Institutional entrepreneurship and change in consumer protection policy in the telecommunications sector: innovations in the text-based analysis approach

Abstract This article analyses the institutional entrepreneurship within independent regulatory agencies (IRAs) as a variable explaining policy change over time and seeks to offer new insight for the identification and analysis of structure-agency relationships. The article contributes to the institutional entrepreneurship research agenda by connecting changes in IRA consumer protection policy to changes in agency leadership (specifically, agency presidents). The method used relies upon a quantitative and qualitative text analysis approach to connect and pinpoint structure-agency dynamics over time. The empirical sections compare and contrast the results obtained through the content analysis of the annual reports issued between 2000 and 2015 by the Italian Communications Authority (Agcom), and illustrate variations between periodic changes to Agcom’s presidency and changes in ideas, strategies and tools in the field of consumer protection in the telecommunications sector.


Introduction: institutional change as an external or an internal variable?
In Europe, the rise and diffusion of Independent Regulatory Agencies (IRAs) have modified the policy game in many policy sectors (Jordana, Levi-Faur, & Fernàndez, 2011). Nevertheless, the interplay over time of institutional and policy change is still far from being fully analysed and understood (Bakir, 2009;Hall, 1993;Maggetti, 2014).
Previous studies on IRAs (Gilardi, 2002(Gilardi, , 2007(Gilardi, , 2008Gilardi & Maggetti, 2011;Maggetti, 2010Maggetti, , 2012Tabellini, 2007) offer many examples of how normative ideas (beliefs) such as independence and credibility are crucial in explaining institutional change, that is to say, the rise of Regulatory State (RS) and the regulatory shift from the political to the technical arena -including IRAs (Righettini & Nesti, 2014). The rational paradigm has highlighted the structural variables of the governance system (Levi-Faur, 2010 in a number of policy areas characterized by large privatization programmes (Lodge, 2014;Thatcher, 2002). However, the rational approach represents the principle-agent relationship as an KEYWORDS institutional entrepreneurship; policy change; consumer protection; text analysis; independent regulatory agency interest-based zero-sum game and obscures the role crucial actors play in interpreting and guiding regulatory change (Gilardi, 2005;Jordana et al., 2011;Maggetti, 2012Maggetti, , 2014Moran, 2003). In contrast, the 'institutional approach' explains the rise and diffusion of new independent and credible organizations somewhat differently, showing how political actors justify and convert ideas of credibility and independence into institutional and policy change (Bakir, 2009).
The 'rhetorical approach' , on the other hand, elects to focus on individual actors (IRA top figures) and analyse their public discourses as soft regulation tools (Lim, 2002;Olson, 2007;Seyranian & Bligh, 2008;Tabellini, 1987). Public discourses, both oral and written, are used as empirical evidence to identify and to track the actor engagement in changing structures and policy outcomes over time (Bligh & Hess, 2007). The rhetorical approach has a long tradition in analysing sounds, strategies and personal leadership techniques (Conger, 1991;Conger & Kanungo, 1998;Fiol, Harris, & House, 1999;Hart, 1987;Lasswell, 1948;Lowi, 1985), but has had only limited application among IRA scholars, essentially due to the instrumental role assigned by the RS theory to these agents in favour of more political principals (Majone, 1994). With the exception of central bank independence reforms, the role of these individual actors, such as presidents and directors, in mobilizing ideas and bringing about concrete changes to regulatory goals, tools and strategy has largely remained obscure. Only recently have studies, among which was the application of discursive institutionalism to policy entrepreneurship and institutional change in central banking reform (Bakir, 2009(Bakir, , 2013, sought to highlight the role institutional entrepreneurs with multiple identities, operating in multiple ideational realms and 'streams' can perform in order to bring about institutional and policy change. However, a critical point of the historical-discursive approach, due to the method used (qualitative), is its difficulty in clearly operationalizing and accurately measuring, through the discourse analysed, changes occurring over time (Bakir, 2009;Campbell, 2002). The application of the discursive approach to institutions 'puts the agency back into institutional change by explaining the dynamics of change in structures through constructive discourse about ideas' (Schmidt, 2008, p. 316), but it needs to further explore and to provide clear empirical evidence to how, when, and to what extent the ideas and the discourse matter are mobilized and utilized endogenously.
