Innovation and ICT in service firms: towards a multidimensional approach for impact assessment. By David Gago and Luis Rubalcaba

J Evol Econ (2007) 17:25–44DOI 10.1007/s00191-006-0030-8R E G U L A R A RT I C L EInnovation and ICT in service firms:towards a multidimensional approachfor impact assessmentDavid Gago & Luis RubalcabaPublished online: 23 September 2006# Springer-Verlag 2006Abstract This article focuses on the multimodal character of innovation in servicesfirms as the analytical framework by which to assess the role of different sources andagents, ICT in particular, in enabling various impacts of innovation. The peculiaritiesof service innovation require a wider approach than that observed for goodsinnovation, which is less focused on non-technological aspects. An ad hoc surveywas carried out in the region of Madrid to test a microeconometric andmultidimensional approach at the firm level. Specific impacts of innovation areexamined by carrying out an ordered probit model with sample selection. Resultsindicate a certain correspondence between the multidimensional nature of serviceinnovation and a preliminary impact assessment. The paper notes that ICT andclients–providers interactions are both important, acting to facilitate different typesof service innovation.Keywords Innovation . Services . Ordered probit model .Information and Communications Technology (ICT) . Impact assessmentJEL Classification O33 . C35 . L8 . L25D. GagoDepartment of Applied Economy, University of Alcalá,Madrid, Spaine-mail: david.gago@telefonica.netL. Rubalcaba (*)Department of Applied Economy, University of Alcalá and Servilab,Madrid, Spaine-mail: luis.rubalcaba@uah.esL. RubalcabaEuropean Network for Research on Services, RESER,Plaza de la Victoria 2, 28802 Alcalá de Henares,Madrid, Spain26D. Gago, L. Rubalcaba1 IntroductionDespite the continuous growth of the service sector in the advanced economies,services have long been perceived as non-innovative or technologically backwardactivities. It was only during the 1990s that the traditional conception of services asinnovation laggards gradually changed. The earlier studies that paved the way forthis shift were mainly focused on the use of technologies by services activities,notably ICT (Information and Communication Technologies) in creative rather thanstandard ways (Miles and Ducatel 1994; OECD 1996; Antonelli 1998). In thisscenario, service innovations were implicit in the hardware components andtransferred when implemented by service industry users. This pattern of innovationdepicted service innovative trajectories as supplier-dominated.ICT somehow helped transform the passive image of services in relation toinnovation so that they have become an important locus for innovative activity withinthe emerging ‘knowledge economy’ (Metcalfe and Miles 2000). The actual innovationand implementation is thus initiated by and implemented throughout the organization,possibly with ‘innovation support’ from outside. Moreover, the service firm may alsoinfluence the innovation process that takes place within a client firm by providingknowledge resources that support the innovation process. In this way, the service firmmay constitute a source of innovation if it plays a major role in initiating anddeveloping innovation in client firms, usually in close interaction with the client firm.In this context, the role of Knowledge Intensive Business Services (KIBS) must behighlighted (Miles et al. 1995; Rubalcaba 1999; Wood 2001) and considered as part ofinnovation systems (Antonelli 1999; Hipp 2000). Despite the relatively minor role ofICT as sources of services innovation in analysis based on European CIS data (Tether2003), ICT act as innovative drivers of services when services innovation is based onnew ways of client–provider co-production or on the more advanced use of businessservices, both KIBS and traditional professional services.The increasing importance of services in relation to innovation has greatlybenefited from the still on-going debate as to whether service innovation should beanalyzed using the same concepts and tools as innovation in manufacturing. Theassimilation approach, which treats services as similar to manufacturing (Coombsand Miles 2000) and epitomises the passive role of services as mere technologyadopters, seems to have been definitely left behind. In an attempt to move away fromwhat might be seen as manufacturing based models of innovation, the more recentapproaches have either sought to highlight the distinctiveness of service innovationin terms of the innovation in manufacturing (the demarcation approach) or broughtto the forefront hitherto overlooked elements of innovation, basically of non-technological content, such as human and organizational capabilities, which are ofrelevance for