Trade barriers in public procurement

Transcription

1 Trade barriers in public procurement Alen Mulabdic Lorenzo Rotunno Preliminary Abstract In this paper, we estimate trade barriers in government procurement, a market which on average accounts for about 13% of GDP. We find that trade frictions are significantly different between public procurement and international markets involving private firms only. We then investigate the role of trade agreements in reducing those trade barriers. Our gravity estimates suggest that joining the EU doubles imports by public authorities, while the trade effect in private markets is relatively lower. Other FTAs with specific provisions on public procurement also have a trade-boosting effect, while the WTO GPA doesn t seem to have any impact. Despite these liberalising effects, public procurement markets remain almost twice as home-biased as private ones. Trade cost worldwide unduly increase domestic public procurement purchases of services, while the effect is less pronounced for manufacturing. Further tests provide suggestive evidence for a political motive behind the higher trade barriers in public procurement. Keywords: Home bias, Public sector, Institutions. JEL codes: F5, H5. We are grateful to seminar participants at Oxford, InsTED Workshop and ETSG Munich for very useful comments. PhD candidate, Dept. of International Economics, The Graduate Institute, Geneva. Case postale 136, 1211 Geneva, Switzerland. alen.mulabdic@graduateinstitute.ch. Aix-Marseille University (Aix-Marseille School of Economics), CNRS & EHESS. Chemin du Château Lafarge, Les Milles, France. Office B987/988; Tel.: +33 (0) lorenzo.rotunno@univ-amu.fr 1

2 1 Introduction Public procurement is a major market, accounting for about 13% of GDP in OECD countries (OECD, 2013a). Given this sheer size, purchases of goods and services by the public sector are under constant scrutiny in an attempt to realise cost savings 1. Ensuring a level playing field for potential suppliers especially at the international level is regarded as an important step towards a more efficient public sector (OECD, 2013b). Governments have thus committed to greater market access in public procurement through the new WTO Government Procurement Agreement (GPA) as well as targeted provisions within regional trade agreements. At the same time, the more or less subtle buy-national provisions included in different fiscal stimulus packages remind us that governments often favour domestic suppliers in their purchasing decision 2. In this paper, we try to assess and quantify trade barriers in public procurement markets. We employ data from the Trade in Value-Added (TiVa) database of the OECD on 56 countries and four years. We take the Public Administration output and the General Government expenditures columns of the inter-country input-output (ICIO) tables to measure bilateral shipments to the public authorities. Importantly, trade values are distinguished by type of goods and services. First, we look at descriptive trends in the data. The public sector spends considerably more on services than on manufacturing compared to the private sector. What s more, public procurement imports as a share of total public procurement are on average less than imports by private firms, although important heterogeneity emerges across countries. To move beyond descriptive evidence and estimate trade barriers in public markets, we apply a canonical gravity framework (see Head and Mayer, 2015 for a review) relating bilateral trade in public and private markets to different variables proxying for trade costs (of their inverse), controlling for multilateral resistance terms. Results show that bilateral trade barriers are different across the public and private markets. Transport costs proxied by distance matter less for public markets than for the private ones in manufacturing, while they hinder more services imports by the governments. We then focus on more policy-driven determinants of trade such as for EU membership, entry into the WTO 1 We use the terms government, public sector and public authorities interchangeably to indicate public institutions that are buyers in the public procurement market 2 The use of exception measures to competitive tendering in government procurement increased in 18% of OECD countries between 2008 and 2011, after the financial crisis. 2

3 agreement on government procurement (GPA) and inclusion of market access provision on public procurement in Free Trade Agreements (FTAs). To alleviate possible endogeneity bias, we follow Baier and Bergstrand (2007) and Bergstrand et al. (2013) and include in the estimating equation bilateral fixed effects to capture time-invariant determinants of trade and time-varying border effects that control for international relative to intranational trade costs. We find a sizeable effect of EU entry on public imports of both manufacturing and services, suggesting that EU directives aimed at opening up public procurement markets, while still far from creating a single market, have been instrumental in increasing public purchases of goods and services from abroad. Estimates suggest that trade in public markets between two countries almost doubles when both are EU members. Besides the special case of the EU, inclusion of liberalising provisions on public procurement in other FTAs has also a significant trade creating effect. The results say that imports goes up by between 20 and 30% when two countries sign an FTA with such provisions. The trade-creating effects of the EU and FTA provisions are more pronounced for service trade, where the effect is significantly higher in public markets than in the private ones. As for the WTO GPA, we do not find any (or, if anything, negative) significant impact on cross-border procurement. Here however, EU membership might actually absorb part of the impact of the WTO GPA given that most members are also EU countries. While these results shed light on the drivers of trade costs (and policies aimed at reducing those) in public markets relative to private ones, they do not provide a measure of protectionism in the two markets. To identify such a measure, we follow two paths well known in the gravity literature. One looks at the (partial) border effect famously introduced by McCallum (1995) and subsequently estimated with different settings and techniques (Anderson and van Wincoop, 2003; Chen, 2004; de Sousa et al., 2012). The idea is to estimate the direct effect of crossing a border on international relative to domestic trade. Since the ICIO data provides the value of internal trade, we can estimate this border effect as the coefficient on a dummy for same-country pairs, after suitably controlling for internal distance and other bilateral determinants of trade as discussed above. All estimates point to a large border effect, confirming the findings of the literature (de Sousa et al., 2012). Comparing public and private gravity, we find that borders in public procurement markets are significantly thicker than in private markets, especially in services. However, panel estimates show that the border effect has declined over time in the public more than in the private sector. 3

