Introduction

Cohesion Policy (CP) is a long-standing and established scheme in the European Union (EU) (Brunazzo, 2016; Cunico et al., 2020). Accounting for the EU’s second largest budget after the Common Agricultural Policy (European Commission, 2021), it aims to sustain harmonic development within the EU by (i) redistributing common resources to foster the development of less developed regions and (ii) supporting the competitiveness of more developed ones (Commission of the European Communities, 1986). CP is implemented through a multi-level governance framework (Hooghe, 1996; Stephenson, 2013), where the EU along with national and regional/local authorities, as well as civil societies and economic associations, are involved in policy formulation and implementation (Carayannis, 2021; Yesilkagit & Blom-Hansen, 2007). Hence, the policy constitutes an open, multi-action, multi-focus, and multi-level scheme where the effects of corrective actions are often delayed (Bachtler & Ferry, 2015; Lion & Martini, 2006; Maggetti & Trein, 2019; Smeriglio et al., 2015; Stephenson, 2013), generating a complex system (Cunico et al., 2021; Zahariadis, 2013).

Given this premise, CP often delivers non-homogenous outcomes (Crescenzi & Giua, 2020; De Rynck & McAleavey, 2001; Fratesi & Wishlade, 2017). Several EU regions face difficulties in spending the allocated resources (i.e. structural funds), exhibiting (often chronic) poor absorption rates (Cunico et al., 2022; European Court of Auditors, 2018, 2021; European Parliament, 2011). For example, concerning the 2014–2020 CP period, the average annual rate of absorption has been lower than the previous year (Makszimov, 2021); at the end of 2021 (with less than 2 years left to invest the available funds over the permissible 10-year time span), some countries were still just above 50% (e.g. Croatia 54%; Slovakia; 57; Malta 58%; Denmark 58%), which entailed the risk of losing the funds if not invested on time (European Court of Auditors, 2021). Given the importance of the policy, this spending issue has captured the attention of public (European Court of Auditors, 2021) and private institutions and think tanks (Álvarez, 2020; Darvas, 2020), as well as of the media (Fotina, 2022; Makszimov, 2021).

In this paper, we advocate that this “incapacity” in absorbing funds may arise from supply-and-demand issues related to (i) public institutions’ capabilities (e.g. staff skills) to develop and manage calls and (ii) locals’ awareness and willingness to apply, respectively. Concerning supply, evidence converges to showcase how insufficient administrative capacity (in terms of “quality of formal and informal institutions”; Smeriglio et al., 2015, p. 1) results in ineffective utilisation of CP funds at regional and/or national levels (Bachtler et al., 2014; Incaltarau et al., 2020; Mendez & Bachtler, 2022; Milio, 2007). Notably, to the best of our knowledge, less attention has been dedicated for analysing how the demand for funds contributes to the observed absorption pattern. Besides aggregate analyses on the characteristics of the organisations and sectors that apply for funding (Bachtrögler et al., 2019) and scattered reflections on beneficiaries, accurate reconstructions of the factors that compose and drive demand are not available. Given the central role in absorbing structural funds, this dearth of research about CP demand is challenging.

In this context, Smeriglio et al. (2015), Wostner (2008), Blom-Hansen (2005), and Bateira and Ferreira (2002) highlight the lack of a connecting link between conceptual frameworks and the daily practices and dynamics in CP implementation. Specifically, an operational construction of how absorption practically unfolds is missing. Indicatively, concerning the supply side, “there is no overall clear idea of what builds administrative capacity and strong institutions for managing CP effectively, or whether specific external factors play a role in this process” (Smeriglio et al., 2015, p. 2). Hence, scarcity and challenges in research conceptualising and quantifying the operative components of supply (e.g. implementation processes and structures) exist (Yesilkagit & Blom-Hansen, 2007). Concerning the demand side, the dearth of studies, as abovementioned, renders the lack of operational analyses on the beneficiaries side even more noticeable. These issues has also been pointed out by Wostner (2008, p. 2) who emphasises the abundance of performance and impact evaluation studies in contrast to the limited amount of research regarding “the systematic properties of the CP implementation systems in general”, as well as by Bachtler et al. (2014, p.1) who highlight the insufficient attention given “to the dynamics of capacity evolution and learning”. Overall, it appears that knowledge of the micro-organisational foundations (i.e. operations, elements, and decisions at the implementation level) that generate CP macro-behaviours (i.e. absorption patterns over time) is lacking. Therefore, we pose the following research question: “How could operations and decisions at the LMA and beneficiaries level (micro-foundations) concur to explain observed absorption patterns (macro-behaviour)?”.

Accurate knowledge of the components constituting CP demand and supply, along with the related interconnecting forces and dynamics, is essential for researchers and policy-makers to overcome absorption performance bottlenecks, design more precise and effective policies, and avoid unintended side effects (e.g. Cunico et al., 2022). Furthermore, understanding how CP works at the micro-operational level, beyond the extant policy theoretical frameworks (Zahariadis, 2013), could significantly improve comprehension of practical aspects that have been neglected so far (Blom-Hansen, 2005), such as agents’ learning (Bateira & Ferreira, 2002) and variables’ interconnectedness (Smeriglio et al., 2015), which have already been proved to be fundamental concepts in regional development literature (Lopes & Franco, 2019; Tiits et al., 2015). Notably, this microscopic perspective could increase the granularity of existing high-level and theoretical models of CP multi-level governance (Maggetti & Trein, 2019).

In this light, the use of systemic approaches for CP analysis is emerging as critical. In these approaches, “the parts are viewed with respect to the whole to which they belong” to examine “the pattern of relationships between the parts and the processes operating within the whole” (Keeves, 2001, p. 2421) and inquire how the interaction between the single micro-elements (i.e. supply-and-demand operations) generate the system macro-behaviour (e.g. funds’ absorption rate), providing a synthetic representation of complexity usually in the form of models (Fiddaman, 2007). Studies about the actors and institutions in the CP system exist, providing a meaningful description of their functions and powers (e.g. Batory & Cartwright, 2011; Piattoni & Polverari, 2016), as well as actual attempts of “unpacking” administrative capacity (Domorenok et al., 2021). However, there is an absence of a CP supply-and-demand study that systemically investigates day-to-day organisational operations and relationships between the diverse agents. To close this gap, we use a mixed-method approach involving field research and computational analysis. Based on a field study, we develop a system dynamics (SD) model capturing supply-and-demand operations involved in CP implementation at the local managing authority (LMA) level. Notably, SD has already been employed to explore local planning and performance management in public administration (Bianchi & Tomaselli, 2015; Cosenz, 2018) and regional development (Samara et al., 2022). Our model elicits concepts and interconnections among agents and resources to describe decision processes, while it integrates existing scattered information within a sole coherent framework (Morrison & Oliva, 2018). We then formalise the model to build a computer environment using information collected through literature, individual interviews, and collective workshops with experts. Formal modelling and simulation experiments enable the mobilisation of knowledge induced from the empirical research by enhancing (i) the conceptual precision and internal validity of our case study and (ii) the theoretical elaboration and exploration through computational experimentation (Davis et al., 2007).

