Outsourcing and Technological Innovations: A Firm-Level Analysis

Transcription

1 DISCUSSION PAPER SERIES IZA DP No Outsourcing and Technological Innovations: A Firm-Level Analysis Ann Bartel Saul Lach Nachum Sicherman February 2008 Forschungsinstitut zur Zuunft der Arbeit Institute for the Study of Labor

2 Outsourcing and Technological Innovations: A Firm-Level Analysis Ann Bartel Columbia University and NBER Saul Lach Hebrew University and CEPR Nachum Sicherman Columbia University and IZA Discussion Paer No February 2008 IZA P.O. Box Bonn Germany Phone: Fax: iza@iza.org Any oinions exressed here are those of the author(s) and not those of IZA. Research ublished in this series may include views on olicy, but the institute itself taes no institutional olicy ositions. The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center and a lace of communication between science, olitics and business. IZA is an indeendent nonrofit organization suorted by Deutsche Post World Net. The center is associated with the University of Bonn and offers a stimulating research environment through its international networ, worshos and conferences, data service, roject suort, research visits and doctoral rogram. IZA engages in (i) original and internationally cometitive research in all fields of labor economics, (ii) develoment of olicy concets, and (iii) dissemination of research results and concets to the interested ublic. IZA Discussion Paers often reresent reliminary wor and are circulated to encourage discussion. Citation of such a aer should account for its rovisional character. A revised version may be available directly from the author.

3 IZA Discussion Paer No February 2008 ABSTRACT Outsourcing and Technological Innovations: A Firm-Level Analysis * This aer resents a dynamic model that analyzes how firms exectations with regards to technological change influence the demand for outsourcing. We show that outsourcing becomes more beneficial to the firm when technology is changing raidly. As the ace of innovations in roduction technology increases, the firm has less time to amortize the sun costs associated with urchasing the new technologies. This maes roducing in-house with the latest technologies relatively more exensive than outsourcing. The model therefore rovides an exlanation for the recent increases in outsourcing that have taen lace in an environment of increased exectations for technological change. We test the redictions of the model using a anel dataset on Sanish firms for the eriod 990 through The emirical results suort the main rediction of the model, namely, that all other things equal, the demand for outsourcing increases with the robability of technological change. JEL Classification: O33, L24, L, J2 Keywords: technological change, outsourcing Corresonding author: Nachum Sicherman Graduate School of Business Columbia University New Yor, NY 0027 USA Nachum.Sicherman@Columbia.edu * The authors gratefully acnowledge the generous suort of grants from the Columbia University Institute for Social and Economic Research and Policy and the Columbia Business School s Center for International Business Education and Research and outstanding research assistance from Ricardo Correa and Cecilia Machado. Saul Lach acnowledges the suort of the Euroean Commission grant CIT5-CT

4 I. Introduction Outsourcing, or the contracting out of activities to subcontractors outside the firm, has become more widesread in recent decades. For examle, Magnani (2006) reorts that the cost share of urchased services in U.S. manufacturing industries grew from 4.4% in 949 to 2% in 998. Similar increases in outsourcing have been observed in Euroe as well. A large theoretical literature has examined the determinants of the decision to outsource, or, more generally, the organization of roduction. Outsourcing arts of the roduction of a final roduct is costly because of the costs associated with searching and finding an aroriate inut sulier and, more imortantly, because contracting with outside suliers may be imerfect if some attributes of the outsourced inuts are not verifiable by third arties. This limits contracting ossibilities and creates a otential hold-u roblem. 2 These ideas go bac to Williamson (975, 985) and Grossman and Hart (986) and have been recently incororated into industry equilibrium models by Grossman and Helman (2002) while their imlications for international trade are analyzed in Antras and Helman (2004). In this aer we resent a dynamic model that analyzes how firms exectations with regards to technological change influences the demand for outsourcing an issue not addressed in revious models while abstracting from other considerations (e.g., transaction costs, secificity, etc.). We show that outsourcing becomes more beneficial to the firm when technology is changing raidly. A firm can buy the latest technology and See Arndt and Kierzowsi (200), Mol (2005) and Abramovsy and Griffith (2006). 2 Baer and Hubbard (2003) consider how information technology in the trucing industry imacts contracting ossibilities and vertical integration. Baccara (2007) uses a general equilibrium model to study how information leaages could affect a firm s outsourcing decision as well as its investments in R&D. 2

