Published on

International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of Computer Science, Neural Networks, Electrical Engineering, Software Engineering, Information Technology, Mechanical Engineering, Chemical Engineering, Plastic Engineering, Food Technology, Textile Engineering, Nano Technology & science, Power Electronics, Electronics & Communication Engineering, Computational mathematics, Image processing, Civil Engineering, Structural Engineering, Environmental Engineering, VLSI Testing & Low Power VLSI Design etc.

Published in: Technology, Business
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide


  1. 1. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071095 | P a g eEnterprise Systems and Labor Productivity: DisentanglingCombination EffectsBenjamin Engelstätter(ZEW, ICT Research Group)AbstractThis study analyzes the relationshipbetween labor productivity and the three widelyestablished enterprise software systems, i. e.Enterprise Resource Planning (ERP), SupplyChain Management (SM) and CustomerRelationship Management (CRM), revealingperformance gains incurred through the varyingcombinations of the systems with an exclusivefocus on complementarities and interactionsamong them. Using German firm-level data, theresults show that the highest productivity gainsbased on enterprise system usage are realizedthrough the use of the three enterprise softwaresystems in concert and not by relying on one ortwo of the systems alone. Concerningcomplementarity the results indicate that SCMand CRM function as complements if the firmsalready have an ERP System running or get thesystem in conjunction. The results stay robust todifferent model specifications andcomplementarity testing procedures.IntroductionEnterprise systems, company-wide suitesof business software devoted to particular processintegration across the value chain, encompass awide range of software products supporting day-to-day business operations and decision-making. Thethree main enterprise systems, Enterprise ResourcePlanning (ERP), Supply Chain Management (SCM)and Customer Relationship Management (CRM),serve many industries in numerous areas. They aredesigned to automate operations from supplymanagement, inventory control, manufacturingscheduling or sales force automation and almost anyother data-oriented management process. In 2006,the global market for ERP systems accounts forapproximately $29 billion (Jacobsen et al. 2007).SAP, the largest global enterprise software vendor,estimates the addressable market for all enterprisesoftware applications to be roughly $ 110 billion2010.Enterprise systems are often adopted toreplace the usually poorly connected legacysoftware and are expected to reduce infrastructuresupport costs (Hendricks et al. 2007). These supportcosts originate from the need for costly programmedinterfaces to connect legacy applications. Also, costsmight occur in form of waiting time since legacysoftware is generally not efficiently programmedand therefore requires more time to execute orderscompared to enterprise systems. In line with thislegacy software can be expected to face moredowntimes than enterprise systems based on itsinefficient programming code. In addition,improvements in operational integration realizedthrough enterprise software can affect the entireorganization and therefore might positively affectfirm performance. ERP systems provide benefits inthe area of transaction automation, SCM systemslead to more sophisticated planning capabilities andCRM systems simplify customer relationshipmanagement.Although many studies (e. g. Aral et al.2006; Hendricks et al. 2007; Hitt et al. 2002) arguethat adopted enterprise systems, in general,positively impact firm performance, performancebenefits based on potential complementarity drivenadvantages resulting from the usage of differententerprise systems in combination are still notinvestigated. Overall, the majority of the existinganalysis lays an exclusive focus on the adoption of asingle system (e. g. Dehning et al. 2007; Hitt et al.2002; Nicolaou 2004), thereby completelydisregarding these complementarity basedperformance gains. This focus might turn out toproduce biased results as enterprise systemsnowadays are often adopted in concert and areexpected to interact, complement or in rare caseseven substitute each other to some extent. The goalsof this study therefore are to disentangle theperformance effects attributed to the combinationsand interactions of enterprise systems using a uniquedatabase consisting of German firms from themanufacturing industry and from service sectors.Based on a production function approach,the results provide empirical evidence about theproductivity gains resulting from the use ofenterprise system usage with respect particularly tothe interacting nature of the systems. Contrary toformer analysis, the highest productivity gain isachieved by using all three enterprise systems inconcert. In addition, a complementary relationshipbetween SCM and CRM is revealed once an ERP isalso in use or acquired in conjunction. The resultsimply that analyzing the influence of the enterprisesystems independently, whereby one disregards
  2. 2. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071096 | P a g epossible benefits stemming from the combination ofsystems, turns out to be insufficient for depicting theperformance effects of enterprise software.The study proceeds as follows. Section 1 givesan overview of the literature and derives hypotheses.Section 2 pictures the basic model and covers theissues concerning complementarity. The dataset ispresented in section 3 whereas section 4 contains theestimation results. Section 1 concludes.1 Background discussion and hypotheses1.1 Benefits of enterprise systems in generalERP systems replace complex interfacesbetween different systems with standardized cross-functional transaction automation. They use a sourceof data that integrates enterprise functions such assales and distribution, materials management,production planning, financial accounting, costcontrol and human resource management (Aral et al.2006). By using an ERP system, order cycle timescan be reduced, which might lead to improvedthroughput, customer response times and deliveryspeeds (Cotteleer and Bendoly 2006; McAfee 2002).In addition, cash-to-cash cycle times and the timeneeded to reconcile financial data at the end of thequarter or year can be reduced through automatedfinancial transactions (Mabert et al. 2000). With anERP system, all enterprise data is collected onceduring