Design of the participatory budget: how to turn citizens into process protagonists of the participatorybudget the

Purpose – In the New Public Governance (NPG) paradigm, citizens play a vital role in the decision-making of public organisations and are fundamental to aligning their expectations with service delivery. Citizen engagement could be realised in the budgeting process by adopting participatory budgeting (PB) even if previous literature on PB does not focus on this tool design issue. Therefore, this study aims to understand which PB institutional design arrangements help enhance citizen participation. Design/methodology/approach – A deductive content analysis and a fuzzy-set qualitative comparative analysis were carried out on the PB regulations of 100 Italian municipalities. Findings – The results suggest that the PB design can be elaborated in different ways that do not always guaranteetheinvolvementofcitizens.Virtuousmunicipalitiesengagecitizensfromthestartoftheprocessandinthemostrelevantdiscussionanddeliberationphases.Asimplelegislativeprovisiondoesnotguaranteearealintroductionofparticipatorygovernance. Originality/value – This study theorises citizen participation in PB and examines it through empirical evidence to define relationships between PB design arrangements and citizen engagement.


Introduction
The traditional model of democracy that has been used and followed for centuries seems to no longer be effective in accomplishing the central rationales of democratic politics that aim to develop a healthy society where people influence policies and actions (Fung and Wright, 2003). Some political choices, related to the New Public Management (NPM) trend (Hood, 1991), have been preferred over the implementation of forms of participation such as democratic involvement and responsiveness (Fung and Wright, 2003), and these are the main criticisms which resulted in a lack of consideration of stakeholders' interests and their participation (Hughes, 2003;Kisner and Vigoda-Gadot, 2017). Institutional design failings could explain lower citizen engagement since there were no processes to directly involve citizens (Fung and Wright, 2003).
This led to the introduction of the New Public Governance (NPG) (Osborne, 2006). This paradigm emphasises the role of citizens, required for governments to develop democratic strategies that could also consider the community and popular control over collective decision-making (Fung, 2015;Wiesel and Modell, 2014). These are only a few reasons why decisions (Fung and Wright, 2003). Therefore, it became necessary to renew institutional design considering the exigency to create new opportunities where public administrations can directly engage citizens in innovative ways, making them part of the discussion (Fischer, 2006). This strategy, known as participatory governance, aims to make public organisations more responsive to citizens and more effective service delivery through building participation and accountability (McGee, 2003, p. 7). This approach has the following characteristics: (1) it offers a practical orientation designed to solve relatively narrow issues (Fung and Wright, 2003); (2) it involves ordinary people who are affected by decisions, which creates bottom-up participation as citizens can provide more perspectives than PA can understand and take into account alone (Fung and Wright, 2003;Grote and Gbikpi, 2002); (3) it provides a deliberative solution generation (Fung and Wright, 2003) that aims to reach consensus based on the exchange of reasoned arguments (Brown and Dillard, 2015). Therefore, the implementation of PG can maximise the combination of the effectiveness in resource allocation with the achievement of social objectives (Cho et al., 2021;Vakulenko et al., 2020) in defining strategic decisions (Fung and Wright, 2003).
PG is justifiable since it allows citizens affected by actions to participate in the decisionmaking (Grote and Gbikpi, 2002), also regarding the budgeting processes (Ebdon and Franklin, 2006), where a participatory approach may be preferred. This choice helps create a relationship that prevents citizens from being disappointed with allocating resources (Ebdon and Franklin, 2006). In addition, citizens' active participation improves steering quality, hence involving more perspectives in the discussion and creating consensus-based on exchanging ideas (Innes and Booher, 1999). This allows people to be central in the relationship with PA (Himmelman, 1994), and consequently, trust increases (Grote and Gbikpi, 2002;Klijn et al., 2010).
In particular, given the relevance of citizen participation in budgeting processes (Ebdon, 2000), it is undoubtedly possible to cite PB. In this process, people are called upon to assume an active role, that is effective in helping the government to identify stakeholders' needs and the services to be provided (Anessi-Pessina et al., 2020). This improves the direct involvement of citizens if the participatory budget is managed so that citizens can submit proposals, explain their ideas and suggest projects. When PB is designed to accomplish the principles required by participatory governance, it creates a public arena for civic discourse and deliberation, in which public needs are discussed, and proposals are developed with involved affected people (Aleksandrov et al., 2018;Patsias et al., 2013).
Therefore, a participatory budget can increase active citizen participation and PA accountability, enhance public trust, improve legitimacy and bring policymaking closer to people's desires (Bartocci et al., 2019;Fung and Wright, 2003). In other cases, the position of citizens is limited (Rios and Rios Insua, 2008;Sintomer et al., 2013) to a merely consultative role, whereas the executive and legislative branches retain all the power (Cabannes, 2004).
Although the literature on PB has been growing exponentially in recent years, most studies have focused on its implementation phase, describing the different cases and logic that this process involves (Jung, 2021;Sintomer et al., 2008) or on institutional contexts (Patsias et al., 2013), leaving the other steps of the process aside, i.e. PB design. Therefore, this study intends to fill this gap by adopting a more practical and process-oriented perspective.