5 roduce intermediate inuts in-house. Firms incur a sun cost when adoting new technologies. Outsourcing, on the other hand, enables the firm to urchase inuts from sulying firms using the latest roduction technology while avoiding the new technology sun costs. The intuition behind the model is that as the ace at which innovations in roduction technology arrive increases, the less time the firm has to amortize the sun costs associated with urchasing the new technologies. This maes roducing in-house with the latest technologies relatively more exensive than outsourcing, The model, therefore, rovides an exlanation for the recent increases in outsourcing that have taen lace in an environment of increased exectations for technological change. We test the redictions of the model using a anel dataset on Sanish firms for the eriod 990 through This dataset is articularly useful for our uroses because, in each survey year, the firms were ased whether they contracted with third arties for the manufacturing of custom-made finished roducts or arts. In addition, the dataset contains information on the firm s R&D exenditures, an ideal roxy for exected technological change. Our econometric analysis controls for unobservable fixed characteristics of the firms and also uses instrumental variables techniques to deal with the otential endogeneity of our technological change measure. The emirical results suort the main rediction of the theoretical model, namely, that all other things equal, the demand for outsourcing increases with the robability of technological change. Ours is the first aer to rovide both a theoretical and emirical analysis of the imact of technological change on outsourcing using firm-level anel data. Magnani (2006), using industry-level data for the U.S., found that outsourcing was facilitated by 3

6 technological diffusion, while Abramovsy and Griffith (2006) found that information and communications technology-intensive firms in the U.K. were more liely to urchase a greater amount of services on the maret. Other research has found evidence that firms engage in outsourcing in resonse to unredictable variations in demand (Abraham and Taylor, 996), to tae advantage of the secialized nowledge of suliers (Abraham and Taylor, 996), and to save on labor costs (Autor, 2003; Diaz-Mora, 2005; and Girma and Gorg, 2004). 3 Part II resents a dynamic model of the relationshi between outsourcing and exected technological change. Part III discusses the data and emirical secifications used to test the redictions of the model. Results are resented in Part IV. Part V concludes. II. A Model of the Demand for Outsourcing A firm roduces a rofit-maximizing amount of outut y using a single inut x. The model focuses on determining how to rocure a given level of inut x. There are two ways of obtaining x. One way is to roduce x in-house using machines each with roductivity according to the in-house roduction function x = () The cost of roducing in-house is C H = x+ s 3 Other researchers (Ono (2007) and Holl (2007)) have studied the effect of agglomeration economies on outsourcing. For emirical studies of the imacts of outsourcing on wages and roductivity, see Feenstra and Hanson (999), Amiti and Wei (2006), Gorg, Hanley and Strobl (2007), and Gorg and Hanley (2007). 4

7 where is the (rental) rice of a machine and s is a sun cost associated with installing and using the machines of a given roductivity (or vintage) for the first time. s is incurred only once. This initial installation cost also reflects training costs. The second way of rocuring x is to buy it in the maret. We call this outsourcing roduction. The cost of this alternative is CO = x where is the unit rice of x. We assume that outsourcing costs are the same across firms, but allow for heterogeneity in the sun cost of in-house roduction by assuming that s is distributed among firms with distribution G(s). Assuming constant returns to scale in roduction we can, without loss of generality, set x = y and write the costs as functions of y (which is observable) C C H O = y = y+ s (2) As in Ono (2000), the firm chooses the rocuring alternative that minimizes costs. Let χ = indicate a firm that outsources its roduction. 4 This simle model imlies 4 The constant returns to scale assumtion imlies that the firm either roduces all y in-house or outsources all of y, therefore it is not otimal to slit a given level of y between in-house and outsourcing. The cost of slitting y between in-house ( y H ) and outsourcing ( O ) is y y ( ) y s, O + + where y = yo + y. H We can write this cost as y ( ) yh + s. It follows that roducing all of y inhouse minimizes costs when >, i.e., when outsourcing is exensive, and, conversely, outsourcing H 5

8 ( ) if CO CH s > y χ = 0 else (3) Suose the rice of x is above the marginal cost of in-house roduction. 5 Firms would then outsource only because the sun cost of in-house roduction is relatively large; firms with low sun costs will not outsource roduction. A larger roduction size decreases the lielihood of outsourcing because the sun cost er unit of roduction decreases. Notice that if we do not allow for heterogeneity in s then all firms of a given size would be either outsourcing or roducing in-house which is in general contrary to the facts. 6 Introducing Technological Change We extend this simle model of demand for outsourcing to allow for technological change in the roduction of x. Our goal is to examine the relationshi between technological change and outsourcing. Since technological change is uncertain and occurs over time, we now introduce dynamics and uncertainty into the model. We focus, for simlicity, on a two-eriod model. < all of y is referred when. Only when = is the firm indifferent between in-house and outsourcing all or arts of its roduction. In reality, however, firms usually outsource art of their roduction, so that we should interret x as one of multile comonent of the final outut roduced by the firm. 5 If the suliers of x have a cost advantage in roducing x and there is cometition among suliers the rice of x could be below the marginal cost of in-house roduction. In this case, all firms will outsource x. 6 We could have also added heterogeneity to the cost of outsourcing. The model would be equivalent because heterogeneity enters additively and what matters for the decision to outsource is the difference between C and C. In this case s may reresent the additional cost of using a standardized inut which O H is absent when x is roduced in-house because then the firm can erfectly tailor the inut to its secific needs. 6

