The role of strategic purchasing in dynamic capability development and deployment: A contingency perspective

2.1 Dynamic capabilities and competitive advantage

The dynamic capabilities perspective extends the essentially static view of resource-based theory by examining how resources are created and refreshed over time in the face of changing business environments (Helfat and Peteraf, 2003; Teece et al., 1997). It is not only concerned with the bundle of VRIN (valuable, rare, inimitable, and non-substitutable) resources, but importantly “the mechanisms by which firms learn and accumulate new skills and capabilities” (Teece et al., 1990, p. 11). As such, the dynamic capabilities perspective emphasizes that competitive advantage does not emerge solely from the possession of resources, but rather the way these resources are deployed and renewed over time in the face of external environmental change (Ambrosini and Bowman, 2009; Augier and Teece, 2007).

A dynamic capability is defined by Zollo and Winter as “a pattern of collective activity through which the organization systematically generates and modifies its operating routines in pursuit of improved effectiveness” (2002, p. 340). As such, dynamic capabilities are predicated on activities that are embedded within repetitious organizational processes. Such activities are also intentional as opposed to ad hoc, oriented on future as well as current needs, and exhibit high levels of path dependency (Ambrosini and Bowman, 2009; Pierce et al., 2002). Furthermore, dynamic capabilities are developed and embedded within the firm, rather than being bought from the market (Eisenhardt and Martin, 2000; Makadok, 2001). Winter (2003) proposes that a firm’s basic functional activities (i.e. the physical and human assets) that permit its existence can be termed operational capabilities or zero-level capabilities. Dynamic capabilities or first-level capabilities are those that give the organization the capacity to understand environments, recognize the value of other resources, and respond through appropriate changes to its zero-level capabilities. In line with Collis’ (1994), notion of meta-capabilities, Winter suggests that, in turn, higher order capabilities act on dynamic capabilities as firms “learn-to-learn” more effectively over time.

A key reason for the growing interest in dynamic capabilities comes from their potential impact on competitive advantage. It is widely contended that, by providing firms with the capacity/mechanisms to learn and accumulate new skills and capabilities in the face of changing environmental conditions, dynamic capabilities provide a universally positive effect on performance (Griffith and Harvey, 2006; Lee et al., 2002; Teece and Pisano, 1994; Teece et al., 1997; Stoelhorst and Van Raaij, 2004). However, whilst the dynamic capabilities perspective has become a highly influential management theory in the past decade, its practical use remains limited due to ill-defined conceptual boundaries and a tendency for researchers to tautologically equate the existence of dynamic capabilities with organizational success and vice versa post hoc (Cepeda and Vera, 2007; Schilke, 2014).

To overcome these problems, it is necessary to identify discreet operational processes in which dynamic capabilities exist (Gruber et al., 2010; Helfat and Winter, 2011). Extant literature points to a number of identifiable routines/processes that embody dynamic capabilities or act as the foundations of dynamic capabilities. These include research and development (Helfat, 1997), acquisition processes (Eisenhardt and Martin, 2000; Karim and Mitchell, 2000), search (Augier and Teece, 2007), new product and service development (Danneels, 2002; Schilke, 2014), and alliance management (Helfat et al., 2007; Schilke, 2014). Following Teece (2007), we can assign these into three key sub-groupings that recognize the capacity to – sense and shape opportunities, seize opportunities, and manage threats and transform the business (see Table I).

Within OM more specifically, there is a small body of work that empirically examines the development and deployment of dynamic capabilities. Foundations of the “seizing” aspect of dynamic capabilities include supply chain agility (Blome et al., 2013; Chiang et al., 2012), logistics fulfilment capabilities (Vaidyanathan and Devaraj, 2008), and supply chain knowledge (Wowak et al., 2013). Foundations of the “transforming” aspect of dynamic capabilities include supply management alignment (Handfield et al., 2015), sustainable global supplier management (Reuter et al., 2010) and supplier sustainability risk management (Foerstl et al., 2010). Considering the “sensing” aspect of dynamic capabilities, Chen et al. (2004) argue that sensing capabilities are predicated on accessing critical upstream resources through open communication with selected suppliers. Revilla and Knoppen (2015) propose two learning processes in a strategic purchasing context – joint sense-making (“sensing”) and joint decision-making (“seizing”) – that constitute a dynamic capability. Joint sense-making refers to the scanning of the environment, the noticing of relevant changes and the development of meaning for the observed changes. Joint problem solving on the other hand refers to the use and implementation of the newly assembled knowledge, in other words the conversion of information into action that eventually propels superior performance.