the initial transaction, stored centrally andupdated in real time. This ensures that all levels ofplanning are based on the same data, allowing theresulting plans to realistically reflect the prevailingoperating conditions of the firm (Hendricks et al.2007). All in all, standardized firm-widetransactions and centrally stored enterprise datagreatly facilitate the governance of the firm(McAfee and Upton 1996; Scott and Vessey 2000).Providing managers with clear oversight of therelative performance of different parts of thecompany, ERP reports can be used to identifynecessary improvements and to take advantage ofmarket opportunities (Mehrjerdi 2010).IT-based SCM systems coordinate andintegrate the flow of information, materials andfinances along the value chain and enhanceoperational and business planning (Dehning et al.2007). With the real-time planning capabilitiesoffered by SCM systems, firms can react quickly tosupply and demand changes and are able to servecustomers in a timely and comprehensive fashion(Cachon and Fisher 2000; Hendricks et al. 2007). Inaddition, SCM systems can directly improveinventory management by reducing inventory levels,holding costs, spoilage and lead times. This resultsin increased profitability by reducing costs, avoidinglost sales and improving customer satisfaction(Cachon and Fisher 2000). SCM systems may alsohave indirect effects on firm performance due tolower coordination, sales, general and administrativecosts as well as improved decision-making orforecasting (Dehning et al. 2007).CRM is a synthesis of customer focusedmanagement and many existing principles fromrelationship marketing (Jancic and Zabkar 2002;Morgan and Hunt 1994; Sheth et al. 2000). The keyfocus of CRM systems is to facilitate the creation oflong-term relationships with customers by providingthe appropriate infrastructure, e. g., enablingeffective sales force automation, centralizedcustomer data warehousing and data mining pairedwith decision support and reporting tools(Engelstätter 2012). Offering a complete view ofcustomer needs and wants, a CRM system is alsoexpected to lead to superior customer loyalty,reduced cost of sales and services and improvedbottom-line profits (Chen 2001). In addition, a CRMsystem reduces duplication in data entry andmaintenance by providing a centralized firm-widedatabase of customer information. This databasereplaces systems maintained by individual salespeople, institutionalizes customer relationships,prevents the loss of organizational customerknowledge when sales staff leaves the firm and canreduce costs by streamlining repetitive transactionsand sales processes (Cohen et al. 2006; Hendricks etal. 2007).1.2 Complementarity between enterprisesoftware applicationsAs the three main enterprise systems showmany linkages among each other, this section willcover the possible combinations of enterprisesystems and their special interactions separately. Forinstance, the integrated data, processes andinterfaces provided by an ERP system, facilitate theeffective implementation of supply chain activities.This planning of internal production activitiesthrough the ERP system can be directly influencedor automated by information inputs from supplychain partners via use of an SCM system (Aral et al.2006). The two-way information exchange betweenERP and SCM systems enables companies tooptimize processes across the product lifecycle.Based on the rapid data transfer between these twoenterprise systems information about resource usageand product life cycles can be accessed bycorrespondent employees in real time as needed.Based on this information room for processimprovements and product enhancements is easy toidentify (Engelstätter2012). Especially the usage ofbest-of-breed SCM applications with ERP, asopposed to relying on the innate SCM functionalityof some ERP systems, provides firms withspecialized SCM features which could even forecastexpected supply and demand. Accordingly, firmsthat use the advanced features of SCM systems inconjunction with ERP systems are expected to
  3. 3. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071097 | P a g erealize complementary benefits and a higher firmperformance as it is argued in Wieder et al. (2006).The centralized customer data provided byCRM systems can be used as source for ERPsystems and is particularly valuable to themanagement of multiple product lines are managed(Hendricks et al. 2007). The flow of informationbetween the two systems will, generally, ease thefirms’ data management effort as the sharedinformation avoids redundant data keeping. Inaddition, the CRM system can utilize the datamining capabilities of ERP systems and datawarehousing to reveal profiles of key customers,customer profitability and purchasing patterns(Conlon 1999). Companies might also gainadditional insights into customer orders, contractsand buying behaviour if they use ERP applicationsin conjunction with CRM systems, possiblyallowing them to establish more focused sales andmarketing strategies based on in-depth analysis ofcustomer behaviour, interests and context. Hence,deploying ERP systems in conjunction with CRMsystems should, based on complementary benefits,increase a firms’ performance.A firm can utilize the insights intocustomer behaviour and demand patterns providedby a CRM system to streamline manufacturing anddistribution, if the CRM system is connected withthe firm’s SCM applications. Especially informationon customer segmentation can provide advice forthe structuring of the purchase of raw materials,scheduling manufacturing, managing inventory andrunning of the supply chain in general, possiblyresulting in reduced costs. A streamlined supplychain based on connected CRM and SCM systems iskey for reaching out to the right customer at theright time and should benefit the companyadditionally through improved resource allocationpotentially reducing inventory stocks. In thiscontext, Mithas et al. (2005) reveal that firms aremore likely to benefit from CRM systems if theyhave a greater supply chain integration, e. g. basedon an SCM system already installed. Therefore,using SCM and CRM systems in conjunction can beexpected to increase firm performance based on acomplementarity relationship.According to (Charkari and Abdolvand2004), the isolated use of SCM or CRM systemsseparately might result in missed opportunities andpoor performance. To put it short, ERP systemsgenerally determine the business processes, the twoother applications optimize these business processesin a specific area, especially by