Combined effects of procurement and collaboration on innovation in public-private-partnerships: a qualitative comparative analysis of 24 infrastructure projects

ABSTRACT

Different public sector innovation literatures tend to focus on either contractual stimuli or collaborative interactions as sources of innovation. This article argues for a combined approach that integrates these literatures. Using an fsQCA design that exploits rich survey and interview data on 24 PPPs in Belgium and the Netherlands, we confirm the combined effect of contractual stimuli and collaboration. Since PPPs are long-term, contractual collaborations, contractual stimuli and collaborative activities (information sharing, network management) complement and even reinforce each other to create novel ideas. Managers in PPPs that only consider contractual stimuli may therefore fail to innovate.

KEYWORDS:

• Public-private partnerships
• PPP
• innovation
• collaborative innovation
• fsQCA

Introduction

The emphasis in public service literature has recently been directed to the search for innovation through collaborative partnerships (Osborne and Brown 2013). In order to meet the growing demands of citizens, address environmental issues such as global warming, and stimulate economic activity, societies across the world are in need of innovative and reliable public infrastructures and services. Due to more traffic, a rising demand for better quality of services and the need for more services, the pressure on existing public infrastructure and services increases and public authorities realize that their own knowledge and resources are insufficient to overcome these issues. Long-term collaboration of governments with the private sector is often beneficial to address these issues, as private partners are able to introduce new knowledge and resources (e.g. finances) into infrastructure projects, hence increasing the feasibility of ambitious infrastructure projects and facilitating the creation of innovative services. Public-private partnerships (PPP) are ubiquitous examples of such collaborations, as they are long-term collaborations between public and private actors that enable the joined development of services and the sharing of risks, costs and resources between the public and private partners (Van Ham and Koppenjan 2001, 598).

PPPs have attracted the attention of innovation scholars as several characteristics of PPPs have been suggested to stimulate innovation: the transfer of risks, the contractual integration of project phases (design, build, maintain/operate), the long-term commitment of the private companies, contractual incentives, design freedom, and the focus on output specifications (Grimsey and Lewis 2004; Leiringer 2006; Rangel and Galende 2010). Yet, research on whether and how PPPs lead to innovation is scarce (Himmel and Siemiatycki 2017), with some studies pointing at positive relations between PPPs and innovation (e.g. Himmel and Siemiatycki 2017) but others also observing negative relations (e.g. Barlow and Martina 2008). More research is, therefore, needed to shed light on the conditions that cause innovation creation by PPPs. This article uses insights from ‘public procurement for innovation’ literature (e.g. Edquist, Vonortas, and Zabala-Iturriagagoitia. 2015) and ‘collaborative innovation’ literature (e.g. Torfing 2019) to contribute to the discussion on how PPPs create innovations. We propose the following research question:

How do conditions related to ‘procurement for innovation’ logics and conditions related to ‘collaborative innovation’ logics stimulate the creation of innovation in PPPs and what is their combined effect on this innovation?

PPPs are ideally suited to study innovation in that they entail both procurement logics and collaboration logics, which both have been shown to foster innovation. Innovation in PPPs is procured as the deliverables of the PPP are contractually defined, but it is also created through long-term collaboration between the public procurer and private contractor. Although a public procurer might stipulate demanded, innovative results in a contract (procurement for innovation), achieving these results is only possible through long-term collaboration (e.g. design, maintenance, exploitation), causing synergies and learning processes, which might in turn generate innovative ideas (collaborative innovation, Torfing 2019). These collaborative interactions are absent in other kinds of arrangements. For instance, pre-commercial procurement (PCP) (Edquist and Zabala-Iturriagagoitia 2014) lacks a long-term commitment between the partners, which might prevent collaborative interactions between the partners that generate new ideas.

These two logics refer to two streams of literature that are linked to innovation, namely the ‘public procurement for innovation’ literature (e.g. Edquist, Vonortas, and Zabala-Iturriagagoitia 2015) and the ‘collaborative innovation’ literature (e.g. Torfing 2019). First, from the perspective of procurement for innovation literature, PPPs are highly structured, long-term contractual arrangements that provide a service or good to a procuring government (Hodge and Greve 2007). In this line of reasoning, innovation is stimulated by contractual incentives of the public procurer. Stimulating innovation occurs through the formulation of the demand in the procurement stages of the process (e.g. Rangel and Galende 2010). As such, it is a demand-side instrument as opposed to a supply-side instrument (Ferrari and Forastieri 2018). Part of this demand can be explicitly formulated towards achieving something innovative (i.e. contract incentives), but the procurement instruments can also allow for more freedom in the development stages by reducing the amount of design restrictions, which fosters creative processes that lead to innovations (i.e. design freedom).

Second, from the perspective of collaborative innovation literature, PPPs are a mode of collaborative governance (Brogaard 2017). Because of their long-term engagement, the public procurers and private contractors in a PPP have to collaborate with each other to generate the outcome. This multi-actor collaboration stimulates innovation, as interaction dynamics create synergy and learning processes which might generate new ideas (Torfing 2019). Studies in PPPs acknowledge these logics. For instance, Koppenjan (2005) and Warsen, Klijn, and Koppenjan (2019) emphasize the importance of interaction between the public and private partners (e.g. information sharing) and enhancing relationships (e.g. network management) between the partners in the early stages of a PPP to establish mutual trust and common understanding.

Recent studies confirm the importance of procurement-related conditions (e.g. contract structure) and collaboration-related conditions (e.g. cooperation and trust) to stimulate innovation through PPPs (Carbonara and Pellegrino 2018). However, most studies that investigate innovation through PPPs look at either procurement for innovation logics (e.g. Barlow and Martina 2008), or collaborative innovation logics (e.g. Brogaard 2017). Little research is oriented towards the combination of both logics. Some of the studies that do so, have only a small number of cases, which limits the generalizability of the studied conditions (e.g. Parrado and Reynaers 2020). Other studies, such as the recent articles of Verweij, Loomans, and Leendertse (2020) and Carbonara and Pellegrino (2020) focus largely on the amount of structural aspects of collaboration-related conditions (the presence of a stakeholder manager in the project, or the length of the concession period as a proxy for the trust and cooperation between partners). Although these studies provide indications that collaboration is necessary in PPPs to foster innovation, they remain vague as to which specific collaborative activities are necessary (e.g. information sharing and network management). These activities are however crucial in multi-actor collaboration as they stimulate learning processes which allow new ideas to emerge (e.g. Sorensen and Torfing 2011; Cinar, Trott, and Simms 2019; Torfing 2019).

