The Role of Institutions and Networks in Firms' Offshoring Decisions

Moriconi, Simone; Peri, Giovanni; Pozzoli, Dario Working Paper The Role of Institutions and Networks in Firms' Offshoring Decisions Working paper, No. 4-2018 Provided in Cooperation with: Department of Economics, Copenhagen Business School (CBS) Suggested Citation: Moriconi, Simone; Peri, Giovanni; Pozzoli, Dario (2018) : The Role of Institutions and Networks in Firms' Offshoring Decisions, Working paper, No. 4-2018, Copenhagen Business School (CBS), Department of Economics, Frederiksberg, https://hdl.handle.net/10398/9658 This Version is available at: https://hdl.handle.net/10419/208584 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/3.0/ Department of Economics Copenhagen Business School Working paper 4-2018 Department of Economics – Porcelænshaven 16A, 1. DK-2000 Frederiksberg The Role of Institutions and Networks in Firms' Offshoring Decisions Simone Moriconi Giovanni Peri Dario Pozzoli The Role of Institutions and Networks in Firms’ Offshoring Decisions∗ Simone Moriconi† , Giovanni Peri‡& Dario Pozzoli§ August 6, 2018 Abstract The offshoring of production by multinational firms has expanded dramatically in recent decades, increasing the potential for economic growth and technological transfers. What determines the location of such offshore production? How do the policies and characteristics of countries affect these decisions? Do firms choose specific countries because of their policies or because they are more familiar with them? In this paper, we use a very rich dataset on Danish firms to analyze how their decisions regarding offshore production depend on institutional characteristics and firm-specific bilateral connections with these countries. We find that institutions that enhance investor protections and reduce corruption increase the probability of offshoring, while those that introduce regulatory constraints in the labor market discourage it. We also show that offshoring activities are more likely for firms that have developed networks in the country of destination. Key words: Offshoring, product market, labor regulations, network, fixed costs. JEL code: F16, J38, J24 ∗Funding from the Danish Research Council (grant number: 7024-00007A) is gratefully acknowledged. We would also like to thank participants at the RES 2018 Annual Conference, the CESifo Area Conference on Global Economy and the 2018 Mid-west International Trade Conference for helpful advice and feedback. We also thank the Tuborg Research Centre for Globalisation and Firms at the Aarhus University School of Business and Social Sciences for granting access to Danish registry data. The registry data build on anonymized micro data sets owned by Statistics Denmark. To enhance the scientific validation of analyses published using DS micro data, the Aarhus University Department of Economics and Business helps researchers obtain access to the data set. The usual disclaimers apply. †Email: [email protected]. I´eseg School of Management and LEM. ‡Email: [email protected]. University of California Davis. §Email: [email protected]. Copenhagen Business School. 1 1 Introduction The rapid increase in offshoring has been one of the most notable trends in the labor markets of several developed countries over the last three decades. The driving forces of firm offshoring have been extensively studied (Bernard and Jensen, 1999). Still, the factors that contribute to determining domestic firm boundaries are highly debated (Barba Navaretti et al., 2011). Firms’ decisions regarding offshore production may be described by heterogeneous firm models of trade such as the one proposed by Melitz (2003). In these theoretical models, the probability of engaging in offshore production is a negative function of the entry cost that the firm has to pay to a destination country to engage in offshore production. High fixed cost of operating a business (e.g. due to lengthy bureacuratic procedures or inefficient financial institutions in the destination country) may discourage firms from offshoring production activities in some countries and direct them to other countries with lower costs. Many empirical studies based on firm-level data show that firms engage in trade activities only if their productivity levels are high enough to cover the fixed costs (see Sofronis K. Clerides, 1998; Bernard and Jensen, 1999; and Greenaway and Kneller, 2004, among many others). None of these studies has identified the factors determining firms’ fixed costs for offshoring production or measured the impact of these costs on firms’ propensity to offshore. Partial exceptions can be identified in the literature focusing on foreign direct investment (FDI). Olney (2013) uses a cross-country database compiled by the OECD to show that US firms have larger FDI presence in countries with more liberal employment protection legislation. On the other hand, Antras et al. (2009) study how the cross-border activities of global firms relate to institutional settings with varying investor protections and levels of capital market development. Their model predicts that arm’s length technology transfers are more common than the deployment of this type of technology through foreign affiliate activity in host countries where investor protections are stronger. Moreover, Antras et al. (2009) show that the share of activity abroad financed by capital flows from multinational parents is decreasing in the quality of investor protections in host economies. Our study is closely related to these studies, and more generally analyzes the role of an ample set of institutions and policies in the destination countries in affecting the extensive and the intensive margin of offshoring from Danish firms. Besides analyzing labor market and investor protection we also consider indices of corruption, measures of excessive bureaucracy and the weak enforcement of standards during the registration process of new companies as those may increase the fixed and operating costs of firms (Djankov et al., 2002). Similarly, lax enforcement, such as difficulties in obtaining or recovering credit, or problems in enforcing contracts may hurt economic activity and reduce returns to investments (Acemoglu 2 et al., 2005). In our analysis we then focus on the role of individuals in the offshoring firm who are familiar with the destination country, namely immigrants from those countries. We estimate if their presence and the size of the group affects the probability and the scale of offshoring in their countries of origin. The presence of a network of foreign workers from specific countries can help firms gather information and navigate the local bureaucracy and culture to comply with burdensome regulations or to obtain better access to credit. More generally, networks can help firms reduce information asymmetries and set-up costs that arise from venturing abroad (Peri and Requena-Silvente, 2010). Therefore, this paper has three main goals. First, we test, among a large set of institutional settings, which ones affect the fixed costs of operating a business abroad and consequently influence firms’ offshoring decisions. Second, we distinguish institutions and policies that reduce propensity to offshore from those that create an ‘offshoring-friendly’ business environment and increase offshoring. Third, we analyze whether firms’ networks in destination countries reduce the fixed costs facilitating bilateral offshoring activites and whether the network of immigrants is more or less effective in promoting offshoring in the presence of some of the institutional features analyzed above. To answer these research questions, we use three main datasets. The first one is the Doing Business database compiled by the World Bank (see Djankov et al., 2002), which is a crosscountry database with information on regulations in over 160 countries for the period 20062012. This dataset covers business regulations