manufacturing as well as for services. Specific service innovation maybe identified in all economic sectors, whether these are goods or service sectors,through the presence of innovative intangibles or, when appropriate, through the“encapsulation” described by Howells (2004).Survey based studies have also followed the evidence gained at the theoreticallevel to put forward a broader conception of innovation in which often neglecteddimensions are considered. The shift of focus has been decisive in shedding someInnovation and ICT in service firms: towards a multidimensional approach for impact assessment27light on the so called productivity paradox (i.e. the fact that productivity measures donot seem to show any impact derived from the massive deployment of ICT), whichis of special relevance in services. In fact, one of their explanatory factors may bethat, unlike manufacturing, innovation in services is often neither represented bynew services nor by process improvements which increase outputs or decreaseinputs. As Licht and Moch (1999) put it, innovation in services is often more closelyconnected to the way products are delivered, such as the loosening of time-and-space restrictions or the increasingly just in time nature of services. This reinforcesthe hierarchy of services as innovative activities.Figure 1 presents the main possible interrelations between ICT and serviceinnovation based on a multidimensional view of innovation in services and theprominent role of co-productions between providers and users, and the comple-mentarities between technological and non-technological components of servicesinnovation in a given innovation system. ICT are drives, facilitators and agents offour non-independent types of services innovation: product and process, organiza-tion, interfaces and co-production schemes, and business services and KIS-relatedinnovation.Bearing these issues in mind, the present paper emphasizes the multidimensionalcharacter of innovation in service industries in the sense that, as has been said, thismay reflect not only enlarged output or a lesser use of inputs (i.e. increases inproductivity levels) but also incorporate other non-material or disembodied natureaspects. As a result, measuring innovation impacts in services by means ofconventional indicators such as sales increases does not seem to be an appropriatemethod to capture fully the peculiarities of service innovation. Impact assessment ofservice innovation should be correspondent to its several dimensions so thatindicators based on the co-productive nature of services (e.g., quality, trust, timing,motivation) may be as important as other traditional indicators (e.g., costs,productivity, employment, turnover, internationalisation). In most of the old andFig. 1 Interactions between ICT and service innovation. Source: Own elaboration28D. Gago, L. Rubalcabanew indicators for impact assessment, we expect to find significant links with thedriving role of users–clients and ICT.The rest of the article is organized as follows. The first section briefly describesthe database, namely the Survey on Innovation 1998–2001 carried out in 2002–2003. In the second section, the analysis focuses on a brief description andjustification of the econometric method used, whereas in the third section somemethodological aspects related to the model are put forward – basically thedescription of the selection bias problem – as well as the final specification adoptedto circumvent this problem and to avoid the harmful effects it may cause on theestimation. The fourth section contains the core results of the essay, as it provides theresults of the estimation as well as the interpretation. We then finish with someconcluding remarks and suggestions for further research.2 The database: the Madrid Survey on Service InnovationThe Madrid Survey on Service Innovation was undertaken between 2002 and 2003.Valid data of 557 enterprises were collected through a mail survey of Madrid servicefirms. The response rate was around 45% and the sample error was below +−5% (tobe exact +−3.99%).Ten different types of services were considered in the survey: hotels andrestaurants, transports, telecommunications, temporary work, engineering andarchitectural services, and security and other ancillary business services. These arethe services that have traditionally been more exposed to innovation (Sirilli andEvangelista 1998). The justification of the choice lies in the fact that one of the mainobjectives of the Survey was to study in depth certain aspects directly linked to thedynamics of innovation, rather than merely studying the percentage of innovativeenterprises, and so it seems plausible to focus the analysis on the service activitiesmore prone to innovation.Amongst other aspects, enterprises surveyed were asked