Sustainable Value Creation: The role of IT innovation persistence

Transcription

1 Associaion for Informaion Sysems AIS Elecronic Library (AISeL) AMCIS 2009 Proceedings Americas Conference on Informaion Sysems (AMCIS) 2009 Susainable Value Creaion: The role of IT innovaion persisence Changling Chen Universiy of Waerloo, Jee-Hae Lim Universiy of Waerloo, Theophanis C. Sraopoulos Universiy of Waerloo, Follow his and addiional works a: hp://aisel.aisne.org/amcis2009 Recommended Ciaion Chen, Changling; Lim, Jee-Hae; and Sraopoulos, Theophanis C., "Susainable Value Creaion: The role of IT innovaion persisence" (2009). AMCIS 2009 Proceedings. 82. hp://aisel.aisne.org/amcis2009/82 This maerial is brough o you by he Americas Conference on Informaion Sysems (AMCIS) a AIS Elecronic Library (AISeL). I has been acceped for inclusion in AMCIS 2009 Proceedings by an auhorized adminisraor of AIS Elecronic Library (AISeL). For more informaion, please conac elibrary@aisne.org.

2 Chen e al. Susainable value creaion: role of IT innovaion Susainable value creaion: The role of IT innovaion persisence Changling Chen Jee-Hae Lim Universiy of Waerloo Universiy of Waerloo clchen@uwaerloo.ca jh2lim@uwaerloo.ca Theophanis C. Sraopoulos Universiy of Waerloo sraop@uwaerloo.ca ABSTRACT: In his research, we invesigae he role of alernaive IT innovaion sraegies (sysemaic, opporunisic, and non-it innovaion) on earnings persisence. Building on agiliy heory we argue ha sysemaic IT innovaion leads o susainable value creaion and abiliy o recover from negaive earnings and recession. Using a sample of large US firms we find ha good (bad) performance of sysemaic IT innovaors is more persisen (ransiory) han non-sysemaic IT innovaors, and are more likely o recover from recession. We conclude ha sysemaic IT innovaors are beer prepared o deal wih he modern hypercompeiive environmen. Keywords Sysemaic IT innovaion, susainable value creaion, earnings persisence, agiliy heory INTRODUCTION Wih he major indusrialized economies already in a recession (Economis 2008), managers are looking for robus sraegies o navigae heir companies hrough he earnings sorm. Guidelines from consuling firms and journaliss abound and mos of hem involve suggesions regarding IT sraegies (Cummings 2008; McKinsey Quarerly 2008). Undersanding wha IT sraegies migh work well during a recession or could help a company bounce back from a sring of negaive earnings is criical for managers and invesors. This knowledge could be ranslaed ino persisen high profiabiliy (earnings) and susainable compeiive advanage. In economic and accouning lieraure, earnings persisence represens he abiliy of firms o preserve heir preferenial compeiive posiion (high earnings) for considerable periods of ime (Sigler 963; Lev 983). Prior sudies have idenified various economic characerisics such as barriers-o-enry, size, produc-ype, capial inensiy, and R&D ha drive earnings persisence (Sigler, 963; Lev, 983; Baginski e al., 999; Ashana and Zhang 2006). Approaching his lieraure from an IT perspecive, we see wo opporuniies o make a posiive conribuion. Firs, IT spending has been rising and i has become he larges capial spending iem for mos companies (Weill e al. 2002). In spie of his and he recogniion of IT enabled sraegies as a source of susainable compeiive advanage (Sambamurhy e al. 2003; Piccoli and Ives 2005), our search of he exan lieraure shows no evidence of prior research linking IT (resources and capabiliies and/or sraegy) o earnings persisence. Second, while prior research on earnings persisence has looked a he role of cerain economic facors on a company's abiliy o bounce back from a sring of negaive earnings (Chamber 2006), no aemp has been made o analyze and compare earnings persisence beween a recession and a recovery period. Hence, he main quesion of our sudy is: Wha is he role of IT on earnings persisence and how his differs during a business cycle? A recen sudy (Sraopoulos and Lim 2007), has defined IT innovaion persisence as a firm s abiliy o differeniae iself from is peers and sand ou in a crowd of compeiors hrough IT innovaions over an exended period of ime. Approaching IT innovaion as a zero-sum game, Sraopoulos and Lim (2007) show ha IT innovaion is a heerogeneously disribued and imperfec mobile capabiliy, hence a source of susainable compeiive advanage. They aribue persisence o he pah dependence of he IT innovaion. This means ha companies ha have been able o disinguish hemselves from heir compeiors hrough IT innovaion in he pas are more likely o repea his in he curren period. In our sudy, we argue ha

3 Chen e al. Susainable value creaion: role of IT innovaion IT innovaion persisence is linked o earnings persisence, and companies ha have developed such an IT innovaion capabiliy are more likely o deal wih recessions and drops in earnings more efficienly han companies ha have no developed similar capabiliies. Our economeric analysis shows ha, a leas wihin he group of large US firms in our sample, he IT capabiliy leads o no only high profiabiliy (abnormal reurn on earnings) bu also high abiliy o recover from earnings drops and economic downurn, as indicaed by he change in heir earnings persisence beween he good and bad financial performance years, or beween he recession and recovery periods. The res of he paper is srucured as follows. In he following secion we review he work on IT business value and IT innovaion. Thereafer, we discuss he heoreical framework used o govern our discussion and o idenify he measures considered in his sudy. The research mehodology and he resuls of an empirical analysis are hen presened. Finally, he implicaions for praciioners and researchers are oulined and conclusions and limiaions of he sudy are presened. LITERATURE REVIEW AND HYPOTHESES DEVELOPMENT Firms oday face inense compeiion in an increasingly urbulen and dynamic global marke (D Aveni 994; Sambamurhy 2000). This manifess in higher volailiy of earnings, increased incidens of performance slumps among Forune 500 companies, and sories of companies wih superior pas performance record having difficuly delivering consisenly superior earnings in recen years (Hamel and Välikangas 2003). This rend has serious implicaions for managers and invesors since earnings persisence (a measure of he company's susainable compeiive advanage) is more desirable han ransiory earnings (a emporary compeiive advanage). In economic and accouning lieraure earnings persisence has been linked o various economic characerisics such as