In this study, we are particularly interested in knowledge scanning, a routine that supports the identification and capture of knowledge and technology within a firm and from its supply base (Tu et al., 2006, p. 695). When exercised strategically, the purchasing function acts as boundary spanner of the firm (Zhang et al., 2011), channeling knowledge flows outside in and inside out (Hult et al., 2000). Consequently, strategic purchasing can leverage its internal involvement in strategic activities and directly influence the resource mix of the firm (Barney, 2012; Chen et al., 2004; Priem and Swink, 2012).

2.2 Enablers of dynamic capabilities

The growing appreciation of the value of dynamic capabilities has led a small number of academics to explore the “positions” that act as internal and external enablers or antecedents to dynamic capability development (Ambrosini and Bowman, 2009). Internal position concerns the firm’s stock of reputational, technological, financial, complementary, and structural assets, whilst external position concerns the institutional environment and market in which the firm operates (Teece et al., 1997). We explore external positions when considering the contingent role of industry sector on dynamic capability development and performance (see section 2.3). In this section, therefore, we focus on the internal positions that may enable dynamic capability development.

Within the literature, there are a number of internal positions (enablers) that arguably impact the development of dynamic capabilities. Blyler and Coff (2003) propose that social capital is an important enabler in “facilitating the acquisition, integration, and release of resources” (p. 678). Social capital, in the form of social ties and internal reputation, enables managers to share information and innovation across functional boundaries (i.e. to be involved in strategic activities), and in doing so contribute more fully to dynamic capability development and deployment. Closely related to this is the argument that leadership from senior managers is critical to dynamic capability development, because it is typically the senior managers within a firm who formulate the organizational routines upon which dynamic capabilities are based (Eisenhardt and Martin, 2000; Helfat et al., 2007; Rindova and Kotha, 2001; Teece, 2007; Tripsas and Gavetti, 2000; Zahra et al., 2006).

The OM studies referenced in Table I also provide enablers to dynamic capabilities. First, the perception of stimuli or pressure from stakeholders enables companies to develop dynamic capabilities. For example, a non-government organization may exert pressure for firms to increase the sustainability of their supply chain. Consequently, firms may develop a dynamic capability relating to sustainable global supplier management (Reuter et al., 2010). Second, alignment of internal stakeholders enables companies to develop the dynamic capability related to supply management alignment (Handfield et al., 2015). Third, strategic purchasing is an enabler of the dynamic capability of supply chain agility (Chiang et al., 2012). In the context of strategic purchasing, and building upon psychology literature which has a long tradition in exploring attitudes at work as they drive behavior (see e.g. Ajzen and Fishbein, 1977; Glasman and Albarracín, 2006; Lindenberg and Steg, 2007), we argue that the increased recognition of the purchasing function by senior management results in greater involvement in company-wide processes, and this expanded role of the purchasing function positively impacts on the dynamic capability creation process.

2.3 The contingent role of industry on dynamic capabilities

Until recently, most academics have assumed a universally positive relationship between dynamic capabilities and competitive advantage (Schilke, 2014). Likewise, the few studies exploring enablers of dynamic capabilities have typically treated antecedents as equally applicable under different contextual conditions. However, a number of researchers now advocate a contingency perspective, where performance outcomes not only rely on the existence of dynamic capabilities, but also on the environmental conditions in which such capabilities are deployed (Ambrosini and Bowman, 2009; Aragon-Correa and Sharma, 2003; Sirmon and Hitt, 2009; Schilke, 2014). In the same vein, we anticipate that these external “positions” are also likely to be applicable as contingency factors in the creation of dynamic capabilities.

Contingency theory holds that organizations must adapt their structures depending on contextual conditions (Donaldson, 2001) and as such the value of different physical and non-physical assets is partly determined by exogenous contextual (or contingency) variables, generally beyond the control of organizations or managers (Sousa and Voss, 2008). These include, but are not limited to, national context and culture, institutional conditions, such as the level of environmental dynamism and uncertainty (e.g. Koufteros et al., 2005), and firm characteristics, such as industry, size, ownership, and structure (e.g. Koufteros et al., 2007).