linking the frontoffice of the enterprise, e. g., sales, marketing,customer services, with the back office, e. g.,operations, logistics, financials, human resources.As an example showing the linkages between thethree applications, let a company be supposed todeliver an order to ten clients the next day. TheSCM solution now calculates that the company canonly deliver to five customers in the given timespan, the ERP can then pull the data from the CRMsystem to determine which orders should be fulfilled(Horwitt 2009). In conclusion, it can be expectedthat firms using all three main enterprise systemswill probably realize larger productivity increasescompared to those companies which rely on lessenterprise software applications or no enterprisesystems at all.2 Research methodology2.1 The basic modelFollowing e. g. Aral et al. (2006) or Hitt etal. (2002), a production function specification isused to estimate the effects of enterprise systemusage on firm performance. Throughout this paper Iuse the performance measure labor productivityYi/Li, i.e. sales over employees. The inputs arecapital (Ki), labor (Li) and other firm characteristics.These additional characteristics, e. g., enterprisesoftware adoption (ESi), capture differences inperformance and can simply be added to theproduction function in its log-log form (Hitt et al.2002). The coefficients of these added terms canthen accordingly be interpreted as percentagedifferences in productivity. In case of softwareadoption, the coefficient captures the enterprisesoftware’s effect on firm-level productivity otherthings being equal (Shin, 2006). Besides softwareadoption I control for East German heritage,industry sector and for ICT capital captured in Xi.The according production function in its log-logform results in:(1) ln (Yi/Li) = c + αln (Ki) + βln (Li) + γ ESi +δXi+ εi2.2 Modeling complementarityTaking the potential complementaritybetween the enterprise software systems intoaccount, this section outlines the definitions andconditions regarding complementarity andsubstitutability for the cases of discrete practices.Throughout this paper, each practice represents theuse of a different enterprise system.Complementarity is defined as in Baumol et al.(1988) and Athey and Stern (1998):Practices xi and xj are considered complementary inthe function f if and only if 2f / xixjis alwayslarger or equal to zero, and larger than zero for atleast one value of ( x1 ,...,xn).Following (Carree et al. 2007), an objectivefunction f is considered as a starting point. Thevalue of f is determined by the practices xp= (p
  4. 4. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071098 | P a g e=1,…,n) with n=3 in the present case of enterprisesoftware system usage. A cross-term specification ofthe objective function f allows testing forcomplementarity or substitutability. This implies thefollowing expression for n equal to 3:(2) f(x1,x2,x3) = θ0 + θ1x1 + θ2x2 + θ12x1x2+ θ3x3+ θ13x1x3+ θ23x2x3 + θ123x1x2x3In the present case of observed enterprisesystem usage, the practices are measureddichotomously, i.e. variables take the value one ifthe practice is used and zero otherwise. In that case,function (2) can be conveniently rewritten in termsof possible combinations of practices (Mohnen andRöller 2005). The collection of possiblecombinations considering three practices is definedin the usual binary order as D ={(0,0,0),(1,0,0),(0,1,0),(0,0,1),(1,1,0),(1,0,1),(0,1,1),(1,1,1)}, with the indicator function for threepractices ID=(r,s,t) being equal to one whenever thecombination is (r,s,t) and zero otherwise. Thefunction f can accordingly be rewritten as:(3)      1 2 31 1 11 2 3 , , , ,0 0 0, ,     rst x x x r s tr s tf x x x IThe conditions of pairwise complementaritybetween practice 1 and 2 then correspond to θ12 =λ110 + λ000 – λ100 – λ010 ≥ 0 and θ12 + θ123 = λ111 + λ001– λ101 – λ011 ≥ 0, with at least one inequality holdingstrictly. Similar inequalities apply for the pairs (1,3)and (2,3). For substitutability, the inequalities arereversed.2.3 Testing for complementarityThe most established and convenientmethod to check for complementarity, even fordiscrete practices, is the standard interaction termapproach as proposed in Athey and Stern (2002).Alternatively, one can use a Wald type test based ona minimum distance estimator derived in Mohnenand Röller (2005) to verify a complementarityrelationship. For this purpose, two independent testsare conducted which test separately forcomplementarity and substitutability. Carree et al.(2007) advanced this method and derived a testwhich decides between complementarity andsubstitutability in one run. Their multiplerestrictions test uses a Likelihood-ratio testprocedure which specifically tests for the twoinequalities derived above but faces a computationaldemanding test-statistic. As the test-statistic followsa mixed chi-square distribution under the nullhypothesis of no complementarity or nosubstitutability, exact p-values need to be computedusing specific weights (Shapiro 1985). A firstappliance of this test in empirical analysis can befound in Belderbos et al. (2006). As the likelihood-ratio test approach is most recent and has not beentested in several different settings and simulation Istick to the well established interaction terms as abaseline in the empirical analysis employing thelikelihood-ratio test merely as robustness check.3 The dataThis study is based on a dataset resultingfrom two computer-aided telephone surveysconducted in 2004 and 2007 by the Centre forEuropean Economic Research (ZEW). Thesesurveys laid a specific focus on the use and diffusionof ICT in German firms. The interviewee was, ingeneral, the chief executive officer of the companieswho could also pass on questions to a correspondingemployee such as, e. g., the head of the ICTdepartment. Each wave of this ICT-datasetoriginally contained information of about 4,000firms with five or more employees which wererepresentatively chosen from important service andmanufacturing sectors in Germany. The data basisfor the sample originates from the credit ratingagency Creditreform. This agency offers the largestdata base on firms available in Germany.Creditreform collects some basic information likeaddress, sector and firm size on all enterprises thatever applied for a bank credit. The selection fromthe population of German firms was stratifiedaccording to three size classes, to industries (sevenbranches of the manufacturing industry and sevenselected service sectors) and to two regions(East/West Germany). As many firms