1 Introduction
Public sector accounting (PSA) reforms are complex and challenging (Polzer et al., 2019; van Helden, 2016). This is even more nuanced for local governments (LGs), where administrative routines are deeply embedded, technical capacity is uneven, and reforms often arrive as top-down mandates from national bodies (Brusca & Montesinos, 2013; Manes Rossi & Aversano, 2015). PSA reforms are change processes that require not only new policies and regulations but also shifts in organizational routines, behaviors, and mindsets (Argento et al., 2018; Caruana et al., 2019). This is particularly true for accrual accounting reforms based on the International Public Sector Accounting Standards (IPSAS), which represent a substantial change from cash-based to accrual accounting (Bergmann et al., 2019; Christiaens et al., 2010; IPSASB, 2014). Numerous factors influence the success of such reforms, ranging from accounting traditions and administrative cultures (Schuler et al., 2023) to international pressures (Argento et al., 2018) and fiscal constraints (Fuchs, 2017). However, one critical and often underestimated factor stands out – the role of individuals enacting these changes (Argento et al., 2018; Hyvönen et al., 2012).
Institutional entrepreneurship (IE) offers a particularly suitable lens for studying PSA reforms. It addresses a fundamental question: how can individuals drive change within constraining institutional structures where reforms are mandated but implementation depends on local actors (Argento et al., 2018)? Unlike general change management approaches, IE explicitly theorizes how embedded actors can transform institutions while being shaped by them (Battilana & D’Aunno, 2009). Drawing on IE, this paper examines how individual actors shape national PSA reforms in Italian local governments. Institutional entrepreneurs are characterized as agents of institutional change (Pacheco et al., 2010) or, as Battilana et al. (2009, p. 70) define them, “change agents who, whether or not they initially intended to change their institutional environment, initiate, and actively participate in the implementation of changes that diverge from existing institutions.” They are motivated by factors such as functional pressure or political developments that make current practices or institutions obsolete (Leca et al., 2008; Oliver, 1992) and their curiosity helps them identify opportunities for change within their field (Hardy & Maguire, 2017). IEs therefore play a prominent role in reform processes, as they can foster the adoption of institutional change (Salignac et al., 2018) and promote acceptance of new practices (Argento et al., 2024).
While institutional entrepreneurship has been widely applied to studying technological standards and new practices (Greenwood & Suddaby, 2006; Maguire et al., 2004; Pacheco et al., 2010), little attention has been paid to PSA reforms as sites of institutional entrepreneurship. This is surprising since IPSAS-based reforms are not merely technical but represent fundamental changes in how public organizations operate and are held accountable (Bergmann et al., 2019; Fuchs, 2017; Gourfinkel, 2021; Manes-Rossi et al., 2016).
Implementation of the ITAS (ITalian Accounting Standards) reform at the subnational level exemplifies institutional change by harmonizing heterogeneous accounting practices into a unified accrual regime and challenging existing routines (Italian Government: Presidency of the Council of Ministers, 2025; Mussari et al., 2021). This represents a substantial gap in the literature, as most IE research focuses on the national level, with little attention to local government (e.g., Argento et al., 2018; Lockett et al., 2012; Xu et al., 2021). This article therefore examines institutional entrepreneurs in Italian LGs, where national reforms face implementation barriers including resource scarcity, limited training, and organizational resistance (Benfante et al., 2024; Fabbrini, 2022). Italian LGs offer a fascinating research field, possessing high autonomy and administrative diversity (Benfante et al., 2024; Steccolini, 2004). In this fragmented environment, national reform goals cannot succeed through legislation alone – local initiative becomes crucial (Mussari & Sorrentino, 2017). Institutional entrepreneurs can therefore interpret, adapt, and promote reform goals while playing a decisive role in local implementation (Argento et al., 2018; Hoogstraaten et al., 2020).
A growing body of literature highlights the importance of IEs in initiating, sustaining, and legitimizing institutional change (Hoogstraaten et al., 2020; Qureshi et al., 2016) and their central role in change processes (Lockett et al., 2012; Mahzouni, 2019; Salonen et al., 2024). In a recent study, Argento et al. (2024, p. 2) found that “to overcome status quo hindrances, especially in the early stages of a change process, the actions of IEs can be crucial.” Given their central role, identifying potential institutional entrepreneurs is both theoretically interesting and practically relevant (Argento et al., 2024; Mahzouni, 2019; Salonen et al., 2024). As Battilana (2006, p. 671) puts it, “Highlighting the profile of individuals who are more likely to act as institutional entrepreneurs will help […] governments to identify the people on whom they can rely locally when they need to implement major changes.”
However, most scholars focus on organizational conditions that foster institutional entrepreneurship (e.g., enabling structures, power asymmetries, and field position) (Hoogstraaten et al., 2020; Pacheco et al., 2010). While this research is highly relevant, little is known about IE characteristics at the individual level, particularly in local government.
This paper addresses this gap by shifting attention from organizational structures to actor-level attributes. The research question guiding this study is: What individual-level characteristics distinguish local government officials who take initiative in PSA reform implementation from those who do not? Building on the theoretical premise that institutional entrepreneurs play a fundamental role in successfully implementing PSA reforms, it explores how such players can be characterized and identified. Specifically, we operationalize three key IE characteristics – agency & motivation, adaptability & change readiness, and legitimacy – as composite survey indices and test their relationship to initiative-taking behavior among local government officials implementing Italy’s ITAS reform. We use ‘agency’ in its sociological and psychological sense, referring to an individual’s capacity to act independently and effect change (Bandura, 2006, 2018; Emirbayer & Mische, 1998), rather than in its organizational or principal-agent connotations often used in public management. We find that agency and motivation is the most decisive predictor of initiative-taking, while legitimacy and adaptability function as conditional enablers. Importantly, individuals with lower agency can still take initiative when these enabling factors combine. Thus, this article contributes to understanding the micro-foundations of institutional change in public sector accounting, demonstrating that individual psychological attributes and relational resources shape reform behavior more than formal position.
