The Short and Long-Run Financial Impact of Corporate Outsourcing Transactions. Ning Gao. B.A. in Accounting, Ren Min University, 1998

Transcription

1 The Short and Long-Run Fnancal Impact of Corporate Outsourcng Transactons by Nng Gao B.A. n Accountng, Ren Mn Unversty, 1998 M.A. n Economcs, Florda State Unversty, 2001 Submtted to the Graduate Faculty of Joseph M. Katz Graduate School of Busness n partal fulfllment of the requrements for the degree of Doctor of Phlosophy Unversty of Pttsburgh 2006

2 UNIVERSITY OF PITTSBURGH Joseph M. Katz Graduate School of Busness Ths dssertaton was presented by Nng Gao It was defended on July 25, 2006 and approved by Steven Husted, PhD, Professor Frederk Schlngemann, PhD, Assocate Professor Shawn Thomas, PhD, Assocate Professor Chad Zutter, PhD, Assstant Professor Dssertaton Advsor: Kuldeep Shastr, PhD, Professor

3 Copyrght by Nng Gao 2006

4 The Short and Long-Run Fnancal Impact of Corporate Outsourcng Transactons Nng Gao, PhD Unversty of Pttsburgh, 2006 Ths dssertaton nvestgates the fnancal mpact of a large sample of outsourcng contracts sgned by corporatons lsted on the US markets from 1990 through We construct a data set that dentfes the outsourcng clent and vendor frms and use ths data set to examne (a) the announcement effects of outsourcng contracts on frm value, (b) the mpact of outsourcng contracts on long-run stock and accountng performance and (c) the mpact of outsourcng contracts on the relaton between clent and vendor frms. v

5 TABLE OF CONTENTS 1.0 INTRODUCTION A REVIEW OF THE LITERATURE THE TESTABLE HYPOTHESES THE SHORT-RUN FINANCIAL IMPACT OF THE ANNOUNCEMENT OF OUTSOURCING TRANSACTIONS Informaton Asymmetry Economes of scale Focus on core competency Flexblty Contract Sze Country of orgn THE LONG-RUN FINANCIAL IMPACT OF THE SIGNING OF OUTSOURCING TRANSACTIONS OUTSOURCING TRANSACTIONS AS STRATEGIC ALLIANCES/PARTNERSHIPS THE METHODOLOGY ESTIMATING THE SHORT-RUN IMPACT OF THE ANNOUNCEMENT OF OUTSOURCING TRANSACTIONS Estmatng Abnormal Returns and Test Statstcs Cross-Sectonal Tests ESTIMATING THE LONG-RUN IMPACT OF THE SIGNING OF OUTSOURCING TRANSACTIONS Calculatng long-run buy-and-hold stock returns Measurng changes n accountng performance v

6 4.2.3 Measurng Abnormal Returns n Calendar Tme MEASURING THE IMPACT OF OUTSOURCING TRANSACTIONS WHEN VIEWED AS STRATEGIC ALLIANCES/PARTNERSHIPS Measurng the Relaton between Clent and Vendor Stock Prces Measurng Correlatons n Accountng Performance THE SAMPLE AND DATA THE SAMPLE A DESCRIPTION OF THE DATA EMPIRICAL RESULTS SHORT-RUN IMPACT OF THE ANNOUNCEMENT OF OUTSOURCING TRANSACTIONS Unvarate results Multvarate results THE LONG-RUN IMPACT OF THE SIGNING OF OUTSOURCING TRANSACTIONS Long-Run Stock Returns for Clent Frms after the Sgnng of Outsourcng Transactons Long-Run Accountng Performance of Clent Frms after the Sgnng of Outsourcng Transactons THE LONG-RUN INTEGRATION OF THE CLIENT AND VENDOR FIRMS The Relaton between Clent and Vendor Stock Returns n the Pre and Post-Sgnng Perod The Relaton between Clent and Vendor Accountng Performance n the Pre and Post-Sgnng Perod CONCLUSIONS REFERENCE v

7 LIST OF FIGURES Fgure 1. Frequency of outsourcng deals Fgure 2. Value of outsourcng deals v

8 LIST OF TABLES Table 1. Descrptve Statstcs for Sgned Contracts, Clent and Vendor Frms Table 2. Descrptve Statstcs for Cancelled Contracts Table 3: Abnormal returns for clent frms around outsourcng contract announcements Table 4: Abnormal returns for vendors around outsourcng contract announcements Table 5: The relaton between clent frms announcement day abnormal stock returns and measures of frm and contracts characterstcs Table 6: The relaton between vendors announcement day abnormal stock returns and measures of frm and contracts characterstcs Table 7: Ex-post long-run holdng perod abnormal returns for clent frms Table 8: Ex-ante holdng-perod abnormal returns for clent frms Table 9: Calendar-tme three-factor model for clent frms Table 10: Regressons of clent frms long-run abnormal stock returns Table 11: Ex-post long-run holdng perod abnormal returns of vendor frms Table 12: Ex-post accountng performance changes for clent frms Table 13: Ex-ante accountng performance changes of clent frms Table 14: Regressons of clent frms long-run accountng performance changes Table 15: Ex-post accountng performance changes of vendor frms Table 16: Regressons of vendor frms long-run abnormal stock returns and accountng performance changes Table 17: The changes n daly stock return cross-correlatons between clent and vendor frms87 Table 18: Cross-sectonal regresson of daly returns cross-correlaton changes Table 19: Long-run buy-and-hold stock return cross-correlatons between clent and vendor frms v

9 Table 20: Granger causalty tests Table 21: Unadusted accountng performance change cross-autocorrelaton test Table 22: Industry-medan adusted accountng performance change cross-autocorrelaton tests93 Table 23: Cross-autocorrelatons of percentage changes n return on assets x

