COMPLEMENTARITY EFFECTS OF R&D AND INFORMATION TECHNOLOGY ON FIRM MARKET

Transcription

1 COMPLEMENTARITY EFFECTS OF R&D AND INFORMATION TECHNOLOGY ON FIRM MARKET VALUE INDRANIL BARDHAN SCHOOL OF MANAGEMENT, SM 41 THE UNIVERSITY OF TEXAS AT DALLAS RICHARDSON, TX , USA VISH V. KRISHNAN RADY SCHOOL OF MANAGEMENT UNIVERSITY OF CALIFORNIA, SAN DIEGO LA JOLLA, CA 92093, USA SHU LIN CRAIG SCHOOL OF BUSINESS CALIFORNIA STATE UNIV, FRESNO FRESNO, CA 93740, USA DECEMBER 7, 2010

2 ABSTRACT We empirically sudy he join impac of R&D and IT invesmens on firm marke value. The mixed evidence in he prior lieraure on he impac of R&D and IT on firm performance brings o mind new quesions of wheher R&D or IT alone are sufficien deerminans of firm marke value. As IT plays an increasingly criical role in he execuion of innovaion-inensive invesmens, i is imporan o undersand he exen o which IT has ransformed he execuion and delivery of R&D. We argue ha complemenariies beween R&D and IT spending will enable fuure growh opions and allow innovaion processes o become more effecive. We esimae he ineracion impac of IT and R&D invesmens on Tobin s q, a forward-looking measure of firms marke-o-book value. We es our empirical model on a recen muli-year, firm-level, archival daase ha spans muliple indusries. Our resuls sugges ha he ineracion effec of R&D and IT on Tobin s q is posiive and saisically significan afer conrolling for oher firm- and indusry-specific effecs. Our sudy addresses a criical gap in he innovaion lieraure which has ignored he possibiliy of ineracion effecs beween R&D and IT and heir impac on firm marke value. Our resuls provide empirical suppor for recen anecdoal evidence wih respec o manner in which IT is revoluionizing R&D-inensive innovaion processes. Keywords: R&D invesmens, IT invesmens, firm performance, ineracion, Tobin s q. 2

3 1. INTRODUCTION A large number of academic sudies have focused on he relaionship beween research and developmen (R&D) spending, produciviy growh, and firm-level performance (Morbey, 1988; Cohen and Klepper, 1996; Cohen, Levin and Mowery, 1987; Elie 1998). While recen sudies sugges ha R&D spending has a significanly posiive impac on firm produciviy, wih a rae of reurn ha is somewha larger ha he rae of reurn on radiional invesmens (Lev and Sougiannis, 1996), ohers have repored ha R&D invesmens do no have a saisically significan impac on financial measures such as sock reurns (Chan e al. 2001). The mixed evidence of he impac of R&D on firm performance brings o mind new quesions: Is R&D alone a sufficien deerminan o explain firm performance? Could oher facors moderae he impac of R&D on firms marke value? Recen anecdoal evidence suggess ha informaion echnology (IT) plays an imporan role in enabling and enhancing he produciviy of R&D processes (Marwaha e al. 2007). Mial and Naul (2009, p. 140) observe ha value from informaion echnology (IT) arises no only direcly hrough changes in he facor inpu mix bu also indirecly hrough IT-enabled augmenaion of non-it inpus and changes in he underlying producion echnology. Their empirical analyses of indusry daa show ha he indirec effecs of IT predominae in IT-inensive indusries while he direc effecs of IT dominae in non-it inensive indusries. Alhough recenly researchers have argued ha IT is changing he naure of innovaion, here is a dearh of empirical evidence on how IT invesmens inerac wih R&D invesmens o enhance overall firm performance and marke value. In his paper, we focus on he role of IT in moderaing he impac of R&D invesmens on firm marke value. We es our research hypoheses on a unique and recen daa se of firm-level, IT spending over an eigh-year period

4 ha allows us o race he impac of IT invesmens on he relaionship beween R&D and firm marke value. Our sudy is he firs aemp a examining he join impac of R&D and IT on firm value using recenly available daa from he Inerne era ha allows us o sudy he impac of newer informaion echnologies on he innovaion-inensive processes. Our resuls suppor our hypohesis ha IT spending moderaes he impac of R&D spending on Tobin s q, a raio ha measures he marke-o-book value of a firm. Our resuls are robus and consisen across a number of model specificaions and economeric esimaion mehods. While he focus of prior research has been on sudying he direc effecs of R&D and IT invesmens, our resuls show ha heir join (or ineracion) effecs are posiive and significan across firms from many indusries spanning a long ime period. Our resuls sugges ha he overall effecs of R&D and IT canno be simply capured by measuring only heir individual direc effecs, bu raher by accouning for heir ineracion effec on firm valuaion. The resuls indicae ha Tobin s q, a forward-looking measure of he sock marke s percepion of a firm s fuure growh poenial, is significanly influenced by he effeciveness of IT-enabled R&D invesmens. 2. RELATED LITERATURE There exiss a large body of lieraure on he impac of R&D on produciviy and firm marke value. However, much of his lieraure does no consider he role of IT invesmens (Griliches and Mairesse, 1984; Englander, Evenson, and Hanazak 1988; Chauvin and Hirschey, 1993; Griliches, 1994; Hall and Mairesse, 1995). We review he relaed lieraure on he impac of R&D and IT on firm performance as a prelude o developing a heory-based model which we es in he nex secion. 4

5 2.1 IMPACT OF R&D AND INFORMATION TECHNOLOGY ON FIRM MARKET VALUE A majoriy of he exan research on IT and firm performance has focused on relaing he impac of IT spending o accouning measures of firm performance. Mos of hese sudies have focused on produciviy measures such as firm oupu (Brynjolfsson and Hi, 1996; Hi and Brynjolfsson, 1996; Barua e al. 1991), while more recen sudies have focused on firm profiabiliy measures. While early sudies on he impac of IT on firm produciviy were equivocal (Rai e al. 1997; Carr 2004), recen sudies have shown ha IT plays a significan role in improving firm and indusry produciviy growh (Brynjolfsson and Hi, 1995; Breshnahan e al. 2002). Recen sudies have also demonsraed he impac of IT on firm profiabiliy using firm-level daa colleced during he las decade (Aral and Weill, 2007; Mihas e al. 2010). However, as noed by Foray, Hall and Mairesse (2007) and Bharadwaj (1999), here are several limiaions in linking he impac of R&D and IT invesmens on accouning-based performance measures. Firs, much like R&D, he rue conribuion of IT invesmens mus be considered based on heir impac on long-run firm performance and risk avoidance. Due o he learning curve and long implemenaion imes associaed wih major IT projecs, invesmens in IT sysems are likely o have a lagged impac on firm business value which is more likely o be refleced in fuure cash flows. Second, many IT invesmens are made o manage or avoid operaional or regulaory/compliance risks. Risk avoidance remains an imporan objecive of many IT invesmens as borne by a spae of IT conrols ha firms implemened in he pos- Sarbanes-Oxley era o comply wih new regulaions relaed o COSO (REF). A recen sudy by Arkal Bardhan and Krishnan (2009) repored ha IT invesmens are associaed wih lower firm risk as measured by he volailiy of sock marke reurns and earnings. For example, in 2003, Wal-Mar iniiaed implemenaion of radio frequency idenificaion (RFID) echnologies wihin

