific Alliance and resuming the idea of a free trade area in the Americas. Looking forward, there is also large scope for “domestic” reforms aiming at improving the resource allocation process across firms, sectors and regions. First, reducing economywide regulatory costs of doing business would help alleviate the burdens that are disproportionately higher for new entrants. In addition, the tax structure needs to be streamlined at least to harmonize taxes across jurisdictions to boost the efficiency of resource allocation process across states and municipalities. Second, transport infrastructure needs to be streamlined which requires improving infrastructure public spending efficiency. According to World Bank (2017), Brazil, as most Latin American countries, has limited fiscal space to increase public investments, so much of what is needed to increase public spending efficiency (in transport and overall infrastructure) lies outside the infrastructure sector and has to do with broader government issues—from pro-competition reforms in construction services sectors, such as engineering and architecture (to help bring costs down) to budgeting rules that no longer solely focus on controlling cash expenditures. Third, business support programs need to be revisited and properly evaluated to identify which incentives are effectively working as well as those that should be eliminated to avoid prolonging the distortive incentive structure that private sector firms often rely on to survive. Fourth, tackling anti- competitive business practices and removing product market regulation that restricts competition represent key anchors to boost domestic competition. In this sense, trade liberalization also plays a role as import competition represents an additional channel to boost domestic competition. Finally, it is key to ensure that the removal of external barriers to integration is accompanied and coordinated with the removal of internal distortions. In addition to the progress achieved in trade facilitation and the initial steps to remove local content requirements, the removal of trade barriers can be accelerated by the elimination of regulatory barriers to export and by an in-depth assessment of NTMs; these are measures that are less costly from a fiscal standpoint and are likely to bring positive impacts on trade and possibly on jobs. As a second step, it is recommended to tackle tariff peaks and tariff escalation in the context of the flexibility Mercosur already offers (i.e. unilaterally) while advancing negotiations on new (PTAs. Meanwhile, the removal of internal market distortions will involve the combination of (medium term) policies encompassing the reduction of regulatory costs of doing business, with focus on the elimination of tax distortions across states, reduction of connectivity costs and boosting domestic competition.   28      References Aghion, P., M. Braun, and J. Fedderke. 2008. “Competition and productivity growth in South Africa”. Economics of Transition, 16(4), 741-768 Aghion, P., N. Bloom, R. Blundell, R.Griffith, and P. Howitt. 2005. “Competition and Innovation: An Inverted-U Relationship”, Quarterly Journal of Economics, 120(2), 701-728 Aghion, P., U. Akcigit, and P.Howitt. 2014 “What do We Learn from Schumpeterian Growth Theory?” Handbook of Economic Growth, Volume 2, pages 515-563. Appy, B. 2017. “Tributação e produtividade no Brasil.” In Anatomia da produtividade no Brasil, edited by Regis Bonelli, Fernando Veloso, and Armando Castelar Pinheiro, 323–60. Rio de Janeiro: Elsevier and FGV/IBRE Aw, B. Y., M.J., Roberts and D.Y. Xu. 2011. “R&D investment, exporting, and productivity dynamics”. American Economic Review, 101(4), 1312-1344 Bernard, A., S. Redding, and P. Schott. 2010. “Multiple-Product Firms and Product Switching.” American Economic Review 100: 70–97. Bustos, P. 2011. “Trade Liberalization, Export, and Technology Upgrading: Evidence on the Impact of Mercosur on Argentinian Firms.” American Economic Review 101 (1): 304–40. Canuto, O., M. Cavallari, and J.G. Reis. 2013. “The Brazilian competitiveness cliff”, Economic Premise n.105, World Bank, February. Cheng, K., S. Rehman, D. Seneviratne, and S. Zhang. 2015. “Reaping the Benefits from Global Value Chains” IMF Working Paper n.o 15/204 Corseuil, C. H., and R.L Moura. 2017. “O Impacto do Simples no Nível de Emprego da Indústria Brasileira.” Mimeo. IPEA, Rio. De Loecker, J.2013. “Detecting learning by exporting”, American Economic Journal: Microeconomics, 5: 1-21. 29      Duhaut, A. and S. Lall. 2017. “Brazil productivity: a look on internal disparities” Background paper to “Jobs and Growth: Brazil’s Productivity Agenda” World Bank. Dutz, M., R. Barroso, J. Bevilaqua Teixeira Basto, X. Cirera, C. Fleischhaker, A. Nucifora, and M. Vijil. 2017. “Business Support Policies in Brazil: Large Spending, Little Impact.” Background Paper to Um Ajuste Justo: Análise da eficiência e equidade do gasto público no Brasil. World Bank, Brasilia. Fay, M., L.A. Andres, C. Fox, U. Narloch, S. Straub, and M. Slawson. 2017. “Rethinking Infrastructure in Latin America and the Caribbean: Spending Better to Achieve More” Directions in Development. Washington, DC: World Bank. Fuentes, R., and V. Mies. 2017. “Technological Absorptive Capacity and Development Stage: Disentangling Barriers to Riches.” Mimeo Garcia-Escribano, M., C. Goes, and I. Karpowicz. 2015. “Filling the Gap: Infrastructure Investment in Brazil” IMF Working Paper N.o 15/180 Grossman, G. and E. Rossi-Hansberg 2008 “Trading Tasks: A Simple Theory of Offshoring” American Economic Review, 98(5):1978-1997. Halpern, L., M. Koren, and A. Szeidl. 