Linkwp1-laursen&foss.pdf

Institut for Industriøkonomi og VirksomhedsstrategiWorking Paper 2000-5 New HRM Practices, Complementarities, and the Keld Laursen and Nicolai J. Foss
Department of Industrial Economics and Strategy, Copenhagen Business School, Howitzvej 60, 2000 Frederiksberg, Denmark 2nd draft, this version: 22 February 2000 Paper prepared for the 3rd Applied Econometrics Conference, Alicante, Spain, Acknowledgements
We are grateful to Bengt-Åke Lundvall, Torben Pedersen and Ammon Salter for
comments made on an earlier version of this paper. In addition, wish to thank the
participants in the DISKO project for allowing us to use the data applied in this paper.
In particular, we wish to thank the persons responsible for carrying out the survey,
including project leader Bengt-Åke Lundvall, Allan Næs Gjerding, Kenneth
Jørgensen, Frank Skov Kristensen, Reinhard Lund, Poul Thøis Madsen, Peter Nielsen
and Søren Nymark.
New HRM Practices, Complementarities and the Abstract
Although organisational structure has sometimes been mentioned in evolutionary
economics as well as in the innovation literature as a possible determinant of innovation
performance, very little systematic theoretical and empirical work exist on this issue. In
this paper, we take our theoretical point of departure in recent work in organisational
economics and elsewhere, on systems of human resource management practices. We put
and develop the argument that just as complementarities between new HRM practices
positively influence financial performance, they will also positively influence innovation
performance. We examine this overall hypothesis by estimating an empirical model of
innovation performance, using data from a Danish survey of 1900 business firms. Using
principal components analysis we identify two HRM systems which are conducive to
innovation. The first is one in which all of our nine HRM variables matter (almost)
equally for the ability to innovate. The second system, which is found to be conducive to
innovation is dominated by performance related pay and to some extent by firm-internal
training. Of our total of nine sectors we find that the four manufacturing sectors correlate
with the first system, while also firms located in ICT intensive service sectors are
associated with the first system. Firms belonging to the wholesale trade sector tend to be
associated with the second system.
JEL classification:
C25, D23, O32
Keywords
Innovation, human resource management practices, organisational complementarities,
evolutionary economics.
I. Introduction: New HRM Practices and Innovation The ongoing re-structuring of management practices designed to cope with an increasinglycomplex and rapidly changing knowledge-based economy has received increasing attentionfrom scholars from a diversity of disciplines and fields (Bowman and Singh, 1993; Zengerand Hesterly, 1997). In particular, much attention has been given to the restructuring of theemployment relation in the form of changed human resource management (hencefort,“HRM”) practices that has accompanied the emergence of firms specialised to competing indynamic, information-rich environments. These practices encompass various types of team-based organisation, continuous (often internal and team-based) learning, decentralisation ofdecision rights and incentives, systems for mobilising employee proposals forimprovements, quality circles, emphasis on internal knowledge dissemination, etc. (Zengerand Hesterly, 1997; Mendelson and Pillai, 1999).
While many of these new practices may not, strictly speaking, be entirely novel, some of the broad generalisations about new HRM practices refer to trends that appear to be trulyrecent. Thus, new HRM practices appear to follow a very steep diffusion curve; they tendto be adopted in a system-like manner rather than as individual components; and they tendto be associated with high innovation performance (Mendelson and Pillai 1999). It is these“stylised facts” that we try to theoretically and empirically address and substantiate in thispaper.
The increased attention paid to new HRM practices has been particularly prevalent in the fields of strategic management, human resource management, and, increasingly, theeconomics of organisation. For example, strategy scholars have argued that humanresources are significant strategic levers, which are particularly likely to the sources ofsustained competitive advantage and that HRM practices should therefore be central tostrategy (Barney, 1991; Barney, 1995). A reason for this is the system-like  or, in theterminology that we shall make use of, “(Edgeworth) complementary”  way in whichHRM practices may connect. The complementary nature of many of the elements of(formal and informal) organisational structure has been examined in an emerging importantliterature in organisational economics (notably Milgrom and Roberts, 1990; Aoki and Dore,1994; Milgrom and Roberts, 1995; Holmström and Roberts, 1998). Insights from thisliterature have made some impact in the human resource management field (Baron andKreps, 1999).
Although the connection between internal organisation issues and innovativeness has certainly never been neglected in the innovation and evolutionary economics literature after all, the increasing bureaucratisation of the R&D function was a key theme inSchumpeter’s later work  it is also fair to say that these literatures are characterised byrelatively scant attention being paid to new (complementary) HRM practices and on howthey influence innovation performance.1 1 The clear exception is constituted by some scholars’ interest in Japanese economic organisation and howthis connects to innovativeness. Thus, Freeman (1988: 335) explicitly notes how in “Japanese management,engineers and workers grew accustomed to thinking of the entire production process as a system and ofthinking in an integrated way about product design and process design,” and he makes systematic reference toquality management, horizontal information flows, and other features of new HRM practices. One could alsoconstruct an argument that already the concern with horizontal information flows in the late 1960s ProjectSAPPHO demonstrates a long-standing awareness of the relation between HRM practices and innovation In fairness, it must be said that the lack of theoretical and empirical treatment of how new HRM practices impact on innovation performance is a fairly general characteristic ofthe literature on HRM practices. Thus, there is an emerging theoretical and empiricalunderstanding of how HRM practices and, in particular, complementarities between theseimpact on productivity and, in turn, on financial performance, but that understanding needsto be extended to also encompass innovation performance. To do this is the purpose of thepaper. Thus, we shall argue and empirically demonstrate that new HRM practices andcomplementarities between these impact on innovation performance, that is, on futurecompetitive advantages. This evidence is, in our view, sufficient reason that evolutionaryeconomists and innovation scholars should take an interest in new HRM practices andcomplementarities between these.
As far as we know, this is the first major empirical examination of the link between innovation performance and complementary new HRM practices. Thus, Gjerding (1997),Michie and Sheehan (1999), and Mendelson and Pillai (1999), while examining the HRM-innovation performance link, do not incorporate considerations of complementarity. Lorenz(1998) analyses complementarities between the use of new HRM practices and so-callednew pay policies, but does not include a measure of performance in the analysis. AndIchniowski, Shaw and Prennushi (1997), while discussing the complementarity-performance (productive efficiency) link, do not deal with innovation performance. Asopposed to this, we link together complementarity and innovation performance.