10 1.0 INTRODUCTION Outsourcng defned as the delegaton of non-core operatons from nternal producton wthn a frm (clent frm) to an external entty that specalzes n that operaton (vendor frm) became a popular buzzword n busness n the md 1990s. Although outsourcng has been a management practce for over 200 years t has receved more attenton n the recent past snce the volume of nternatonal outsourcng has grown rapdly. 1 Proponents of outsourcng argue that ths actvty helps clent frms by provdng them wth the ablty to () purchase ntellectual captal that may otherwse not be avalable to them, () focus more on core competences, () better antcpate future costs and (v) lower costs. Ths mples that outsourcng helps US frms become more proftable, thereby beneftng shareholders. 2 Opponents of nternatonal outsourcng argue that ths actvty hurts the economy of the Unted States of Amerca, reduces the qualty of servce provded and eopardzes securty. For example, n the 2004 presdental campagn, Democratc canddate John Kerry clamed that nternatonal outsourcng was a maor cause of unemployment n the US and blasted companes and Chef Executve Offcers (CEOs) that outsource callng them Benedct Arnold Corporatons and CEOs and argued for nterventon by the US government. Ths vew seems to be shared by the Amercan publc snce a Zoghy Internatonal poll n 2004 reported that 71 percent of those polled beleved that outsourcng obs overseas hurts the US economy and 62 percent thnk that the government should mpose legslatve restrctons on outsourcng. In contrast, a Wall Street Journal poll of economsts reports that only 16 percent of those surveyed saw outsourcng as havng a sgnfcant mpact on obs. As a matter of fact, many suggested that outsourced obs are replaced by better payng obs n another 1 For example, there was not much publcty when Eastman Kodak Company contracted away the management of ts data centers to IBM Corporaton n 1989 or when Amercan Express set up ts back-offce operaton n Inda more than a decade ago. 2 For example, on December 9, 2002 when Anthem Inc. sgned an nformaton technology (IT) outsourcng contract wth Afflated Computer Servces, Inc., the stock of Anthem Inc, dsplayed a market-adusted abnormal return of 1.14 percent. 1

11 area/ndustry and that outsourcng s ust a new way of dong nternatonal trade. For example, Drezner (2004) reports that although 70,000 computer programmers lost ther obs between 1999 and 2003, 115,000 computer software engneers found hgher payng obs wthn the same perod. In addton, the McKnsey Global Insttute has estmated that for every dollar spent on outsourcng to Inda, the US reaps between $1.12 and $1.14 n benefts. As stated prevously, outsourcng has been part of busness practce for many years but has garnered much more attenton recently because of both the hstorcal and forecasted growth n the number and volume of these transactons. Fgures 1 and 2 provde a hstorcal perspectve on the annual number and total value of outsourcng transactons sgned by frms wth stocks lsted n the US over the perod startng n January 1990 and endng n December As can be seen from these fgures, the number of outsourcng deals per year has grown from 7 n 1990 to 216 n 2003 wth a peak of 241 n The correspondng fgures for the total value of outsourcng deals sgned per year (n 2000 dollars) are $ mllon, $27.78 bllon and $61.98 bllon (n 2002), respectvely. 3 Deavers (1997) argues that ths ncrease n outsourcng actvtes s a result of rapd technologcal change, ncreased rsk and the search for flexblty, greater emphass on core competences and globalzaton. In terms of future growth, the McKnsey Global Insttute estmates that the sze of ths market s gong to grow at a rate between 30 and 40 percent per year for the next fve years. In addton, a recent report by the busness consultant company INPUT states that outsourcng expendtures on the Informaton Technology (IT) sector alone s expected to grow from $10 bllon n fscal year 2005 to approxmately $18 bllon n fscal year Gven the magntude of the market for these transactons and the controversy surroundng offshore outsourcng, a natural queston that arses s as follows: what s the magntude of the benefts of outsourcng to the clent and vendor frms and what are ts determnants? The purpose of ths paper s to examne the short and long-run fnancal mpact of outsourcng transactons on clent and vendor frms over the 1990 to 2003 perod. Specfcally, we frst analyze the mpact of announcements of outsourcng contracts on the stock prces of the clent and vendor frms and provde an examnaton of the cross-sectonal determnants of the announcement effect. Second, we study the mpact of outsourcng contracts on the long-run 3 The numbers reported n Fgures 1 and 2 are based on the outsourcng sample used n ths study. See Secton 5 for more detals on our sample. 2

12 accountng and stock performance of the clent frms. Thrd, we analyze the mpact of outsourcng contracts on the relaton between the long-run accountng and stock performance of the clent and vendor frms. Our results ndcate that the average abnormal return for clent frms on the announcement day of outsourcng contracts (day 0) s an nsgnfcant percent. On the other hand, multvarate tests ndcate that the announcement s assocated wth postve day 0 abnormal returns. Specfcally, we fnd that the abnormal return decrease wth clent frm sze and clent frm flexblty whle they ncrease wth the relatve sze of the vendor frm to the clent frm and clent frm opacty. Based on these results, we conclude that the short-run of outsourcng transactons can be partally attrbuted nformaton asymmetry. In addton, our results are consstent wth the hypothess that outsourcng s undertaken to take advantage of economes of scale, to have the ablty to focus more on core competences and to provde more operatng flexblty. We also fnd that the rvals of the clent frms experence a negatve day 0 abnormal return. Ths suggests that n the eyes of the market, sgnng the outsourcng contract provdes the clent frm wth compettve advantages vs-à-vs the rest of the ndustry. Wth respect to vendor frms we fnd that they gan percent on the announcement day. Ths gan decreases wth vendor sze and ncreases wth the sze of the deal. We also fnd those deals that are renewals of prevously sgned deals have a smaller mpact on day 0 stock prces. Overall ths suggests support for hypothess that the short-term effect of the announcements of outsourcng transactons on vendor frms are partally drven by nformaton asymmetry. In addton, the results also ndcate support for the hypotheses that the sources of gans to vendor frms from outsourcng are economes of scale and the amount the contract enhances the vendor s revenue stream. We also fnd that the rvals of vendor frms experence a postve abnormal return on day 0. Ths result s consstent wth the good news assocated wth the announcement havng a contagon effect n the ndustry. Gven that vendors experence a postve abnormal return when t s learned that they have been awarded an outsourcng contract, one would expect the opposte on an announcement of a contract beng cancelled by the clent frm or when the clent frm announces that the vendor was not chosen to be the partner n a partcular transacton. Our results are consstent wth ths 3