Within dynamic capabilities research, institutional conditions and, to a much lesser extent, firm characteristics have been explored as possible contingency variables influencing the development and value of dynamic capabilities. The majority of extant literature suggests that dynamic capabilities are most valuable when deployed in dynamic environmental contexts (i.e. high uncertainty, frequent change, high complexity) and are less valuable in environments characterized by low levels of dynamism. This logic is predicated on the fact that dynamic capabilities are expensive to develop and therefore only worth investing in when environmental conditions necessitate their existence and regular use – in highly dynamic environments (Winter, 2003; Zollo and Winter, 2002).

Building on the idea of different environmental contexts is the potential contingent role of industry in influencing both the development of dynamic capabilities and its value to organizations (Aragon-Correa and Sharma, 2003; Winter, 2003), an area that has, to date, remained largely unexplored (Schilke, 2014). For example, from a conceptual standpoint, Winter (2003) argues that the decision to develop and deploy dynamic capabilities will be contingent on the pace of change within a particular industry. Aragon-Correa and Sharma (2003) argue that the business environment, including the type of industry and its characteristics, and the way that managers interpret such business environments, may play a contingent role in the development of various dynamic capabilities. Miller and Shamsie (1996) demonstrate that knowledge-based capabilities are most valuable in uncertain environments, whereas property-based capabilities are more valuable in stable environments. As such, the characteristics of the industry are likely to be crucial to the value of different kinds of capabilities. Exploring this issue further, Sirmon and Hitt (2009) propose that the lack of manufacturing in service-based firms elevates the importance of human dynamic capabilities (i.e. the skills, education, experience, and knowledge of employees and managers), especially in light of the high environmental dynamism faced by many such organizations. This view is supported by the work of Wowak et al. (2013), which finds significantly stronger relationships between supply chain knowledge and performance for service-based firms as opposed to those operating in manufacturing contexts in their meta-analysis of 35 supply chain management studies. Sampson and Spring (2012) elaborate the bidirectional processes between buyers and suppliers of services, and consequently an increased role for purchasing. More precisely, buyers can play different roles, ranging from providing input materials to competing with suppliers. Given this variety in roles, there are more opportunities for innovation, and consequently dynamic capabilities may gain importance.

2.4 Conceptual model and hypotheses

Figure 1 illustrates our conceptual model linking purchasing recognition, purchasing involvement, knowledge scanning, and two forms of performance – cost and innovation.

Our model is based on the premise that in order to generate improvements in cost and innovation performance, purchasing must effectively develop and deploy knowledge scanning, a dynamic capability that enables the identification and capture of internal and external knowledge and technology. Purchasing accomplishes this when it is recognized as a strategic function by senior management as well as those within other functions (purchasing recognition). This attitude elicits corresponding behavior (Ajzen and Fishbein, 1977; Glasman and Albarracín, 2006), in the form of the purchasing function’s actual involvement in key strategic activities such as new product and service development, organization-wide process improvement, and strategic planning (purchasing involvement). Further, from a contingency perspective, we argue that the positive relationship between the elements of our model will be stronger for service-based as opposed to manufacturing-based firms. We now examine each element of our model in turn.

Purchasing is increasingly seen as a strategic function within organizations. Evidence of this strategic role can be seen in the increasing value placed on purchasing’s views by senior management and recognition of purchasing’s equality with other organizational functions (Carr and Smeltzer, 1997; Cousins et al., 2006). Evidence also stems from the tendency to centralize the function in order to increase added value and minimize total cost of purchasing (Pagell, 2004). Centralization is possible for more generic activities, while context-dependent purchasing decisions necessarily have to remain at a local level. Inherent to centralization is top management’s attitude to viewing purchasing as a strategic role in their company (Barney, 2012; Foerstl et al., 2013). In other words, first there is an organizational restructuring reflecting a novel attitude on the status of the purchasing function, and after that, there is the actual change in daily behavior; i.e. how the purchasing function adds value in on-going processes. Finally, evidence can be found in companies devoted to the development of customized solutions, where project managers and engineers typically lead the contact with suppliers, and purchasing is involved in a later stage to execute prior agreements. These companies have seen a restructuring where purchasing has become the key boundary spanning function, responsible for inward and outward channeling of initiatives (Jia et al., 2014).