barriers-o-enry, size, produc-ype, capial inensiy, and R&D (Sigler, 963; Lev, 983; Baginski e al., 999; Ashana and Zhang 2006). In his sudy we inroduce a company s approach o IT innovaion (IT innovaion sraegy) as a deerminan of earnings persisence. In erms of IT innovaion sraegies firms can be classified as innovaive or non-innovaive. Among he group of IT-innovaive firms, a disincion has been made beween hose ha are sysemaic and hose ha are opporunisic innovaors in erms of aribues and heir approach o IT innovaion sraegy (Swanson and Ramiller 997; Sraopoulos and Lim 2007). Sysemaic IT innovaors seem o be aking a long-erm and sraegic approach o IT innovaion. Opporunisic IT innovaors, on he oher hand, will end o ake a shor-erm approach driven by fashion or fads (Abrahamson and Rosenkopf 997; Swanson and Ramiller 997). Using a cross-secional daa se of large US firms from , Sraopoulos and Lim (2007) have shown ha IT innovaion capabiliy is heerogeneously disribued and no easy o imiae, and over an exended period of ime, i is more likely for sysemaic IT innovaors o coninue being sysemaic innovaors han for opporunisic and non-it innovaors o become sysemaic innovaors. In he following paragraphs we build on he IT innovaion persisence of he sysemaic IT innovaors in order o esablish he heoreical link beween IT innovaion sraegies and earnings persisence. Hypoheses Developmen H: IT Innovaion Sraegies and Value Creaion. I is an empirical fac ha some firms ouperform ohers and sraegy heories ry o explain his (Powell 2003). In fac he purpose of firm sraegies and measure of success is is abiliy o creae value. This occurs when a firm generaes earnings, which are greaer han is cos of capial. Residual income (abnormal ROE) is such a measure of performance ha capures value creaion ha has been adoped in accouning (Felham and Ohlson 995; Ohlson 200; Cheng 2005) and sraegic managemen lieraure (Hawawini e al. SMJ 2005). Since he main objecive of our sudy is o consider he effec of IT innovaion sraegies, i makes sense ha we should consider a measure of performance ha reflecs value creaion. Following he suggesion of Kohli and Grover (2008; p. 30) we sar by uncovering he capabiliies required for value and hen idenify wha i akes o build hem. Business agiliy is a criical facor for success in oday s dynamic compeiive environmen (D Aveni 994; Brown and Eisenhard 997; Sambamurhy 2000; Sambamurhy e al. 2003). In our sudy we will adop he following definiion for business agiliy proposed by van Ooserhou e al. (2007: 53-54): "Business agiliy is he abiliy o sense highly uncerain exernal and inernal changes, and respond o hem reacively or proacively, based on innovaion of he inernal operaional processes, involving he cusomer in In he res of our discussion we use he erms sysemaic/opporunisic IT innovaors and sysemaic/opporunisic IT innovaion sraegy inerchangeably. 2

4 Chen e al. Susainable value creaion: role of IT innovaion exploraion and exploiaion aciviies, while leveraging he capabiliies of parners in he business nework." (van Ooserhou e al. 2007) This definiion capures he hree dimensions of agiliy: cusomer, parnering, and operaional agiliy. Sambamurhy e al. (2003) argue ha firms ha can bundle hese hree dimensions of agiliy are more likely o ouperform heir compeiors by launching a higher number and more complex compeiive acions. Piccoli and Ives (2005) claim ha sraegic IS research has been invigoraed by recen sudies (Sambamurhy e al. 2003; Weill e al. 2002) ha have linked IT wih business agiliy. Building on he work of Sambamurhy e al. (2003) and Bharadwaj e al. (999), we see he following link beween dimensions of business agiliy and IT capabiliies: The building of exernal IT linkages capabiliy refers o connecions beween he firm and business parners ha nurure he sharing of knowledge (Konsynski and McFarlan 990; Zaheer and Venkaraman 994) and prepare he firm o explore IT opporuniies wihin he firm as well as wih exernal parners (Feeny and Wilcocks 998; Bharadwaj e al. 999). This capabiliy is he foundaion for he developmen of cusomer and parner agiliy. A company s operaional agiliy and abiliy o innovae on is inernal operaional processes will res on he building of IT business parnerships, i.e., abiliy o foser relaionships beween IT providers and IT users. IT business parnerships are criical for developing IT innovaion (Reich and Kaars-Brown 2003). Finally, a firm s abiliy o sense highly uncerain exernal and inernal changes, and respond o hem reacively or proacively will be relaed o he Business IT sraegic capabiliy. Swanson and Rasmiller (997) argue ha firms ha sysemaically innovae wih IT seem o be more auned o he idiosyncrasies of heir inernal environmen (IT business parnership capabiliy) as well as he exernal environmen (exernal IT linkages), and hey end o promoe ineracive learning (IT business parnership capabiliy). Such IT-innovaive firms are also more likely o succeed if senior managemen perceives IT innovaion as an imporan capabiliy and suppors i (IT managemen and IT sraegic hinking capabiliy). Therefore companies ha have been sysemaic IT innovaors are more likely o have hese capabiliies for developing business agiliy han opporunisic and non-it innovaors. Thus we develop he following hypohesis: H: Sysemaic IT innovaive firms are more posiively associaed wih abnormal ROE han opporunisic and non- IT innovaive firms. H2: IT innovaion sraegies and susainable value creaion Value creaion ha resiss he erosion from compeiors acions is consisen wih Porer s (985) definiion of susainable compeiive advanage, and i should be disinguished from occasional spikes in value creaion ha are associaed wih a emporary compeiive advanage. Managers and invesors recognize he imporance of earnings persisence, a manifesaion of susainable value creaion (Collins and Kohari, 989; Barh e al., 999; Ashana and Zhang 2006). However, earnings persisence is differenly affeced by good news and bad news; good news is defined as posiive abnormal ROE, and bad news is defined as non-posiive abnormal ROE. Prior sudies have linked earnings persisence o various economic characerisics and have explored heir conribuion o persisence of good news and ransiory bad news (Sigler, 963; Lev, 983; Chamber, 996; Baginski e al., 999; Ashana and Zhang 2006). Here we inroduce IT innovaion sraegies as an explanaory variable of persisence of posiive earnings and ransiory negaive earnings. According o Resource Based View (RBV), only resources and capabiliies ha are valuable, heerogeneously disribued, and