13 hypothess. Specfcally, we fnd that vendors experence a negatve day 0 abnormal return around such announcements. In the case of clent frms, long-run stock and accountng performance s consstent wth the results reported above for short-run stock performance. Specfcally, we fnd that the buyand-hold returns (BHARs) for clent frm stocks n the 3-year perod pror to the outsourcng transacton are not sgnfcantly dfferent from the BHARs for a control group of frms. On the other hand, BHARs for clent frms n the post-outsourcng perod are sgnfcantly larger than that for the control group. The same result hold for measures of accountng performance ncludng sales per employee, ncome per employee, gross proft margn, net proft margn, and asset turnover. Fnally our results are consstent wth the vew that outsourcng transactons create a strategc partnershp between clent and vendor frms. Specfcally we fnd that clent and vendor frm stock returns are not correlated pror to outsourcng transactons but the correlaton becomes sgnfcantly postve n the post-outsourcng perod. The same result holds true for changes n accountng performance when accountng performance s measured by return on assets and asset turnover. The remander of ths dssertaton s organzed as follows. Secton 2 revews the lterature of the fnancal mpact of outsourcng. Secton 3 dentfes the specfc hypotheses that are tested here. Secton 4 descrbes the methodology used n the tests of the hypotheses, whle Secton 5 contans a descrpton of the sample and data. The emprcal results and the nterpretaton of the results are contaned n Secton 6 and Secton 7 concludes. 4

14 2.0 A REVIEW OF THE LITERATURE The early outsourcng lterature employs the economcs of transacton cost model of the frm as the prmary theoretcal lens to examne outsourcng arrangements. 4 Accordng to Coase (1937), the lmts of the frm are determned by the relatve producton costs nsde the frm as compared wth the costs of usng the market (outsourcng). Wllamson (1985) attrbutes transactons costs to suppler hold-ups after customers have already nvested n relatonshp specfc assets. Thus, when hold-up problems are very costly, nternal herarches are more advantageous to external relatonshps (outsourcng). Grossman and Helpman (2002) theoretcally examne a frm s decson to produce nhouse or outsource. They model the make or buy decson as a trade-off between dseconomes of scope and the transacton costs that stem from search frctons and ncomplete contracts. They fnd that where the cost advantage of specalzed component producers s large and ther barganng power vs-à-vs specalzed fnal producers s great, outsourcng s more lkely to emerge n a stable equlbrum the greater s the substtutablty between varetes of fnal goods. The transacton cost theory of the frm suggests that as frms evolve and compare the costs of an nternal versus external herarchy, they wll engage n outsourcng as they start to recognze that the producton costs of managng ther own nternal operatons may be reduced by outsourcng due to the effect of consderable economes of scale. 5 Specfcally, by outsourcng, clent frms can beneft from economes of scale when there are large specalst vendor frms that can provde the servces at lower average cost.. Another possble reason for frms recognzng the value of outsourcng could be a result of an external shock to ther operatng envronment. For example, Sharpe (1997) suggests that 4 For more detals on the theory of the frm based on the economcs of transactons costs, see Coase (1937) and Wllamson (1979). 5 For example, see Abraham and Taylor (1996), Sharpe (1997), Deavers (1997) and McCarthy and Anagnostou (2004). 5

15 deregulaton of telecommuncatons companes n the 1980s and publc utlty companes n the 1990s caused a change n the operatng envronment of these companes resultng n ther embracng outsourcng as a more effcent way of dong busness. Specfcally, Sharpe states that outsourcng dd not emerge as the consequence of a sudden techncal breakthrough, nor dd t grow out of a best sellng book by a well-known management guru. Rather, t was the result of market forces that emerged n response to demands for more effcent ways to address organzatonal compettveness. Qunn and Hlmer (1994) make a core competency argument for the use of outsourcng. Specfcally, they suggest that the potental gans offered by outsourcng are optmzed when outsourcng enhances a core competency busness strategy. Thus, when correctly or optmally combned, core competency and extensve outsourcng strateges provde more flexblty, greater effcency and better responsveness to customer needs at lower costs. Most mportantly, they suggest that strategc outsourcng provdes organzatons wth a compettve edge n the long run. There s anecdotal evdence that corporate decson makers act n a way consstent wth ths strategc outsourcng theory. For example, at the tme of sgnng a fve-year outsourcng contract worth $2.1 bllon n July 1991 by LaBarge Inc. and McDonnell Douglas Co., Wllam Maender, LaBarge s vce presdent, sad that t s an effort to mprove operatng effcency, control costs, standardze practces among plants and comply wth the demandng government reportng requrements of manufacturng for the defense market. In another example when Afflated Computer Servces Inc. (ACS) announced the renewal of a comprehensve IT outsourcng agreement wth Afflated Health Servces of Mt. Vernon, Washngton n January 2000, Tom Ltaker, CFO of Afflated Health Servces sad that "ACS has the experence and expertse to ntegrate technology wth our busness and help us reach our goals n the future. We expect ACS' broad range of servces and ndustry knowledge to brng us effcency and cost effectveness n operatons as we focus on provdng the best possble care to our consttuents." In summary, the ratonale for frms to outsource s that ths actvty helps organzatons ncrease operatng effcency, reduce operatng costs, whle achevng an ncreased focus on core competences. Ths would suggest that clent and vendor frms should beneft n the short-run and long-run from sgnng these outsourcng contracts. 6