9 Suose that technology (i.e., the roductivity of machines) in the first eriod is given by. In the second eriod, roductivity can either increase to 2, 2 >, with robability λ or remain at with the comlementary robability, i.e., 2 with robability λ = with robability λ At the beginning of each eriod, the firm decides to outsource or to roduce inhouse given the observed technology level and rices. We denote the rice of x in eriod by. We assume that x is sulied by firms using the latest technologies. As roduction technology imroves, firms can charge a lower rice of x and still mae rofits. Whether they will do this or not deends on the cometitive environment. We do not model the suly side here but allow the rice of x to change as technology imroves. We let 2 be the rice of x when a technological change occurs in the second eriod, and 2 be the rice of x when technology does not change. Because the rice of machines does not change, their quality-adjusted rice imroves. declines as technology We roceed bacwards. Suose there is technological change in the second eriod. The firm faces three alternatives: it can outsource, it can roduce in-house with the old technology or it can ay the sun cost and ugrade to the new vintage of machines. To simlify the analysis we mae the assumtion that ugrading always dominates eeing the old technology. This requires a not too large sun cost, namely 2 ( ) s y, which we assume to hold, i.e., 2 7

10 2 G( s) = for s y (4) 2 In this case, the firm s decision is essentially as in the static model analyzed in the revious section: to outsource or to roduce in-house. Because roducing in-house requires incurring the sun cost, the firm s decision in the first eriod does not affect its current decision. Then, as in (3), when there is a technological imrovement, the demand for outsourcing is ( 2λ ) if s y 2 χ2 = 0 else (5) If there is no technological change, the firm s decision deends on its outsourcing decision in the first eriod. If the firm roduced in-house in the first eriod, it already aid the sun cost s for the use of the technology and therefore it will outsource only if y 2 y. > But, if the firm outsourced in the first eriod, the cost of using the old technology in-house is y+ s which needs to be comared to the cost of outsourcing, 2 y. This gives the following outsourcing decision in eriod 2 in the absence of technological change, ( ) if χ s 2 y χ2 = 0 else (6) Because of the resence of sun costs, the decision to outsource during the first eriod affects future decisions if technological change does not occur. Secifically, a firm 8

11 that did not outsource during the first eriod, χ = 0, will be less liely to outsource during the second eriod if no technological change occurs. In the first eriod, the firm chooses to outsource or roduce in-house by comaring the exected discounted costs of each alternative. These costs are, C( χ = ) = y+ β( λ) Min 2y, y+ s + βλmin 2λ y, y+ s 2 C( χ = 0) = y+ s+ β( λ) Min 2y, y + βλmin 2λ y, y + s 2 where β is the discount factor. The difference between these two costs is C( χ = ) C( χ = 0) = y( β( λ) ) s+ β( λ) Min 2 y s, 0 β( λ) Min 2 y, 0 which can be written as, ( ) ( 2 ) ys y 0 C( χ = ) C( χ = 0) = y+ β( λ) ys, 0 y s (7) ( ) ( ) ( ) 2 2 ( ) ( β( λ) ), ( 2 ) y s y s We can rule out the first case because it is reasonable to assume that the rice of x in eriod 2 (the last eriod) cannot be below the marginal cost of in-house roduction, 9

12 i.e., 7 2 (8) The firm decides to outsource in the first eriod whenever C( χ = ) C( χ = 0) < 0. The decision to outsource in the first eriod therefore deends on current and future rices of x as well as on the size of sun costs and the robability of technological change, ( ) β λ ( 2 ) ( 2 ) if s y+ ( ) y and s y χ ( ) = if s and s y 0 else y ( ) ( 2 ) β λ (9) From (7) we see that, all other things equal, an increase in λ decreases the cost of outsourcing relative to that of in-house roduction maing the firm more liely to decide to outsource in the first eriod. This is the main message of the model: when the lielihood of technological change increases, firms will be more reluctant to urchase the technology to roduce in-house because it will soon be obsolete. Ugrading the technology involves incurring a sun cost di novo. The more frequent the innovations arrive, the less time the firm has to amortize these sun costs. The firm can use outsourcing to obtain x from sulying firms using the latest technology and avoid the sun costs. 7 This assumtion deends on the tye of technology and cometition among the firms roducing x. 0