imperfecly mobile will confer a susained compeiive advanage (Barney 99; Maa e al. 995). Sraopoulos and Lim (2007) have shown ha IT innovaion is a pah dependen process, and hence no easily replicaed. Companies ha have been able o disinguish hemselves from heir compeiors hrough IT innovaion, and hey have been able o repea his over an exended ime period (i.e., Sysemaic IT innovaors) are more likely o coninue being sysemaic innovaors han for opporunisic and non-it innovaors o become sysemaic innovaors. Therefore, sysemaic IT innovaion sraegy is more likely o lead o susainable value creaion (posiive earnings) han non-sysemaic IT innovaion sraegies. In addiion, we have seen ha sysemaic IT innovaors are more likely o have developed he IT capabiliies (IT business parnerships, exernal IT linkage, as well as IT managemen and IT sraegic hinking) needed for business agiliy. Firms wih high levels of cusomer, parnering, and operaional agiliy are beer posiioned o deec and exploi marke opporuniies (Sambamurhy e al. 2003). Therefore, sysemaic innovaors are more likely and beer prepared o respond o negaive earnings shocks han opporunisic and non-it innovaors. Thus, we conjecure: H2a: Sysemaic IT innovaive firms are more likely o have posiive persisen abnormal ROE (good news) han opporunisic and non- IT innovaive firms. 3

5 Chen e al. Susainable value creaion: role of IT innovaion H2b: Sysemaic IT innovaive firms are more likely o have less persisen (ransiory) negaive abnormal ROE (bad news) han opporunisic and non- IT innovaive firms. H3: IT innovaion sraegies and susainable value creaion during Business Cycles An aricle wih he ile "How o Think Sraegically in a Recession - Sraegy for he Recession" was published in Harvard Managemen Updae shorly afer he 9/ aacks. The aricle (Zook and Rigby 200) suggesed ha managers should consider o wha exen heir company was prepared o deal wih he following scenarios/challenges: Is your company prepared o deal wih supply chain disrupions? Will you be able o survive in a price war (prices drop drasically)? Is your company prepared for a global recession? While he lis is no exhausive, i cerainly capures he main challenges ha a company is likely o face during a recession. Failure o deal wih hese challenges would have a derimenal effec on he company abiliy o creae and susain value. In order o deal wih he firs of hese challenges, companies will need o focus on wha Weill e al. (2003) call supply-side iniiaives (parnering agiliy) and operaional iniiaives (operaional agiliy). Leveraging heir exernal IT linkage capabiliies and business IT parnership capabiliies, companies can reach he level of agiliy ha Dell aained in he days afer he 9/ aacks. Dell, which has buil one of he world's bes supply chain neworks, charered an airliner o fly pars from Taiwan o is Texas facory, ran facories day and nigh, and convered hree 8-wheel rucks ino mobile echnology and suppor faciliies in order o supply 24,000 compuers o New York Ciy and Washingon, D.C (Zook and Rigby 200) The second and hird challenge relae o he following wo inerwined paerns ha companies are more likely o face during an economic recession: Firs, cusomers end o become more price sensiive. Second, companies are more likely o engage in price wars (Chou and Chen 2004). During a period of economic expansion, boh high- and low-cos producers can be profiable. However, during he recession, when prices are falling, only he low-cos producers will be profiable (Zook and Rigby 200). Inel has cu prices on is microprocessors by 35%. Dell halved is prices and sill makes money no so for some of is compeiors. The speed of he economy's decline from is high poin eigheen monhs ago will soon underscore he imporance of relaive cos posiion. Companies ha have developed wha Weill e al. (2003) call demand-side iniiaives (cusomer agiliy) and operaional iniiaives (operaional agiliy) are more likely o succeed n such an environmen of increased price elasiciy of cusomers and price wars. Operaional iniiaives leading o improved cos posiion will be possible for companies ha can leverage heir business IT parnership capabiliies and exernal IT linkages wih parners. The former in order o exploi opporuniies o reduces inefficiencies in heir value chain and idenify opporuniies o cu cos, and he laer o improve he efficiency of heir supply chain. However, hese operaional changes should be done in way ha does no sacrifice he service ha he company offers o is cusomers. Such cusomer agiliy will be feasible for companies ha can leverage he exernal IT linkages wih cusomers. As we have seen, Sysemaic IT innovaors are more likely o have developed he IT capabiliies (IT business parnerships and exernal IT linkage) needed for a company o respond o he challenges of an economic recession. Therefore, sysemaic innovaors are more likely and beer prepared o weaher an economic recession han opporunisic and non-it innovaors. Therefore we conjecure ha: H3a: Sysemaic IT innovaive firms are more likely o have persisen abnormal ROE in recovery period han opporunisic and non- IT innovaive firms. H3b: Sysemaic IT innovaive firms are more likely o have less persisen (ransiory) abnormal ROE in recession period han opporunisic and non- IT innovaive firms. METHOD 3. Sample To idenify firms approach o IT innovaion wihin an indusry, we followed he mehodology used by Sraopoulos and Lim (2007). We obained daa from InformaionWeek 500 (IW500) from 997 o Conducing a deailed survey of IT 2 This period was chosen because i conains he approximae boundaries of a complee cycle in managers percepion regarding innovaion wih IT (Sraopoulos and Lim 2007). The period was marked by a sequence of evens ha one can 4