The following section develops the theoretical framework, drawing on institutional entrepreneurship to identify key characteristics associated with reform agency and formulate hypotheses. Section 3 outlines the Italian context, examining the fragmented accounting environment and institutional trajectory leading to the ITAS reform. Section 4 presents the methodological approach: survey design, variable operationalization, and analytical strategy. Section 5 reports the empirical findings and robustness measurements. Finally, Section 6 discusses the implications of the results, reflects on limitations, and concludes by highlighting this study’s contribution to the literature on local government accounting reform and institutional entrepreneurship.
2 Theoretical Framework & Hypotheses
New institutional theory provides a useful lens to understand both the persistence and transformation of institutions in the public sector, emphasizing how organizational behavior is shaped not only by efficiency concerns but also by embedded rules, norms, and cognitive structures (Carpenter & Feroz, 2001; DiMaggio & Powell, 1983). PSA reforms are not only technical changes but institutional ones, involving shifts in administrative traditions, professional identities, and policy logic (Argento et al., 2018; Jorge et al., 2020).
This paper draws on institutional entrepreneurship theory (Hardy & Maguire, 2017). We identify three categories of individual-level characteristics that the IE literature associates with actors who drive institutional change: agency & motivation (the internal drive to initiate change), adaptability & change readiness (psychological preparedness for change), and legitimacy (the relational resources enabling support for initiatives). Together, these address the core challenge of institutional entrepreneurship: how embedded actors can initiate change, navigate uncertainty, and mobilize support within constraining structures (Seo & Creed, 2002). We also consider autonomy and professional role as structural moderators (Battilana, 2006), before deriving hypotheses.
2.1 Agency & Motivation
Agency is a well-established concept across the social sciences, referring to an individual’s capacity to act independently and shape outcomes. In psychology, Bandura (2018, p. 130) defines an agent as one who can “[…] intentionally produce certain effects by one’s actions.” This emphasizes that humans possess the capacity to “shape their environment and life courses.” In sociology, agency denotes the ability to act within and upon social structures (Emirbayer & Mische, 1998; Giddens, 1984; Leca et al., 2008). This understanding differs from public management literature, where ‘agency’ often refers to governmental organizations or principal-agent relationships. In institutional theory, agency specifically concerns an actor’s ability to drive institutional change within constraints (Battilana et al., 2009).
Within this discussion, many scholars have raised the issue of embedded agency (Emirbayer & Mische, 1998; Leca & Naccache, 2006), which, in essence, asks how individuals can change an institution if their behavior, actions, and rationality are shaped by the institution they want to change (Holm, 1995). In their seminal paper, Seo and Creed (2002) referred to this problem as the “paradox of embedded agency.” This is central to understanding agency in IE, as institutional entrepreneurs must navigate existing structures and norms while simultaneously seeking to transform them to achieve new goals (Battilana & D’Aunno, 2009).
Institutional entrepreneurs exercise agency by recognizing opportunities within their fields, mobilizing resources, and strategically influencing stakeholders to support their initiatives (Hoogstraaten et al., 2020). This strategic agency involves deliberate action, such as building coalitions, leveraging social skills, and positioning themselves as credible change agents in gaining legitimacy and support for reform efforts (Fligstein, 1997).
Agency in IE is also closely tied to motivation, as institutional entrepreneurs are often driven by personal values, career aspirations, and a commitment to advance institutional goals (Pacheco et al., 2010). Motivation shapes the intentionality of agency, as actors interpret past experiences, respond to present constraints, and work toward envisioned futures. This aligns with Bandura’s (2018) concept of forethought – the capacity to motivate and guide oneself by adopting goals and visualizing likely outcomes – reflecting what Emirbayer and Mische (1998) describe as a temporally embedded perspective on agency. This perspective emphasizes that agency is not merely reactive but is shaped by an actor’s aspirations and capacity to imagine alternative institutional possibilities (Battilana et al., 2009).
By conceptualizing agency in this way, this article analyzes the complex interplay between individual initiative, strategic action, and motivational forces that enable actors to initiate change within established institutions (Dorado, 2005; Lowndes, 2005). Institutional entrepreneurs, therefore, are not merely passive actors within institutions; they are proactive agents who use their strategic and motivational resources to alter institutional landscapes, aligning stakeholder interests to drive change even within constraining structures (Fligstein, 1997; Garud et al., 2007).
2.2 Adaptability & Change Readiness
Change readiness refers to the preparedness of individuals or organizations to embrace changes and effectively implement those changes (Armenakis et al., 1993). Holt and Vardaman (2013) define change readiness as the extent to which those involved in change are capable and motivated to carry out that change.
Most research on change readiness focuses on organizations (Armenakis et al., 1993; Weiner, 2009), but some extends to individuals (Holt & Vardaman, 2013; Rafferty et al., 2013). This paper focuses on the change readiness of single actors in the form of potential institutional entrepreneurs. Change readiness represents a precondition to either engaging in or resisting change initiatives (Rafferty et al., 2013). This is emphasized by Lyons et al. (2009, p. 461), stating that “[it] is an important construct because an individual’s change readiness will be related to their actual adoption or acceptance of the change initiatives at hand.”