in the following areas: starting a business, registering property rights, obtaining credit, protecting minority investors, paying taxes, trading across borders, enforcing contracts and resolving insolvency. This database also includes information on labor market regulations regarding hiring practices, hours worked, redundancy rules and minimum wage provisions. We develop a country-specific measure for corruption from the Worldwide Governance Indicators (WGI) project for the period 20062012 (Kraay, 2010). The third data source is a Danish employer-employee matched dataset that covers the universe of individuals and firms in the manufacturing sector for the same period. This dataset is especially well suited for studying firms’ offshoring decisions, since it allows international trade transactions to be measured at the firm level (Hummels et al., 2014a) rather than at the country level (Olney, 2013). This dataset also includes extensive information on firms’ workforce characteristics, which provides an excellent opportunity for identifying the networks of the foreign workers in each firm. Our main results suggest that business and labor market regulations have opposite effects on offshoring activities. Business regulations are beneficial for the extensive margin of offshoring. A lack of credit coverage and the inadequate protection of creditors’ property 3 rights result in high fixed costs for offshoring. Accordingly, business regulations that enhance credit coverage and resolve insolvency issues in destination countries tend to increase domestic firms’ propensity to offshore. This is an interesting finding that substantiates a widely spread negative view about business regulations (see e.g. Djankov et al., 2002) and suggests that such regulations may be needed to ensure business quality. Conversely, labor market regulations have a negative impact on the propensity to offshore: stringent measures of employment protection increase firms’ labor costs, which affect negatively firms’ extensive margin of offshoring. This result is in line with the general view that employment protection regulations increase labor market frictions and/or labor costs. Furthermore, we find that a lack of control of corruption in the destination country highly reduces firms’ probability of offshoring to that destination. The impact of all these institutions on firms’ intensive margins, conditional on offshoring, is never statistically significant. This is consistent with the hypothesis that regulations and corruption affect the fixed, rather than the variable, costs of accessing a foreign market. We also find that a firm’s network, measured by the share of workers from a foreign country, has a significant positive effect on the extensive margin of offshoring in that country, consistent with its role in reducing fixed costs of offshoring.1 Several refinements of the network variable confirm the role played by foreign workers in promoting offshoring activities at the bilateral level. We also find that the positive impact of networks is magnified (attenuated) in destination markets with high levels of credit risk (corruption). In this respect it looks like a thick network of immigrants can in part compensate for high credit risk, possibly ensuring a level of trust and knowledg that reduces credit risk. On the other hand immigrants may warn their firms of high level of corruption in their own countries so that the negative effect of corruption on offshoring is enhanced in firms with a larger network from the country. In the next section we present a conceptual framework for the fixed costs of offshoring. Our empirical strategy is explained in Section 3. The data and summary statistics are then discussed in Section 4. We present our results in Section 5 and conclude in Section 6. The figures and tables are provided in the Appendix. 1Our findings are different from those in Ottaviano et al. (2018), who find a negative effect of network at the bilateral level. That paper focuses on service trade and emphasizes that imported service tasks can be substituted by immigrants working in the firm. We explore instead the offshoring of production activities for a representative sample of firms in the manufacturing industry, where the information channle may be more important than task substytuion. 4 2 Theoretical Intuition In this section, we present a simple theoretical model that describes the economic mechanism that this paper focuses on. We consider a multi-country economy, with a continuum of countries i, j ∈[0,1]. There are two sectors in this economy. One sector provides a single homogeneous good. This good is used as the numeraire, and its price is set at 1. This good is produced under perfect competition. The second sector supplies a differentiated good under monopolistic competition: each firm is a monopoly for the variety of good that it produces, and varieties are imperfect substitutes. All the countries produce both goods, which can be freely traded. 2.1 Demand We assume the world is populated by a unit measure of consumers with identical preferences. The utility function of these consumers is increasing in the consumption of the homogeneous good xoand in the quantity q(x) of each variety xof the differentiated good, where Xis the set of all the available varieties: U=x1−µ oZx∈X q(x)σ−1 σdi(µσ −1+σ) .(1) σ > 1 is the elasticity of substitution between varieties, µis the share of income devoted to consumption of the differentiated good (so 1 −µrepresents the consumer’s expenditures devoted to xo). Consumers choose the demand for the differentiated good that maximizes their utility (1) subject to their budget constraint. In this typical Dixit-Stiglitz framework, the (inverse) demand function for a single variety is p(x) = Aq(x)−1 σ, where A =P1−σ µ−(1/σ) .(2) The inverse demand function (2) features an index Aof the market size of the differentiated sector, which is increasing in the price index P: P=Zx∈X p(x)1−σdx1/1−σ .(3) 5 2.2 Offshoring and supply Regarding the supply of the differentiated good, in each country, there is a continuum of firms z; these firms are heterogeneous in their productivity θz∈[0,1] and produce a single variety of the differentiated product. The production technology of firms includes both headquarter tasks, h, and manufacturing tasks, m. Headquarter services are performed locally and thus identify the home country of the firm. The manufacturing tasks are supplied everywhere and can be performed in different countries, i.e., offshored abroad. We assume that labor is the only factor of production in the economy. Labor is supplied inelastically in all countries and used both to produce the homogeneous good and to perform the tasks necessary to produce the differentiated good. Perfect competition in the homogeneous sector and free trade imply that wage rates are set at their reservation level and equalized across countries, so that they can be set equal to one. Let us now consider a firm in country ithat offshores production to foreign country j. We use the following Cobb-Douglas production function: xij(z) = θz(h)1/2(λjm)1/2.(4) The output of firm zfrom country idepends on its productivity, θz2, the local headquarter’s inputs, h, and the manufacturing inputs, m, which are offshored to country j.λj>1 is the efficiency of the labor inputs available in country jin terms of the performance of manufacturing tasks. We assume that λi= 1, so λj>1 implies that firms can access more efficient manufacturing inputs by offshoring production. Finally, note that the production function (4) features technology that is equally intensive for manufacturing and headquarter services. All firms incur a positive fixed cost f > 1 when they start production. An additional positive fixed cost, rj>0, is paid by firms when they offshore production to country j, e.g., due to a different functioning of the institutions there (such as weaker enforcement of judicial institutions, higher regulatory constraints, and higher levels of corruption), information frictions, and linguistic barriers. However, some of these entry costs are attenuated when a firm has a specific network in country j(e.g. created by the firm’s workers who originate from country j), 0 < φzj ≤1.3 2As the productivity distribution is country-specific, we should also add subscript i. However, for the sake of simplicity, we omit it, and as we focus on country i, this does not reduce expositional clarity. 