2015. “Imported Inputs and Productivity.” American Economic Review 105 (12): 3660–3703 Henn, C., N.Papageorgiou, and N.Spatafora. 2015 “Export Quality in Advanced and Developing Economies: Evidence from a New Dataset”. WTO Working Paper ERSD-2015-02 Hollweg, C. and N. Rocha. 2017. “GVCs Participation and Deep Integration in Brazil” Mimeo Hummels, D. L., and G. Schaur. 2013. "Time as a Trade Barrier." American Economic Review, 103 (7): 2935-59. IMF.2015. “Making Public Investment More Efficient” IMF Policy Paper June Kannebley Jr, S., R. Remedio, and G. Oliveira. 2017. “Uma avaliacao empirica das praticas de antidumping sobre a produtividade e o poder de mercado das firmas industriais brasileiras” Revista Brasileira de Comercio Exterior (133): Outubro/Novembro Kitzmuller, M., and M. Licetti. 2012. Competition Policy: Encouraging Thriving Markets for Development. Viewpoint, Public Policy for the Private Sector, Note# 331(Note num. 331). Klapper, L., L. Laeven, and R. Rajan. 2006. “Entry Regulation as a Barrier to Entrepreneurship”. Journal of Financial Economics 82, 591-629. Lazzarini, S., A. Musacchio, R. Bandeira-de-Mello, and R. Marcon. 2014. “What Do State-Owned Development Banks Do? Evidence from BNDES, 2002–09.” World Development 66: 237–53. 30      Lederman, D., J. Messina, S. Pienknagura and J. Rigolini. 2014. Latin American Entrepreneurs: Many Firms but Little Innovation. Washington, DC: World Bank. Linde, J. and A. Pescatori. 2017. “The Macroeconomic Effects of Trade Tariffs: Revisiting the Lerner Symmetry Result,” IMF Working Paper 17/151, IMF, Washington, DC. Lisboa, M.,N. Menezes Filho, and A. Schor. 2010. “The effects of trade liberalization on productivity growth in Brazil: competition or technology?” Revista Brasileira de Economia vol.64 no.3 Rio de Janeiro Sept. 2010 Melitz, M. 2003. “The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity.” Econometrica 71: 1695–1725. Melitz, M., and G. Ottaviano. 2008. “Market Size, Trade, and Productivity.” Review of Economic Studies 75: 295–316. Muendler, M. 2004. “Trade, Technology, and Productivity: A Study of Brazilian Manufacturers, 1986-1998”, mimeo. OECD. 2015. “Participation of Developing Countries in Global Value Chains: Implications for Trade and Trade-Related Policies” OECD Trade Policy Paper No. 179 Pavcnik, N. 2002. “Trade Liberalization, Exit, and Productivity Improvement: Evidence from Chilean Plants.” Review of Economic Studies 69: 245–76. Pires, J.C, and N. Russel. 2017. “Assessing Firm-Support Programs in Brazil” IADB Office of Evaluation and Oversight, OVE, Washington, DC. Piza, C. 2016. “Revisiting the Impact of the Brazilian SIMPLESProgram on Firms’ Formalization Rates.” Policy Research Working Paper 7605, World Bank, Washington, DC. Restuccia, D. and R. Rogerson. 2017. “The Causes and Costs of Misallocation,” Journal of Economic Perspectives 31 (3): 151-74. Rios, S and P. Motta Veiga. 2014 “Política comercial e produtividade no Brasil: evolução recente e recomendações de politica.” Revista Brasileira de Comercio Exterior 120. Rossi Jr, J.L. and P. Cavalcanti Ferreira. 1999. “Evolução da Produtividade Industrial Brasileira e Abertura Comercial.” Texto para Discussao IPEA # 0651 Schor, A. 2004. “Heterogeneous productivity response to tariff reduction: Evidence from Brazilian manufacturing firms.” Journal of Development Economics, 75:373–396. Stone, S., J. Messent and D. Flaig (2015), "Emerging Policy Issues: Localisation Barriers to Trade",OECD Trade Policy Papers, No. 180 31      Sturgeon, T., L. Chagas, and J. Barnes. 2017 “Inovar Auto: Evaluating Brazil’s Automotive Industrial Policy to Meet the Challenges of Global Value Chains” Background Paper to Um Ajuste Justo: Análise da eficiência e equidade do gasto público no Brasil. World Bank, Brasilia. Taglioni, D. and D.Winkler. 2016. “Making Global Value Chains Work for Development. Trade and Development”. Washington, DC: World Bank. © World Bank. Topalova, P., and A. Khandelwal. 2011. “Trade Liberalization and Firm Productivity: The Case of India.” Review of Economics and Statistics 93 (3): 995–1009. Volpe Martincus, C., J. Carballo, and A. Graziano (2015). “Customs,” Journal of International Economics 96 (1). Winkler, D. 2017 “Productivity Spillovers from Global Value Chain Integration and the Role of Mediating Factors: Evidence from Enterprise Data” Background paper to “Jobs and Growth: Brazil’s Productivity Agenda” World Bank 32      Annex 1. Methodology for the estimation of the relationship between GVC integration and labor productivity This annex draws from the analysis presented in Winkler (2017). Using a cross-section of around 12,700 domestic manufacturing firms in 22 low- and middle-income countries from the World Bank’s Enterprise Surveys, the analysis presented in Winkler (2017) explores the relationship between global value chain (GVC) integration and labor productivity for Brazil and the full sample. Besides the standard variables from the Enterprise Survey, the database was merged with two sectoral measures of structural integration in GVCs, as captured by the measures of Eigenvector centrality: BONwin (buyer’s perspective) and BONwout (seller’s perspective). The measures were computed by Taglioni and Santoni (2014) and are based on value added trade data. The measures express the idea that the influence of a node in a trade network is proportional to the influence of its neighbors: the node’s eigenvector centrality is largely determined by the eigenvector centrality of its neighbors. The analysis assesses for the full sample and Brazil separately how GVC integration and other firm characteristics are correlated with a domestic firm’s average labor productivity. The baseline equation is estimated by ordinary least squares (OLS) and takes the following form: lnlpirst= α + βACcst + ζlncapintirst + Dr + Ds + Dt + εirst (1) lnlpirst indicates the labor productivity for firm i in subnational region r, sector s and at time t in natural logarithms, defined as value added per worker. AC designates a firm’s absorptive capacity and includes the measures as defined in the table below. lncapintirst is capital intensity in natural logarithms. The equation also includes subnational region, sector and time fixed effects. Standard errors are robust to heteroscedasticity and clustered at the country-sector level. The analysis is particularly interested in whether labor productivity for domestic firms is higher in sectors that are characterized by higher GVC integration. The variable BON_d is then introduced; it is a dummy that takes the value of 1 if a sector’s GVC integration equals or exceeds the median across all countries, and 0 otherwise.49 The analysis is replicated for Brazilian firms only which allows to compare the relationship of firm characteristics, including the extent of GVC integration at the sector level, with labor productivity.50 Box A1.1 Measures of absorptive capacity • gap = firm’s LP relative to median LP of multinational firms in sector in natural logarithms; a higher number indicates a lower gap. • tech = firm’s technology indicator, defined as iso + tech_for + website + email, where tech ϵ {0, 1, 2, 3, 4}. iso = 1 if firm owns internationally-recognized quality certification and 0 otherwise, tech_for = 1 if firm uses technology licensed from foreign firms and 0 otherwise, website = 1 if firm uses own website to communicate with clients or suppliers, email = 1 if firm uses email to communicate with clients or suppliers. The technology indicator serves as a proxy for R&D intensity which is unavailable.                                                              49 The continuous BON variables got dropped from the Brazil regressions which required using a GVC dummy variable. 50 Our data set includes 851 domestic firms and 59 foreign firms in Brazil (see Annex 1). 33      • skills = firm’s share of high-skilled labor in firm’s total labor force. • size = firm’s total number of permanent and temporary employees in natural logarithms. • aggl = region’s total number of manufacturing and services firms as percentage of a country’s total number of manufacturing and services firms. This measure is a proxy for urbanization economies (locational advantages) and covers both domestic and foreign firms. • exp = firm’s share of direct or indirect exports in firm sales. • imp = firm’s share of imported inputs in total inputs.   Results for the full sample. Table A1.1 reports the correlation between firm characteristics and labor productivity for all domestic manufacturing firms in the sample, following equation (1). The baseline regression in column 1 suggests that capital intensity is positively correlated with labor productivity which holds across all specifications. Columns 2 and 3 focus on GVC integration as buyer and seller. GVC integration is captured by a dummy which is 1 if the sectoral measure of structural integration into GVCs equals or exceeds the median value across all countries for that sector, and 0 otherwise. None of the GVC dummies show a statistically significant correlation with labor productivity in the full sample. Table A1.1. Productivity and Firm Characteristics, All Domestic Manufacturing Firms, OLS Dependent variable: lnlpirst GVC dummy = 1 if sectoral GVC integration >= median across all countries Continuous GVC integration measure (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) lncapintirst 0.2580*** 0.2580***0.2580*** 0.0379*** 0.2403*** 0.2577*** 0.2526*** 0.2579*** 0.2526***0.0348*** 0.2580*** 0.2578*** 0.0324*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) BONwincst -0.0353 -0.0950 0.7058 -4.7013 (0.327) (0.284) (0.568) (0.128) BONwoutcst -0.0275 0.1520* 3.7733 23.1776*** (0.526) (0.090) (0.303) (0.001) gapirst 0.8626*** 0.8578*** 0.8620*** (0.000) (0.000) (0.000) techirst 0.2226*** 0.0313*** 0.0342*** (0.000) (0.006) (0.002) skillsirst -0.0141 0.0092 0.0089 (0.806) (0.588) (0.601) sizeirst 0.1267*** 0.0167*** 0.0140** (0.000) (0.007) (0.025) expirst 0.4301*** 0.0299 0.0243 (0.000) (0.540) (0.562) impirst 0.0031*** 0.0006** 0.0007** (0.000) (0.044) (0.014) constantt 5.9605*** 5.9939*** 5.9733*** 8.2047*** 5.8319*** 4.5451*** 5.3609*** 5.8357*** 4.5561*** 8.5954*** 5.8501*** 5.4570*** 6.2394*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Obs. 7,716 7,716 7,716 7,160 7,716 7,621 7,716 7,670 7,670 7,001 7,716 7,716 7,001 R-squared 0.48 0.48 0.48 0.92 0.50 0.48 0.49 0.49 0.48 0.93 0.48 0.48 0.93 Source: Own computations based on Farole and Winkler (2012) and Taglioni and Santoni (2014). Note: p*<0.1, p**<0.05, p***<0.01 (p-values in parentheses). All regressions include sector, subnational region, and year fixed effects. Standard errors are clustered at the country-sector level. Most other firm characteristics, however, are positively correlated with labor productivity, including a lower technology gap relative to multinational firms in a sector (column 4), a higher technology intensity (column 5), a larger firm size (column 7), a higher export share (column 8), and a higher share of imported inputs (column 9). Technology gap, in particular, explains the 34      largest portion of a firm’s labor productivity, as indicated by the high R-squared of 0.92. Surprisingly, a higher skill intensity is uncorrelated with labor productivity in the full sample (column 6). Controlling for all firm characteristics simultaneously reveals that most of the firm characteristics continue to be positively