6 Chen e al. Susainable value creaion: role of IT innovaion execuives, IW500 publishes an annual lis of he larges US companies ranked by heir innovaive use of IT, businessechnology sraegies, and deploymen of invesmens in IT archiecure, infrasrucure, business, and e-business applicaion. The end resul is an annual lis of IW500 firms ha are classified as IT innovaors due o heir demonsraion of a consisen paern of echnological, procedural, and organizaional innovaion. (InformaionWeek 500 Research Repors various issues ). The publicly available daa from IW500 source have been commonly used in a number of sudies in he pas (Bharadwaj e al. 999; Bharadwaj 200; Brynjolfsson and Hi 996; Hi and Brynjolfsson 996; Lichenberg 995; Sanhanam and Harono 2003; Sraopoulos and Lim 2007). [Inser Table ] Using he lising of IT innovaive firms from IW500, he iniial sep produced 4,000 records of firms (firm-enries) ha have aained he IT innovaor saus on any year from However, among hose firms here were 588 enries associaed wih privae, non-profi, foreign, or bankrup firms for which we would no be able o produce financial performance daa. Eliminaing hese enries reduced our records o 3,42 corresponding o,067 unique firms from IW500. For our second sep, we used daa from Hoover s Handbook of American Business (Hoovers), which has involved in he preparaion of he IW500 repor as a provider of corporae financial daa, as well as idenified he lis of non-it innovaive firms ha are also perceived o be he op compeiors for each of he IT-innovaive firms (Sraopoulos and Lim 2007). In fac, we mached he op hree compeiors for each IT innovaive firms; specifically, a furher manual review was conduced for eliminaing 07 compeiors wih missing eiher size or indusry value from Hoover s. The final daa conain 960 firms by 622 IT-innovaive firms from IW500 and 338 non-it- innovaive firms from Hoover s. We resric he sample o include firm-year observaions wih financial daa from Compusa. A cross-mach wih he Global Vanage Key (GVKey) file from Compusa resuled in he eliminaion of anoher 26 firms. Thus, 934 unique firms remained in our sample, giving us a oal of 4,670 firm-year observaions for our esing period 2000~2004. Our es requires he iniial hree-year daa for he period 997~999 o derive he sysemaic IT innovaion daa for each firm. We furher eliminae,393 firm-year observaions of missing curren and one-year ahead ROE, and 84 firm-year observaions of op and boom one percen exreme values in ROE and abnormal ROE values. This leaves us a oal of 3,93 firm-year observaions for he esing period 2000~2004, represening 72 unique firms (abou an average of 639 firms in each year). In our analysis, we apply our IT innovaion daa from he sample period 2000~2004 o forecas one-year lead abnormal ROE for he forecas period 200~2005. According o Naional Bureau of Economic Research (NBER), he echnical definiion of a recession is ha of wo consecuive quarers of GDP decline; based on his he US economy officially peaked (enered he recession) in March 200 wih he burs of he do-com bubble and collapse of NASDAQ. The economy reached is rough (end of recession) in November 200 (NBER 200, 2003a, 2003b). However, he period from he fall of 200 ill he summer of 2003 was marked by several evens indicaing he economy was no in recovery sae ye. Firs, he sock marke and he economy suffered afer he Sepember, 200 aacks. Second, indusries ha los jobs during he recession coninued shrinking during he recovery. Third, srucural changes (due o such facors as ousourcing) in he labour marke mean ha we were observing permanen job losses raher han emporary layoffs. In general, economy wide figures (unemploymen saisics) and he sock marke did no fully recover ill he summer of 2003 (Groshen and Poer 2003; NBER 2003b; Wikipedia). In he conex of our empirical analysis we generae forecass for 200 o 2003 (period of recession) and 2004~2005 (period of recovery) 3 based on sample daa from 2000~2002 and 2003~2004 respecively. A oal of,954 firm-year observaions of our sample come from and,239 firm-year observaions from 2003~ Explanaory Variables Given he naure of our hypoheses, we idenify firms in erms of heir approach o IT innovaion as Sysemaic, opporunisic, and non-it innovaors. Following a similar ye more conservaive approach han Sraopoulos and Lim (2007); we creae four-year rolling windows (i.e., , , ), and wihin each one of hese windows, we classify a firm as a sysemaic innovaor (SYS=) if hey have been recognized in IW500 more han hree consecuive imes in he four-year rolling window ending wih curren year. 4 Non IT-innovaors are firms ha have no been recognized in any argue has driven his percepion: he elecommunicaions deregulaion in 996, he Y2K problem, he do com boom, he crash of NASDAQ, and 9/. 3 The recovery period coninued ill he fall of We limi our analysis o he years 2004 and 2005 because we use he same sample daa se of IT innovaors as in Sraopoulos and Lim (2007) and his daa se ends in We classify a firm as a sysemaic IT innovaor if i has appeared in he lis of IW500 in more han wo consecuive years in a four year rolling window. In oher words, we replicaed our empirical analysis using he same definiion as in Sraopoulos and Lim (2007) and our resuls were he same. 5

7 Chen e al. Susainable value creaion: role of IT innovaion of he four years. We classify he res as opporunisic innovaors. Opporunisic innovaors are firms ha have appeared sporadically in IW500; hey have no been able o demonsrae a paern of coninuiy/persisence of IT innovaion in a fouryear rolling window. Based on his classificaion, our seleced sample produced he following resuls: 536 sysemaic firmyear observaions (SYS=);,323 opporunisic firm-year observaions (OPP=); and,334 non-it innovaive firm-year observaions (SYS=0 and OPP=0). 3.3 Measures of Firm Performance Following Cheng (2005) and Hawawini e al. (2002), we measure firm performance, firm abnormal ROE (ROEA), which equals residual income scaled by book value of equiy. Prior research suggess ha a firm s value is deermined by curren book value and he presen value of expeced fuure residual income, where residual income is he difference beween acual income and expeced income deermined by a firm s marke required cos of equiy (Ohlson, 995). Following Cheng (2005), we derive he indusry cos of equiy from he 48 indusry risk premiums documened by Fama and French (997, able 6, page 72~73). The indusry cos of equiy is he sum of he annualized one-monh T-bill yield and he Fama and French (997) indusry equiy premium. The indusry equiy premium is esimaed from he hree-facor model as sudied in Fama and French (997). Firms in a same indusry shares he same indusry cos of capial in he calculaion of abnormal ROE. To disinguish differenial firm abnormal ROE beyond indusry average abnormal ROE, we compue he indusry average abnormal ROE weighed by book value of equiy (ROEAIN) and hen ake he difference beween firm abnormal ROE and indusry average ROE for firm differenial abnormal ROE (ROEFM). We predic ha for sysemaic IT innovaive firms, heir curren posiive abnormal ROE is persisen, ye heir curren negaive abnormal ROE is ransiory. We define an indicaor variable (GN) o disinguish posiive abnormal ROE (GN= if ROEA>0) from negaive abnormal ROE (GN=0 if ROEA<=0). We also expec sysemaic IT innovaive firms will be more likely o recover from losses during recession years (RECESS=), and more likely o mainain heir good performance in recovery periods (RECESS=0). 