Services Trade Restrictiveness and Manufacturing Productivity: The Role of Institutions

Transcription

1 Services Trade Restrictiveness and Manufacturing Productivity: The Role of Institutions Cosimo Beverelli Matteo Fiorini Bernard Hoekman WTO EUI EUI and CEPR ITSG 10th December / 31

2 Motivation Producer services have a crucial input function for manufacturing production Descriptive evidence Trade is an important channel for improving access to intermediate inputs Services trade openness is relevant for productivity in downstream industries (Arnold et al., 2011; forthcoming) Proximity burden in services trade: exporters often must perform some stages of their economic activity in the importing country (Modes 3 & 4) Exporter payoffs as well as the quality of their output is a function of the institutional environment of the country where demand is located and the service is performed Potential role of formal economic institutions (rule of law, control of corruption, regulatory quality etc) in shaping the indirect effect of services trade openness Raw evidence 2 / 31

3 Motivation Producer services have a crucial input function for manufacturing production Descriptive evidence Trade is an important channel for improving access to intermediate inputs Services trade openness is relevant for productivity in downstream industries (Arnold et al., 2011; forthcoming) Proximity burden in services trade: exporters often must perform some stages of their economic activity in the importing country (Modes 3 & 4) Exporter payoffs as well as the quality of their output is a function of the institutional environment of the country where demand is located and the service is performed Potential role of formal economic institutions (rule of law, control of corruption, regulatory quality etc) in shaping the indirect effect of services trade openness Raw evidence 2 / 31

4 Research objectives 1 Revisit the empirical assessment of the indirect effect of services trade policy on downstream productivity within a cross-country framework 2 Test the role of formal institutions in shaping the effect 3 Propose a framework to rationalize the role of institutions 3 / 31

5 Preview of results Reducing services trade restrictiveness has a positive impact on productivity in downstream manufacturing sectors This effect is crucially moderated by country-level institutional capacity Stronger effect in country with good institutions (high control of corruption / rule of law / regulatory quality) Quantification: an example - Canada and Tanzania have the same IO-weighted average services trade restrictiveness (CST RI) - If both move to the less restricted level of the UK ( 1/2 of a sd decrease), the resulting average increase in manufacturing productivity is equal to 16.7% in Canada and only 3.9% in Tanzania The empirical result can be rationalized with a theoretical framework where the role of institutions is linked to services specificities (intangibility) 4 / 31

6 Literature I Indirect effect of outcomes and policies in the upstream (services) industry on downstream (manufacturing) sectors Firm-level studies - Arnold et al. (2008): 10 SSA Countries - Fernandes and Paunov (2011): Chile - Arnold et al. (2011): Czech Republic - Forlani (2012): France - Duggan et al. (2013): Indonesia - Arnold et al. (forthcoming): India - Hoekman and Shepherd (forthcoming): 119 developing countries Industry-level studies - Barone and Cingano (2011): OECD countries - Bourlès et al. (2013): OECD countries - Hoekman and Shepherd (forthcoming): wide country coverage 5 / 31

7 Literature II Conditionality in the benefits from trade openness - Rodriguez and Rodrik (2001) - Winters and Masters (2013) The role of institutions For economic development - Acemoglu et al. (2001; 2005) - Rodrik et al. (2004) For trade - As triggers of comparative advantage: Nunn (2007); Levchenko (2007); Costinot (2009) - As determinants of bilateral trade flows: Anderson and Marcouiller (2002) - As determinants of offshore and FDI decisions at the firm level: Bernard et al. (2010); Antràs and Helpman (2004); Grossman and Helpman (2005) - As determinants of the services trade policy stance (van der Marel, 2014) and of the coverage of services policy commitments made in trade agreements (van der Marel and Miroudot, 2014) 6 / 31

8 Methodology Rajan and Zingales (1998) approach Underlying assumption: the effect of upstream trade restrictiveness on downstream productivity varies with the degree of input penetration Composite Service Trade Restrictiveness Index (CSTRI) in country (i) and downstream manufacturing sector (j): CST RI ij s ST RI is w ijs - ST RI is: Index of Service s Trade Restrictiveness in country i - w ijs: a measure of input penetration of service s into downstream sector j in country i 7 / 31