as necessarywere asked until all strata were filled.Besides detailed information on the use ofICT, the dataset contains additional informationabout total sales, the firms’ workforce, the totalinvestments and various other variables. Thequestionnaire also covered the usage level of thethree main enterprise software applications ERP,SCM and CRM. The level of usage could be none,minor or broad. Minor usage could be referring tosoftware which is only employed by a small numberof employees, departments or subsidies whereasbroad usage might describe software packages inuse by the entire firm (Engelstätter 2012). However,the questionnaire did not make any distinctionsbetween these categories and provides no additionalexplanations for minor and broad usage. Withoutadditional information, an interpretation of aproductivity effect due to minor software use incomparison to no or broad usage cannot beillustrated appropriately. Accordingly, I constructeda dummy variable for the use of each softwareapplication, which takes the value one if a firm usesthe software at least to a minor degree or broadlyand zero otherwise.The survey in 2007 covers total sales andthe number of employees for 2006 only. However,the answers on enterprise system usage in that
  5. 5. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071099 | P a g esurvey relate to the year 2007. Since the survey isorganized as a panel dataset, I use the softwareusage level reported in 2004 to construct the dummyvariables necessary. This two-year differencebetween software use and firm performance forms awell-defined temporal sequence which should beadequate to measure the productivity effects ofenterprise systems, given that multiple analysesconcluded that enterprise software needs about twoyears to generate some kind of performance effect(e. g. Matolcsy et al. 2005; Nicolaou 2004; Nicolaouet al. 2003).Due to panel mortality and item-nonresponse, matching the data for the two periodsresults in nearly 1,000 observations. By droppingthe banking sector, I obtained 927 observations formy final data set. As there is no data available tomeasure the firms’ physical capital stock, I follow e.g. Bertschek et al. (2006) and Hall et al. (2009) byusing the gross investment figures as an empiricalproxy for the capital stock. A potential problem withthis method arises as some firms report zeroinvestments in 2004, although the occurrence ofzero investments is explained by Bond and VanReenen (2007) and is therefore not surprising at all.However, with an econometric specification of theproduction function in logarithmic values for thefactor inputs, these firms must be excluded from thefinal sample. To do away with this problem and toavoid losing further observations, I follow theapproach used in Ohnemus (2007) and determinethe value of investment for the firms reporting zeroinvestments as the 10 percent quantile of theirrespective industry and size class.Table 1 shows the descriptive statistics forthe variables of the production function estimationin this study. Table 1 also includes two additionalvariables, namely export activity and the existenceof a works council which will be used for additionalanalysis in the next section. Sales, labor and thelabor productivity ratio refer to the year 2006, allother inputs, like e. g. capital, refer to the year 2004.In addition, Table 1also provides the descriptivestatistics for the industry affiliations of the firms inthe final sample.Table 1: Summary statisticsVariable Mean Std. Dev. DV4Output (sales) 46,667.2 169,785.8labor productivity1192.2 229.7capital22282.6 9843.0labor3245.7 916.5ln (output) 8.879 1.914ln (labor productivity) 4.891 0.811ln (capital) 5.366 2.230ln (labor) 3.987 1.644share of computer workers 0.467 0.328East Germany 0.273 yesexport share 0.557 yesworks council 0.368 yesconsumer goods 0.093 yeschemical industry 0.052 yesother raw materials 0.081 yesmetal and machine construction 0.118 yeselectrical engineering 0.076 yesprecision instruments 0.066 yesautomobile 0.073 yescomplete manufacturing sector 0.559 yeswhole sale trade 0.054 yesretail trade 0.051 yestransport and postal services 0.065 yeselectronic data transfer 0.097 yes
  6. 6. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071100 | P a g etechnical services 0.102 yesother business-related services 0.073 yesNumber of observations 927Notes: 1Sales per employee (in 2006) in €1,000. 2Capital is proxied by gross investment in €1,000.3Labor is measured in amount of total employees. 4Dummy variableSource: ZEW ICT survey 2004, 2007 and own calculations.In 2006, mean sales amount to € 46,667,200 andthe average firm size results in 246 employees. For2004, the mean investment is € 2,382,600. Themean share of workers mainly using a personalcomputer for their work, as proxy for the ICTcapital of the firm, is around 47 percent in the usedsample. This ICT capital measure should beincluded in the estimation to reduce omittedvariable bias, as working with a computer can beexpected to positively impact productivity(Greenan and Mairesse, 1996; Astrostic andNyugen, 2005). Nearly 27 percent of the firms arelocated in East Germany. Export activity isreported by 56 percent of the firms and 47 percentof the sample firms have established a workscouncil. Regarding the industry affiliation of thefirms, the largest share of 12 percent does businessin metal and machine construction, whereas only asmall number is associated with the wholesale orretail trade industry (5 percent each).Table 2 provides the summary statisticsfor the dummy variables of enterprise system usageand shows additional statistics of the firms usingthese systems. In addition, it lays a specific focuson the group of firms using either no enterprisesystems at all or the full suite of enterprise systems.It is striking that around one quarter of firms (28percent) in the sample use all three enterprisesystems in concert or no enterprise software at all(24 percent). The use of ERP is widespread (64percent), around 44 percent apply SCM softwareand about 51 percent of the firms have adoptedCRM software. Providing initial descriptiveevidence by comparing the average laborproductivity for each group with the overall samplemean reported in Table 2, it stands out that thefirms’ labor productivity exceeds the averageproductivity of € 192.200 once they adopt somekind of enterprise software. Especially firms usingall three systems in concert achieve the highestaverage labor productivity, reaching € 234.200.Unsurprisingly, companies using no enterprisesystems at all