Furthermore, in our analysis the HRM “systems” (a particular combination of HRMpractices) emerge out of an analysis (principal components analysis), while Ichniowski,Shaw and Prennushi (1997), assume their four different systems from the outset. Arguably,Ichniowski, Shaw and Prennushi (1997) are able to define fine-grained controls since theyfocus on HRM complementarities found in steel finishing lines only. However, thedrawback is that the conclusions drawn do not concern the entire economy as such. Incontrast, we test hypotheses that articulate the HRM-innovation link on a large Danish dataset, the DISKO data base, which contains cross-sectional information on the HRM practicesand innovation performances of 1,900 Danish firms in both manufacturing and non-manufacturing industries.
We contribute to several literatures. Thus, our finding that complementarity obtains in HRM practices provides further empirical support for theoretical work on complementarityin organisational economics and elsewhere. Our investigation of the links betweencomplementary HRM practices and innovation performance contributes to the strategycontent literature as well as to the innovation literature. However, we see the present paperas most directly linking up with work in evolutionary economics and innovation studies.
Much of this work has had an aggregate focus in which the internal organisation of the firmhas been left out, and where interest has primarily centred on issues such as appropriability,firm size, market structure, complementary assets, etc. as determinants of innovationperformance. The findings in this paper may be taken as an indication of the importance ofinternal factors for the understanding of innovation (while not denying the importance ofother factors).
The design of the paper is the following. We begin by reviewing recent work on complementarities in organisational economics, and argue that for a number of reasons, performance. However, exceptions may always be found, and we think it is a fair judgement that otherdeterminants of innovation performance, such as appropriability, market structure, control of complementaryasserts, etc. have played bigger roles in the literatures.
evolutionary economists and innovation should take an interest in this concept.
Complementarities allow us to better understand what is “systemic” about technologies andfirms, and how performance is influenced by this. We argue that complementarities allowus to understand the clustering of HRM practices in firms. Moreover, complementaritiesbetween HRM practices influence not only the firm’s profits, but also, we argue, itsinnovation performance (“Complementarity, New HRM Practices and InnovationPerformance: Theoretical Considerations”). We then specify a simple model that allows usto test these ideas on the data set represented by the DISKO database. We apply a probitmodel as the relevant means of estimation. Using principal-components analysis, weidentify two HRM systems that are both conducive to innovation. The first is one in whichall of nine HRM variables matter (almost) equally for the ability to innovate. The secondsystem is dominated by performance related pay and to some extent by firm-internaltraining as well. Hence, we conclude that the application of HRM practices do matter forthe likelihood of a firm being an innovator. Furthermore, since the two HRM systems werestrongly significant in explaining innovation performance, while only two individualpractices (out of nine) are found to be significant, we find support for the hypothesis of theimportance of Edgeworth complementarities between certain HRM practices within each ofthe two HRM systems (“Empirical Analysis”).
II. Complementarity, New HRM Practices and Innovation The HRM-Innovativeness Link: a Black Box?
Contributions that not only mention but actually theoretically and empirically address the link between HRM practices and innovation performance are surprisingly few innumber. To be sure, there is a large, somewhat heterogeneous, literature on themanagement of innovation and technology. However, much of this literature is largelytaken up with strategy issues connected to the exogenous dynamics of technology (e.g.,technology life cycles), large-scale organisational issues, and questions relating toappropriability (e.g., Tushman and Moore, 1988). Of course, beginning with Burns andStalker (1961) the organisational behaviour field has stressed the link between “organic”organisational structures and innovation performance. A recent stream of pertinentorganisational behaviour has been prompted by March’s (1991) distinction between“exploration” and “exploitation.” However, it is not too unfair to say that more precise theoretical identifications of the mechanisms underlying the hypothesised links between HRM practices andinnovativeness are virtually non-existing. This is true of the technology management andorganisational behaviour literatures. To offer further illustrative examples, Baron andKreps’ (1999) recent economics inspired treatise on human resource management does nottreat innovativeness as a relevant performance variable. Michie and Sheehan (1999), whileempirically finding a link between HRM practices and innovation performance, do not offera theory of this link. Virtually all of the economics literature on the firm level determinantsof innovation have dealt with issues such as the famous debates on the relation betweenfirm size and innovation performance (Acs and Audretsch, 1988; Cohen and Klepper,1996). The organisational factors which may mediate any such relations have largely been black-boxed. Finally, while the emerging evolutionary economics literature on the firm(e.g., Kogut and Zander, 1992; Dosi and Marengo, 1994; Henderson, 1994; Granstrand,Patel and Pavitt, 1997; Pavitt, 1998; Laamanen and Autio, 2000) has certainly stressedcomplementarities between diverse technologies and the learning this may stimulate, theorganisational requirements that stimulating and reaping benefits from suchcomplementarities may demand have seldom been investigated in any detail.
In sum, therefore, while many contributes have noted a link between new HRM practices and innovation performance, and while some contributions have stressed the linkbetween complementary knowledge stocks and innovation performance, no contributionsappear (as far as we know) to have put forward theoretical arguments asserting a linkbetween complementary new HRM practices and innovation performance. However, asindicated, various literatures do contain ideas that are pertinent to the understanding of thelink between HRM practices, complementarities between these and innovationperformance. We briefly discuss such ideas in the following.
Complementarities
One of the most important strides forward in the economics of organisation during the last decade is the increasing use that has been made of the notion of Edgeworthcomplementarities (Milgrom and Roberts, 1990; Milgrom, Qian and Roberts, 1991; Aokiand Dore, 1994; Holmström and Milgrom, 1994; Milgrom and Roberts, 1995; Ichniowski,Shaw and Prennushi, 1997; Holmström and Roberts, 1998; Baron and Kreps, 1999). Nodoubt, the high priests of this movement have been Paul Milgrom and John Roberts. Asthey define it, complementarity between activities obtains if “… doing more of one thingincreases the returns to doing (more of) the others” (Milgrom and Roberts, 1995: 181).