Theoretical framework
Based on Ebdon and Franklin (2006), a theoretical framework is created to analyse how PB institutional design arrangements can contribute to citizen participation. Ebdon and Franklin's model was selected because it is a comprehensive and systematic framework that includes all essential variables for developing and implementing citizen participation in the budgeting process. According to Ebdon and Franklin (2006), their model identified the IJPSM 35,3 following four elements that influence citizen participation in budgeting processes: (1) environment, (2) process design, (3) mechanisms and (4) goals and outcomes.
Looking at the environment, scholars state that the adequacy of the PB design partly depends on some contextual elements considered critical for greater participation (Ebdon, 2000;Ebdon and Franklin, 2006 ). They are the local form of government, the presence of legal requirements and the political actors involved in the process. For example, council-managed cities are more likely to include citizens and seek their input in the budgeting process. In addition, political and societal factors (Ebdon, 2000;Ebdon and Franklin, 2006) may influence the level of participation. Indeed, participation is enhanced by the larger cities and more heterogeneity in the population. Finally, another element that must be taken into account is digitalisation. Even if digital tools simplify access to information and several steps in the process, they may produce digital inequalities (Ragnedda and Muschert, 2013).
The process design is analysed by focusing on four different aspects (Ebdon and Franklin, 2006). First, timing relates to the importance of involving citizens earlier on in the process. This is relevant because citizens are less likely to influence outcomes if their input is received late in the process. The second aspect is the type of budgeting allocation, considering that citizen participation could concern resources that have already been defined and have a different nature, such as capital or operating funds. The third characteristic is the adequate selection of participants. The advantages and disadvantages of each method adopted to choose participants in a budgeting process are identified as follows; for example, open discussion with a higher number of participants is a helpful method even if community representativeness should also be ensured. The last aspect concerns the need to structure the process to identify the citizens' preferences.
The mechanism element focuses on the methods used to involve citizens: organising public meetings involves more citizens, even though they may not represent the entire community. On the other hand, focus groups could be more effective and centred on a specific subject, even if this method produces a non-representative form and influences decisions if not well-managed. In this view, citizens with an advisory role of LGs may be more successful (Ebdon and Franklin, 2006). These mechanisms have several advantages and disadvantages and may also be integrated with digital tools.
Finally, the goal and outcomes element relates to the purpose of the process. Ebdon and Franklin (2006) identified the following five outcomes: (1) informing decision-making, (2) educating participants on the budget, (3) gaining support for budget proposals, (4) influencing decision-making and (5) enhancing trust and creating a sense of community. These outputs could lead to higher or lower citizen engagement, considering the type of communication that each goal implies (e.g. one-way communication in informing decisionmaking or two-way communication creating a sense of community).
Although the environment and goal and outcomes are quite relevant in the Ebdon and Franklin model (2006), in this study, the primary attention is drawn to the process design and mechanisms elements that are related to the point of view adopted (design stage) (Jan van Helden et al., 2008). In particular, the design process has been split into three out of four characteristics identified by the original model. The first characteristic is the initial stages of citizen involvement, which details the adequate timing. The second characteristic is the type of budget dedicated to PB that reasonably reproduces the budget. The last element is defining community preferences, a broad and relevant concept in the participatory process. However, adopting Ebdon and Franklin's view, this framework is focused on if and how people are selected in this crucial phase, echoing the original idea of adequate participant selection included in the Ebdon and Franklin model (2006). Mechanisms are analysed in each process design stage. The environment characteristics are only used to contextualise the analysis, whereas the element goal and outcomes have not been included because they should be evaluated in subsequent stages (Jan van Helden et al., 2008). The framework is presented in Figure 1.
Focusing on the process design, the authors included the PB stages (initial information, intervention areas, projects proposal, evaluation phase and voting) and their peculiarities (resources volume and participation tools) in the characteristics inspired by the original model (initial stages of citizen involvement, type of budget dedicated to PB and defining community preferences). Then, to understand the possibilities and different configurations that can be adopted in each category, existing literature was analysed to describe different PB experiences (Aleksandrov et al., 2018;Allegretti and Antunes, 2014;Bartocci et al., 2016;Bassoli, 2012;Rios and Rios Insua, 2008;Sintomer et al., 2013). Therefore, a set of general categories has been identified, with different configurations depending on the underlying logic.
The first characteristic identified in the framework (initial stages of citizens' involvement) includes the three categories that determine the factor for earlier citizen involvement: initial information, intervention area and project proposals.