seem to fall behind in terms of laborproductivity (€ 167.000). In addition, it seems to bethe case that large firms choose to use all threesystems together. These firms engage the servicesof an average of 566 employees, as shown in Table2 However, as the available sample suggestssmaller firms also seem to have confidence in thefull suite of enterprise software applications asabout 43 percent of the firms which use all threeenterprise systems have 100 or less employees (notreported).Table 2: Means and shares for the enterprise systemsNo system All systems ERP SCM CRMShare of entire population 0.236 0.278 0.636 0.442 0.513Labor Productivity Mean 167.0(269.0)234.2(245.8)205.6(215.8)217.1(219.8)204.8(229.4)Size Mean 43.1(70.1)566.3(1605.1)359.5(1132.1)441.3(1319.7)357.8(1229.6)share of manufacturing sector 0.457 0.667 0.625 0.676 0.538Notes: Standard errors in parentheses.Source: ZEW ICT survey 2004, 2007 and own calculations.4 Empirical results4.1 Returns to enterprise systems in generalTable 3 reports the basic estimationresults, using the regression formulation describedin equation (1). The firm-level labor productivity isregressed on production input variables and anindicator of each enterprise software applicationwith additional controlling for industry and thegeographical region of East Germany. As thecurrent analysis focuses particularly on the impactsof enterprise systems on firm performance, I willdiscuss other factors influencing labor productivityonly briefly. Overall, the input factors labor, capitaland the share of computers per worker turn out tobe highly significant at the one percent level inevery single one of the following specifications,indicating a high impact of all three factors on laborproductivity.In order to provide a suitable baseline, I firstlyintroduce the three enterprise systems one by one in
  7. 7. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071101 | P a g ethe production function estimation. Overall, I findthat firms using ERP or SCM show greaterperformance in terms of labor productivity thanfirms without these systems. Both point estimatesare statistically significant at the one percent level.For instance, the estimate of 0.152 in Column (1)of Table 3 indicates that ERP users demonstrate agreater labor productivity, averaging 15.2 percentabove those firms which do not use ERP. Bothcoefficients show a similar order of magnitude,with SCM having the highest impact. Thecoefficient of CRM, on the other hand, turns out tobe only weakly significant at the ten percent levelas shown in Column (3).Table 3: Returns of enterprise systems evaluated individuallyDependent Variable:Labor Productivity (1) (2) (3)ln (labor) -0.111***(0.026)-0.111***(0.026)-0.102***(0.026)ln (capital) 0.125***(0.019)0.124***(0.019)0.127***(0.019)share of computerworkers0.584***(0.106)0.593***(0.106)0.593***(0.108)ERP 0.152***(0.055)- -SCM - 0.164***(0.051)-CRM - - 0.082*(0.049)Constant 5.697***(0.166)5.703***(0.166)5.727***(0.167)Control variables Industry,EastIndustry,EastIndustry,EastR20.234 0.235 0.230Number of Observations 927Notes: *** p<0.01; ** p<0.05; * p<0.1; robust standard errors in parentheses.Source: ZEW ICT survey 2004, 2007 and own calculations.As all three systems are expected to contribute toperformance enhancement, omitting one of themcould upwardly bias all the returns of the observedone (Aral et al. 2006). In order to check for thispotentially omitted variable bias and to give a firstinsight of the interacting nature of enterprisesystems, Table 4 reports the regression resultstaking all possible combinations of enterprisesystems into account. If the assumption of omittedvariable bias is true, one would expect a decrease insignificance and in the size of the coefficients onceall enterprise systems are integrated together in theregression. Overall, ERP and SCM stay significantif one additionally controls for CRM. Abstractingfrom SCM leads to a higher significance of ERP, asshown in Column (2) of Table 4. The coefficient ofCRM stays insignificant in all regressions.Moreover, every coefficient decreases in magnitudeonce the adoption of another enterprise system iscontrolled for. This result confirms the assumptionof omitted variable bias implying that once theenterprise systems are considered simultaneously inthe regressions the performance impact of oneenterprise system is to some extent explained bythe other one with SCM explaining the major partin the available dataset. The bias might even turnout to be larger if the systems are used incombination as potential performance gainsresulting from a complementary relationship amongthese interacting systems are not explicitlyrevealed. The possible complementarity relationbetween the systems is addressed in the nextsection.Table 4: Returns to enterprise systemsDependent Variable:Labor Productivity (1) (2) (3) (4)ln (labor) -0.119***(0.026)-0.112***(0.026)-0.111***(0.026)-0.119***(0.026)ln (capital) 0.120***(0.019)0.123***(0.019)0.123***(0.019)0.120***(0.019)share of computer 0.584*** 0.570*** 0.584*** 0.566***
  8. 8. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071102 | P a g eworkers (0.105) (0.107) (0.107) (0.106)ERP 0.120**(0.055)0.138**(0.057)- 0.118**(0.057)SCM 0.139***(0.051)- 0.152***(0.054)0.135**(0.053)CRM - 0.049(0.050)0.034(0.052)0.012(0.053)Constant 5.669***(0.166)5.688***(0.167)5.697***(0.167)5.667***(0.167)Control variables Industry,EastIndustry,EastIndustry,EastIndustry,EastR20.239 0.234 0.236 0.239Number of Observations 927Notes: *** p<0.01; ** p<0.05; * p<0.1; robust standard errors in parentheses.Source: ZEW ICT survey 2004, 2007 and own calculations.The large impact of SCM is not surprising at all,bearing in mind the potential benefits a SCMsystem has to offer and the positive impacts ofSCM on firm performance, as reported by manystudies focusing on SCM usage, e g. Dehning et al.