Formally, this will be seen to closely correspond to mixed-partial derivatives of a pay-offfunction with standard assumptions about smoothness of this function. However, asMilgrom and Roberts argue, drawing on the mathematical field of lattice theory, the notionof complementarity is not wedded to the conventional differentiable framework.2Mathematically, complementarity between a set of variables obtains when a functioncontaining the relevant variables as arguments is supermodular.3 There are many reasons why we think evolutionary economists and innovation scholars should take an interest in the notion of complementarities (and the associatedformalisms). On the most fundamental level, it provides an understanding of those systemicfeatures of technologies that have traditionally interested these scholars (e.g., nationalsystems of innovation, technology systems). The other side of the coin is thatcomplementarity is an important source of path-dependence: successful change has toinvolve many, perhaps all, relevant variables of a system. On the other hand, 2 In terms of the intuition of the notion of complementarity, the notion represents a strong possibleconceptualisation of notion such as “synergy,” “(organisational) fit,” and “consistency” (Porter, 1996; Baronand Kreps, 1999).
3 Given a real-valued function f on a lattice X, f is supermodular and its arguments are complements if for anyx and y in X, f (x) – f (x ∧ y) ≤ f (x ∨ y) – f (y) (Milgrom and Roberts, 1995: 183). A lattice (X, ≥) is a set (X)with a partial order (≥) with the property that for any x and y in X, there is a smallest element (x ∨ y) that islarger than x and y and largest element (x ∧ y) that is smaller than both.
complementarities are an important source of self-propelled change (cf. Milgrom, Qian andRoberts, 1991), that is, “cumulative change.”4 On the method level, it should appeal that complementarities (and the underlying mathematical lattice theory) do not involve the drastic divisibility and concavityassumptions that have often been criticised by evolutionary economists (e.g., Nelson,1980). At the level of the firm, the notion of complementarity may assist in theunderstanding of diversification patterns (Granstrand, Patel and Pavitt, 1997)  forexample, it implies that firms will find most profitable new activities (or technologies) inareas that are complementary to newly increased activities (technologies). As we shallfurther argue, the notion of complementarity is also helpful for understanding the linksbetween organisational variables  specifically, what is here called “new human resourcemanagement practices”  and innovation performance.
Complementarities and New HRM Practices
Although ideas on complementarity is applicable to virtually any social system, the paradigm case of Milgrom and Roberts’ work on the subject is the basic redefinitions ofstrategies, organisation and management in manufacturing firms that have taken during thelast decades. This is sometimes conceptualised as a transition from a mass productionsystem (or, Fordism) to a lean production system, the latter involving flexibility, trust-basedrelationships, speedy production and delivery, core competencies, etc. In Milgrom andRoberts’ interpretation the diverse characteristics of modern manufacturing arecomplementary parts in an interlocking system whose emergence has been prompted byfundamental changes in technology (IT, flexible machines) and tastes (broader productlines with more frequent product introductions).
A strongly simplified representation of Milgrom and Roberts’ (1990) reasoning is to stipulate that a firm’s profits depend on the frequency with which it product innovates, α,and on the frequency of its process improvements, β: π (α, β). Following Milgrom andRoberts, we assume that π is supermodular in α and β. The costs of undertaking productinnovations and process improvements depend on, for example, the level of training of thefirm’s workforce, the assignment of problem-solving rights to the shop floor, the use ofteams, etc. – in short, a host of variables related to new HRM practices. For example, thelower the costs of increasing the training of the workforce, the more process improvementswill result. We assume there are n HRM variables, (µ1 …, µi,… µn).
We also assume that the cost of producing a particular level of innovations, ε are E (ε, s), where s is a crucial parameter which influences at least some of the cost-influencingvariables in such a way that increasing s implies that the costs of increasing these variablesare reduced. What might this parameter be? In Milgrom and Roberts (1990; 1995), there areactually two such (shift) parameters, which they take to represent the (increased) use ofcomputer-aided design equipment and the (lower) costs of applying computer-numericcontrolled machinery. In this context it is important to note that in our empirical analysiswe take the exogenous shift for given, while focussing on the effects of such a shift.
Whatever the specific rationale we may represent the costs of undertaking productinnovations and process improvements as R (µ1 …, µi,… µn; λ; ε). λ represents otherpossible costs of innovations and process improvements. Such costs might be related to the 4 As Milgrom and Roberts (1995: 187) point out, a “… movement of a whole system of complementaryvariables, once begun, tends to continue,” thus providing an aspect of the understanding of co-evolution.
external environment of firms in terms of costs of vertical disintegration5, or to the fact thatfirms face different costs of innovating, due to differences in technological opportunitiesacross sectors. We assume that (-R) is supermodular. This means that increasing µi reducesadditional costs of undertaking product innovation or process improvements. Moreover,increasing s reduces the costs of increasing µi.
Since both π and R are supermodular, the overall objective function, Π (α, β; µ1 …, µi,… µn; λ; ε; s) = π (α, β) - R (µ1 …, µi,… µn; λ) - E (ε, s), will also be supermodular.6 This implies that, for example, a reduction in the costs of transmitting information leads to an increase in the intensity of processing, storing and
transmitting information, which in turn will give rise to more product innovation and
process improvements and a greater use of new HRM practices. In other words, ideas on
complementarity can account for what Mendelsson and Pillai (1999) call “high IQ
organisations,” that is, firms that combine a high level of innovativeness with widespread
use of IT and new HRM practices, and that the emergence of such firms have primarily
been prompted by the falling costs of processing information. However, it still remains to
be explained why new HRM practices matter to innovation performance, that is, why (α, β)
depend on (µ1 …, µi,… µn). Moreover, it also remains to be explained why
complementarity of new HRM practices matters to innovativeness.
Innovation, Complementarities and New HRM Practices
To repeat, “new HRM practices” is the overall label put on a host of contemporary changes in the organisation of the employment relation, referring to team-basedorganisation, continuous (often team-based) learning, decentralisation of decision rightsand incentives, emphasis on internal knowledge dissemination, etc. While there may bestrong productivity effects and flexibility advantages of such new HRM practices  asdocumented by Ichniowski, Shaw, and Prennushi (1997) and Mendelsson and Pillai (1999),respectively  our main emphasis is on the impact of innovation performance, in particularproduct innovation.
New HRM practices can be conducive to innovative activity for a number of reasons.