The first category identified was "initial information", which is how PB is initially explained and how citizens meet it. It is essential to raise awareness of the process among citizens, making the information accessible and understandable to all participants (Aleksandrov et al., 2018). As illustrated by several case studies (Allegretti and Antunes, 2014), most of the information is provided online, helping to reach more people (Sintomer et al., 2013). However, from a community-building perspective, making direct information and discussion available to citizens is essential for their empowerment. Thus, open-discussions, focus groups and public meetings could be adequate. In rare cases, PB information is provided through advertising (e.g. using brochures) (Sintomer et al., 2013). This category, stressing the PG principle of communication improvement, captures the need for earlier citizen involvement, as Ebdon and Franklin (2006) highlighted.
The second category is the "intervention area" PA may limit the themes on which citizens can submit proposals or define the specific municipalities where proposals must be submitted (Sintomer et al., 2013). In some cases, citizens are already consulted to specify areas, which increases their role in the process. Citizen engagement in this phase is coherent with the necessity of identifying people's desires and involving them as soon as possible to make their role relevant in the decision-making (Ebdon and Franklin, 2006). When people are engaged at an early stage, such as defining the intervention area, it is possible to anticipate what they are more likely to support (Verweij et al., 2013). People could be involved through live public meetings, advisory committees or surveys.
"Project proposals" (the third category) is essential for citizen engagement as they are asked to suggest ways of improving general or more specific issues (Bartocci et al., 2016). Citizens must submit or even co-design project proposals as they acquire expertise through their day-to-day experience, making them knowledgeable about what needs improvement (Sintomer et al., 2013). In some cases, PA may draw up an initial list of projects (Rios and Rios Insua, 2008). According to the PG principles, the presentation of projects by citizens, especially if they collaborate with PA, is fundamental to increase their participation since they appreciate the practical orientation of the process (Fung and Wright, 2003). Presenting their projects means citizens can adapt or complement the agenda during decision-making (Verweij et al., 2013).
The fourth category represents the original characteristic named type of budget dedicated to PB. It consists of the "volume of the resources" that PA makes available for citizen proposals and are allocable as capital or operating funds. The resources can be defined as ex ante, where citizens cannot exceed the predefined amount in their recommendations (Sintomer et al., 2013). Available resources can be defined as a percentage or a specific amount of the municipality's budget. However, resources are not usually predefined (Bartocci et al., 2016) Design of the participatory budget instead, they are defined ex-post once proposals have been submitted. The type of resources used may influence participation since not determining it could be interpreted as depending on the quality of choices and suggestions made by citizens, making them the determining criterion for selection.
The following two categories are included in the defining community preferences characteristic as they represent the main stages when people can make their views known.
The fifth category is the "evaluation phase". After being presented, proposals are evaluated and analysed. A technical committee usually carries out this assessment with the expertise to judge the recommendations and create a final list of projects voted during the final consultation phase (Allegretti and Antunes, 2014). An evaluation may also be conducted by a committee of citizens or by both citizens and technicians together. In this phase, citizens' presence may be considered a sign of representation (Ebdon and Franklin, 2006), in a typically quite technical stage. However, with an emphasis on mechanisms that can be used, the employment of multiple approaches, such as focus groups or advisory committees of both citizens and technicians, may enhance trust and build a sense of community (Ebdon and Franklin, 2006), thus reaching a deliberative solution as suggested by Fung and Wright (2003).
The sixth category relates to "voting" Voting is relevant and allows citizens to express their preferences. Therefore, most PB experiences involve proposals discussion and voting by citizens. In some cases, only people who are residents in a municipality can vote, whereas in other cases, city users may also acquire the right to vote, which ensures universal participation (Allegretti and Antunes, 2014). The age of citizens allowed to vote is another issue. The right to vote in political elections is afforded to people aged 18 and above. However, as for other public consultations, voting rights can be extended to younger people (Bassoli, 2012).
Nevertheless, some PB experiences require proposals to be voted by elected delegates only, rather than all citizens, enabling a more thorough discussion. In some forms of PB, proposals are voted directly by public administrators (Sintomer et al., 2013). Different mechanisms can be used depending on how people are selected to vote. The voting phase is essential, allowing an order of preference to be created for all the projects submitted at the start of PB, as previously discussed according to the PG principles (Fung and Wright, 2003).
The last category concerns' participation tools', namely how citizens can participate in discussions (Bartocci et al., 2016). Most debates occurred during public meetings and in similar settings, such as workshops and focus groups (Bassoli, 2012), where citizens could directly discuss proposals. However, widespread access to Internet and communication technology (ICT) tools has partially changed how people participate in this process (Kolsaker and Lee-Kelley, 2008). ICT tools may reduce the time allocation of the closing discussion, reducing the debate to the defence of individual projects rather than allowing it to stimulate a dialogue among people (Sintomer et al., 2013). Therefore, in some cases, a mixed approach was preferred.