13 Emirical Imlications We use this simle model of the demand for outsourcing to generate a reducedform equation for the demand for outsourcing in any eriod. There are three different threshold values in (9) that determine the demand for outsourcing in eriod : y y + 2 ) ( λ β, y 2 and ) ( λ β y. Because demand for outsourcing deends on s being above or below these thresholds, their relative magnitudes are imortant. Among the different ways these three thresholds can be raned, only two configurations are feasible, namely, ) ( ) ( 2 2 λ β λ β + < > < > which can we written more comactly as ) ( 2 λ β < > Thus, ( ) ( ) 2 ( ) β λ > < defines two rice regions which determine demand for outsourcing in the first eriod,

14 ( ) ( 2 ) ( ) y y if s β( λ) when y β( λ) χ = if s ( ) y+ β( λ) ( ) y when ( ) y 0 else ( ) y 2 2 β ( λ ) which imlies ( ) ( 2 ) ( ) G β( λ) y when β( λ) E( χ) = G( ( ) y+ β( λ) ( ) y) when ( ) ( ) 2 2 β ( λ ) Notice that, for given rices, G ( ) increases with λ so that the demand for outsourcing in eriod is increasing in the robability of technological change λ. The demand for outsourcing in eriod 2 is given by ( χ2) = λ ( χ2 2 = ) + ( λ) E( χ T = 0, χ = ) E( χ ) + E( χ T = 0, χ = 0)( E( χ )) E E T where T 2 is an indicator for technological change occurring in eriod 2. In this simle model, E( χ2 T2 = 0, χ = 0) = 0 because a firm that did not outsource in the first eriod will continue roducing in-house in the second eriod if does not change. Exected outsourcing demand in the second eriod is therefore E( χ2) = λ G 2λ y + ( λ) G 2 y E( ) 2 χ (0) 2

15 Given rices, E( χ 2) is an increasing function of λ when the rice margin after an innovation occurs, λ is no larger than the rice margin when there is no 2, 2 technology change, This sufficient condition is liely to hold under many 2. maret structures because the number of otential outsourcers is larger after an innovation occurs romting sulying firms to lower their marus in order to cature a larger maret share. III. Data and Emirical Secification We use the Encuesta sobre Estrategias Emresariales (ESEE, or Survey on Business Strategies), a anel of aroximately 800 Sanish manufacturing comanies, surveyed annually since 990. Data are currently available through The survey is conducted by the Fundacion SEPI with the suort of the Ministry of Industry, Tourism and Trade. We use data from the 990, 994, 998 and 2002 surveys. 8 In each survey, the firms were ased if they contracted with third arties for the manufacture of custom-made finished roducts or arts, and if so, the value of the outsourced roducts or arts. We use this information to create two indicators of outsourcing, a dummy for whether or not the firm did engage in outsourcing, and the value of outsourcing divided by total costs. 9 Table shows, by industry sector and overall, the ercentage of firms that reorted outsourcing at least some art of roduction during the time eriod and the mean value of the outsourced roduction as a 8 Budgetary constraints revented us from urchasing data for all of the years between 990 and Loez (2002) describes the evolution of outsourcing of services and of roduction in Sanish manufacturing firms using a sub-samle of the ESSE data for the eriod He finds significant differences in outsourcing between small and large firms and a ositive effect of outsourcing on roductivity. 3

16 ercentage of total cost. On average, 4% of firms reorted that they outsourced roduction during this time eriod. The outsourcing ercentage rose from 35% in 990 to 43% in There is significant variation in the lielihood of outsourcing across industries ranging from a low of 6% for Drins to a high of 6% for Machinery and Mechanical Goods. The value of the outsourced roduction as a ercentage of total costs is aroximately 5 ercent during this time eriod; for firms that did outsource roduction, the mean value of outsourced roduction as a ercentage of total costs is ercent. In the revious section, we derived the demand for outsourcing in eriod t as a function of the robability of technological change, outut, and the various inut rices and technology arameters. Notice that the demand function differs between eriods and 2 because of the dynamic nature of the roblem and the finiteness of the model. We will ignore this issue in the emirical alication because we imlicitly assume that the data-generating rocess has been going on for some time so that the relationshi between outsourcing and its drivers stabilizes over time. Under the assumtion that inut rices and technology arameters are common to all firms in an industry, these can be catured by industry dummies (ID). Thus, E( χt) = h( λ, yt, ID t ) () Equation () is a reduced-form exression because all the arguments in the h function are exogenous in the model. 0 The main imlication of the theoretical model is that, all 0 Equation (0) is a dynamic equation that could also serve as the basis for estimating the relationshi between outsourcing and λ. It is, however, secific to a 2-eriod model. In a model with 3 or more eriods the dynamics are much more comlicated. We therefore refer to estimate reduced form equations. 4

