Does off shoring aff ect industry employment ? Evidence from a wide European panel of countries

Th is paper contributes to the literature on the possible impact of international outsourcing on domestic labour markets. We focus on off shoring-employment relationship. Th e analysis is performed for a wide European panel, composed of 27 EU countries and 13 manufacturing sectors, observed in the period 1995-2009. Th anks to the use of input-output tables from the WIOD project, we measure the intensity of off shoring in the sectors, as well as its decomposition into the domestic and foreign components. Th eoretical background for our analysis is rooted in recent trade-in-tasks models of international trade. Our empirical results, based on the estimation of panel data model, suggest that indeed domestic employment in EU manufacturing can be pushed down by increased off shoring. More specifi cally, low skill workers are the ones to be aff ected the most because of shrinking labour demand at home.


INTRODUCTION
Th e pattern of international trade has changed substantially in the recent decades.Many tasks, especially those which are relatively low demanding in terms of skills and rather repetitive (so called 'routine tasks' -see Grossman and Rossi-Hansberg, 2008) have been moved abroad in order to exploit cost advantage of foreign countries and/or cross-country productivity diff erentials.It is estimated that nowadays more than half of the world's manufacturing imports are intermediate goods (primary goods, parts and components, and semi-fi nished products), and more than 70% of the world's services imports are intermediate services (OECD, 2013) so the eff ects of cross-border disintegration of production 1 cannot be ignored by the researchers.
At the same time, many labour markets in developed countries have been experiencing rapid changes concerning changing labour demand and/or wage structure (falling demand for less skill workers, a rise in wage inequality -documented by, among others: Feenstra and Hanson, 1999;Feenstra, 2000, Feenstra, 2010, Acemoglu and Autor, 2011).
Given these tendencies, a common fear expressed by employees and policy makers, mainly from the developed countries, was the following: how do trade opening and off shoring to low wage countries aff ect employment (and wages) in the countries from which the tasks are moved (off shored)?Th reats of negative impact of international outsourcing, especially in terms of cross-border labour substitution (and/or lowering wages) of the low skilled workers at home, have become a policy issue and an important theme in applied -policy oriented research debate.(OECD, 2007;WTO&ILO, 2007, 2011;WTO, 2008WTO, , 2013)).Political worries started to dominate the debate on possibly 'worrying' implications of trade integration with less developed countries (eg. a paper by A. Wood entitled "How trade hurt unskilled workers", JEP, 1995 or Blinder's, 2006 estimates of 30-40% of all U.S. manufacturing jobs being potentially off shorable).Since then off shoring undoubtedly has been one of the leading research themes in international economics underlying its signifi cant role in the global economy (surveyed in: OECD, 2007;Feenstra 2010).
Yet, there are still some research gaps to be fulfi lled.New possibilities in empirical analysis have emerged as a result of a new release of world input-output tables (WIOD).It is now possible to go beyond countrylevel studies (see Crinò, 2009 for a survey on the fi rst wave of related empirical studies).We now dispose of comparable input-output statistics needed for the calculation of precise off shoring measures and obtaining a match between outsourcing and labour market data for many countries.
In this paper we focus on trends observable in European countries.Th e eff ects of off shoring can be exhibited on wages, even though its impact might be actually smaller than conventionally perceived (see Parteka and Wolszczak-Derlacz, 2015 for recent EU-focused evidence).Countries diff er in labour market institutions, so it can happen that in economies with rigid wages, the impact of off shoring on domestic labour markets might actually pass through the eff ect on employment rather than on wages.In order to analyse this issue, in this paper we deal with the eff ects of off shoring on employment in EU27 sample (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009).
Hence, the main objective of this paper is to answer the following research question: does off shoring of tasks from European countries aff ect their employment levels?Moreover, we are interested to test if eff ects of off shoring on employment diff er across distinct groups of workers (divided according to the skill level).In particular, can off shoring be blamed for shrinking labour demand for only these workers whose activities are off shored (typically performing low/medium skill intensive tasks)?
Th e rest of the paper is structured as follows.In Section 2 we provide a brief literature review.Section 3 presents the theoretical background for our empirical analysis.It is based on the data described in Section 4. In Section 5 we try to answer the key question: is there any eff ect of off shoring on employment in EU manufacturing industries?In order to address this issue we estimate an empirical model based on augmented labour demand function.Finally, section 6 concludes.
Our empirical results, based on the estimation of panel data model, suggest that indeed domestic employment can be aff ected by growth in off shoring.More specifi cally, low skill workers are the ones to be aff ected the most because of shrinking labour demand at home.