The rest of the paper is structured as follows. In the second section, theories on absorption capacity are briefly reviewed. The third section outlines our methodological approach. Then, the conceptual SD model portraying the CP implementation system is developed (fourth section). In the fifth section, we present simulations that unravel different compelling aspects of the system. Finally, the sixth section concludes with insights and discusses how our research contributes to the existing theory and stimulates future research.

Literature Background

During the policy cycles, researchers, practitioners, and evaluators have investigated the causes behind some regions’ poor performance (Smeriglio et al., 2015); a widely accepted explanation is the lack of absorption capacity (European Court of Auditors, 2018; European Parliament, 2011; Georgescu, 2008; Incaltarau et al., 2020; Milio, 2007; Surubaru, 2017a; Zaman & Cristea, 2011). However, absorption capacity is quite a broad concept (Wostner, 2008). It is defined as the extent to which a managing authority (e.g. local authority, state) is capable of effectively and efficiently spending the funds allocated from the policy (Jurevičienė & Pileckaitė, 2013; Martin, 2010). More specifically, it pertains to the quality of development and conduction of the entire policy cycle (Fitzgerald & Promè, 1996): from the initial phases of planning, project generation, and selection, through the implementation period, until the monitoring and evaluation tasks (Wostner, 2008). Therefore, absorption capacity can be related to (i) the supply side, referring to the public administration (i.e. LMAs) capacity of handling CP resources to facilitate the effectuation of eligible expenditure (Wostner, 2008), and (ii) the demand side, namely, the ability by project applicants to develop and conduct adequate applications and projects (Jurevičienė & Pileckaitė, 2013) by fulfilling the necessary requirements (Wostner, 2008).

The supply side of the absorption capacity can be then divided into three main elements (European Jurevičienė & Pileckaitė, 2013; European Parliament, 2011; Wostner, 2008): (i) the macroeconomic absorption capacity, (ii) the financial absorption capacity, and (ii) the administrative capacity. The first one is measured and defined as a threshold in terms of GDP; in the case of CP, the transfer of EU funds is limited to the 4% of the respective national GDP (Jurevičienė & Pileckaitė, 2013). The second one refers to the capacity of co-financing the EU-funded projects and programmes, guaranteeing that these national or regional contributions are maintained for the whole duration of the multi-annual policy, and timely and promptly collecting these contributions from other partners involved in the programmes and projects implementations (Jurevičienė & Pileckaitė, 2013). Finally, administrative capacity involves the resources influencing the internal capability of an institution to perform tasks (Coremans & Meissner, 2018); it is specifically defined as “the ability and skill of central and local authorities to prepare suitable plans, programmes and projects in due time, to decide on programmes and projects, to arrange the coordination among principal partners, to cope with the administrative and reporting requirements, and to finance and supervise implementation properly, avoiding irregularities as far as possible” (Boeckhout et al., 2002, p. 2). Researchers and evaluators have emphasised local administrative capacity as a crucial explanatory factor of policy performance (Bachtler et al., 2014; Incaltarau et al., 2020; Milio, 2007; Surubaru, 2017a) within a broader debate on the quality of local governance (Charron, 2016; Mendez & Bachtler, 2022; Surubaru, 2017b).

As already mentioned, knowledge on the demand side of absorption is rather limited, although funds’ applicants often come into discussion given their centrality in the implementation process. In general, research efforts that address beneficiaries and applicants tend to focus on (i) assessing the CP impact on their organisations and their characteristics (e.g. how firms that receive funds improve compared to those not receiving or the determinants of the heterogeneity among the firms benefitting from CP funds; Bachtrögler et al., 2019; Piątkowski, 2020) and (ii) beneficiaries’ influence on CP implementation procedures and priorities’ definitions (e.g. NGO clientelism; Surubaru, 2017a, b). However, to the best of our knowledge, the aspects (e.g. barriers and drivers) that shape the actual demand for CP funds are still rather obscure since only few scattered and unstructured surveys on problems encountered by applicants and beneficiaries were conducted (e.g. Jaliu & Rǎdulescu, 2013; Jurevičienė & Pileckaitė, 2013; Zaman & Cristea, 2011), leading to elements or problems beneficiaries may face during their CP journey. Limited access to data, such as information on implemented projects and applicants, may be among the causes of this gap (Rinaldi, 2020). Specifically, collecting primary data on funds’ actual and potential CP beneficiaries is a challenging task for scholars; except for public institutions, private ones are also involved, which seem to be fragmented particularly in Europe (e.g. Daskalakis et al., 2013).

In general, the literature has predominantly adopted what could be identified as a top-down approach when investigating CP implementation. Specifically, it has mostly investigated the theoretical constructs, high-level factors, and trends that play a crucial role in affecting absorption rates. Thus, research on absorption rate has been mostly conducted through macro-approaches, such as statistical analyses, using either available CP-related datasets (e.g. Bachtrögler et al., 2019; Dicharry, 2022; Dotti, 2016; Incaltarau et al., 2020; Mendez & Bachtler, 2022; Surubaru, 2017a; Tosun, 2014) or conceptual efforts (e.g. Mendez & Bachtler, 2011; Milio, 2007). This body of literature has shaped the current knowledge boundaries, which provide a shared understanding of the macro-factors that influence CP absorption. However, macro-approaches tend to overlook or fall short of capturing the complexity and dynamism of CP. Indicatively, static and high-level analyses may not be able to explain why some regions demonstrate chronically low absorption trends, while others improve their performances (Cunico et al., 2022). To that end, it is important to reflect upon CP in a more granular and operational manner by examining how the micro (individual) aspects build up the meso (organisational) levels which, in turn, affect the observed macro (systemic) behaviours (Cunico et al., 2022; Domorenok et al., 2021). This approach will attempt to reconstruct the operations at the base of the LMA implementation of CP, through field research and analysis of existing literature, and then incorporate them into a coherent model to explain the macro-behaviour. Particular attention is devoted to the “system” level that has been studied less from an operative perspective (e.g. applicants’ side). Such type of models that integrate different perspectives have already drawn attention and been used within the regional development research (e.g. Lopes & Franco, 2019).