about the existence/nonexistence of innovation, thus distinguishing among product and process on the onehand, and organizational innovation on the other. In this respect, it is worthmentioning that some scholars have long stressed the difficulties in determining theorientation of innovation in services. A good example is Hipp and Tether (2002),who argue that there is often a close nexus between the service provided, the processof provision and the organization of provision, such that it is often difficult to changeone without impacting on the others.This paper explicitly considers this peculiarity of service innovation by presentingproduct and process innovation results jointly, but aside from organizationalinnovation, as it underpins impacts of a different nature. In fact, the specific focusof the paper on impacts of innovation has called for a particular presentation of thetraditional categories of innovation, so as to tackle the objectives of the analysis,bearing in mind that they are so much interrelated. In order to ensure properunderstanding of the categories by respondents in our survey, explanations andprecise guidelines about what is termed product, process or organizationalinnovation were given, including examples of each type.Innovation and ICT in service firms: towards a multidimensional approach for impact assessment29The survey, which accomplishes the first ever comprehensive attempt to measureservice innovation impacts in Spain, is based on the widely accepted European CIS IIImethodological framework. Nevertheless, some particular issues that did not deservespecial attention in the Community Survey are analyzed in detail in this survey.In the first place, the Madrid survey emphasizes the key role played by theInformation and Communication Technologies (ICT) as sources of serviceinnovation, echoing the aforementioned major services use as well as the evidenceon the ample complementarities between investments in ICT and innovations(Brynjolfsson and Hitt 2002).Secondly, the Survey is organized under a graduation of product, process andorganizational innovation impacts on a Likert scale, which means that the interest notonly lies on the mere occurrence/non occurrence of innovation, but also on thedegree of such impacts. This is in accordance with the Oslo Manual (differenteditions), which clearly distinguishes amongst ‘new products and processes andsignificant improvements in such products or processes’.On the other hand, a complete coverage of agents that may promote innovationhas been considered in the Survey, thus distinguishing amongst the enterprise andlinked entities (such as the competition, professional associations, etc.), providers,clients and institutions linked to the Public Administration.Following some previous attempts such as the Canadian Service Innovation Survey(1996) or the Mannheim Innovation Panel for the Service Sector (MIP-S), the impactsof innovation captured by the Survey go beyond the traditional effects on productivityand costs to allow for others, such as service quality, employment and skills or market/product growth. In relation to this, the Survey also includes a specific question onorganizational innovation, which is based on the capabilities induced by innovation tocentralize/de-centralize service tasks, enterprise re-location, task outsourcing, higherlevels of enterprise co-operation either by strategic alliances or networking.On the other hand, unlike the still large neglect of services as arenas forgovernment intervention policy (Howells 2000), the Survey is somehow policyoriented, in the sense that it includes an assessment (according to the enterprisessurveyed) of the alternative policies at the disposal of public bodies to promoteservice innovation. However, these policy issues are not discussed here since theyare not of importance in the present paper.In short, the pioneering character of the surveys is due to the followingarguments: inclusion of the services more prone to innovation (more appropriatecoverage of the innovation phenomenon); particular treatment of ICT as sources ofinnovation; distinction amongst levels of innovation impacts; a complete coverage ofagents promoting innovation; systematization of innovation impacts; specifictreatment of organizational impact; applied orientation of the survey: analysis ofalternative policies to promote innovation.3 The modelThe econometric model attempts to measure the influence of different variables onvarious dimensions of service innovation. The enterprises surveyed were required to30D. Gago, L. Rubalcabarank different impacts on innovation according to a Likert scale, which ranges from 1to 5, the value 1 being of ‘negligible important’, while 5 is ‘strategically important’.Thus, impacts of product and process innovation were gathered