In short, this article contributes to the discussion on how PPPs can stimulate innovation and which conditions are responsible for this (e.g. Rangel and Galende 2010; Himmel and Siemiatycki 2017). It contributes to this discussion on two aspects. First, it brings together insights from ‘public procurement for innovation’ literature and ‘collaborative innovation’ literature to theorize how procurement-related and collaboration-related conditions cause innovation creation in PPPs. Second, the article examines the combined effect of these conditions on innovation created in PPPs, using a fsQCA design that exploits rich survey and interview data of 24 infrastructure PPPs in Belgium and the Netherlands, which exceeds earlier qualitative studies that were limited by their number of cases.

In the remainder of the article, we start by introducing the concept of innovation and describing the ‘procurement for innovation’ logics and ‘collaborative innovation’ logics. We subsequently present the selected conditions related to these logics, and our cases and methodology. We then outline the results of our fuzzy set Qualitative Comparative Analysis (fsQCA) and discuss these results using in-depth qualitative data. In our discussion and conclusion, we reflect on the results and formulate implications for research and practice.

Two logics to enhance innovation in public-private partnerships

Innovation

Rogers (2003, 12) definition of innovation as an ‘idea, practice, or object that is perceived as new by an individual or other unit of adoption’ is widely shared in the innovation literature (De Vries, Bekkers, and Tummers 2016). We restrict ourselves in this section to how we perceive innovation in this study using Rogers’ definition. For a broader discussion on the concept of innovation, see the literature review of De Vries, Bekkers, and Tummers (2016).

Two dimensions of innovation can be extracted from Rogers’ definition: 1) innovation is something that is perceived as new, and 2) innovation is something that needs to be adopted. The first dimension of this definition ties into the case dependency of innovation. The same idea, practice or object might be regarded as very innovative in one case, but as not innovative at all in another case. Thus, we need to look at how the actors in the project perceive this idea, practice or object in order to say something about its innovativeness for that case. Second, since innovation is something that needs to be adopted, measuring innovation before it is totally implemented, for instance by using innovation evaluation scores of bids (Himmel and Siemiatycki 2017), only displays how public procurers evaluate the innovativeness of the proposed solution, but not of the implemented one. It also confines the measurement of innovation as something that is determined by the public procurer, and ignores how the private contractor perceives this innovation. Furthermore, innovation literature confirms the difficulty of quantifying innovation in an absolute way because of the difficulty of measuring something new, given the lack of standards against which to compare it, and the ambiguity in the meaning of the word ‘new’ (i.e., new for whom) (Smith 2006). Hence, we consider innovation as the perceived newness of an implemented solution, evaluated by both the public procurer and the private contractor.

We focus on product and service innovations and process innovations (De Vries, Bekkers, and Tummers 2016), both of which are innovations developed within the PPP itself, and not within the involved organizations. Product and service innovations focus on the output of the collaborative process (new product or service) (Damanpour and Schneider 2009), whereas process innovations are innovations in the way the output is generated (Walker 2014). Examples of the former are design innovations and the use of innovative building materials in the infrastructure, whereas smart construction phasing and modular construction systems are examples of process innovations. These examples also indicate that, due to the specific nature of the financial schemes and risk allocations, PPPs are more prone to produce modest and piecemeal innovations instead of more systemic and radical innovations since the latter imply higher risks (Van den Hurk and Siemiatycki 2018; Himmel and Siemiatycki 2017).

Procurement for innovation logics

Public-private partnerships are methods of procurement (Grimsey and Lewis 2007), in which the properties of the procurement process affect the outcome of the project. Public procurement for innovation refers to the demand-side rationale in which an order is placed by the public sector to fulfil particular needs (Edquist, Vonortas, and Zabala-Iturriagagoitia 2015). Public procurers are viewed as lead users who translate perceived social needs (e.g. policy goals) into new market demands through innovation (Edler and Georghiou 2007; Hueskes, Verhoest, and Block 2017; Ferrari and Forastieri 2018).

The instruments that lead to this translation of (innovation) demands are synthesized by Uyarra et al. (2014). Two of the most widely recognized instruments are 1) the level of tender specification (i.e. level of detail of demands in tenders) and 2) the kind of incentives for supplying innovative solutions (i.e. contractual incentives to provide innovation). These instruments are, according to this literature, crucial to generate innovation as they direct contractors towards creative thinking (since innovation is part of the procurer’s demands) and provide room for exploration (since the procurer’s demands are less explicitly stipulated).

Together, these conditions provide stimuli for creativity and innovation. They do this in two ways. First, rigid specifications push away innovative organizations because they might think that the procurer is not amenable to innovative solutions (Leiringer 2006; Uyarra et al. 2014). The absence of rigid specifications to stimulate innovation might be especially important in PPPs because of the long-term engagement of the partners, which means that partners have more time to experiment and are less risk averse in comparison to similar short-term projects. These conditions refer to what we will call ‘design freedom’. Second, ‘[…] by placing a sophisticated demand upon market, […] public procurers can introduce strong incentives for private providers to come up with new solutions’ (Lember, Kattel, and Kalvet 2015, 405). Because of the inherent fuzziness of the demand in long-term projects such as PPPs, this should give more direction to the private partners about what the procurer actually wants. This set of conditions pertains to what we will call ‘stimulating tender award criteria’.

Design freedom

Design freedom is defined as the lack of restrictions in the design of the project in the form of planning restrictions, physical restrictions, spatial planning procedures, and reference designs (Leiringer 2006). These restrictions are necessary for setting specific boundaries within which the private contractor has to operate (Tadelis and Bajari 2006). By defining such restrictions, the public procurer can control the behaviour and performance of the private contractor in order to ensure the desired outcomes. However, rigid restrictions and specifications can be harmful for the innovative potential of the project because innovative organizations are less likely to engage in such contracts (Uyarra et al. 2014); moreover, they impede the creative search for solutions of the contractor (Matthews 2005). A lack of rigid restrictions increases the design freedom of the contractor and hence stimulates innovations in the project.