With respect to process innovations/improvements, one notable feature of many new HRMpractices is that they increase decentralisation, in the sense that problem-solving rights aredelegated to the shopfloor. Accomplished in the right way, this amounts to delegating rightsin such a way that they are co-located with the pertinent knowledge, much of which isinherently tacit. In other words, increased delegation may better allow for the discovery andutilisation of local knowledge in the organisation, particularly when there are incentives inplace that foster such discovery (Hayek, 1948; Jensen and Meckling, 1992). Indeed, muchof the ability of Japanese firms to engage in ongoing, incremental process innovation turnson a successful co-location of problem rights and localised knowledge combined withappropriate pecuniary and non-pecuniary incentives (Aoki and Dore, 1994).
5 In fact, Milgrom and Roberts (1995: 196) have experimented with extensions of their model, where theyincluded increased vertical disintegration to the pattern of changes. This analysis is not however, explicitlydocumented in the paper.
6 Formally, this also requires that the feasible values of the choice variables, (α, β; µ1 …, µi,… µn; λ; ε), lie ina sublattice in R n+2.
Relatedly, the increased use of teams that is an important component in the package of new HRM practices also means that better use can be made of local knowledge, leadingto improvements in processes and perhaps also to minor product improvements. But teamscan do something more, since they are often composed of different human resource inputs.
In other words, teams often bring together knowledge that hitherto existed separately,potentially resulting in non-trivial process improvements (when teams are on the shopfloor) or “new combinations” that lead to novel products (Schumpeter, 1912/1934)(whenteams are in product development departments). Training of the workforce may be expectedto be a force pulling in the direction of a higher rate of process improvements and possiblyalso product innovations, depending on the type, amount and quality of the relevanttraining. Generally, increased knowledge diffusion, for example, through job rotation, andincreased information dissemination, for example, through IT, may also be expected toprovide a positive contribution to the firm’s innovation performance, for obvious reasons.
Thus, there are reasons to expect that the adoption of new HRM practices leads to increased innovation performance. Arguably, the adoption of a single such practice mayprovide a contribution to innovative performance. For example, rewarding shop flooremployees for minor process improvements is likely to increase such incrementalinnovation activity, more or less regardless of the specific firm in which the reward systemis implemented. However, other practices may not be expected to have significant impacton innovation performance, if implemented in isolation. At least to the extent thatimplementing new HRM practices is associated with extra effort or with disutility ofchanging to new routines, etc., employees will have to be somehow compensated. Thus, wewould expect many new HRM practices to work well (in terms of both profits andinnovation performance) only if accompanied with new, typically more incentive-based,remuneration schemes. Evidence appears to support this (Ichniowski, Shaw and Prennushi,1997).
In general, we should on a priori grounds expect new work practices to be most conducive to innovation performance when adopted, not in isolation, but as a system ofmutually reinforcing practices. The arguments in favour of this are relativelystraightforward. For example, the benefits from giving shop floor employees moreproblem-solving rights will likely depend positively on the level of training of suchemployees. The converse is also likely to hold: employees may invest more in upgradingtheir skills if they are also given the extensive problem-solving rights (i.e., actually utilisethose skills), particularly if they are given the right (intrinsic or extrinsic) motivation.
Relatedly, rotation and job-related training may be complements in terms of their impact oninnovative activity. All such practices are likely to be complements to various incentive-based remuneration schemes (whether based on individual, team or firm performance),profit sharing arrangements, and promotion schemes (Zenger and Hesterly, 1997).
In sum, while individual new HRM practices may be expected have some positive impact on innovation performance, theory would lead us to expect that because ofcomplementarities between these practices, systems of HRM practices will be significantlymore conducive to innovation than individual practices. In the following, we empiricallyexamine these ideas.
Based on the discussion above the rate of innovation may be specified as follows: a = f (β z,β x).
Here, a is the level of innovation, β1 and β2 are parameter vectors, and z is a set of(exogenous) determinants of innovation, related to the application of human resourcemanagement practices, while x is a set of other variables explaining innovative performanceacross business firms. The variables included in the vector x, can be said to be standardvariables in explaining innovation performance (Geroski, 1990; Kleinknecht, 1996). Thismodel can be made operational in the following way: where Ai expresses the firms’ ability to innovate. If the firm in question is a non-innovator,the variable takes the value of 0, if the firm has introduced (in the period 1993-95) aproduct or service, new to the firm the value is 1, if the firm has introduced a product that isnew in a Danish context over the period, the value is 2, while the value for this variable is 3if the firm has introduced a product (or service) which is new to the world. Hence, only thefinal category qualifies for being an innovation in the strict sense of the word. Our sampleincludes 934 non-innovators, 733 firms that produced products/services that were new onlyto the firm itself, 127 firms that produced products/services that were new to the nationalmarket, while 103 firms introduced products/services that were new to the world.
As is common in studies aiming at explaining innovative performance (e.g. Geroski, 1990; Michie and Sheehan, 1999) we control for firm size (SIZE) and for sectoral affiliation(SECT). We include four size categories and nine sectors. For what concerns the sectoralclassification, we apply the Pavitt taxonomy and the four corresponding sectors formanufacturing firms. For the service firms in our sample, we construct five additionalsectors. Explanations of the sectoral classification that we apply may be found inAppendices 1 and 2 to this paper. As argued by Geroski (1990), such sectoral controls canbe interpreted to capture differences in technological opportunities facing firms located indifferent sectors.
Other control variables include whether or not the firm in question has increased its vertical interaction with other firms, being it either upstream or downstream (LINK). Thisvariable is supposed to pick up the effect of interactions with suppliers and users forinnovation performance as stressed by e.g. Lundvall (1988) and von Hippel (1988).
EXREL expresses whether the firm has increased its interaction with knowledge institutions, including technical support institutions, consultancies or with universities. Inthis context it can be noted that Brouwer and Kleinknecht (1996) found that firms whichhad consulted an innovation centre were more likely to innovate than were other firms.
Although both LINK and EXREL concern whether the firms have increased their externallinkages, we interpret these variables more broadly as measuring the strength of therespective linkages. The reason for this is that we argue that respondents who have stronglinkages with external partner are very likely to answer that they have increased interactionwith partners. Finally, we control for whether or not the firm is a subsidiary of a larger firm.