4 The border effect provides a direct but partial measure of protectionism as it does not take into account the underlying tendency of private and public agents to source goods and services locally. The concept of home bias overcomes this shortcoming as it measures the amount of actual internal trade relative to counterfactual internal trade in an hypothetical frictionless trade scenario. Following Anderson and Yotov (2010b) and more recently Agnosteva et al. (2014), we exploit the structure of the gravity model to estimate an index of Constructed Home Bias. Results confirm what anecdotal evidence suggests: home bias in public procurement is huge and higher on average than in the private markets. This is true for both manufacturing and services, although the difference is higher for services. Consequently, when we aggregate at the country-level, the difference is even starker: governments are much more home biased than firms in their purchasing decisions. We then explore the role of political economy factors in explaining such a protectionist bias in public procurement. Theory and anecdotal evidence suggest two distinct but related forces to explain the higher propensity of governments to favour domestic over foreign suppliers of goods and services. One is the level of transparency and corruption in public procurement procedures, which is often related to the broader quality of public institutions. The idea, evoked in efforts at the international level to make awards of public contracts more open and transparent, is that foreign firms have greater difficulties than domestic ones to strike deals with more corrupt and opaque public authorities. Another type of political-economy trade barriers specific to the public sector can be the extent to which the incumbent feels the need to please the electorate, which is often referred to the level of political or electoral accountability. The buy-national provisions included in stimulus packages are often politically justified with the need to gain political support from the electorate or special interest groups. This type of pork-barrel procurement is relevant to public authorities when choosing their suppliers, while it shouldn t be important for private firms. To test for these hypothesis, we augment the gravity equation by interacting the same-country dummy with different measures of corruption and fragmentation of the political system, the latter being proxies for the level of political accountability in our sample composed mainly of democracies. Results show that corruption does affect the protectionist level of public and private markets equally. Therefore, there doesn t seem to be a specific effect on public procurement. Having in mind the inherent measurement errors 4

5 of these country-level indicators, these findings resonates well with the ample anecdotal evidence on corruption scandals related to public contracts between governments and foreign firms. When dealing with corrupt officials or opaque procedures, domestic firms do not seem to have a competitive hedge over their foreign rivals. We do find a differential effect of political accountability, as measured by an index of legislative fractionalisatiion, on trade barriers in public procurement. Results are however significant mainly for manufacturing goods and less robust using similar indicators, such as the share of parliament seats held by the largest government party or a broader index of political constraints from Henisz (2000). Taken together, this gives rather suggestive evidence that public authorities favour more domestic supplies over foreign ones when the government is institutionally more unstable and hence needs to cater more to political incentives. This paper contributes to the relatively scant academic literature on public procurement in the context of international trade. Baldwin (1970) was the first to formally analyse the role of government expenditures in a traditional factor proportions model of international trade. His findings that discrimination in public expenditure is inconsequential for trade flows and specialisation were revised and confirmed only partly in oligopolistic settings (Miyagiwa, 1991) and with imperfect information (McAfee and McMillan, 1989). Back to general equilibrium models but in a setting with increasing returns to scale, Brulhart and Trionfetti (2004) find that trade barriers in government expenditure can actually change the patterns of specialisation, while Trionfetti (2001) identifies a significant impact of home-biased public procurement on agglomeration following trade liberalisation. In all these papers, home bias in the public sector is treated as parameter. I thus extend the existing work by estimating home bias in the public sector, in absolute terms and relative to the private sector. A few papers have tried to quantify trade barriers in public procurement. Trionfetti (2000) uses domestic input-output tables, which are part of the ICIO data used here, for seven European countries to compare import penetration ratios across public and private sectors. We go beyond these descriptive evidence and employ a gravity model to estimate trade barriers in public procurement. The empirical framework of the gravity equation should further filter out all measurement errors that inevitably affects the ICIO data. Riker (2013) exploits comparable ICIO data from WIOD (Timmer, 2012) to estimate the share of foreign value added in government consumption and compare it to the same same share 5