6 is supply chain which allowed i o rack invenory a any given ime (Mendelson, 2007). The RFID iniiaive resuled in reduced sockous and lower losses (from hef and perishable iems) as producs move across he supply chain, and provides an example of a firm leveraging is IT invesmens o manage supply chain disrupion risks, which may no be refleced in greaer profiabiliy in he shor run. Similarly, he impac of IT on oher facors such as firm flexibiliy, agiliy, and growh poenial, may no be fully represened in accouning measures (such as profiabiliy) in sudies linking he impac of IT invesmens o firm performance. On he oher hand, measures such as Tobin s q reflec he ex ane marke valuaion of he level and risk of fuure firm cash flows (Ben-Horim and Callen, 1989; Smirlock e al. 1986). Second, R&D and IT invesmens provide firms wih significan growh opions ha are ypically no accouned for in presen reurns. While pas sudies have primarily focused on he cos-reducion and ransacional auomaion benefis associaed wih IT invesmens, more recen sudies have explored he revenue growh poenial associaed wih IT invesmens, especially in he Inerne era (Bardhan e al. 2004; Sambamurhy e al. 2003). New ypes of web-based sysems are expeced o have far greaer ransformaional poenial compared o heir predecessor sysems (Aral and Weill, 2007; McAfee and Brynjolfsson, 2008; Mihas e al. 2010). In his respec, boh R&D and IT invesmens are associaed wih significan inangible value in he form of fuure growh opions ha can be enabled by hese invesmens (Brynjolfsson, Yang and Hi, 1999). Jus as R&D expendiures have been reaed as an imporan deerminan of a firm s inangible asses, in erms of creaing superior knowledge, innovaion, and echnological capabiliies, researchers have increasingly sared o pay more aenion o he sof benefis associaed wih IT invesmens such as improvemens in qualiy, cusomer service, and sraegic flexibiliy associaed wih IT infrasrucure (Bharadwaj e al. 1999; refs). A financial measure 6

7 such a Tobin s q which measures he value of a firm based on is fuure earnings relaive o curren book value is a beer indicaor of he growh opions associaed wih IT spending. We argue ha he associaion beween IT invesmens and firm inangible value, wih respec o is enablemen of echnology-enabled growh opions, suppors our raionale for using Tobin s q as a measure of firm performance in his sudy. In heir seminal work on he impac of IT asses on firm performance, Bharadwaj e al. (1999) show ha IT expendiures accouned for a significan porion of he variance in Tobin s q based on heir analyses of firm-level daa from 1988 o Their resuls sugges ha he IT is posiively associaed wih Tobin s q, even afer conrolling for firm-specific facors such as R&D and adverising expendiures, and firm size, and indusry-specific facors such as indusry srucure, capial inensiy, and regulaion. Their work provides a foundaion for our research on he ineracion effecs of R&D and IT, in he conex of a significan shif in he naure of ITenabled innovaion in he Inerne era. 2.2 COMPLEMENTARITY EFFECTS OF IT AND R&D A review of he prior lieraure shows ha he impac of R&D and IT on firm marke value has been reaed separaely in a vas majoriy of he sudies. To he bes of our knowledge, none of hese sudies have explored he complemenary effec of IT and R&D invesmens on firm marke value. We believe ha, while he iniial focus on measuring he direc impac of IT in he iniial sages of indusry-wide compuerizaion in he 1990s was useful, i is even more imporan o develop a more nuanced undersanding of he impac of IT invesmens by virue of heir complemenariies wih discreionary invesmens in oher processes. R&D represens one such business process which enails making complemenary invesmens in informaion echnology.

8 We seek o undersand specific mechanisms hrough which IT can impac firm value hrough improvemens in he effeciveness of innovaion or R&D processes. The knowledgebased view of he firm suggess ha innovaion processes are criical o generae new knowledge in he execuion of R&D projecs (Kogu and Zander, 1992; Nonaka and Takeuch 1995). In order o leverage aci and explici knowledge ha resides wihin and ouside firms boundaries, firms mus build exensive capabiliies in idenifying and processing he informaion ha resides wihin he workplace and can frequenly involve exernal parners (Sakakibara, 2001; Cohen and Levinhal, 1990). Research on organizaional learning suggess ha his knowledge can be capured hrough IT-enabled rouines which allow firms o leverage ino he knowledge-base of is parner nework (Nicholls-Nixon and Woo, 2003). IT can help firms build high bandwidh channels wih heir parners and cusomers o sense aci and emerging cusomer/supplier informaion as a feedback loop ino heir R&D processes. For example, unil recenly, P&G used physical mockups of producs on shelves when hey engaged consumer focus groups or reailers in he developmen of new producs. Wih he adven of new virual realiy and simulaion ools, P&G now leverages is echnology ceners o provide hree-dimensional views of he sorefron which allows cusomers o provide immediae feedback on produc placemen and packaging decisions. Implemenaion of hese IT soluions has reduced he ime o creae a produc mockup from six weeks o a few days, and is now used in almos 80% of P&G s R&D iniiaives (Bloch and Lempres, 2008). This example shows how IT complemens invesmen in R&D processes and improves he execuion of R&D projecs. An alernae pahway of measuring he impac of IT is o focus on is role as an enabler of knowledge-generaing R&D processes. New ypes of IT, such as produc lifecycle managemen (PLM) sofware, enable produc design eams o collaborae 8

9 across iner-organizaional boundaries, gaher and share design requiremens, conduc design ieraions, verify and es produc designs, and faciliae final design hand-offs o oher deparmens (Adler, 1995; McGrah and Iansi 1998). Such web-based ools provide an informaion rich medium ha suppors collaboraion by faciliaing synchronous communicaion wihin and across R&D eams (Bardhan 2007). These ools also provide efficien daa sorage, elecronic rerieval and reuse of produc designs, and allow R&D eams o compress he overall produc developmen ime by reducing laency. Improvemens in design qualiy arise from he abiliy o share design ideas beween R&D eam members elecronically and conduc real-ime version conrol, which enables engineers o rack design defecs and implemen design changes more efficienly (Banker, Bardhan and Asdemir, 2006). A recen case sudy by Mendelson (2007) describes how echnology developmens are fueling he life sciences indusry hrough an informaion revoluion wih significan informaion processing, daa analysis, and sorage requiremens. New ypes of informaion echnologies can increase he efficiency and effeciveness of R&D in he life sciences indusry in many ways. While IT can accelerae drug discovery and developmen hrough produciviy improvemens, i can also significanly improve value chain opimizaion by enabling new developmen processes and organizaional changes. Mendelson (2007, page 20) suggess ha effecive use of informaion for decision making is an imporan driver of value chain opimizaion since increasing he success rae of drug discovery and esing can have subsanial benefis. For example, by inegraing informaion on he oxiciy of poenial leads while a drug arge is being evaluaed, an effecive decision suppor sysem can screen poenial arges ou of consideraion which would have oherwise progressed only o be rejeced in a laer esing phase. By enabling informaion exchange across differen pars of he drug developmen value chain, and managing