Stimulating tender award criteria

‘Stimulating tender award criteria’ refers to the criteria that the procuring authority uses – before contract close – for assessing the tenders in order to ensure that desirable outcomes are attained (Georghiou et al. 2014). Merely focusing on price is insufficient for stimulating the private partners in the PPP to be innovative. For example, Georghiou et al. (2014) reported that 60% of the surveyed firms in their procurement study perceived the sole evaluation on price of tenders as significantly reducing the potential of innovation in the PPP projects. Stimulating tender award criteria are part of a larger group of control instruments with which the procuring authority attempts to intentionally influence the behaviours of private actors to achieve the goals of the public procurer (Hueskes, Verhoest, and Block 2017). These procurement instruments enable the public procurer to stimulate markets in adopting innovations (Edler and Georghiou 2007), which are subsequently applied in cases where the perceived needs for the innovation are high. The use of innovation measures in tender evaluations should, therefore, stimulate the likelihood of innovation in the later project.

Collaborative innovation logics

Innovation is not only produced by demand instruments, but is also determined by the interaction between the procurer and the contractor (Edler et al. 2015). The customer depends on information about the supplier’s abilities and expertise, whereas the supplier relies on information about the customer’s needs (von Hippel 1986). The more the partners exchange information with one another or are encouraged to productively interact with each other, the higher the likelihood of achieving innovative outcomes. Recent research of Brogaard (2017) for instance, already indicated a link between ‘collaborative innovation’ and ‘innovation in PPPs’.

The collaborative innovation conditions are established in recent public administration research into the generative mechanisms for collaborative innovation (Ansell and Torfing 2014). Following these authors, two of the most important generative mechanisms are synergy and learning. Synergy refers to the combination of skills, perspectives, and resources of actors with the purpose of jointly creating something new (Lasker, Weiss, and Miller 2001); by contrast, learning refers to the process of using prior interpretations to construct new interpretations that direct people’s actions (Mezirow 2000), which is stimulated in group interactions (Van den Bossche et al. 2011). Both of these mechanisms have been linked to PPP construction projects (Grotenberg and Arwin 2018). To create synergistic dynamics and learning processes, individuals need to exchange information with each other and they need to be encouraged to interact with each other (Sorensen and Torfing 2011). Two conditions are therefore linked to these mechanisms, namely information sharing (i.e. the informal exchange of information) and network management (i.e. the management of collaborative interactions).

Information sharing

This study considers the relevance of voluntary information sharing between public and private actors in the project for innovation, because it refers to the sharing of knowledge and experiences to stimulate mutual learning (Jean and Rashman 2018) and to inducing synergetic opportunities by combining different ideas and perspectives (Wegrich 2018). The exchange of various opinions, perspectives, experiences, and knowledge, stimulates transformative learning between the partners, which generates novel solutions (Torfing 2019). Even in rather simple collaborative arrangements with only public organizations, a lack of communication and knowledge sharing can be a barrier for innovation (Cinar, Trott, and Simms 2019). In more complex arrangements such as PPPs, we expect an even larger impact on the innovative capacity of the partnership because the perspectives and knowledge of actors in a cross-sectoral collaboration are more diverse due to the varied backgrounds of the stakeholders (Hartley and Benington 2006). Alam, Kabir, and Chaudhri (2014) concluded that the sharing of knowledge and information about the PPP project between the partners increased their collaboration and encouraged innovative ideas.

Network management

The effectiveness of network management on innovation in collaborations has been reported in multiple studies (see e.g. Stevens and Verhoest 2016). Network management is defined as ‘the deliberate attempt to govern processes in networks’ (Klijn, Steijn, and Edelenbos 2010, 1065). This can relate to the intention to connect people in the partnership (connecting network management) and the intention to discover various opinions and interpretations (exploring network management) (Klijn, Steijn, and Edelenbos 2010). Examples of the first strategy are the efforts of actors to align opposing interests, whereas attempts that render the visibility of different assumptions illustrates the exploring strategy. These network management strategies directly match the generative mechanisms of learning and synergy because the connections between the stakeholders at which network management is aimed, constitute ties of social interactions, which are the vehicles of learning and synergy processes.

Hypotheses

This study conceptualizes the relation between the aforementioned conditions and innovation in terms of set relations (Schneider and Wagemann 2012). While we will later introduce Qualitative Comparative Analysis as a set-theoretic method in more detail (cf. section ‘fuzzy-set Qualitative Comparative Analysis’), it is important at this point to understand that set relations are typically understood in terms of sufficiency and necessity, and not in terms of significance, and that this has implications for thinking about the relations between conditions.

A condition is necessary if the outcome cannot be produced without that condition, whereas a condition is sufficient not only when it consistently leads to the outcome, but also when the outcome can be produced by combinations of other conditions. Schneider and Wagemann state that ‘arguments about set relations are pervasive in the social sciences, but this is not always obvious’ (2012, 1). We will now argue that debates on the relations between procurement for innovation logics and collaborative innovation logics in the PPP literature implicitly use set-theoretic argumentation, which are, however, hardly translated into set-theoretic research designs. For reasons explained in the methodological section, we consider how (combinations of) conditions lead to ‘high levels’ of innovation in PPPs.

First of all, although the literature recognizes the potential benefits of combining procurement and collaboration (Edler et al. 2015), studies that combine both of these logics underline the importance of procurement as opposed to collaboration. It has been argued that large DBFM projects are little conducive of collaborative behaviour because of the strict separation of public and private management roles (Verweij, Loomans, and Leendertse 2020). Recent empirical evidence of Parrado and Reynaers (2020) on innovation in three PPPs reveals that the partners in the PPP predominantly follow the contract at the expense of collaborative efforts. Research from Carbonara and Pellegrino (2020) confirms this view, as the authors are unable to find a significant positive relationship between what the authors call ‘network structure’ (i.e. length of concession period, indicating the level of trust between the partners) and innovation through PPPs. This suggests that the ‘procurement for innovation’ logics might be necessary for high levels of innovation.