On an individual level, change readiness comprises various psychological factors such as the recognition of the necessity for change, openness to new ideas, agreement on the proposed change initiative, confidence in the benefits of change, and a willingness to adapt (Holt & Vardaman, 2013; Lyons et al., 2009; Rafferty et al., 2013; Weiner, 2009). According to Rafferty et al. (2013), readiness for change is a multifaceted construct that involves not only the individual’s cognitive and emotional receptivity to change but also their perception of the organization’s capacity to execute the proposed initiatives. In the same vein, Holt and Vardaman (2013) argue that readiness for change should be understood as more than just a willingness to adopt new practices, as it also involves behavioral commitment and cognitive alignment with the goals of the change. In their longitudinal study, Cunningham et al. (2002) emphasize that psychological readiness for change is significantly correlated with behavioral factors, such as active participation in change-related activities and a reduction in resistance, which are essential for sustainable change.
In the context of IE, change readiness is particularly critical, since institutional entrepreneurs rely on their own preparedness and that of their organization to enact and sustain reforms (Battilana et al., 2009). As a result, for institutional entrepreneurs, assessing change readiness involves evaluating their organization’s preparedness to launch initiatives, confidence in realizing reform goals, and the perceived flexibility to navigate challenges during implementation (Holt & Vardaman, 2013; Salignac et al., 2018).
2.3 Legitimacy
While the concept of legitimacy has traditionally been applied at the organizational level (Suchman, 1995), a growing body of literature recognizes its relevance at the individual level, particularly in institutional entrepreneurship (Lawrence & Suddaby, 2006; Pacheco et al., 2010; Tost, 2011). Institutional entrepreneurs act within and across organizational boundaries, requiring them to build personal legitimacy to advocate for change effectively (Deephouse & Suchman, 2008; Greenwood & Suddaby, 2006). As Pacheco et al. (2010, p. 986) note, “success [of a change initiative] may be dependent upon the legitimacy that the institutional entrepreneur enjoys.” This article builds on the framework developed by Suchman (1995), who distinguishes three dimensions of legitimacy – cognitive, normative, and pragmatic – and applies them to the individual level.
Cognitive legitimacy pertains to perceived expertise and competence. Stakeholders are more likely to support changes initiated by individuals perceived as knowledgeable and capable (Hoogstraaten et al., 2020). Tost’s (2011) integrative model proposes that legitimacy is fundamentally a social judgment process applying to individuals whose knowledge and skills align with stakeholder expectations. Similarly, DiMaggio (1988) emphasizes that perceived credibility is essential for individuals attempting to introduce institutional change.
Normative legitimacy involves alignment with professional and institutional norms. While Suchman (1995) uses the term “moral,” “normative” is more appropriate in the IE context, as it refers to field-specific norms rather than broad ethical values (Lawrence & Suddaby, 2006). Maguire et al. (2004) show that legitimacy in IE is primarily about professional standards and relevance. This alignment is particularly important as institutional entrepreneurs must navigate established norms to gain acceptance, even while working to reform these structures.
Pragmatic legitimacy reflects the perceived practical benefits of an actor’s initiatives. Institutional entrepreneurs must demonstrate that their proposed changes provide tangible benefits to stakeholders (Maguire et al., 2004). Drori and Honig (2013) emphasize that individual actors must build pragmatic legitimacy by showing their initiatives align with stakeholder needs and add measurable value. This dimension establishes the utility of their efforts, encouraging others to support or participate in proposed changes.
In sum, individual legitimacy can be understood as a reflection of perceived expertise (cognitive), alignment with professional norms (normative), and practical benefit (pragmatic). This transfer from the organizational to the individual level is empirically supported by studies of actors who drive institutional change by utilizing personal legitimacy to gain stakeholder support (Maguire et al., 2004; Tost, 2011).
2.4 Autonomy & Professional Roles
Autonomy refers to an individual’s perceived discretion to make decisions and act independently within their organizational role (Mutch, 2007). It encompasses the scope of authority, flexibility in decision-making, and the degree of control over professional tasks (Tummers & Bekkers, 2014). In the context of PSA reforms, autonomy becomes particularly relevant, as frontline actors are required to interpret and implement changes in ways that go beyond routine compliance (Argento et al., 2018; Battilana, 2006). Institutional entrepreneurs who perceive themselves as having greater autonomy are more likely to identify opportunities for innovation and exercise discretion in adapting or promoting reform agendas (Pacheco et al., 2010).
Research on street-level bureaucracy similarly underscores that autonomy is a structural feature and a subjective experience that shapes behavior (Lipsky, 2010; Tummers & Bekkers, 2014). Even those with less senior roles may act entrepreneurially if they perceive room to maneuver within their institutional environment (Tummers & Bekkers, 2014). Conversely, highly centralized or hierarchical structures may inhibit such behavior, particularly if autonomy is minimal or highly constrained by formal procedures and oversight mechanisms. Autonomy thus operates not merely as a resource but as a contextual moderator, enabling proactive behavior when aligned with motivational factors and supportive organizational conditions (Tummers et al., 2015).
The professional role refers to an individual’s formal occupational status within the organizational hierarchy – such as accounting operator, officer, or director in the context of this article. The theory suggests that senior civil servants, leading managers, or politically appointed reform leaders possess not only strategic oversight but also the formal discretion to evoke power, allocate resources, set priorities, and mandate organizational changes (DiMaggio & Powell, 1983; Phillips et al., 2000). Such authority would give institutional entrepreneurs higher levels of influence that are unavailable to those lower down the hierarchy. Accordingly, Battilana (2006) argues that the higher placed an individual is in the formal organizational hierarchy, the more likely this person is to function as an institutional entrepreneur. However, actors in the middle or even lower-tier positions can also act entrepreneurially, especially when they possess high motivation, legitimacy, or organizational support (Deephouse & Suchman, 2008; Tummers & Bekkers, 2014).