3The combined assumptions of f > 1, rj>0 and 0 < φzj ≤1 imply that the network benefits do not 6 The profit function of an offshoring firm is πij(z) = pxij(z)−(h+m)−(f+rj−φzj).(5) We now substitute demand (2) and the production function (4) into (5). After some simplifications, we obtain: πij(z) = Aθz(h)1/2(λjm)1/2 −1+σ σ−(h+m)−(f+rj−φzj).(6) The firm chooses hand mto maximize (6). From the first order conditions, we obtain the input demand for the offshoring firm: m∗(z) = h∗(z) = A 2σσ−1 σσ λ σ−1 2 jθσ−1 z(7) where m∗=h∗follows because headquarters and manufacturing services have the same intensity in the production function. If we substitute (7) back into (6), then we obtain the equilibrium profits of the offshoring firm: π∗ ij(z) = 21−σAσσ−1 σσ1 σ−1λ 1 2(σ−1) jθ(σ−1) z−(f+rj−φzj).(8) From the condition π∗ ij(z)>0, we derive the productivity threshold faced by firms that offshore production abroad: θ > ˆ θzj, where ˆ θzj =2σ (σ−1)Aσ −1+σγj1/2(σ(f+rj−φzj)) 1 −1+σ(9) Only firms characterized by a productivity level θ > ˆ θzj that is sufficiently high will find it profitable to offshore production to country j. According to (9), ˆ θzj depends on the size of the differentiated good sector (A); therefore, we define the following lemma. Lemma :More firms will offshore production when the size of the differentiated sector increases, i.e., dˆ θzj/dA < 0. When the size of the differentiated sector increases, more firms will be needed to satisfy the increased demand for consumption. Put differently, as ˆ θzj decreases, more firms will find completely offset the fixed costs. These assumptions are necessary to guarantee economies of scale for the firms. 7 Using data from the Foreign Trade Statistics Register, we construct a firm-level measure of offshoring. Following Hummels et al. (2014b), we use a “narrow offshoring” measure that is constructed as the sum of imports for the same product category as firm exports and domestic sales.9This narrow measure of offshoring rule out the imports of raw materials that are inconsistent with standard definitions of offshoring.10 We look at two aspects of offshoring. First, we analyze whether regulations and networks affect the decisions of firms to begin offshoring production activities in a destination market (i.e., the extensive margin of offshoring). Second, in order to corroborate the interpretation on the results obtained for the extensive margin, we also estimate the impact of institutions and networks on firms’ values of bilateral offshoring, conditional on firms already offshoring in a destination market (i.e., the intensive margin of offshoring). Figure 2 presents evidence on the distribution of offshoring across manufacturing industries in Denmark. We report the twelve industries with the highest share of offshoring firms. Offshoring is common in routine, lower-skilled industries such as motor vehicles, machinery and equipment, basic metals and textiles, where more than forty percent of firms offshore.11 We deem these descriptive statistics to be reasonable because they are consistent with existing findings that the offshoring of routine, lower-skilled jobs is relatively common (Hummels et al., 2014b), and they indicate that our measure is successfully capturing offshoring in the data. [Insert Figure 2 about here] The most popular offshoring destinations over the period 2006-2012 are reported in Figure 3. Most of the neighbouring countries are included in the top 12 destinations of Danish offshoring. The EU is in fact the recipient of 58 percent of Danish offshoring, with Norway, Germany and Sweden representing the top three destinations. These patterns are consistent with the known fact that most offshoring takes place among developed countries. Many of the neighboring Eastern European countries are included in the top 20 destinations of Danish offshoring. Poland, Latvia, Lithuania and the Czech Republic, in fact, represent some of the largest recipients of Danish offshoring. [Insert Figure 3 about here] 9The first 6-digits of the Combined Nomenclature in the Foreign Trade Statistics Register are aggregated to the 4-digit level to considerably improve consistency over time. 10In the sensitivity analysis, we use an alternative ”broad offshoring” measure that includes all the imports of a given firm in a given year, independently of the product code. 11These findings are consistent with a distinct measure of offshoring based on survey data for Denmark (Bernard et al., 2017a). 14 4.1 Descriptive statistics Table 2 reports the descriptive statistics of the main variables used in the empirical analysis. According to our destination-specific definition of offshoring, the average offshoring rate at the firm level across all destination countries is 3 percent. Focusing on the intensive margin of offshoring, we see that the average volume of offshoring to each destination, conditional on offshoring, is approximately 30,000 dkk at the firm level. [Insert Table 2 about here] Given the detailed employer-employee matched data, we are able to control for an extensive set of firm and workforce-level characteristics. First, we are able to measure our network variable, φijt−1, as the firm share of foreign workers from country jat time t−1. Second, we can extend our specification with the following firm level controls: productivity, size, capital intensity, the number of offshoring destinations, multi-establishment and foreign ownership dummies. We also consider employees’ nationality, gender, occupation, average age, education, tenure, and work experience. The summary statistics for the networks and control variables are reported at the bottom of Table 1. Table 3 reports the main descriptive statistics for the measures of fixed costs associated with the institutions of the country of destination. These costs are derived from the indicators of labor, business and credit regulations collected by the World Bank at the country level. Specifically, the labor regulations include the following: i) whether fixed term contracts are prohibited for permanent tasks; ii) the maximum number of working days per week, calculated as 7-the maximum allowed number of working days; iii) whether employers must notify or consult a third party before a collective dismissal of employees; and iv) the minimum wage (measured as the ratio of the minimum wage relative to median wages). We combine the labor regulations with the principal component analysis (PCA) to calculate an index that summarizes the institutional costs due to labor market rigidity. The business regulations include the following: i) the time required to start a business (in days); ii) the time required to register property (in days); iii) the time required to prepare and pay taxes (in hours); and iv) the time required to export goods (in days). The corresponding PCA index is also reported in the same table. Finally the credit related costs in our empirical analysis include the following: i) the lack of private credit bureau coverage (percent of adults), computed as 100-private credit bureau coverage; ii) the lack of an investors’ protection index, calculated as 10-business extent of disclosure index (the disclosure index ranges from 0=less disclosure to 10=more disclosure); iii) the cost of enforcing contracts (percent of claims)12 and viii) the 12The cost is recorded as a percentage of the claim, which is assumed to be equivalent to 200 percent 15 