10 he coordinaion of workflow, IT suppors beer decision-making capabiliies across he enire value chain raher han wihin individual silos. High-performance compuing infrasrucure can also suppor high-hroughpu screening wherein lead-arge drug pairs can be analyzed simulaneously, hereby compressing he biological and chemical phases of he drug discovery value chain ino one parallel phase. This example provides one insance of he use of IT o suppor and improve he effeciveness and efficiency of R&D processes wihin he pharmaceuical indusry which can reduce overall developmen imes and lead o higher success raes in drug discovery and developmen. Our observaions on he effec of IT on innovaion are based on a field sudy ha we conduced wih a bio-pharmaceuical firm o moivae our hypoheses regarding he role of IT in improving he effeciveness and produciviy of R&D invesmens (see Appendix). A recen sudy by McKinsey repors ha pharmaceuical companies, which use IT in clinical rials processes increased heir overall produciviy by improving he speed, qualiy and coss associaed wih hese processes (Marwaha e al. 2007). Esimaed savings from IT-driven iniiaives ha improve he overall efficiency of clinical rials is esimaed o be in he range of $50 million o $100 million. Table 1 provides a useful perspecive on he role of IT in four areas of R&D relaed o drug discovery and developmen during clinical rials: improving resource allocaion for inegraed, enerprise-wide planning of clinical rials; beer daa managemen hrough elecronic case repor forms; enabling easier access o researchers hough elecronic daa capure ools by providing sandardized inerfaces; and providing greaer visibiliy across he clinical rials process by eliminaing bolenecks. Advances in developmen of high-hroughpu screening and simulaion sofware and developmen of unified IT sysems have grealy improved he efficacy of he drug discovery and developmen process (Mendelson, 2007). 10

11 These examples provide anecdoal evidence on how IT can be used o enable R&D processes and improve he execuion of R&D projecs in several indusries. In he nex secion, we draw upon exising heory o develop research hypoheses relaed o he complemenary effecs of R&D and I and heir impac on firm marke value. 3. RESEARCH HYPOTHESES Prior work by economiss has indicaed ha firms ha inves more in heir own R&D are beer able o exploi exernally-generaed knowledge compared o firms wih lower R&D expendiures (Mowery 1983, Evenson and Kislev, 1973). Cohen and Levinhal (1989) observe ha firms inves in R&D for wo reasons: o generae new knowledge and develop absorpive capaciy which represens he abiliy o recognize, assimilae and exploi knowledge embedded wihin a firm s business processes and rouines. While he economics lieraure has primarily focused on he role of innovaion in he heory of economic produciviy growh and social welfare, he finance lieraure has sudied invesor percepions o R&D announcemens and expendiures (Chan e al. 2001; Hirschey and Weygand 1985). Similarly, he accouning lieraure is sparse and has primarily focused on invesors cognizance of he capial aspecs of R&D wih he noable excepion of Lev and Sougiannis (1996) who sudied he effec of R&D capial on firm earnings and sock reurns for he period. In his sudy, we argue ha invesmens in R&D will be associaed wih a higher valuaion on growh opions ha are creaed hrough improvemens in he produc porfolio which allow firms o command greaer marke value. These fuure growh opions can be joinly enabled by pruden invesmens in R&D and IT which provide he innovaion capabiliies necessary for susained, compeiive advanage. Prior research on he impac of IT spending on firm performance has provided empirical evidence on he relaionship beween IT invesmens and improvemens in firm labor

12 produciviy, oal facor produciviy, and oupu growh, using longiudinal daa colleced in he lae eighies hrough he mid-1990s (Barua e al. 1995; Brynjolfsson and Hi 1996; Bresnahan e al. 2002). More recen sudies have focused on he pahways hrough which IT invesmens can creae value, using he resource-based view of he firm as he heoreical framework o sudy he impac of IT-enabled capabiliies on firm performance (Kohli and Devaraj, 2003; Banker e al. 2006; McAfee and Brynjolfsson 2008). A recen sudy by Aral and Weill (2007) has provided empirical evidence of he link beween sraegic and infrasrucure IT invesmens and various measures of firm performance, such as ne margin, new produc innovaion and Tobin s q. Mihas e al. (2010) show ha IT invesmens are posiively associaed wih firm profiabiliy hrough heir impac on sales growh and cos reducion iniiaives. However, hese sudies have focused on he direc impac of IT invesmens wihou considering he poenial complemenariies beween IT and oher discreionary invesmens such as R&D, and heir join effec on firm performance. An alernae pahway o measure he impac of IT is o focus on is role as a complemenary resource for innovaion-cenric R&D processes. In oher words, can IT make R&D invesmens more producive? An imporan issue in improving R&D produciviy is he capabiliy o faciliae seamless communicaion among virual produc design eams (Loch and Terwiesch 1998). New ypes of informaion echnologies, such as produc lifecycle managemen (PLM) sofware and collaboraion ools, provide an informaion rich medium ha suppors collaboraion by faciliaing synchronous communicaion wihin and across R&D eams (Bardhan 2007). Improvemens in design qualiy arise from he abiliy o share design ideas beween eam members elecronically and conduc real-ime version conrol, which enables engineers o rack design defecs and implemen design changes more efficienly. These 12

13 informaion echnologies play a criical role in enhancing he produciviy of R&D processes by reducing he overall ime o marke and produc developmen cos (Banker e al. 2006). The abiliy o exploi and disseminae knowledge is an imporan componen of innovaion. Cohen and Levinhal (1990, p. 128) coined he erm absorpive capaciy as he abiliy o recognize he value of new informaion, assimilae i, and apply i o commercial ends Their research sugges ha absorpive capaciy may be creaed as a resuls of a firm s R&D invesmens which allows he firm o assimilae and use new knowledge. However, absorpive capaciy refers no only o he acquisiion and assimilaion of new knowledge by a firm bu also o he firm s abiliy o exploi i. Hence, an organizaion s absorpive capaciy depends no only on is inerface wih he exernal environmen (cusomers, parners, suppliers, ec.) bu also on he ransfer of knowledge across and wihin organizaional silos ha may be quie removed from he original locus of innovaion. Cohen and Levinhal (1990) noe ha in order o undersand he sources of a firm s absorpive capaciy, i is necessary o focus on he srucure of communicaion beween he exernal environmen and he firm, as well as among sub-unis of he firm and he characer and disribuion of experise wihin he firm. We argue ha IT plays a cenral role in improving a firm s absorpive capaciy for innovaion aciviies by providing a plaform o exchange and disseminae informaion beween and across R&D eams ha may be disribued across differen funcions of he organizaion. IT miigaes he negaive effecs of disance by providing he richness of a communicaion medium for projec eam members o collaborae wih each oher (Massey and Monoya-Weiss, 2006). In he weny-firs cenury, global produc developmen is characerized by disribued eams in muliple counries ha are suppored by an IT-enabled digial infrasrucure ha suppors a collaboraive R&D process (Banker e al. 2006). To miigae he negaive effecs of disance and