While agency & motivation, adaptability & change readiness, and legitimacy are theorized as core attributes of institutional entrepreneurs, autonomy and professional roles are conceptualized as contextual enablers that shape the conditions under which these individual traits can translate into initiative-taking (e.g., Leca et al., 2008). Autonomy captures perceived discretion in decision-making (Tummers & Bekkers, 2014), while role reflects the formal hierarchical position, both of which may influence how individuals navigate reform dynamics.
2.5 Hypotheses
Building on the theoretical considerations outlined in Sections 2.1 to 2.4, this study advances a set of hypotheses linking individual-level attributes to initiative-taking behavior during the implementation of public sector accounting reforms. Specifically, the following propositions are formulated:
H1: Individuals who exhibit higher levels of agency and motivation are more likely to report taking the initiative in reform implementation than those with lower levels.
H2: Individuals who demonstrate greater adaptability and readiness for change are more likely to report initiative-taking than those who perceive themselves as less adaptable or less ready for change.
H3: Individuals who perceive themselves as possessing higher levels of legitimacy within their organization are more likely to report initiative-taking than those with lower perceptions of legitimacy.
In addition to these key psychological characteristics, this study examines the moderating influence of two structural factors: perceived autonomy in decision-making and formal professional role within the organization. Perceived autonomy reflects the individual’s discretion to make independent decisions, while professional role captures the formal hierarchical position that may enable or constrain such discretion. Figure 1 illustrates the full theoretical framework, showing how the three individual-level characteristics – organized along motivational, psychological, and relational dimensions – are hypothesized to predict initiative-taking, with autonomy and professional role acting as structural moderators. This framework guides the empirical analysis that follows.
Figure 1
Theoretical Framework.
3 The Italian Context
Italy is a unitary state with a complex and multi-layered local government system, defined in Article 114 of the Italian Constitution and in the Legislative Decree 267/2000 (Testo Unico degli Enti Locali – TUEL). Accordingly, the levels of local government include municipalities (comuni), provinces (province), regions (regioni) and metropolitan cities (città metropolitane) (Valdesalici & Trettel, 2023). In addition, unions of municipalities (unioni di comuni) constitute a collaborative organizational form that aggregates smaller municipalities to provide services jointly. As of 2023, Italy comprises nearly 7,900 municipalities, over 100 provinces, 14 metropolitan cities, and 20 regions, creating a highly diverse institutional landscape with significant heterogeneity in administrative capacity, fiscal autonomy, and governance structures (Benfante et al., 2024). Such structural diversity has posed particular challenges for public financial management and has directly influenced the trajectory and effectiveness of accounting reforms (Manes Rossi & Aversano, 2015). In particular, the decentralization of competencies and the varying degrees of technical readiness among local governments have led to heterogeneous implementation paths, complicating the harmonization of accounting systems and financial reporting practices across the country (Mussari & Sorrentino, 2017). Earlier reforms in the 1990s, inspired by New Public Management, failed due to fragmented implementation and lack of a unified framework (Manes Rossi et al., 2019).
The Italian government embarked on a series of reform steps to address these challenges, culminating in a constitutional amendment in 2001 to prioritize accounting harmonization (Mussari et al., 2021). However, it was not until 2009 that a new legislative framework was adopted to introduce a national standard-setting body and initiate a coordinated move towards full accrual accounting (Grandis & Mattei, 2012). The new legislation required adopting a common set of accounting standards for all public sector entities and introducing accrual accounting for those entities still operating on cash-based accounting (Columbano et al., 2023). These developments reflected internal and external pressures, particularly the 2011/85/EU Directive, which emphasized the need for reliable, accrual-based fiscal data at all levels of government (Sforza & Cimini, 2017).
In response, Italy launched several initiatives under the EU Structural Reform Support Program (SRSP) to consolidate its PSA framework. A key outcome has been the introduction of the ITAS reform, with its central goal of “equipping public administrations with a single accrual accounting system” by the year 2026 (Italian Government: Presidency of the Council of Ministers, 2025). This latest initiative seeks to unify accounting practices across the entire public sector in alignment with international standards such as IPSAS and the prospective EPSAS (Mazzotta et al., 2022). By 2025, this initiative had produced a complete set of outputs, including 18 Italian Accounting Standards (ITAS), a single conceptual framework, and a national chart of accounts (Italian Government: Presidency of the Council of Ministers, 2025). As part of a pilot phase of the implementation of the National Recovery and Resilience Plan, certain pilot entities have been selected that include almost all local governments, i.e.,: regions, provinces, metropolitan cities and municipalities with a resident population equal to or greater than five thousand (Italian Government: Presidency of the Council of Ministers, 2024).
In this evolving context, the role of institutional entrepreneurs at the local level is strategically important as these individuals can mediate between national reform mandates and the practical constraints of local implementation (Salonen et al., 2024), particularly in contexts marked by technical limitations, organizational inertia, or cultural resistance to change (Argento et al., 2024). Their presence will likely determine whether these reforms achieve a sustainable impact across Italy’s fragmented local government landscape. As Battilana (2006, p. 671) argues, “[…] when governments want to implement major public reforms, they need to identify the people on whom they can rely to locally initiate major public reforms.” Understanding the profile of individuals likely to function as institutional entrepreneurs thus becomes crucial, not only for advancing reforms on paper but for enacting them locally.
4 Research Design and Methodology
A cross-sectional survey design was employed to explore individual-level characteristics of potential institutional entrepreneurs in the context of the ITAS reform. This approach is well suited for capturing psychological and behavioral attributes – such as agency, motivation, adaptability, and legitimacy perceptions – at a specific moment during the reform process (Hall, 2008; Setia, 2023). A cross-sectional design also allows for the systematic comparison of actor-level dispositions across diverse institutional contexts (Setia, 2023), which suits the fragmented PSA environment of Italian local government.