rate of insolvency, calculated as 100-the recovery rate (cents on the dollar). The PCA index of credit risk is obtained by combining these credit related items. The last row of Table 2 shows the main descriptive statistics of our country-specific corruption index, which is based on the inverse of the estimated control of corruption indicator (Kraay, 2010). [Insert Table 3 about here] Figure 4 shows how the regulations and our measure of corruption described in Table 3 changed from 2006 to 2012 across different destination countries by plotting the index in a given area in 2006 against the same index in 2012 for a given economy. A few things are worth noting. First, there are substantial time-series variations in each index as the large majority of destination countries do not lie on the 45-degree line, which is exactly the temporal variation we will exploit in our empirical analysis. Second, whereas most of the destination countries have experienced a reduction in business regulations and credit risk over the sample period, the trends for the index of labor market rigidity and our measure of corruption are less clear-cut. [Insert Figure 4 about here] Figure 5 shows the destination countries with the highest index along each of the four dimensions described in Table 3. It is interesting to note that 5 destinations with the highest index of business regulations (Angola, Chad, Haiti, Iraq, Ukraine and Venezuela) are among the countries with the highest credit risk scores, as reported in the two second and third panels, respectively. Furthermore, the second and fourth panels of Figure 5 reveal that the 5 destinations with the highest corruption index are also among the countries featuring the highest credit risk. We do not observe the same overlap across destinations if we compare the country ranking in terms of the index of labor market rigidity to the other three rankings. This result is confirmed by the fact that there is a relatively low and negative correlation between the index of labor market rigidity and the other three indexes, whereas the index of business regulation is highly and positively correlated with the index of credit risk and the corruption index.13 [Insert Figure 5 about here] of income per capita or 5,000 dollars, whichever is greater. Three types of costs are recorded: court costs, enforcement costs and average attorney fees. 13The correlation coefficient between the index of labor market rigidity and the index of business regulation (index of credit risk) is -0.22 (-0.26). The correlation coefficient between the index of business regulation and the index of credit risk (corruption index) is 0.56 (0.70). 16 Our main hypothesis suggests that regulations entailing higher institutional costs, higher levels of perceived corruption or a weaker network will reduce a firm’s extensive margin for offshore production abroad. To provide preliminary insights into these relationships, Figure 6 shows separate scatter plots of our indexes of labor market rigidity, business regulations, credit risk, corruption and networks against the extensive margin of offshoring at the destination-year level. A statistically significant negative (positive) relationship is evident between our indexes (the network) and the average share of firms that offshore at the yeardestination level. Consistent with our hypothesis, the number of offshoring firms decreases (increases) when the institutional fixed costs (networks) increase. Furthermore, Figure 7 plots our indexes and networks variable against the average log of offshoring values at the year-destination level. The intensive margin of offshoring does not correlate with either the institutions or the network. This result does not contradict our theory, which predicts a negative (positive) relationship between institutional costs (networks) and the extensive margin of offshoring only. It is encouraging that such significant relationships emerge in the data, even though they are raw. Next, we examine whether these results hold in a more rigorous empirical specification that controls for a large set of confounding factors. [Insert Figures 6 and 7 about here] 5 Results In what follows, we first discuss the main results regarding the impact of institutions and networks on firms’ extensive and intensive margins of offshoring. They are obtained from the estimation of equation (12). We then investigate how institutions and networks interact with each other. Then, we show extensions, checks and refinements of the main results. 5.1 Main results Table 4 presents the baseline results showing the impact of regulations and networks on the extensive margin of offshoring. In column 1, we report a specification showing the impact of regulations, corruption and networks on the extensive margin of offshoring after controlling for only industry by year, firm and municipality fixed effects. In this basic specification, we see that the presence of employment protection measures, weaker credit coverage, a lower resolving insolvency index and higher corruption in the destination market at time t−1 leads to a significant decrease in the bilateral probability that a firm will offshore to that 17 destination at time t.14 None of the business regulations are significantly associated with the bilateral probability of offshoring.15 In contrast, our network variable is positively associated with the firm’s extensive margin at the bilateral level. In Columns 2-4, we sequentially include firm-level controls (such as labor productivity) and destination country fixed effects in the specification. Furthermore the network variable is instrumented for by using its shiftshare prediction, as described in equation (13).16 The coefficients for the institutions remain unchanged after controlling for the numerous firm characteristics and fixed effects, whereas the instrumented coefficient estimated on the network variable is of smaller size, which is consistent with the upward bias hypothesis highlighted in the methodological section. The same coefficients are never statistically significant when we consider the intensive margin, conditional on the decision to offshore (see Columns 5-8).17 We interpret all these results as being consistent with our two main hypotheses, which are highlighted in the methodological section. On one hand, in line with Hypothesis (1), the presence of restrictive employment protection legislations, high levels of perceived corruption, and the lack of credit coverage and credit solvency represent proxies for the institutional costs of offshoring to a certain destination country. These are some of the fixed costs that hinder the early stages of firms’ offshoring (i.e., firms’ extensive margin) but do not affect the volume of offshoring activities, conditional on having accessed a specific foreign market. On the other hand, consistent with Hypothesis (2), a firm’s network in the country of destination helps the firm reduce the overall fixed costs (Peri and Requena-Silvente, 2010), which, in turn, positively affects the extensive margin but not the offshoring volume, conditional on offshoring. Furthermore, the estimated positive effect of labor productivity at the firm level on offshoring is intuitive and consistent with a Melitz-type theoretical framework when we focus on the extensive margin of the offshoring decision. [Insert Table 4 about here] To simplify the interpretation of our main results, we now present the effects of regulations identified by using the aggregate PCA indexes, which are calculated as highlighted in 14To simplify the comparison across the magnitudes involved in each estimation, all the regression tables show standardized coefficients estimated on the z-score of each explanatory variable. 15This result is likely due to the fact that our credit variables are highly correlated with business regulations. If we estimate a simpler specification in which credit and business regulations are entered separately in the extensive margin equation, we find that they are all significantly and negatively associated with the bilateral probability of offshoring. 