(2007) or Hendricks et al. (2007). For CRM usage,however, there is nearly no evidence of positiveimpacts. The performance effect of ERP, on theother hand, might already be generated muchearlier. Thus, it does not show high significanceanymore, as most firms install their ERP systemfirst and adopt SCM and CRM applications a fewyears later. Unfortunately, the data does not provideany information about the date of purchase orimplementation of the enterprise systems.However, without an ERP system already installed,the firms would need to feed their SCM and CRMsystems from legacy systems, often in form ofspreadsheets spread out over different departmentsand subsidiaries of the firm. Without a reliableinformation basis in form of an ERP system theinformation from all areas of the company cannotbe accessed quickly, which could result in bottlenecks and idle times.4.2 Complementarity and interactionbetween the enterprise systemsTo test for complementarity Table 5reports the interaction terms controlling for theadoption of any two systems together in Column(1) to (3). Indicating no complementarity betweenany two enterprise software applications at a firstglance no interaction term turns out to besignificant in the regressions containing twodifferent enterprise systems. Even the performanceimpact of the enterprise systems individually dropsdown to zero with the exception of ERP whichshows a small impact in Column (2). However, thespecifications in Column (1) to (3) only control fortwo systems at a time neglecting potential influenceof the third system. Accordingly, the specificationused in Column (4) controls for all possibleinteractions between the three software systems. Inthis specification, the influence of ERP turns out tobe significant, indicating an increase in laborproductivity for firms using this kind of system.Striking, however, is the significance, even only atthe ten percent level, of the interaction term whichcaptures the use of all three software systemstogether. According to this term, firms using ERPand additionally employing SCM and CRMsystems demonstrate a considerable increase inlabor productivity. Complementarity between thethree different systems seems not to be directlyrealized through the adoption of only two systems,but SCM and CRM complement each other onceERP is already in use or implemented inconjunction. The result confirms thecomplementary benefits outlined in section 1 thatutilizing SCM and CRM without an ERP systemrunning might not be as useful as otherwise.Linking front and back office an ERP systemprovides the necessary infrastructure to feed theneeded data to the SCM and CRM system. Theperformance effects generated through the use of anERP system are increased even further once thefirms adopt the other two systems. Therefore,adopting the full suite of enterprise systems turnsout to be most useful for firms if they can rely onthe needed data infrastructure provided by an ERPsystem in first place.As specification (1) of section 2 is ratherparsimonious in terms of variables which affectlabor productivity besides enterprise systems andtraditional inputs, the results might to some extentbe driven by unobserved heterogeneity with respectto firm characteristics. Therefore, a reduction of theheterogeneity bias by introducing more variablesaffecting productivity seems reasonable, especiallyas the evidence is based on weak significance. Forthis kind of robustness check I introduce twoadditional dummy variables in the estimation, thefirst one taking the value one if the firm reports any
  9. 9. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071103 | P a g eexporting activity and zero elsewise. Compared tofirms that are only active in their home market,firms engaged in export activities are more exposedto international market pressure. Those firms aregenerally used to making quick adjustments inresponse to changes in the market environment.They rely on a highly flexible workforce asworldwide demand may change more rapidly anddrastically than domestic demand. Therefore,export activity is argued to positively impact laborproductivity (Baldwin and Gu 2004; Bernard andJensen 2004). The second dummy variableintroduced takes the value one if the firm hasestablished a works council. With an establishedworks council, employees are expected to show ahigher degree of identification with their enterpriseand the corresponding decisions made, encouragingthem to feel more committed to the company andconsequently do a better job. In addition, employeeparticipation in decision-making might balanceproduction more effectively to eliminate bottle-necks or interruptions of the production process.Hence, the establishment of a works council shouldalso lead to higher labor productivity (Zwick 2003).Column (5) of Table 5 reports the result onceworks council and export activity is controlled for.As expected, both estimates show a significantpositive impact on labor productivity. Concerningcomplementarity between the enterprise systems,the two newly introduced variables do not affectthe relationship much. Both relevant coefficients,ERP and the interaction term of all three systems,decrease in size but still show a significant impacton labor productivity.Table 5: Returns to enterprise systems – interaction between the systemsDependent Variable:Labor Productivity (1) (2) (3) (4) (5)ln (labor) -0.120***(0.026)-0.112***(0.026)-0.110***(0.026)-0.121***(0.026)-0.146***(0.028)ln (capital) 0.120***(0.019)0.123***(0.019)0.122***(0.019)0.120***(0.019)0.110***(0.019)share of computerworkers0.568***(0.105)0.571***(0.107)0.582***(0.107)0.572***(0.107)0.572***(0.106)export activity - - - - 0.148***(0.057)works council - - - - 0.171**(0.069)ERP 0.107(0.068)0.135*(0.072)- 0.185**(0.084)0.154*(0.085)SCM 0.103(0.085)- 0.072(0.076)0.171(0.116)0.139(0.118)CRM - 0.044(0.089)-0.023(0.070)0.046(0.121)0.045(0.123)ERP & SCM 0.049(0.104)- - -0.196(0.151)-0.186(0.151)ERP & CRM - 0.007(0.106)- -0.179(0.149)-0.189(0.150)SCM & CRM - - 0.138(0.100)-0.134(0.180)-0.110(0.183)All three systems - - - 0.419*(0.219)0.408*(0.220)Constant 5.678***(0.169)5.690***(0.168)5.711***(0.167)5.664***(0.169)5.574***(0.171)Control variables Industry,EastIndustry,EastIndustry,EastIndustry,EastIndustry,EastR20.239 0.234 0.237 0.244 0.255Number of Observations 927Notes: *** p<0.01; ** p<0.05; * p<0.1; robust standard errors in parentheses.Source: ZEW ICT survey 2004, 2007 and own calculations.4.3 Alternative procedure to test forcomplementarityAs a robustness check I conduct thelikelihood-ratio test from Lokshin et al. (2007).Table 6 reports the computed log-likelihood valuesof the unconstrained and constrained models, thelikelihood-ratio statistics and p-values. I rely on themethod developed by Shapiro (1985) to generate