4.1 Survey Methodology
The survey was grounded in existing theoretical frameworks from the literature on institutional entrepreneurship, including agency (e.g., Battilana et al., 2009; Pacheco et al., 2010), change readiness (e.g., Holt & Vardaman, 2013; Rafferty et al., 2013), and legitimacy (e.g., Suchman, 1995; Tost, 2011). The questionnaire included Likert-scale and categorical items alongside open-ended prompts to capture subjective experiences and interpretations of reform processes. This structure supports the study’s conceptual emphasis on the actor-level of institutional change, which seeks to understand how individuals contribute to implementing reform through their perceptions, actions, and interactions (Argento et al., 2018; Bandura, 2018).
The survey was distributed online between September and December 2024 to public officials from local government entities involved in the pilot phase of Italy’s accrual accounting reform (ITAS reform). It was sent to a total of 578 entities, including 472 municipalities, 30 provinces, 7 metropolitan cities, 19 regions and autonomous provinces, and 50 mountain or island communities and unions of municipalities. The sample was deliberately drawn from local governments that voluntarily participated in an ITAS online training course, under the assumption that such engagement may indicate a higher likelihood of reform-mindedness and initiative-taking. This purposeful sampling strategy follows the recommendation of Palinkas et al. (2015) to select information-rich cases in implementation research to support analytical depth.
A total of 47 local government entities completed the survey in full, corresponding to a response rate of 8%. The final sample comprised 37 municipalities, 5 provinces, 3 mountain or island communities, 1 region, and 1 union of municipalities.
The questionnaire aimed to capture the respondents’ perceptions, attitudes, and behavior regarding the reform, with a focus on roles, autonomy, agency, motivation, adaptability, and perceived legitimacy within their organizational context. To ensure validity, the survey underwent a critical pre-test with Italian public service managers selected exclusively from outside the target population, avoiding any overlap with the main sample. The entire process of the survey design and execution was conducted through the Qualtrics XM® Experience Management Software.
4.2 Analytical Strategy
A classification tree model was employed to explore the conditions under which civil servants are more likely to take initiative in the context of the ITAS reform. The dependent variable, initiative-taking (Q8), captures whether respondents reported taking the initiative to shape reform implementation (1 = yes, 0 = no). Classification tree models can be helpful in analyzing complex interactions among variables in reform implementation, particularly for small sample sizes (Siciliano, 1999). Unlike logistic regression, which assumes a linear relationship between predictors and the log odds of the outcome, classification trees make no such parametric assumptions (Kang et al., 2021). They sequentially partition the data based on the variable that best separates the dependent outcome – in this case, initiative-taking behavior. The model thus identifies interaction structures among predictors and provides intuitive pathways through a decision-tree model.
The tree was trained using a recursive partitioning algorithm, with a complexity parameter (cp) of 0.01 and a minimum split size of ten observations to balance model flexibility with interpretability. This approach is well suited to explore non-linear interactions and conditional patterns among variables, particularly when the sample size is modest (n = 47) (Kang et al., 2021) and the predictors include a combination of ordinal indices and categorical variables as in our dataset (Kuhn & Johnson, 2013).
The model comprises three composite indices, each representing the mean of selected indicators from the survey:
Agency & Motivation (Q7_1, Q7_2, Q7_3, Q12_1, Q12_2) (→ AgencyMotiv)
Adaptability & Change Readiness (Q10_1, Q10_2, Q10_3, Q12_3) (→ AdaptChange), and
Legitimacy (Q11_1 to Q11_5).
Regarding predictors, the model also included the respondent’s professional role (Q3) and their self-reported autonomy (Q5). After pruning based on the complexity parameter (cp = 0.01), the final tree included four variables: AgencyMotiv, AdaptChange, Legitimacy, and Autonomy. Notably, the variable “Role” was excluded during the pruning process, indicating limited marginal contribution to the classification performance. Table 1 provides an overview of all variables included in the classification tree model. The table distinguishes between the dependent variable, the three theoretically derived composite indices, and the two predictor variables, “Autonomy” and “Role.”
Table 1
Description of variables.
| VARIABLE DESCRIPTION | VARIABLE NAME (SURVEY CODE) | TYPE | SCALE/QUESTION DESIGN | DESCRIPTION |
|---|---|---|---|---|
| Initiative-taking (dependent variable) | Q8 | Binary | 1 = yes, 0 = no | Respondent reports taking the initiative to influence reform implementation. |
| AgencyMotiv Index | Q7_1, Q7_2, Q7_3, Q12_1, Q12_2 | Composite Index | 5-point Likert (1 = strongly disagree to 5 = strongly agree) | Self-perceived agency, motivation, leadership, and risk orientation. |
| AdaptChange Index | Q10_1, Q10_2, Q10_3, Q12_3 | Composite Index | 5-point Likert (1 = strongly disagree to 5 = strongly agree) | Perceived organizational flexibility, change readiness, and value alignment. |
| Legitimacy Index | Q11_1 to Q11_5 | Composite Index | 5-point Likert (1 = strongly disagree to 5 = strongly agree) | Frequency of being consulted, recognized, and supported by peers/superiors. |
| Autonomy | Q5 | Ordinal | 1 = limited to 5 = full autonomy | Self-reported decision-making authority. |
| Professional Role | Q3 | Categorical | Categorical: 1) Accounting Operator/Instructor 2) Accounting Officer 3) Director/Municipal Secretary | An occupational category within the administration. |
The questionnaire included additional items such as training motivation and demographic information. Those items were included in previous model iterations but yielded no added value to the explanatory power of the classification tree model. Accordingly, they were excluded from future analytical steps.