This article aims to contribute to the institutional entrepreneurship research agenda by connecting institutional changes and policy change in IRAs in Italy. As will be seen in greater detail in the following sections, this article provides a different and hitherto unexamined perspective on institutional and policy change in IRAs in the telecommunications sector, by focusing on the effects of agency changes (presidency) on the annual accountability reports of an Italian Independent Regulatory Authority (Agcom). We ascertain, through a mixedmethod approach, when, if and how the idea of consumer protection in TLC increases in salience in the authority's public discourse and gains prominence in the regulatory agenda over time. In doing this, we seek to look at the institutional entrepreneurship to operationalize agency in the structure-agency debate with special reference to policy change in IRAs (Bakir & Jarvis, 2018;Campbell, 2002;Schmidt, 2010Schmidt, , 2011Schumpeter, 1942).

The discourse perspective on institutional and policy change
The discursive dimension is crucial in the literature on the institutional entrepreneurship, as evidenced both by organizational studies (Leca, Battilana, & Boxenbaum, 2008) and by the discursive institutionalism approach (Bakir, 2009;Peters, 1999;Radaelli, Dente, & Dossi, 2012) and the way in which they consider change and structure-agency dynamics. In the aim to address the issues encountered by the historical approach in understanding the structure-agency dynamic over time (Campbell, 2004;Peters, 1999), they endogenise innovation and change, arguing that ideas and discourse can explain not only institutional persistence and path dependence but also institutional change over time (Schmidt, 2001(Schmidt, , 2008(Schmidt, , 2011. The discursive dimension indicates the opportunity to link change in regulatory discourse to changes in the selection and prioritization of regulatory strategies and tools (Schmidt, 2010). In order to do this, discourse can be empirically observed at two levels: at the interactive, micro level, which takes place between individual actors through the regulative conversation and argumentation by which certain ideas succeed or fail to be translated into policy programmes and outcomes (Howart & Torfin, 2005), and at the macro level, that of the different forms of communicative and programmatic public discourse (Schmidt, 2008). Public discourse is a unique communicative tool that represents the evolution of normative action and moreover is an action in itself (Habermas, 1996). The persistent difficulties associated with the study of how new policy ideas are mobilized, and how they shape institutions, organizations and policies, at both the micro and macro levels, seem to be due primarily to the lack of causal mechanisms linking institutional change to policy change and to the lack of measurement of institutional change itself (Campbell, 2002(Campbell, , 2004. Discursive analysis can raise methodological problems of causal explication and concept operationalization, principally deriving from a lack of distinction between different types of discourses and ideas (cognitive frames, policy frames, programmatic and normative ideas) and therefore from their different functionality and dynamics in the process of institutional and policy change (Campbell, 2004). To fill this gap, the present article seeks to provide additional insight into the discursive dimension of the institutional entrepreneurship approach, by using discourse not merely to prove that discourse affects policy-making but, rather, to analyse variations between institutional change and policy change (Bakir, 2009). We focus here on one particular (and therefore limited) type of written discourse -namely the annual accountability reports of a single IRA, which have both a communicative content, providing an accounting to other actors, organizations and stakeholders in the governance system, and a programmatic content, related to the salience of specific regulatory issues, to that which has been achieved and to the goals that have yet to be accomplished. The accountability reports can be used to identify changes that occur over time in the context of relative structural (de jure) independence and functional stability (delegation), which is precisely the case of Agcom, the Italian independent regulator for the telecommunications sector, in the period considered.