16The first-stage results, which are reported in the bottom panel of Table 4, show that the instrument has a significant positive impact on our network variable. 17Very similar results are obtained by using an alternative measure of intensive margin, which is calculated as the share of destination-specific imports in total imports. 18 the previous section. Column 1 of Table 5 includes the coefficients estimated for the index of labor market rigidity, business regulations and credit risk in the most complete specification for the extensive margin. These results are in line with the findings obtained with the disaggregated measures: although business regulations do not affect the extensive margin, an increase in labor market rigidity and credit risk in the destination country significantly reduces the firm’s bilateral probability of offshoring.18 Quantitatively, our regression analysis suggests that a one standard deviation increase in the index of labor market rigidity19 at time t−1 will lead to a 0.007 decrease in the probability of offshoring, i.e., a 20 percent decrease in the firm’s extensive margin of offshoring.20 The impact of credit risk is about half this much: the firm’s bilateral probability of offshoring is reduced by 10 percent at time twhen there is a one standard deviation increase in the related index at time t−1.21 Consistent with the results reported in Table 4, our measure of corruption negatively affects the probability of offshoring: a one standard deviation increase in the corruption index decreases the extensive margin by approximately 3 percent.22 Furthermore, the presence of a network in the country of destination at the municipality in which the firm is localized at time t−1 positively affects the extensive margin. According to the instrumented coefficient, a one standard deviation increase in the firm’s network leads to a 0.010 increase in the probability of offshoring, which corresponds to an increase of approximately 32 percent in the extensive margin. We deem that all the effects estimated on our measures of institutional costs and networks are sizeable from an economic point of view. In fact, these effects have a similar or stronger predictive power than the one estimated on productivity: a one standard deviation increase in the firm’s productivity is in fact associated with only a 3 percent increase in the probability of offshoring. 18As mentioned in Section 4.1, destinations with high credit risk also tend to have numerous business regulations, which may explain the lack of significance observed for the index of business regulations. In fact, when we add these indexes separately in the main specification, the estimated coefficients are all negative and statistically significant. These additional results are available on request from the authors. 19This approximately corresponds to an increase in labor market rigidity from the level of Austria, whose average index is at the median of the distribution, to that of Portugal, whose average index is at the 95th percentile of the distribution. 20This figure is calculated by dividing our standardized coefficient by 100 for the average firm’s probability of offshoring reported at the bottom of Table 5. 21This approximately corresponds to an increase in credit risk from the level of Germany, whose average index is at the median of the distribution, to that of Macedonia, whose average index is at the 95th percentile of the distribution. 22This approximately corresponds to an increase in the corruption index from the level of France, whose average index is at the 25th percentile of the distribution, to that of Azerbaijan, whose average index is at the 90th percentile of the distribution. 19 5.2 Refinements of the main results: the role of interactions In the next step, we proceed by exploring whether the institutional indexes interact with the network variable in column 2 of Table 5 .23 Such exercise reveals whether the presence of a network of immigrants plays a stronger role in countries where local institutions are particularly poor. This may be because in cases of poor institutional features the network provides information to reduce risk or to increase knowledge. We find a positive and significant interaction between the network variable and the credit risk index. This suggests that the positive effect of a network is magnified in destinations characterized by high credit risk. Networks are more effective in promoting offshoring to those destination markets with higher credit risk. This result suggests that a local network may act as a substitute for institutions that guarantee creditors (i.e., reduce the costs associated with credit risk) and increase the firm’s bilateral probability of offshoring to a certain destination. On the other hand we do not find significant effect of interactions between networks and the other two indices. Furthermore, the measure of corruption interacts negatively with the network variable: the positive effect of networks is in fact reduced in destinations characterized by high levels of corruption. Networks are therefore less effective in promoting offshoring in corrupt environments and vice versa. To simplify the interpretation of all these findings, we plot the estimated marginal effect of our network variable on the extensive margin against the whole distribution of the two indexes of credit risk and corruption. Specifically, the top (bottom) panels of Figure 8 report the estimated marginal effects of our network against the standardized index of credit risk (corruption) by setting the indexes of business regulations, labor market rigidity and corruption (credit risk) at respectively the 25th, 50th and 75th percentile of the distribution of these indexes.24 These plots confirm the substitutability (complementarity) between our network variable and the index of credit risk (the corruption index), given that the im23The interactions of our regulation indexes with the network variable are instrumented by their interactions with the shift share prediction described in equation (13). 24The standard errors of these marginal effects are calculated to account for all the covariance terms involved in the fully interacted specification, as in the following interaction specification: Offijmt =α+γ1φijt−1+β1index labrigjt−1+β2index busregjt−1+ β3index credriskjt−1+β4index corrupjt−1+γ2(φijt−1)(index labrigjt−1) + γ3(φijt−1) (index busregjt−1) + γ4(φijt−1)(index credriskjt−1) + γ5(φijt−1)(index corrupjt−1) + X0 it−1ζ+θi+λj+ ηm+ρt+ijt The marginal effects of our network variable φijt−1reported in the top panels of Figure 8 are calculated as follows: dOffijmt dφijt−1 =γ1+γ2¯ index labrigjt−1+γ3¯ index busregjt−1+γ4index credriskjt−1+γ5¯ index corrupjt−1 20 pact of the former increases (decreases) as the credit risk index (corruption index) in the destination country becomes higher for given levels of the other indexes. Interestingly, the marginal effects plotted in the top (bottom) panels of Figure 8 reset to a lower (higher) level when we select higher moments of the corruption (credit risk) index. Furthermore, the estimated coefficients for our network variable retain their statistical significance along the whole distribution of the two indexes, i.e., credit risk and corruption. [Insert Figure 8 about here] Similar results regarding the interaction effects are obtained in a simpler specification, in which we only include the variable for the presence of labor employment protection and a PCA index of credit risk calculated by combining the variables on credit coverage and resolving insolvency (see column 3 of Table 5). [Insert Table 5 about here] Consistently our two main hypothesis, none of the coefficients discussed in Table 5 is precisely estimated in the regression for the intensive margins (columns 4-6).25 5.3 Refinements of the main results: alternate measure of networks Our main analysis uses the bilateral share of foreign workers at the firm level to measure whether a firm has a network in the destination country. This section examines whether our results hold using alternative definitions for networks. Specifically, columns 3 and 4 of Table 6 use the firm’s bilateral share of white-collar foreign workers at time t−1 as the network variable. Columns 5 and 6 focus on the share of blue-collar foreign workers. For these where ¯ index labrigjt−1,¯ index busregjt−1and ¯ index corrupjt−1are alternatively set at either the 25th, 50th or 75th percentile of the corresponding index distribution. The variance in the marginal effects is calculated by considering the covariance across all the different indexes included in the main specification. 