Purchasing recognition can be viewed as a reputational asset (i.e. a form of social capital) that makes subsequent involvement in the strategic dialogue of the firm and in organization-wide processes related to new product and service design, process improvement, and strategic planning more likely. By introducing this sequential relationship between the two dimensions of strategic purchasing, we thus differentiate our research from other studies that treat both dimensions as reflective. Therefore, our first hypothesis is:

We suggest that behavior in the form of the purchasing function’s involvement in intra-organizational and cross-functional activities acts as an internal factor or “position” (see Ambrosini and Bowman, 2009) to enable the development of dynamic capabilities. In the context of strategic purchasing, a key dynamic capability is the role of purchasing in acquiring and assimilating knowledge and technology resources from within the firm and across its supply network, and applying it to commercial ends (Barney, 2012; Chen et al., 2004; Hult et al., 2000; Priem and Swink, 2012). This can be termed knowledge scanning (Tu et al., 2006) and refers to mechanisms such as market tracking, benchmarking, technology assessments, and the capacity to research the supply base (Giunipero et al., 2006). Knowledge scanning builds upon actual purchasing behavior, rather than on the attitudinal dimension of strategic purchasing, in line with a behavioral view on learning (Levitt and March, 1988). Thus, we propose the following hypothesis:

In line with the general arguments set out in extant dynamic capabilities research, we argue that knowledge scanning permits the purchasing function to leverage its internal involvement in order to drive key performance outcomes of cost (Tu et al., 2006) and innovation performance (Malhotra et al., 2005; Stock et al., 2002). We look to gain a more comprehensive evaluation of performance outcomes (Volberda et al., 2010) by expanding on the predominant financial focus of previous dynamic capabilities studies. The combined focus on cost and innovation performance is increasingly common in studies on dynamic capabilities based on knowledge resources (e.g. Revilla and Villena, 2012; Sáenz et al., 2014). In other words, whilst dynamic capabilities may lead to operational improvements by reducing costs or using resources more efficiently (Tu et al., 2006; Malhotra et al., 2005), they may also impact on innovation performance in delivering greater levels of innovation and in bringing such products/services to market more quickly (Handfield et al., 1999; Visser et al., 2010). It is important to note that there are OM scholars who have an alternative perspective on the strategic contribution of purchasing. Ramsay (2001), for example, argues from a resource-based perspective that the purchasing function is inherently an operational rather than strategic function and is therefore unlikely to contribute to competitive advantage. However, the author does acknowledge that the identification and development of unknown suppliers (something akin to a dynamic capability) has the potential to generate completive advantage. OM scholars have responded to these challenges with further evidence from the field of strategic management or resource-based view (RBV) more specifically to support the potential role of purchasing as a strategic function, alongside empirical examples demonstrating the contribution of purchasing to competitive advantage (For example, Mol, 2003; Nair et al., 2015; Pressey et al., 2009; Steinle and Schiele, 2008). A balance between these views is found in work exploring the future of purchasing and supply (Carter et al., 2000) which predicts a shift towards smaller, more strategic purchasing functions, with much more tactical activities being outsourced.

Another important consideration is that it is often difficult to unambiguously demonstrate positive linkages between dynamic capabilities and firm-level performance (Helfat et al., 2007). This is because first, internally, dynamic capabilities do not operate in isolation but combine to restore and refresh a firm’s resource mix, and second, overall performance is influenced by external dynamics (e.g. competition, market growth, uncertainty, etc.) in addition to dynamic capabilities. Therefore, we narrow down performance to the purchase category level to enable us to more accurately assess the direct impact of knowledge scanning – as a purchasing capability - on purchasing performance. Thus, we envisage the following hypothesis:

Figure 1 shows that we expect the disturbance terms of the two latent performance variables to correlate. This is because we acknowledge that performance may be affected by many variables, not only those included in our research model. The correlation is in line with studies that model a general “performance” construct through a broad range of reflective indicators. Based on our earlier exposition of the contingency perspective as it applies to dynamic capabilities, we expect to see service-based firms exhibiting stronger relationships between purchasing recognition and involvement; between purchasing involvement and dynamic capability development; and between dynamic capabilities and the two facets of performance investigated here. Thus, we propose the following hypothesis:

3.1 Survey instrument

The survey instrument was developed iteratively through a number of phases. First, extensive literature review was the basis for development of all constructs and their operationalization was based on previously validated survey instruments. The survey had two parts: one part focusing on the organization level and the second part focusing on the purchase category level. In the second part, respondents had to select a specific purchase category and describe this, before proceeding to answer questions related to the category level. Since the survey was to be administered in ten countries (see next section), we addressed construct equivalence across those countries by actively involving all members of the research team, representing their respective countries, in the development and final selection of constructs and their operationalization.