3.4 Conrol Variables Based upon recen sudies of he deerminans of IT invesmens, we conrol for he effecs of known firm-characerisics ha are associaed wih firm performance. Previous sudies (Lev, 983; Cheng 2005) find ha large firms are more likely o have persisen good financial performance relaive o small firms and hey aribue his o he compeiive advanage of large firms. We measure firm size (SIZE) as he naural log of oal asses. Second, research and developmen inensiy is viewed as a proxy for a firm s inangible capial, which may enhance firm performance (Li and Wong 2005). We measure R&D (RND) as a firm s R&D expendiures divided by sales Callen e al. 2005; Darrough and Rangan 2005; Bharadwaj e al. 999; Hi and Brynjolfsson, 996. Prior sudies (Ashana and Zhang 2006) have found ha firms and indusries R&D inensiies are boh posiively correlaed wih persisence of abnormal earnings. Following prior sudies, we also include marke share (MS) in our analysis, where MS is defined as he raio of a firm s sales o he oal indusry sale. The lieraure shows mixed findings regarding boh he sign and he saisical significance of he marke share-firm performance relaionship (Szymanski e al. 993). This may be because of he inclusion of some proxy variables for power, which may weaken he saisical associaion beween marke share and firm performance (e.g., Hi and Brynjolfsson 996). In addiion, firms wih high marke shares may derive no exra negoiaion power benefis, and as a resul, he effors devoed o increasing marke share (e.g., markeing coss and capaciy build-up) may no pay off (e.g., Szymanski e al. 993). Hence, we canno predic he sign of marke share (MS). Following Cheng (2005), we compue he indusry concenraion raio (CR) for each indusry formed by 2-digi SIC codes, which equals he squared marke shared in an indusry. Cohen and Levin (989) find ha indusry concenraion is conversely relaed o indusry compeiiveness (Cohen and Levin 989). However, Cheng (2005) argues ha a high indusry concenraion represens high barriers o enry and can induce and susain economic rens by reducing hreas from ouside compeiion. Thus, we do no have prediced sign for he indusry concenraion variable (CR). 3.5 Research Model Prior research suggesed ha he coefficien α should be posiive in he firs-order auoregressive (AR) model, meaning ha he accouning income of wo adjacen fiscal years should be posiively correlaed (e.g. Lev, 983, and Kormendi and Lipe, 987). Thus nex period abnormal ROE is a funcion of curren abnormal ROE: 6

8 Chen e al. Susainable value creaion: role of IT innovaion ROEA = # + " $ ROEA +! + 0 () where ROEA is he abnormal ROE, is year subscrip. For abbreviaion, firm subscrip is omied from all equaions. The coefficien α capures he persisence of curren abnormal ROE. Following Cheng (2005), we decompose curren abnormal ROE ino indusry average abnormal ROE (ROEAIN j, ), and firm differenial ROE (ROEAFM ): ROEA = ROEAIN, + ROEAFM (2) j where j sands for he indusry o which each firm belongs. Because indusry abnormal ROE varies wih indusry characerisics and firm differenial ROE varies wih firm characerisics, hese wo componens migh have differen persisence (Cheng 2005). Our research focus is he persisence of firm specific ROE. Therefore, he isolaion of firm differenial ROE from indusry average ROE inroduces conrol on indusry specific abnormal ROE and enables our ess o address how he IT innovaion sraegy of each firm affecs heir abnormal ROE performance a he firm level. From equaion 2 and equaion, we can derive equaion 3 as follows: ROEA = # + " ROEAIN + " $ ROEAFM +! j, (3) where ROEAIN is indusry abnormal ROE, compued as weighed (by book value) average firm abnormal ROE wihin a same indusry formed by wo-digi SIC codes, ROEAFM is firm abnormal ROE compued as he difference beween ROEA and ROEAIN, j is indusry subscrip. Following Cheng (2005), we allow differen coefficiens for indusry ROE (α 0 ) and for firm ROE (α ) o reflec heir differen persisence. We expec boh α 0 and α posiive based on he findings of Cheng (2005). To es our firs hypohesis, we add sysemaic and opporunisic IT innovaive variable (SYS, and OPP ) in he regression o es if firms ha sysemaically innovae wih IT ouperform opporunisic and non-it innovaive firms: ROEA + + " % SIZE = $ + # ROEAIN 0 + " % RND 2 0 j, + # % ROEAFM + " % MS 3 + " % CR 4 + # % SYS + # OPP where SYS equals one if a firm has been classified as sysemaic IT innovaor in he four year window (including curren year ), and zero oherwise; OPP equals one if he firm has been classified as opporunisic IT innovaor in he four year window (including curren year ), and zero oherwise. Our conrol variables include SIZE, he naural logarihm of oal asses; RND, he percenage of R&D invesmen over sales; MS, he marke share, equals he percenage of a firm s sale over he oal sales of an indusry formed by wo-digi SIC code; CR, indusry concenraion raio (Herfindahl index), equals he sum of squared marke shared in he indusry. Our firs hypohesis predics ha he abnormal ROE of sysemaic IT innovaive firms is higher relaive o opporunisic and non-it innovaive firms. So he expeced sign of he coefficien α 2 of he variable SYS is posiive. While we do no have predicion of he coefficien α 3 of he variable OPP, we expec he coefficien α 2 > α 3. Our second hypohesis (2a and 2b) predics differen persisence of he abnormal ROE of sysemaic IT innovaive firms relaive o opporunisic and non-it innovaive firms, condiioning on good/bad news and recession/recovery period. Therefore, we inroduce he wo-way ineracion erms of firm abnormal ROE (ROEAFM ) wih indicaor variable SYS and OPP, as well as wih our conrol variables: ROEA + + " % SIZE = $ + # ROEAIN j, + # % ROEAFM +! + # % ROEAFM % SYS + # % ROEAFM % OPP 4 + " % RND + " % ROEAFM % RND + " " % MS " % CR 4 % ROEAFM % MS + " 2 + # % SYS + # % OPP + " % ROEAFM * SIZE % ROEAFM % CR +! Hypohesis wo (2a and 2b) requires regressions by good (GN=) and bad (GN=0) news subsamples, which are formed by he sign of ROEAFM (GN= if ROEAFM >0, oherwise GN=0). For good news observaions, hypohesis 2.a predic α 4 >0 and α 4 > α 5. This means ha sysemaic IT innovaive firms are more likely o coninue heir curren good performance in he nex year (have more persisen good performance) relaive o opporunisic and non-it innovaive firms. For bad news observaions, hypohesis 2.b predic ha α 4 < 0 and α 4 < α 5. This means ha sysemaic IT innovaive firms are more likely o urn around heir curren bad performance in he nex year (have more ransiory bad performance) relaive o opporunisic or non-it innovaive firms. 2 3 (4) (5) 7