Methodological Approach

In the literature, primary data about the LMA and beneficiaries’ decisions and operations is scarce. More importantly, there is a dearth of theorising on how an emergent epiphenomenon (e.g. funds’ absorption rate) is connected to micro-behavior. To address this gap, we use a mixed-method approach combining empirical evidence and computational analysis. First, to build our theoretical framework, as described by Glaser and Strauss (1967), we used a field study to identify the constructs necessary to represent the phenomenon under analysis, elicit their conceptual properties, and articulate the hypotheses regarding the causal relationships among those constructs.

In our field study, we directly observed CP policy implementation operations in the Emilia-Romagna region (Italy) by examining real-world data on regional procedures and related reports. This region was selected due to data availability and their willingness to share data that were not publicly available. More importantly, we selected the case of Emilia-Romagna for its properties as an ideal type for theorising since the region crystallises the essential traits to explain the phenomenon under study (Swedberg, 2005). Notably, this choice appeared to be reasonable given that Emilia-Romagna’s absorption behaviour follows the prototypical trends (Aiello et al., 2019b; Aivazidou et al., 2020), demonstrating sufficient CP funds absorption over the years (Aivazidou et al., 2020; Arbolino et al., 2016; Corte dei Conti, 2017; Hooghe, 1998). Emilia-Romagna is a region in central Italy with a GDP per capita of 115% of the EU average and a contained unemployment rate (7.7%) (Eurostat, 2022), while it has experienced relatively stable governance over the years. Thus, given its limited deviation from the average European regional conditions and good absorption performances, the Emilia-Romagna region could act as a representative case of the local CP implementation process to foster the generalisation of our results (Aiello et al., 2019b; Aivazidou et al., 2020).

The data collected were then complemented with additional qualitative data (Appendix A), which are instrumental in looking into a partially unexplored research context (Eisenhardt & Graebner, 2007) as it is the collection of processes through which LMAs and beneficiaries interact. First, we performed sixteen semi-structured interviews (audio-recorded), which are recommended for exploring new and complex issues (Barriball & While, 1994). For the interviews, we selected a variety of experts in CP (e.g. researchers, funds’ applicants, consulting companies supporting potential beneficiaries, LMA officers, CEOs, journalists). The interviewees were identified within the context of PERCEIVE, a European research project on CP, in which the Emilia-Romagna region was also involved. This allowed us to access stakeholders and information at diverse levels of CP implementation. The aim of the interviews was to shed light on the LMA and beneficiaries’ priorities, operations, and practical concerns, which are elements that are not usually disclosed in reports, official documents, and literature. The interviews were continued until theoretical saturation was achieved (i.e. till additional interviews do not deliver any substantial insights; Eisenhardt, 1989). Second, two audio-recorded workshops were conducted following participatory modelling principles (Vennix, 1996). Academicians, practitioners, EU officials, and representatives of other LMAs from other European countries attended the workshops (all involved in the PERCEIVE research project). The workshops served to expand and corroborate data and information previously collected. Specifically, workshop participants were given printed versions of the model developed up to that point in time and asked to comment and elaborate on the model’s structure (i.e. on its cause–effect relationships). This process of shared knowledge co-creation was important (Vennix, 1996) as it went beyond the knowledge gained through the interviews and rendered the elaboration of additional constructs and relationships, as well as the validation of the developed structure, possible.

Then, to increase the reliability of the empirical information collected through the field research and assess its coherence with existent knowledge, the data collected were then integrated and triangulated with the extant literature to structure the CP operations, decisions, and other relevant functioning elements (Eisenhardt, 1989). Specifically, this procedure helped to connect and harmonise the developed set of operational micro-variables with the existing macro-constructs. Lastly, the data were utilised to create our qualitative causal model of CP implementation that includes a set of variables and a nexus of interconnections among variables.

To increase our confidence in the qualitative model of CP implementation, we have built a simulation model to apply computational analysis. The use of simulation aims at two goals. First, a computational representation of the underlying cause–effect logic of the qualitative model enables the elaboration of simple, yet logically precise, and comprehensive theorisation from inductive field research. This improves internal validity in the description of longitudinal, nonlinear, and complex phenomena (Davis et al., 2007), such as CP (Cunico et al., 2021; Zahariadis, 2013). Secondly, comparing the model simulations with the observed macro-behaviour of absorption renders points of contact between our model and the empirical pattern under study, thereby enhancing the opportunity to eventually falsify or validate our theorisation (Barlas, 1996; Sterman, 2000). The more our description of decision processes, which is grounded in field research, produces a simulated absorption behaviour that is similar to the empirically observed one, the more we increase the confidence that the model is a good candidate to represent the phenomenon. To build the simulation model, we employed the SD approach.

SD is an established methodology to study the nonlinear behaviour of complex systems over time (Sterman, 2000), including socio-economic ones (Yücel & Chiong Meza, 2008), such as the CP absorption trends. In particular, SD articulates the problem under study into a system map comprising cause–effect relationships, stock and flow variables, and feedback loops (Sterman, 2000). Then, the map is formalised and quantified into a set of differential equations to simulate the developed model (Sterman, 2000; Sticha & Axelrad, 2016). In this way, all relationships in the model are supported by quotes, real-world data, and/or literature (Kim & Andersen, 2012; Kopainsky & Luna-Reyes, 2008).

Simulation is essential for controlling the behaviour of complex systems to significantly reduce time and costs related to the process of experimental verifications (Samara et al., 2022). Thus, the computational model allows us to simulate how the constructs in the model affect regional absorption patterns. In contrast to other quantitative methods, SD allows for the integration of qualitative data in the analysis, further filling the gap between CP implementation and evaluation (Hoerner & Stephenson, 2012). Hence, SD contributes to the problem’s comprehension from an operational and dynamic perspective (Mowlanapour et al., 2021; Sterman, 2000). In this respect, SD has established its potential to serve as a bridge between theory and practice (Olaya, 2015) and explore the micro-components and foundations of systemic macro-behaviours (Fiddaman, 2007). The model was calibrated using official data, experts’ educated guesses, and specific quantification which were provided by our research conducted in the Emilia-Romagna region. This was essential for testing the model outputs against the actual absorption trends of Emilia-Romagna (Appendix B) to peform the aforementioned formal validation of the model’s reliability and ability to capture real-world dynamics (Barlas, 1996; Sterman, 2000). Further validation tests have been conducted to increase the confidence in the model and its outputs, namely, parameter confirmation tests, dimensional consistency, formal inspections, walkthroughs, extreme conditions tests, behaviour sensitivity tests, and modified-behaviour predictions (Barlas, 1996; Sterman, 2000). To guarantee transparency and promote replicability, the complete mathematical model (accessible with Vensim software) is provided in the supplementary material, along with the related documentation as described by Martinez-Moyano (2012) and Sterman and Rahmandad (2012).