The effect of this variable is however, ambiguous. On the one hand, firms with centralisedR&D departments might not wish their subsidiaries to be innovative, as such a procedurecould be seen as hampering economies of scale in R&D. On the other hand, as argued byHarris and Trainor (1995), subsidiary firms might benefit from the larger resource base andexperience of the parent firm. Some early empirical studies (e.g. Howells, 1984) found anegative effect of this variable on innovation performance, while more recent studiesdetected a positive effect (Harris and Trainor, 1995; Love, Ashcroft and Dunlop, 1996).
i . HRMPi are our new HRM variables, that is, those variables that are key to the analysis. We include nine binary variables pertaining to new HRMpractices, which express whether firms apply (i) interdisciplinary workgroups, (ii) qualitycircles, (iii) systems for collection of employee proposals, (iv) planned job rotation, (v)delegation of responsibility, (vi) integration of functions, (vii) performance related pay,(viii) firm-internal training, and finally, whether or not the firm in questions uses (ix) firm-external training.
However, as argued earlier, work on complementarities suggest that HRM practices are more effective when they are applied in systems relative to when they are applied alone.
Hence, we will estimate models where HRM practices enter the equation to be estimated incertain configurations or systems: where the notation is the same as in Equation (3). HRMS j i . HRMSi denote HRM systems, made up by configurations of our nine HRM practices.7 Subsequently, we shall comparethe estimations made, when applying the HRMPs individually, and when they appear in aHRM system.
Concerning the signs of the parameters for each variable, we expect all sign to be positive, except for the SIZE and SECT variables. In those cases, the interpretation has to bemade relative to the other size and sector categories. For what concerns size, we expectlarger firms to be more likely to innovate, while we expect the likelihood of innovation atthe level of the sector to correspond to what is normally thought of as a high-tech/low-techtypology.
The main source of data for this paper is the DISKO database. The database is based on a questionnaire which aims at tracing the relationship between technical andorganisational innovation in a way that permits an analysis of new principles for workorganisation and their implications for the use and development of the employee’s 7 The way in which the HRM practices are transformed into “systems” will be explained in the section below.
Table 1: Descriptive statistics for a set of DISKO variables (n=1897)
Systems for collection of employee proposals At least two HRMP applied (HRMPTWO) At least three HRMPs applied (HRMPTHREE) qualifications in firms in the Danish private business sector. The survey was carried out bythe DISKO project at Aalborg University in 1996. The questionnaire was submitted to anational sample of 4,000 firms selected among manufacturing firms with at least 20 full-time employees and non-manufacturing firms with at least 10 full-time employees.8Furthermore, all Danish firms with at least 100 employees were included in the sample, i.e.
a total of 913 firms. The resulting numbers of respondents were 684 manufacturing and1,216 non-manufacturing firms, corresponding to response rates of, respectively, 52 percent and 45 per cent.9 The first descriptive analysis of the survey can be found in Gjerding(1997). The database is held by Statistics Denmark, and the data on the firms in thedatabase, can be linked to regular register data, also held by Statistics Denmark. In our casewe have obtained data on the size of the firms in the sample from regular register data.
8 In the stratification of the sample, firms with less than 10 employees were excluded from the analysis.
However in our analysis, we have a size category containing firms smaller than 10 employees. The reason forthis is that when the sample was stratified, size was measured at a given point in time. However, in this paperwe measure size as the number of full time employees over a full year.
9 The full questionnaire is available in English, as an appendix to Lund and Gjerding (1996, Appendix 1) Table 1 displays descriptive statistics for our explanatory variables. It can be seen from the Table 1 that between 39 and 84 per cent of the firms in our sample apply each oneof the nine HRMPs, described above. 39 per cent apply performance related pay, while 84per cent apply delegation of responsibility. 80 per cent of the firms apply at least one of theHRMPs, while 65 per cent apply at least three such practices. For what concerns thedistribution on sectors and across size categories, it can be seen that none of the groups areeither extremely large nor are there any extremely small groups. Since the analysis containsmany different variables, each reflecting different aspects of HRMPs, we have chosen touse principal components analysis in order to reduce the amount of variables in theregression analysis to be carried out subsequently. The principal components technique, aform of factor analysis, estimates linear combinations of the underlying variables, in thiscase the indices of various work practices, that “explain” the highest possible fraction of theremaining variance in the data set. Thus, the first principal component is estimated toexplain the highest possible fraction of the total variance, the second principal componentthe highest possible fraction of the variance not explained by the first principal component,etc. By maximising the “explained residual variance” in each round, the first m (< n)principal components will explain a relatively large proportion of the total variance.
An economic interpretation of the factor loadings is that the “typical” pattern is one in which some of the above mentioned work practices play a major role. Accordingly, weinterpret each of the sets of factor loadings as “HRM systems”. The sets of factor loadingare reported in Table 2. It can be seen from Table 2 that we include six principalcomponents in the analysis, since these six components explain 80 per cent of the totalvariance. Hence, we only miss out 20 per cent of the total variance by applying theprincipal components technique.
An example of a HRM system is FL1 from Table 2. In this case the factor loadings are all positive and have all approximately the same size (factor loadings of about 0.3-0.4),except for firm external training, which is about half the size of the other factor loadings.
Nevertheless, FL1 expresses a HRM system in which eight of our nine HRMPs are equallyimportant. Note that each individual firm which scores high on FL1 is not necessarilyapplying all HRMPs simultaneously. However, it does imply that a firm, which scoreshighly on FL1 applies several of the HRMPs. Hence this system (FL1) is one in which allpractices are applied in just about equal proportions. In the same manner FL2 is dominated Table 2: Factor loadings for six organisational variables
by firm-external training (factor loading of 0.85), but to some extent by delegation ofresponsibility (factor loading of 0.34).10 Another example of a specific system orconfiguration is FL5 which dominated by performance related pay, but also to some extentby firm-internal training.
Since our dependent variable is a discrete variable we apply a probit model as the means of estimation. Hence, the method is maximum likelihood estimation (MLE), whichprovides a means of choosing an asymptotically efficient estimator for a set of parameters(for an exposition of the properties of ML estimators, see Greene, 1997, p. 129). AlthoughMLE has been criticised for having less than optimal small sample properties (may bebiased, since the MLE of the variance in sampling from a normal distribution is biaseddownwards), we do not consider this to be a major problem, given the fact that our samplecontains about 1,900 firms.