  10. 10. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071104 | P a g ethe needed weights. Overall, the likelihood-ratiostatistics turn out to be small in the first two casesand the test rejects the hypothesis of pairwisecomplementarity for the combinations of ERP andeither SCM or CRM. However, in the third case,the test reports a highly significantcomplementarity relationship between SCM andCRM. The complementarity relationship isunconditional on ERP adoption indicating thatcomplementary benefits are realized even if an ERPsystem is not installed. Overall, this result isslightly different to the results obtained based onthe interaction term specification. However, thisrecent testing method still lacks empirical evidence.Accordingly, one has to treat the results withappropriate care. As this robustness check is notsufficient I also conducted the likelihood-ratiotesting procedure of for two practices, namely SCMand CRM, but only if ERP is also in use. Theappropriate one-sided t-test reports acomplementarity relationship significant at the onepercent level and thus strengthens the results ofTable 5.Table 6: Multiple restrictions test for complementarityComplemen-tarity RelationUncon-strainedInequalityConstrained ≥ 0InequalityConstrained ≤ 0EqualityConstrainedLR-Statistic P-ValueERP – SCM -658.475 -659.319 -659.804 -660.653 2.668 0.117ERP – CRM -658.475 -659.694 -659.499 -660.709 2.420 0.134SCM – CRM -658.475 -658.768 -662.311 -662.604 7.672 0.010Note: In the LR tests the null corresponds to the value in italics and the alternative corresponds to the equalityconstraint. In order to conclude in favor of complementarity or substitability the Log-Likelihood value with theinequality constraints should be significantly larger than the Log-Likelihood value with the equality constraint.Source: ZEW ICT survey 2004, 2007 and own calculations.5 ConclusionWith the first wave of implementationsdating back to the early 1990s, enterprise systemsare nowadays widely spread and broadly acceptedamong industries and firms of all sizes. Their exactinfluence on firm performance, however, remainsto be discovered and revealed as it is still unknownwhether the interactions between various enterprisesystems affect performance in a different way thanthe reliance on a single application.The paper approaches this question from adifferent and novel angle as it focuses explicitly ondisentangling the productivity impacts caused bythe combinations of different enterprise systems.The results provide empirical evidence of theimpacts of the three major enterprise systems onlabor productivity and display possiblecomplementarities among the systems. It is shownthat SCM and CRM function as complements,especially if an ERP system is already in place andprovides the necessary IT-infrastructure for bothenterprise software applications. Thecomplementarity relationship turns out to be robusteven if a non-parsimonious labor productivityspecification is used for inference. Hence, theproductivity gains caused by enterprise systemimplementation are not only generated by the usageof one single system, but rather augmented andincreased by adopting the three major enterprisesystems together. In consequence, previousestimations of the productivity impacts due toenterprise software usage might be biased as longas any interaction effects of the systems are nottaken into account and pictured adequately.This analysis faces a few caveats which areprimarily related to data restriction. First, apotential short-coming of this analysis might be thefact that I could only check for the effects ofadopting different enterprise software systemswithout controlling for other obstacles interferingwith enterprise system usage. This may be adrawback since not only the adoption of otherenterprise systems might influence the productivitygains of one system. Special IT-training or thequality of updates and maintenance, to name only afew examples, may also affect the performanceeffects of enterprise software. The dataset alsodoesn’t contain information on the usage ofEnterprise Application Integration (EAI) software.This software enhances the communicationbetween the different enterprise systems andenables them to check for redundant informationthereby increasing the performance of the systems.Without information about EAI the estimatedimpacts of enterprise systems might bedownwardly biased as with EAI software runningthe impacts of enterprise software on laborproductivity can be expected to be larger. However,due to data constraints I have to pass theexploration of the productivity effects generated byEAI software usage on to future research withnewly obtained data. Besides EAI software, it mayalso be the case that some companies might haveless legacy systems compared to other firms,allowing them to implement next generationenterprise software faster and consequentlyrealizing performance benefits earlier. Futureavailability of new data may provide evidence even
  11. 11. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071105 | P a g efor these cases. Lastly, a potential endogeneity biascannot be excluded completely. Although based ona well-defined temporal sequence incorporating atwo year lag between software usage andperformance measured it is not clear if the firmspossess the software to increase performance orbecause they are already performing well acquiringthe software as an additional asset. However, theworks council captures such strategic firmdecisions to some extent as it has the option toconfirm or deny certain decisions. Accordingly, asthe estimations control for the establishment of aworks council the mentioned bias due tounobserved strategic decisions of firms can beexpected to be negligibly small in size.6 Appendix Chapter 1Table 6: Distribution of the combinations of enterprise systemsFrequency Percentno enterprise system 219 23.62Only ERP 121 13.05Only SCM 23 2.48Only CRM 54 5.83Only ERP and SCM 88 9.49Only ERP and CRM 123 13.27Only CRM and SCM 41 4.42All three systems 258 27.83number of observations 927 100Source: ZEW ICT survey 2004.Survey-Question for enterprise system usage:Which application or system do you use in your daily business routine? Please state if the application or systemis used to minor degree, broadly or not at for planning and controlling, e. g. SAP/R3 (ERP system from SAP)customer relationship management (CRM)supply chain management (SCM)Table 8: Distribution of enterprise systems usage (in percent)ERP SCM CRMno usage 36.35 55.77 47.65minor usage 16.83 25.46 32.79broad usage 46.82 18.77 18.55number of observations 927 927 927Source: ZEW ICT survey 2004.References[1]. Aral S, Brynjolfsson E, Wu D J (2006)Which came first, IT or productivity? Thevirtuous cycle of investment and use inenterprise systems.Proceedings of the 27thConference on Information Systems,Milwaukee[2]. Aral S, Aral S, Brynjolfsson E, Wu L(2011) Three-Way Complementarities:Performance Pay, Human ResourceAnalytics and Information Technology.Management Science – Articles inAdvance: 1-19.[3]. Atrostic B K, Nguyen S (2005) ComputerInvestment, Computer Networks andProductivity. Working Papers 05-01,Center for Economic Studies, U.S. CensusBureau[4]. Athey S, Stern S (1998) An empiricalframework for testing theories aboutcomplementarity in organizational design.NBER working paper 6600[5]. Athey S, Stern S (2002) The Impact ofInformation Technology on EmergencyHealth Care Outcomes. The RANDJournal of Economics 33 (3): 399-432