Finally, it is essential to transparently outline the epistemological stances that this work adopts. In line with the usual applications of SD (Lane, 1994, 1999, 2001; Lane & Oliva, 1998; Schwaninger, 2004), this research adopts a principally functionalist view of the social world (Burrell & Morgan, 1979). Namely, it assumes that reality can be objectively viewed; as models are simplifications of the world (Sterman, 2002), it is important to develop a model that is sufficiently useful to provide valid insights on a specific issue (Sterman, 2002). Model validity stems from formal tests and constitutes a continuous process to build confidence in models’ results (Lane, 1999).

Modelling Cohesion Policy Implementation

In the SD language, the arrows link causes with their effects. A positive arrow polarity implies that an increase (decrease) in a cause results in an increase (decreases) in the effect. In contrast, a negative polarity indicates that the effect decreases (increases) when the cause increases (decreases). A series of cause–effect relationships in a closed sequence generate a feedback loop mechanism. Reinforcing loops lead to exponential growth patterns, while balancing loops produce goal-seeking behaviours, driving a system towards a self-balancing state.

Conceptual System Map

To describe the model, we present the key causal dynamics and feedback loops in the CP system map (Fig. 1). A central element of the system is the main pipeline of the CP implementation (black links) (Bachtler et al., 2014; Cunico et al., 2021; Zaman & Cristea, 2011), through which the structural funds are processed, absorbed, and transformed into local projects. After the CP budget and regulations have been approved, before an LMA can initiate the management of CP funds, a number of necessary institutional steps need to be completed (Milio, 2007). Namely, the LMA operational programme (OP) approval (which refers to the detailed plan through which an LMA or a member state spends the allocated funds) (European Court of Auditors, 2018; Georgescu, 2008) and the member state–EU partnership agreement need to be signed (European Court of Auditors, 2018; Piattoni & Polverari, 2016). These steps require a considerable amount of time (“delays before LMA starts managing EU funds”). Then, funds (“total funds available”) are allocated by the LMA (“funds allocated in calls”) by issuing calls to which potential beneficiaries can apply based on their available resources (“applications”) (Wostner, 2008). Their applications are then processed. Specifically, they are evaluated, and, if the assessment is satisfactory and funds are available, an official contract is signed (“approved projects”). The duration of these processes depends on the “LMAs administrative capacity”; the more capable the LMA is, the faster it allocates the funds in the calls and processes the submitted applications. As long as the projects are implemented and completed (“projects completed”), they are refunded increasing the funds’ absorption rate that is calculated based on the “total funds available”. The role of the LMA is crucial again in terms of the time needed to perform tasks (Milio, 2007; Tosun, 2014). In this case, if the LMA is efficient, it may need less time to monitor and assess the completed projects, certify the project’s expenditures, and reimburse the beneficiaries, thus raising the absorption rate.

Fig. 1
figure 1

System map of Cohesion Policy implementation (links: black, main pipeline; green, internal LMA operations; pink, external factors; blue, beneficiaries’ dynamics; orange, funds’ absorption; red, LMA policies. Zones: blue, potential beneficiaries’ demand; green, administrative capacity (supply); yellow, financial absorption capacity (supply)). For colours, please refer to the online version of the manuscript

Full size image

LMA administrative capacity (Milio, 2007; Tosun, 2014) is affected by four operational dimensions: “LMA management stability” (Gandolfo, 2014), “LMA equipment and capabilities” (Bachtler et al., 2014), “LMA staff” (Jaliu & Rǎdulescu, 2013), and “LMA staff skills” (Jaliu & Rǎdulescu, 2013; Zaman & Cristea, 2011). While the first variable is external to the LMA organisation control (pink link), the last three are under the control of the LMA management (green links). Interestingly, not only do the completed projects increase the absorption rate, but they also tend to increase the LMA’s and beneficiaries’ capacity due to the accumulation of experience, generating two important learning loops (Bachtler et al., 2014). Specifically, the LMA staff are expected to improve their skills, and subsequently the LMA’s administrative capacity, and perform more efficiently CP-specific bureaucratic duties, thus being faster in processing the various implementation pipeline phases (i.e. “funds allocated in calls”, “approved projects”, “funds absorption rate”) (green links) (Bateira & Ferreira, 2002). Moreover, LMAs with improved administrative capacity can draft better calls and offer enhanced support to potential beneficiaries willing to apply (i.e. more technical assistance and wider spread of information about funding opportunities), thus increasing the overall number of “applications” to a specific call. At the same time, the completion of a significant number of projects can trigger a word-of-mouth effect, through which more potential beneficiaries are informed about CP opportunities, and lead to a pool of potential applicants with improved CP skills due to the learning effect. Both mechanisms are expected to increase the number of “applications” since more aware beneficiaries may consider a project proposal submission as simpler and more economical (blue links). Additionally, given the improved abilities, it is plausible to expect that the quality of the submitted proposals increases, leading, in turn, to a potentially higher number of approved projects. However, the described LMA and applicants’ learning loops can be interrupted if operational discontinuities occur (i.e. in case of changes in regulations and procedures, part of the set of skills already developed might not be meaningful anymore and new abilities and knowledge might be needed).

The absorption gap can be caused by a low “funds absorption rate” generated by a poor “LMA administrative capacity” (supply) or an insufficient rate of “applications” (demand) that satisfy the calls’ supply (i.e. part of the calls may remain unfulfilled). If the LMA identifies discrepancies between the actual and the ideal absorption rate (“absorption gap”) (orange links), it might opt for actions to correct this gap (red links) (Appendix C). Better communication activities to increase awareness of funds’ opportunities among potential applicants, or an increase in the number and pace of calls for funds issued, could foster demand and improve the absorption rate. These actions are considered low-demanding policies as the LMA can implement them quite autonomously. Alternatively, an extension in the calls’ scope may raise the number of potential applicants that are interested in submitting a proposal. This policy tends to require more LMA effort since a revised call might need to receive local government approval. However, these actions tend to address only minor absorption gaps. In extreme cases, when the absorption rate is significantly below the targets, the LMA may attempt to put in action more drastic strategies to increase the absorption rate, demanding much more effort as not only do they require the intervention of the local government but also of additional multiple actors (e.g. national government, EU, audit authorities). These strategies refer to regional co-finance reduction, retrospective projects’ use, and lower acceptance standards (Corte dei Conti, 2017; European Court of Auditors, 2018). The first strategy entails a controlled reduction of the local and/or national share of contribution to the structural funds to increase the percentage of EU funds’ absorbed (Aivazidou et al., 2020). Retrospective projects are those already accepted and initiated under other national or local financial schemes and then moved within the CP framework (Aivazidou et al., 2020). As these projects are already at a higher level of development, this policy assists LMAs in increasing the amount of funds absorbed rapidly. Finally, lowering the evaluation standards could increase the applications’ acceptance rate in case there is a significant number of applications rejected due to low-quality level (Cunico et al., 2020).