LITERATURE REVIEW
Th e literature on international outsourcing-labour market interactions started to gain ground in international economic literature in 1980s and 1990s when the researchers noticed the rising importance of international production networks, global disintegration of production process and resulting changes in global structure of trade (an increase of trade in parts and components -its fi rst wave is well documented in Arndt and Kierzkowski, 2001).To give an example, it is documented (Feenstra and Jensen, 2012) that the share of imported inputs in total use of intermediates in the US manufacturing increased from approx.. 6% in 1980 to over 27% in 2009, when intermediate inputs already accounted for two thirds of world trade (Acemoglu et al., 2012, p.1).
Since then many studies and analysis in the fi eld of international economics have dealt with the interaction between the increasing importance of off shoring on employment or earnings (see Acemoglu and Autor, 2011, Crinò, 2009or Görg, 2011 for a review).Off shoring involves moving parts of production abroad, often from more developed countries (called the "North" or "the West") to less developed ones ("the South" or "the East") which aff ects skill composition of labour demand and relative wages (see, among others: Katz and Margo, 2014 Consequently, many papers expressed fears on negative labour market eff ects of such global production sharing, experienced by advanced economies (eg.Feenstra and Hanson, 1996;Deardorff , 2001).Import competition with countries such as China has been perceived as a threat to local labour markets (eg, the US - Autor et al, 2013).Given the focus of our paper, we shall focus here on the evidence concerning the eff ects of off shoring on employment and/or papers using similar sector level data as we do.
On the empirical ground, there seems to be an agreement among scholars that a rise in employment in favor of the high skilled in some developed countries such as U.S. or the U.K. has been the result of an outward shift in relative skilled labour demand (Bound and Johnson, 1992;Katz and Murphy, 1992;Autor et al., 2008).Among the determinants of such a shift, many studies pointed out the role played by so-called skill biased technological change (SBTS, eg.Acemoglu, 2002;Acemoglu et al., 2012).However, other infl uential papers argue that it is indeed trade in intermediate inputs and material off shoring that matters for domestic labour markets (among others: Feenstra andHanson, 1999, 2001).
Studies on negative relationships between trade and domestic employment include (among others): Sachs et al. (1994), Freeman and Katz (1991), Revenga (1992), Bernard, Jensen and Schott (2006) or Federico (2014).Görg (2011) in his summarizing discussion on the eff ects of globalization and off shoring on jobs in developed countries concludes that off shoring may indeed lead to higher job turnover in the short run.However, he argues that in the long run, there is no sign that trade or off shoring leads to higher unemployment (or lower employment) overall, even though employment of low-skilled workers may be aff ected.
Sector level data (mainly coming from EU KLEMS -of which WIOD is a substantial extension2 ) has been used by other researchers, but there is still a wide research gap to fulfi ll.Lo Turco and Parteka (2011), Michaels et al. (2014), Polgár andWörz (2010), Wolszczak-Derlacz (2013) all focused on trade-labour interactions but without looking explicitly into the outsourcing practices.
Among the recent analysis performed with the use of WIOD data, few papers deal explicitly with offshoring -labour market nexus 3 .Schwörer (2013) analyse (mainly in a descriptive manner) off shoring and domestic outsourcing practise only in nine European countries between 1995 and 2008.Additionally he concludes that estimated establishment's productivity gains are driven in particular by gains of multinational fi rms.Foster-McGregor et al. (2013) adopt a fairly standard approach and estimate translog cost function to check the impact of off shoring on the cost shares of workers with diff erent skills.Th ey found that the largest impacts have been observed for medium-skilled workers.Timmer et al. (2013) use WIOD's data to calculate a new concept based on the value added that countries contribute to the production of fi nal manufacturing goods, called 'global value chain (GVC) income.Finally, in the recent study, Parteka and Wolszczak-Derlacz (2015) analyse the impact of off shoring on wages (specifi cally on the wage growth and wage convergence) in manufacturing sectors from the EU27 sample of countries.Interestingly, it is found that even though off shoring is associated with the decline in the growth of wages of less skilled workers, the magnitude of this eff ect is small and economically negligible.Th e set of countries and sectors employed in this paper is the same as in Parteka and Wolszczak-Derlacz (2015) study on off shoring-wage nexus, so the two studies can be treated as complements.