Second, we used the translation, review, adjudication, pretesting, and documentation (TRAPD) procedure as the basis for translation of the source English instrument to the local languages. The TRAPD procedure is increasingly seen as both theoretically and practically superior to the traditional translation/back-translation method (Douglas and Craig, 2006; Harkness et al., 2003). Special attention was paid to consistent translation of key grammatical characteristics of the response scale (such as: not using a neutral middle category point in scales where possible; assuring symmetry of scales with a neutral middle category; providing brief labels for all scale categories and not just for the anchors; using balanced requests; and assuring congruence between the unipolar and bipolar nature of theoretical concepts and scales), as they impact significantly on the quality of the instrument and may be subject to inconsistent translation (Saris and Gallhofer, 2014).

Third, the translated survey instruments were locally tested with practitioners in each country in order to capture potential differences in local interpretations of the construct. Modifications were centrally coordinated in order to refine the same source document so as to allow clear translations in all local contexts.

3.2 Data collection

The data used to examine our hypotheses were collected in the first round of the International Purchasing Survey, a collaborative project involving researchers from European and North American institutions in ten countries (Canada, Finland, France, Germany, Italy, the Netherlands, Spain, Sweden, UK, and the USA of America). Sampling criteria were pre-agreed and unified among the participating researchers and as a result 681 purchasing directors or equivalent positions responded to the survey, which was administered online. An internet survey offers higher levels of accuracy and reduces missing values due to either the respondent or data entry mistakes (Boyer et al., 2002). One academic partner was in charge of managing the online tool for data collection in all countries and the related data management.

In this study, we are concerned with dynamic capability development and its effect on performance, and the contingent effect of industry on these relationships. Such impact is likely to be less applicable for low-spend and indirect purchase categories. We, therefore, focus on companies that report on purchase categories that represent at least 20 percent of total purchasing spend and are classified as direct spend. Of the net sample of 681 companies and their respective selected purchase category, a total of 322 (n=108 for service-based firms; n=201 for manufacturing-based firms; n=8 agriculture, n=4 other, n=1 missing) fit these criteria. From these, we use the 309 firms that are either manufacturing-based or service-based as the basis for our statistical analysis. Key characteristics of both sub-samples are provided in Table II.

3.3 Measures

All constructs used in our hypotheses were taken from literature, and their operationalization was based on validated survey instruments (see Table III). Purchasing recognition, purchasing involvement and knowledge scanning operate at the level of the purchasing organization. Both performance constructs operate at the level of the purchasing category. Such an embedded design reduces the risk of spurious effects due to omitted but relevant levels of analysis (Rothaermel and Hess, 2007). Purchasing recognition is defined as the recognition from senior management that the purchasing function is of strategic importance to the organization. Items tapping purchasing recognition, an attitude of top management, are reflective and include the extent to which top management is supportive of efforts to improve the purchasing department; purchasing’s views are considered important by most top managers; and purchasing is recognized as an equal partner with other functions of the top management team (Carr and Smeltzer, 1997; Cousins et al., 2006). Purchasing involvement is defined as the involvement of purchasing in key strategic activities. Items tapping purchasing involvement, actual behavior by the purchasing function, are reflective and include the extent to which purchasing: recommends and initiates changes in products/services based on supply market analysis; participates actively in new product/service design; and participates actively in organization-wide process improvement (Carr and Smeltzer, 1997; Cousins et al., 2006). Response scales for both constructs are six-point Likert scales (totally disagree, considerably disagree, slightly disagree, slightly agree, considerably agree, and totally agree).