correlated with labor productivity, although their coefficient sizes become smaller due to the large explanatory power of technology gap (column 10). Interestingly, firms operating in sectors with a high GVC integration on the selling side now show a positive relationship with labor productivity. This probably explains why a higher export share at the firm level is no longer significantly correlated with labor productivity. Columns 11 to 13 use the continuous GVC measures instead of the GVC dummy. While the GVC measures show no correlation when considered individually, structural integration into GVCs on the selling side shows a statistically significant effect when all control variables are included (column 13). Reassuringly, the coefficients of the other firm characteristics are similar in terms of size and significance. Table A1.2 shows the correlation between firm characteristics and labor productivity for the Brazilian subsample only. Capital intensity is more strongly correlated with labor productivity compared to the full sample in Table A1.1. Firms operating in sectors with a higher GVC integration as a buyer also show a higher labor productivity (column 2). The model could not be estimated for GVC integration as a seller, since all sectors in the Brazilian sample were assigned a dummy of 1, i.e. exceeded the median GVC integration across the 22 countries in the full sample. Table A1.2: Productivity and Firm Characteristics, Brazilian Manufacturing Firms, OLS Dependent variable: lnlpirst GVC dummy = 1 if sectoral GVC integration >= median across all countries (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) lncapintirst 0.5661*** 0.5661*** 0.5661*** 0.0000 0.5518*** 0.5649*** 0.5634*** 0.5654*** 0.5651*** 0.0000 (0.000) (0.000) (0.000) (0.189) (0.000) (0.000) (0.000) (0.000) (0.000) (0.194) BONwincst 0.7005*** 1.6295*** (0.000) (0.000) BONwoutcst 0.0000 0.0000 (.) (.) gapirst 1.0000*** 1.0000*** (0.000) (0.000) techirst 0.1700** -0.0000 (0.039) (0.180) skillsirst -0.0513 0.0000 (0.773) (0.366) sizeirst 0.0791 -0.0000 (0.249) (0.723) expirst 0.1779 -0.0000 (0.342) (0.514) impirst 0.0009 0.0000 (0.758) (0.429) constantt 3.2524*** 3.2524*** 3.2524*** 9.6447*** 3.1709*** 3.3272*** 3.0796*** 3.2538*** 3.2456*** 9.6447*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Obs. 844 844 844 704 844 840 844 843 843 698 R-squared 0.48 0.48 0.48 0.92 0.50 0.48 0.49 0.49 0.51 1.00 • Source: Own computations based on Farole and Winkler (2012) and Taglioni and Santoni (2014). • Note: p*<0.1, p**<0.05, p***<0.01 (p-values in parentheses). All regressions include sector, subnational region, and year fixed effects. Standard errors are clustered at the country-sector level. 35      Among other firm characteristics, technology gap explains a larger portion of labor productivity in Brazil (column 4) compared to the full country sample. A higher technology intensity, by contrast, is less strongly correlated with labor productivity in Brazil (column 5). Surprisingly, other characteristics like firm size, export share, or imported input share are uncorrelated with labor productivity in Brazil, which is in contrast to the findings using the full sample. The full specification in column 10 confirms the positive correlation between labor productivity and a lower technology gap as well as for firms operating in sectors that are highly integrated on the buying side. Technology intensity, however, no longer shows a statistically significant correlation. In order to allow for sectoral variation of the GVC dummy on the selling side in Brazil, Table A1.3 replicates the regressions using an alternative GVC dummy which takes the value of 1 if the measure of structural integration into GVCs equals or exceeds the 75th percentile across all countries for that sector, and 0 otherwise. In other words, the threshold to be assigned a dummy of 1 is higher for sectors in Table A1.3 compared to Table A1.2. Table A1.3. Productivity and Firm Characteristics, Alternative GVC Dummy, Domestic Manufacturing Firms, OLS Dependent variable: lnlpirst GVC dummy = 1 if sectoral GVC integration >= 75 pctl across all countries Brazilian sample Full sample (1) (2) (3) (4) (5) (6) lncapintirst 0.5661*** 0.5661*** 0.0000 0.2580*** 0.2580*** 0.0330*** (0.000) (0.000) (0.194) (0.000) (0.000) (0.000) BONwincst 0.0908** 0.8218*** 0.0131 0.0907 (0.016) (0.000) (0.768) (0.347) BONwoutcst 0.7005*** 1.0793*** 0.0627 0.3233** (0.000) (0.000) (0.213) (0.037) gapirst 1.0000*** 0.8604*** (0.000) (0.000) techirst -0.0000 0.0344*** (0.180) (0.002) skillsirst 0.0000 0.0050 (0.366) (0.768) sizeirst -0.0000 0.0132** (0.723) (0.042) expirst -0.0000 0.0091 (0.514) (0.819) impirst 0.0000 0.0007** (0.429) (0.014) constantt 3.2524*** 3.2524*** 9.6447*** 5.9588*** 5.9614*** 8.6132*** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Obs. 844 844 698 7,716 7,716 7,001 R-squared 0.50 0.50 1.00 0.48 0.48 0.93 Source: Own computations based on Farole and Winkler (2012) and Taglioni and Santoni (2014). Note: p*<0.1, p**<0.05, p***<0.01 (p-values in parentheses). All regressions include sector, subnational region, and year fixed effects. Standard errors are clustered at the country-sector level. The results show that Brazilian firms with a high GVC integration as a buyer continue to show higher labor productivity, but their coefficient size is lower using the 75th percentile threshold (column 1). In other words, there seem to be decreasing labor productivity gains to increased GVC integration on the buying side in Brazil. Focusing on firms that operate in sectors with a high GVC 36      integration as a seller now also shows a positive correlation with labor productivity (column 2). The coefficient size is much larger than for GVC integration on the buying side. Taking into account all firm level controls in column 3 confirms the positive relationship for Brazilian firms that are highly integrated on the buying and selling sides. Besides GVC integration, technology gap is the only other firm characteristic in the Brazilian sample that explains differences in labor productivity. The results for the full country sample show that none of the individual effects of GVC integration is significant (columns 4 and 5). Only when all firm characteristics are being simultaneously controlled for do firms operating in sectors with high GVC integration on the selling side show significantly higher productivity levels (column 6), but the coefficient size is smaller compared to the Brazil sample only. In addition, firms with a lower productivity gap, higher technology intensity, larger firm size, and larger share of imported inputs are more productive. The findings for the full country sample are in line with the results using the GVC integration dummies with the median threshold (Table A1.1, column 10). In summary, Brazilian firms operating in sectors with high GVC integration as a buyer show higher productivity levels using the median GVC integration level as threshold. Since all sectors in Brazil show a high extent of GVC integration as a seller using the median threshold, the 75th percentile is used as alternative threshold. The results show that operating in sectors with very high GVC integration as a seller explains a large portion of Brazilian firms’ productivity. While very high GVC integration on the buying side continues to be positively related to labor productivity, the coefficient size becomes smaller. 37      Annex 2. Methodology for the estimation of the impact of deep integration through FTAs on GVC trade This annex draws from the analysis presented in Hollweg and Rocha (2017). The relationship between deep agreements and GVC-related trade is formally estimated using a structural gravity equation. An augmented gravity equation is estimated for 189 countries, using data from 1990 to 2014, to investigate the relationship between the depth of an agreement and GVC-related trade. The depth of an agreement is captured by the number of legally enforceable provisions that it includes. This methodology has been extensively used by economists to test empirically the determinants of trade flows, and to estimate the effect of preferential trade opening on trade flows. Data set. The New World Bank data set on the content of PTAs, is an extension of the Horn et al. (2010) and WTO (2011) data sets and contains 280 PTAs signed by 180 countries between 1980 and 2015. The methodology of Horn et al. (2010) is followed in order to define the content and the legal enforceability of PTAs. As a first step, a set of 51 policy areas covered in PTAs is identified. These areas can be classified into two different groups. The first group is represented by WTO+ provisions which fall under the current mandate of the WTO and are already subject to some form of commitment in WTO agreements. The second group of policy areas, which is denoted as WTO- X provisions, includes those obligations that are outside the current mandate of the WTO. Table A2.1 lists the 51 policy areas that are identified. Table A2.1: Provisions included in PTAs WTO+ provisions WTO-X provisions Public FTA Industrial Competition Policy Political Dialogue Approximation of Legislation Procurement FTA Agriculture STE Movement of Capital Social Matters Innovation Policies Customs TRIMs Investment Financial Assistance Audio Visual Export Taxes SPS IPR Cultural Cooperation Health AD CVM Environmental Laws Anti-Corruption Illicit Drugs TBT TRIPs Information Society Taxation Human Rights State Aid GATS Regional Cooperation Data Protection Mining Agriculture Education and Training Money Laundering Visa and Asylum Industrial Cooperation Terrorism Labor Market Regulation Public Administration Illegal Immigration Economic Policy Dialogue Statistics Nuclear Safety Research and Technology Consumer Protection Energy SME Civil Protection Note: Provisions in blue are included in the Mercosur agreement and are in force. Provisions in red are included in Mercosur but are not in force 38      The legal enforceability of the PTA obligations is established according to the language used in the text of the agreements. In other words, it is assumed that commitments expressed with a clear, specific and imperative legal language, can more successfully be invoked by a complainant in a dispute settlement proceeding, and therefore are more likely to be legally enforceable. In contrast, unclearly formulated legal language might be related with policy areas that are covered but that might not be legally enforceable. Gravity equations are derived from models that seek to explain or predict the relationship between a (dependent) variable (in this case bilateral GVC-related trade) and a set of other (independent or explanatory) variables whose values can be estimated (in this case elements of deep integration). Endogeneity occurs when both the variable being explained (the left-hand side variable in the equation) and the explanatory variable (the right-hand side variable in the equation) may be determined by a third factor not in the model. For example, firms that want to invest in a country may also lobby for free trade agreements. Consequently, a free trade agreement may not increase FDI, but both FDI and FTAs may both come about due to perceived economic benefits of firms and their political lobbying efforts. In order to control for endogeneity and for the existence of