9 Chen e al. Susainable value creaion: role of IT innovaion Noe ha while according o hypohesis one, he abnormal ROE of sysemaic IT innovaive firms is higher relaive o opporunisic IT innovaive firms and non-it innovaive firms, i.e., α 2 > 0, and α 2 > α 3. However, when grouping firms by heir curren abnormal ROE performance, he good performance of sysemaic IT innovaive firms relaive o opporunisic or non-it innovaive firms is parially capured by he inercep! 0. Specifically, he coefficien esimae of! 0 in he regression of good news sample is greaer han he! 0 in he regression of he bad news sample. In oher words, he effec of variable SYS on ROEA + is parly subsumed by he good and bad news sample spli. Therefore, we may no observe he predicion of hypohesis one (α 2 > 0, and α 2 > α 3) in he good/bad news regression. Hypohesis 3a predics ha for he recession period (RECESS=), he coefficien α 4 < 0, and α 4 < α 5. This means ha sysemaic IT innovaive firms are more likely o recover from he recession in he nex year (have more ransiory abnormal ROE) relaive o opporunisic or non IT innovaive firms. Hypohesis 3b predics ha for recovery period (RECESS=0), he coefficien α 4 > 0, and α 4 > α 5. This means ha sysemaic IT innovaive firms are more likely o coninue heir good performance in he recovery period in he nex year (have more persisen abnormal ROE) relaive o opporunisic or non IT innovaive firms. DISCUSSION OF RESULTS Descripive saisics for all variables are shown Panel A of Table 2. The curren abnormal ROE (ROEA ) has a mean of 0.02, which is greaer han he close-o-zero mean of one-year ahead abnormal ROE (ROEA + ). However, heir median values are he same (0.03). The indusry abnormal ROE (ROEAIN ) has a mean of 0.03 and he differenial firm abnormal ROE (ROEAFM ) has a mean of The mean sysemaic IT innovaion variable (SYS ) is 0.7, indicaing ha 7 percen of our sample is caegorized as sysemaic IT innovaors in he period 2000~2004. The indicaor variable of opporunisic IT innovaors (OPP ) has a mean of 0.4, meaning ha 4% of our sample is caegorized as opporunisic IT innovaors in he period The good news indicaor (GN) has a mean of 0.64, suggesing ha 64 percen of our observaions have posiive curren abnormal ROE. The recession indicaor (RECESS) has a mean of 0.6, presening ha 6 percen of our sample falls ino he recession period. The disribuion of all conrol variables size (SIZE), R&D (RND), marke share (MS), and indusry concenraion raio (CR) have no obvious ouliers. Our examinaion of he sandard errors and size of he coefficiens also shows ha hey are no sensiive o he inclusion or exclusion of he highly correlaed variables, indicaing mulicollineariy is unlikely o be problemaic (Hosmer and Lemeshow 989). [Inser Table 2] Panel B of Table 2 presens he correlaion coefficiens for all variables in he mulivariae model. While marke share (MS) and indusry concenraion (CR) are highly correlaed wih a correlaion coefficien of 0.7, none of he correlaions beween he conrol variables and our key esing variables (he abnormal ROE variables ROEA and ROEA +, he good news indicaor GN, and he recession indicaor RECESS) are above 0.20, and he highes variance inflaion facor (VIF) in our regression is To faciliae he inerpreaion of our coefficiens, we ake he difference beween each conrol variable (SIZE, RND, MS, CR) and is corresponding mean value. So he coefficien of variable ROEA capures he sample average persisence of abnormal ROE. 4. Univariae Analysis To provide preliminary evidence on our proposed hypoheses, we firs calculae average abnormal ROE for he curren and nex period +. If sysemaic IT innovaive firms ouperform han non-sysemaic IT innovaive firms (including boh he opporunisic and non IT innovaors), he average abnormal ROE of sysemaic IT innovaive firms should be higher compared o non-sysemaic IT innovaive firms. Consisen wih our predicion, Panel A of Table 3 shows ha sysemaic IT innovaive firms have greaer abnormal ROE in boh curren period and nex period +, for no only he full sample bu also subsamples pariioned by good/bad news and recession/recovery periods. All differences are saisically significan a less han 5 percen level excep for he difference in curren ROE of he good news subsample. This loss of significance for he good news subsample is likely o be because he good news pariion is based on curren abnormal ROE, hus i subsumes he posiive effec of sysemaic IT innovaion on curren abnormal ROE. [Inser Table 3] We also conduc frequency analysis o reveal how sysemaic IT innovaion affecs he auocorrelaion beween curren and fuure abnormal ROE. Specifically, his analysis addresses our research quesion wheher i is more likely for sysemaic IT innovaive firms o coninue heir curren good performance in he fuure, or alernaively, o rever heir curren bad performance in he fuure. If a firm has wo consecuive posiive (negaive) abnormal ROE in year and +, i is caegorized 8

10 Chen e al. Susainable value creaion: role of IT innovaion as he consisen group; oherwise i is caegorized as he inconsisen group. Since hese resuls are descripive in naure and do no accoun for conrol variables, hey only provide preliminary evidence on he relaive performance of sysemaic IT innovaors versus non-sysemaic IT innovaors. The Panel B of Table 3 shows ha for our full sample, more sysemaic IT innovaive firms (8%) have consisen abnormal ROE relaive o he non-sysemaic IT innovaive firms (78%). When grouping observaions by good/bad news or recession/recovery periods, we find ha sysemaic IT innovaive firms having good news or in he recovery period are more likely o have a persisen abnormal ROE of same signs in wo consecuive years. And he Chi-square saisics boh he good news (4.97) and recovery (3.59) subsamples are significan a 3 and 6 percen level respecively. In conras, while Chi-square saisics are insignifican, sysemaic IT innovaive firms experiencing bad news or in he recession period appear o more likely have non-persisen abnormal ROE of differen signs in wo consecuive years. In summary, our univariae resul indicaes ha sysemaic IT innovaive firms are more likely o mainain heir curren good abnormal ROE performance (in recovery period) and flip heir curren bad abnormal ROE performance (in recession period). 