6 for household consumption. His descriptive evidence suggests that foreign suppliers are able to penetrate more public procurement markets indirectly by serving domestic firms. While the value-added decomposition of gross shipments is relevant to the political economy explanations behind home bias in the public sector, its treatment in a bilateral structural gravity framework is still not well-established 3. Our analysis draws extensively from the large literature on the gravity model of trade (Head and Mayer, 2015; Anderson, 2011) to estimate trade barriers in the public sector. In doing so, we do not attempt to develop a fully-fledged theoretical model that explain, for instance, the allocation of public and private expenditures across sectors. Owing to the separability between allocation of resources within and across countries that is common to many models of trade, we are able to infer trade costs in a conditional general equilibrium setting (Anderson and van Wincoop, 2004), i.e. taking as given the allocations of resources across type of goods and services in the public and private sectors. Further, our work expands the literature on the partial equilibrium effects of trade agreements (Baier and Bergstrand, 2007; Bergstrand et al., 2013) and its provisions (Kohl et al., 2013; Dür et al., 2014) by focusing on trade where the public sector is the buyer 4. Within the vast literature on empirical models of trade, we contribute also to a sub-field that looks at the role of institutions and corruption in hindering trade. Anderson and Marcouiller (2002) incorporate the role of institutions in an empirical model of import demand and finds that poor institutions act as a significant tax on trade. Subsequent work has confirmed this finding in different gravity-like empirical frameworks (Berkowitz and Moenius, 2011; Francois and Manchin, 2013). On the role of political accountability, empirical work has found a strong and positive effects of democratisation on trade, highlighting different channels from the tendency of more open and democratic societies to impose less restrictions on free mobility of goods and services (Milner and Mukherjee, 2009; Eichengreen and Leblang, 2008) to the higher trust that is given to a counterpart from a democratic country (Yu, 2010; Levchenko, 2007). We extend this strand of the 3 Noguera (2012) derives a structural gravity equation for value-added exports. In future versions of this work, we plan to extend his framework in order to answer our research questions from a value-added perspective. 4 A theoretically-consistent estimate of the comparative statics effect of trade agreements requires to specify the full general equilibrium model as changes in trade costs generally affect the allocation of resources across sectors. Different assumptions on the underlying structure of the economy can nevertheless lead to a common formulation of the comparative statics effect of a change in trade costs as recently reviewed by Costinot and Rodrguez-Clare (2015). Egger et al. (2011) estimate the full trade effect of FTAs. 6

7 literature by exploring how the quality of institutions and political accountability affect a direct (although partial) measure of protectionism, the border effect. Importantly, we focus on direct imports by the government, for which political economy determinants should be more relevant than for imports by private firms. The rest of the paper proceeds as follows. In section 2, we briefly discuss the choice of the gravity equation as our theoretical framework underlying the empirical analysis. In section 3, we discuss the empirical strategy and describe the data. Section 4 presents preliminary results from the gravity equation and the home bias indexes. Finally, in section 5 we identify the policy implications of our results and directions for future work. 2 Theoretical framework In this section, we present our theoretical framework, justify its choice, and describe how we bring it to the data. In devising our empirical strategy, we aim to define a simple framework that allows us to identify trade barriers in public procurement across countries. The gravity model can serve this purpose. It has been used for decades to infer determinants of bilateral trade and it is consistent with many general equilibrium models of trade (Head and Mayer, 2015). Here we argue that the gravity equation can be used also to explore trade barriers in public procurement. To see this, let us work with what is perhaps the simplest theoretical framework that delivers gravity, the national product differentiation assumption due to Armington (1969) 5. Each country is endowed with a differentiated variety of a type k of goods and services. As Anderson (1979) shows, this assumption coupled with CES preferences or technology delivers a gravity equation. To better capture the procurement of goods and service, we consider shipments of intermediate inputs. The private and public market in each country j source inputs of type k originated from country i. Crucially, varieties are thus differentiated by the type s {p, r} 5 The theoretical framework outlined here, being based on the gravity equation, can be derived from a number of assumptions on the demand and supply sides of the model (Head and Mayer, 2015). Ricardian comparative advantage models á la Eaton and Kortum (2002) and monopolistic competition models with Dixit-Stiglitz-type assumptions deliver a gravity equation. Larch and Lechthaler (2013), for instance, use a monopolistic competition framework to estimate the welfare maximising share of domestic public procurement. 7

8 of buyer, where p stands for public and r denotes private market. One way to think about this assumption is that firms are completely specialised in either the public or private market. Let X k,s ij denote the value of shipments of good or service k from country i to market (public or private) s of country j. Trade is subject to a variable cost factor t k,s ij > 1 of iceberg type. Given factory gate prices of p k,s i, destination prices are p k,s ij p k,s i t k,s ij. Let denote public or private expenditure on good type k in country j and Y k,s i that suppliers in i derive from selling good k to market s. E k,s j the income Governments choose their optimal demand for intermediate input k from country i in order to minimise costs subject to a CES technology, which, for simplicity, is assumed to be equal across public and private markets. The different varieties of intermediate inputs are thus assembled in a composite public good that is transferred to consumers 6. Consumers derive utility from this public good and a private good aggregate sold by the private firms. Invoking the trade separability assumption (Anderson and van Wincoop, 2004), we do not need to impose a specific functional form to preferences, but only require that the allocation of resources to private and public goods can be separated from the allocation of income and expenditures within good type (k, s) across countries 7. Under this assumption, the government s problem can be solved in two steps. First, public authorities chooses the optimal mix of spending across type k and sourcing country i, taking as given aggregate expenditure for each good type k and hence total public expenditure and consequently taxation. In a second step, it chooses the level of aggregate expenditure and thus taxation that maximises household s utility (Larch and Lechthaler, 2013). Separability implies that only the first step determines bilateral trade flows. Crucially, taxation does not affect bilateral trade flows under the conditional general equilibrium (Anderson and van Wincoop, 2004), as long as it does not come from border tariffs, which we assume throughout. While this limits the scope of the theory, it enables us to focus on trade costs. Given this structure, the CES demand function for intermediate goods is: (1) X k,s ij = where P k,s j ( p k,s i t k,s ij P k,s j [ ( i p k,s i ) 1 σ k E k,s j ) ] 1 σ t k,s k 1 σ k ij is the market s price index of good type k, i.e. the unit 6 Private firms provide the private good aggregate under perfect competition 7 Cobb-Douglas preferences across public and private good aggregates satisfy this condition. 8