Research design 4.1 Research method
The research question is addressed through a qualitative-quantitative approach, allowing the author to proceed to an in-depth analysis of the PB institutional arrangements used in a sample of Italian LGs, as is already done in another similar search (Covaleski et al., 2017).
The research is divided into two different steps. First, a deductive content analysis was performed (Elo and Kyng€ as, 2008;Krippendorff, 2013), considering the framework developed ( Figure 1). The second step concerns a qualitative comparative analysis (QCA), allowing a detailed study of how each condition can contribute to a particular result and how different combinations of conditions lead to the same development (Rihoux and Ragin, 2008).

IJPSM 35,3
It fits with the need to analyse various PB designs. The considered conditions are literatureinformed and based on the above-mentioned framework ( Figure 1). The authors chose the fuzzy set qualitative comparative analysis (fsQCA) approach to capture PB complexity, which overcomes an important limitation of other QCA techniques, namely the necessity to deal with binary variables (Greckhamer et al., 2018).

Sample selection and data collection
This study's sample consists of 100 Italian LGs, all of which adopted PB from 2014 to 2019. The sampling selected is purposive (Flick, 2018), particularly suitable for content analysis and QCA (Elo and Kyng€ as, 2008;Greckhamer et al., 2018). An Italian sample was chosen for various reasons. First, Italy is one of the European countries where the participatory budget tool is most widespread. After its initial implementation in Porto Alegre in 1989, a participatory budget was adopted throughout Europe, with various models achieving different purposes (Sintomer et al., 2013). Second, LGs in Italy generally suffer from a reduction in voting in elections (Cerruto, 2012) due to disenchantment with PA. The implementation of processes that enhance citizen participation is a legal requirement [1]. An existing study has already analysed the state of PB in Italy (Bartocci et al., 2016) from 1994 to 2014. Following the cited research, the selection came from online research using two leading search engines, Google and Yahoo. The keywords used were: "participatory budget regulation" and "local government". The use of online research is justified since the web is one of PA's most widely used communication channels . It is also reasonable to assume that PB, a voluntary practice, is highly publicised by LGs that adopt it (Bartocci et al., 2016). The authors collected data of PB experiences that fulfilled the study's criteria, namely implementation in the period analysed.
From the results obtained, the Italian LGs sample can be classified by geographical area and the number of inhabitants based on ISTAT [2] and the Italian Ministry of Internal Affairs [3] classifications. Considering these two criteria, the sample composition is summarised in Table 1. For the content analysis, data was retrieved from the PB regulations that are available online on the official local government websites. They were manually analysed to classify various experiences based on the framework defined a priori, ensuring trustworthiness during the process and implementing validity measures and reliability checks (Krippendorff, 2013). In this study, all authors were involved in the analysis of the participatory budget regulations to improve the objectivity of the overall process (Elo and Kyng€ as, 2008). All authors then compared results, and, in case of disagreement, a majority vote was taken. Thanks to the detailed ex ante definition of the criteria used in coding, the authors coded the LG documents homogeneously. The main problems were found for the categories "projects proposal", and "voting" since these phases have a complex structure and are thus the most difficult to define. A majority vote was taken once choices were discussed in these cases (12 for projects proposals and 9 for voting). To highlight how the findings in each condition relate to one another and compare the different participatory processes across the Italian LGs, the results obtained from the content analysis were used for the second step of the research as fsQCA.