Institutional entrepreneurship in regulation: concept definition and research questions
The institutional entrepreneurship theory aims at explain, 'how actors can shape the institutions while being constrained by them' (Leca et al., 2008). We refer to institutional entrepreneur as 'actor who have an interest in particular institutional arrangement and who leverage organizational resources for policy change' (Garud, Hardy, & McGuire, 2007, p. 957). Institutional entrepreneurship is here identified through the analysis of the evolution over time of the content of the annual reports through which a certain IRA, under different presidencies, tell the story of the continuity or change of regulatory ideas, strategies and tools delivered for reaching specific policy goals. In organizational studies the institutional entrepreneur leverages resources to create new institutions (norms, rules and regulations) or to transform existing ones (Bakir, 2009;Garud et al., 2007;Mcguire, Hardy, & Lawrence, 2004;Mintrom & Norman, 2009).
There are various definitions of policy and institutional entrepreneurs used in the literature (see Bakir & Jarvis, 2017). Policy entrepreneurs can be identified by their efforts to set governmental agenda to promote ideas that they favour and capture key decision-makers when they couple problem, solution and politics streams (Kingdon, 1995). The institutional entrepreneur can draw upon existing rules, procedures, ideas and frames to narrate and theorize change in ways that provide other actors with reasons to cooperate and to produce change (Campbell, 2004;Garud et al., 2007;Schmidt, 2011). The success of the institutional and policy entrepreneur depends on his or her ability not only to influence and frame new practices and agenda setting and but also to seize opportunities for change (Kingdon, 1995;Mackenzie, 2004;Mintrom, 1995;Provost, 2003;Roberts & King, 1996). The term 'policy entrepreneurship' is functionally defined, but it is also used to indicate an ideational entrepreneur who promotes and disseminates regulatory ideas and, on behalf of the entire organization, communicates with the general public and with other actors involved in diverse regulatory fields with regard to what has been done and what future regulatory plans lie ahead (Gupta, 2009;Righettini, 2011). Institutional entrepreneurs provide public representation of changes to the institutional mission, of the mission's embedded values and regulative choices, and of decisions that are to be made or have already been made regarding regulatory strategies and instruments (Bakir, 2009). More recently, scholars have come to consider the role of single actors and of the heads of independent and semi-independent European agencies to be among those specific factors that can affect 'the developmental trajectory' of agency structure and policy over time (Bakir, 2009;Groenleer, 2007, p. 364). Institutional entrepreneurship can therefore affect the ways in which organizations represent or legitimate themselves externally, by stressing certain ideas (their political independence) over others (the political relevance involved in regulatory policy-making processes) (Groenleer, 2007;Righettini, 2011;Sbalchiero & Righettini, 2017). IRA presidential addresses are particularly relevant when institutional heads serve for long terms (for example, seven years in the Italian context), given that these longer terms allow us to understand whether any continuity -or change -identified is in fact due to presidential background and activities, or is due to other factors. Entrepreneurship also has a central role in the institutionalization of IRAs (Groenleer, 2007;Selznick, 1957), and contributes to the definition of social missions and to 'agency positioning' in the regulatory network with respect to the majority of relevant actors in the network, including other regulatory agencies (Groenleer, 2007, p. 367).
Thus institutional entrepreneurship here is defined both structurally and functionally. Structurally, as office holding, with exclusive accountability and communication tasks, and functionally, as a set of opportunities and resources to be activated (Bianculli, Fernandez & Jordana, 2015;Puppis, Maggetti, Gilardi, Papadopulos, & Biela, 2013;Righettini & Grimaldi, 2015). The IRA, and its presidency, can therefore act as entrepreneurs for institutional and policy change, and their communication and programmatic resources allow them to deliver written and oral public discourses through which to give an accounting of change (or lack thereof). In so doing, the IRA and its president promote or obscure certain values and ideas that have prompted and inspired future regulatory policies (Schmidt, 2010;Selznick, 1957). We apply a qualitative/quantitative analysis in order to study the endogenous structure-agency dynamic in the field of consumer protection policy through the annual reports issued by Italy's telecommunications Independent Regulatory Agency. Endogenous change in salience and content over time can be identified through the written annual reports delivered by IRAs and accompanied by the speeches given by the authority's presidents. We argue that this partial and limited insight into the IRA's activity could very well shed light on the dynamics of the interplay between institutional change and policy change over time. The IRA's annual report is thus a unique accountability and communicative tool that sums up and represents a set of goals regarding the policies, rules, activity, procedures, and strategies in the regulatory agenda.