Moreover, we identified four external factors affecting implementation operations absorption rate (pink links). Firstly, in some regions, absorption decelerates (or even halts) for a certain period (Aivazidou et al., 2020), indicating that the necessary conditions for processing the structural funds might be missing. In particular, a decreased (or even static) rate could be caused by a low political efficiency (“political efficiency and stability”), namely, a low (or even zero) temporary capacity of the political system to provide directions and fulfil its duties (i.e. defining and approving priorities and regulations for issuing calls) (Gandolfo, 2014; Milio, 2007, 2008). Secondly, the public national and/or regional authorities are expected to co-finance OP; if they are unable to obtain the necessary funds (“OP co-finance availability”) (European Commission, 2013; Jurevičienė & Pileckaitė, 2013), then the absorption pace can be slowed down (or even stopped). Thirdly, “LMA management stability” may be problematic in case an LMA manager has not been appointed or because changes occur quite often (e.g. highly skilled staff turnover), not offering managers enough time to learn how to perform their administrative duties (Bachtler et al., 2014; Hagemann, 2019; Zaman & Cristea, 2011). In this situation, it may also be challenging for the staff to adapt to the new conditions and complete regular procedures. Finally, operational programmes may not recognise the needs of local communities, and the calls in which locals may not be interested (i.e. poor “OP quality”, which refers to the capacity of the developed OP to match the local needs and drive demand towards the planned objectives). In this case, the pool of potential beneficiaries interested in applying for funds may become narrower and, thus, negatively affect the number of applications for funds, rendering, for example, the expansion of the calls’ scope necessary.

Furthermore, we highlighted specific areas of the system map to connect conceptually the model with the high-level theory on CP demand and supply (in the second section). The blue zone represents the demand side of the absorption capacity, while the green and yellow zones refer to the supply side. More specifically, the green areas indicate the administrative capacity of the CP system. The yellow area focuses on the financial absorption capacity. Given the emphasis on the LMA analysis level, the macroeconomic capacity, which constitutes a supranational political criterion adopted to define the maximum amount of funds to be allocated, is considered out of the model’s boundaries. In the map, demand and supply meet where beneficiaries interact with LMAs (i.e. approved/completed projects’ processes). Similarly, financial absorption capacity, which refers to the OP co-finance, affects the ability of potential applicants to find economic resources to co-finance their project; hence, the yellow and blue zones partially overlap.

Zooming into Demand

As already mentioned, although literature emphasises the supply side of absorption capacity, knowledge about the mechanisms underpinning the demand for CP funds is limited. Thus, mobilising the empirical evidence collected through our field research and triangulating it with the information described in the literature (Jaliu & Rǎdulescu, 2013; Jurevičienė & Pileckaitė, 2013), this subsection focuses on this latter component of the system. Specifically, the variable “applications” is deconstructed to elucidate the decision path that transforms potential beneficiaries’ demand into actual applications submitted (Fig. 2). To model the decision process, a basic decision analysis approach has been used (Goodwin & Wright, 2004). In more detail, the total pool of potential beneficiaries (“demand potential”) includes private and public organisations that may also cooperate through public institutional coordination or public–private partnerships. To apply for funding, we argue that potential beneficiaries may pass through several decision steps; firstly, they have to be informed about the opportunity of financing (“opportunity awareness”) and to be interested in the call (“interest in opportunity”). Then, they should decide if it is financially viable to apply (“cost efficiency”). Indicatively, they should consider whether (i) the cost of preparing an application is affordable, (ii) they can raise their share of project co-finance at a cost acceptable for them, and (iii) potential delays in receiving reimbursement once the project is completed are not discouraging. Concerning costs, research has shown how micro and small firms, which constitute the entrepreneurial texture in several European regions, tend, in practice, to rely on their own funds and may be more reluctant than larger companies to seek financing outside their usual practices (e.g. family funds, micro-loans), independently of their economic convenience (Daskalakis et al., 2013). In this light, the agents in the cost evaluation phase (and potentially in the whole process of funds’ application) should be treated as boundedly rational ones (Cyert & March, 1963; Gavetti & Rivkin, 2007). Finally, organisations should also be willing to take the risk of applying and potentially go through the CP implementation process (“risk acceptance”). It should be noted that these steps imply that applicants possess the necessary skills to apply for CP funds and manage the related projects.

Fig. 2
figure 2

Potential beneficiaries’ decision steps

Full size image

Scenario Analysis

Identifying and quantifying the operational components of CP implementation allow to simulate diverse scenarios for assessing their impact on the system’s macro-behaviours in a controlled environment. This section reports the simulation outcomes of diverse supply-and-demand scenarios and their explorations. Particularly, we focus on specific issues that have been highlighted as fruitful directions for research (Bachtler, 2013) and policy action (Aiello et al., 2019a; European Parliament, 2011), namely, (i) the organisational and procedural impact on absorption, (ii) the role of LMA administrative capacity and skills’ learning and the significance of LMA adaptability, and (iii) the effects and responses to low demand scenarios.Footnote 1 Given that previous policy cycles (2000–2006, 2007–2013, 2014–2020) were used to calibrate the model (Appendix B), all “what if…?” scenarios are simulated for the current policy cycle (2021–2027), retaining the previous cycles’ assumptions.