Analysis and findings
Looking at the created framework, the considered elements are: environment, process design and mechanism.
For environment, Italian characteristics may influence the process design and, consequently, the level of participation. Firstly, Italy is composed of territorial entities in the form of Mayor-council government that does not always favour the emergence of participatory events (Sønderskov, 2019). As mentioned, Italian legislation allows LGs to introduce participatory processes, leaving a wide choice about how achieving it.
Looking at the political and social culture, there is a gap between Northern and Southern regions that produce inequalities leading to a lack of participation (Cerruto, 2012) that arose, e.g. from the digitalisation process (Ragnedda and Muschert, 2013).
An in-depth analysis of the process design and mechanisms elements is conducted through both content analysis and fsQCA while considering the primary purpose of the present study.

Deductive content analysis
The analysis starts with a deductive content analysis wherein data were assessed using various categories defined ex ante (Elo and Kyng€ as, 2008). The content was detailed by identifying different configurations (Dey, 2003). Specifically, in this study, the content analysis categories used have been defined in the framework ( Figure 1) and applied to PB regulations.
The content analysis results are shown in Table 2, which displays the absolute frequencies and percentages of the various categories.
To reach as many people as possible, most LGs provide various communication channels using mixed methodologies, live events (e.g. assemblies or public meetings) and digital tools (78% of the sample in the initial information category), which capture different types of citizens. In most cases, the community did the proposal presentation (76%). In the virtuous context, citizens elaborate on projects thanks to PA support (23%). In contrast to findings obtained from 1994 to 2014 (Bartocci et al., 2016), more than two-thirds of the sample (70%) make the volume of the resources available for PB known before the process is begun.
Once citizens have proposed their projects, in some cases (86%), they are evaluated by a technical committee to assess the economic and technological projects' feasibility. In the other cases (10%), projects are assessed by a committee of technicians and citizens who can merge similar projects and create new ones.
After the evaluation stage, all economically and technically feasible projects are presented. In 86% of cases, voting is extended to all citizens who want to participate (residents or cityusers). In 3% of the sample, the vote is by a citizens' delegation. In 77 municipalities out of 89 where citizens can vote (86% þ 3%), this right is provided to teenagers (14 or 16 years old) to involve young people in the political decision-making in their city.
Two main participation tools are used to vote during the consultation. Most voting is done on regular "election days" in polling stations (51%). However, due to different levels of digitalisation, 17% of municipalities use only ICT tools, while 32% allow both in-person and online participation.

Fuzzy-set qualitative comparative analysis
A fsQCA phase started with coding activity, assigning the variables a value between 0.0 and 1.0, depending on their degree of membership (Rihoux and Ragin, 2008). Therefore, each IJPSM 35,3 category from Figure 1 has been measured using appropriate theoretical and substantive knowledge to identify the thresholds, determining which cases can be meaningfully considered as fully in vs fully out the set of citizen engagement (Greckhamer et al., 2018).
The categories identified in the framework were divided into six conditions to develop the coding scheme. "Participation tools" was eliminated since it could not be differently weighted, and the "age" of the voters was deleted from the coding system as it is not always available. These choices were made since QCA researchers suggest combining substitutable conditions because too many variables can pose a unique problem. Therefore, the six resulting conditions are: (1) Initial Information, (2) Intervention Area, (3) Projects Proposal, (4) Volume of the Resources, (5) Evaluation Phase and (6) Voting. The coding value is assigned considering citizen involvement. (0) was assigned in the absence of citizen participation, and (1) was assigned in cases where citizens are fully involved and interact with PA. Some