9 cost that market s faces to buy a bundle k of intermediate input varieties. The term σ k > 1 is the elasticity of substitution between varieties of intermediate good class k and is assumed to be equal across public and private market. Using market clearance on the supply side, Y k,s i = ( ) 1 σ p k,s j i t k,s ij /P k,s k j E k,s j to solve for the exogenous factory prices (see, e.g., Anderson and Yotov, 2010a), we obtain the structural gravity model for each buyer s {p, r}: (2) (3) (4) ( ( P k,s j Π k,s i X k,s ij ) 1 σ k ) 1 σ k ( = Ek,s j Y k,s i t k,s ij Y k,s = i = j ( t k,s ij Π k,s i ( t k,s ij P k,s j P k,s j ) 1 σ k Π k,s i Y k,s i Y k,s ) 1 σ k E k,s j Y k,s ) 1 σ k where Y k,s i Y k,s i denotes world income generated from supplies of good k to buyer s. The Π i s terms are referred to as sellers incidence or inward multilateral resistance, while the price indexes P j s are suitably re-interpreted as buyers incidence or outward multilateral resistance (Anderson and van Wincoop, 2003; Anderson and Yotov, 2010b). These terms summarise the average trade resistance between one country and the rest of the world. The system can be solved for the P j s and Π i s terms (up to a scalar) given data on income and expenditure and estimates of the trade cost vector {t ij }. The structural gravity model provides a theoretically-consistent index of home bias, defined as the amount of predicted internal trade relative, given trade costs, relative to the same internal flow that would arise in a frictionless benchmark. In absence of trade barriers (t ij = 1 i, j), trade flows are proportional to income and expenditures shares: X k,s i,i (t ij = 1) = Y k,s i E k,s i /Y k,s. The Constructed Home Bias (CHB) index (Anderson and Yotov, 2010b) is thus: (5) CHB k,s i ( Π k,s i t k,s ii P k,s j ) 1 σ k This index summarises how trade costs around the world inflates domestic shipments over international trade. It thus provide a specific measure of protectionism that can be computed for both the private and public market. 9

10 As Agnosteva et al. (2014) argue, the CHB index is comparable across type of good and service, countries and over time, does not depend on normalisations nor on estimates of σ k. Importantly, it can be estimated given the structure of the gravity equation. As such, the estimated index is meant to capture central tendency in the data and hence share the good empirical properties of the gravity equation 8. Importantly, we implicitly assume that the elasticity of substitution is equal across buyers, which justifies the comparison of CHB s across private and public market. 3 Empirical strategy and data Given data on the value of bilateral sectoral shipments X k,s ij and proxies for the trade cost function t k,s ij, the parameters of the gravity equation in (2) can be consistently estimated. We follow common practice in the literature and use country-year importer and exporter fixed-effects in our panel regressions to control for multilateral resistance terms and the income and expenditure effect. equation (2) is thus: (6) X k,s ij,t = exp ( m k,s j,t + ek,s i,t + αk,s t Adding a time subscript, the empirical counterpart of ) T k,s ij,t + ε k,s ij,t where T k,s ij,t is the associated vector of coefficients: t ks ij,t is the matrix of possibly time-varying bilateral trade cost variables and α ( ) k,s t exp. To avoid collinearity and consistently with the structural gravity model in (2) (see Anderson and Yotov, 2010b), we normalise exp( m k,s ARG,t ) = 1 P k,s ARG,t = 1 in all our estimations9. α k,s t T k,s ij,t In specifying the trade cost function, we follow two approaches. First, we include time-invariant determinant of trade barriers that have extensively used in the literature together with time-variant and policy-driven variables, including measures that should capture changes in trade barriers specific to the public procurement market. Specifically, 8 Another approach to measure trade cost is to solve the gravity equation in 2 for bilateral trade costs t ij s (Novy, 2013). This measure however does not directly relate to the concept of home bias and, more importantly, is based on actual data. As we expect important measurement error to affect our data, we consider this approach in our application inferior to the CHB one. ( 9 Given our structural interpretation of the model, the normalisation implies that ( ) E k,s j,t /E k,s ARG,t exp. It follows that P ARG,t = 1. m k,s j,t P k,s j,t ) 1 σk = 10