17 other things equal, the demand for outsourcing increases with the exogenous robability of technological change λ. In order to test this imlication we need a roxy for λ. We will roxy λ by an binary variable indicating whether a firm engages in R&D, and if it does, we will also use the amount of R&D exenditures. The rationale for this roxy is that a firm that engages in R&D is more liely to exerience technological change (new rocesses and/or new roducts) as comared to a firm that does not engage in R&D. Table shows that during the time eriod under study, 36 ercent of the firms in the samle engaged in R&D, with this ercentage ranging from a low of 3% for Editing and Printing to 67% for Chemicals. We mae searability and linearity assumtions on the function h, χ = β r + β y + δ ID + u (2) it r it y it j j ij it where r it is our measure of R&D (a dummy for being engaged in R&D and/or the ratio of R&D exenditures to sales) and ID ij = if firm i is in industry j and zero otherwise. By using the two measures of r it we are able to study whether the extensive or the intensive margin of R&D is more imortant in exlaining outsourcing. We will estimate this equation for all years using clustered standard errors at the firm level to account for heterosedasticity and arbitrary serial correlation. Although the model assumes λ to be exogenous, a serious concern is that the robability that the firm will exerience technological change is determined by unobserved factors that also affect the decision to outsource. More innovative firms may 5

18 be doing more R&D and may also be adoting new roduction methods that require more outsourcing, i.e., λ, or its R&D roxy, may be endogenous in the outsourcing equation. The estimated R&D coefficient will therefore not cature the causal effect of technological change on outsourcing. We address this concern by allowing for a timeinvariant comonent in u it to affect both the decision to engage in R&D (and therefore λ) and the decision to outsource, i.e., u it = µ i + ηit. Consistent estimates under this assumtion can be obtained from time-differencing equation (2) ( rit rit ) + β y ( yit yit ) + ηit ηit χ χ = β (3) it it r and estimating this equation by OLS eriod by eriod. These estimates are consistent rovided R&D is strictly exogenous. The strict exogeneity assumtion is quite strong because it recludes any correlation between shocs that affected outsourcing in revious eriods and current R&D exenditures. We therefore mae a weaer assumtion that allows for outsourcing to affect future R&D exenditures, E( η r, r,...) = 0 (4) it it it Under this assumtion, we can use the lag of R&D as an instrument for r it r it in equation (3) to obtain consistent estimates of the arameters. Equation (2) includes, in addition to the roxy for technological change, the firm s annual sales (Sales) as well as a vector of industry dummies. We also add two variables that measure the firm s current technological intensity, whether it uses 6

19 comuterized digital machine tools (com_dmt) or uses robotics (com_robotic). In order to control for factors other than technological change that may contribute to outsourcing, we also include a set of variables that have been the focus of revious research on the determinants of outsourcing. Since firms may use outsourcing as a way of economizing on labor costs (see Abraham and Taylor, 996), we include the firm s average labor cost defined as total annual sending (wages and benefits) on staff divided by total emloyment (wage). Outsourcing may also be used to smooth the worload of the core worforce during eas of demand (Abraham and Taylor, 996; Holl, 2007). Hence, we add a measure of caacity utilization (caacity) defined as the average ercentage of the standard roduction caacity used during the year. Small firms would be exected to be more liely to outsource because it may not be otimal for them to carry out all stes in the roduction rocess because of the costs of maintaining secialized equiment or sills in-house (Abraham and Taylor, 996). Hence we control for the size of the firm using four categories for number of emloyees. Another factor that can increase the roensity to outsource is the volatility in demand for the roduct (Abraham and Taylor, 996; Holl, 2007). We control for this by adding dummy variables for whether the main maret the comany serves exanded (maret_exand) or declined (maret_decline) during the year. It has also been argued that older firms are more liely to outsource because they have had time to learn about the quality and reliability of otential subcontractors (Holl, 2007). We therefore include the age of the firm (age). Whether the firm rimarily roduces standardized roducts or custom roducts should also affect the roensity to outsource with firms focusing on roduct variety being more These two variables may also be correlated with technological change and their inclusion in the equation 7