16 There are a few studes that have examned the short-run stock prce mpact of outsourcng transactons. One of the earlest papers s one by Hayes, Hunton and Reck (2000) n whch they analyze the effect of the announcement to outsource all or a porton of a frm s nformaton systems (IS) functons on the market value of the clent frm. The frst hypothess they test s that IS outsourcng announcements wll have a greater postve mpact on the market value of smaller frms as compared to the market value of larger frms, because there tends to be more nformaton asymmetry about smaller frms. Specfcally, announcements made by small frms are expected to yeld a greater market response because of the bgger surprse they generate. Second, they hypothesze that the mpact of IS outsourcng announcements would have a greater postve mpact on the market values of servce frms as compared to the market values of non-servce frms. Ths hypothess s based on the argument that there s more nformaton asymmetry about servce frms snce standard fnancal-reportng systems do not capture many factors (such as ntellectual captal and other soft assets) mportant to servce ndustres and that servce frms allocate a hgher proporton of ther resources to nformaton technology as compared to non-servce frms. Ths mples that the reacton to the announcement should be more postve because t s more of a surprse and s more meanngful to the frm. Usng a sample of clent frms that announce IS outsourcng arrangements from 1990 through 1997, they fnd no statstcally sgnfcant stock prce change for a two-day event wndow for ther complete sample of clent frms. 6 For ther sub-sample of small frms, they report weakly sgnfcant postve abnormal stock returns for the two-day event wndow. For a one-day wndow (the day after the announcement day), they report sgnfcant postve abnormal stock returns for both small frms and servce frms. Fnally, they fnd support for both ther hypotheses n a multvarate regresson analyss. There are several other papers that follow Hayes et. al. (2000) and employ the event study methodology to examne the short-run mpact of the announcement of outsourcng transactons. For example, Farag and Krshnan (2003) examne IT outsourcng deals announced between January 1994 and August 2001 and fnd that there are postve announcement effects for outsourcng decsons by frms n the IT and servce ndustres. They also fnd that the stock prce react postvely to the announcement of strategc sourcng proects but not to cost cuttng proects. Gellrch and Gewald (2005) examne outsourcng decsons by frms n the fnancal 6 The two-day event wndow conssts of the announcement day and the day after the announcement day. 7

17 servces ndustry and fnd that there are postve announcement effects assocated wth large deals, deals that nvolve experenced vendors and those that nvolve the IT functon. There s a paucty of emprcal evdence on the long-run effects of outsourcng transactons. One excepton s Glma and Görg (2004) who examne the mpact of outsourcng decsons by UK manufacturng frms on the labor productvty. They fnd that outsourcng s assocated wth mproved labor productvty n the long-run. A maorty of the emprcal lterature on outsourcng focused on only one party n the contract, that s, ether the clent or vendor frm. In contrast, on the theoretcal level the lterature recognzed outsourcng transactons as a proect assgned by the clent frm (the prncpal) to the vendor frm (the agent). Ths could lead to potental agency problems snce the two partes do not have perfectly matched goal. 7 These agency ssues can be mtgated by developng contracts between the two partes that specfy the relatonshp between the two partes and defne the performance metrcs that can be used by the prncpal to montor the agent. 8 More recently, t has been argued n the lterature that rather than vewng outsourcng transactons as smple transactonal contracts, they should be vewed as strategc allances/partnershps between the clent and vendor frms. 9 For example, Natovch (2003) uses a case study of a proect falure to argue that a contract-drven approach to mtgate vendor rsk n outsourcng transactons may not be as effectve as a partnershp approach to sharng rsk. 10 Ths strategc allance vew of outsourcng transactons has sgnfcant mplcatons for the future relaton between the accountng and stock prce performance of the clent and vendor frms. Specfcally, one would expect the performance metrcs for the two frms to become more correlated wth each other after the sgnng of the outsourcng transacton. Ths dssertaton extends the extant fnance and IS lterature on outsourcng n a number of ways. Frst, t provdes a more complete pcture of the mpact of the announcement of the sgnng of outsourcng transacton by lookng at both the clent and vendor frms. Second, t 7 For example, see Banker and Kemerer (1992) 8 For example, see Chaudhury, Nam and Rao (1995). 9 For example, see Gallvan and Oh (1999). 10 The case study n Natovch (2003) s that of an outsourcng contract between a telecommuncatons company, Bezeq (the clent) and a software company, AMS (the vendor) to develop the code for a new bllng system that was sgned n September Bezeq cancelled the contract n August 1999 and termnated the proect before even a sngle lne of code was delvered. 8

18 analyzes the long-run mpact of outsourcng transactons on clent frms, an area not explored prevously. Fnally, t provdes a drect test of the strategc allance hypothess. 9

19 3.0 THE TESTABLE HYPOTHESES 3.1 THE SHORT-RUN FINANCIAL IMPACT OF THE ANNOUNCEMENT OF OUTSOURCING TRANSACTIONS The argument presented n the prevous secton would suggest that the announcement of outsourcng transactons should be assocated wth an ncrease n the clent and vendor frms stock prces. Agan, the prevous lterature suggests that the magntude of the stock prce ncrease should vary across frms based on a number of factors that nclude frm sze, economes of scale nvolved, the degree of focus on core competences, flexblty provded by the outsourcng, the sze of the contract and the locaton of the vendor frm Informaton Asymmetry It s well recognzed n the lterature that the amount of publcly avalable nformaton s not equal for frms of dfferent szes. Snce large frms are more closely followed by meda and analysts than small frms one would expect less nformaton asymmetry to be assocated wth large frms. Ths would mply that announcements by smaller frms should be assocated wth a larger announcement-day movement n stock prce. Ths suggests that the market reacton of the clent (vendor) frm s stock to corporate outsourcng announcements wll be a decreasng functon of the sze of the clent (vendor) frm Economes of scale It was argued earler that outsourcng takes advantage of the economes of scale avalable through the vendor frm by savng overall producton costs and mprovng operatng effcency 10