11 the trade cost function is specified as follows: (7) (8) (9) t ks ij,t exp ( 4 m=1 β k,s m DIST m ij + β k,s 5 CONT IG ij + β k,s 6 COLONY ij + β k,s 7 LEGAL ij ) ( ) exp β k,s 8 LANG ij + β k,s 9 SMCT Y ij + β k,s 10 EU ij,t + β k,s 11 GP A ij,t + ( ) exp β k,s 12 P ROC ij,t + β k,s 13 DEP T HF T A ij,t The variables DIST m are the the log of population-weighted bilateral distance (Mayer and Zignago, 2011) within each quartile m of the distance distribution in the sample to identify possible non-linear effects of transportation costs (see, e.g., Eaton and Kortum, 2002). CONT IG is a dummy equal to one if the two countries in the pair share a border, COLONY equals one if the two countries share colonial history, LEGAL is a dummy for common legal origin and LANG equals one if the two countries share an official language. These variables are sourced from CEPII (Mayer and Zignago, 2011) 10. The SMCT Y indicator equals one if i = j, i.e. if the the trade flow is internal. The coefficient β 9 thus identifies the (partial) border effect, i.e. how much trade within national borders is different from trade with other countries. In comparing trade barriers across public and private markets, we look at this coefficient as a first measure of protectionism. The other determinants of trade are policy variables. EU equals one if the two countries in the pair are both EU members at time t, while GP A captures membership in the WTO agreement on government procurement 11. The last two variables are meant to identify the effect of FTAs. P ROC equals one if the two countries are part of an FTA that includes market access provisions concerning public procurement. To identify the effect of these provisions net of the overall effect of forming an FTA, we control for the average number of provisions (trade and non-trade related) included in the agreement with DEP T HF T A. The first objective is to compare estimates of the trade cost function across public and private markets, for each type of sourced good k. We focus on estimates of the coefficient on the SMCT Y dummy as it picks up the border effect and hence can be taken as a first indicator of protectionism. 10 Population-weighted bilateral distances are appropriately recomputed taking distances between the 25 most populous cities in each country, with the Rest-of-the-World aggregate being one country. 11 While the agreement entered officially into force in 1996, it was firstly singed in We thus assume that the countries that entered the agreement in 1996 were already de facto members in 1995, the first year of our panel. 11

12 We then seek to see whether policy efforts to liberalise public procurement markets have contributed to increase public imports. Specifically, we test if the coefficients on the EU, GP A and P ROC dummies are positive and significant and if they are higher for the public market than the private one. To this end, the trade cost specification in (7) is likely to provide inconsistent estimate of these partial effects as it does not control for unobserved time-invariant heterogeneity that can drive both the propensity to increase cooperation through various agreements and trade flows (Baier and Bergstrand, 2007). Furthermore, changes in trade costs can shift international trade relative to domestic shipments affecting the likelihood that two countries engage in trade agreements. To control for unobserved heterogeneity and time-varying changes in cost of international relative intranational trade, we follow Bergstrand et al. (2013) and add the following specification: (10) (11) t ks,f E ij,t ( exp β k,s ( T exp ) 1 EU ij,t + β k,s 2 GP A ij,t + β k,s 3 P ROC ij,t + β k,s 4 DEP T HF T A ij,t ) t=2000 β k,s t SMCT Y ij + γ k,s ij where the γ s terms are undirected bilateral fixed effects that capture unobserved and time-invariant determinants of trade costs 12. The SMCT Y indicator is now interacted with time dummies. The coefficients β t s identify changes in the border effect over time relative to the initial year (1995), whose effect is absorbed in the country-pair dummies. Our interest is in the (partial) effect of trade agreements. We isolate the effect of EU membership as almost half of the countries in our sample are members of the EU and the agreement is arguably the deepest form of economic integration among states, including specific directives on public procurement. Two policy variables should pick up trade liberalisation specific to public procurement markets. One is the dummy for membership in the WTO GPA. Started in 1996 and revised in 2014, the agreement aims to ensure national treatment to foreign firms in public procurement markets, although each member defines the areas of commitments (e.g. different public entities, goods vs. service) that can thus vary substantially across countries. Importantly and unlike most of the WTO agreement, the GPA is plurilateral, meaning that it binds only its signatories having de facto the 12 Collinearity requires further restrictions on the set of bilateral fixed effects γ ij s. As in Agnosteva et al. (2014), we suppress the time-invariant internal trade cost dummies so that the estimates of international time-invariant trade costs are relative to a geometric mean of the two countries internal trade cost: exp(γ k,s ij ) = [ t s,k ij / ( t s,k ii ts,k ij ) 1/2 ]. 12