9 Empirical specifications Baseline specification y ij,t = α + βcst RI ij,t 1 + γ x ij,t 1 + δ i + δ j + ɛ ij - y: labor productivity - x: column vector of the other relevant regressors Interaction model with institutional capacity y ij,t = α + βcst RI ij,t 1 + κ(cst RI ij,t 1 IC i,t 1 )+ + γ x ij,t 1 + δ i + δ j + ɛ ij - IC i: a measure of institutional capacity in country i - The main effect of IC i is accounted for by δ i 8 / 31

10 Empirical specifications Baseline specification y ij,t = α + βcst RI ij,t 1 + γ x ij,t 1 + δ i + δ j + ɛ ij - y: labor productivity - x: column vector of the other relevant regressors Interaction model with institutional capacity y ij,t = α + βcst RI ij,t 1 + κ(cst RI ij,t 1 IC i,t 1 )+ + γ x ij,t 1 + δ i + δ j + ɛ ij - IC i: a measure of institutional capacity in country i - The main effect of IC i is accounted for by δ i 8 / 31

11 Identification strategy I Unobserved heterogeneity and omitted variables bias Solution: - We include country fixed effects and sector dummies to account for countryand industry-specific characteristics - We control for (country-industry)-specific downstream trade restrictiveness including (log) tariffs, ln(1 + τ ij), in the vector x 9 / 31

12 Identification strategy II Endogeneity of the input penetration weights w ijs : With respect to the dependent variable: reverse causality bias (higher productivity as a determinant of higher producer services consumption) With respect to services trade restrictiveness: the treatment (STRI) might affect w ijs s and therefore the allocation of observations (ij s) across treatment and control groups Solution to both issues: We apply US IO coefficients to all countries removing the US from the sample Underlying assumption: US coefficients as representative of the technological relationships between industries rather than reflecting US-specific determinants (Rajan and Zingales, 1998; Barone and Cingano, 2011; Bourlès et al., 2013) 10 / 31

13 Identification strategy III Endogeneity of services trade restrictiveness STRI: Reverse causality bias due to lobbying: 2 possible cases 1 Low productivity industries can lobby for less restrictiveness in upstream trade positive sign of the bias the results would be a lower bound for the impact of reducing services trade restrictiveness on manufacturing productivity 2 High productivity industries can lobby for less restrictiveness in upstream trade undetermined sign of the bias problematic IV solution: Instrument for ST RI i using the weighted average of ST RI k, k i with Weights given by the similarity index in GDP per capita between i and k { } GDP 2 { } i GDP 2 k SI ik 1 GDP i + GDP k GDP i + GDP k k belonging to a different geographical region than i 11 / 31

14 Data Productivity in manufacturing - Output per worker in Source: Unido Industrial Statistics 2015 Release (ISIC Rev. 3, 2 digits) Services trade restrictiveness - STRI for 4 sectors (Finance + Insurance / Telecommunications / Transports / Professional) - Retail excluded: not relevant as an input, dirty concordance see Barone and Cingano (2011) - Source: World Bank STRI data (proxy for year 2006) Input penetration weights - Technical coefficients from the US input-output table - Source: OECD STAN IO Data 2005 Institutional capacity - Control of Corruption in Source: Worldwide Governance Indicators Tariffs - Applied tariff, simple average Source: UNCTAD TRAINS 12 / 31

15 Summary statistics Estimation dataset: 57 countries and up to 18 manufacturing sectors Details 912 observations Summary statistics Variable mean median sd min max y ij CST RI ij IC i ln(1 + τ ij ) / 31

16 Main empirical results Dependent variable: y ij (log of labor productivity) All modes Mode 3 (1) (2) (3) (4) CST RI * (0.024) (0.038) (0.021) (0.032) CST RI IC *** *** (0.014) (0.012) Observations R-squared Robust standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 ln(1 + τ ij ), country fixed effects and sector dummies always included IC = control of corruption (WGI), rescaled from 1 (least) to 5 (most controlled) 14 / 31

17 Marginal effects (Mode 3 case) Effect on log productivity Control of Corruption in year 2007 Estimated Impact 95% CI Mean Moderator The effect of CST RI has the expected negative sign (for 95% of our sample) Reducing services trade restrictiveness has a positive effect on productivity The magnitude of the effect increases with institutional capacity The effect is statistically significant at the 0.05 level for 60% of sample For weak institutional capacity, the effect is not statistically different from 0 15 / 31