The estimations of our model can be found in Table 3. It can be seen from the table that large firms are more likely to innovate than small firms (e.g. in model i), althoughfirms sized 51-100 employees innovates slightly more than firms larger than 100employees. Given that our dependent variable is not a measure of the frequency ofinnovation this finding is not surprising, but should be controlled for.
It can be seen from Table 3 that the likelihood of firms being innovators, given their sectoral affiliation, can be ranked as follows: (1) specialised suppliers, (2) ICT (Informationand Communication Technology) intensive services, (3) science based, (4) wholesale trade,(5) scale intensive, (6) supplier dominated, (7) scale intensive services, (8) specialisedservices, and (9) crafts. Such a ranking must be said to correspond with intuition, related towhat is high-tech or low-tech.
The results also confirm that firm’s external linkages are important to innovation, since both the parameters for vertical linkages (LINK) and for other knowledge linkages(EXREL) are significantly different from zero. It can be noted however, that upstream orupstream linkages are particularly important, given the high parameter for this variable. Thelatter finding is in line with the predictions of Lundvall (1988) and von Hippel (1988) andwith the empirical findings of Rothwell et al. (1974) and Malerba (1992). The variable forbeing a subsidiary has a positive sign, but is not significant.
Using the principal components tool, described above, we identify two HRM systems which are conducive to innovation.11 The first is FL1 from Table 2, in which all of our nineHRM variables (interdisciplinary workgroups, quality circles, systems for collection ofemployee proposals, planned job rotation, delegation of responsibility, integration offunctions, performance related pay, firm-internal training, and finally, whether or not thefirm in questions uses firm-external training) matter (almost) equally for firm’s ability toinnovate. The second system, which is found to be conducive to innovation (FL5 fromTable 2) is dominated by performance related pay and to some extent by firm-internal 10 Admittedly, it is a weakness of the principal components methodology that the size of each factor loading chosen, for one to conclude that an underlying variable is “important”, is somewhat arbitrary.
11 Other examples of principal components regression include Arvanitis and Hollenstein (1996), in which theeffects on innovation performance of various sources of innovation are examined. In the field of internationaleconomics, Dalum, Laursen and Verspagen (1999) analysed the effect of international patterns ofspecialisation on economic growth, while applying the methodology.
Table 3: Probit regression, explaining innovative performance across 1897 Danish firms
training. Hence we can –as a first step– conclude that HRMPs matter for the ability of firmsto innovate.
Concerning our HRMP complementarity hypothesis, it can be seen from Table 3 (model ii) that only performance related pay (HRMP7) and firm-internal training (HRMP8) Table 4: Correlations amongst HRM systems and the firm’s sectoral affiliation
are individually significant of the total of nine human resource management practices.
However, when all HRMPs are combined into a single variable (a “system”), this“synthetic” variable (FL1) is highly significant. We take this as evidence of theexistence of Edgeworth complementarities between the HRMPs in our analysis.
However, while all HRMPs (except for firm-external training) are complementarywith respect to innovation performance for one group of firms, for another group offirms, complementarity between firm-internal training and performance related payappear to the important factor for firms’ ability to innovate.
Another way of gauging HRMP complementarities is to look at whether it is sufficient to apply at least two (or one) HRMPs, rather than it being necessary toapply several practices together. In Table 3, model iii, we test the hypothesis ofwhether having at least two HRMP, against the alternative hypothesis of applyingthree or more HRMPs at the same time. Although having at least two practice isindividually significant (not shown for reasons of space) it is not significant, whentaken together with a variable expressing whether or not each firm apply three ormore HRM practices (HRMPTHREE). In contrast HRMPTHREE has a positive signand is highly significant. We take this as further evidence of the importance ofEdgeworth complementarities between new HRM practices with respect todetermining innovation performance.
The final part of our analysis is devoted to the assessment of whether sectoral regularities in the application of the two (successful) HRM systems can be detected.
According to Table 4, we find that of our total of nine sectors, the four manufacturingsectors correlate with the first system, while also firms located in ICT intensiveservice sectors are associated with the first system. Firms belonging to the wholesaletrade sector tend to be associated with the second system. Hence it seems fair toconclude in general, that sectoral regularities in the effect of HRMPcomplementarities on innovation performance, can be detected.
We began by observing a number of stylised facts pertaining to the ongoing changesin the nature of the employment relation  often conceptualised in the term, “newHRM practices , to the apparently systemic nature of these practices, and to their adoption by innovative firms. Building on earlier fundamental work, we argued thatthe notion of complementarities (and the associated theorising and formalisms) washelpful for allowing us to construct explanations of these stylised facts. In particular,we argued that while the adoption of individual HRM practices may be expected topositively influence innovation performance, an adoption of a package ofcomplementary HRM practices could be expected to impact on innovationperformance to a much higher degree.
In our empirical analysis of these overall ideas, we began by finding that strong linkages to users or suppliers is conducive to innovation (while controlling for sizeand sectoral affiliation). Moreover, strong linkages to knowledge institutions,including technical support institutions, consultancies or with universities, was alsofound to be conducive to innovation. With respect to the application of new HRMpractices we applied principal components analysis in order to compress theinformation from the survey and in order to identify possible patterns of HRMpractices. Using this tool we identified two HRM systems which are conducive toinnovation. The first is one in which all of our nine HRM variables matter (almost)equally for the ability to innovate. The second system, which was found to beconducive to innovation is dominated by performance related pay and to some extentby firm-internal training as well. Hence, we conclude that the application of HRMpractices do matter for the likelihood of a firm being an innovator. Furthermore, sincethe two HRM systems were strongly significant in explaining innovationperformance, while only two individual practices (out of nine) were found to besignificant, we found support for the hypothesis of the importance of Edgeworthcomplementarities between certain HRM practices within each of the two HRMsystems.