20 for the clent frm. It has also been suggested that larger frms enoy better economes of scale than ther smaller counterparts. 11 Thus, one would expect the market reacton of the clent frm s stock to corporate outsourcng announcements to be the smallest for contracts between large clents and small vendors and the largest for contracts between small clents and large vendors Focus on core competency It was argued earler that by focusng on ts core competences and strategcally outsourcng other actvtes, a frm can create unque value for ther customers and that, n turn, can help the frm mantan ts compettve edge n the long run. Therefore, ths ncrease focus n core competences should translate nto hgher stock values for the clent frm. We use opacty as a measure of the degree to whch a frm has focused on core competences. Specfcally, opacty measures whether a frm s earnngs are more dependent on the realzaton of future growth opportuntes than on assets already n place. The specfc measure of opacty used here s net plant, property and equpment (PPE) dvded by total assets (PPE/TA) wth frms havng hgh values of PPE/TA beng consdered more transparent and beng more focused on core competences. Ths suggests that the market reacton of the clent frm s stock to corporate outsourcng announcements wll be a decreasng functon of opacty (PPE/TA) Flexblty Outsourcng can also help provde greater flexblty, especally n the purchase of rapdly developng new technologes, fashon goods, or the myrad components of complex systems. 13 It 11 For example, see Sharpe (1997). 12 One possble problem wth ths concluson s that our measure of opacty may ust be a proxy for frms n the busness servce ndustry snce such frms tend to have a low PPE to total assets raton. To ensure that are results are not drven by ths factor, we examne frms that do not belong to the busness servce ndustry separately. Our results ndcate that frms wth low opacty n ths group experence a day 0 abnormal return of percent wth an assocated sgnfcance level of 10 percent. Ths suggests that the argument that clent frms outsource to focus on core competences s not drven by frms n the busness servce ndustry. 13 For example, see Carlson (1989) and Domberger (1998). 11

21 allows companes to ncorporate the latest technology and respond to changes n busness envronment more quckly and at a lower cost than vertcally ntegrated organzatons. Flexblty s measured here by two lqudty ratos - the current rato and the quck rato. Specfcally, the hgher a frm s lqudty rato, the better able they are to meet short-term oblgatons and the greater ther fnancal flexblty. Snce clent frms wth low lqudty ratos have less fnancal flexblty, we would expect them to beneft more from outsourcng. Therefore, the market reacton of the clent frm s stock to corporate outsourcng announcements wll be decreasng functon of frm lqudty Contract Sze It can be argued that a larger contract sze represents potentally larger cost savngs for the clent frm and a potentally larger revenue stream for the vendor frm. Ths would suggest that the market reacton of both the clent frm s and vendor frm s stocks to outsourcng contract announcements ncreases wth the sze of the contract Country of orgn A number of groups have suggested that nternatonal outsourcng s not good for the US economy because t results n a frm substtutng hgher payng obs n the US for lower payng obs n another country. Ths would mply that the announcement of nternatonal outsourcng contracts should be assocated wth a larger market reacton n the clent frm s stock prce as compared to the announcement of domestc outsourcng contracts. On the other hand, t has been argued that outsourcng of any sort s a response by frms facng a complex change n ts cost boundares and the choce between domestc and nternatonal vendor frms s solely dependent on whch frm provdes the best strategc ft. 14 In ths scenaro, the market reacton of both the clent frm s stock to outsourcng contract announcements should not be dependent on the country of orgn of the vendor. 14 For example, see Deavers (1997) 12

22 3.2 THE LONG-RUN FINANCIAL IMPACT OF THE SIGNING OF OUTSOURCING TRANSACTIONS As stated prevously, the second obectve of ths dssertaton s to examne the mpact of outsourcng transactons on the long-run performance of the clent frms. In the prevous secton we argued that the announcement of outsourcng transactons should be assocated wth an ncrease n the clent frm s stock prce. Ths hypotheszed ncrease n stock prce results from an expectaton n the market that the clent frm wll be operatng more effcently n the future. Thus one would expect that clent frms would exhbt abnormally postve long-run accountng and stock prce performance. In addton, snce the mpact of outsourcng s hypotheszed to be more postve for smaller, more opaque and less flexble clent frms, we would expect the same to be true wth long-run abnormal performance. 3.3 OUTSOURCING TRANSACTIONS AS STRATEGIC ALLIANCES/PARTNERSHIPS As stated earler, t has been argued n the lterature that rather than vewng outsourcng transactons as smple transactonal contracts, they should be vewed as strategc allances/partnershps between the clent and vendor frms. Ths strategc allance vew of outsourcng transactons has sgnfcant mplcatons for the future relaton between the accountng and stock prce performance of the clent and vendor frms. Specfcally, one would expect the performance metrcs for the two frms to become more correlated wth each other after the sgnng of the outsourcng transacton. 13