25We have also investigated whether there exists a significant interaction between the institution variables. Namely we have analyzed whether the presence of rigidities/imperfections in one dimension is magnified by the presence of rigidities in another dimension. The individual effects, estimated in this specification are consistent with the findings reported in column 1, and none of the interaction is significant at standard levels. These results, which are available on request from the authors, suggest that the linear specification is reasonably good for our institutional variables and each factor affects offshoring independently from the other. 21 estimations, we construct analogous instruments using the predicted immigration of whiteand blue-collar workers in the municipality in which the firm is located at time t−1. The results obtained by using these two narrower definitions of networks are qualitatively similar to the main results re-reported in columns 1 and 2 for comparison purposes, i.e., networks significantly affect the extensive margin of offshoring but not its intensive margin. In terms of the magnitudes of these coefficients, the positive effect of the network provided by whitecollar workers is stronger than that reported for the blue-collar workers. Specifically, a one standard deviation increase in the share of white-collar foreign workers doubles the bilateral probability of offshoring, whereas the same effect estimated on the blue-collar network is half as much as the one estimated in the main analysis, i.e., approximately 15 percent. These additional findings reveal that foreign workers in white-collar occupations (managers, middle managers and professionals) bring with them destination-specific knowledge that can promote offshoring activities in their country of origin more effectively compared to their blue-collar counterparts. Similarly, in columns 7 and 8 of Table 6, we estimate the impact of hiring managers from domestic companies that offshore to a specific destination country. Our results show that domestic firms poaching managers with destination-specific offshoring experience are 14 percent more likely to offshore to the same destination country at time t but do not feature higher offshoring volumes, conditional on offshoring. This poaching effect on offshoring is consistent with the findings reported in Mion et al. (2017), according to which export experience gained by managers in previous firms improves the export performance of their current firm. All these results combined together reveal that workers with destinationspecific knowledge, which is either acquired from their country of origin or gained in previous companies, help domestic firms reduce the fixed costs of offshoring to a specific destination country, especially if they are in a white collar occupation. Our main results on the network variable are also confirmed, when we use the baseyear method to address the endogeneity of the network variable instead of relying on the shift-share instrumental variable approach. In columns 8 and 9, our network variable is measured as the firm-level bilateral share of foreign workers at the base-year, which should be predetermined with respect to the current offshoring decision. The base year is generally set at 2005, i.e., one year before the estimation sample starts. If a firm enters the market and imports at any time during our sample period, we treat its first year of existence as the pre-sample year and focus our estimation on the subsequent changes in offshoring. The results obtained from this alternative approach confirm the main results for the instrumental variable (IV), i.e., networks impact the bilateral probability of offshoring but not offshoring volumes. The size of the coefficient estimated on the extensive margin is twice as large as that obtained in the main analysis: a one standard deviation increase in the base year share 22 of foreign workers increases the current bilateral probability of offshoring by 60 percent. Finally, columns 11 and 12 of Table 6 focus on the bilateral share of foreign workers in the municipality in which the firm is located at time t−1 rather than the number of foreign workers in the firm. The instrument, however, is still the shift-share prediction at the municipality level. Using the share of municipality-level immigrants is less specific, but the advantage is that firm-specific endogeneity issues are less severe. In addition, immigration may influence offshoring decisions not only directly through the hiring of foreign employees but also more generally by promoting the development of informal networks with foreign workers employed in other companies in the local labor market (which is proxied by the municipality in which the firm is located). The results obtained with this broader measure of networks are almost identical to the baseline results: a one standard deviation increase in the bilateral share of foreign workers within the municipality at year t−1 increases the extensive margin by 33 percent and the corresponding effect on the intensive margin is not statistically significant. Ultimately, the results in Table 6 demonstrate that our findings on the network effects are robust to a variety of alternate definitions. [Insert Table 6 about here] 5.4 Additional sensitivity analysis In this section we evaluate how the coefficients estimated for the institution and network variables change when we, for example, focus on specific sub-samples of firms or when we use alternative specifications for our index variables. To simplify the presentation of these sensitivity exercises, we only discuss the results obtained from the specification used in columns 1 and 5 of Table 5, i.e., the one featuring the PCA indexes of institutional costs (i.e., labor market rigidity, business regulations, credit risk and corruption) and the network variable. 