The final part of our analysis was devoted to assess whether sectoral regularities in the application of the two (successful) HRM systems could be detected. Of ourtotal of nine sectors we found that the four manufacturing sectors correlate with thefirst system, while also firms located in ICT intensive service sectors are associatedwith the first system. Firms belonging to the wholesale trade sector tend to beassociated with the second system. Theoretical analysis has focussed almostexclusively on identifying organisational practices and complementarities betweensuch practices, invariant to the type of activity in question (e.g. Milgrom and Roberts,1995). Hence, in order to inform future theoretical research in the field, furtherempirical research should be devoted to the more detailed unfolding of sectoralregularities in the effect of HRM practice complementarities on innovationperformance.
The Sectoral Classification Applied in this Paper Pavitt (1984), identifies differences in the importance of different sources ofinnovation according to which broad sector the individual firm belongs. Thetaxonomy of firms, according to principal activity, emerged out of a statisticalanalysis of more than 2000 post-war innovations in Britain and was explained by thesources of technology; the nature of users needs; and means of appropriation. Fourtypes of firms were identified accordingly, namely supplier dominated firms, scale-intensive firms, specialised suppliers and science-based firms. Supplier dominatedfirms are typically small. Most technology comes from suppliers of equipment andmaterial. Scale intensive firms are found in bulk materials and assembly. Theirinternal sources of technology are production engineering and R&D departments.
External sources of technology include mainly interactive learning with specialisedsuppliers, but also inputs from science-based firms are of some importance.
Specialised suppliers are small firms, which are producers of production equipmentand control instrumentation. Their main internal sources are primarily design anddevelopment. External sources are users (science-based and scale-intensive firms).
Science based firms are found in the chemical and electronic sectors. Their maininternal sources of technology are internal R&D and production engineering.
Important external sources of technology include universities, but also specialisedsuppliers.
Since the Pavitt taxonomy was created mainly with the manufacturing sector in mind (although our crafts sector [see below] could be included in the supplierdominated sector, if one were to follow the original Pavitt taxonomy), and since weare conducting an analysis of firms in both manufacturing as well as in services, wehave added five additional service sectors. ICT (Information and CommunicationTechnology) intensive services are firms providing business services and financialservices. Wholesale trade consists of firms selling bulk materials or machines. Scaleintensive services consists of typically large firms in the transport industries, cleaningservice as well as of supermarkets and warehouses. Specialised services is made up ofsmaller firms including miscellaneous shops, hotels and restaurants, taxi companiesetc. Crafts consists of firms in construction industries, as well as of automobile repairshops.
For a detailed assignment of all industries into our nine sectors, see Appendix 2 to The Assignment of Industries Into Nine Sectoral Categories 1 Production etc. of meat and meat products 43 Sale of motor vehicles, motorcycles etc.
44 Maintenance and repair of motor vehicles 46 Ws. of agricul. raw materials, live animals 6 Manufacture of textiles and textile products 7 Mfr. of wearing apparel; dressing etc. of fur 49 Ws. of wood and construction materials 50 Ws. of other raw mat. and semimanufactures 51 Ws. of machinery, equipment and supplies 10 Mfr. of pulp, paper and paper products 52 Commission trade and other wholesale trade 53 Re. sale of food in non-specialised stores 12 Publishing activities, excl. newspapers 54 Re. sale of food in specialised stores 14 Mfr. of refined petroleum products etc.
56 Retail sale of phar. goods, cosmetic art. etc.
57 Re. sale of clothing, footwear etc.
16 Mfr. of paints, soap, cosmetics, etc.
58 Re. sale of furniture, household appliances 60 Repair of personal and household goods 19 Mfr. of glass and ceramic goods etc.
20 Mfr. of cement, bricks, concrete ind. etc.
22 Mfr. construction materials of metal etc.
23 Mfr. of hand tools, metal packaging etc.
65 Freight transport by road and via pipelines 24 Mfr. of marine engines, compressors etc.
25 Mfr. of other general purpose machinery 26 Mfr. of agricultural and forestry machinery 68 Cargo handling, harbours etc.; travel agencies 27 Mfr. of machinery for industries etc.
28 Mfr. of domestic appliances n.e.c.
29 Mfr. of office machinery and computers 30 Mfr. of radio and communication equipment etc. SCIB 72 Activities auxiliary to financial intermediates 31 Mfr. of medical and optical instruments etc.
32 Building and repairing of ships and boats 33 Mfr. of transport equipment excl. ships, etc.
75 Renting of machinery and equipment etc.
35 Mfr. of toys, gold and silver articles etc.
79 Accounting, book-keeping and auditing activities ITIS 38 Install. of electrical wiring and fittings 80 Consulting engineers, architects etc.
SCAI = Scale intensive firms; SDOM = Supplier dominated firms; SCIB = Science based firms; SPEC= Specialised suppliers; CRAF = Crafts; WTRA = Whole sale trade; SSER = Specialised services;SCIS = Scale intensive services; ITIS = ICT intensive services.
Acs, Z.J. and D.D. Audretsch, 1988, Innovation in Large and Small Firms: An Empirical Analysis, American Economic Review 78, 678-690.
Aoki, M. and R. Dore, 1994, The Japanese Firm: The Sources of Competitive Strength (Oxford University Press, Oxford).
Arvanitis, S. and H. Hollenstein, 1996, Industrial Innovation in Switzerland: a model- based Analysis with Survey Data, in: A. Kleinknecht (Editor) Determinants ofInnovation: The Message from New Indicators (Macmillan Press Ltd,London).
Barney, J.B., 1991, Firm Resources and Sustained Competitive Advantage , Journal Barney, J.B., 1995, Looking Inside for Competitive Advantage, Academy of Baron, J.N. and D.M. Kreps, 1999, Strategic Human Resources: Frameworks for General Managers (John Wiley, New York).
Bowman, E.H. and H. Singh, 1993, Corporate Restructuring: Reconfiguring the Firm, Strategic Management Journal 14, 5-14.
Brouwer, E. and A. Kleinknecht, 1996, Determinants of Innovation: A Microeconometric Analysis of Three Alternative Innovation OutputIndicators, in: A. Kleinknecht (Editor) Determinants of Innovation: TheMessage from New Indicators (Macmillan Press Ltd, London).
Burns, T. and G.M. Stalker, 1961, The Management of Innovation (Tavistock, Cohen, W.M. and S. Klepper, 1996, A Reprise of Size and R&D, Economic Journal Dalum, B., K. Laursen and B. Verspagen, 1999, Does Specialization Matter for Growth?, Industrial and Corporate Change 8, 267-288.