13 same structure of an FTA. Countries have also explicitly included provisions on public procurement markets in FTAs. We thus draw from the database of Kohl et al. (2013), who coded the legal texts of FTAs to extrapolate the areas that are covered in the agreement, including public procurement. While they use the legal wording of the text to differentiate between covered and legally enforceable provisions, in our sample of countries there are very few cases of FTAs that cover public procurement but not in a legally enforceable way. We thus use a single indicator, P ROC, which equals one if the two countries partners in an FTA with provisions on public procurement. Having an FTA with provisions on public procurement may nonetheless proxy for any general trade-creating effect of FTAs. We thus include an index, DEP T HF T A, which equals the average share of covered and enforced provisions 13. Our next objective is to quantify protectionism in the public procurement market and try to identify its determinants relative to protectionism in private markets. The border effect gives a partial measure of protectionism as it estimates how much countries trade more with themselves than with foreign countries, controlling for other bilateral trade barriers and multilateral resistance effects. We then test if this effect is significantly different across public and private markets. We then ask: which factors can explain differences in protectionism in public procurement (relative to private markets) across countries? To answer this question, we look at institutional aspects that can create a wedge in trade barriers between the public and private markets. First, the lack of transparency and corruption behaviour in the public sector is generally associated with low competition in public procurement markets and, the argument goes, possibly with less openness to foreign suppliers (Arrowsmith and Anderson, 2011). While poor institutions have been shown to hinder trade in general (Anderson and Marcouiller, 2002), they should play a greater role in public procurement markets where suppliers often deal directly with public officials. As an inverse measure of quality of institutions, we use the control of corruption index from the World Governance Indicators (Kaufmann et al., 2010). Given the difficulty in measuring the propensity to engage in corruption behaviour by public officials in our cross-country setting, we rely on the WGI 13 There are 17 provisions in total, 13 under the legal WTO mandate (so-called WTO+ ) and 4 outside WTO legal reach ( WTOx ). We first take the average between the shares of WTO+ and WTOx provisions that are covered and do the same for the shares of each type of provision that is enforceable, as in Kohl et al. (2013). Since we do not distinguish between covered and enforceable provisions on public procurement, DEP T HF T A is then the mean between average coverage and average enforceability. 13

14 corruption index, which, however, is highly correlated with the other WGI index on the quality of institutions. As a slightly different political economy mechanism possibly behind protectionism in public procurement, we investigate the role of political incentives of the national government. The idea here is that public procurement can favour domestic firms in order to (at least apparently) boost the local economy in order to captivate the electorate or domestic special interest groups. The incumbent has more incentives to use public procurement as a pork-barrel politics tool when it is more accountable to voters. Crucially, we don not expect private firms to source more inputs locally as a result of political incentives, unless prompted to do so indirectly through government policies. While democratic institutions are generally synonymous of greater political accountability, we are not able to use this measure in this paper since very few countries are not democratic in our sample. Conditional on having democratic institutions, a measure of accountability is the level of political constraints faced by the executive. For instance, a highly fractionalised legislature makes the incumbent government highly accountable to the different parties in the parliament or to the electorate in order to acquire direct legitimacy. Under these circumstances, public purchases can be directed more to local suppliers to satisfy their political requests. We thus use the probability that two legislators picked at random are from different parties (F RAC) and a broader index of political constraints (P OLCON, which considers, inter alia, political alignment between the executive and the legislature) from Henisz (2000) as two measures of political accountability. As a robustness and inverse measure of accountability, we also employ the share of parliament seats held by the largest government party (GOV 1V OT E) sourced from the Database on Political Institutions (Beck et al., 2001). To test for the importance of these two mechanisms in explaining protectionism in public procurement, we add an interaction term between the SM CT Y dummy and the different proxies for institutional quality and political accountability. If these political economy arguments play a role, we should observe significant variation of the extent to which public procurement buys more nationally across countries according to the quality of institutions and the level of political accountability. In other words, national borders to public procurement markets should be thicker in countries with poor institutions and high political accountability. To estimate these effects, we see if and how this variation differs between the public market and the private one. In all specifications, we also include interaction with the country s GDP and GDP per capita as measures of market 14

15 size and the level of economic development. Both can affect the amount of intranational relative to international trade and are arguably correlated with political and institutional characteristics. To take into account how trade barriers around the world create protectionism, we next estimate the CHB. Differently from the border effect, the CHB measures how trade costs shifts up internal tirade relative to a frictionless benchmark. To estimate it, we manipulate the gravity equation in (2) as follows (see Agnosteva et al., 2014): (12) CHB k,s i,t = Y k,s t X k,s ij,t E k,s j,t Y k,s i,t = tk,s ii,t P k,s j,t Πk,s i,t 1 σ k The estimated CHB is thus given by the predicted values of the gravity model rescaled by sectoral expenditures and incomes. The trade cost specification with fixed effects in (10) is chosen as it controls for all time-invariant bilateral determinants of trade 14. We first obtain CHB for each class of good k (and market s) and then consistently aggregate those tot he country level suing expenditure shares as weights. This approach gives consistent estimates of the CHB index if the gravity equation is correctly specified, i.e. if the country-specific fixed effects are consistent estimates of their theoretical counterparts. Fally (2013) shows that this automatically holds true when the Poisson pseudo-maximum-likelihood (PPML) estimator advocated by Silva and Tenreyro (2006) is employed and income and expenditure are consistent with bilateral trade flows (i.e., Y k,s i,t = j Xk,s ij,t ; Ek,s j,t = i Xk,s ij,t ). The peculiar properties of the estimator implies that the actual income and expenditure values equal the predicted ones, which should normally be used as both are endogenous in the full general equilibrium model. We thus employ the PPML estimator, which has the added advantages of controlling for heteroskedasticity in the data and statistically dealing with zero trade flows. The gravity equation in (6) is estimated separately for each market s and sector k, although covariances of the estimated coefficients are taken into account when testing significance of the difference in coefficients across public and private markets. 14 Note that under that specification t k,s ii,t = tk,s jj,t i, j, i.e. the border effect is equal across countries. Differences in estimated CHB s across countries come thus from differences in the multilateral resistance terms. 15