on five maindimensions. At the same time, every impact is made up of different manifestations:A)Impact on productivity and enterprise costs. A.1) Costs savings. A.2)Enhancement in labor productivity. A.3) Higher levels of employees’motivation. A.4) Increasing production capacity.B)Product or market expansion. B.1) Increasing income revenue; B.2) Morevariety of services B.3) Presence in other regions. B.4) Higher degree ofinternationalization.C)Employment and skills. C.1) Acceleration of the employment generationprocess. C.2) Capital/employment substitution. C.3) Higher use of skilledlabor. C.4) Higher use of non skilled labor.D)Service quality. D.1) Flexibility in adjusting to customers needs. D.2) Deliveryspeed. D.3) Temporal availability. D.4) Service user friendliness. D.5) Reliability.E)Environmental impact. E.1) Fulfillment of ecological and sanitary standards andregulations.On the other hand, impacts brought about by organizational innovation are listedas follows: A) Change in the number of employees. B) Expansion in the number ofpremises/establishments (multi-location). C) Decentralization of tasks. D) Taskspecialization. E) Promotion of networking/strategic alliances. F) Higher levels ofdepartment autonomy. G) Outsourcing of routine tasks. H) Outsourcing of non-routine (advanced) tasks. I) Enterprise re-location. J) Relocation of certain activities(partial relocation).Independent variables depict a priori sources/agents prompting impacts ofinnovation of any kind. The influence of enterprise size is captured through thevariable employment, which is approached by the number of employees by 31stDecember 2001 and expressed in logarithms in order to avoid scale problems thatmay cause non-convergence in this type of model. The ICT variable, on the otherhand, measures the investments on ICT as a driving and enabling source of differentinnovation impacts. The variable is of qualitative character, and is built using theaforementioned Likert scale, ranging from zero to five. Value zero is interpreted inthe sense that the enterprise has carried out no ICT investments whatsoever, andassessment of the influence on innovation is thus not possible. By contrast, valuefive implies that the enterprise has indeed carried out ICT investments that areconsidered as ‘strategically important’ in terms of sources of innovation. The sameapplies for the software variable, which is included aside from ICT on the grounds ofan a priori differential impact of this ICT component in relation to the others.Inclusion of the variables international clients and domestic clients may bejustified on rather different grounds. The first must be interpreted as the importanceattached to international clients as a source of innovation, while the latter is referredto the role played by domestic clients. Consequently, the inclusion of both variablesis intended to grasp the contribution that external knowledge may have on differentdimensions of service innovation. The same rationale lies behind the variable thatreaps the role played by computer services providers.Innovation and ICT in service firms: towards a multidimensional approach for impact assessment31The variable character is built by classifying service activities into four differenttypes, according to its higher (or lesser) dynamic character, ranging from one (thelesser dynamic category) to four (the most advanced). Seven dummy variables areincluded in the model so as to control for sectoral effects. The twelve serviceactivities from the survey are reduced to eight groups1 for the sake of clarity.Consequently, final groups are the following: 1) Hotels (which is made up of hotels,restaurants and travel agents); 2) Transport (road transport); 3) Telecom (made up bytelecommunications and computer services); 4) Consultancy (management consul-tancy and financial intermediation); 5) Engineering (architectural and engineeringservices); 6) Advertising; 7) Personal and 8) Security. The latter category is used asthe reference to avoid the so-called dummy trap.3.1 Justification of the model and methodologyIn order to assess the importance of the different dimensions of innovation, anordered probit model with selection bias correction is used, as it is the model thatbetter fits the profile and characteristics of the data.The ordered probit model is based on the following specification:yà ¼ β0xii þ "i; "i % N ½0; 1Šyi ¼ 0 if yÃi ¼ μ01 if μ0 < yà μi12 if μ1 < yà μð1Þi23 if μ2 < yà μi3::::::::::::::::::::::J if yÃiiμJThe observed counterpart to yà is yii. The variance of ɛi is assumed to be onesince, as long as yÃ, β and ɛii are unobserved, no scaling of the underlying model canbe deduced from the observed data. The ordered probit model was developed byZavoina and McElvey (1975). Since the μs are free parameters, there is