5.4.1 Non-linear specifications While our model predicts that lower institutional costs (stronger networks) in the destination markets are positively associated with the firms’ bilateral extensive margin of offshoring, it is silent on the curvature of these relationships. To explore the existence of non-linearities in the effects discussed so far, we consider an additional specification, in which we add the squares of the institutional indexes and of the network variable. The results are reported in the first two columns Table 7. Our main results are confirmed: in all the specifications, 23 Head, K. and Ries, J. (1998). Immigration and Trade Creation: Econometric Evidence from Canada. Canadian Journal of Economics, 31(1):47–62. Hummels, D., Jorgensen, R., Munch, J., and Xiang, C. (2014a). The Wage Effects of Offshoring: Evidence from Danish Matched Worker-Firm Data. American Economic Review, 104(6):1597–1629. Hummels, D., Jorgensen, R., Munch, J., and Xiang, C. (2014b). The Wage Effects of Offshoring: Evidence from Danish Matched Worker-Firm Data. American Economic Review, 104(6):1597–1629. Kraay, Aart Kaufmann, D. M. M. (2010). The worldwide governance indicators : methodology and analytical issues. The World Bank. Melitz, M. J. (2003). The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry Productivity. Econometrica, 71(6):1695–1725. Mion, G., Opromolla, L. D., and Sforza, A. (2017). The diffusion of knowledge via managers’ mobility. CESifo Working Paper Series No. 6256. Olney, W. W. (2013). A Race to the Bottom? Employment Protection and Foreign Direct Investment. Journal of International Economics, 91(2):191 – 203. Ottaviano, G. I., Peri, G., and Wright, G. C. (2018). Immigration, Trade and Productivity in Services: Evidence from U.K. Firms. Journal of International Economics, 112:88 – 108. Peri, G. and Requena-Silvente, F. (2010). The Trade Creation Effect of Immigrants: Evidence from the Remarkable Case of Spain. Canadian Journal of Economics, 43(4):1433– 1459. Rauch, J. E. and Trindade, V. (2002). Ethnic Chinese Networks In International Trade. The Review of Economics and Statistics, 84(1):116–130. Sofronis K. Clerides, Saul Lach, J. R. T. (1998). Is Learning by Exporting Important? Micro-Dynamic Evidence from Colombia, Mexico, and Morocco. The Quarterly Journal of Economics, 113(3):903–947. Viscusi, W., Harrington, J. E., and Vernon, J. M. (2005). Economics of Regulation and Antitrust, 4th Edition, volume 1. The MIT Press, 4 edition. 30 Figure 1: Bilateral share of foreign workers and its instrument 0 .05 .1 .15 Share of foreign workwers 0 .01 .02 .03 .04 .05 .06 .07 .08 .09 .1 Predicted share of foreign workwers Notes: The bilateral share of foreign workers in a given municipality and year is reported on the vertical axis. The predicted bilateral share (IV) of foreign workers in a given municipality and year is reported on the horizontal axis. 31 Figure 2: Sectors with the largest share of offshoring firms on average (2006-2012) 0 .1 .2 .3 .4 Share of offshoring firms within industry Motor Vehicles Machinery and Equip. Textiles Electrical Equipment Plastic and Rubber Wholesale (Motor Vehicles) Food Fabricated Metal Basic Metal Other Manufacturing Furniture Wholesale (excl. Motor Vehicles) Notes: Authors’ calculation using data from the Danish Integrated Database for Labor Market Research. 32 Figure 3: Destination countries with the largest share of offshoring firms on average (20062012) 0 .1 .2 .3 Share of offshoring firms Germany Sweden Norway Great Britain Netherlands France Finland Belgium Italy Poland Notes: Authors’ calculation using data from the Danish Integrated Database for Labor Market Research. 33 Figure 4: Regulations and corruption by destination country in 2006 and 2012 AE AG AL AM AO AR AT AU AZ BE BF BG BI BJ BN BO BTBW BY BZ CA CG CHCI CL CM CO CR CV CZ DE DJ DM DO DZ EC EE EG ES ET FI FJ FR GA GB GD GE GH GM GN GQ GR GTGW HK HN HU IE IL IQ IR IS ITJM JO KE KG KN KR KW KZ LA LB LC LK LS LT LV MA MD MG ML MN MR MU MY NA NE NI NL NO NP NZ OM PA PE PH PL PT PY RS SA SC SE SG SK SN ST SV SY SZ TDTG TH TJ TN TR TT TZ UA UG UY UZ VC VE VN YE ZA ZM -2 -1 0 1 2 Index of labor market rigidity, 2012 -2 -1 0 1 2 Index of labor market rigidity, 2006 AE AG AL AM AR AT AU AZ BE BF BG BI BJ BO BT BW BZ CA CH CI CL CM CO CR CV CZ DE DJ DM DO DZ EC EE EG ES ET FI FJ FR GA GB GD GE GH GM GN GQ GR GT HK HN HU IE IL IR IS IT JM JO KE KG KN KR KW KZ LA LB LC LK LS LT LV MA MD MG ML MN MR MU MY NA NE NI NL NO NP NZOM PA PE PH PL PT PY RS SA SC SE SG SK SN ST SV SY SZ TH TJ TN TR TT TZUG UY UZ VC VN YE ZA ZM -2 -1 0 1 2 Index of business regulations, 2012 -2 -1 0 1 2 Index of business regulations, 2006 AE AG AL AM AO AR AT AU AZ BE BF BG BI BJ BN BO BT BW BY BZ CA CG CH CI CL CM CO CR CV CZ DE DJ DM DO DZ EC EE EG ES ET FI FJ FR GA GB GD GE GH GM GN GQ GR GT GW HK HN HU IE IL IQ IR IS IT JM JO KE KG KN KR KW KZ LA LB LC LK LS LT LV MA MD MG ML MN MR MU MY NA NE NI NL NO NP NZ OM PA PE PH PL PT PY RS SA SC SE SG SK SN ST SV SYSZ TD TG TH TJ TN TR TT TZ UA UG UY UZ VC VE VN YE ZA ZM -4 -2 0 2 Index of credit risk, 2012 -4 -2 0 2 Index of credit risk, 2006 AL AM AO AR AZ BF BG BI BJ BN BO BT BY BZ CF CG CI CM CO CR CV CZ DJ DM DO DZEC EG ET FJ GA GD GE GH GM GN GQ GR GT GW HN HT HU IL IQ IR IT JM JO KE KG KH KM KN KR KW KZ LA LB LK LS LT LV MA MD MGML MN MR MU MW MY MZ NA NE NI NP OM PA PE PH PL PY RS RW SA SC SK SL SN SR ST SV SY SZ TD TG TH TJ TN TR TT TZ UA UG UZ VC VE VN YE ZA ZM ZW -1 -.5 0 .5 1 1.5 Corruption index, 2012 -1 -.5 0 .5 1 1.5 Corruption index, 2006 Notes: The index of labor market rigidity is estimated by combining all the labor regulations (limits on fixed term contracts; limits on working days per week; employment protection measures; and the minimum wage) with principal component analysis. The index of business regulations is estimated by combining all the business regulations (the time it takes to open a business; the time it takes to register property; the time it takes to pay taxes; the time it takes to export goods) with principal component analysis. The index of credit risk is estimated by combining all the credit regulations (100credit coverage; 10-investor protection; enforcing contracts; 100-resolving insolvency) with principal component analysis. The corruption index is the inverse of the estimated control of the corruption indicator. 34 Figure 5: Destination countries with the highest regulation and corruption indexes 0 .5 1 1.5 2 Index of labor market rigidity NP NE LR HN TZ EE ZA UA PY SN ES LV 0 .5 1 1.5 2 Index of business regulations VE UA TG TD HT GW BN AO IQ CG BO KZ 0 .5 1 1.5 2 Index of credit risk HT ST RW LA CV AO BF UA VE TD PH IQ 0 .5 1 1.5 Corruption index GQ TD ZW IQ AO SD KG KH HT LA CF BI Notes: The index of labor market rigidity is estimated by combining all the labor regulations (limits on fixed term contracts; limits on working days per week; employment protection measures; and the minimum wage) with principal component analysis. The index of business regulations is estimated by combining all the business regulations (the time it takes to open a business; the time it takes to register property; the time it takes to pay taxes; the time it takes to export goods) with principal component analysis. The index of credit risk is estimated by combining all the credit regulations (100credit coverage; 10-investor protection; enforcing contracts; 100-resolving insolvency) with principal component analysis. The corruption index is the inverse of the estimated control of the corruption indicator. 35 Figure 6: Extensive margin of offshoring, institutions and network 0 .1 .2 .3 .4 .5 .6 Extensive margin of offshoring -3 -2 -1 0 1 2 Index of labor market rigidity 0 .1 .2 .3 .4 .5 .6 Extensive margin of offshoring -2 -1 0 1 2 Index of business regulations 0 .1 .2 .3 .4 .5 .6 Extensive margin of offshoring -4 -2 0 2 Index of credit risk 0 .1 .2 .3 .4 Extensive margin of offshoring -3 -2 -1 0 1 2 Corruption index 0 .1 .2 .3 .4 Extensive margin of offshoring 0 .005 .01 .015 .02 Network Notes: The share of firms that offshore at the year-destination level are reported on the vertical axis. The index of regulations or the network variable at the year-destination level are reported on the horizontal axis. A network is measured as the firm-level bilateral share of foreign workers. 36 Figure 7: Intensive margin of offshoring, institutions and networks 0 5 10 15 20 Intensive margin of offshoring -3 -2 -1 0 1 2 Index of labor market rigidity 0 5 10 15 20 Intensive margin of offshoring -2 -1 0 1 2 Index of business regulations 0 5 10 15 20 Intensive margin of offshoring -4 -2 0 2 Index of credit risk 0 5 10 15 20 Intensive margin of offshoring -3 -2 -1 0 1 2 Corruption index 0 5 10 15 20 Intensive margin of offshoring 0 .005 .01 .015 .02 Network Notes: The share of firms that offshore at the year-destination level are reported on the vertical axis. The index of regulations or the network variable at the year-destination level are reported on the horizontal axis. A network is measured as the firm-level bilateral share of foreign workers. 