Hypothesis 1: Stimulating tender award criteria (STAC) and design freedom (DF) are necessary for high levels of innovation in PPPs.

However, Parrado and Reynaers (2020) indicate that collaborative behaviour between the private and public actor occasionally spurred innovation, but this was only possible if the innovation did not clash with contractual clauses. Other authors have suggested the complementarity of procurement and collaboration in PPPs. Poppo and Zenger (2002) show that formal contracts are combined with high amounts of relational management. Similarly, Roberts and Siemiatycki (2015) show how the combination of project management (procurement-related) and process management (collaboration-related) in PPPs allows for a coherent end goal while facilitating meaningful collaboration, which together leads to successful projects. This suggests that, at the very least, we would expect to observe the presence of both logics in PPPs with high levels of innovation.

Hypothesis 2: The combined presence of stimulating tender award criteria (STAC), design freedom (DF), information sharing (INFOS) and network management (NM) is sufficient for high levels of innovation in PPPs.¹

Cases and methodology

Public-private partnerships in Belgium and the Netherlands

We used data from 24 PPPs in Belgium and the Netherlands, in the form of Design, Build and Maintain (DBM), Design, Build, Finance and Maintain (DBFM) and Design, Build, Finance, Maintain and Operate (DBFMO). Before we started the data collection, a project database was created with 71 DBM/DBFM(O) projects which included all the Dutch and Belgian projects which had achieved a contract close between 2007 and 2015. We selected projects based on three characteristics: policy sector, contract type and size. The selected projects reflected the variation across the projects in the database. Two types of projects were therefore selected: transport infrastructure projects (railways, roads, and sluices) and social infrastructure projects (swimming pools, prisons, and government buildings). We selected projects in Belgium and the Netherlands because they are quite similar. This has two reasons. First, PPPs are politically supported by government in a similar manner in Belgium and the Netherlands. Second, Belgium adopted some of the manuals and instruments (e.g. contract templates) from the Netherlands when it started exploring the possibilities of PPPs, which resulted in similar projects in the two countries. Furthermore, as PPP projects are rather scarce in both countries, the scope of comparative research is limited when only focusing on one country.

We collected data through 71 semi-structured interviews of 74 professionals who were closely involved in one or more of the stages of the projects. These professionals included both public procurers and private contractors. Prior to each interview, each professional was sent a survey in which the concepts of theoretical interests were addressed in a standardized manner. The interviews were transcribed and coded in Nvivo, which added rich qualitative data.² Both the survey and interview materials were used in the operationalization of conditions (see section on calibration). Each of the conditions present in the transcribed interviews were coded in NVivo, after which a table was produced with synthesized information about each of the conditions (per interview). This synthesis of the interviews not only provided information from the transcribed interviews (i.e. quotes), but also provided information about the interviewer’s assessment on what was said by the respondent and extra contextual information, needed to correctly interpret the answers of the respondents. This rich interview data was therefore complementary to the more standardized survey data, and both data sets were used in the calibration process. The interviews are in addition used to illustrate more concretely how conditions came to have an effect. In doing so, we benefited from QCA’s ability to shed light on both cross-case patterns in the data and stay close to the original case data (Schneider and Wagemann 2010). The QCA method is further discussed in the next sections.

Fuzzy-set qualitative comparative analysis (QCA)

Qualitative Comparative Analysis (QCA) is a set-theoretic and case-based method that uses insights from multiple cases (Ragin 2008). It helps to think of QCA as a rigorous comparative approach that seeks to meet two apparently contradicting goals (Ragin 1987): do justice to the complexity of each case by gaining in-depth knowledge, but also revealing regularities and patterns across cases that are to some extent generalizable. Familiarity with cases is very important, both before, during and after the analytical moment of QCA (Schneider and Wagemann 2010).

As the method is based on the principles of set theory, formal logic, and Boolean and fuzzy algebra, QCA uses a specific terminology: the word ‘condition’ is used instead of ‘independent variable’, ‘outcome’ instead of ‘dependent variable’, and results are called ‘solution terms (or formula)’ (Schneider and Wagemann 2010). In QCA, cases receive membership scores for each condition and outcome, which are all sets. These membership scores, or ‘case scores’, have to be calibrated to reflect the presence of a case in a certain set of conditions or outcomes. This article utilizes fuzzy-set QCA (fsQCA) in order to allow conditions to display different degrees of membership in sets. A thorough introduction to the fsQCA method is outside the scope of this article, yet a detailed overview can be found in Schneider and Wagemann (2010).

The choice to apply fsQCA is theoretically and empirically grounded. First, theoretically, we demonstrated that we have reasons to assume that the conditions we use will have a combined effect on our outcome (innovation) (Schneider and Wagemann 2012, 78). Second, empirically, the advantage of using fsQCA is that we are able to examine these potential combined effects on a medium-sized dataset of 24 observations, which is extremely small for a regression analysis and very large for in-depth case studies. Different from the correlation coefficient in regression analysis, QCA provides two central measures as parameters of fit, namely consistency and coverage. Consistency reflects the degree to which cases sharing a combination of conditions have the same outcome. Coverage indicates the extent to which the outcome is covered by a condition or a solution term (Ragin 2008).

Operationalization and calibration of outcome and conditions

Operationalization

We consider innovation as the perceived newness of an implemented solution, evaluated by both the public procurer and the private contractor. These solutions can be technologies, products or maintenance solutions (Hueskes and Verhoest 2016). We make a distinction between innovations related to technologies, materials, smart designs, adjustments to the environment and sustainability solutions. We asked both the public procurer and the private contractor to indicate on a bipolar 10-point scale if they thought that there were no innovative solutions or there were many innovative solutions in their project. We did this for all five of the mentioned solutions. Table A2 of the supplementary material shows the items used for this question. In the interviews, we asked the public procurer and private contractor for examples of these innovations, in order to ensure that there were concrete examples of such solutions.