zero trade flows, the following structural gravity regression is estimated for a set of 189 countries between 1990 and 2014 using Poisson pseudo maximum‐likelihood (PPML)51: ℎ (1) Where is a measure of GVC-related trade between country i and j at time t and it is captured with gross trade flows in parts and components from the UN Broad Economic Categories classification (BEC)52 ℎ is a measure of the depth PTAs. A statistically significant and positive coefficient β1 implies that signing a deeper agreement is associated with greater GVC- related trade. This variable is calculated as the number of enforceable provisions that are included in a certain agreement. The are a series of fixed effects: i for importer, j for exporter and t is year. Finally, εijt is the error term. The following calculations are based on the results from the gravity estimation presented above. The following scenarios in terms of depth of a PTA are considered in order to assess the potential impact of a deeper Mercosur and of a potential agreement between the Pacific Alliance and Mercosur on GVC-related trade:  Mercosur 17 scenario: all the 17 provisions that are currently covered in Mercosur enter into force. This scenario shows Brazil’s currently “money left on the table” by not including disciplines such as customs, export taxes, public procurement, TRIPs, stated aid, countervailing measures, TRIMs and state enterprises, as well as competition policy, movement of capital and IPR provisions in force.                                                              51 See Piermatini and Yotov (2016). 52 Parts and components include non-fuel BEC intermediates (111,121,21,22,42 and 53). 39       Deepest within Pacific Alliance scenario: Mercosur is renegotiated as deep as “Colombia – Mexico” which is the deepest bilateral agreement within the pacific alliance countries with a total of 19 legally enforceable provisions. In this scenario disciplines such as investment and visa and asylum are included in addition to the “Mercosur 17” scenario.  Deepest outside Pacific Alliance scenario: Mercosur is renegotiated as deep as “Peru – Republic of Korea” which is the deepest bilateral agreement of a pacific alliance member, including a total of 30 disciplines. Additional disciplines that are not present in the previous scenarios include labor market regulation, consumer protection, cultural cooperation, research and technology and agriculture. Improvements in the level of depth of Mercosur will further increase Brazil’s GVC-related trade with Mercosur partners. Rough calculations based on the gravity estimations suggest that if Mercosur had all 17 legally enforceable provisions currently in force, Brazil exports in parts and components to other Mercosur members would increase 22 percent on average53 (US$2,287 million), while imports in intermediates form other Mercosur members would increase around 37 percent more (US$2,660 million). In particular, exports to Argentina would increase on average US$1,800 million, while imports would increase around US$1,916 million. In the same way, exports to Paraguay would increase about US$343 million, as imports would increase on average US$304 million (see column a of Table A2.2). These results represent a lower bound and imply the entering into force of disciplines that are already covered. Re-negotiation of a deeper Mercosur could increase GVC related trade up to US$30 billion, where exports to Mercosur members would growth on average 55 percent (US$5,654 million) whereas imports from other Mercosur members would almost double to US$14,332 million (see columns b and c of Table A2.2). Table A2.2: Change in Brazil's GVC-related trade within Mercosur under different scenarios (a) (b) (c) "Mercosur 17" "Deepest within PA" "Deepest outside PA" scenario scenario scenario Depth 17 19 30 (USD'MM) orts GVC-related Exports 22.2% 26.7% 54.9% Argentina $1,802 $2,171 $4,456 Paraguay $343 $413 $848 Uruguay $142 $171 $350 Total Exports $2,287 $2,755 $5,654 GVC-related Imports 36.7% 44.7% 97.9% Argentina $1,916 $2,333 $5,106 Paraguay $304 $370 $811 Uruguay $440 $536 $1,172 Total Imports $2,660 $3,239 $7,089 Notes: Depth increases from 6 to 17 in the “Mercosur 17” scenario, to 20 in the “Deepest within PA” scenario and to 30 in the “Deepest outside PA” scenario.                                                              . . ∗ 53 The following formula is used to calculate the percentage change in GVC related trade 1 . . ∗ 0.222 40      Annex 3. Methodology for the estimation of the association between intensity of competition and productivity growth in the Brazilian manufacturing industry Following the standard in the literature, we measure market power using the price-cost margin (PCM), which is a Lerner Index. The PCM measure margins (i.e., the difference between price and marginal cost) as proportion of price. In the absence of information on price and marginal cost, the extent of pricing power in an industry is proxied by the difference between value added and labor costs as a proportion of output (all measured in current prices), as follows: value added labor costs PCM ≃ , 1 output where j denotes the sector and t denotes the respective year (varying from 2007 to 2014). Gross output, valued added and labor costs are all extracted from PIA sectoral tabulations at CNAE 3 digit level for manufacturing industry (sectoral tabulations are available at SIDRA IBGE system at https://sidra.ibge.gov.br/pesquisa/pia-empresa/tabelas). All nominal values were deflated to 2007 values using Brazil’s CPI as reported by the OECD. Due to lack of data, financial costs of capital are not included in the average costs. However, Aghion et al (2005) show that excluding costs of capital from the Lerner measure does not affect the results given that these costs are relatively small and constant over time. Changes in PCM within a sector drive changes in productivity, while the different levels of PCMs across sectors are not indicative of differences in productivity levels. Typically, the capital stock or cost and the capital rent as a fraction of value added does not change dramatically from year to year within one sector. We use real labor productivity growth as our measure of productivity growth. We calculate real labor productivity by sector j as real valued added in sector j (CNAE 3 digit) per worker. We use the average number of employees in sector j in the year t– also extracted from PIA tabulation - as the measure of employment. Using contemporaneous values of the measures to evaluate the relation between market power and productivity growth could be problematic. Higher margins could be the result, rather than cause, of innovation and changes in productivity growth. Similarly, the cost-advantage gained from innovation could translate into higher margins. We, therefore, address this problem by relating PCMs from the preceding year (denoted as “[t-1]”) with changes in contemporaneous productivity growth as done in other studies (e.g., Aghion, Braun and Fedderke, 2008). Exceptional growth in labor productivity can occur independently from competition firm’s innovation efforts. The analysis therefore accounts for productivity shocks that occur economy-wide at specific points in time and for differences across industries in the growth rates of productivity that are unrelated to competition levels and do not change over time by including industry and year fixed effects. Recent studies (e.g., Aghion et al 2005, 2008) have shown that the relationship between market power and productivity growth could be non-linear and so we allow for that by including the squared term of PCM in the regression analysis. Based on Aghion, Braun and Fedderke (2008), we estimated the following fixed effect regression: 41      ln LP , /LP , α βPCM γ PCM θ sector + δ time ∈ 2 where ∆LP , , /LP , is defined as the growth rate of real labor productivity in sector j, from year t-1 to t. The term PCM denotes the one year lagged mark-up in (sub) sector j, as computed in equation (1). The sector level observations are not assumed to be independent within each year, so that we compute significance levels using errors that are clustered at the year level. If competition spurs productivity growth, we would expect a negative coefficient for PCM. Table A3.1 shows the average values of real labor productivity growth and price cost margin in Brazilian manufacturing industry. Table A3.1 Average (real) labor productivity growth and price cost margin in Brazilian manufacturing industries Productivity Price-cost margin growth 2008 0.0688 0.1469 2009 -0.0618 0.1432 2010 0.0926 0.1509 2011 -0.0217 0.1432 2012 0.0056 0.1389 2013 0.0316 0.1426 2014 -0.0481 0.1291 Total 0.0096 0.1421 Note: Real labor productivity = real value added per worker. Valued added is taken from Table 1855. The average number of employees in the year is taken from Table 1841. Nominal values of value added are deflated to 2007 values using the CPI as reported by the OECD. PCM = (Value added - labor costs)/gross output. Labor costs equals personnel expenses relating to salaries, withdrawals and other remunerations and is taken from Table 1844. Gross output is the gross production value and is taken from Table 1855. Results exclude influential outliers. 42          % 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 Roasting and milling of coffee 8.00 1.43 Biofuels Tanning and other leather preparations Artificial and synthetic fibers Electromedical and electrotherapeutic… Footwear (of any material) Manufacture and refining of sugar Luggage and of leather articles Fish preservation and manufacture of fish… Tobacco products Articles of concrete/cement/plaster and… Cement Installation of machinery and equipment Footwear Textile yarn and fabrics and textile finishing Machinery and equipment for mineral… Pharmaceutical and veterinary Other food products Wearing apparel and accessories Furniture Jewelery/costume jewelery and allied products Preserved fruit and vegetables Plastics products Printing and publishing Medical/dental and optical instruments and… Construction of vessels Reconditioning and recovery of engines for… Maintenance and repair of machinery and… Milling and manufacture of grains and seeds for which the interaction coefficient was statistically significant. Organic chemicals Other non-metallic mineral products and stones Ceramic products Aircraft Other fabrics Musical instruments Preparation and spinning of textile fibers Paint/varnish/lacquer and allied products Wood sawing Engines/pumps/compressors and… Toys and games Other textile products Machinery and equipment for specific… Various chemical products Weaving (except knitting) Wood products manufacturing (except… 43  Builders' carpentry and joinery Fishing and sporting articles Machine tools Household appliances Printing and reproduction services Other fabricated metal products nec Rubber products Generators/transformers and electric motors Lamps and lighting equipment Batteries and accumulators Miscellaneous paper and paperboard products Tractors and machinery and equipment for… Electronic components Cabins/coachwork and trailers for motor… Knitted and crocheted fabrics Pharmaceuticals Pesticides and household disinfectants Optical/photographic and cinematographic… Paper and paperboard containers and coated… Forging/stamping/powder metallurgy and… Non-alcoholic