14 ime on eam communicaion, companies have increasingly urned o IT ools. Based on heir sudy on he effec of compuer-mediaed communicaion (CMC) echnologies across a crosssecion of 277 high-ech firms in he US, Song e al. (2007) repored ha he CMC echnologies were associaed wih greaer knowledge disseminaion among R&D saff of produc developmen eams. IT invesmens help R&D managers improve he effeciveness of heir innovaion processes by providing greaer capabiliies o harness he knowledge embedded across firm boundaries and enable easier (and faser) access o criical produc daa ha help o reduce ime o marke and overall produc developmen coss. In a recen aricle, Brynjolfsson and Schrage (2009) describe how IT is revoluionizing he way ha innovaion ges done. As a caalys for complemenary changes, Brynjolfsson (2010) argues ha IT is seing off a revoluion in innovaion on four dimensions simulaneously: measuremen, experimenaion, sharing, and replicaion. By enabling all four of hese changes ogeher, companies are using IT o creae a new kind of R&D. For insance, Amazon.com uses IT-based experimenaion o conduc A/B experimens ess of is web pages ha deliver differen versions of he same page a he same ime o differen visiors, monior cusomer experience, and follow hrough. Google runs abou 200 o 300 experimens on any given day o evaluae cusomer reacion o differen ypes of online experimens which allows hem o conduc quasi-naural experimens on cusomer behavior which informs heir R&D processes. Wih respec o informaion sharing, Brynjolfsson (2010) highlighs Cisco where IT was used o develop a Wiki poral ha provided heir Macinosh users wih innovaive ways o insall new sofware and ge heir Macs o work wih he company s Linux priners. IT also makes i easier o replicae and scale up innovaions once hey have been idenified (Brynjolfsson and Schrage, 2010). For example, CVS Caremark Corporaion idenified 14

15 a novel way o implemen online prescripion ordering a one of is pharmacies ha resuled in a significan jump in cusomer saisfacion. CVS was able o use IT o quickly scale up and replicae his business process innovaion by embedding i in an enerprise echnology sysem and replicaed i o 4000 oher pharmacies wihin one year. In his manner, IT can foser a culure of experimenaion and innovaion where several ideas can be esed simulaneously o sudy real-ime changes in cusomer behavior. The knowledge gleaned from hese experimens can be used o design new innovaions ha can be scaled and replicaed quickly across muliple locaions and funcions hrough digial echnologies. In ligh of he complemenary effecs of IT on R&D innovaions, we argue ha IT moderaes he effeciveness of R&D by providing he echnology infrasrucure ha allows beer coordinaion among innovaion processes. In oher words, IT enables digial capabiliies ha have a posiive ineracion effec on he relaionship beween R&D and firm marke value. Hence, our sole research hypohesis can be framed as follows: The ineracion of R&D and IT invesmens has a posiive impac on firm marke value. Our concepual research model and hypoheses are described in Figure ECONOMETRIC MODELS We esimae he impac of firm- and indusry-specific facors on firm marke value using a series of hierarchical regression models. Firs, we esimae he effecs of he firm-specific and indusry-level conrol variables using a baseline model which is specified in equaion (1). TOBINQ α TOBINQ 5 = α + α ADVT α MKTSHARE α CAPEX 2 + α SALES + α INDCONC ε + α SIZE 4 + (1)

16 Firm-specific conrol variables include adverising inensiy (ADVT), capial expendiures (CAPEX), annual sales in dollars (SALES), and firm size measured as he logarihm of oal number of employees (SIZE). Indusry-specific conrol variables include he marke share (MKTSHARE) and indusry concenraion (INDCONC) as measured by he Herfindahl index. We also include he lagged value of he firm s Tobin s q raio in he previous year (TOBINQ -1 ) o conrol for he effec of prior year s Tobin s q on our esimaion in year. Noe ha he model in (1) represens a baseline of he variance in firms Tobin s q ha is explained by a combinaion of firm- and indusry-specific conrol variables. Nex, we inroduce firm-level R&D (RD) and IT inensiy which represens R&D and IT spending (in dollars) as a percenage of firm asses, respecively. Here, we esimae he impac of firm R&D and IT spending on Tobin s q, in addiion o he effec of he conrol variables esimaed specified in equaion (1). This model represens he main effecs of firm-level R&D and IT inensiy and is specified in equaion (2) as follows: TOBINQ β SIZE 6 = β + β RD + β TOBINQ β IT 2 + β ADVT + β MKTSHARE β CAPEX + β INDCONC β SALES 5 + ε + (2) Nex, we esimae he ineracion effecs of he R&D and IT on firm Tobin s q. Wih he addiion of he cross-produc of he R&D and IT variables o he main effecs model, he ineracion model is specified in equaion (3) as follows. TOBINQ + δ SIZE 7 = δ + δ RD 0 + δ TOBINQ δ IT δ RD xit + δ MKTSHARE δ ADVT + δ 10 4 INDCONC + δ CAPEX 5 + ε + δ SALES 6 (3) In muli-period panel daa analyses, i is well known ha he sandard errors of ordinary leas squares (OLS) esimaion can be biased when he residuals are no independen, resuling in 16

17 eiher over- or under-esimaion of he rue variabiliy of he coefficien esimaes. The possible dependence among he residuals may resul from wihin-firm auocorrelaion in he errors (imeseries dependence) or from wihin-year, across-firm correlaion in he errors (cross-secional dependence). To accoun for ime-series as well as cross-secional dependence of he residuals, we compue sandard errors ha are robus o wo-way clusering along boh dimensions of year and firm as described in Cameron, Gelbach and Miller (2006) and Peersen (2009). Following Cameron, Gelbach and Miller (2006), we separaely esimae hree OLS variance marices each wih one-way clusering by firm, year, and he inersecion beween firm and year, respecively. We hen compue he variance marix which is robus o wo-way clusering of firm and year by adding he firs wo variance marices and subracing he hird one. The cluser-robus sandard errors of he regression are he square roo of diagonal elemens of he resuling marix. Following Cameron, Gelbach and Miller (2006), he cluser-robus variance marix of regression coefficien esimaors is compued as (4) where each of he hree componens is esimaed separaely using he SAS saisical sofware. The cluser-robus sandard errors of he regression are he square roo of diagonal elemens of his marix. The firs componen is he OLS variance marix esimae compued by clusering on firm, where is an N x N indicaor marix wih ij h enry equal o one if he i h and j h observaion belong o he same firm, and equal o zero oherwise. The second componen is he OLS variance marix esimae compued by clusering on year where is an N x N indicaor marix wih ij h enry equal o one if he i h and j h observaion belong o he same year, and equal

18 o zero oherwise. The hird componen is he OLS variance marix esimae compued by clusering on he inersecion beween firm and year where is an N x N indicaor marix wih ij h enry equal o one if he i h and j h observaion belong o he same firm and year and equal o zero oherwise. This wo-way clusering procedure provides unbiased esimaes of he sandard errors of our regression models ha accoun for firm- and ime-specific effecs (Peerson, 2009) in our economeric esimaion. We noe ha Tobin s q measures he marke value of a firm based on is sock price a he end of each of he years in our eigh-year sample period, while decisions regarding discreionary invesmens such as R&D and IT spending are ypically made when annual budges are deermined a he beginning of he year. I is unlikely ha such invesmens may be codeermined wih firms marke valuaion leading o he problem of reverse causaliy beween our independen and dependen variables. In oher words, since IT and R&D invesmen decisions are made a he beginning of he year (or in he prior year in some cases) while Tobin s q measures he firm marke value a he end of he year, he poenial for endogeneiy is minimal. Furhermore, by incorporaing a lagged value of he dependen variable (i.e. TOBINQ -1 ) as an independen variable in our model specificaion in equaions (1) hrough (3), we explicily accoun for he possibiliy of reverse causaliy. Tha is, wha if firms wih greaer marke value are more likely o make greaer levels of R&D and IT invesmens which may lead o higher Tobin s q in he curren year? Our specificaion of a lagged dependen variable conrols for he impac of TOBINQ -1 on TOBINQ, and our coefficiens of R&D, IT and heir ineracion effec are esimaed afer accouning for his possibiliy. 1 1 We also used an insrumenal variable approach o esimae our models using wo-sage leas squares (2SLS) regressions following he approach models specified by Lev and Sougiannis (1996). Our coefficien esimaes are similar o he resuls repored in his paper and sugges ha endogeneiy is no of significan concern. 18