The procurement for innovation logics are operationalized through two conditions, design freedom and stimulating tender award criteria. First, design freedom is defined as the lack of restrictions in the design of the project in the form of planning restrictions, physical restrictions, and restrictions due to previous outcomes, spatial planning procedures, and reference designs (Leiringer 2006). We asked the respondents in the interviews to indicate whether there were restrictions present in the design of the project in the form of five types of restrictions: 1) planning restrictions, 2) physical restrictions, 3) restrictions due to previous outcomes, 4) spatial planning procedures, and 5) reference designs. Second, stimulating tender award criteria are defined as the criteria the procuring authority uses – before contract close – for assessing the tenders in order to ensure that desirable outcomes are attained (Georghiou et al. 2014). To measure these criteria, we used a 7-point Likert scale with which we asked the respondents to indicate how much they agreed with the following statement: ‘In the tender award criteria, market actors could make a difference by proposing creative solutions’ (see Table A2, supplementary material)

The collaborative innovation logics are operationalized through the conditions information sharing and network management. For information sharing, we used the voluntary information sharing between public and private actors in the project. We used a 10-point scale to measure the answers of the respondents on three questions. The respondents were asked 1) to what extent the contract partners were willing to share relevant information with each other, 2) to what extent the contract partners were keeping the other partners posted about events or changes that might be relevant for these partners, and 3) to what extent the private or public partners shared all relevant information with their organization. Network management was defined as ‘the deliberate attempt to govern processes in networks’ (Klijn, Steijn, and Edelenbos 2010, 1065). We used validated survey scales of Klijn, Steijn, and Edelenbos (2010) which refer to two types of strategies, namely 1) exploring strategies, and 2) connecting strategies (see Table A2, supplementary material).

Calibration

The membership scores of cases in each set are generated in the calibration procedure based on the assessment of the cases and theory (Schneider and Wagemann 2012). Three anchor points define a set: full membership (score of 1), full non-membership (score of 0), and a crossover point (score of 0.5). Furthermore, most conditions are considered multi-dimensional, and are therefore composed of several items. Tables A1 and A2 of the supplementary material illustrate the calibrated data set, the type of data used for measuring and calibrating the outcome and the conditions. Depending on the richness of the data the surveys and interviews provided, we worked with survey items, interview data, or both to calibrate our conditions.

Our reliance on multiple data types (survey and interview), multiple sources (different respondents per project), and multiple items poses a paradox. On the one hand, triangulation creates for a rich data environment on which to rely for assigning and cross-checking membership scores against cases, thus doing justice to the case-sensitive nature of the QCA method. On the other hand, the diversity in data complicates the development of a calibration procedure that is systematic across conditions. To address this issue, we developed a conservative calibration procedure in which we placed strict requirements on cases before they could be incorporated in a set. Additionally, we conducted a robustness check for the presence of our outcome, on which we elaborate in the next section of the article.

The conservative calibration standard imposed strict demands on cases before they could be included in a set. In practice, cases had to exhibit high levels of the outcome or a certain condition before they could be part of a set. In other words, a case that was present in a set, would indicate that it had a high level of the particular condition or the outcome of that set (i.e., high levels of innovation, high levels of information sharing, …). We relied on different levels of calibration to ensure the quality of the data for the condition. For detailed accounts of the calibration procedure per condition, we refer to Table A6 of the supplementary material. We will only discuss in more detail the calibration process with respect to our outcome, high levels of innovation.

We used both quantitative (survey) and qualitative (interview) data to calibrate the outcome. The calibration procedure comprised out of two levels, namely the calibration at the level of the respondents and the calibration at the level of the cases. We calibrated the individual scores of the respondents because each case constituted several respondents (up to four). The calibration of the scores of the individual respondents subsequently allowed for the calibration of the case-scores. As a qualitative check on the survey answers, we asked the respondents to elaborate on their survey answers in the interviews, which were evaluated by the case knowledge of the researchers. We checked how many examples of innovations the respondents could give to verify their survey answers. Once the case scores were calibrated for the survey and interview answers of the respondents (see Table A6 of the supplementary material for the calibration rules), we combined the two case scores following specific rules (see also Table A6 of the supplementary material) to obtain one case score that describes the extent of the perceived innovation in the project.

The calibration procedures that we adopted, resulted in the qualitative selection of specific answers of case respondents. Only the answers of private actor respondents for the conditions ‘design freedom’ and ‘stimulating tender award criteria’ were included because the private respondents were most susceptible to those conditions and could provide the most correct and consistent answers to our questions. Marx and Dusa (2011) propose a threshold table with probability measurements in which the probability of generating solution paths on random data cannot be greater than 10%. With four conditions and 24 cases, the probability of generating results on random data is 2%, which is well below the threshold that the authors suggest.

Results

The analyses were performed using the fs/QCA 3.0 package. After the calibration procedure, a truth table was constructed, which lists all the logically possible combinations of causal conditions (configurations), and sorts all the cases along these combinations (Ragin 2008). Each possible combination of conditions (2 ^k; k = number of conditions) is presented as a row in a truth table (see Table A4 of the supplementary material). According to standards of best practice (Schneider and Wagemann 2010), we first present and discuss the results for the analysis of necessary conditions, after which we turn to the analysis of sufficient conditions. In Table 1 we report the number of cases with high levels of innovation (scores above the cross over point of 0.5) and the number of cases with low levels of innovation.

Table 1. Set membership of cases for the outcome

Levels of innovation in the projects		Number of cases
Low levels of innovation	Below 0.5	12
High levels of innovation	Above 0.5	13

First, we examined the necessity of the conditions to explain innovation (see Table 2). For necessary conditions, a consistency threshold of at least 0.90 is advised (Schneider and Wagemann 2012). Table 2 indicates that neither the presence nor the absence (~) of any of the conditions is necessary for the outcome. We also explored the necessity of conditions for the absence of the outcome, for which similar results were visible (see Table A3, supplementary material).