Content analysis Categories
No of LG % Design of the participatory budget eventual middle conditions have been coded considering partial citizen involvement (0.6 or 0.75) or no involvement (0.25). Finally, some conditions, whose effect may be ambiguous, have been considered neutral (0.5). The coding scheme with the assigned coding values is presented in Table 3, whereas the LGs scores are shown in Table A1. Authors compare the different LGs results using fsQCA software (Ragin and Davey, 2016). They construct a truth table, a matrix of conditions and their occurrences in the sample (Rihoux and Ragin, 2008). Using this software minimises data and identifies the simplest set of conditions that can account for the selected outcome (Ragin and Davey, 2016).

Initial information
According to literature (Rihoux and Ragin, 2008;Greckhamer et al., 2018), the authors chose to analyse both conditions linked to the outcome's presence and absence and select a consistency threshold higher than 0.75 and near to 1 (0.95). Table 4 represents the truth table considering the presence of the outcome.
The results show the parsimonious solution representing the minimal sets of conditions in the truth table with perfect consistency. These two combinations of states are (INT_AREA * VOT) OR (EV * VOT).
In plain language, according to the sample data, the equation states that full citizen engagement is achievable in two cases: (1) When LGs involved citizens from the beginning of the process, collaborating with them from the outset to define the issues to be focused on in PB and then providing them with the task of expressing a final order of preference on the projects to be realised.
(2) When citizens are involved in the evaluation phase. This is not common since this phase is quite technical. After the evaluation, citizens have to express a final decision about the projects to be realised.
The results presented in Table 4 show that the overall solution had a consistency level of 0.88, meaning that the degree to which the combination of conditions guaranteed the correct results was relatively high. Furthermore, the solution had high coverage of 0.76, which means that almost 76% of the cases with a high level of citizen engagement showed these two combinations of conditions. The results regarding the conditions linked to the absence of the outcome are shown in Table 5.
The findings show the parsimonious solutions that are: (int_area * vr * ev) OR (ev * vot). These equations state that citizen engagement is not achievable when LGs: (1) Avoid involving citizens both in the first decisions on the intervention areas and the definition of resources, followed by a technical evaluation made by only technicians; (2) Preclude the citizens from any form of participation in decision-making and deliberation phases.
The overall solution had a high consistency level of 0.99 and a coverage level of 0.67. Almost 67% of the cases with a low citizen engagement showed one of these two combinations.