19 In addiion o he wo-way robus clusering approach described earlier in his secion, we also deployed a random effecs esimaion model wih auoregressive lag srucure (i.e. AR(1)) o esimae he main and ineracion effecs of R&D and IT on Tobin s q. Random effecs models uilizie variaion no only wihin each firm hrough ime bu also he variaion beween firms, and provide more efficien parameer esimaes while accouning for unobserved ime-invarian indusry or firm-level facors if hey are uncorrelaed wih explanaory variables in he model. We esimae our models in equaions (1) hru (3) using random effec esimaion echniques which views firm-specific consan erms as randomly disribued across cross-secional unis as in our panel daa (Balagi 2001; Greene 2000, pp ). We esimaed he random effecs models wih AR(1) errors using he AR(1) XTREG procedure in he STATA saisical sofware. We performed diagnosic checks o ensure he sabiliy of our resuls and did no deec any significan problems (Belsley, Kuh and Welsch 1980). We also checked for muli-collineariy in our models and ascerained ha variance inflaion facors were wihin he accepable hreshold. In addiion o esimaing he ineracion effec of RD x IT on Tobin s q, using he woway clusered error regression and random effecs wih AR(1) esimaion models described earlier, we also esimae a firs-difference model which incorporaes he absolue values of he dependen and independen variables a ime, as well as he change in hese values beween year and -1. In oher words, we esimae he impac of year-over-year change in our independen variables, including R&D, IT and heir ineracion effec, on change in firm Tobin s q. We posi ha he moderaing effec of IT on change in R&D spending beween years -1 and (i.e. ΔRD x ΔIT ) will be associaed wih a significan impac on change in firm Tobin s q (i.e. Δ(TOBINQ) ).

20 ΔTOBINQ γ CAPEX + γ INDCONC = γ + γ RD γ SALES + ε + γ ΔRD γ SIZE + γ IT γ ΔIT 4 + γ ΔRD + γ Δ( IND _ TOBINQ) 5 + γ xδit 11 + γ ADVT 6 MKTSHARE (5) ΔTOBINQ γ CAPEX + γ TOBINQ = γ + γ RD γ SALES + γ + γ ΔRD 2 9 INDCONC + γ SIZE + ε + γ IT γ ΔIT 4 + γ ΔRD + γ Δ( IND _ TOBINQ) 5 + γ xδit 11 + γ ADVT 6 MKTSHARE (6) Model (5) provides a simplified approach for esimaing he year-over-year change in Tobin s q due o a corresponding change in R&D and IT inensiies and he moderaing effec of IT on change in R&D spending (Hausman, Hall and Griliches, 1984). The coefficien γ 5 provides an esimae of he ineracion effec of change in R&D and IT on change in TobinQ beween years -1 and. This approach explicily accouns for bias due o missing variables in our esimaion models by modeling he effec of differences in he values of independen and dependen variables over a muli-year ime period (Forman, Ghose and Goldfarb, 2009). Our esimaion model in equaion (6) is very similar o equaion (5) wih he only difference being ha we explicily conrol for he effec of he lagged value of Tobin s q in he previous year (i.e. TOBINQ -1 ). This allows us o esimae he impac of he change in TOBINQ due o he ineracion effecs of corresponding changes in R&D and IT, and afer conrolling for he impac of he Tobin s q in he prior year. 5. RESEARCH DATA We use daa from hree sources in his sudy. Firs, we obained muli-year, archival daa on firm-level IT spending from an inernaional research firm ha is well-known for is IT daa and research services. This proprieary daabase was obained under a non-disclosure agreemen 20

21 ha proecs he confidenialiy of he daa. The daa was colleced hrough an annual survey ha is adminisered o chief informaion officers (CIOs) and oher senior IT execuives of large, global firms wih he goal of collecing objecive merics on IT invesmens. The research firm collecs archival IT invesmen daa, along wih oher IT invesmen-relaed informaion, as par of is annual, worldwide IT benchmarking survey. IT invesmens include all hardware, sofware, personnel, raining, disaser recovery, faciliies, and any oher coss associaed wih supporing he IT environmen, including he daa cener, deskop/wan/lan server, voice and daa nework, help desk, applicaion developmen and mainenance, finance, and adminisraion. In his sudy, we resric our locus of ineres o he subse of firms for which firm-level IT and R&D spending daa are available for an eigh-year span from 1997 o 2004 period. Daa on R&D invesmens as well as firm- and indusry-specific financial and accouning merics were consruced from he Sandard & Poor s COMPUSTAT daabase. We noe ha he adverising daa mainained in Compusa was limied because several firms do no repor heir adverising expendiures for a few years in our panel. Hence, we supplemened his daa using daa obained from he TNS Media Inelligence daabase which collecs firm-level adverising daa for he period from 2002 onwards. This daabase is widely used by researchers in markeing since i racks spending on new media including inerne adverising expendiures. Our panel daa se consiss of 208 firms for which we had firm-level daa on our main variables of ineres, namely R&D, IT, and adverising expendiures, as well as firm asses. All oher variables including Tobin s q, marke share, and indusry concenraion were consruced from available daa in Compusa. Our iniial sample conained 4,356 firm-year observaions from 567 firms colleced in he years 1997 hrough Nex, we remove: (1) 2,591 observaions wih missing IT spending in year, (2) 926 observaions wih missing R&D

22 spending in year, (3) 7 observaions wih missing daa for oal asses in eiher year or -1, (4) 50 observaions wih missing daa for marke value used in compuing Tobin s q in eiher year or -1. The resuling sample conains 962 observaions from 258 firms. We require a minimum of en observaions per year for each wo-digi NAICS indusry. This consrain ensures ha we have enough observaions per indusry o compue indusry-adjused R&D and IT spending measures. The size of our sample is reduced o 703 observaions from 210 firms. Finally, we remove indusries in which all firms repor zero R&D spending for year. Hence, he final sample consiss of 693 observaions from 208 firms spanning 8 years as shown in Table 2. The firms in our daa se span muliple indusries which include NAICS codes 32, 33, 42, 44, and 51. Five indusry caegories represen a high proporion of he firms in our sample: Chemical manufacuring, Meal and meal producs manufacuring, Machinery manufacuring, Compuer & elecronic producs manufacuring, and Transporaion equipmen manufacuring. Table 3 provides our sample disribuion by indusry. Definiions of all model variables are provided in Table 4. We noe ha Tobin s q, he dependen variable of ineres, is measured as oal firm liabiliies plus oal marke value of common equiy, divided by oal book value of asses of he firm (Hall, 1999; Bharadwaj e al. 1999). We sandardize R&D and IT inensiy by wo-digi NAICS by subracing he indusry mean from firm-level R&D (and IT) spending and dividing by he indusry sandard deviaion, respecively. While we repor boh unsandardized and sandardized values of R&D and IT spending as a percen of firm sales in our descripive saisics, we use only he sandardized values in our economeric analysis in he nex secion. Definiions of oher independen and conrol variables of ineres, such as adverising, capial expendiures, sales, firm size, marke 22