Table 2. Analysis of the necessary conditions

Presence of high levels of innovation
Conditions	Consistency	Coverage
High levels of information sharing (INFOS)	.726860	.649109
Absence of high levels of information sharing (~INFOS)	.452813	.427959
High levels of network management (NM)	.606171	.715203
Absence of high levels of network management (~NM)	.603448	.453615
High levels of design freedom (DF)	.788566	.578562
Absence of high levels of design freedom (~DF)	.361161	.443207
High levels of stimulating tender award criteria (STAC)	.727768	.668333
Absence of high levels of stimulating tender award criteria (~STAC)	.451906	.415000

We then reviewed the sufficient conditions for the presence of high levels of innovation. We constructed the truth table for the conditions that were assumed to explain the high levels of innovation in PPPs (see Table A4, supplementary material). Particularly in small- and medium-N studies, no empirical evidence is available for all possible combinations, or rows (16 or 2⁴ in our study). We followed standards of practice. First, we only included rows with at least one case that was relevant for the empirical analysis (Ragin 2008), thereby resulting in 11 rows (see Table A4, supplementary material). Second, we selected only those combinations with a raw consistency level of 0.80 or higher (Schneider and Wagemann 2012). Third, we observed a substantial drop in the raw consistency between the lowest consistency level of the selected paths (above 0.80) and the highest consistency level of paths that were not selected (below 0.80), which was also an indication that the threshold was reached (Schneider and Wagemann 2012, 128).

The results of our intermediate solution are illustrated in Table 3. The intermediate solution generates solution paths based on theoretical assumptions. Our directional expectations are outlined in Hypothesis 2, and essentially mean that we expect that the selected conditions will all be present in the solutions paths. The intermediate solution generated one distinct solution path. The path in Table 3 was present for 7 of the cases. Our coverage score of 0.52 reveals that more than half of the cases are described by the path in Table 3. With a consistency level of ca. 0.90, the path accurately depicts the presence of high levels of innovation in those cases. There were no tied prime implicants, hence there was no model ambiguity.

Table 3. Intermediate solution for high levels of innovation

	Consistency	Raw coverage	Unique coverage	# Cases in path
INFOS * NM * DF	.895899	.515427	.515427	7
Solution consistency	.895899
Solution coverage	.515427

Maggetti and Levi-Faur (2014) argue that solution paths for fsQCA always have to be interpreted using the intermediate, complex and parsimonious solutions. The parsimonious solution generated an identical solution path and consistency/coverage scores (see Table A5, supplementary material). There were no complex solution paths generated by our analysis. Furthermore, no contradictory cases emerged (i.e. cases that exhibit the solution path but not the outcome). The results in Table 3 can be described as follows:

PPPs that display high levels of information sharing (INFOS), network management (NM) and design freedom (DF), have high levels of innovation.

We applied a robustness check to ensure that the paths we observed in our data were adequately robust to withstand some small changes in the calibration procedure. We adjusted the calibration for the outcome by making the interview data more important than the survey data (i.e. starting from the interview data and correcting with the survey data). Our previous calibration rules only corrected the survey data with the interview data when the data were not interpretable using only the survey data (i.e., answers both above and below the cross over point for the same case). This approach yielded an identical solution path of INFOS * NM * DF for the presence of innovation, with a coverage score of 0.42 and a consistency score of 0.84, hence confirming the explaining power of this path for the presence of high levels of innovation. Additionally, even when we dropped the raw consistency level in the truth table slightly below the recommended point of 0.75 (Schneider and Wagemann 2012, 127), we still observed our identified solution path.

A qualitative analysis of the interviews sheds light on the underlying mechanisms of this solution path. From the interviews, it seems that the contract conditions were not always desirable anymore for the procurer at the time of the construction process. Innovations often spontaneously emerged from interactions between partners because of random events in the phases after contract close. The following quote illustrates this:

The point was that there was a bid on the old output specifications. But after the procurement phase, the discovery came that in specific places, we actually wanted something else. There are several concrete examples of this. […] One example is a self-service system for multiple things. […] This was initially not included, but it was eventually included in this manner [after contract close, ed.]. Everything was squeezed into a mobile device. But such devices did not exist. It had to be designed.

Furthermore, the interviews indicate that all of the respondents in the cases covered by the solution path experienced the collaboration between the partners as positive. Words such as constructive, positive, pleasant and smooth collaboration were frequent answers to this question. The interviews exhibited that the openness towards each other, together with the design freedom, fosters creativity and innovation in the project.

The collaboration was constructive. It was really a municipality that was open for ideas. Even though we had a contract with each other, they were always curious if we had some new techniques [and] new innovations in our expertise.

The interviews also seem to indicate a mutually reinforcing effect of design freedom, information sharing and network management. In cases where a detailed design was drafted after contract close, the induced design freedom for the private partners created opportunities for information sharing, exploring other’s ideas and connecting people in the partnership, which led to innovative outcomes. The following quote from a public partner of one of the cases covered by the solution path indicates the mutually reinforcing nature of these innovation mechanisms:

Our initial program with demands wasn’t very detailed. The detailed design [which came after contract close, ed.] was however drafted in detail. At that moment, the plans were discussed thoroughly and there were some changes left and right, [for example, an expansion of the cafeteria of the adjacent sport complex, ed.].[…] The collaboration with [the private contractor] was very open. Even before the exploitation phase, the project was very open. We had a steering committee in which we and [the private contractor] were represented to discuss issues that were suggested in work meetings. […] As our location is adjacent to [a river], we didn’t know exactly how to keep this site dry. [The private contractor] found something that could work with a draining system. Although we knew this problem was evident because our consulting firm had already confronted us with it in the procurement phase, the innovation was established especially in the detailed design.

Discussion

Contributing to the recent literature on innovation creation through PPPs, we aimed to address what makes that PPPs create innovations. Scholars in fragmented literatures have pointed at the importance of conditions related to ‘procurement for innovation’ logics and conditions related to ‘collaborative innovation’ logics. Yet we lacked an empirical understanding on whether and how these conditions combine in facilitating innovations, and the relative importance of conditions within these combinations. Scholars underline the necessity of procurement-related conditions for innovation in PPPs, as the strict separation of public and private management roles is considered a limiting factor for collaborative behaviour (Verweij, Teisman, and Gerrits 2017), as is the tendency of actors to primarily follow the contract at the expense of collaborative efforts (Parrado and Reynaers 2020). However, scholars have also suggested that collaborative conditions complement procurement-related conditions to create innovation in PPPs (Poppo and Zenger 2002; Roberts and Siemiatycki 2015; Parrado and Reynaers 2020).