20 liely to outsource some of their roduction; we therefore add a dummy variable indicating that the firm roduces standardized roducts that are, in most cases, the same for all buyers (std_roduct). Finally, we control for the firm s exort roensity, the value of its exorts divided by its sales (exort), and whether the firm has any foreign ownershi (foreign_own). Both of these factors have been included in rior studies of outsourcing (Girma and Gorg, 2004; Diaz-Mora, 2005; and Holl, 2007). IV. Results We estimated Equation (2) using two deendent variables: whether or not the firm engaged in outsourcing and the value of the firm s outsourced roduction as a ercentage of total costs. Results for the first deendent variable are shown in Table 2 and for the second in Table 3. Each table rovides the coefficients on the R&D dummy using various aroaches to estimating Equation (2). The coefficients on the other variables included in these regressions are shown in Aendix Tables A- and A-2. Columns (), (2) and (3) in Table 2 use OLS, robit and logit, resectively, to estimate Equation (2). We find remarably similar results across these three columns; firms that engage in R&D are -2 ercent more liely to outsource some art of their roduction. When we control for firm-secific time-invariant unobservable factors in Column (4), firms that engage in R&D are 9.5% more liely to outsource roduction. Using first-differences in column (5), the coefficient on R&D falls to but is still very significant. Column (6) adds the R&D/sales ratio and we find that what matters is whether the firm engages in R&D, not its R&D-intensity. Finally, in column (7), we use could weaen the measured effect of the R&D variable. 8

21 the first lag of R&D as an instrument for the growth rate of R&D. 2 Although the coefficient on R&D is no longer significant because of the doubling of the standard error, the magnitude of the coefficient is virtually identical to the coefficient reorted in column (5). The results in Aendix Table A- show that, in the cross-section, high wage firms, older firms, larger firms, firms that are in a roduct maret that has exerienced an increase or a decrease (as comared to no change), firms that use comuterized digital machine tools or robotics, are all more liely to outsource. Foreign-owned firms are less liely to outsource. These results, however, change dramatically when we use fixed effects or first differences: none of these additional regressors are significant while the R&D variable remains significant. 3 In Table 3, the deendent variable is the value of the outsourced roduction divided by total costs. Since 60 ercent of the observations are zeroes, we use tobit regressions. The results in the first column indicate that the robability of outsourcing is 9 ercent higher for firms that engage in R&D; this effect is very close to the results observed in columns (), (2) and (3) in Table 2. Column () in Table 3 also shows that, for firms that do outsource, the ratio of outsourcing to total costs is 0.03 higher for firms that engage in R&D. Using the mean value of the deendent variable for firms that do outsource (see Table ), the coefficient translates to an ercent increase in outsourcing. The regressions in Table 3 were also estimated using standard random 2 We also tried using the first lag of the growth rate in R&D as an instrument and obtained very similar results but less recise -- to those reorted in column (7). 3 The excetion is the exort roensity variable which is ositive and significant in the fixed effects and first differences secifications, but not significant in the first differenced regression with IV. 9

22 effects tobit, a method of random effects tobit that controls for unobserved effects 4, and instrumental variables tobit. Although the imact of R&D on outsourcing is smaller using random effects, the coefficients are still significant. The results using IV tobit are very strong; the coefficients on R&D are significant for both the extensive and the intensive margins of outsourcing. V. Conclusions A large theoretical literature has examined the determinants of the decision to outsource, or, more generally, the organization of roduction. But revious models have ignored the role layed by technological change. In this aer we resent a model that fills this void. We show that outsourcing becomes more beneficial to the firm when technology is changing raidly. The intuition behind the model is that the more frequently innovations in technology arrive, the less time the firm has to amortize the sun costs associated with an obsolete technology. Outsourcing enables the firm to urchase from sulying firms that are using the latest technology and avoid the sun costs. Our model, therefore, lins technological change and outsourcing and exlains why outsourcing has been increasing in recent decades when the ace of technological change has accelerated. We test the redictions of the model using a anel dataset on Sanish firms for the time eriod 990 through Our econometric analysis controls for unobservable fixed characteristics of the firms and also uses instrumental variables techniques to deal with the otential endogeneity of our technological change measure. The emirical results 4 We follow the Chamberlain-lie aroach suggested in section 8.2 of Chater 6 in Wooldridge (2002) and include the means (over time) of all the regressors to account for the correlation between the regressors 20