23 4.0 THE METHODOLOGY 4.1 ESTIMATING THE SHORT-RUN IMPACT OF THE ANNOUNCEMENT OF OUTSOURCING TRANSACTIONS Estmatng Abnormal Returns and Test Statstcs We estmate the short-run mpact of the announcement of outsourcng transactons usng the methodology outlned n Brown and Warner (1985) to calculate abnormal returns on stocks around a day of nterest. Specfcally, for all the outsourcng deals n our sample, we defne the contract announcement day as event day zero. 15 A tradng date that s t days before the announcement day s denoted as day t, whle a tradng day t days after the announcement s denoted as day +t. The analyss s based on stock returns over a perod startng at day -250 and endng at day +10 (-250, +10). The frst 240 days n ths perod (-250, -11) are desgnated as the estmaton perod, and the followng 21 days (-10, +10) are desgnated as the announcement perod. The market model s used to adust for market-wde rsk factors. Specfcally, for any securty, the abnormal or excess return over each of the t =-10,...,10 days around the tme of the contract announcement s defned as, A t = R t ˆ α + ˆ β R ( mt ) (1) where A t s the abnormal return for securty at event day t R, t s return on securty at event day t and R mt s return on Center for Research n Securty Prce (CRSP) value-weghted ndex at event day t. The coeffcents $α and $ β are Ordnary Least Squares (OLS) estmates 15 A descrpton of our sample and the methods used to dentfy the announcement date are provded n Secton 5. 14

24 from the regresson of the return on securty on the CRSP value-weghted ndex over the estmaton perod. The statstcal sgnfcance of the abnormal returns s assessed n two ways dependng on whether the returns under consderaton are one event day or multple event days n the announcement perod. For a sngle event day, the t statstc s calculated as the rato of that event day s mean excess return to ts estmated standard devaton. Specfcally, let N t represent the number of sample securtes whose excess returns are avalable on event day t. Then we can express the test statstc for any sngle event day t as, where A Sˆ( A ) t / t (2) A t = 1 N t N t = 1 A t (3) and Sˆ( A ) = t ( A = 11 ( t= The test statstc has a Student-t dstrbuton. A A) t 11 A t t= ) / 239 For tests over mult-day ntervals, our test statstc s calculated as follows. The measure of abnormal performance of any securty between any two days d 1 and d 2 announcement perod s gven by the cumulatve abnormal returns, (4) (5) n the d CAR A t = 2 t= d 1 (6) The mean cumulatve abnormal return between days d 1 and d 2 for the N securtes n the sample s gven by, 1 CAR = N 15 N = 1 CAR The test statstc s defned as the rato of the mean cumulatve abnormal return to ts estmated standard devaton, whch s expressed as, (7)

25 CAR d 2 t= d1 Sˆ 2 ( A t ) (8) where Sˆ( A t ) as a Student-t. n the denomnator s defned as n equaton (4). Ths test statstc s also dstrbuted Cross-Sectonal Tests Cross-sectonal tests are conducted to ascertan what frm and contract characterstcs have an mpact on the stock prce reacton to announcements of outsourcng transactons. Specfcally, we regress the clent (vendor) frms day 0 abnormal returns on measures of frm and contract characterstcs. The regresson for clent frms s of the form: where A = α + βcsze + ϕlqudty + χcszevsvsze + η Re newal + γforegnv A s the clent s day 0 abnormal return, market value of equty (MVE), sze, C Re latvedealsze costs of goods sold or by MVE, equpment (PPE) to total assets, Re newal contract, CSzevsVSze + δc RelatveDealSze CSze + θearly + φopacty + πinteractve + ε (9) s ether the log of clent frm sales or s the log of vendor sze dvded by clent frm s the log of contract value per year dvded ether by clent frm s Opacty Lqudty s the log of clent frm s rato of property plant and s the log of clent frm s current or quck rato, s a dummy varable that equals 1 when the contract s a renewal of a exstng ForegnV s a dummy varable that takes on a value of 1 f the vendor frm of the contract s a foregn frm, Early Interactve occurs before 1998 and varables. 16 s a dummy varable tthat takes on a value of 1 f the deal s the nteracton of the tme dummy wth other ndependent 16 All varables based on accountng data are measured at the end of the year precedng the contract announcement year. 16

26 Based on arguments presented prevously, we would expect the coeffcents of clent sze ( β ), relatve clent to vendor sze ( χ ), opacty ( φ ), flexblty ( ϕ ) and the renewal dummy ( η ) to be negatve, whle that of deal sze (δ ) to be postve. 17 The coeffcent of the foregn vendor dummy ( γ ) s an emprcal queston. The regresson for the vendor frms s of the form: A = κ + VSze ι + λv RelatveDealSze + ν Re newal + ϑearly + ρinteractve + ϖ + οgov (10) where and entty, and A s the vendor VSze s day 0 abnormal return, s the log of vendor frm sales or MVE Gov s a dummy varable that equals 1 when vendor sgns a contract wth a governmental Early s a dummy varable that takes on the value of 1 when the deal occurs before 1998 Interactve s the nteracton of the tme dummy wth other ndependent varables. 18 Based on arguments presented prevously, we would expect the coeffcents of vendor sze (τ ), the renewal dummy (ν ) and the government dummy ( ο ) to be negatve, whle that of deal sze ( λ ) to be postve ESTIMATING THE LONG-RUN IMPACT OF THE SIGNING OF OUTSOURCING TRANSACTIONS We use three technques to analyze the long-run fnancal mpact of the sgnng of outsourcng transactons. Frst, we examne buy-and hold returns for the clent stock from the effectve day of the contract to the three-year annversary of the effectve date. 20 Second, we compute changes n accountng performance measures from one year before to three years after the effectve date of 17 The predcted negatve sgn for the renewal dummy s based on the argument that announcement of renewals should have a smaller market mpact than the announcement of new contracts. 18 All varables based on accountng data are measured at the end of the year precedng the contract announcement year. 19 The predcted negatve sgn for the government dummy s based on the argument that an outsourcng transacton wth the government does not represent a strategc allance. 20 If the frm gets delsted pror to the three-year annversary, we use the delstng date as the endng date for the buy-and-hold perod. 17