The QCA method was applied in this article, which is well suited to shed light on the relative importance of conditions by distinguishing necessary from sufficient conditions. As QCA is a case-based method, it uses insights from multiple cases. This study uses different data sources, which does justice to the complexity of each case, and returns to the interview materials to make sense of the observed patterns. Yet QCA’s key operations rely on Boolean algebra, where each case is reduced to a series of conditions. This study, too, is very transparent in the calibration process (see Table A6, supplementary material) as a standard of good practice to allow for replication (Schneider and Wagemann 2010). As such, we were able to study a large number of cases (24 projects), which creates opportunities for cautious generalization of the results. Future studies can formulate and apply propositions to cases that share a reasonable number of features with the cases observed here (transport and social infrastructure PPPs and/or DBM/DBFM(O) projects in Western democracies) (Ragin 1987).

Rejecting our first hypotheses, results showed that neither of the ‘procurement for innovation’ logics (design freedom and stimulating tender award criteria) were necessary for the presence of high levels of innovation. Additionally, and largely supporting our second hypothesis, we found that design freedom, information sharing and network management together lead to high levels of innovation in PPPs. The fact that procurement-related conditions are not necessary for innovation in PPPs, introduces more nuance in the discussion about the benefits of contractual stimuli to stimulate innovation in PPPs. It seems that designing the proper contract (i.e. a contract which formulates explicit innovation demands and ensures design freedom) is not enough to create innovation in PPPs. As the qualitative results from the interviews show, conditions related to the contract are especially important in the early phases of the project, while a lot of situational dynamics are created in the collaborative phases of the project. This might explain why the effect of contract related conditions on innovation is weakened in later stages of the project.

Still, the results indicate that the combination of design freedom, information sharing and network management is sufficient to create high levels of innovation in PPPs. This confirms literature that suggests that procurement and collaboration complement each other (e.g. Poppo and Zenger 2002; Edelenbos and Teisman 2008; Roberts and Siemiatycki 2015). Especially after contract close, collaborative innovation logics come into play, and they can enhance innovation during the execution of the project. Interview data confirms this, as all of the respondents in the cases covered by the solution path experienced the collaboration between the partners as positive, which is contradictory to the non-collaborative nature of large PPP projects that the literature suggests (Verweij 2015; Verweij, Teisman and Gerrits 2017). As such, positive collaboration caused by high levels of information sharing, exploring ideas and connecting individuals (i.e. network management) causes, together with the presence of design freedom, synergy and learning processes to occur in the stages after contract close.

We could, however, not fully accept our second hypothesis. The interviews hinted at the importance of the specific circumstances under which PPPs work, which might explain why ‘stimulating tender award criteria’ is not part of the solution path. The rationale behind stimulating tender award criteria is to influence the behaviour of the private partners at the start of the project to produce innovative outcomes (Hueskes, Verhoest, and Block 2017). However, as PPPs are long-term collaborative arrangements between actors operating in relatively diverse settings, the likelihood of having to adapt to unforeseen circumstances is a lot higher in comparison to traditional procurement of products and services.³ A consequence might be that the effect of stimulating conditions introduced at the start of the project (e.g. stimulating tender award criteria) diminishes throughout the lifespan of the project as dynamic changes require other stimuli (e.g. design freedom, information sharing and network management) to achieve innovation. These stimuli have in common that they facilitate creative discovery and exploration, which are crucial properties of innovation creation (Crosby and Torfing 2017). The results therefore emphasize the special importance of collaborative innovation logics in long-term projects such as PPPs.

The fact that PPPs are long-term, collaborative arrangements also supports an additional, compelling mechanism, as the assessment of the interviews indicates that design freedom, information sharing and network management not only complement but also reinforce each other. Research has argued that innovation is created by establishing an open and flexible environment in which stakeholders can collaborate with each other and which fosters synergy and learning dynamics (Ansell and Torfing 2014; Torfing 2019). Design freedom, information sharing and network management are mutually reinforcing when design freedom is an enabler for flexible contract interpretation, exploited through thorough information sharing, exploring ideas and connecting individuals (i.e. network management). Design freedom facilitates the formulation of a detailed design after contract close, during which information sharing and network management are particularly important for aligning the stakeholders’ visions about that detailed design, thus spurring learning and synergy dynamics. By enhancing the flexibility in contract interpretation through design freedom, exploration activities such as learning and synergy, which are stimulated by collaboration (information sharing and network management), are stimulated.

Conclusion

This article aimed to contribute to the current literature on two aspects. First, it brings together insights from ‘public procurement for innovation’ literature and ‘collaborative innovation’ literature to theorize how procurement and collaboration cause innovation creation in PPPs. Second, the article examines the combined effect of these conditions on innovation creation in PPPs, using a fsQCA design that exploits rich survey and interview data of 24 infrastructure PPPs in Belgium and the Netherlands, which exceeds earlier qualitative studies that were limited by their number of cases.

Multiple explanations for innovation processes were developed in this article. We started from the premise that both ‘procurement for innovation’ logics and ‘collaborative innovation’ logics might cause the creation of innovation in PPPs. None of these conditions were however necessary to create high levels of innovation. Particularly, the conditions related to procurement for innovation logics were not necessary, which goes against the assumptions in the literature that procurement for innovation is more important in PPPs than collaborative innovation. However, procurement and collaboration seem to have combined effects on innovation in PPPs. Design freedom complements and reinforces information sharing and network management in the phases after contract close and exhibit effects on the innovations. Innovation in PPPs is a process situated in a dynamic environment and influenced by combinations of multiple types of conditions, acting on multiple points throughout the lifetime of the project.