THEORETICAL BACKGROUND
Recent theoretical literature which explains how international fragmentation of production may aff ect labour markets, and the structure of employment in particular, is built upon the so-called 'new-new trade theory' and the Melitz model (see the extensive review on heterogeneous fi rms and trade in: Melitz and Redding, 2014).
As the theoretical background for our analysis we shall use the model of production fragmentation (conceptualized in terms of tradable tasks) proposed within this stream by Grossman andRossi-Hansberg (2008, from now on GRH, 2008).Before their work, theoretical models dealing with off shoring -labour markets nexus emphasized mainly the eff ects of changing labour demand on wages.Consequently, low skilled labour from developed countries had been seen to be a loser from trade integration process, mainly because of the eff ect on trade-induced demand shift and changes in the relative wages of the low skilled and rising wage inequality (Wood, 1995;Autor et al, 2008;Feenstra, 2000;Feenstra and Hanson, 1996).GRH (2008) propose less conventional approach and decompose the impact of off shoring into three eff ects: productivity eff ect, relative price eff ect, and labor supply eff ect.Th e interplay between these eff ects creates the net result which is not necessarily negative for the domestic market (from which some of the tasks are off shored elsewhere in the world).
GRH ( 2008) approach can be summarized as follows.Th ey deal with domestic (home) and foreign (abroad) labour force composed of workers with diff erent skills.Production process is conceptualized in terms of tradable tasks that can be undertaken at home or abroad -mainly in order to take advantage of lower foreign wages.Diff erent tasks can be performed by low skilled workers ("ls tasks") and by high skilled workers ("hs tasks").Off shoring usually involves performing low-skill tasks abroad, as they can be moved abroad easily.However, it is connected with some additional costs.Th e cost of off shoring varies with the nature of tasks (tasks easily off shorable versus tasks which are diffi cult to be off shored), and the marginal task performed at home has to balance the off shoring costs.