nosignificance to the unit distance between the set of observed values of y. Theymerely provide the ranking. Estimates are obtained by maximum likelihood. Theprobabilities which enter the log likelihood are:ÂÃProb½yi ¼ jŠ ¼ Pr ob yà in the rank jð2ÞiA pseudo R2 based on the formula used by Zavoina and McElvey (1975) in theirpaper on the ordered probit model is computed as a measure of the model goodnessof fit: !yÃE¼ yf ¼ b0x þ 1;ð3Þy1 Groups have been constructed according to the evident affinities amongst the service activities.32D. Gago, L. Rubalcabawhere λ is the inverse of Mills ratio, defined as the quotient between density anddistribution function. Mathematically:À Áf x0bλ ¼σÀ Áð4ÞF x0bσHence, the pseudo R2 may be specified as:R2 ¼ ½varðy f Þ =ð1 þ varðy f ÞÞŠð5Þ3.1.1 The selection bias problemThe selection bias problem includes different truncation phenomena, that is, sampleextractions where the variable of interest is not used as the guide for the sampleselection. In other words, the selection of the truncated population distribution fromwhich the sample was extracted was undertaken using a different variable from whatis labelled as the observed variable (the one under study). In the model here devised,enterprises answering questions regarding the impacts of innovation are exclusivelythose where the impact is, to some extent, visible, and the sample subject to study isobtained from a dychotomic variable (i.e. the existence/non-existence of impacts oninnovations).Selection bias causes severe estimation troubles in the event that the problem isnot dealt with properly, As it may be deduced from the earlier explanation, theobserved variable y, which assesses the degree of the innovative impact according toa Likert scale), is not randomly selected from the population, but is determined bytaking other variable that captures the existence/non-existence of the innovativeimpact (labelled z*) as reference. Variable z* can only take zero (no innovation) andone (innovation) values, so that only when the variable reports one is it thenconsidered in the experiment. In this context, if the observed variable is regarded asa random variable (despite the fact it is obtained from variable z*), estimators maybe biased. The general solution to circumvent this selection bias problem lies inbuilding up an auxiliary model of the process generating the variable z*.The model, once the selection bias has been accounted for, may be expressed asfollows:yà ¼ β0x þ ɛ;zà ¼ α0v þ u;ð6Þ"; u % N 0; 0; σ2"; σ2μ; ρVariable z* (as well as y*), may not be directly observed. Conversely, thecounterpart z may be observed, and expressed in this way:z ¼ 1 if zÃ>0ð7Þz ¼ 0 if zÃ≤ 0Accordingly, y values (the observed counterpart of y*) may be observed if andonly if z =1.Innovation and ICT in service firms: towards a multidimensional approach for impact assessment334 Results4.1 Product and process innovationResults for every manifestation of product and process service innovation are listedin tables one to five of the Statistical Annex, grouped around the five dimensionsintroduced above, namely: impact on productivity and costs, product or marketexpansion, employment and skills, service quality and environment.Overall, it must be stressed that variables that turn out to be statisticallysignificant are quite similar, not only amongst impacts pertaining to the samedimension, but also across impacts attached to different dimensions. In fact,variables such as presence of international clients, importance attached to ICT,importance attached to software, and to computer services suppliers as a source ofinnovation are statistically significant (up to 10 per cent levels), in more than 50 percent of the regressions undertaken. The clustering of the effects on virtually the samevariables for every service innovation dimension grants the results a higher degree ofrobustness and a certain degree of internal coherence. The homogeneity in thevariables that report statistically significant results is reflected on very close pseudoICT2 for all dimensions of innovation (0.45 in most cases).Amongst the variables included in the regression, ICT clearly stands out from therest, as it yields positive significant effects in sixteen out of eighteen regressions, andmostly at a 1 per cent level. This result must be interpreted in line with the recentfocus on the role of ICT as a source of innovation in services (Van Ark et al. 2003)and the tight connection between the fast pace of ICT developments and theemergence of new forms of service delivery and the creation of new services (Gago2001), which is amplified by the use of these technologies from the services side.Conversely, it is worth emphasizing the lack of statistical significance in twovariables that, a priori, are expected to influence the