Supply-Oriented Scenarios

We first explore how the structure and management of the implementation process affect the absorption pattern. To this end, we modify the time that the LMAs need to process the project applications to simulate increased delays (i.e. higher average evaluation time) in the implementation pipeline. Compared to the baseline case (blue line: 2.5 months), we simulate three alternative scenarios (red line: 5 months; green line: 7.5 months; orange line: 1 year). Figure 3 reports the absorption rate behaviours for the different evaluation times. When the time required to complete a step in the funding pipeline increases, this delay acts as a bottleneck reducing the absorption rate. For example, the delay could be generated due to bureaucratic variations in CP regulations that render the time needed by an LMA to assess project proposals more time-demanding. Given the importance of a regular and rapid flow of projects through the pipeline, the Commission and the national and/or regional authorities should work towards smoothening, to the best possible extent, the CP procedural steps to promote funds’ absorption.

Fig. 3
figure 3

Absorption rate (dimensionless) depending on evaluation time (blue line, 2.5 months (baseline); red line, 5 months; green line, 7.5 months; orange line, 1 year)

Full size image

In general, simulations showcase how reductions in administrative capacity are translated into higher processing times through the implementation pipeline, thereby lowering the absorption rates. In this respect, it is worth reconstructing the possible roots of this phenomenon. For example, an increased staff turnover rate may negatively affect staff skills. Indeed, experienced and motivated staff accumulate critical abilities to process the different tasks of CP implementation over time through learning. If staff members change often, this knowledge could be lost, considering that recently hired employees may require time to acquire the necessary skills. When the turnover rate is considerably high, newcomers might not even have time to learn.

Figures 4, 5, and 6 illustrate the system’s behaviour (i.e. staff skills, evaluation time, absorption rate) when simulating the aforementioned scenarios. Defining the blue line as the average turnover rate, the green line depicts a six times quicker rate, while the red line represents the same increased turnover rate along with a lower LMA ability to learn (i.e. reduced by 20% compared to the baseline to replicate the case in which experience accumulation is missing). As retiring staff members entail a loss of experience, an increase in the turnover rate leads to a considerable reduction in the staff skills’ level. This lack may subsequently lead to a rise in the time needed to complete the different CP implementation steps and, finally, a reduction in the absorption rate. Moreover, when an increased turnover rate is combined with a low LMA learning ability, the negative impact on the LMA staff skills is even amplified. Therefore, a greater LMA learning ability can partially compensate for increased staff turnover, thus limiting the skills’ loss. However, as this compensation is a time-consuming procedure, efforts should be placed on the long-term LMA commitment to increase their learning capacity for improving performance.

Fig. 4
figure 4

LMA staff skills (dimensionless) depending on staff turnover rate and LMA learning ability (blue line, average turnover rate (baseline); red line, increased turnover rate; green line, increased turnover rate and lower learning ability)

Full size image
Fig. 5
figure 5

Evaluation time (year) depending on staff turnover rate and LMA learning ability (blue line, average turnover rate (baseline); red line, increased turnover rate; green line, increased turnover rate and lower learning ability)

Full size image
Fig. 6
figure 6

Absorption rate (dimensionless) depending on staff turnover rate and LMA learning ability (blue line, average turnover rate (baseline); red line, increased turnover rate; green line, increased turnover rate and lower learning ability)

Full size image

In this context, learning and adaptability emerge as necessary organisational capabilities for obtaining high-performing absorption results. Therefore, it is essential to explore more precisely how these organisational features affect absorption trends. For example, legislative and operational disruptions in CP implementation (e.g. modifications in procedures to manage funds and projects among CP cycles) that occur between a previous CP cycle and a new one may render the experience previously accumulated by the staff rather useless, especially when the regulations change radically, leading to a significant loss of skills.

Exploring this direction, Fig. 7 reports the simulated behaviour of the LMA staff skills in different scenarios. More specifically, the blue line represents the baseline case where LMA adjusts to usual minimal changes among different policy cycles (i.e. the slight decrease in skills visible every 7 years). Both red and green scenarios simulate considerable differences in regulations and procedures between policy cycles 2014–2020 and 2021–2027, to the point in which 90% of old accumulated skills cannot be utilised anymore. The red line indicates an LMA with high adaptability and resilience (especially due to their learning, skills development, and training capacity) that rapidly recovers from the shock. On the contrary, the green line refers to a less adaptable LMA generated by a lower capacity to increase skills, struggling to recover after the shock. The more resilient LMA will recover more quickly than the poorly resilient one, which needs considerably more time. This loss of expertise further increases the time needed by the LMA staff to perform their tasks across the CP implementation pipeline (Fig. 8), finally affecting the funds’ absorption (Fig. 9). While the absorption rate pattern is marginally decreased in the case of the resilient LMA, leading to the same final absorption as in the baseline case, both the behaviour over time and the final level of the non-resilient LMA’s absorption are significantly lower compared to the baseline scenario. Overall, it is recommended that a certain degree of continuity in regulation and operational procedures should be maintained, while the organisational resilience of LMAs, especially in terms of skills’ improvement, should be strengthened to be prepared for any possible shock and avoid delays and low absorption. This is expected to contribute towards improving understanding of the relevance and impact of learning dynamics as demanded by Bachtler et al. (2014).

Fig. 7
figure 7

LMA staff skills (dimensionless) depending on LMA adaptability under legislative disruptions (blue line, average LMA (baseline); red line, highly resilient LMA; green line, poorly resilient LMA)

Full size image
Fig. 8
figure 8

Evaluation time (year) depending on LMA adaptability under legislative disruptions (blue line, average LMA (baseline); red line, highly resilient LMA; green line, poorly resilient LMA)

Full size image
Fig. 9
figure 9

Absorption rate (dimensionless) depending on LMA adaptability under legislative disruptions (blue line, average LMA (baseline); red line, highly resilient LMA; green line, poorly resilient LMA)

Full size image

Demand-Oriented Scenarios

Concerning the demand side, we focus on the decisions and actions of potential beneficiaries applying for funds. Specifically, we explore the factors affecting the decision process, such as the potential applicants’ awareness, their interest in the calls, their convenience to apply, the economic affordability of their contribution, and the risk of applying. We suggest that each decision step acts as a “conditio sine qua non”; if any of these steps is overlooked, then the potential applicant may not apply. To explore their impact on the CP system, a scenario in which the average interest of potential beneficiaries in the calls is decreased is simulated (Fig. 10). This may occur in case the LMAs do not adjust their proposed calls adequately based on the local needs. The baseline values (blue line) are calculated based on the model’s calibration for the Emilia-Romagna region (as a high-performing LMA), to which the average number of applications submitted per call (i.e. calls’ attractiveness) is approximately twice the available places. The red, green, and orange lines represent a decrease in attractiveness by 11%, 17%, and 33%, respectively. Results showcase how a reduction in the calls’ attractiveness could generate a decrease in the absorption rate. However, though, in the red scenario, the consequences of lower attractiveness are less visible given the initial high pool of interested organisations, this buffer becomes exhausted in the rest scenarios, leading to significantly decreased absorption trends. In general, this behaviour indicates how an obstacle in the decision process can considerably diminish the total demand, thus lessening the funds’ absorption rate. In fact, situations in which potential applicants are discouraged during the decision procedure are rather common in the literature. Indicatively, potential applicants may (i) encounter incapacities to co-finance their proposals (Georgescu, 2008; Jurevičienė & Pileckaitė, 2013; Zaman & Cristea, 2011), (ii) be unaware of the calls’ existence (Barberio et al., 2017), or (iii) consider investing time and resources to develop an application as inconvenient (Smętkowsk et al., 2018; Tatar, 2010; Wostner, 2008).