Discussion
The analyses showed that Italian LGs adopt a moderately diversified design where both similarities and differences can be found and have different impacts on citizen engagement. From the QCA, the initial information has never appeared as a determining condition for the success or failure of involving citizens. In fact, 78% of the sample provides this information online and through public assemblies and workshops, leading to greater participation and accessibility. The proposals of projects by citizens have not been a IJPSM 35,3 determining condition because most LGs allowed citizens to propose their projects and influence resource allocation, either alone or in cooperation with PA. However, considering the tables of truth (Tables 4 and 5), it can be observed that the predefinition of the projects to vote for made by PA is present in only one city (Castellanza) where PB is organised by involving citizens in the choice of the project to be implemented from a shortlist of proposals, indicated mainly by the PA, with only eventual additions made by the citizens.
Looking at Table 4, the combinations of conditions that are more likely to produce citizen engagement are identified. The first solution emphasises an earlier involvement of citizens whose opinions are collected to define the intervention areas. A clear example of this is Pavia which organises an initial guideline review phase with citizens to define the "rules of the game" together and propose general ideas and issues. However, when Table 2 is considered, in a large part of the sample (72%), intervention areas are defined by PA for several reasons. Firstly, LGs in Italy must plan and schedule their activities in a strategic document voted by  Capital letters are used to identify the presence of a condition, whereas lowercase letters are used to identify the absence of a condition  Capital letters are used to identify the presence of a condition, whereas lowercase letters are used to identify the absence of a condition . For this reason, implemented projects must comply with this document, which is why the City Council's identification of the intervention area is a widespread practice. Secondly, it may also be a desire to limit discussion to suitable and appropriate proposals. The first solution also required that the right to vote be provided to citizens to achieve citizen engagement. The citizenry is called to participate in most cases (65 residents, 21 city-users). City users are mainly involved in bigger cities (e.g. Rome, Milan and Bologna), where university students and other categories can typically vote. When citizens are engaged in the voting stage (a total of 89 LGs composed of 65 residents, 21 city users and three delegates), the right to vote is usually given to teenagers to involve young people in political decision-making in their city. In a few cases (e.g. Verbania), a parallel participatory budget edition was organised for younger citizens (from 15 to 25 years old) to improve their engagement in local political life.
When the other combinations of conditions that could lead to citizen engagement are taken into account, a relevant issue is related to the involvement of citizens in the evaluation phase. This happens in several cities, such as Pavia, Parma, Milano and Bologna, which present a co-design phase. In this context, interested citizens and representatives of proposals assessed as feasible/implementable better define their content, technical, regulatory and budget constraints in constant interaction with the relevant public administrators' (PB regulation in Parma). This co-design process can be primarily implemented when it is possible to merge "similar ideas into a single project, discussing them with the proposers" (PB regulation in Rescaldina). As in the previous combination of conditions, in these cases also, citizens are involved in the voting phase, guaranteeing them the possibility of defining an order of preference. These two results align with participatory governance's principles, considering that people have been involved in the deliberation process and allowing them, as the primary beneficiaries of these choices, to focus on themes chosen by themselves that are closer to them. The logic of community-building emerges from this (Bartocci et al., 2019).
Considering the conditions that limit citizen engagement (Table 5), the absence of citizens involved in the crucial phases of discussion and deliberation (i.e. the evaluation and voting phases, according to Figure 1) strongly influences the non-achievement of citizen engagement. These lacks are incompatible with PG principles, for which a deliberative solution is essential. However, while the absence of citizen involvement is not too widespread in the voting phase, such as in Carloforte (Sardinia), where the City Council approves the list of the projects that must be implemented, the lack of citizen engagement in the evaluation phase is more common, considered as the most technical step in the process. Hence, specific skills and expertise are required, which are uncommon among citizens. For example, in a few LGs (e.g. Palazzolo sull'Oglio), projects must be presented consistently with the United Nations 2030 Agenda for Sustainable Development. Therefore, in these cases, the evaluation phase focuses not only on technical and financial aspects but also on the sustainable impact of proposed projects. The other combination of conditions suggests that citizens' noninvolvement is related to their lack of influence on decisions at the earliest stages of the process. The impossibility of defining the areas of intervention and the volume of the resources, together with the lack of involvement in the evaluation phase, demonstrate a substantial delay in citizen participation that precludes decisiveness in the resource allocation process, leading to a lower willingness to participate (Ebdon and Franklin, 2006).
A final remark could be dedicated to the participation tools. As seen in 17% of LGs analysed, a fundamental way of improving participation is to use ICT technologies that promote citizens' active involvement and strengthen accountability and transparency, which is essential for this type of PA (Br as and Dowley, 2021). However, in 51% of LGs, collaboration is also stimulated by public meetings and workshops, where people can freely discuss projects. On these occasions, people who are not familiar with ICT technologies can IJPSM 35,3 also participate in the process. In this sense, combining these methods might ensure greater involvement and avoid the negative effects of digital inequalities (Ragnedda and Muschert, 2013).
Therefore, even if this study identifies some standard features that do not influence the differences in citizen participation, some differences that make it possible to bring out the best practices are also observed. These are mainly related to introducing co-design and coevaluation phases for proposals, guaranteeing earlier citizen involvement to have a better chance of influencing resource allocation.