23 indicate that firms doing R&D are between 6 and 0 ercent more liely to outsource than firms not engaged in R&D. This is consistent with the main rediction of the theoretical model that the demand for outsourcing increases with the robability of technological change. Interestingly, while the existing literature has found evidence that other variables lay a role in the decision to outsource, we find no such evidence here when accounting for firm effects. and the unobserved firm effects. 2

24 References Abraham, K. and S. Taylor (996). Firms Use of Outside Contractors: Theory and Evidence. Journal of Labor Economics 4: Abramovsy, L. and R. Griffith (2006). Outsourcing and Offshoring of Business Services: How Imortant is ICT? Journal of the Euroean Economic Association 4: Amiti, M. and S.J. Wei (2006). Service Offshoring and Productivity: Evidence from the United States, NBER Woring Paer 926. Antras, P. and E. Helman (2004). Global Sourcing, Journal of Political Economy, 2: Arndt, S.W. and H. Kierzowsi (200). Fragmentation: New Production Patterns in the World Economy (New Yor: Oxford University Press). Autor, D. (2002). Outsourcing at Will: Unjust Dismissal Doctrine and the Growth of Temorary Hel Emloyment. Journal of Labor Economics Baccara, M. (2007). Outsourcing, Information Leaage and Consulting Firms, The Rand Journal of Economics, 38: Baer, G. and Hubbard, T. (2003): Mae versus Buy in Trucing: Asset Ownershi, Job Design and Information, American Economic Review, June: Diaz-Mora, C. (2005): Determinants of Outsourcing Production: A Dynamic Panel Data Aroach for Manufacturing Industries. Foundation for Alied Economics Studies woring aer. Feenstra, R.C. and G.H. Hanson (999). The Imact of Outsourcing and High- Technology Caital on Wages: Estimates for the United States, , Quarterly Journal of Economics 4: Girma and Gorg (2004). Outsourcing, Foreign Ownershi, and Productivity: Evidence from UK Establishment-level Data, Review of International Economics 2: Gorg, H., A. Hanley and E. Strobl (2007). Productivity Effects of International Outsourcing: Evidence from Plant-level Data, Canadian Journal of Economics. Gorg, H. and A. Hanley (2007). International Services Outsourcing and Innovation: An Emirical Investigation. woring aer. Grossman, S. and O. Hart (986). The Costs and Benefits of Ownershi: A Theory of Vertical and Lateral Integration, Journal of Political Economy 94(4): Grossman, S. and E. Helman (2002). Integration versus Outsourcing in Industry Equilibrium, Quarterly Journal of Economics 7:

25 Holl, A. (2007). Production Subcontracting and Location, Foundation for Alied Economics Studies woring aer. Loez, A. (2002). Subcontratacion de servicios y roduccion: evidencia ar alas emresas manufactureras esanolas, Econmia Indsutrial, 348: Magnani, E (2006). Technological Diffusion, the Diffusion of Sill and the Growth of Outsourcing in US Manufacturing, Economics of Innovation and New Technology Mol, M. (2005). Does Being R&D-Intensive Still Discourage Outsourcing? Evidence from Dutch Manufacturing, Research Policy 34: Ono, Y. (2007). Outsourcing Business Services and the Scoe of Local Marets, Regional Science and Urban Economics 37: Williamson, O. (975). Marets and Hierarchies: Analysis and Antitrust Imlications, New Yor: The Free Press. Williamson, O. (985). The Economic Institutions of Caitalism: Firms, Marets, Relational Contracting, New Yor: The Free Press. Wooldridge, J. (2002). Econometric Analysis of Cross Section and Panel Data, Cambridge: MIT Press. 23

26 Table Mean Values of Outsourcing and R&D, By Industry Sector ( ) Industry Proortion of Firms Value of Outsourcing Proortion of Firms Outsourcing Production Divided by Total Costs Engaging in R&D All If >0 Meat-rocessing industry Foodstuffs and tobacco Drins Textiles Leather and footware Wood industry Paer Editing and Printing Chemicals Rubber and lastics Non-metallic minerals roducts Iron and steel Metallic Products Machinery and mechanical goods Office machinery, comuters, rocessing, Electrical and electronic machinery Motor vehicles Other transort material Furniture Other manufacturing industries All Industries ( ) (N=289) (N=876) (N=776) (N=708)