4.2 Regression Resuls: Hypohesis One We firs run a regression of equaion 3 as our base model and find ha he one-year ahead abnormal ROE (ROEA + ) is posiively relaed o boh curren indusry abnormal ROE (ROEAIN ) and curren firm differenial abnormal ROE (ROEAFM ). Table 4 repor he regression saisics, he variable ROEAIN has a coefficien of 0.85 and he variable ROEAFM has a coefficien of 0.52, all significan a less han % level and comparable o he coefficien repored in Cheng (2005, Table 3, page 98). [Inser Table 4] We hen add he sysemaic IT innovaion indicaor variable (SYS ) and he opporunisic IT innovaion indicaor variable (OPP ) in he regression o es our hypohesis one. Consisen wih our predicions, we find a significan posiive associaion beween ROEA + and SYS for boh regressions ha wih and wihou conrol variables (coefficien = 0.03, -sa = 2.34), and he coefficien of SYS is more posiive han he coefficien of OPP (α 2 - α 3 =0.04, -sa=2.07). These findings sugges ha sysemaic IT innovaors have higher abnormal ROE han opporunisic and non-it innovaors. We also find ha he high R&D firms are associaed wih significanly lower abnormal ROE relaive o low R&D firms as indicaed by he negaive coefficien of RND (coefficien = -0.0, -sa = -4.02). The oher conrol variables of size (SIZE), marke share (MS) and indusry concenraion raio (CR) are all insignifican. 4.3 Regression Resuls: Hypohesis Two Good News versus Bad News Our hypohesis wo predic ha he persisence of abnormal ROE for sysemaic IT innovaive firms depends on he level of curren abnormal ROE: if curren abnormal ROE is posiive (good news), i is more likely o susain in he nex period; if curren abnormal ROE is negaive (bad news), i is more likely o be ransiory and evaporae in he nex period. Therefore, o es our hypohesis wo, we esimae equaion 5 by good and bad news pariion of our sample. However, before spliing sample by good and bad news, wih our full sample, we find ha he coefficien of SYS is significanly posiive (coefficien = 0.03, -sa = 2.46) and greaer han he coefficien of OPP (α 2 - α 3 =0.04, -sa=2.58). This means ha our hypohesis one sill holds afer including all he ineracion erms beween ROEAFM and our independen variables SYS and OPP, and our four conrol variables. The coefficien of he ineracion erm ROEAFM *SYS is insignificanly differen from zero (-sa = -0.67), meaning ha before condiioning on he good/bad news and recession/recovery, he overall persisence of firm abnormal ROE for he sysemaic IT innovaors is insignificanly differen from ha of he non-it innovaors. However, he coefficien of he ineracion erm ROEAFM *OPP is significanly negaive (coefficien = -0.3, -sa = -2.88), suggesing ha before condiioning on he good/bad news, he overall persisence of firm abnormal ROE for he opporunisic IT innovaors is significanly lower han ha of he non-it innovaors. Consequenly, he overall persisence of firm abnormal ROE of he sysemaic IT innovaors is significanly higher han ha of he opporunisic IT innovaors (α 4 - α 5 = 0.09, =2.06). [Inser Table 5] Afer spliing he sample by good and bad news, consisen wih our hypohesis wo (2a and 2b), we find ha he coefficien of he ineracion erm ROEAFM *SYS is significanly posiive for he good news subsample (coefficien = 0.4, -sa = 2.43), bu negaive for he bad news subsample (coefficien = -0.35, -sa = -2.22). This means ha sysemaic IT innovaors exhibi more persisence good performance (posiive abnormal ROE) and more ransiory bad performance (negaive abnormal ROE) relaive o non-it innovaors. In he bad news sample regression, we also find evidence ha when sysemaic IT innovaors are confroned wih a bad performance in he curren year, hey are more likely o urnaround and achieve beer performance han he opporunisic IT innovaors (α 4 - α 5 = -0.20, =.59). However, we did no find significan 9

11 Chen e al. Susainable value creaion: role of IT innovaion evidence (in he good news sample regression) ha sysemaic IT innovaors have more persisen good performance relaive o opporunisic IT innovaors (α 4 - α 5 = 0.0, =0.0). 4.4 Regression Resuls: Hypohesis Three Economic Cycle Effec Our hypohesis hree (3a and 3b) predic ha sysemaic IT innovaors have persisen abnormal ROE in he recovery period and less persisen abnormal ROE in he recession period, hus we esimae equaion 5 by he recession and recovery period pariion. Consisen wih our predicion, we find ha he coefficien of he ineracion erm ROEAFM *SYS is significanly posiive for he recovery subperiod (coefficien = 0.4, -sa = 2.74), bu negaive for he recession subperiod (coefficien = , -sa = -2.03). This means ha sysemaic IT innovaors have more persisen (ransiory) abnormal ROE relaive o non-it innovaors in recovery (recession) period. In he recovery period regression, we also find evidence ha sysemaic IT innovaors are more likely o have more persisen abnormal ROE han opporunisic IT innovaors (α 4 - α 5 = 0.6, =8.03). However, we did no find significan evidence (in he recession period regression) ha sysemaic IT innovaors have more ransiory abnormal ROE relaive o opporunisic IT innovaors (α 4 - α 5 = -0.07, =0.49). To es he join effec of good/bad news and recovery/recession on he persisence of abnormal ROE of sysemaic IT innovaors, we conduc furher analysis and esimae equaion 5 by four subsamples, namely, good news observaions in recession period, and bad news observaions in recession period, good news observaions in recovery period, and bad news observaions in recovery period. The resuls of Table 6 reveal ha he ransiory abnormal ROE associaed wih sysemaic IT innovaors in he recession period, is mainly for he bad news firms. The coefficien esimae of he ineracion erm ROEAFM *SYS for he recession and bad news subsample is -0.4 (-sa = -2.09) and he difference beween he coefficien esimae ROEAFM *SYS and ROEAFM *OPP is (-sa = 3.65). We find no significan evidence ha he sysemaic IT innovaors have more persisen abnormal ROE relaive o opporunisic and non-it innovaors in he subsample of recession and good news. [Inser Table 6] On he oher hand, he regression of he good news and recovery period subsample seems perplexing. Alhough sysemaic IT innovaors show more persisen abnormal ROE han non-it innovaors (α 4 =0.9, -sa = 3.69), opposie o our predicion, we find ha sysemaic IT innovaors have less persisen abnormal ROE relaive o opporunisic IT innovaors (α 4 - α 5 = - 0.8, =8.32). However, hese resuls make sense for he following wo reasons:. While he recovery period exends ill he fall of 2007, our forecas period covers only he firs wo years of recovery ( ). We limied our analysis o he years because we use he same