27 the contract. The specfc accountng performance measures used are sales effcency, ncome effcency, gross proft margn, net proft margn, asset turnover and return on assets. Fnally, we use the Fama and French (1993) three-factor model to determne the 3-year abnormal return on portfolos on clent frms. Ths approach s smlar to analyzng buy-and-hold returns but s consdered superor snce t elmnates the problem of cross-sectonal dependence among sample frms and yelds more robust test statstcs. Each one the three set of tests are descrbed n more detals n the followng sectons Calculatng long-run buy-and-hold stock returns For the purposes of calculatng buy-and-hold returns for the clent frm stocks, we defne the contract effectve day as event day zero. A tradng date t days before the effectve day s denoted as day t and that t days after day 0 s denoted as day +t. We follow each clent (vendor) frm from day zero untl the earler of ts delstng date or the date of the contract s thrd annversary. We defne a year as twelve 21-tradng day ntervals (252 days). A three-year wndow of 756 tradng days s used n order to facltate comparsons wth other studes. The percentage buyand-hold return for frm s defned as: R T T = [ (1 + r ) 1] where T r s the earler of the delstng date or the end of the three-year wndow, and t t= 1 t (11) s the return for frm on datet. For each frm n the sample, we choose a matchng control frm usng a varaton of the matchng procedure suggested by Barber and Lyon (1996). They suggest that f the tests of nterest are desgned to detect abnormal performance followng an event, sample frms should be matched wth control frms based on pre-event performance as of year pror to the event year. Barber and Lyon (1997) provde evdence that the procedure of matchng sample frms to control frms of smlar szes and book-to-market ratos gves well-specfed test statstcs and yelds more powerful, and unbased test statstcs than other matchng procedures. As suggested by Barber and Lyon, our matchng procedure attempts to match clent (vendor) frms wth control frms on the bass of sze and book-to-market effects. 18

28 Specfcally, sze-matched frms are selected from all publc companes (excludng the sample frms) at the end of the year pror to the contract effectve year. The sze matched frm s the frm closest n market captalzaton to the clent (vendor) frm. When matchng on sze and book-to-market ratos, we select the subset of frms that have market equty values wthn 30% of the market equty value of the sample frm. Ths subset s then ranked agan accordng to bookto-market ratos. The sze and book-to-market matched frm s the frm wth the book-to-market rato, measured at the end of the year pror to the contract effectve year, whch s closest to the clent (vendor) frm s rato. Matched frms are ncluded for each sample frm for the full 3-year holdng perod or untl the date of delstng. If a matchng frm s delsted before the endng date for ts correspondng sample frm, a second matchng frm s splced n after the delstng date of the frst matchng frm. The replacement frm s the non-sample frm wth sze (sze and book-tomarket) at the orgnal rankng mmedately next to the orgnal matchng frm. Ths matchng methodology helps reduce any bas from survvorshp. Buy-and-hold abnormal returns are defned as the average of equally weghted pared dfferences between buy-and-hold returns for the sample frms and those for the matchng control frms. That average equals 0 f the sample frms do not have a better long-run stock performance compared to ther control frms. The average T -year abnormal buy-and-hold return s measured as, where N 1 Rτ = ( RT RmT) N = 1 R T s the percentage buy-and-hold return on frm for holdng perod T, RmT (12) s the percentage buy-and-hold return on frm s matchng frm for the same holdng perod T, and N s the number of frms n the sample. Based on our prevous dscusson, we would expect ths abnormal buy-and hold return to be postve and sgnfcant for our sample of clent frms. In addton, we regress the abnormal buy-and hold return on the same set of ndependent varables used n the multvarate analyss n Secton to determne what frms and contract characterstcs contrbute to the correlaton change. 19

29 4.2.2 Measurng changes n accountng performance From both academc and anecdotal evdence, the often cted motvaton for outsourcng s to ncrease the focus on the core operaton of the clent frms, thus mprovng ther operatng effcency. To examne whether operatng effcency mproves after outsourcng transactons, we examne changes n accountng performance/effcency measures around the sgnng of outsourcng transactons. Specfcally, operatng effcency s proxed by sx measures: sales effcency (SALEFF) defned as sales dvded by the number of employees, ncome effcency (IEFF) defned as operatng ncome dvded by the number of employees, 21 gross proft margn (GRSMRGN) defned as one mnus the rato of costs of goods sold to sales, net proft margn (NETMRGN) defned as the rato of net ncome to sales, asset turnover (TURNOVER) defned as the rato of sales to total assets and return on assets (ROA) defned as the rato of operatng ncome to total assets. Agan, f we defne the year the contract s effectve as year zero, we calculate these sx ratos for our sample frms and ther matchng frms one year pror to the contract effectve year (year -1) and three years after (year 3). For each year, we defne the adusted performance measure as the sample frm s rato mnus the benchmark rato. Because of skewness and the potental nfluence of outlers when usng accountng ratos, we follow Loughran and Rtter (1997) and focus on medan values. Two groups of benchmarks are used n the analyss. To control for ndustry effects, the operatng effcency measures are adusted by subtractng the medan value of the correspondng measures for all frms n the prmary two-dgt SIC ndustry n whch the frm was actve one year before the event. A two-dgt ndustry defnton s used because Clarke (1989) has shown that the two-dgt defnton captures smlartes among frms as effectvely as ndustry defntons based on three- or four- dgt SIC groupngs. A matchng-frm approach s used to facltate comparsons of ndustry-adusted operatng effcency. The specfc matchng algorthm follows the matchng method of Fee and Thomas (2004). In the case of each effcency proxy, we frst dentfy all frms that are not n our sample wth the same frst two-dgt SIC code as our sample frm, asset sze at the end of year SALEFF and IEFF are deflated by normalzng year -1 numbers to unty. Ths mples that the values n other year are expressed as a fracton of the value n the base year. 20