5 Results
The first part of this section presents a comprehensive overview of the results from the classification tree analysis, which includes a step-by-step description of how the model evolved. To support the results and to increase the robustness of this methodology, bootstrap aggregation was performed, which is discussed in the second part of this section.
5.1 Decision Tree Analysis: Predicting Initiative-Taking
The final tree presented in Figure 2 was constructed based on a sample of 47 observations. The model sequentially split the data using four predictors: the Agency & Motivation Index (AgencyMotiv), the Legitimacy Index, the Adaptability & Change Readiness Index (AdaptChange), and the Autonomy factor. While the variable “Role” was included in the input data, it was not selected in the final tree owing to its comparatively low contribution to reducing classification error.
Figure 2
Classification Tree.
Figure 2 presents the classification tree, which begins at the root node containing all 47 observations, thus representing 100% of the sample, with a base classification rate of 66%. This means that 31 out of 47 respondents (i.e., 66%) did not report taking the initiative. The tree then sequentially partitions the data to best differentiate between those who did and did not report initiative-taking (i.e., Q8 = 1 for “yes,” and Q8 = 0 for “no”).
Each node in the figure follows a consistent format: the node number is displayed at the top; the predicted classification (0 or 1, i.e., yes or no to Q8) appears first within the box; the central figure shows the distribution of Q8 responses in percentage, with “no” always listed on the left and “yes” on the right; and the bottom figure indicates the percentage of the total sample represented by that node (n = 47). The following subsection provides a step-by-step interpretation of the four main splits in the tree and discusses the implications of each branching point for understanding initiative-taking behavior.
5.2 Interpretation of Tree Splits
1) AgencyMotiv – First Split
The first and most influential split occurs at a threshold of 4.7 on the Agency & Motivation Index. The value of 4.7 relates to the index score with a maximum of 5, where 1 = strongly disagree and 5 = strongly agree. This index aggregates five items (Q7_1 to Q7_3 and Q12_1, Q12_2, with a maximum of 5) that capture perceived leadership capability, risk orientation, and motivation. Thus, individuals with lower agency scores (<4.7) were mostly classified as non-initiative-takers (node 2: 74% “no”, 26% “yes”), while those with higher AgencyMotiv scores (≥4.7) were more likely to report initiative (node 3: 75% “yes”, 25% “no”).
2) Legitimacy – Second Level, Node 2
For individuals with lower AgencyMotiv scores, a further split is made based on their Legitimacy Index score (Q11_1 to Q11_5), with a threshold of 3.5, again with a maximum of 5. Respondents scoring below 3.5 on the legitimacy index (node 4) show the lowest initiative-taking: 89 percent reported not taking the initiative. In contrast, those with legitimacy scores above 3.5 (node 5) are more mixed: 60 percent “no” vs. 40 percent “yes.”
3) Autonomy – Third Level, Node 5
Among respondents with medium to high legitimacy and low agency motivation, the next split is made using the Autonomy variable (Q5), which was treated as a categorical factor with five levels (Limited, Mixed, Collaborative, Delegated, and Full Autonomy). Interestingly, the model selected only two autonomy levels for further differentiation: Mixed and Delegated. This choice suggests that the likelihood of initiative-taking differs markedly at certain levels of perceived decision-making authority. In node 10 (where these autonomy levels converge), only one out of ten respondents took the initiative. This is theoretically coherent (Tummers & Bekkers, 2014): Initiative-taking is less likely when autonomy is either minimal or bounded by collaborative constraints – particularly in bureaucratic environments or when it is very high but not combined with other enabling conditions (Battilana et al., 2009).
4) AdaptChange – Final Split, Node 11
Finally, among individuals with mid-level agency, moderate to high legitimacy, and specific autonomy levels (i.e., Mixed or Delegated), the model splits further by Adaptability & Change Readiness at a threshold of 2.9 (Q10_1, Q10_2, Q10_3, Q12_3). Here, those scoring above 2.9 in the index (node 23) are very likely to report initiative-taking (88% “yes”), while those below are not (node 22: 100% “no”). This indicates that readiness to adapt and an openness to change are crucial for initiative-taking.
The tree model findings reveal the following patterns: Agency & Motivation is the most decisive factor in distinguishing initiative-takers from non-initiative-takers, confirming theoretical expectations that self-perceived leadership, curiosity, and risk tolerance are key antecedents of proactive behavior (Hardy & Maguire, 2017; Pacheco et al., 2010). Legitimacy, defined as feeling supported and recognized by peers and supervisors (e.g., Suchman, 1995), is a secondary condition, particularly relevant for those with lower agency. Autonomy operates as a contextual moderator; its contribution is not linear, and only specific configurations (e.g., Mixed or Delegated) lead to a meaningful differentiation (e.g., Tummers & Bekkers, 2014). Lastly, Adaptability & Change Readiness plays a final but critical filtering role among respondents who would otherwise not be considered initiative takers, thus determining who ultimately crosses the threshold into proactive behavior.
The final tree consists of five internal and six terminal nodes, achieving a relative error of 0.375 and a cross-validated error of 0.9375. The tree includes no surrogate variables in the final splits, confirming the selected variables offer clear, non-overlapping information about variation in the dependent variable. Therefore, it successfully captures meaningful structure and differentiates initiative-takers from non-takers using theoretically grounded variables. However, the cross-validated error was higher (0.9375), highlighting sensitivity to the small sample and potential overfitting. This trade-off is common in models that balance interpretability and predictive accuracy in exploratory, small-N settings (Kang et al., 2021). To capture these effects better and check the classification tree’s robustness, bootstrap aggregation was performed, which is explained in detail in the next section.