In a previous study, we sought to discover whether changes in the personal characteristics of Agcom presidents cause changes (and if so, to what degree) in the discourse and content connected with consumer and user protection issues (dependent variable) (Sbalchiero & Righettini, 2017). Given that a change of presidency is a process whereby a new institutional entrepreneur enters the regulatory arena, we expected that his or her presence would lead to specific consequences in terms of the selection of ideas, strategies and policy instruments relating to consumer protection issues in the telecommunications sector over time. We found empirical evidence indicating that changes to the presidency were in fact accompanied by changes in the salience, attention and content of presidential programmatic discourse related to consumer protection issues (Sbalchiero & Righettini, 2017). In this article, we ask whether changes in presidency also lead to wider regulatory changes in terms of goals, strategies and tools. In other words, we ask if, and to what degree, the periodic changes to the Agcom presidency are reflected in the authority's annual accountability reports, and if those documents reveal changes in the degree of entrepreneurship, specifically with regard to the implementation of consumer protection policies.
Analysis of the texts of the annual accountability reports demonstrates the role institutional entrepreneurs play in fostering regulatory ideas, activities and policy strategies over time. IRA annual reports define the 'epistemological borders' and the 'repertoire of acceptable (and expectable) actions' (Schmidt & Radaelli, 2004:11) embedded within the agency's regulative policy. In this study, as previously mentioned, we analyse the content dynamic of Agcom public annual reports from 2000 to 2015, using a quantitative/qualitative approach. The accountability report is thus used to assess variations between institutional entrepreneurship and changes in regulatory ideas, strategies and tools.
The comparison of Agcom presidencies is taken as a case study of an IRA that is quite similar to those of the other European member states due to its structural isomorphism and the shared legal European supranational context in the TLC sector. Moreover, this particular choice of case study should allow us, moving forward into the future, to test hypotheses, methods and techniques in a comparative perspective.

Methods
Where, and how, can we ascertain and operationalize change that occurs over time in agency-structure relationships regarding consumer and user protection at the state level? The degree of institutional and policy entrepreneurship is operationalized here through the salience and attention given to the consumer protection issue empirically assessable by analysing the words appearing most frequently in IRA public discourses under different presidencies, a technique that allows us to at least partially resolve not only the problem of potential bias in policy historical process tracing but also that of identifying the complex variability of idea-related content over time. Our hypothesis is that it should be possible to assess variations between institutional entrepreneurship and policy change through the content of the accountability discourse, as evidenced by the language used in the authority's annual reports. This textual approach is not proposed as a stand-alone analysis, but rather could be a valid complement to historical and process tracing methods, thus providing higher degrees of reliability and additional evidence that may also be used to identify causal chains.
Thus in this study, to explore and to demonstrate the variations between policy change and institutional entrepreneurship in IRA accountability reports, we have applied statistical analysis to textual data in order to focus on the words, and categories of words, used by Agcom under different presidencies to develop ideas and strategies of consumer protection over time. Although textual analysis, involving scholars from different disciplines, has already proved fruitful in the analysis of numerous annual messages and political discourses (Bara & Weale, 2007;Bolasco, Galli, dé, & Giuliano, 2006;Cortelazzo & Gambarotto, 2013;Cortelazzo & Tuzzi, 2007;Schonhardt-Bailey, 2005;Tuzzi, Popescu, & Altmann, 2010), those studies have generally ignored a crucial aspect of DI, namely, how changes in institutional entrepreneurship may determine specific consequences in the selection of consumer-related policy issues. We examined in detail the content of the authority's annual reports, Agcom's main accountability tool, which are also directly related to the authority's annual presidential speeches, delivered to the Italian Parliament.