37 Figure 8: Marginal effects of networks on the extensive margin of offshoring ******************* -.02 0 .02 .04 Marginal effect of network (1) -2 -1 0 1 2 Index of credit risk (standardized) ******************* -.02 0 .02 .04 Marginal effect of network (2) -2 -1 0 1 2 Index of credit risk (standardized) ******************* -.02 0 .02 .04 Marginal effect of network (3) -2 -1 0 1 2 Index of credit risk (standardized) ********* ******** -.02 0 .02 .04 Marginal effect of network (1) -2 -1 0 1 2 Corruption index ******************* -.02 0 .02 .04 Marginal effect of network (2) -2 -1 0 1 2 Corruption index ******************* -.02 0 .02 .04 Marginal effect of network (3) -2 -1 0 1 2 Corruption index Notes: The marginal effects of networks are estimated from the interaction specification as reported in column (3) of Table 4. In the top (bottom) panels, the marginal effect of networks (1) is calculated by interacting our network variable with the index of credit risk (corruption) and setting the index of labor market rigidity, business regulations and corruption (credit risk) at the 25th percentile of their distributions. In the top (bottom) panels, the marginal effect of networks (2) is calculated by interacting our network variable with the index of credit risk (corruption) and setting the index of labor market rigidity, business regulations and corruption (credit risk) at the median of their distributions. In the top (bottom) panels, the marginal effect of networks (3) is calculated by interacting our network variable with the index of credit risk (corruption) and setting the index of labor market rigidity, business regulations and corruption (credit risk) at the 75th percentile of their distributions. “*” indicates significance at the 95% level. 38 Table 1: Pre-trends tests ∆ Ext. Margin Offshoring ∆ Int. Margin Offshoring ∆ Share of Non-EU Img (1995-2000) (1995-2000) (2006-2012) (1) (2) (3) ∆ Non-EU Img IV (2006-2012) 0.638 22.722 0.885*** (2.615) (41.863) (0.233) N 15,425 15,425 15,425 R-sq 0.712 0.716 0.070 Notes: In columns 1 and 2 the dependent variable is the pre-sample trend (i.e. the change from 1995 to 2000) in destinationspecific offshoring outcomes at the municipality level. In column 3 the dependent variable is the long-run change (2006 to 2012) in the bilateral share of foreign workers at the municipality level. Regressions also include municipality fixed effects. Standard errors in parentheses are clustered at the municipality level. Significance levels: ***1%, **5%, *10%. Significance levels: ***1%, **5%, *10%. 39 Table 8: Institutions, network and firm’s offshoring, results by industry Labor intensive industries Capital intensive industries Service industries trading merchandise [1] [2] [3] [4] [5] [6] Extensive Intensive Extensive Intensive Extensive Intensive Index of labor market rigidityt−1–0.009342*** –0.124779 –0.001055** –0.279290 –0.002664*** 0.013290 (0.000973) (0.137045) (0.000407) (0.253183) (0.000356) (0.035068) Index of business regulationst−10.000887 0.098059 0.001116 –0.415847 –0.000453 0.010822 (0.000953) (0.087409) (0.000788) (0.282987) (0.000890) (0.038427) Index of credit riskt−1–0.004611*** –0.140198 –0.001201** 0.235629 –0.004396*** 0.008424 (0.001203) (0.096101) (0.000626) (0.169324) (0.000600) (0.032060) Corruption indext−1–0.000758** 0.147094 –0.001208** –0.242397 –0.008368*** 0.114904 (0.000322) (0.105655) (0.000480) (0.251844) (0.000814) (0.097453) Networkt−10.011317*** –0.048889 0.021110*** 0.214269 0.027426*** 0.111188 (0.003315) (0.134705) (0.003630) (0.326278) (0.004196) (0.189153) Mean Y 0.019 10.378 0.041 10.126 0.012 10.626 R-sq 0.192 0.334 0.128 0.274 0.097 0.431 N 529,444 10,116 874,406 36163 2,383,597 28,937 Notes: Dependent variable in columns 1-3 (4-6) is the extensive (intensive) margin of offshoring. All regressions coefficients estimated on the z-scores of each explanatory variable and refer to the most complete specification. Index of labor market rigidity is estimated by combining all labor regulations with principal component analysis. Index of business regulations is estimated by combining all business regulations with principal component analysis. Index of credit risk is estimated by combining all credit regulations with principal component analysis. We classify as “labor intensive” those industries with average capital intensity below the sample mean in the middle of the sample period. Standard errors are clustered at the municipality-destination-year level. Significance levels: ***1%, **5%, *10%. 46 Table 9: Institutions, network and firm’s offshoring, results by sub-samples Exporting firms Developed destination countries Developing destination countries [1] [2] [3] [4] [5] [6] Extensive Intensive Extensive Intensive Extensive Intensive Index of labor market rigidityt−1–0.007597*** 0.036211 –0.052060** –0.426927 –0.006979*** 0.022112 (0.000892) (0.082354) (0.023135) (0.698491) (0.000749) (0.100866) Index of business regulationst−10.000759 –0.091759 –0.006992 –0.238628 0.001003 –0.104912 (0.000703) (0.076027) (0.011348) (0.274805) (0.000697) (0.090465) Index of credit riskt−1–0.003755*** 0.096030 –0.025886** 0.390200 –0.004936*** 0.204307 (0.000934) (0.090833) (0.011655) (0.309809) (0.000851) (0.124140) Corruption indext−1–0.001998** –0.021581 –0.027655** 0.219605 –0.002221** –0.184417 (0.000774) (0.128689) (0.010583) (0.230004) (0.000597) (0.201678) Networkt−10.012378*** –0.128508 0.078192** –0.869828 0.009048*** –0.103029 (0.002497) (0.122212) (0.033512) (0.728292) (0.002038) (0.146180) Mean Y 0.037 10.176 0.152 10.144 0.014 10.244 R-sq 0.187 0.285 0.266 0.311 0.097 0.260 N 1,261,507 46,070 193,433 29,287 1,210,417 16,752 Notes: Dependent variable in columns 1-3 (4-6) is the extensive (intensive) margin of offshoring. All regressions coefficients estimated on the z-scores of each explanatory variable and refer to the most complete specification. Index of labor market rigidity is estimated by combining all labor regulations with principal component analysis. Index of business regulations is estimated by combining all business regulations with principal component analysis. Index of credit risk is estimated by combining all credit regulations with principal component analysis. Standard errors are clustered at the municipality-destination-year level. Significance levels: ***1%, **5%, *10%. 47 Table 10: Institutions, network and firm’s offshoring, alternate offshoring variables Broad offshoring FDI-based definition [1] [2] [3] Extensive Intensive Extensive Index of labor market rigidityt−1–0.034499*** –0.270267 –0.000125* (0.002112) (0.164527) (0.000068) Index of business regulationst−10.003986 0.000223 0.000146 (0.003412) (0.041549) (0.000123) Index of credit riskt−1–0.014048*** 0.017132 –0.000348** (0.00278) (0.047458) (0.000141) Corruption indext−1–0.001931** –0.114114 –0.000826 (0.000809) (0.090234) (0.000498) Networkt−10.032439*** 0.110731 0.001125** (0.005400) (0.079005) (0.000641) Mean Y 0.101 10.176 0.010 R-sq 0.413 0.351 0.106 N 1,403,850 156,761 1,403,850 Notes: In column (1) the dependent variable is a dummy variable equal to 1 if firm has non zero offshoring volumes (broad definition). In column (2) the dependent variable is the log of offshoring volumes (broad definition). In column (3) the dependent variable is the extensive margin based on information on outward FDI activity. All regressions show coefficients estimated on the z-scores of each explanatory variable and refer to the most complete specification. Index of labor market rigidity is estimated by combining all labor regulations with principal component analysis. Index of business regulations is estimated by combining all business regulations with principal component analysis. Index of credit risk is estimated by combining all credit regulations with principal component analysis. Standard errors are clustered at the municipality-destination-year level. Significance levels: ***1%, **5%, *10%. 48