Dosi, G. and L. Marengo, 1994, Some Elements of an Evolutionary Theory of Organizational Competences, in: R.W. Englander (Editor) EvolutionaryConcepts in Contemporary Economics (University of Michigan Press., AnnArbor).
Freeman, C., 1988, Japan: a New National System of Innovation?, in: G. Dosi, C.
Freeman, R. Nelson, G. Silverberg and L.L.G. Soete (Editors), TechnicalChange and Economic Theory (Pinter Publishers, London).
Geroski, P.A., 1990, Innovation, Technological Opportunity, and Market Structure, Gjerding, A.N., (Editor), 1997, Den fleksible virksomhed: Omstillingspres og fornyelse i dansk erhvervsliv (Erhvervsudviklingsrådet, Copenhagen).
Granstrand, O., P. Patel and K.L.R. Pavitt, 1997, Multi-Technology Corporations: Why They Have 'Distributed' Rather than 'Distinctive Core' Competencies,California Management Review 39, 8-25.
Greene, W.H., 1997, Econometric Analysis (Prentice-Hall, Upper Saddle River, New Harris, R.I.D. and M. Trainor, 1995, Innovations and R&D in Northern Ireland Manufacturing: A Schumpeterian Approach, Regional-Studies 29, 593-604.
Hayek, F.A.v., 1948, The Use of Knowledge in Society, in: Individualism and Economic Order (University of Chicago Press, Chicago).
Henderson, R., 1994, The Evolution of Integrative Capabilities: Innovation in Cardiovascular Drug Discovery, Industrial and Corporate Change 3, 607-630.
Holmström, B. and P. Milgrom, 1994 , The firm as an incentive system , American Holmström, B. and J. Roberts, 1998, The Boundaries of the Firm Revisited, Journal Howells, J., 1984, The Localisation of Research and Development: Some Observations and Evidence from Britain, Regional Studies 18, 13-29.
Ichniowski, C., K. Shaw and G. Prennushi, 1997, The Effects of Human Resource Management Practices on Productivity: A Study of Steel Finishing Lines,American Economic Review 87, 291-313.
Jensen, M.C. and W.H. Meckling, 1992, Specific and General Knowledge and Organizational Structure, in: L. Werin and H. Wijkander (Editors), ContractEconomics (Blackwell, Oxford).
Kleinknecht, A., (Editor), 1996, Determinants of Innovation: The Message from New Indicators (Macmillan Press Ltd, London).
Kogut, B. and U. Zander, 1992, Knowledge of the Firm, Combinative Capabilities, and the Replication of Technology, Organization Science 3, 383-397.
Laamanen, T. and E. Autio, 2000, Dynamic Complementarities and Technology Acquisition, in: N.J. Foss and P.L. Robertson (Editors), Resources,Technology, and Strategy (Routledge, London).
Lorentz, E., 1998, Organisational Innovation, Governance Structure and Innovative Capacity in British and French Industry, mimeo (University of Compiégne,Compiégne).
Love, J.H., B. Ashcroft and S. Dunlop, 1996, Corporate Structure, Ownership and the Likelihood of Innovation, Applied Economics 28, 737-746.
Lund, R. and A.N. Gjerding, 1996, The flexible company innovation, work organisation and human resource management, DRUID Working Paper no. 17(IKE Group/DRUID, Department of Business Studies, Aalborg).
Lundvall, B.-Å., 1988, Innovation as an Interactive Process - from user-producer Interaction to National Systems of Innovation, in: G. Dosi, C. Freeman, R.
Nelson, G. Silverberg and L.L.G. Soete (Editors), Technical Change andEconomic Theory (Pinter Publishers, London).
Malerba, F., 1992, Learning by Firms and Incremental Technical Change, Economic March, J.G., 1991, Exploration and Exploitation in Organizational Learning, Mendelson, H. and R.R. Pillai, 1999, Information Age Organizations, Dynamics, and Performance, Journal of Economic Behavior and Organization 38, 253-281.
Michie, J. and M. Sheehan, 1999, HRM Practices, R&D Expenditure and Innovative Investment: Evidence from the UK's 1990 Workplace Industrial RelationsSurvey, Industrial and Corporate Change 8, 211-234.
Milgrom, P., Y. Qian and J. Roberts, 1991, Complementarities, Momentum, and the Evolution of Modern Manufacturing, American Economic Review (PaP) 81,85-89.
Milgrom, P. and J. Roberts, 1990, The Economics of Modern Manufacturing Technology, Strategy and Organization, American Economic Review 80,511-528.
Milgrom, P. and J. Roberts, 1995, Complementarities and fit: Strategy, structure, and organizational change in manufacturing, Journal of Accounting andEconomics 19, 179-208.
Nelson, R., 1980, Production Sets Technological Knowledge and R and D: Fragile and Overworked Constructs for Analysis of Productivity Growth, AmericanEconomic Review 70, 62-67.
Pavitt, K.L.R., 1984, Sectoral Patterns of Technical Change: Towards a Taxonomy and a Theory, Research Policy 13, 343-373.
Pavitt, K.L.R., 1998, Technologies, Products and Organization in the Innovating Firm: What Adam Smith Tells us and Joseph Schumpeter Doesn't, Industrialand Corporate Change 7, 433-452.
Porter, M.E., 1996, What is Strategy?, Harvard Business Review 74, 61-80.
Rothwell, R., C. Freeman, P. Jervis, A. Robertson and J. Townsend, 1974, SAPPHO Updated -Project SAPPHO Phase 2., Research Policy 3, 258-291.
Schumpeter, J.A., 1912/1934, The Theory of Economic Development: An Inquiry into Profits, Capital, Credit, Interest and the Business Cycle (OxfordUniversity Press, London).
Tushman, M.L. and W.L. Moore, (Editors), 1988, Readings in the Management of Innovation (Ballinger Publishing Co., Cambridge, Massachusetts).
von Hippel, E., 1988, The Sources of Innovation (Oxford University Press, New York Zenger, T. and W.S. Hesterly, 1997, The Disaggregation of Corporations: Selective Intervention, High-Powered Incentives and Molecular Units, OrganizationScience 8, 209-222.

Source: http://mrp.economia.unife.it/paper/Laursen_Foss_April_2000.pdf