3
Most of the papers prepared within WIOD project (www.wiod.org)present a selected theme (eg.supply chain trade, price of services, value added content of trade, energy intensities).
Originally GRH (2008) focus on the eff ect of off shoring on wages.When the technology of off shoring improves, the cost of performing the set of low-skill tasks abroad declines.Consequently, domestic low-skill labour focuses now on tasks characterised by greater productivity (because less demanding low-skill tasks with lower productivity have been off shored).It means that on average the productivity of tasks performed at home increases.What is the eff ect on wages at home?Wages are tied to average productivity, so its rise results in increasing wages of low skilled workers at home.In other words, the model highlights the case where off shoring increases the wage of workers whose jobs are off shored.GRH ( 2008) demonstrate that such an eff ect is plausible, even though it is contrary to the common view that off shoring necessarily pushes down domestic low skilled wages.
Hence, the productivity eff ect can be summarized as follows: productivity rising due to off shoring puts an upward pressure on wages of the low-skilled.However, in GRH ( 2008) framework there is also another parallel eff ect: because of the drop in the costs of production due to off shoring, the relative price of lowskilled labour intensive good should drop, harming the wages of domestic low-skilled workers (relative price eff ect).On top of that, as the domestic demand for some low-skilled workers falls because of more intense off shoring activity, their wages also fall (labor supply eff ect).
Th e net result of off shoring depends on the interplay between these three aforementioned forces.GRH (2008) show that domestic low skilled workers can either loose or gain as a result of off shoring -the net eff ect is ambiguous.Analogous reasoning can be performed for high-skilled workers or any other type of labour.Th e novelty of GRH ( 2008) approach lies in stressing the importance of the productivity eff ect that paradoxically benefi ts these workers whose tasks are easily off shored.Consequently, off shoring of tasks can generate shared gains for all domestic factors, in contrast to the traditional view of the mechanism linked strictly to labour supply eff ect.
Recently, Wright (2014) has extended the GRH ( 2008) wage-focused model to conceptualise the impact of off shoring on employment.He decomposes the demand for labor into three channels.A negative displacement eff ect leads to a direct decline in the demand for employment at home (fi rms move tasks overseas).However, there is also an output eff ect which can aff ect the domestic employment positively -it is connected with productivity gains from off shoring and higher aggregate production.Finally, Wright (2014) describes a substitution eff ect of off shoring: potential substitution between high-skill factor and low-skill factor and/or between domestic and foreign tasks can have ambiguous eff ects on domestic employment.
In order to test this model, Wright (2014) estimates the labour demand function for the set of the US manufacturing industries in the period 2001-2007 and checks explicitly the impact of increasing off shoring to China.He fi nds that as a result of off shoring, the employment of US low-skill workers dropped but at the same time it was compensated by the increase in high-skill workers' employment.

THE DATA AND THE MEASUREMENT OF OFFSHORING
In our work we use sector level data on fl ows of intermediate goods coming from input-output tables, complied within the World Input Output Database project (WIOD -see Timmer et al. 2012 for the details).We match them with socioeconomic accounts data on employment, wages and capital -also from WIOD.Our panel is composed of 13 manufacturing sectors (listed in Table 1) in EU27 countries 4 Due to data availability our analysis covers the years 1995-2009.
Off shoring index (OFF) employed in our research is in the fi rst instance calculated in a traditional way (Feenstra and Hanson, 1999;Hijzen and Swaim, 2007) as the ratio of the value of imported inputs to the size of the industry: where i denotes country from which the tasks are off shored (called also home or domestic country), j and k refer to sectors and t is the time period; mi denotes imported intermediates and VA is the value added.Such an index refl ects so-called 'broad' off shoring (see also Parteka and Wolszczak-Derlacz, 2015 for the application): it relates the sum of inputs imported by industry j from all industries diff erent than j to the added value of industry j.As far as the sources of mi are concerned, we consider imported intermediates from the whole world (so-called global off shoring).Source: own calculation.
Furthermore, following Castellani et al. (2013) we decompose the off shoring index into its international and domestic component: where di is the value of inputs coming from domestic sectors.Th e fi rst expression is the ratio of imported inputs to the domestic ones (IntOUT -international outsourcing) while the second refl ects the intensity of domestic outsourcing (DomOUT).As argued by Castellani et al. (2013), off shoring indices calculated broadly (as in eq. 1) can overestimate the role of imported parts and components, ignoring the role played by structural changes within the domestic economy, refl ected in DomOUT component.
Th e values of off shoring indices for diff erent industries in the analysed EU27 sample (weighted averages) in the fi rst (1995) and last (2009) year of our analysis are reported in Table 1.Off shoring (whether measured by general off shoring index, OFF or by more precise international outsourcing index, IntOUT) rose substantially over the 1995-2009 period in all the analysed industries.Th e highest growth was noted in 'Electrical and optical equipment' (the rise of OFF from 0.6 to 1.13).In case of domestic outsourcing the situation is diff erent: the change in DomOUT is much smaller and three sectors even experienced a drop in domestic outsourcing activity (namely: 'Food, beverages and tobacco', 'Textiles and textile products', 'Leather, leather and footwear products'; note that a drop in DomOUT in these sectors went in line with a growth in IntOUT).Still, in general the intensity of domestic outsourcing is much higher than that characterising international outsourcing activity.