beverages Dairy products Communication equipment Pig iron and ferro-alloys Alcoholic beverages Other general-purpose machinery and… Measuring/testing and checking apparatus… Electrical equipment and apparatus Electricity distribution and control equipment Cutlery/wickerwork and tools Casting Knitted and crocheted articles Steel tubes (except seamless) Audio and video equipment Glass and glass products Petroleum products Paper and paperboard Parts and accessories for motor vehicles Soap/detergent and perfumes Industrial processing of tobacco Fire arms/ammunition and military equipment Computer equipment and peripherals Reproduction of recordings Tanks/reservoirs/metallic and boiler Cars/vans and buses Coke Inorganic chemicals Figure A3.1. Estimated “effect” on productivity growth from a 10 percent decrease in average markup by (3 digit) sector Metallurgy of non-ferrous metals Pulp and paper pulp Note: productivity growth dividend at sector level correspond to the coefficient of the (3-digit level) sector interaction with the PCMt-1 variable. The figure shows only the sectors Steel Resins and elastomers Trucks and buses Transport equipment Vegetable and animal fats and oils Magnetic and optical media 7.54     Annex 4. Is regional development credit effective in increasing local productivity? This annex draws from the analysis presented in Duhaut and Lall (2017). To assess whether regional development credit is effective in increasing local productivity the areas that received credits between 2008 and 2014 and those that do not are compared. To understand the impact of the allocation of BNDES credits and Constitutional Funds on local areas labor productivity, the first analysis compares the performance of areas with similar characteristics before the program. As areas benefiting from credits might further differ in trends and characteristics, weights based on the propensity score method developed by Rosenbaum and Robin (1983) are used. Finally, the effects of the BNDES Credits and Constitutional Funds on local labor productivity are estimated using a difference-in-difference approach that controls for unobservable and observable invariant characteristics. The first step recovers the propensity for each area to have its productive units being allocated BNDEs Credits (first type of credits), Constitutional Funds (CF, second type of credits), or both (third type of credits). This is done using a multinomial logit:     ∑ where the variables used are the deflated GDP and population growth between 2002 and 2006, per capita GDP in 2006, and variables capturing local natural advantages such as distance to state capital, temperature in the winter and pluviometry. This analysis allows for an understanding of the type of areas that receive funds and to match areas that exhibit the same dynamic before the analyzed period (the parallel trend assumption). This approach generates weights to correct for the difference in the covariates in the different type of credits groups (inverse probability weights). As the inverse probability weight approach is known for being unstable for high values of the propensity score (Frolich, 2004), the weights are trimmed at the 95th centile on the right. This means that, based on the weights, areas that are not likely at all to receive any of the three types of funds are dropped from the sample, which is a potential cause of bias in the estimation. Finally, a weighted differences-in-differences estimator is used to recover the impact of receiving one of the types of credits or the average treatment effect on the treated of each type of credit, compared to receiving nothing. This estimator uses the inverse probability weights to estimate the following equation: log ∗ Where log denotes the logarithm of the productivity index ,  represents the time dummies for 2008 - at the start of the observed credits- and 2014 -at the end- and represents the type of credit received. The parameter is thus the differences-in-differences parameter. This approach compares minimally comparable areas before and after receiving funds and credits or not. It allows for filtering the effect of unobservable and time invariant factors that might influence local labor productivity. The table presented includes leads: if the credits attributed during the time-period considered influence the productivity but not the other way around, indicator variables for the type of credits received before the period considered should not be significant. 44      This approach recovers a causal relationship under several hypotheses. First, no unobserved elements correlated to the credits would also influence the outcome during the time period. Similarly, if credits allocated to firms from another minimally comparable area influence local productivity in one area, the parameter will not measure the effects of the credits themselves. One important issue is the lack of information on past credit allocation. If considered unobserved past credit allocation do not have long-lasting effects on productivity, such that the difference in changes in productivity between minimally comparable areas between 2008 and 2014 can be attributed to the credits only. Given the lack of data on previous credits, one should be careful when interpreting the results: if the credits previously attributed have long lasting impact, the coefficients could be biased downward or upward, depending on which minimally comparable areas they were allocated to. In the limiting case where the current credits allocated have no impact on current performance, the estimated parameter just measure the effect of past credits, or other time varying unobserved heterogeneity. 45