Methodologically, we relied on the fsQCA method, which benefited from the triangulation of different data collection methods (survey and interviews). This approach allowed us to propose claims about both the combined effects of procurement for innovation and collaborative innovation logics in 24 cases, and the reinforcing nature of these conditions. By going back to the interviews, we were also able to better understand and explain these results. Furthermore, our calibration procedures were based on in-depth descriptions of the conditions, available through the interviews, which made a rigorous calibration possible and allowed us to perform a robustness check on the results. Moreover, because of the specific properties of our case selection and the QCA method, we believe that cautious generalization of our results to similar projects (i.e. transport and social infrastructure PPPs and/or DBM/DBFM(O) projects in Western democracies) is possible.

Our results have important implications for theory and practice. First, innovation through PPPs does not solely depend on contractual stimulation. Researchers and practitioners need to consider innovation as something that is not simply controlled for by procurement-related conditions at the start of a project. Innovations might emerge from dealing with random challenges in the design and construction processes. Hence, both public and private managers need to be aware of these innovation opportunities during the processes after contract close, instead of solely relying on the stipulated conditions in the contract. Private managers need to be open for feedback of the public partner during the construction phases and public managers need to recognize the importance of real collaboration with the private partner to develop innovative ideas, instead of just ‘demanding’ innovation through the contract. Policy makers need to be aware that setting up PPPs to create innovative services only works if the public and private partners are willing to invest time and resources into collaborative activities. Our research indicates that PPPs that do not stimulate these collaborative activities are less likely to generate innovation.

Second, design freedom, information sharing and network management not only complement but also reinforce each other in enhancing innovation. Designing an environment in which exploration can occur (through design freedom), reinforces the potential impact of information sharing and network management on generating innovation. Public and private managers need to be aware that stimulating design freedom opens the door for additional innovation dynamics that work in conjunction with design freedom (i.e. exchange of information, exploring ideas and connecting people).

However, our research is not without limitations. First, we conducted a study on 24 cases, which restricted us in the level of detail that we could obtain. Consequentially, our research was more focused on the presence of the conditions and the outcome itself, and less on the mechanisms that resulted in these observations. In particular, the collaborative innovation mechanisms (e.g. synergy and learning) are not adequately understood in PPPs. Future research should consider these mechanisms more directly, for instance through longitudinal research or process tracing. Second, we concluded that innovation in PPPs evolves in a dynamic environment, which creates boundaries for which combinations of conditions work (i.e., produce high levels of innovation). However, due to our research design, we were unable to map the changes in the contracts or the relationships between the public and private partners. Future research should thus focus on the dynamical nature of PPPs and the way this affects the procurement and collaborative mechanisms and conditions. Lastly, not only conditions in the project itself can affect how innovation occurs, but also environmental conditions such as the financial context or the type of industry involved in the process. Future studies on innovation through PPPs should recognize the specific environment in which PPPs occur and how this can affect the innovation process itself.

Acknowledgments

This article would not have been possible without the intensive collaboration with Marlies Hueskes who played a crucial role in the collection of the data and provided us with advice and assistance with the case-specific details, the calibration procedures, and the interpretation of the results of our analysis. We wish also to express our gratitude to our colleagues at the Erasmus University of Rotterdam, prof. Erik-Hans Klijn, prof. Joop Koppenjan and Rianne Warsen, for their contribution to the data collection and the insights they provided to the calibration and analysis of the data.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/14719037.2020.1867228.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This research was funded by a Ph.D. grant of the Hermes Foundation represented by the Flemish Agency for Innovation & Entrepreneurship (VLAIO) and is part of a project that has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 726840. The research also benefitted from research funding within the BRAIN (Belgian Research Action through Interdisciplinary Networks) program, in particular the interuniversity PSI-CO project “Public Sector Innovation through Collaboration”, and is part of the GOVTRUST Centre of Excellence.

Notes on contributors

Chesney Callens

Chesney Callens is PhD candidate at the research group on Politics and Public Governance of the University of Antwerp (Belgium). He holds a master’s degree in Public Administration and Management from the University of Ghent (Belgium). His research interests are collaborative and network governance, innovation management and collaborative innovation, and strategic management.

Koen Verhoest

Koen Verhoest is research professor in Comparative Public Administration and Globalization at the Department of Political Science, in the research Group on Politics & Public Governance, University of Antwerp (Belgium). He is also the main promotor of the GOVTRUST Centre of Excellence on 'Trust in Multi-level Governance', one of the 13 UAntwerp Centres of Excellence. His main research interest is on the organizational aspects of public tasks/regulation and their governance in multi-level and multi-actor contexts, including the autonomy, control and coordination of (regulatory and other) agencies, the governance of liberalized markets, and the governance of (public-private) partnerships and other forms of collaboration. Moreover, he studies these aspects in an international comparative perspective, enabling to study the impact of globalization and multi-level contexts on the governance and performance of these organizational forms.

Jan Boon

Jan Boon is postdoctoral researcher at the research group on Politics and Public Governance of the University of Antwerp (Belgium). He holds a doctorate in social sciences (political science), a master's degree in political communication (University of Antwerp) and a master-after-master in multilingual business communication (Ghent University, Belgium). In 2019, he obtained a senior postdoc fellowship from the Research Foundation for his research proposal on “Reputation and Structural Reforms of Public Organizations: Explaining Temporal Dynamics”. His research interests are strategic management and reputation; organizational change, learning behavior and innovation; and machine learning and artificial intelligence.

Notes

1. In QCA terms: conditions related to procurement for innovation logics and collaborative innovation logics are an insufficient but necessary part of a solution which is itself unnecessary but sufficient for innovation in PPP. These conditions are called INUS conditions (Schneider and Wagemann 2012, 79).

2. The dataset was also used in the recent article of Warsen, Klijn, and Koppenjan (2019), in which the authors focused on how certain conditions influenced the performance of PPP projects. In our article, we focus exclusively on the influence of conditions on innovation in PPPs, not on PPP performance.

3. A couple of instruments are commonly used in PPP projects to adapt the contract to dynamics in the environment. First, changes in the environments of and relationships between the actors can incite the partners to open-up the contract again in formal renegotiations (Cruz and Marques 2013). Second, less drastic and more common alignments include on the one hand contractual adjustments or expansions (e.g. by utilizing the flexibility in the contract to modify contract conditions). On the other hand, there might also be room for the partners to reinterpret the contract conditions, or for informal agreements between the partners that adjust small, technical, and operational issues.