23 share, and indusry concenraion, are also provided in Table 4, and are consisen wih heir measuremen in he accouning and economics lieraure. We presen descripive saisics on our model variables in Panel A of Table 5. The mean and median values of Tobin s q are equal o 2.35 and 1.67, respecively. Similarly, mean R&D and IT spending inensiies are equal o 4.8% and 2.6% as a % of firm sales. By definiion, he mean, indusry-adjused values of R&D and IT inensiies are equal o zero. Similarly, mean adverising expendiures are approximaely 1.2% of firm sales, while capial expendiures accoun for 5.4% of oal firm asses and mean annual revenues are abou 1.06 imes he value of oal firm asses. Nex, we presen he Spearman/Pearson correlaion marix in Panel B of Table 5. The correlaion coefficiens beween he independen variables are generally below 0.40 and do no indicae he presence of muli-collineariy in our esimaion models. 6. RESULTS The esimaion resuls of our baseline model in equaion (1) are shown in he column iled Model 1 of Table 6. The resuls sugges ha he lagged value of Tobin s q in he previous year is he only deerminan of Tobin s q in he curren year. Firm-specific adverising and capial expendiures and indusry srucure variables such as marke share and indusry concenraion are no significan deerminans of Tobin s q. Our esimaion resuls of equaion (2), where we include our key variables of ineres, are presened in he column iled Model 2. We observe ha R&D has a significan, posiive associaion wih Tobin s q wih a coefficien value of 0.20 (p < 0.05). This resul suggess ha every 1% increase in firm R&D spending above is indusry mean is associaed wih a 0.2% increase in Tobin s q. Similarly, we observe ha IT spending has a posiive associaion wih Tobin s q wih a coefficien value of (p <

24 0.10). This resul suggess ha every 1% increase in firm IT spending above is sandardized indusry mean is associaed wih a 0.078% increase in Tobin s q. The las column of Table 6 provides he esimaion resuls of he ineracion effecs model where we include he ineracion erm, RD x IT, in equaion (3). The resuls indicae ha is R&D is no only significan a p < 0.01, bu is ineracion effec wih IT is also significan wih an esimaed coefficien value of (p < 0.05). In oher words, a 1% increase in he ineracion effec resuls in a corresponding 0.109% increase in firm Tobin s q. These resuls suppor our hypohesis wih respec o he posiive effec of he ineracion of R&D and IT on firm marke value. We also observe ha he increase in explained variance of he ineracion effecs model is saisically significan and greaer han he main effecs model (F-saisic = 7.47; p < 0.01). We noe ha while he main effec of IT spending in Model 3 is posiive, i is no saisically significan a p-value < All significan coefficiens are also economically significan based on he magniude of heir esimaes. Nex we esimae he same models in (1) (3) using random effecs esimaion echniques wih AR(1) sandard errors. The resuls of he random effecs esimaion are shown in Table 7 and are very consisen wih our earlier resuls in Table 6. We find ha he main effec of R&D and he ineracion effec of R&D x IT are boh posiive and srongly significan a p < We also noe ha ADVT has a posiive associaion wih firm Tobin s q in he main and ineracion effec models. Our panel daa esimaion using wo ypes of esimaion mehods confirm ha he ineracion effec of R&D and IT inensiy has a significan and posiive impac on firm Tobin s q. Nex, we esimae he firs-difference models specified in equaions (5) and (6) using random effecs esimaion and presen hese resuls in Table 8. The esimaed coefficiens 24

25 represen he impac of a uni change in he difference beween he values of independen variables on he change in Tobin s q beween he years -1 and. The firs column provides he esimaed coefficiens of he firs-difference model wihou conrolling for he effec of Tobin s q in he prior year, while we explicily conrol for TOBINQ -1 in he second column. Our resuls sugges ha he ineracion effec of Δ(R&D) x Δ(IT) is posiive and significan in boh models wih a p-value < When we conrol for he firm s Tobin s q in he prior year, we find ha he impac of a one uni change in he ineracion effec of Δ(R&D) and Δ(IT) represens a change of 0.204% in Tobin s q beween he years -1 and. Hence, our firs-difference esimaion resuls suppor our earlier resuls (using he acual levels of IT and R&D inensiy) wih respec o he moderaing effec of on he relaionship beween R&D and Tobin s q. Nex, in order o obain a numerical sense of he impac of R&D and IT on Tobin s q, we conduc a simple, univariae analysis o esimae he differences in Tobin s q across groups of firms wih differen R&D and IT spending profiles. We presen he resuls of our univariae ess in Table 9. We classify firms as high and low spenders based on heir R&D and IT invesmen profiles, relaive o heir wo-digi NAICS indusry median value. Accordingly, Panel A of Table 9 compares he Tobin s q for wo groups of firms: High R&D, high IT Spenders versus Low R&D, high IT spenders. We observe ha he difference in Tobin s q (z-value = 5.96) is saisically significan a a p-value < On average, our resuls indicae ha, high R&D and IT spenders ouperform he laer group in erms of heir marke-o-book raios. Similarly, in Panel B, we compare he differences beween wo groups of firms: High R&D, Low IT versus Low R&D, Low IT spenders. Again, he difference in Tobin s q across hese wo groups of firms is saisically significan wih he former ouperforming he laer. Panels A and B are paricularly ineresing since hey compare wo groups ha are characerized by differences in

26 heir levels of R&D spending while holding IT spending consan. In boh cases, we see ha higher R&D spending is associaed wih greaer Tobin s q. Nex, we conras he differences in performance across firms in Panels C and D. The primary difference beween he wo groups in panel C is due o Δ(IT) since boh groups are high R&D spenders. We observe ha he difference in Tobin s q (z-value = 4.79) is saisically significan a p-value < Similarly, he difference beween firms in panel D is due o Δ(IT) since boh groups exhibi low R&D spending. Again, he difference in Tobin s q (z-value = 5.90) is saisically significan a p-value < Panel E compares firms ha are high spenders of R&D and IT agains firms ha are low spenders. As expeced he former ouperforms he laer in erms of Tobin s q. Finally, in Panel F, we compare firms ha are high R&D, low IT spenders agains firms ha are low R&D, high IT spenders. Here, we observe ha he differences are no saisically significan since he effec of R&D and IT spending is mixed across groups. Our univariae resuls suppor our observaions from he regression esimaion in erms of he ineracion effec of R&D and IT spending on firm Tobin s q. We noe ha changes in R&D spending while holding IT consan, and vice versa, are significanly associaed wih greaer Tobin s q. 6.1 DISCUSSION We now discuss he several implicaions of our research. Firs, R&D spending has a significan posiive impac on firm Tobin s q which indicaes is associaion wih he marke expecaion for fuure growh. This resul is consisen wih prior finance and accouning research which shows ha R&D spending is valued a a premium by sock marke analyss in erms of heir conribuion o firm marke value (Lev and Sougiannis, 1996; Hall e al. 2007). Second, our 26