Fig. 10
figure 10

Absorption rate (dimensionless) depending on factors affecting beneficiaries’ demand (blue line, average calls’ attractiveness (baseline); red line, 11% reduction; green line, 17% reduction; orange line, 33% reduction)

Full size image

In this light, to respond to low beneficiaries’ demand, LMAs should work towards increasing applications for regional calls. In this scenario, an LMA may launch several policies rendering calls more appealing, for example, through media promotion and reduction of bureaucracy, economic barriers’ limitations, and risk. Figure 11 presents the absorption rate behaviour in response to different scenarios of policy intervention. The blue line reports the baseline case in which demand is sufficient. The black line describes a case of low demand, yet without LMA action to improve absorption, while the red line simulates a case in which the LMA acts towards increasing the promotion of LMA funds (soft policy) to tackle low demand. Both orange and green lines illustrate scenarios in which an LMA extends the scope of their calls (moderate policy) to broaden the audience of potentially interested beneficiaries. Notably, the orange line depicts a timely adaptation, while the green line refers to a delayed L MA action (e.g. the LMA may not immediately perceive that there is a demand issue). Overall, policies intervening on the features of the call (i.e. call’s scope extension) have a better impact, for example, with respect to those aimed to reinforce media promotion. Of course, even if the calls’ scope extension can be considered more efficient, risks may ensue to shifting from the original OP purposes, thereby accepting projects that do not actually correspond to the CP strategic goals (Cunico et al., 2020). In addition, given that timely actions lead to better absorption performance, LMAs should monitor with proper indicators the state of the system to intervene on time (Cunico et al., 2021). Thus, to tackle efficiently low beneficiaries’ demand in both short and medium terms, LMAs should work towards increasing their flexibility and adaptability.

Fig. 11
figure 11

Absorption rate (dimensionless) depending on beneficiaries’ demand and soft/moderate LMA policies (blue line, sufficient demand (baseline); black line, low demand and no LMA action; red line, low demand and media promotion; orange line, low demand and timely calls’ scope extension; green line, low demand and delayed calls’ scope extension)

Full size image

Discussion

As initially presented, extant literature mostly assists in identifying the macro-categories and constructs of CP implementation but often fails to capture the organisational mechanisms that enhance or hinder CP implementation at an LMA level (Smeriglio et al., 2015; Wostner, 2008). In this context, a bottom-up approach, which exposes how the interactions among the operational elements of the implementation process interact over time to generate the observed absorption patterns, is missing. This lack of knowledge is critical not only because it limits our reflections and discussions on a phenomenon of interest but also because it generates blindspots for policy-makers in the policy design process. Indicatively, Cunico et al. (2022) report how European Commission’s attempt to promote improvements of administrative capacity in LMAs with low absorption through the policy of the “decommitments” (i.e. allocated funds which are not spent by the LMAs on time are lost) could generate a number of potential undesirable side effects. Namely, as a consequence of the policy, LMAs may be pushed to use some controversial policies (Appendix C) to inflate the absorption rate in order to comply with the Commission’s requirements instead of working on improving their administrative capacity. In this respect, an in-depth understanding of the micro-mechanisms that underpin both policy functioning and emerging behavioural dynamics may intervene to prevent such shortcomings.

In this light, the central contribution of our research is to unveil the deep causal and operational structure that lies behind the observed absorption patterns. We capitalise on our field research and its formalisation through the SD language to construct and simulate a dynamic causal model (Fig. 1) that can be used to explain and interpret how the micro-elements, decisions, and operations in CP concur to generate the real-world absorption. Moreover, not only does our effort focus on the micro-aspects of absorption but also tries to tie the new insights with existing macro-constructs discussed in the literature.

Notably, analyses based on the nexus of cause–effect relationships could facilitate the understanding of systems’ processes and dynamics among policy-makers (Black, 2013), even within the public administration domain (Bianchi & Tomaselli, 2015; Cosenz, 2018). Our model (Fig. 1) provides a novel representation (“boundary object”Footnote 2; Black, 2013, p. 76) to analyse CP at an operational level. Thus, the proposed system map, where macro-theories are supported by micro-concepts and procedures, could support practitioners, policy-makers, and experts in understanding the system’s complexity and, subsequently, enhance the design of efficient funding schemes and policy interventions. In this context, our research is expected to stimulate a comprehensive discussion on CP implementation, shifting from theoretical contexts to more practical ones, to bridge supply (i.e. LMAs) and demand (i.e. potential beneficiaries) operations (e.g. Aiello et al., 2019b; Domorenok et al., 2021). Whereas research on CP demand has been minoritarian, this research highlights how important it is to focus also on the analysis of the determinants of demand for funds to achieve a more comprehensive knowledge of the CP functioning.

In more detail, the construction of an operational model allowed to quantify the CP system and perform scenario analyses. As the CP system evolves across the implementation pipeline, simulation outcomes highlight that performing tasks with adequate speed and efficiency is crucial in all steps. Notably, our findings support the positions of those arguing that regional institutions and regional development policies cannot be studied in isolation as they are part of a more complex and interdependent regional ecosystem (Lopes & Franco, 2019; Papamichail et al., 2022). In this respect, high LMA administrative capacity (e.g. technical and equipment capabilities, quality top-level management, staff skills’ capacity, and staff numbers) is essential for maximising absorption performance. The outcomes further confirm the findings of the studies that emphasise the important role of administrative capacity in public service performance (Andrews et al., 2017), especially in the EU policy context (Coremans & Meissner, 2018). In particular, simulations suggest that LMAs should be adaptable, retain accumulated skills, and promote learning to be best prepared for any challenge and/or disruption during the policy cycles. These strategies are in line with those identified in the literature for effective public administration (Farazmand, 2009). At the same time, the model’s simulations stress how each decision step of the potential beneficiaries’ decision process could affect absorption patterns and, thus, should be carefully considered by LMAs. This is not surprising, as previous research has already shown that the characteristics of the local private sector are central for regional development (Ejdemo & Örtqvist, 2021).