16 To bring this empirical artillery to the data, we need information on bilateral trade that involves the public sector as a buyer. Other studies that investigate trade barriers in public procurement employ data from inter-country input-output tables (Riker, 2013; Messerlin and Mirodout, 2012) as this can split public expenditures from national accounts across type of goods and services purchases and country of origin. We thus follow this route and employ data from the TiVa initiative of the OECD. Similarly to other ICIO database (e.g., Timmer, 2012), the TiVa database harmonises national IO tables and combines them with information from national accounts and bilateral trade statistics in goods and services to obtain an international input-output table (see OECD-WTO, 2012 for details). The estimation procedure allocates output from each country and sector to intermediate usage (by all sectors) or final demand across countries. While far from perfect and inevitably rife with measurement errors (especially compared with official trade statistics), this type of data is the only one that enables international comparison of public expenditures across countries and sectors. The data covers 57 countries (OECD members plus other major developing countries) plus a Rest-of-the-World aggregate for four years (1995, 2000, 2005, 2009) and 37 manufacturing and service sectors 15. To determine the perimeter of the public sector, we need to define public procurement. The OECD defines public procurement as intermediate consumption (goods and services purchased by governments for their own use, such as accounting or IT services), gross fixed capital formation (acquisition of capital excluding sales of fixed assets, such as building new roads) and social transfers in kind via market producers (goods and services produced by market producers, purchased by government and supplied to households) (OECD, 2013a, p.130). Intermediate consumption should be recorded in the Public Administration column vector of the Input-Output matrix. The TiVa database does not provide a split between public and private gross fixed capital formation, while social transfers in kind should in principle be recorded in the General Government Expenditure component of final demand. The problem with this latter variable is that for many countries it usually includes compensation to public employees and hence would automatically inflate our measures of protectionism in public procurement. To proxy for sector-level compensation of public employees, we distribute the country-level compensation (sourced from the World 15 Thailand never reports imports in public procurement (no imports are reported also in the Government Expenditures column). We thus add it to the RoW aggregate. Our final sample hence includes 56 countries. 16

17 Development Indicators) to each sector proportionally to its share of domestic government expenditures for final demand. We thus subtract this constructed series of public wages from the original domestic government purchases. Albeit imperfect, this adjustment should isolate the social transfer component in the Government Expenditures column of the ICIO tables. Our final measure of government purchases (and hence imports) is thus the sum of the adjusted Government Expenditures and Public Administration column vectors of the ICIO table 16. Once public procurement is defined, we identify a private market that is suitable for comparisons. The other columns in the ICIO table are the most immediate and comparable definition of private procurement. We thus take the sum across intermediate inputs columns of the ICIO table as our measure of trade involving private firms as buyers. This choice can nevertheless lead to overlap with public procurement to the extent to which public authorities operate outside the Public Administration sector as it is usually the case in some service sectors such as Health and Education. Since we do not have information about the share of public procurement done outside the Public Administration sector (which can arguably vary substantially across countries), we stick to our definitions bearing in mind that any overlap between the public and private markets should work against finding significant differences in trade barriers between the two. Before turning to the empirical estimates, we investigate descriptive trends in the data. The objective here is twofold: (1) To identify patterns of expenditures across sectors in public and private markets as these affect estimated home bias at the country-level; (2) To have a first look at trade barriers by looking at import penetration ratio. To ease exposition, we aggregate the original 37 sectors to 18 sectors according to an OECD concordance. First, we compute the sector expenditure share for each country in the public and private markets as defined above. Figure 1 reports the average of these shares across countries in 1995 and 2009 (the first and last year of our sample). One pattern stands out: public procurement is largely about services. Importantly, the share of pubic purchases on various services is higher than the share of expenditure by private firms, with the the difference being particularly high in the Construction, Health and Education and Real Estate sectors. To uncover heterogeneity across countries, we sum public and private purchases over 16 We believe this definition is also more relevant to current negotiations on market access to public procurement and it is endorsed by some statistical offices (see Dey-Chowdhury and Tily, 2007 for the UK). 17