27 resuls sugges ha IT spending has a marginal, posiive impac on Tobin s q when we only esimae he main effecs of R&D and IT. However, our resuls also sugges ha afer accouning for ineracion effecs, he posiive impac of IT spending on Tobin s q disappears. Third, our analyses suggess ha IT moderaes he impac of R&D spending on Tobin s q, and his moderaion effec is posiive and saisically significan as observed in our relaively large sample of firms from a muliude of indusries and a ime period ha spans he do-com boom and bus cycle, i.e. pre- and pos-y2k. While prior research has focused solely on he individual effecs of R&D and IT, our resuls reveal a new dimension by measuring he ineracion beween hese wo ypes of complemenary invesmens. Our resuls suppor our hypohesis wih respec o he posiive, complemenary effecs of R&D and IT on Tobin s q, and provide empirical evidence o refue recen observaions in he praciioner lieraure ha quesion he value of R&D invesmens (Kandybihn and Kihn, 2004). Our resuls have several implicaions for pracice and research. Our resuls imply ha IT and R&D invesmens mus be well coordinaed and are mos effecive for firms ha adop a lean approach. The findings of our sudy imply ha he IT can play a criical role in enhancing firms marke value by improving he effeciveness and efficiency of R&D processes hrough heir effec as a lever o reduce coordinaion coss, increasing ransparency, and make beer resource allocaion decisions. From a managerial perspecive, an imporan implicaion of our sudy is o focus on he role of IT in improving innovaion capabiliies in R&D organizaions. Opporuniies o use IT in all phases of he produc innovaion lifecycle should be explored. For insance, IT sysems should be deployed o idenify and sense emerging cusomer needs which will feed he produc ideaion phase. IT can be used o improve projec governance hrough creaion of appropriae projec managemen sofware and disciplined sage-gae processes ha

28 sreamline projec workflows. From a research dimension, our sudy provides a fresh perspecive ino he drivers of firm financial performance and provides a new causal pah ha explains how IT can moderae he impac of R&D on firm marke value. I addresses a gap in he lieraure which has hereofore ignored he possibiliy of ineracion effecs when sudying he relaive impac of IT and R&D on firm performance. 7. CONCLUSIONS One pahway hrough which he ineracion of R&D and IT spending can impac firm marke value is hrough is impac on firm innovaion capabiliies. We argue ha invesmens in R&D and IT provide firms wih greaer absorpive capaciy o generae fuure growh hrough improvemens in heir produc developmen porfolio. Such fuure growh opions allow firms o command greaer marke value. Considering ha innovaion has been a fundamenal source of echnological change and produciviy growh during he las wo decades, here is no doub ha R&D is a key driver of such produciviy improvemens in he economy. In his sudy, we focus on he role of IT-enabled innovaion and he imporance of IT invesmens in moderaing he impac of R&D spending on firm Tobin s q. Considering ha quanum srides in compuing have been made during he IT revoluion since he early 1990s, i is imporan o (a) undersand he specific role of IT in improving he innovaion capabiliies of R&D processes, and (b) measure he join impac of R&D and IT invesmens on firm marke value. While he lieraure has focused on he impac of R&D or IT spending separaely, our sudy focuses on he criical quesion: Do complemenariies beween IT and R&D spending increase firm marke value? We propose and empirically es our research model using a relaively recen daa se ha reflecs he significan echnological changes in knowledge-based indusries since he firs wave of Inerne-based, commercial echnologies in he lae 1990s. Our sudy shows ha R&D has a 28

29 significan, posiive impac on firm marke value as measured using Tobin s q. Our resuls indicae ha IT spending moderaes he impac of R&D spending on Tobin s q. While R&D invesmens are considered risky, smar IT invesmens may lower he volailiy of hese risks by providing greaer ransparency ino execuion of R&D projecs, and improving he coordinaion across muliple projecs. Fuure research will focus on evaluaing wheher hese effecs are sronger among firms in IT-inensive indusries compared o non-it inensive indusries. Our sudy has several limiaions. Firs, due o he secondary naure of he daa, we do no have insigh ino he specific ypes of IT invesmens ha firms in our sample have made during he ime horizon in our sudy. Fuure research should focus on he ypes of IT invesmens made including he allocaion beween various componens of IT spending such as sofware applicaions, consuling, and infrasrucure invesmens. Second, since produc developmen imes in some indusries are long, i is possible ha he ime period in our sudy did no allow us o observe he full effecs of R&D and IT spending. Third, we use daa on large, global firms. This limis generalizaion of our findings o similar firms, and furher exploraion wih daa from smaller firms is needed.

30 References Adler, P.S. Inerdeparmenal inerdependence and coordinaion: The case of he design / manufacuring inerface, Organizaion Science (6:2), 1995, pp Aral, S., and Weill, P. IT Asses, Organizaional Capabiliies & Firm Performance: How Resource Allocaions and Organizaional Differences Explain Performance Variaion, Organizaion Science (18:5), 2007, pp Arkal G., Bardhan, I.R., and Krishnan, V., Informaion echnology risk and reurn rade-off: New evidence from recen daa 2009, Working Paper, The Universiy of Texas a Dallas, Richardson, TX. Banker, R.D., Bardhan, I.R., and Asdemir, O. Undersanding he Impac of Collaboraion Sofware on Produc Design and Developmen, Informaion Sysems Research (17:4), 2006, pp Bardhan, I.R. Toward a Theory o Sudy he Use of Collaboraive Produc Commerce for Produc Developmen, Informaion Technology and Managemen (8:2), 2007, pp Barua, A., Kriebel, C.H., and Mukhopadhyay, T. An Economic Analysis of Sraegic Informaion Technology Invesmens, MIS Quarerly (15:3), 1991, pp Ben-Horim, M., and Callen, J.L. The cos of capial, Macaulay s duraion, and Tobin s q, Journal of Financial Research 12, 1989, Bharadwaj, A.S., Bharadwaj, S.G., and Konsynsk B.R., Informaion Technology Effecs on Firm Perormance as measured by Tobin s q Managemen Science (45:6), 1999, Bloch, M., and Lempres, E.C. From Inernal Service Provider o Sraegic Parner, McKinsey Quarerly, July 2008, pp Bresnahan T. F., Brynjolfsson, E., and Hi, L.M. Informaion Technology, Workplace Organizaion, and he Demand for Skilled Labor: Firm-level Evidence, Quarerly Journal of Economics, 117 (2002), Brynjolfsson, E., and L. Hi, Informaion Technology as a Facor of Producion: The Role of Differences among Firms, Economics of Innovaion and New Technology, Special Issue on Informaion Technology and Produciviy Paradox, (3:4), 1995, Brynjolfsson, E., and Hi, L. Paradox Los? Firm-level Evidence on he Reurns o Informaion Sysems Spending, Managemen Science (42:4), 1996, pp Brynjolfsson, E., The four ways IT is revoluionizing Innovaion MIT Sloan Managemen Review, 2010 (51:3), 2010, pp Brynjolfsson, E., and Schrage, M., The new faser face of innovaion MIT Sloan Managemen Review Execuive Adviser, Augus 17, 2009, hp://sloanreview.mi.edu/execuiveadviser/aricles/2009/3/5139/he-new-faser-face-of-innovaion/. Cameron, A.C., Gelbach, J.B., and Miller, D.L., Robus inference wih muli-way clusering 2006, Working Paper, Universiy of California, Davis, CA. 30