Conclusion
With the advent of NPG, citizens assume a vital role in the decision-making of public organisations (Osborne, 2006). From the 2000s, public sector governance moved from a strategy focused on outputs to one more aimed at process outcomes to align better citizens' expectations and service delivery (Hughes, 2003;Kisner and Vigoda-Gadot, 2017). This idea leads PA to adopt a participatory governance strategy (Fung and Wright, 2003). Therefore, in this scenario, the engagement of citizens becomes fundamental, especially in the budgeting process (Ebdon and Franklin, 2006). Precisely, a participatory budget, which can engage citizens in the political life of the PA (Anessi-Pessina et al., 2020), aligns citizens' expectations with the implementation of effective policies (Allegretti and Antunes, 2014;Sintomer et al., 2013).
When PB literature is studied (Aleksandrov et al., 2018;Bartocci et al., 2019), it can be seen that a gap exists, as no studies are aimed at understanding how designed institutional arrangements can improve the levels of citizen involvement. Therefore, to understand which designed features are useful in enhancing citizen participation, this study focuses on adopting a participatory budget in Italian LGs. In this context, the existing legal provision (Legislative Decree 267/2000) could influence the adoption of PB that could be viewed as a way to accomplish legal requirements. This adoption could occur since the PB design is often the result of political compromises and leaders' dominant intentions (Kuruppu et al., 2016) that could culminate in organising a participatory budget monologically, limiting citizen empowerment (Aleksandrov et al., 2018). Therefore, it is essential to examine PB design in Italy to understand which elements increase the citizenry's involvement rather than being designed in a merely performative manner to comply with a general requirement without any real interest in fostering collaboration (Cohen et al., 2019).
The results show a complex picture of PB design in Italy, characterised by similarities and differences. Virtuous cases (see, e.g. Pavia, Parma and Bologna) have structured a process that engaged citizens early to enable their participation to weigh in on allocation choices. This has been done by involving people in defining the "rules of the game" and then giving them the possibility of participating in the primary phases, with deliberative processes based on close collaboration with LGs according to PG principles (Fung and Wright, 2003).
On the other hand, the lack of citizens involved in the preliminary stages, neglecting the adequate timing that enables them to influence decisions (Ebdon and Franklin, 2006), or in both the phases that represent the heart of the deliberative and decision-making (i.e. evaluation and voting) led to an adoption of the Participatory Budget which has nothing to do with a process aimed at recreating a close relationship between citizens and PA.
Answering the RQ, this study identifies early citizen involvement in the definition of the area of interventions combined with the possibility they can vote or absolute citizenry participation in the deliberative and decision-making stages (i.e. evaluation and voting) as conditions leading to citizen engagement.
From the findings analysis, there are virtuous examples of PB, while there are also cases in which participatory budget is used as a mere instrument to satisfy legislation. However, the Design of the participatory budget latter aspect, which does not seem to satisfy expectations fully, must also be seen as an opportunity, as compliance with legislation in the public sector has sometimes been one of the main stimuli for administrations to adopt innovative tools which, otherwise, would not have been implemented spontaneously (Grandis and Paoloni, 2007).

Value, limitations and future lines of research
This study shows that adopting PB does not guarantee a relationship with citizens and that some of its configurations may enhance citizen engagement more than others. Therefore, this study contributes to the current debate about the participatory budget as a participatory governance implementation tool (Baiocchi, 2003;Patsias et al., 2013), adding a new point of view focused on PB design stages and their peculiarities. Regarding practical implications, this research provides an in-depth analysis of the different configurations of PB, with some practical examples. Therefore, this investigation could be helpful in assisting politicians to decide on how to design and develop the entire process to improve trust. While compliance with the law explains PB implementation, the willingness of LGs to draw on the full potential of this tool makes any study seeking to define some "best practices" beneficial for practitioners. This paper presents some limitations linked to the authors' choice to focus on the design stage (Jan van Helden et al., 2008). In fact, this analysis neglects all the following stages, which may reduce the accuracy of the investigation in identifying the participatory budget's "best practices". A further limitation of this study concerns the decision about which participation tools to use, given all the advantages and disadvantages of each means chosen, and therefore remains partly unexplored in this work. In particular, the impact of digitalisation in PB could be addressed in future research. Another aspect of being further developed is focusing on some environmental elements (i.e. politicians, public managers or consultants) that can promote the implementation of PB initiatives. Finally, to overcome the short-term period analysed, it would be possible to examine more longitudinal case studies in Italian local governments with a long PB experience, such as Grottammare, or relatively new experiences, such as the ones here analysed.

Design of the participatory budget
LGs