sample daa se of IT innovaors as in Sraopoulos and Lim (2007) and his daa se ends in In absolue levels, he sysemaic IT innovaors significanly ouperform he opporunisic IT innovaors by 2% in firm differenial abnormal ROE in he good news recovery sample (SYS - OPP = 0.02, = 3.6). Since i is more difficul o mainain high growh in abnormal ROE because compeiion will drive away high abnormal ROE (Cheng, 2005), his may conribue o lower persisence of he abnormal ROE of sysemaic IT innovaors relaive o ha of he opporunisic IT innovaors. Collecively, our findings largely suppor our hree hypoheses, suggesing ha sysemaic IT innovaors have higher abnormal ROE, and more persisence (ransiory) good (bad) performance relaive o opporunisic and non-it innovaors. Moreover, he persisence of abnormal ROE of sysemaic IT innovaors is higher (lower) han ha of opporunisic and non- IT innovaors in more regular recovery period (chaoic recession period). SUMMARY AND CONCLUSION 5. Summary & Implicaions Overall, our findings sugges ha sysemaic IT innovaion sraegy leads o value creaion and his value creaion resiss he erosion from compeiors acions. Based on our sample of large US firms, sysemaic IT innovaors enjoy higher abnormal ROE han opporunisic and non-it innovaors, and overall persisence of firm abnormal ROE of he sysemaic IT innovaors is significanly higher han ha of he opporunisic IT innovaors. During a recession, when sysemaic IT innovaors are confroned wih a bad performance in he curren year, hey are more likely o urnaround and achieve beer performance han opporunisic IT innovaors. Based on he evidence from he recession of he early 2000s, we conclude ha companies ha had adoped a sysemaic approach o IT innovaion have demonsraed remarkable business agiliy. They are beer prepared o deal wih economic recessions and he urbulence of a modern hypercompeiive environmen. Our findings complemen prior evidence ha has shown ha successful companies end o press heir advanage (inves) during recession as a preparaion for recovery (Dobbs e al. 2002). 0

12 Chen e al. Susainable value creaion: role of IT innovaion The sudy has several implicaions. Firs, we have inroduced a new economic characerisic (IT innovaion sraegy) in he lieraure of drivers of earnings persisence. Analyss and invesors ineresed in selecing companies wih long-erm poenial (earnings persisence) may wan o consider he company s IT innovaion sraegy in heir evaluaion. Second, we evaluaed he role of IT innovaion sraegy during he recen business cycle, as well as in he conex of good news and bad news. This prepares a pah for more fuure research on he role of IT innovaion sraegies in he conex of he curren global economic crisis. Third, alhough he resuls indicae ha sysemaic IT innovaion leads o improved firm earning persisence, he underlying mechanisms for developing a successful sysemaic IT innovaion sraegy require furher invesigaion. Fourh, wih budge cus on cener sage of IT agendas, forward looking managers are exploring ways o use IT in order o cu coss in a way ha will help hem leap ahead of weakened compeiors (Kaplan, Robers, and Sikes 2008; Kaplan and Sikes 2008; McGee and Soas 2008). To hese managers, he message of our sudy is ha a long-erm sysemaic approach o IT innovaion is beer han rying o follow he mos recen echnological rend or new killer applicaion. 5.2 Limiaions Our main source, IW500 does no provide he complee lis of firms ha were considered as IT innovaive bu which were subsequenly rejeced. In oher words, we have access o a runcaed sample of only he firms ha have aained IT innovaion saus. 5 As menioned in he descripion of he daa se, we ried o miigae his limiaion by leveraging IS and managemen lieraure as well as he guidelines followed by IW500. We looked a he op compeiors and we used he same source (Hoover s) ha was used by IW500 for he compilaion of heir daa se. We limied he number of compeiors o a maximum of hree in order o ensure ha only companies ha could have been included in he IW500 pool of candidaes were included in our conrol group. Anoher limiaion ha warrans menion is he selecion of he conrol groups. We use op hree compeiors from Hoover s, which was acively involved in he compleion of IW500. Despie aemps o mach he conrol sample based on indusry ype and firm size, for some of firms in IT innovaive sample, an appropriae conrol firm of similar size, bu could be also IT innovaive firms in a differen ime period. In addiion, our lack of knowledge abou he IT capabiliy of he conrol sample precludes any direc comparison of he wo groups on he naure and qualiy of heir IT resources. Finally, for our analysis we have used he same sample daa se period ( ) as in Sraopoulos and Lim (2007) and his limis our abiliy o forecas in he recovery period o he years 2004 and According o he NBER he recovery period lased unil he end of Fuure Research Alhough he resuls indicae ha IT innovaion persisence leads o improve firm s earning persisence, he underlying mechanisms hrough which his can be achieved furher. Facors such as IT inensiy, curren sae of IT, and several new areas of IT innovaion could play a criical role in deermining he success of IT innovaion persisence. Percepions regarding he number of associaed agens, such as cusomers, suppliers, or inermediaes, he capabiliies of heir IT experise, or he ype of arrangemens could also influence earning persisence. Examining such addiional facors migh be a fruiful research endeavour in he fuure. The period of analysis offers anoher opporuniy for fuure expansion of our work. Our analysis in based on he recession in he early 2000s, which was more localized in naure (mosly US) and i was driven in par by he 9/ evens. Fuure sudies will have o replicae his wih daa from he curren crisis which is global in naure and driven by he financial crisis. 5.4 Conclusion While, IT innovaion may no be suiable and feasible for all firms, companies ha have developed a sysemaic approach o IT innovaion are beer prepared o deal wih he modern hypercompeiive environmen. Sysemaic IT innovaion sraegy leads o he creaion of susainable value creaion. ACKNOWLEDGMENTS We hank he anonymous reviewers of he Americas Conference on Informaion Sysems for heir feedback. We wish o acknowledge he financial suppor provided by he Social Sciences and Humaniies Research Council of Canada (SSHRC) and he Universiy of Waerloo. 5 This is a common limiaion applying o all sudies (Bharadwaj 2000; Sanhanam and Harono 2003; Sraopoulos and Lim 2007, and numerous ohers) ha relied on he IW500 daa se.