Lastly, this research could be framed as a contribution to the existing CP conceptual constructs and frameworks (Fig 12; solid black boxes) that typically gloss over micro-processes. The insights gained in this research could be used to further deconstruct these macro-constructs, showcasing how micro-foundations of the CP system (Fig. 12; dotted red boxes) build up on each other to constitute the macro-concepts. Specifically, to explain macro-dynamics, we emphasise how administrative capacity could be further deconstructed into management stability, institution quality, and staff capacity, while the latter may be divided into equipment, staff skills, and staff numbers. These components have already been discussed in the model development section, whereas institutional quality corresponds to the “political efficiency and stability” construct in our model (Fig. 1). This new name has been selected to abstract and generalise the construct and, further, to connect our model with existing research on quality of governance. In fact, in this case, institution quality is intended as the extent to which a public government performs their required activities in an impartial, uncorrupt, effective, continuous, and high-quality manner (Charron & Lapuente, 2013; Charron et al., 2015; Mendez & Bachtler, 2022)). At the same time, we offer a preliminary comprehension of the drivers and barriers of the decision process of potential beneficiaries applying for CP funds.

Fig. 12
figure 12

Absorption capacity framework (black area, literature evidence; red area, field research and modelling evidence)

Full size image

Conclusions

Research Contributions

The developed SD model on CP implementation crystallises the knowledge and expertise of academics, policy-makers, and involved stakeholders, which was collected through field research, dedicated focus groups, and the study of relevant literature. The model reconstructs the mechanisms at an LMA level through which CP structural funds are allocated and managed. We suggest that the model advances the knowledge of the drivers and barriers of funds’ absorption by offering an “operationalised” perspective for explaining the CP supply-and-demand dynamics. In this way, the model provides a preliminary contribution towards defining the operations, elements, and decisions (micro-foundations) that concur to explain the CP absorption rates (macro-behaviours). It should be seen as one of the first attempts to add, complement with, and merge a bottom-up perspective of the study of CP to a body of knowledge that has mostly tended to adopt macro, high-level, and top-down viewpoints. Moreover, our analysis and developed model further includes the demand side of CP in the picture, which has been often neglected, highlighting that the beneficiaries are fundamental components of the policy and overlooking them could be problematic for the researchers and practitioners who are interested in understanding and improving regions’ absorption behaviours. To this aim, based on an operational theorisation and the proposed cause–effect model, our work contributes towards supporting the implementation of CP and overcoming the current literature limitations by rendering available conceptual and practical tools to reflect on scenarios and conduct what-if analyses of funds’ absorption.

From a technical perspective, SD has been proved to be an effective method to analyse CP operations and deliver insights. The proposed modelling approach could be added to the toolbox of CP practitioners desiring to investigate practical aspects of policy implementation. Since SD has been successfully used to study regional development (e.g. Samara et al., 2022), we anticipate that our research showcases even further the benefits of this method. Overall, this work contributes to the portfolio of CP models (Bradley & Untiedt, 2007) by adding new operational and microscopic perspectives to the dominating macroeconomic approaches mainly focused on assessing CP impacts (López-Rodríguez & Faíña, 2014; Yesilkagit & Blom-Hansen, 2007).

Policy Implications

As our model identifies the LMA organisational characteristics for improving CP implementation performance, policy-makers may cross-compare the current outcomes with previous findings in public administration literature to draw effective strategies that could enhance administrative capacity (Farazmand, 2009). More specifically, our work further highlights how CP is a complex dynamic system and that any action aiming to improve absorption should be considered through this lens as neglecting them could lead to unintended side effects (Meadows, 2009; Sterman, 2000). For example, as abovementioned, Cunico et al. (2022) report how the decommitments policy implemented to foster funds absorption could end up decreasing the total amount of resources invested (other reflections about side effects within CP can be found at Davies & Polverari, 2011; Mendez & Bachtler, 2011). Therefore, in addition to the policy reflections outlined in this research, there are two major points that policy-makers should consider while deliberating over CP. First, policies should be intended as actions alterating complex systems. In this sense, policy-makers should take into consideration not only the direct impact of the actions, but also the complex dynamics that they trigger and the effects that they can generate in components of the system that are perceived to be distant or irrelevant. This type of more profound evaluation could reduce the likelihood of undesired side effects (policies enacted in social systems often produce counterintuitive outcomes; Forrester, 1971). On the contrary, the leverages of the system, intended as variables in a system where a modest shift can produce major changes (Chan et al., 2020; Meadows, 2009), should be explored to identify the policies that can deliver the greatest desired change with the lowest effort. Second, given that complex socio-technical systems tend to be inertial (e.g. Papachristos, 2014; Schill et al., 2019), they may require not only just one policy but a set of timely coordinated actions to drive CP to the desired state and behaviour. In general, the developed model could be used to explore viable synergies among alternative policies.

Ideas for Future Research

Overall, our system map assists in identifying the CP system sectors; future research should, in general, add further knowledge by focusing on the links of our model, each one considered as a unique research item. For example, capitalising on our research work, further efforts may empirically test and quantify the factors influencing potential beneficiaries’ decisions to assess their impact on the process. In addition, the developed model could be tested against different European regions to search for similarities and differences in structure and behaviour. More generally, our study renders available a repertoire of hypotheses for further empirical testing concerning the micro-foundations of absorption macro-behaviours. Indicatively, further empirical studies may envisage cross-sectional analyses that will explore the correlation among absorption rates and organisational characteristics and attitudes of LMAs in different regions, for example, by testing the impact of the investment of LMAs in calls and tenders’ promotion on absorption rates. From a broader perspective, this testing may go in the direction of further accumulating knowledge about LMA staff capabilities and skills to promote knowledge convergence and cohesion among EU regions (Erdil et al., 2022).

In general, we encourage the use of data collection processes and methods beyond the main macro-approaches, mostly used thus far to explore CP implementation, to expand, improve, or disconfirm the findings of this work. Nevertheless, we recognise that availability of empirical data is a key problem in researching CP implementation. Therefore, researchers should work towards developing more robust datasets, in particular for what concerns micro-aspects. When data are unavailable, experts’ educated guesses are used as an established approach for operative modelling (Sterman, 2000), but this should be seen as a temporary solution towards more precise calibrations; indicatively, real-data calibration of inputs could further minimise output errors.