31 Carr, N.G. Does IT Maer? Informaion Technology and he Corrosion of Compeiive Advanage, Harvard Business School Press, Chan, L.K.C., Lakonishok, J., and Sougiannis, T. The Sock Marke Valuaion of Research and Developmen Expendiure, Journal of Finance (56:6), 2001, pp Chauvin, K.W., and Hirschey, M. Adverising, R&D Expendiures, and he Marke Value of he Firm, Financial Managemen (22:4), 1993, pp Cohen, W.M., and Klepper, S. A Reprise of Size and R&D, The Economic Journal (106), 1996, pp Cohen, W.M., Levin, R.J., and Mowery, D.C. Firm Size and R&D Inensiy: A Reexaminaion, The Journal of Indusrial Economics (15:4), 1987, pp Cohen, W., and Levinhal, D. Absorpive Capaciy: A new perspecive on learning and Innovaion, Adminisraive Science Quarerly (35), 1990, pp Englander, A., Evenson, R., and Hanazak M. R&D, Innovaion, and he Toal Facor Produciviy Slowdown, OECD Economic Sudies (11), Auumn 1988, pp Elie, J.E. R&D and Global Manufacuring Performance, Managemen Science (44:1), 1998, pp Evenson, R.E. and Kislev, Y., Research and Produciviy in Whea and Maize, Journal of Poliical Economy, 81 (1973), Foray, D., Hall, B., and Mairesse, J. Pifalls in esimaing he Reurns o corporae R&D using accouning daa Firs European Conference on Knowledge for Growh, Ocober 8-9, Griliches, Z. Produciviy, R&D, and he Daa Consrain, American Economic Review (84:1), 1994, pp Griliches, Z., and Mairesse, J. Produciviy and R&D a he Firm Level, In Zvi Griliches, ed., R&D, Paens, and Produciviy, Chicago: Universiy of Chicago Press, Hirschey, M. and Weygand, J. J., Amorizaion Policy for Adverising and Research and Developmen Expendiures, Journal of Accouning Research, 23 (1) (1985), Hall, B., and Mairesse, J. Exploring he Relaionship beween R&D and Produciviy in French Manufacuring Firms, Journal of Economerics (65:1), 1995, pp Hall, B.H. Innovaion and marke value Naional Bureau of Economic Research, Working paper 6984, , Cambridge, MA. Hi, L. and Brynjolfsson, E., Produciviy, Profi and Consumer Welfare: Three Differen Measures of Informaion Technology's Value, MIS Quarerly 20 (2) (1996), Kandybin, A.., and Kihn, M. Raising Your Reurn on Innovaion Invesmen, Resilience Repor by Booz Allen Hamilon Kogu, B., and Zander, U. Knowledge of he firm, combinaive capabiliies, and he replicaion of echnology, Organizaion Science (3), 1992, pp Kohl R., and Devaraj, S. Performance Impacs of Informaion Technology: Is Acual Usage he Missing Link? Managemen Science (49:3), 2003, pp

32 Lev, B. and Sougiannis, T. The Capializaion, Amorizaion and Value-Relevance of R&D, Journal of Accouning and Economics (21:1), 1996, pp Loch, C.H. and C. Terwiesch, Communicaion and Uncerainy in Concurren Engineering, Managemen Science, 44 (8) (1998), Marwaha, S., Pail, S., and Singh, N. Using IT o Speed Up Clinical Trials, McKinsey Quarerly, Spring 2007, pp Massey, A.P., M.M. Monoya-Weiss Unraveling he emporal fabric of knowledge conversion: a model of media selecion and use. MIS Quarerly 30(1), McAfee, A., and Brynjolfsson, E. Informaion Technology ha makes a Compeiive Difference Harvard Business Review, July-Augus 2008, pp McGrah, M. E. and Iansi M. Envisioning IT-Enabled Innovaion, PRTM Insigh, Mendelson, H., Value disciplines: Three case examples Sanford Graduae School of Business, 2007, Palo Alo, CA. Mihas, S., Tafi A., Bardhan, I., and Goh, J.M. Informaion Technology and Firm Profiabiliy: Mechanisms and Empirical Evidence, Working Paper, Universiy of Maryland, Mial, N. and Naul, B.R. Invesmens in Informaion Technology: Indirec Effecs and Informaion Technology Inensiy Informaion Sysems Research (20:1), 2009, Morbey, G.K. R&D: Is Relaionship o Company Performance, Journal of Produc Innovaion Managemen (5), 1988, pp Nicholls-Nixon, C. L., and Woo, C.Y. Technology sourcing and oupu of esablished firms in a regime of encompassing echnological change, Sraegic Managemen Journal (24:7), 2003, pp Nonaka, I. and Takeuch H. The Knowledge-creaing Company, Oxford Universiy Press, New York, NY, Peersen, M.A., Esimaing Sandard Errors in Finance Panel Daa Ses: Comparing Approaches Review of Financial Sudies (22:1), 2009, Ra A., Panayakun R., and Panayakun N. Technology Invesmens and Business Performance, Communicaions of he ACM (40:7), 1997, pp Sakakibara, M., and Branseer, L. Do Sronger Paens Induce More Innovaion? Evidence from he 1988 Japanese Paen Law Reforms RAND Journal of Economics (32:1), Spring 2001, pp Sambamurhy, V., A. Bharadwaj, and V. Grover, Shaping Agiliy hrough Digial Opions: Reconcepualizing he Role of Informaion Technology in Conemporary Firms, MIS Quarerly (27:2), 2003, Song, M., H. Berends, H. van der Bij, M. Weggeman The effec of IT and co-locaion on knowledge disseminaion. Journal of Produc Innovaion Managemen 24, Smirlock, M., Gilligan, T., and Marshall, W., Tobin s q and he srucure-performance relaionship: commen American Economic Review 74 (December), 1984,

33 Table 1. The Role of Informaion Technology in Drug Discovery and Developmen Design, Planning Sar -Up Managing Trial Closeou Repor Key aciviies benefiing from IT Designing proocols Creaing regulaory documens Disribuing drugs, informaion Daa collecion, Selecing sie Monioring progress, adverse evens Tracking paien enrollmen Tracking clinical response forms Enering, verifying daa Processing clinicalresource forms Reconciling invesigaor s queries Reporing Oucomes IT Focus Clinical Daa Managemen Daabase of invesigaors o suppor design of elecronic forms Sandard inerface o inegrae hird pary sysems Greaer compuing power allows scieniss o design randomized rials Auomaed daa checks o minimize queries Paien Safey Documen Managemen Clinical Trials Managemen Projec Managemen Modular design, consrucion of consen and case repor forms Sandard daa models o work wih vendors Sysem o conver sudy designs o elecronic forms and daabase wih minimal rework Real-ime monioring Single daa reposiory wih version conrol, workflow managemen Elecronic invoicing Auomaed drug supply work flow Paien mgm. More effecive rials managed hru large-scale daabases and simulaion sofware Repor builder Source: Marwaha, Pail & Singh, 2007.

34 Figure 1. Concepual Research Model Noe: All model variables have been indusry-adjused by subracing he indusry mean and dividing by he indusry sandard deviaion. Table 2. Sample Selecion Mehodology Firm-year observaions obained from he original sample covering years from 1997 o 2004 Firm-Year Observaions Firms 4, Afer: Removing hose wih missing IT spending daa in year 1, Removing hose wih missing R&D spending daa in year 1, Removing hose wih missing Toal Asses in eiher year or 1, year -1 Removing hose wih insufficien daa for compuing Tobin s Q in year or -1 Requiring a leas en observaions per year for each indusry a wo-digi secor level using NAICS codes Removing indusries in which all firms repor zero R&D spending for year