27 Table 2 The Imact of R&D on the Lielihood of Outsourcing Production, a Indeendent Variable () OLS (2) Probit (3) Logit (4) Logit (with fixed effects) *** (5) First Differences (6) First Differences (7) First Differences and IV R&D Dummy b 0.27*** (0.06) 0.203*** (0.07) 0.25*** (0.08) (0.033) *** (0.023) *** (0.025) (0.04) R&D/Sales (0.5979) Observations R 2 /Pseudo R 2 LR(chi 2 ) a Deendent Variable: dummy variable equals one if firm outsourced some of its roduction; zero otherwise. All regressions include year dummies, industry dummies, and the set of control variables listed in the text. The comlete regression results are given in the Aendix. Huber-White robust standard errors are in arentheses. b Coefficients shown in columns (2) through (7) are the marginal effects calculated at the mean. * Significant at 0%. ** Significant at 5%; ***Significant at %. 25

28 Table 3 The Imact of R&D on Share of Outsourcing Exenditures in Total Costs, a Random Effect Tobit Tobit Standard Unobserved Effects b IV c Tobit Marginal effects for the robability of outsourcing 0.092*** (0.06) 0.048*** (0.0) ** (0.05) *** (0.042) Marginal effects for the exected value of outsourcing exenditures divided by costs, conditional on outsourcing being ositive *** (0.002) *** (0.002) * (0.002) 0.040** (0.006) Observations Number of grous a Deendent Variable: Value of outsourcing divided by total costs. All regressions include year dummies, industry dummies and the set of control variables described in the text. The comlete regression results are shown in the Aendix. Huber-White robust standard errors are in arentheses. b See Wooldridge (2002),. 540 for a discussion of unobserved effects Tobit models. This regression includes means over time of indeendent variables. c The instrument used for the R&D dummy is its one eriod lag. * Significant at 0%. ** Significant at 5%; ***Significant at %. 26

29 Aendix Table A- Deendent Variable: The lielihood of outsourcing Full Regression Estimates of Regressions (3), (4), (5) and (7) in Table 2 Logit Logit Fixed Effect First Dif. First Diff. & IV variable dy/dx Std. Err. dy/dx Std. Err. coef. Std. Err. coef. Std. Err. RandD dummy 0.25*** *** *** Year dummy *** *** *** *** *** *** *** sales roduct_standard foreign_own ** exort_roensity * * If using comuter digital machine tools * If using robotic * age *** age * Emloyees: *** *** *** wage 0.003*** maret_exand *** maret_decline * caacity Industry Dummies (omitted: Meatrocessing) Foodstuffs and tobacco Drins Textiles *** Leather and footware *** Wood industry 0.209*** Paer Editing and Printing 0.487*** Chemical industry 0.520** Rubber and lastics 0.305*** Non-metallic minerals roducts Iron and steel Metallic Products *** Machinery and mechanical goods *** Office machinery, comuters, rocessing, *** Eletrical and electronic machinery and m *** Motor vehicals *** Other transort material *** 0.06 Furniture *** Other manufacturing industries *** * Significant at 0%. **Significant at 5%; ***Significant at % 27

30 Aendix Table A-2 Deendent Variable: Share of Outsourcing Exenditures in Total Costs Full Regression Results for Table 3 Random Effect tobit tobit standard unobserved effects IV tobit Variable coeff. Std. Err. coeff. Std. Err. coeff. Std. Err. coeff. Std. Err. RandD dummy *** Year dummy *** ** *** *** * *** sales ** *** *** roduct_standard foreign_own *** ** *** exort_roensity * ** If using comuter digital machine tools If using robotic ** age age Emloyees: * * wage 0.004*** *** *** maret_exand 0.024*** *** * ** maret_decline caacity * ** ** Industry Dummies (omitted: Meat-rocessing) Foodstuffs and tobacco * Drins Textiles 0.795*** *** *** *** Leather and footware 0.387*** *** *** *** Wood industry 0.006*** ** ** ** Paer * ** Editing and Printing *** *** *** *** Chemical industry ** * Rubber and lastics 0.239*** ** ** *** Non-metallic minerals roducts * Iron and steel Metallic Products 0.489*** *** *** *** Machinery and mechanical goods *** *** *** *** Office machinery, comuters, rocessing, 0.530*** ** ** *** Eletrical and electronic machinery and m 0.540*** *** *** *** Motor vehicals 0.355*** *** *** *** Other transort material 0.905*** *** *** *** Furniture 0.274*** *** *** *** Other manufacturing industries 0.583*** *** *** *** Mean values of: RandD sales ** roduct_standard 0.003*

31 foreign_own exort_roensity ** If using comuter digital machine tools If using robotic age age emloyment wage maret_exand maret_decline caacity constant *** ** * *** * Significant at 0%. **Significant at 5%; ***Significant at % 29