appreciation of innovation, suchas the higher or lesser advanced character of the service, and also the sectoraldummies.As far as disaggregated results for different dimensions are concerned, the firstdimension refers to the most traditional and conventional effects of innovation,namely those occurring on level costs — productivity and production capacity. Theonly two variables bearing statistically significant coefficients are the ICT andcomputer services providers. Positive coefficients are obtained for ICT in allregressions.Computer service providers also report positive coefficients, ranging from 0.083at ‘employees motivation’ up to 0.119 at the ‘saving costs’ manifestation. Thesoftware variable is highly significant on all manifestations, except on ‘employees’motivation’. Moreover, ‘employees’ productivity’ and ‘higher production capacity’are better explained by software rather than by generic ICT. The intuition behind thisresult may be that software systems (which are included in the ICT variable as a partof it) are endowed with a higher degree of flexibility to respond adequately to thechanging profile of the current economic context. As a consequence, implementationof software systems may enhance productivity levels by the presence of networkexternalities (Licht and Moch 1999, op. cit.).34D. Gago, L. RubalcabaThe outstanding role of clients on the productivity/costs dimension is obvious.Not only is this result evident for international clients, but also for domesticclients, even though with very distinctive patterns. Hence, international clientsseem to have played a major role as factors promoting ‘costs savings’, whiledomestic customers have especially triggered the achievement of ‘higherproductivity’ and ‘higher production capacity’. On the other side, the influenceon innovation dimensions exerted by variables character and employment is quiteselective but salient in those cases in which it turns out to be statisticallysignificant. From the latter it may be concluded that the more innovative a serviceis, the higher the chances are that, as a consequence, it may increase its productioncapacity. As far as the other innovation effects on this first dimension areconcerned, the evidence is elusive. On the other hand, enterprise size approachedby employment does affect positively cost saving achievements, in such a way thatthe bigger (the smaller) the size of the enterprise, the higher (the lesser) the chancesthat cost savings may be obtained.On the other side, dummy variables accounting for sectoral effects show nosignificant effects as far as this dimension of innovation is concerned. The onlyexception is made up by ‘productivity of employees’, where the activities Hotels andtravel agents, Transports (both statistically significant at 5 per cent), Telecommu-nications and Computer services (at 10 per cent) display a significant effect.Eventually, the pseudo R2 are quite similar, being the category ‘higher productioncapacity’ the one reporting the highest goodness of fit (0.47), and ‘employees’motivation’ the lowest (0.40).The second dimension is clustered around four different manifestationsrelated to market or product expansion issues, and thereby the most traditionalaspects of innovation linked to productivity and costs are set aside.2 Resultsseem again to emphasize the role of ICT as carriers of different realms ofinnovation, thus confirming theoretical evidence on their influence on scope(approached by ‘higher variety of services’) and scale economies, the latter both interms of the most traditional perspective (‘increasing income revenues’) and themost sophisticated conception linked to geographical expansion (‘presence in otherregions’ and ‘higher levels of internationalization’). Scale economies seem toflourish only at their conventional conception, that is, associated with a productionincrease.Computer services providers command some explanatory power on a higherachievement of scale economies, but not on scope economies. On the other side, theinnovative content of the services reported here seems to play a more substantial roleon this dimension linked to market/product enlargement than that associated withproductivity enhancement. In fact, coefficients are statistically significant at the 5 percent level for ‘higher income revenues’ and ‘presence in other regions’, and at 10 percent for ‘variety of services’. Larger enterprises are relatively more prone to create ahigher variety of services as a particular manifestation of this second dimension.2 In any case, manifestations must not be considered as watertight departments in the sense that in somepoints may be heavily intertwined. In fact, market expansion may be underpinned by productivitymotivations.Document Outline

• • ÿ
    • ÿ
    • • ÿ
    • ÿ
      • ÿ
    • ÿ