Th e model
Our empirical specifi cation is derived from the theoretical framework described briefl y in Section 3 (for the exact mathematical notations of the theoretical model please refer to GRH, 2008 andWright, 2014).Specifi cally, the model to be estimated with our data is based on augmented labour demand function which can be expressed in logs as 5 : The first step in our empirical analysis is to check for unit roots in the panel.We employ Fisher's panel unit root test which does not require a balanced panel.We run the test for the variables and their differences and the null hypothesis of unit root is rejected in most of the cases at standard levels of confidence.Detail results are available from authors upon request.We thank an anonymous referee for pointing this out.
where, as before, i refers to country, j to sector and t to time period.Employment in domestic sector (Emp, measured in terms of total hours worked by persons engaged in the sector) is our dependent variable.It is a function of: w -real wage (labour compensation per hour in 2005 USD), k -capital per labour ratio and outsourcing (off shoring), measured by: OFF, IntOUT and DomOUT (as defi ned in eq. 1 and eq.2).
We use fi xed-eff ects (within) model with robust standard errors (adjusted for clusters in groups relating to country-sector specifi c eff ects).Th e choice of within estimator is confi rmed by the Hausman test.Additionally, the regression is run using the fi rst-diff erence specifi cation which eliminates invariant countrysector specifi c features (Wright, 2014)

Results
Firstly, the regression ( 5) is run for overall employment in the sector, without distinguishing between diff erent types of labour force (the results are reported in Table 2).Th en, we repeat the estimations separately for three diff erent types of workers: high, low and medium-skill (the results are presented in Table 3) 6 .Th is gives us the possibility to distinguish between heterogeneous (across various skill groups of workers) responses of employment to off shoring (outsourcing) activity.
In all specifi cations we obtain negative and statistically signifi cant parameter associated with wages (w): ceteris paribus, the higher the change in wage, the lower the change in labour demand.Additionally, sectors with higher growth of capital/labour ratio (k) are on average characterised by lower employment growth (note negative and statistically signifi cant parameter on k) which may suggest the replacement of workers by the technology (so called skill-biased technological change - Acemoglu, 2002).
We are especially interested in the relationship between off shoring indices and employment.In the general case (Table 2), all parameters connected with production fragmentation, either measured by general offshoring index (column 1 in Table 2) or by international (IntOUT) and domesting outsourcing (DomOUT) (column 2 in Table 2), are negatively associated with overall employment in the sector.Th is might suggest that, indeed, outsourcing of tasks from a given sector, either domestically or abroad, can reduce industry employment.Such an outcome would be in line with the results of (for instance) Federico (2014).
However, when the estimations are performed separately for the subgroups of workers divided by skills (Table 3), there are some noteworthy changes.Off shoring (OFF) aff ects negatively only employment of medium and low skill workers (columns 3 and 5 in Table 3).Th e negative and statistically signifi cant eff ect of international outsourcing (IntOUT) on employment is sustained only in case of low-skill labour (column 6 in Table 3).Such a result is in line with the common view that the low skilled are the ones to loose from global production sharing.In case of high-skill workers, we are not able to detect statistically signifi cant correlation between off shoring measures and employment.

6
The division into skill groups comes from WIOD and is based on the education level of workers according to the International Standard Classification of Education (ISCED).Low skilled workers are those with: primary education, first stage of basic education or lower secondary or second stage of basic education; medium-skilled: (upper) secondary or post-secondary education; high-skilled: first or second level of tertiary education.

Table 1
Off shoring indices -manufacturing industries, EU27 countries, 1995 and 2009 OFF, IntOUT and DomOUT explained in the main text.Weighted averages across the EU27 countries, weighted according to the sector size (total hours worked by persons engaged). Notes: while year dummies ( )