One methodological objection could conceivably be that public accountability discourse only registers what has already happened, and that therefore it is not a causal explanation of change but rather a result of change. To this objection, we can answer that, as rhetorical leadership studies demonstrate, communicative discourse in the regulatory arena often has additional normative and innovative effects. Moreover, governance studies demonstrate that the annual reports of IRAs with no democratic legitimacy are not only an opportunity to give an account of what has been done and what remains to be done, but are also the result of interactions and mediations with other policy actors, such as regulated interests and stakeholders (Bianculli, Fernàndez, & Jordana, 2015). Therefore, annual accountability reports have both communicative and programmatic content, which are delivered through and which may be analysed in order to identify variations between institutional entrepreneurship and policy change.
To that end, we examined the actual dimensions of the consumer protection issue as it was portrayed in the sixteen annual Agcom reports delivered between 2000 and 2015. From a methodological point of view, the analysis relies on the text analysis approach, an umbrella term referring to qualitative and quantitative perspectives (Bernard & Ryan, 1998;Creswell, 2013;Johnson, Onwuegbuzie, & Turner, 2007) and to the process of analysing and interpreting information embedded in a text (Tashakkori & Teddlie, 2003). The corpus was pre-processed using Taltac2 1 dedicated software (Bolasco, Baiocchi, & Morrone, 2010), 1 taltac2 is a software programme developed by a research team from the University 'la Sapienza' of Rome. and the textual data was processed with R 2 (R development core team, 2010) and Iramuteq 3 software (Ratinaud, 2009;Ratinaud & Déjean, 2009;Ratinaud & Marchand, 2012).
In any annual report, the specific topics relating to consumers can emerge anywhere in the text. Consequently, topic-based extraction and classification of portions of text often proves to be useful in pooling together all those segments referring to the specific topic to be analysed. Attempts aimed at the combined implementation of this procedure, involving the automatic extraction of portions of texts referring to the same topic, have already proven fruitful (Sbalchiero & Righettini, 2017;Sbalchiero & Tuzzi, 2016). Thus in the first step of the analysis, the algorithm split texts into Elementary Context Units (ECUs) of approximately similar length, generally sentences, or fragments of sentences, delimited by punctuation marks. The software then created a sub-corpus by extracting only those ECUs containing key words related to consumers (users, consumers, consumatori, consumatore, utente, utenti). The sub-corpus 'consumers' created in this fashion included 7912 portions of text (Elementary Context Units -ECUs), and contained N = 294,170 occurrences and V = 16,190 different words. The relationship between the corpus and the sub-corpus 'consumers' indicated the 'annual report coverage' of the 'consumer protection' issue. The effect of this selection procedure led to both a reduction of available textual data and to more homogeneous portions of text (ECUs), reducing the length of the corpus and discarding portions where annual reports dealt with other topics. This procedure not only allowed us to perform the analysis and to explore the role, meaning and evolution of consumer protection issues in the annual reports studied, but also permitted us to avoid researcher bias in the selection process.
In the second step, we applied Reinert's method (Reinert, 1983), consisting of the classification of portions of texts (ECUs) into categories, i.e. 'lexical worlds' (Reinert, 1990(Reinert, , 1993(Reinert, , 1999(Reinert, , 2001, semantic areas each having a specific vocabulary (essentially a set of words), identified without any a priori information (Sbalchiero & Tuzzi, 2016). The algorithm checked the occurrence and co-occurrences of words 4 in each ECU and reported results in a matrix (words × ECU). The occurrence and co-occurrence of words in the ECUs thus formed the basis used to assess similarity among units, summarized by means of a classification procedure. The main advantage of this step is the ability to quickly investigate the corpus and to thereby obtain an automatic classification of ECUs into a limited number of 'lexical worlds' , each of which depicts a particular semantic area. Again, this topic-based classification, providing an automatic assessment of the primary content, shielded us from analysts' bias (Sbalchiero & Righettini, 2017).