Knowledge scanning is defined as a mechanism that enables firms to identify and capture relevant external and internal knowledge and technology. Items tapping knowledge scanning are reflective and include the extent to which firms: seek to learn from tracking new market trends in the supply industry; seek to learn from benchmarking best practices in purchasing; seek to learn from trying out new technologies; and seek to learn from their suppliers. These items are drawn from the knowledge scanning dimension of the absorptive capacity construct as developed by Tu et al. (2006). The response scale for knowledge scanning is a six-point Likert scale (not at all, a very small extent, a small extent, a large extent, a very large extent, completely).

The variables used to measure performance were cost performance and innovation performance. For both variables, we used two items that converged well; a scale with few items performing well is preferable to a scale with more items that are non-convergent. In other words, it is common for later studies that build upon the relatively novel confirmatory factor analysis (CFA) to conclude that earlier developed scales, which were validated with other methods, must be narrowed down in order to converge and demonstrate predictive power (Saris et al., 2013). Cost performance highlights one of the most distinct dimensions of operational capabilities (Boyer and Lewis, 2002), and items tapping costs include: purchasing price; and costs of the purchasing process. Innovation performance, on the other hand, focuses on novel strategies to compete in the market, and items tapping this construct included: the supplier time-to-market for new or improved product/services; and the level of innovation in products/services from suppliers. In this study, we did not have access to objective performance data and therefore relied on senior manager´s perceptions of performance, in line with Chen et al. (2004) and numerous other OM studies. Response scales for both performance measures are seven-point Likert scales that request the respondent to compare a performance facet with management targets. Response categories are all labeled and range from “much worse than target,” to “on target,” to “much better than target.”

3.4 Analytical procedures

We employed a two-step process of analysis to evaluate first the measurement model and then the structural model (Anderson and Gerbing, 1988). The measurement model test built upon CFA, while the structural model test built upon path analysis. We performed both tests with maximum likelihood estimation within Lisrel 8.72 for structural equation modeling (SEM). Common practice for both kinds of tests is to base the accept/reject decision on a range of test statistics (e.g. AGFI, GFI, SRMR, NFI, CFI, RMSEA) (Hu and Bentler, 1998), all of which have the shortcoming of being highly dependent upon the power of the test (Saris et al., 2009). In other words, the standard test and fit measures can only detect misspecifications for which the test is sensitive (high power). As a consequence, rejection of the model may be due to very small misspecifications for which the test is sensitive. Acceptance of the model, on the other hand, does not necessarily mean that the model is correct but may indicate a lack of power of the test. Therefore, we supplemented the standard test statistics with an alternative procedure to iterate between evaluation of misspecifications (i.e. relevant parameters that have been omitted from the model, or modeled parameters that are not present in the data) and subsequent partial modifications of the model in line with the procedure of Saris et al. (2009). The analysis of misspecifications is supported by modification indexes (MI) and expected parameter changes provided by Lisrel. Cut-off sizes to consider misspecification are 0.40 for factor loadings, 0.10 for causal effects, 0.10 for correlations, and 0.05 for mean structures (Van der Veld et al., 2009).

Rules of thumb for cut-off values are arbitrary and context dependent (Lance et al., 2006). We aimed to overcome this on the one hand by complementing the standard test and fit measures by the misspecification analysis, and, on the other hand, by testing alternative path models related to H1. In other words, we gathered multiples pieces of information to judge model fit (Lance et al., 2006). In order to test the alternative path models, we had to introduce an instrumental variable (Hamilton and Nickerson, 2003), as elaborated in section 4.2.1.

Given that H5a-H5d aim to contrast findings for manufacturing vs services sub-samples, we performed equivalence tests while evaluating the measurement model and the structural model. Measurement equivalence implies that the relations between the observed variables and the latent variables are identical across sub-samples (Drasgow, 1984). Without the assessment of measurement equivalence, it is hard to know if findings reflect “true” similarities and differences between selected groups rather than the spurious effect of cognitive or socio-cultural differences in response to a survey (Mullen, 1995). Awareness and testing of equivalence, however, remains limited within the empirical OM literature working with latent variables (Knoppen et al., 2015). Testing for measurement model invariance involves testing the equivalence of measurement models across sub-samples. We followed the bottom-to-top test strategy, which means that we started with the weakest constraints and proceeded to the most severe. With this strategy, configural equivalence may be established when the same measurement model fits the data in the different groups; in other words, when items load significantly on the same factors across groups and the correlations between the latent constructs are significantly less than one, guaranteeing discriminant validity. Metric equivalence may be subsequently established when the factor loadings (i.e. the slopes of the measurement model) across the different groups are the same. It implies that relationships between the evaluated construct and other constructs can be compared across groups (Steenkamp and Baumgartner, 1998).