5.3 Robustness Check: Bootstrap Aggregation
Bagging (bootstrap aggregating) was employed as a secondary test to support the classification tree findings and increase model robustness. Bagging is an ensemble method generating multiple decision trees on random bootstrap samples of the dataset and aggregating predictions to reduce variance and mitigate overfitting (Breiman, 1996; Kuhn & Johnson, 2013). This technique is particularly suitable in exploratory, small-N settings, where individual trees may be overly sensitive to sample-specific fluctuations (James et al., 2021). Unlike single decision trees, which offer interpretability but are prone to overfitting, bagging improves generalizability by combining the results of many individually unstable models trained on different data samples (Dietterich, 2000). It serves here as a powerful diagnostic tool for assessing the robustness of theoretical constructs operationalized in public administration settings (James et al., 2021).
Bootstrap aggregation was implemented using 1,000 bootstrapped trees trained on the same five predictors: AgencyMotiv, AdaptChange, Legitimacy, Autonomy, and Role. All predictors were kept in their raw form (ordinal or categorical), and the binary outcome variable reflected whether respondents reported initiative-taking. To assess the contribution of each variable, we computed the average variable importance, which quantifies how much each predictor improves classification accuracy across all bootstrapped trees (James et al., 2021). The results are presented in Figure 3 below and confirm that AgencyMotiv was the most influential predictor (importance score: 12.3), followed by AdaptChange (11.6), Legitimacy (11.6), and Autonomy (9.7). Role showed the lowest contribution (8.4), reinforcing its limited explanatory power observed in the single-tree model. These scores reflect the relative contribution of each predictor to improving classification performance across the ensemble, with higher values indicating a stronger average impact on decision accuracy.
Figure 3
Variable Importance from Bagging Model.
These results indicate that AgencyMotiv was the most influential variable, mirroring its role in the classification tree analysis. This finding reinforces theoretical expectations that internalized motivation, perceived leadership capacity, willingness to take risks, and curiosity are critical antecedents of initiative-taking behavior in public sector reform contexts (Battilana & D’Aunno, 2009; Emirbayer & Mische, 1998).
AdaptChange and Legitimacy both demonstrated substantial contributions to the model, with importance scores virtually equivalent to AgencyMotiv. This underscores the relevance of openness to change, value alignment, and perceived organizational flexibility as critical psychological enablers of initiative-taking (Holt & Vardaman, 2013; Rafferty et al., 2013). Likewise, Legitimacy – operationalized as the extent to which individuals feel recognized, consulted, and supported (Tost, 2011) – emerges as a meaningful predictor. These findings align with the literature, suggesting that reform engagement is shaped not only by character traits such as agency, but also by perceptions of social embeddedness and change readiness (Cunningham et al., 2002; Suchman, 1995).
Autonomy also contributed to the classification model, but to a lesser extent. Its lower variable importance score (see Figure 3) indicates that perceived decision-making autonomy functions more as a contextual moderator than a primary driver of initiative-taking. This supports earlier research on the autonomy of street-level bureaucracy, which suggests that certain levels of discretion can enable proactive behavior but more likely so when coupled with strong motivation or legitimacy (Lipsky, 2010; Tummers & Bekkers, 2014).
Again, Professional Role emerged as the least informative variable in the model. This result corroborates its exclusion from the final classification tree and indicates a degree of redundancy with other individual-level attributes such as autonomy and agency. It also reinforces the view that official hierarchical position alone is not a reliable predictor of reform-oriented behavior when individual characteristics and relational factors are considered (Battilana, 2006).
6 Discussion & Conclusion
This study explored the individual-level characteristics that predict initiative-taking among local government officials involved in implementing Italy’s accrual accounting reform. Drawing on institutional entrepreneurship theory, the paper hypothesized that individuals who take initiative exhibit higher levels of agency and motivation (H1), greater adaptability and readiness for change (H2), and higher perceived legitimacy (H3) compared to non-initiative-takers.
A central contribution of this study is operationalizing institutional entrepreneurship at the individual level. While IE theory has frequently emphasized the importance of agency, legitimacy, and change orientation in conceptual terms (Battilana et al., 2009; DiMaggio, 1988; Pacheco et al., 2010; Suchman, 1995), few empirical studies have translated these into measurable constructs grounded in actor-level perceptions.
This paper addresses this gap by developing and applying three composite indices – agency & motivation, adaptability & change readiness, and legitimacy – transformed into validated survey items. These indices enabled a theoretically informed, empirically robust analysis of how different key characteristics of IEs manifest in local government contexts. Additionally, two structural moderators, autonomy and professional role, were incorporated to assess how perceived and formal organizational positioning shapes entrepreneurial initiative.
The empirical findings strongly support the central role of agency and motivation, confirming H1. Across the classification tree and bootstrap aggregation approach, this construct emerged as the most decisive factor in distinguishing initiative-takers from non-initiative-takers. This aligns with the theoretical expectation that institutional entrepreneurship is rooted in an individual’s capacity to act strategically within existing constraints, guided by a sense of purpose, leadership, and willingness to create and sustain change (Bandura, 2018; Dorado, 2005).
Support for H2 (adaptability & change readiness) was also observed, albeit in a more nuanced manner. While not as dominant as agency, adaptability functioned as a conditional enabler – particularly effective in distinguishing initiative-taking among individuals with mid-level scores in other key characteristics. This reinforces the idea that psychological readiness to change, including openness to reform values and alignment with organizational goals, is a facilitating condition that interacts with deeper motivational factors (Holt & Vardaman, 2013; Rafferty et al., 2013).