30 between 25% and 200% of the sample frm, and the effcency proxy between 90% and 110% of the sample frm. From these frms we choose as the matchng frm the company wth the effcency proxy closest to that of our sample frm. If no frm meets these crtera, then we relax the effcency proxy screen to between 70% and 130% of the sample frm. If no matchng frm s avalable ths tme, we relax the ndustry screen to requre only a one-dgt SIC code match, and then relax the effcency proxy requrement accordngly. If there s stll no match, we elmnate the ndustry matchng requrement and match on sze and the effcency proxy. Fnally, f no match s avalable after elmnatng the ndustry matchng requrement, we elmnate the sze requrement and match purely on the effcency proxy. In all but 4 percent of the cases, clent frms have matches at the two-dgt SIC level. The correspondng fgures based on matchng on one-dgt SIC codes, matchng on sze and the effcency proxy and matchng on the effcency proxy alone are 1.9, 2 and 0.1 percent, respectvely. Based on our prevous dscusson, we would expect the clent frms to show mprovements n operatng effcency from the pre-outsourcng perod to the post-outsourcng perod. Specfcally, we would expect ncome effcency, sales effcency, gross proft margn, net proft margn, asset turnover and return on assets to ncrease from year -1 to year +3. Fnally, we regress the changes n accountng performance on the same set of ndependent varables used n the multvarate analyss n Secton to determne what frms and contract characterstcs contrbute to the change n operatng effcency Measurng Abnormal Returns n Calendar Tme In ths secton we dscuss a calendar-tme approach to calculatng abnormal returns after the sgnng of the outsourcng transacton for a portfolo of clent frms. As opposed to an event-tme approach such as calculatng buy-and-hold returns, the calendar-tme approach offers some advantages. Frst, ths approach elmnates the problem of cross-sectonal dependence among sample frms because the returns on sample frms are aggregated nto a sngle portfolo. Second, the calendar-tme portfolo methods yeld more robust test. We use the three-factor model developed by Fama and French (1993). The calendar-tme perod consdered n the analyss represents a 36-month post-event wndow. For each calendar month, we calculate the return on a portfolo composed of frms that has an event wthn our 21

31 calendar wndow. The calendar-tme return on ths portfolo s used to estmate the followng regresson: where R pt R = α + β ( R R ) + s SMB + h HML + ε ft R pt R s the smple monthly return on the calendar-tme portfolo, ft on three-month Treasury blls, R mt mt ft t t t (13) s the monthly return s the return on the CRSP value-weghted market ndex, SMBt s the dfference n the returns of value-weghted portfolos of small stocks and bg stocks, and HMLt s the dfference n the returns of value-weghted portfolos of hgh book-to-market stocks and low book-to-market stocks. The estmate of the ntercept term α provdes an estmate of the abnormal return on a portfolo of clent frms. Based on our prevous dscusson we would expect ths abnormal return to be postve and sgnfcant for the portfolo of clent frms. 4.3 MEASURING THE IMPACT OF OUTSOURCING TRANSACTIONS WHEN VIEWED AS STRATEGIC ALLIANCES/PARTNERSHIPS In ths secton we dscuss the methodology used to analyze the evoluton of the relaton between clent and vendor frms from a tme perod n the pre-contract sgnng perod to a tme perod n the post-contract sgnng perod to determne f the sgnng of the outsourcng contract results n a closer lnk between the performances of the two partes nvolved n the contract Measurng the Relaton between Clent and Vendor Stock Prces One mplcaton of vewng an outsourcng transacton as a strategc allance between the clent and vendor frms s that the correlaton between the stock returns for the frms should ncrease from the pre-effectve to the post-effectve perod. 22 For purposes of mplementaton, the exact specfcaton of the pre-effectve and post-effectve perods depends on whether the vendor 22 The effectve date of a contract s the start date of the contract. 22

32 appears once (sngle contract vendors) or multple tmes (seral contract vendors) n 3-year (756 tradng day) wndows wthn our sample perod. For sngle contract vendors, the pre-perod (post-perod) s defned as the 378 tradng days before (after) the contract effectve date (day 0). For the seral contract vendors, the pre-perod s defned as the 756 tradng days before day 0 whle the post-perod s the day after the contract effectve day (day +1) to one day before the same vendor takes another contract. Cross-autocorrelaton coeffcents are calculated usng the Pearson product-moment correlaton method appled to daly stock returns for the clent and vendor frms. In order to mnmze the effect of non-synchronous tradng on cross-autocorrelaton, returns for both clent and vendor frm stocks on a partcular tradng day are excluded from the computaton of correlaton f ether the clent or vendor stocks dd not trade on that date. The cross-autocorrelaton coeffcent for each par of clent and vendor frm stocks n the pre-effectve perod s denoted by ρ ( r, r ;) whle that n the post-effectve perod s denoted PRE c v by ρ ( r, r ;) where r ( r ) s the return on the clent (vendor) frm stock. If the outsourcng POST c v c v transacton does create a strategc allance between the clent and vendor frms, one would expect the change n cross-autocorrelaton to be postve. We also regress the changes n crossautocorrelatons on the same set of ndependent varables used n Secton to determne what frms and contract characterstcs contrbute to the correlaton change. Another mplcaton of the strategc allance hypothess s that the outsourcng transacton can result n a lead-lag relaton between the clent and vendor frm stocks n the post-effectve perod where none exsted n the pre-effectve perod. We use a vector autoregresson (VAR) approach wth a lead-lag of one day to test whether there s a lead-lag relaton between clent and vendor stock returns. 23 The specfc equatons estmated are gven as: r V, t = α,0 + a,1rv, t 1 + a,2rc, t 1 + ν, t r C, t = β,0 + b,1rv, t 1 + b,2rc, t 1 + μ, t (14a) (14b) where rv, t( r C, t) represents the returns on the vendor (clent) frm assocated wth contract. If the daly stock returns of the clent frms lead those of the vendor frms, one would expect 23 The order of the VAR model s determned by usng both the Akake nformaton crteron (AIC) and the Bayes nformaton crteron (BIC). Both crtera suggest a lead-lag of one day. 23