Public sector innovation in collaborative networks: the key role of learning

Dealing with wicked problems such as security or public health requires policymakers to come up with public-sector innovation (Osbourne & Brown, 2005). To do so, new governance structures such as collaborative networks are necessary, because wicked problems most often involve multiple stakeholders with different points of view (Head & Alford, 2005; Klijn & Koppenjan, 2016). Collaborative networks gather multiple organizations to solve problems that are difficult or impossible to solve by a unique organization (Agranoff, 2006; Agranoff & McGuire, 2001). Collaborative networks are generally assumed to generate regular interactions between multiple stakeholders with potentially conflicting interests and different resources, which would create a space for ideas circulation (Koebele, 2019). In turn, collaborative networks are supposed to foster knowledge acquisition and opinion changes among participants (i.e., individual learning), as well as goal convergence and consensus building (i.e., collective learning) about innovative processes, products or concepts. While the factors of learning in collaborative networks have been thoroughly investigated in the literature (Riche et al., 2020), not much has been said about whether and how learning actually turns into innovation. How does learning amongst the participants to a collaborative network turn into innovative solutions to wicked problems? To answer this research question, we compare eight Belgian collaborative networks dealing with wicked problems in four policy sectors: social services, environment, health and home affairs. The data collected during 78 semi-structured interviews with the network participants is investigated with a thematic analysis (Braun & Clarke, 2006; Miles & Huberman, 1994) using NVivo. For each network, we determine whether learning and innovations occurred, as well as the specific types of learning observed – instrumental, relational, and political. Then, we systematically compare the following conditions influencing the emergence of public-sector innovation: exogenous conditions (political and economic context) (Heikkila & Gerlak, 2013), collective conditions (diversity of the participants, trust, social relations) (Koebele, 2019), and conditions related to the structure of the network itself (formal and informal rules, procedural fairness) (Riche et al., 2020). The comparative design allows us to explore what type of learning, with what network conditions, lead to public sector innovation. The overall finding of this paper is that learning in collaborative networks does not trigger innovation solutions to wicked problems per se, nor in all conditions. On the one hand, not all types of learning lead to innovation and some types of learning lead to some types of innovations. On the other hand, learning leads to innovation only under certain conditions related to the network. These findings have theoretical and practical implications related to the conditions of public-sector innovation in collaborative settings through learning (Riche et al., 2020), about the successes and ‘failures’ of policy learning (Dunlop & Radaelli, 2017), as well as about the metagovernance of collaborative networks (Sørensen & Torfing, 2007) and the ability of the latter to lead to innovative outcomes (Dowding, 1995; Oh & Bush, 2016). We discuss these implications and conclude with an agenda for future research.