For the structural model invariance test, the condition of metric measurement invariance should be fulfilled for all five constructs of our model. The structural invariance test involves constraining the model of the second sample to use the same estimated path coefficients as the first sample (or vice versa). When the model fit decreases, the structural paths are not equivalent, whereas when the fit remains the same, the structural paths are equivalent. For both measurement model and structural model equivalence tests, we used multi-group confirmatory factor analysis and the approach to SEM outlined above. In other words, we combined the comparison of test statistics of subsequent test steps with an analysis of misspecifications.

4.1 Results of the measurement model test

To answer H1-H4, we performed initial analyses on dimensionality for the pooled database. For identification purposes, the factor structures of the five constructs of our research model were jointly analyzed, as correlated first-order constructs. Fit statistics are satisfactory (χ²=85.60; df=67; χ² /df=1.28; RMSEA=0.027; NFI=0.96; CFI=0.99). More importantly, the analysis of misspecifications does not point to misspecifications. Table III shows the standardized loadings and random errors. Additionally, convergent validity was assessed through the average variance extracted (AVE) (Anderson and Gerbing, 1988), which ranged from 47 to 68 percent (Table III). The complementary misspecifications analysis that we have performed did not point to alternative better fitting models.

We assessed discriminant validity by comparing χ² for a constrained CFA (where the interfactor correlation was set to 1, indicating they are the same construct) with χ² for an unconstrained model (where the interfactor correlation was free). All χ² differences are significant, providing support for discriminant validity (Anderson and Gerbing, 1988). After establishing validity, composite reliability (CR) was assessed in line with Fornell and Larcker (1981). Composite reliabilities range between 64 and 86 percent, as Table III shows. CFA results, which are satisfactory in our case, are an important complement to indicate acceptable reliability levels for the constructs (Bollen, 1989).

Given that we relied on a single-respondent design, we controlled for common method bias in two ways: a priori through the design of the study and a posteriori through statistical control (Podsakoff et al., 2003). Regarding the survey design, the research project was labeled as a broad overview of purchasing management and purchasing practices adoption. Therefore, respondents’ attention was not drawn to the relationships being targeted in this study. Questions including items and constructs related to each other in the general model were also separated in the questionnaire in order to prevent respondents from developing their own theories about possible cause-effect relationships. Furthermore, the questionnaire was carefully created and pretested and respondents were assured of strict confidentiality. Finally, we used different scales and formats for the independent and the dependent measures. As a matter of statistical control, we compared the baseline model with five correlated first-order factors, as reported before, with a model where all items were also allowed to load on a sixth latent common method variance factor. The latter loadings were all insignificant, while the loadings on the theoretical constructs remained significant. Moreover, we applied Harman’s single factor procedure; i.e. all items from the five main constructs were included in an exploratory factor analysis to examine the un-rotated factors solution and determine the number of factors that are necessary to account for the variance in the variables. In this analysis, no single factor emerged and no general factor accounted for the majority of the co-variance among the measures. Given these findings, common method bias is less likely to be present our analysis. Finally, for the pooled database, the correlations between the first-order constructs are shown in Table IV.

Subsequently, we proceeded to contrast the measurement models of the two sub-samples defined by industry. The outcomes of the configural equivalence tests are satisfactory (χ²=177.12; df=134; χ²/df=1.32; RMSEA=0.040; NFI=0.93; CFI=0.98). More importantly, the analysis does not point to misspecifications. The subsequent metric equivalence test also reports satisfactory outcomes (χ²=187.94; DF=143; χ²/DF=1.31; RMSEA=0.040; NFI=0.93; CFI=0.98) and again no misspecifications are detected with the misspecifications test. Consequently, it is appropriate to pool the data for estimation of an overall structural model as well as to compare structural models of service companies vs manufacturing companies in the next section.

4.2 Results of the path model test

In this section we first develop a path model test for the pooled database, followed by a path model test with sub-groups defined by sector.