H3 (legitimacy) received moderate empirical support. While not the strongest predictor, the legitimacy index ranked closely behind AgencyMotiv in importance, appearing as the second-level split in the classification tree, reinforcing its conditional role, particularly for individuals with lower agency or motivation. This suggests legitimacy acts as a relational resource that compensates for limited personal agency, enabling actors to gain support, recognition, or influence through their embeddedness in social structures (Battilana, 2006; Qureshi et al., 2016). This aligns with theoretical accounts of legitimacy as a vital component of institutional entrepreneurship, especially in contexts where authority or autonomy is constrained (Suchman, 1995; Tost, 2011).
The two additional factors – autonomy and the professional role – were included in the analysis as structural or contextual enablers. The results show that autonomy contributes meaningfully, but in non-linear ways: Only certain levels of perceived decision-making authority are associated with higher initiative-taking and only in specific combinations with other traits. This supports recent studies that emphasize the subjective and contextual nature of autonomy in an individual’s actions (Tummers & Bekkers, 2014). By contrast, professional role had the weakest explanatory power, suggesting that hierarchical status alone does not reliably predict proactive reform behavior when individual character traits and perceived autonomy are considered (Battilana, 2006).
The model revealed an important secondary pathway beyond high agency. While individuals scoring above 4.7 on AgencyMotiv had a 75 percent likelihood of taking initiative, the analysis showed that lower agency scores do not preclude initiative-taking. In fact, individuals with lower agency reached an even higher likelihood (88 percent) when three conditions were met: they perceived high legitimacy from peers and supervisors (above 3.5), experienced mixed or delegated autonomy in decision-making, and demonstrated strong adaptability and readiness for change (above 2.9). This pattern confirms that agency, while important, is not the sole determinant of institutional entrepreneurship. When individuals are embedded in supportive organizational contexts (legitimacy), have appropriate decision-making discretion (autonomy), and are psychologically ready for change (adaptability), these factors can effectively compensate for lower intrinsic agency. This combination enables them to take initiative in reform implementation despite not scoring high on traditional agency measures.
These findings highlight that initiative-taking in public sector accounting reforms is shaped more by actor-level attributes than by formal authority or role. This has important implications for Italy‘s ongoing ITAS implementation at the local government level. If reform success depends not only on technical capacity but also on the motivational and cognitive engagement of individuals (Argento et al., 2018; Benfante et al., 2024), then policy strategies should invest in cultivating agency, readiness, and perceived legitimacy among frontline staff (Battilana, 2006). As Bandura (2018) emphasizes, agentic capacities are modifiable, providing a foundation for such targeted interventions rather than relying solely on top-down directives. Moreover, the results resonate with collective institutional entrepreneurship (Argento et al., 2024). The presence of motivated, adaptable, and legitimate actors at the local level may form the micro-foundations of broader collective action, especially in fragmented institutional settings such as Italian local government accounting.
As with any exploratory study, our research has a number of limitations. First, the sample size was modest (n = 47), reflecting both the specificity of the pilot reform context and the challenges of accessing individual-level data in a PSA context in Italian local government. While the models were optimized for small-N analysis and yielded stable patterns, the findings have primarily exploratory character, and we suggest they are validated in future studies using larger or more diverse samples. Additionally, the study relies on self-reported measures, which, while theoretically grounded, are subject to perceptual bias.
Overall, this study offers conceptual and empirical insights on the key characteristics of potential institutional entrepreneurs, that proactively engage with public sector reforms. By operationalizing core elements of IE and testing them empirically, it lays the groundwork for more fine-grained, actor-centered analyses of institutional change in public sector accounting at the local government level.
Publication Note
This publication note explains the unusual circumstances surrounding the publication history of this manuscript and documents changes made during the revision process.
The manuscript was originally submitted to Lex Localis – Journal of Local Self-Government in March 2025 under the title “Taking the Initiative: Institutional Entrepreneurs and Accounting Reform in Italian Local Government” and was forwarded to peer reviewers shortly thereafter. After an extended period without communication regarding the review status, we attempted to contact the journal editors on multiple occasions but received no response. In September 2025, we were informed that ownership of the journal had been transferred, though the message provided gave rise to suspicions regarding the authenticity and credibility of the new owner.
Upon investigating the journal’s status, we discovered that Lex Localis had ceased operations and that submitted manuscripts were no longer adequately protected. Once we became aware of this situation, we immediately withdrew our manuscript and formally requested deletion of all associated files and data. However, we subsequently learned that portions of our original submission may have been accessed and plagiarized by other parties during the period when the journal’s systems were inadequately secured.
Following withdrawal, the manuscript was revised in response to peer reviewer feedback obtained through subsequent submission to the Yearbook of Swiss Administrative Sciences. As part of these revisions, the title was changed to “Identifying Institutional Entrepreneurs: Initiative-Taking and Individual Characteristics in Italian Public Sector Accounting Reform” to more accurately reflect the paper’s focus on identifying and characterizing institutional entrepreneurs at the individual level. This title change, along with other substantive revisions, distinguishes the current version from the original submission.
We have taken all reasonable measures to protect the integrity of our work, including monitoring for unauthorized use of our content. Nevertheless, we cannot entirely rule out the possibility that fragments from our original submission – bearing the earlier title – may appear elsewhere. We wanted to make readers aware of these circumstances and assure them that any such instances would reflect the compromised security of the previous journal.
Acknowledgements
We are extremely grateful to all the local government public officials in Italy who have taken the time to complete the survey. Without their support, this research paper would not have been possible.
Competing Interests
The authors have no competing interests to declare.