Seeking to identify variation across annual reports, and specifically any variation occurring in relation to changes of Agcom presidents, we then mapped the main content and words, thus creating lexical correspondence analysis graphs representing changes in Agcom's consumer orientations. Correspondence analysis (Greenacre, 2007;Lebart, Morineau, & Warwick, 1984;Murtagh, 2005) can be applied to a contingency table, in our case words x lexical worlds, transforming the frequencies of words into coordinates on a multidimensional 2 R is a popular language and environment for statistical computing available as free software under the terms of the Free Software Foundation's GnU General Public license in source code form. 3 iRaMuteQ stands for 'interface de R pour les analyses Multidimensionnelles de textes et de Questionnaires' [in english, 'R interface for multi-dimensional analysis of text and questionnaire'].
Cartesian axis system. This displays the lexical worlds (as semantic clusters of words) and words in a low-dimensional space by mapping an appropriate distance (the χ 2 distance) into a specific Euclidean distance (Murtagh, 2005) and, subsequently, on Cartesian planes (Sbalchiero & Tuzzi, 2016). The positions of these lexical worlds on the graph are determined by the degree of similarity in terms of their lexical profiles, described by the words of the same semantic area of the plot.

Empirical results. From presidential speeches to annual reports
In order to understand whether, and to what degree, changes in Agcom leadership affect changes in IRA consumer protection salience and issue definition over time, in a previous study (Sbalchiero & Righettini, 2017) we analysed the evolution of the rhetoric of the authority's annual Presidents' messages to the Italian Parliament delivered by Agcom presidents as 'obligatory' rhetoric (Smith, 2000, p. 82). Those previous analyses illustrated that, in a context of both stable formal independence from the central government and sufficient regulatory power, different actors play diverse ideational roles to promote certain ideas and policy programmes (Sbalchiero & Righettini, 2017). On the one hand, the distribution of the messages to Parliament depicted a clear chronological pattern, that is to say, the time variable appeared to be one of the most important factors constrained by rules, routines and ritual. On the other hand, as other scholars have demonstrated (Cortelazzo & Tuzzi, 2007;Tuzzi et al., 2010), despite the restraints deriving from the formal rules and the solemn nature of the messages, each Agcom President delivered (or failed to deliver) specific, unique programmatic content in the field of consumer and user protection, which proved individual entrepreneurship to be the most important factor, far more important than the time variable. This, in other words, signified that the marginally incremental accommodation of ideas in annual messages to Parliament over time is regularly affected by significant deviations occurring with every change in the Agcom presidency (Baumgartner, Jones, & Mortensen, 2014;Sbalchiero & Righettini, 2017). Lexical similarities and continuity between consecutive presidents were in part highlighted by those analyses, but during the period examined this continuity was somewhat reduced by the diverse topics and weights of policy-oriented speeches that mirrored differing policy ideas and choices with regard to regulatory agenda and targets, end users and overall goals (Schmidt, 2010). Although the main results of that study have already been discussed elsewhere (Sbalchiero & Righettini, 2017), it is worth mentioning that several different profiles emerged. In terms of regulatory ideas, goals and tools, in their addresses to Parliament Agcom presidents adopted different communication strategies, which reflected not only their diverse backgrounds 5 but also the different degrees and levels of engagement with consumer protection policy. Moving forward from those results, here, however, we analyse the authority's annual reports, which are not only Agcom's main accountability tools, but additionally are directly related to the agency's annual messages to Parliament. Given the previous results, our expectation was that the content of the annual reports would also reflect changes to the 5 Before becoming president of the agcom, enzo cheli was a professor of constitutional law at a number of italian Universities and served as President of the italian constitutional court; corrado calabro was an administrative magistrate, a member of the council of State, and lead a number of central Government Ministries. Marcello cardani was a professor of political economy at the Bocconi University, and as an economist has held many offices in the european commission, where he was member of commissioner Mario Monti's cabinet on the internal Market and was Head of commissioner Monti's cabinet on competition Policy. authority's presidency, similar to the changes found in the content of presidential speeches to Parliament. In order to assess the variations between change in agency leadership and regulative change, we specifically focused analysis on the levels of salience and content given to consumer protection issues in Agcom's annual reports. Applying computer-assisted