  12. 12. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071106 | P a g e[6]. Baldwin J R, Gu W (2004) Export-marketParticipation and ProductivityPerformance in Canadian Manufacturing.Canadian Journal of Economics 36 (3):634-657[7]. Baumol W, Panzar J C, Willig R D (1988)Contestable Markets and the Theory ofIndustry Structure. Harcourt BraceJovanovich, San Diego[8]. Belderbos R, Carree M, Lokshin B, (2006)Complementarity in R&D cooperationstrategies. Review of IndustrialOrganization 28 (4): 401-426[9]. Bernard A B, Jensen J B (2004) Exportingand Productivity in the USA. OxfordReview of Economic Policy 20: 343-357[10]. Bertschek I, Fryges H, Kaiser U (2000)B2B or Not to Be: Does B2B E-Commerce increase Labour Productivity?International Journal of the Economics ofBusiness 13 (3): 387–405[11]. Bond S, Van Reenen J (2007)Microeconomic Models of Investment andEmployment. Handbook of Econometrics6 (1): 4417-4498[12]. Cachon G, Fisher M (2000) Supply ChainInventory Management and the Value ofshared Information. Management Science46 (8): 1032-1048[13]. Carree M, Lokshin B, Belderbos R (2007)A note on testing for complementarity andsubstitutability in the case of multiplepractices. Journal of Productivity Analysis35: 263-269.[14]. Charkari N M, Abdolvand N (2004) AProposed Model in Integrating SCM,CRM and ERP. In: The First InternationalConference on Information andKnowledge Technology: 18-24[15]. Chen I J (2001) Planning for ERPSystems: Analysis and future Trends.Business Process Management 7 (5): 374-386[16]. Cohen M A, Agrawal N, Agrawal V(2006) Winning in the Aftermarket.Harvard Business Review 84 (5): 129-138[17]. Conlon G (1999) Wired Executive:Growing Sales from existing Customers.Sales and Marketing Management 135[18]. Cotteleer M J, Bendoly E (2006) Orderlead-time Improvement followingEnterprise-IT Implementation: Anempirical Study. Management InformationSystems Quarterly 30 (3): 643-660[19]. Dehning B, Richardson V J, Zmud R W(2007) The financial Performance Effectsof IT-Based Supply Chain ManagementSystems in manufacturing Firms. Journalof Operations Management 25 (1): 806-824[20]. Engelstätter B (2011) Its not all aboutPerformance Gains – Enterprise Softwareand Innovations. Economics of Innovationand New Technology 21 (3): 223-245.[21]. Greenan N, Mairesse J (2000). ComputersAnd Productivity In France: SomeEvidence. Economics of Innovation andNew Technology 9 (3): 275-315.[22]. Hall B H, Lotti F, Mairesse J (2009)Innovation and Productivity in SMEs:Empirical Evidence for Italy. SmallBusiness Economics 33: 13-33[23]. Hendricks K B, Singhal V R, Stratman J K(2007) The Impact of Enterprise Systemson Corporate Performance: A Study ofERP, SCM and CRM SystemImplementations. Journal of OperationsManagement 25 (1): 65-82[24]. Hitt L M, Wou D J, Zhou X (2002)Investment in Enterprise ResourcePlanning: Business Impact andProductivity Measures. Journal ofManagement Information Systems 19 (1):71-98[25]. Jacobsen S, Shepherd J, D’Aquila M,Carter K (2007) The ERP Market SizingReport, 2006-2011. AMR Research 20495[26]. Jancic Z, Zabkar V (2002) Interpersonalvs. Personal Exchanges in MarketingRelationships, Journal of MarketingManagement 18 (7-8): 657-671[27]. Mabert V A, Soni A K, VenkataramananM A (2000) Enterprise resource planningsurvey of US manufacturing firms.Production & Inventory ManagementJournal 41 (20): 52-58[28]. Matolcsy Z P, Booth P, Wieder B (2005)Economic Benefits of Enterprise ResourcePlanning Systems: Some empiricalEvidence. Accounting and Finance 45 (3):439-456[29]. McAfee A (2002) The Impact ofEnterprise Information TechnologyAdoption on Operational Performance: AnEmpirical Investigation. Productions andOperations Management 11 (1): 33-53[30]. McAfee A, Upton D (1996) VandelayIndustries. Harvard Business School Case# 9-697-037. Harvard Business SchoolPublishing, Boston[31]. Mehrjerdi Y Z (2010) Enterprise resourceplanning: risk and benefit analysis.Business Strategy Series 11 (5): 308-324[32]. Milgrom P, Roberts J (1990) TheEconomics of Modern Manufacturing:Technology, Strategy, and Organization.American Economic Review 80 (3): 511-528[33]. Mithas S, Krishnan M S, Fornell C (2005)Why do Customer Relationship
  13. 13. Benjamin Engelstätter / International Journal of Engineering Research and Applications(IJERA) ISSN: 2248-9622 www.ijera.comVol. 3, Issue 3, May-Jun 2013, pp.1095-11071107 | P a g eManagement Applications affectCustomer Satisfaction? Journal ofMarketing 69 (4): 201-209[34]. Mohnen P, Röller L-H (2005)Complementarities in Innovation Policy.European Economic Review 49 (6):1431-1450[35]. Morgan R M, Hunt S D (1994) TheCommitment-trust Theory of RelationshipMarketing. Journal of Marketing 58: 20-38[36]. Nicolaou A I (2004) Firm PerformanceEffects in Relation to the Implementationand Use of Enterprise Resource PlanningSystems. Journal of Information Systems18 (2): 79-105[37]. Nicolaou A I, Stratopoulus T, Dehning B(2003) Financial Analysis of PotentialBenefits from ERP Systems Adoption.Journal of Business and InformationTechnology 2 (1): 40-50[38]. Ohnemus J (2007) Does IT OutsourcingIncrease Firm Success? An EmpiricalAssessment using Firm-Level Data. ZEWDiscussion Paper 07-087[39]. Scott J E, Vessey, I (2000) ImplementingEnterprise Resource Planning Systems:The Role of Learning from Failure.Information Systems Frontiers 2000 2 (2):213-232[40]. Shapiro A (1985) Asymptotic Distributionof Test Statistics in the Analysis ofMoment Structures under InequalityConstraints. Biometrika 72 (1): 133-144[41]. Sheth J N, Sisodia R S, Sharma A (2000)The Antecedents and Consequences ofcustomercentric Marketing. Journal of theAcademy of Marketing Science 28 (1):55-66[42]. Shin I (2006) Adoption of EnterpriseApplication Software and FirmPerformance. Small Business Economics26 (3): 241-256[43]. Wieder B, Booth P, Matolcsy Z P,Ossimitz M-L (2006) The Impact of ERPSystems on Firm and Business ProcessPerformance. Journal of EnterpriseInformation Management 19 (1): 13-29[44]. Zwick T (2003) Works Councils and theProductivity Impact of Direct EmployeeParticipation. ZEW Discussion Paper 03-47