18 IV estimation All modes Mode 3 (1) (2) (3) (4) CST RI * (0.072) (0.061) (0.052) (0.058) CST RI IC *** *** (0.019) (0.017) Observations R-squared First-stage F statistics CST RI (p-value) CSTRI IC (p-value) Underid SW Chi-sq statistics CST RI (p-value) CSTRI IC (p-value) Weak id SW F statistics CST RI CSTRI IC Stock-Wright LM S statistics Chi-sq (p-value) Robust standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 ln(1 + τ ij), country fixed effects and sector dummies always included SW refers to Sanderson and Windmeijer (forthcoming) 16 / 31

19 Quantification Austria, Canada, Italy and Tanzania - Similar CST RI - Different institutional capacity Thought experiment: apply the less restricted services trade regime of the UK Country decrease in CST RI IC rank effect Austria 1/2 sd 6 th +18.2% Canada 1/2 sd 7 th +16.7% Italy 1/2 sd 25 th +7.3% Tanzania 1/2 sd 43 th +3.9% 17 / 31

20 Random STRIs Define CST RI ij = s ŜT RI is w ijs, where ŜT RI is is a random draw from a uniform distribution with support [0, 100] Perform 100,000 interaction model regressions, each with different, randomly constructed ŜT RI is and estimate marginal effects at 39 values of the Control of Corruption variable For each of those 39 values, take the average of the 100,000 estimated marginal effects and of the corresponding confidence intervals Effect on log productivity Control of Corruption in year 2007 Estimated Impact 95% CI Mean Moderator 18 / 31

21 Alternative moderator variables Moderator (M) Rule of Law Reg. Quality GDP per capita All Modes Mode 3 All Modes Mode 3 All Modes Mode 3 CST RI * * (0.024) (0.021) (0.025) (0.021) (0.024) (0.020) CST RI M *** *** *** *** (0.014) (0.012) (0.014) (0.012) (0.000) (0.000) Observations R-squared Robust standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 ln(1 + τ ij), country fixed effects and sector dummies always included Rule of law and regulatory quality are from the Worldwide Governance Database (2007) GDP per capita (2007) is from WDI 19 / 31

22 Additional tariff controls Import protection for other manufacturing sectors k j should also matter for sector j productivity We include the variable Tariff k τ ik w ijk - τ ik : log of 1 + effectively applied tariffs by country i in manufacturing sector k j - w ijk : input penetration coefficients of k in j from the US IO table All results are qualitatively and quantitatively unchanged Like the coefficient on own tariffs, the coefficient on Tariff has always the expected negative sign (higher tariffs in upstream manufacturing sectors reduce productivity in downstream manufacturing) 20 / 31

23 Alternative input penetration measures I Replace share of intermediate consumption (from US IO table) with - US technical coefficients: ratio between cost of services input and value of total output of the downstream sector - US Leontief coefficients: accounting for indirect linkages IO weights Technical Leontief All modes Mode 3 All modes Mode 3 (1) (2) (3) (4) (5) (6) (7) (8) CST RI ** (0.052) (0.081) (0.043) (0.075) (0.082) (0.133) (0.062) (0.144) CST RI IC *** *** *** ** (0.027) (0.026) (0.042) (0.049) Observations R-squared Robust standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 ln(1 + τ ij), country fixed effects and sector dummies always included 21 / 31

24 Alternative input penetration measures II Replace share of intermediate consumption derived from US IO table with share of intermediate consumption derived from Chinese IO table All modes Mode 3 (1) (2) (3) (4) CST RI ** (0.050) (0.090) (0.043) (0.083) CST RI IC *** ** (0.032) (0.030) Observations R-squared Robust standard errors in parentheses * p<0.10, ** p<0.05, *** p<0.01 ln(1 + τ ij ), country fixed effects and sector dummies always included China excluded from the estimation sample 22 / 31

25 Alternative input penetration measures III Non-country specific IO weights Country-specific measures of input intensity may carry an idiosyncratic component related to the trade restrictiveness regime Following Ciccone and Papaioannou (2006), we: 1 Regressed country-industry specific IO weights on country dummies, industry dummies and industry dummies interacted with country-level STRI, for each services sector s 2 Used fitted values, evaluated at country c with least regulation (US, or Sweden in a robustness check), as instrument for w js Although the statistical significance of the estimated coefficients is reduced (especially in the case where c is set equal to Sweden), their signs and magnitudes are in line with the baseline results 23 / 31