18 manufacturing and service sectors 17 and compare the purchases share across countries. In Figure 2 we plot the services share of public and private purchases, manufacturing and primary sector shares being the excluded category. The share of public procurement directed at services is higher than that of manufacturing (and primary products) for almost all countries in the sample, the average being around 65% in both 1995 and As the cross-country averages in Figure 1 suggested, public markets are buy relative more services than private ones, with very few countries (e.g. Romania in 1995 and Brunei in 2009) being the exception. We then turn to import penetration ratios defined as the value of imports divided by total expenditures. While partial and purely descriptive, the measure has been used extensively to assess openness to trade, including in public procurement markets (Messerlin and Mirodout, 2012). We first take the average import penetration ratios across countries for each supplying sector. As Figure 3 shows, openness is sensibly (and not surprisingly) lower in service sectors than in manufacturing ones. Importantly, the average openness is lower in public markets then private ones for most of supplying sectors, although the difference is often not large. This trend can however mask substantial heterogeneity across countries. As before, we thus compute the import penetration ratios for public and private markets in the aggregate service and manufacturing sectors. Figure 4 reports the ratio of public to private import penetration ratios for the service and manufacturing sectors. A value greater than one suggests that public markets are more open the private ones. At the beginning of the sample, public markets are less open than private ones in services for most countries, while the picture is more nuanced in manufacturing. Relative openness of public procurement markets to services increases for most countries over time, while it slightly decreases on average for manufacturing goods. While purely illustrative, this descriptive analysis delivers some relevant messages to the subsequent more formal empirical analysis. Public procurement is really about services, which are generally less traded than manufacturing goods. These two observations alone 17 Services are: Electricity, gas and water supply ( electricity ), Construction ( construct ), Wholesale and retail trade; Hotels and restaurants ( retail ), Transport and storage, post and telecommunications ( transport ), Financial intermediation ( finance ), Real estate, renting and business activities ( realestate ), Community, social and personal services ( pahealthedu ). Manufacturing sectors are all other sectors except primary sectors Agriculture, hunting forestry and fishing ( agri ) and Mining and quarrying ( mineral ). 18

19 naturally increase home bias in the public sector at the country level as services are weighted more in the public expenditure basket than in the private one. Crucially, the gravity framework outlined in section 2 cannot capture this composition effect because expenditure and income allocations within countries across sectors are taken as given. Still, the analysis of import penetration ratios suggests that already within services, public markets are less open than private ones, although substantial heterogeneity across countries and over time emerges. The ensuing empirical analysis aims to exploit this variation. 4 Empirical results In this section, we discuss the estimates from the gravity equation (6), its extensions with border effect interaction and the estimated CHB. The objective is to estimate trade barriers in public procurement markets (relative to private markets) by applying the empirical framework described in section 3. To make the analysis clearer and in line with the descriptive evidence, we sum up bilateral trade values over supplying sectors in a Manufacturing (including also primary sectors) and a Services aggregate. For each specification and supplying sector, we report the estimates for the private market next to the ones for the public one to ease comparison. Coefficients in bold are significantly different across the two markets. 4.1 The partial effect of trade agreements Table 1 reports the gravity estimates. Columns (1) to (4) report the estimates for the manufacturing sector. The first two columns show the pooled gravity estimates of the trade cost function in (7). Coefficients of the time-invariant determinants of trade costs have the expected sign and most of them are statistically significant. Distance has the usual depressing role on bilateral trade, regardless of whether the purchaser is a private or public entity. The depressing effect is rather on the lower end of the range of distance effects found in the literature (Disdier and Head, 2008) and, interestingly, it is significantly lower (in absolute terms) for public purchases. No clear patterns emerges across the distance distributions as the coefficients for the different quartiles are not significantly different from each other. Concerning the other time-invariant variables, sharing a 19

20 border or an official language increase bilateral trade in manufacturing irrespective of the type of purchaser. The coefficient on the SM CT Y dummy gives a first indication of protectionist trade barriers in manufacturing, which are significantly higher for public procurement markets. All other trade cost variables (expect for distance) are switched off for same-country pairs, so that the excluded category includes country-pairs that have, for instance, no contiguous border, common language and are no EU members. The estimates imply that, relative to this category, public procurement purchases from local suppliers are something around 35 times higher, the increase being significantly lower in private markets. Time-varying and policy-driven determinants of trade are included here merely as controls. Omitted variable bias is likely to plague their partial effects on trade. In Columns (3) and (4), we include (undirected) country-pair fixed effects and interaction terms between the SM CT Y dummy and time dummies in order to identify arguably unbiased effects of trade policy. The bilateral fixed-effects control for time-invariant unobserved heterogeneity that could drive both trade and the propensity of signing trade agreements. The time-varying border effects control for shocks that can affect intranational relative to international trade costs and hence influence the likelihood of trade agreements. Entry in the EU increases significantly cross-border public procurement of manufacturing goods. Public procurement trade (i.e. manufacturing imports by the government and manufacturing exports to the partner s government) doubles after that the two countries are part of the EU. This effect is significantly higher than the increase observed in manufacturing trade involving private markets, suggesting that EU cross-border public procurement, while still far from a true single market, has opened up the public sector compared to what would have happened otherwise. GPA membership does not seem to increase public procurement trade, probably also because most of the effect might be captured by EU countries, which represent more than half of the members. Furthermore, the dummy variable might not be able to identify the quite substantial heterogeneity in market access commitments among signatories. FTAs seem to be more instrumental in opening up public procurement markets. When two countries sign an FTAs including provisions on public procurement, bilateral manufacturing imports by public authorities goes up by 25%, controlling for the formation of an FTA per se and the inclusion of other provisions. If anything, the formation of an FTA which includes provisions on areas other than public procurement decreases openness in public markets. Importantly, this trade-creating effect of public procurement provisions 20