6.1 Managerial implications

Our study offers a number of important insights for practitioners. The purchasing function has matured over the past decade to become increasingly involved in transversal processes and the strategic direction of the firm. This paper has examined how companies may develop a purchasing function’s knowledge capabilities related to supply market scanning in order to further leverage its strategic potential. We have empirically established the impact of purchasing recognition and subsequent involvement on transversal value chain processes, the development of knowledge capabilities, as well as the effect of such capabilities on performance. Such findings can be used by practitioners to argue the case for purchasing’s strategic value in broad terms, but more specifically to emphasize the value of dynamic capabilities, such as knowledge scanning, in the process of developing (and hiring into) strategic purchasing functions. However, it is also important to note that the link between dynamic capabilities and competitive advantage may not be deterministic (Helfat et al., 2007). Whilst the resource base of a firm is altered as a result of dynamic capabilities, the resulting resource stock may lead to a number of different outcomes relative to competitors, including sustainable competitive advantage, temporary competitive advantage, competitive parity, or in some cases even failure (Helfat et al., 2007; Rindova and Kotha, 2001).

Our empirical results highlight that everything starts with top management´s recognition of the strategic role of purchasing. In other words, the right attitude has to precede purchasing’s actual involvement in product/service design or other organization-wide improvement processes. It is also evident that this “attitude-into-action” allows purchasing to develop critical knowledge capabilities, which in turn improve performance. Critically, it is not only in traditional cost performance that results are seen, but also innovation performance, related to the time-to-market of new or improved products/services. For practitioners, our decision to include both cost and innovation performance as key outcome variables is useful because it allows a case for dynamic capability development and deployment to be made with multiple performance objectives in mind.

Our study points to a particular opportunity for strategic purchasing within service firms, where many functions remain relatively operational. When purchasing functions within service firms are recognized for their strategic importance, their involvement in strategic activities and resulting impact of dynamic capability development appears to be especially high. Finally, the identification of a reinforcing effect between cost performance and purchasing´s actual involvement in company-wide processes within manufacturing-based companies only at present, suggests an opportunity for service-based companies to also improve learning effects. In the words of Vargo and Lusch (2004) “Outcomes are not something to be maximized but something to learn from” (p. 6). For practitioners, this finding reiterates the need to place greater emphasis on learning from past projects with suppliers in order to support further dynamic capability development and ultimately performance. Figure 5 provides a practitioner summary model identifying the differing impacts of purchasing-derived dynamic capabilities for both manufacturing and service organizations.

6.2 Limitations and future research opportunities

Whilst making a number of valuable contributions to both academics and practitioners, the limitations of our study give rise to a number of potential research avenues. First, in line with the majority of OM survey-based work, our study adopted single-respondent approach to data collection. Whilst efforts were made in the design of the survey to minimize potential single-respondent bias, we acknowledge that future best practice in survey research will include broader perspectives on the key outcome variables. We would therefore welcome research that develops performance measures to more adequately capture the outcome of the knowledge scanning dynamic capability and thus increase the variance fit of our model. Psychometric properties of perceptual performance measures, such as the ones adapted from the literature for our study, may be improved by having multiple-informant studies (Ketokivi and Schroeder, 2004). Alternatively, secondary data on performance offer a valuable complementary perspective to the perceptual view in our survey data.

Second, we believe that further research is necessary to shed more light on the return effect that emerged from our data, and the potential role of environmental dynamism. Literature on learning and dynamic capabilities can help to shape research questions and hypotheses that involve reciprocal effects. In other words, we recommend a research design that includes the reflexive experience of managers, where performance outcomes are potentially back translated to purchasing involvement in company-wide processes. Such a reflexive process is not necessarily conscious, explicit, or purposive (Hargadon and Fanelli, 200). Consequently, we would suggest in-depth longitudinal case studies or ethnographic studies may be the most appropriate methodologies. A focus on concrete organizational projects, and their post-mortems, may facilitate data gathering from multiple angles on one and the same phenomenon, in this case the potential feedback loop from performance to involvement. While not a tradition in empirical OM research, other fields have already embraced the related nonrecursive analytical models (e.g. Finkel, 1985, in political science and Neyer et al., 2014, in personality research).

Finally, further research is also merited to expand our understanding of the difference between service-based and manufacturing-based organizations in how they reflect upon past experience. Do different degrees of dynamism impact on the speed or duration of reflection and therefore capabilities development, for example? Service firms still have a longer way to go in this regard, but their horizon looks bright.