26 Alternative input penetration measures IV Random IO weights Instead of randomizing the treatment (STRI), we randomized the IO weights (in the spirit of Keller, 1998) Marginal effects, averaged across 100,000 regressions, not different from zero IO linkages matter 24 / 31

27 The set up Two countries i {1, 2} Identical economic structure: industry Y producing with input x y i = f(q i ) i (1) - q i: quality of the input x as available in country i and f positive and increasing - Without international trade in x, q 1 = q 2 = 0 Institutional setting of country i: capacity to minimize the exposure of the economic agents active within its territory to harmful unexpected changes in the operating environment International supply of x - Continuum of heterogeneous exporters outside the {1, 2}-country system and indexed by productivity parameter ϕ in [0, 1] - Exporters have to choose where to export x among the potential destination countries 25 / 31

28 Intangibility and institutions Allow for x to be intangible proximity burden: exporters must undertake activities in the territory of the selected destination country - The duration of this performance stage is given by the degree of intangibility τ [0, 1] During the stage of services performance in the importing country i, the foreign firm (exporter) confronts unexpected shocks in the operating environment Shocks follow a homogeneous Poisson process with rate parameter θ i - θ captures the institutional setting in country i, with high values of θ being associated with fragile institutions Expected payoff of exporting x with intangibility τ to country i given by: - g positive, increasing and g(0) > 1 Similarly, expected quality of x exported to i is: - k positive, increasing and k(0) > 1 E[π i (ϕ)] = g(ϕ) θ i τ (2) E[q i (ϕ)] = k(ϕ) θ i τ (3) 26 / 31

29 Imperfect observability Institutional capacity of potential destination countries is not perfectly observable to exporters - Exporters observe a signal ϑ i instead of θ i Productivity ϕ determines the precision with which an exporter can estimate the true value of θ Signals are independently distributed according to non-standard uniform probability density functions: ϑ i U[q 1 (θ i, ϕ), q 2 (θ i, ϕ)] i (4) - q 1 θ iϕ and q 2 (θ i 1)ϕ / 31

30 Assume δ θ 2 θ 1 > 0 Closed and open regimes I An exporter with productivity ϕ chooses country 1 if and only if g(ϕ) ϑ 1 τ > g(ϕ) ϑ 2 τ ϑ 1 < ϑ 2 (5) Removing restrictions for trade in x Result 1 If x possesses some degree of intangibility (τ > 0), then y 1 > y 2 > f(0) 28 / 31

31 Closed and open regimes II Mechanism: two effects of institutions Selection effect (ex-ante) Better exporters get more precise signals export to the best country with higher probability Performance effect (ex-post) Given export decisions exporters perform a higher quality service in the country with better institutions 29 / 31

32 Conclusions Reduction in trade restrictiveness upstream has a positive impact on downstream manufacturing productivity This effect is bigger in countries with better formal economic institutions The role of institutions can be linked to services trade specificities (proximity burden) and rationalized through a selection and a performance effect 30 / 31

33 Appendix Services shares of intermediate consumption ISIC Rev. 3, 2 digit manufacturing industries Transport Telecom Finance Business Source: 2005 US IO Table from OECD STAN IO Database Back to Motivation 1 / 3

34 Appendix Log of labor productivity (Y/L) CSTRI High control of corruption (CC) Linear fit if High CC (slope coeff. = with s.e. = 0.020) Low control of corruption (CC) Linear fit if Low CC (slope coeff. = with s.e. = 0.021) Back to Motivation 2 / 3

35 Appendix Country Sector Albania Kyrgyz Rep Austria Lebanese Rep Belgium Lithuania 20 Botswana Malawi Brazil Malaysia 23 Bulgaria Mauritius 24 Burundi Mongolia 25 Canada Morocco 26 Chile Netherlands 27 China New Zealand 28 Colombia Oman 29 Czech Republic Peru 30 Denmark Poland 31 Ecuador Portugal 32 Ethiopia Qatar 33 Finland Romania 34 France Saudi Arabia 35 Georgia South Africa Germany Spain Greece Sri Lanka Hungary Sweden India Tanzania Indonesia Turkey Ireland Ukraine Italy United Kingdom Japan Uruguay Jordan Viet Nam Korea, Rep. Yemen Kuwait Sectors are ISIC Rev. 3 manufacturing industries Back to Summary statistics 3 / 3