content analysis to the annual reports issued between 2000 and 2015, we explored the most significant changes with regard to the topic of consumer protection. The category 'consumers' was composed of key words related to consumers (users, consumers, consumatori, consumatore, utente, utenti). In order to make an empirical statement about the association grade between the consumer category and annual reports, we used the chi-square correlation measure 6 in order to determine whether the consumer category in the authority's annual reports presented any significant differences over time (Figure 1). Figure 1 indicates a discontinuity in associations between the consumer category and annual reports, due to changes in the authority's presidency. This change also seems to be partially independent from the external pressures (Bakir, 2009) deriving from the European Union's Telecoms Package directives (2002) that both fostered and called for greater attention to and salience of a new European regulatory paradigm for consumer and user protection policies (Lodge, 2014). European consumer and user protection law in the TLC sector represents the most significant 'external event' or challenge to state member discourses and policy programmes in the field (Schmidt & Radaelli, 2004), the framework of which is necessarily mediated by domestic institutional filters and entrepreneurship (Campbell, 2004). This is particularly true in Italy, due to delays in its domestic consumer protection policies and to high levels of regulatory fragmentation.
During the period 2000-2004 (Cheli presidency) the consumer category is negatively correlated to annual reports, as Figure 1 demonstrates: given that there is an under-representation 6 the chi-square, measuring the divergence of the observed data from the values that would be expected under the null hypothesis of no association, shows the strength of association: if the correlation is greater than zero, then the consumer category is positively correlated (or directly related) to the annual report. if the correlation is less than zero, then it is negatively correlated (or inversely related). on the contrary, small values (or zero) do not support a significant association, i.e. observed counts and expected counts would be similar. of consumer-related topics in annual reports issued during these years, users and consumers are not yet in the institutional agenda or targets of specific policies. During the period 2004-2012 (Calabrò presidency) however, the figures change, presenting values that are close to zero, suggesting that the consumer protection issue is constant but that there was neither under-representation nor over-representation of consumer-related issues in the annual reports from that period. Finally, during the most recent period 2013-2015 (Cardani presidency), we observe a significant positive association, indicating the increasing representation of the consumer category in annual reports, due to its greater importance in the Agcom institutional agenda. This very interesting clue led us to continue with the analysis of this matter by exploring the depiction of consumer protection, in terms of content, meaning and evolution, in the annual reports released by the authority. While Figure 1 shows the occurrences of a set of words (category), by measuring its association with the sixteen Agcom annual reports delivered between 2000 and 2015, in Figure 2 we measure the contexts in terms of total 'space' dedicated to the topic in annual reports, that is say, where the category 'consumers' appeared. In the first case, the unit of analysis is the overall category of words, in the second case, however, we utilized fragments of texts (ECUs), which are sentences, or fragments of sentences, i.e. portions of texts containing key words relating to consumers (users, consumers, consumatori, consumatore, utente, utenti). Figure 2. coverage of 'consumer protection' issues in agcom annual reports (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). Source: annual reports issued by the italian communication authority agcom (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). note: each year (2000-2015) corresponds to an annual report issued by the italian communications authority (agcom).
Borrowing from the concept of 'media coverage' (Maggetti, 2012), we refer to the 'annual report coverage' to indicate the degree of attention given to the issue of user and consumer protection as compared to the total of all regulatory issues dealt with by the annual reports.
Thus the attention given to the issue of consumer and user protection in the telecommunication sector has clearly increased, as evidenced by the fact that the annual report coverage for the topic of consumer protection has grown over time in the Agcom reports studied. As shown in Figure 3, in the first annual report only 6.08% of the corpus related in any way to consumers, while over the following years coverage increased steadily, reaching overall average values of above 7.44%, and amounting to 11.89% in the last annual report studied (2015).  (2000)(2001)(2002)(2003)(2004). 'consumer protection: an idea without a specific policy agenda' . Source: sub-corpus 'consumer Protection' extracted from the annual reports of the italian communication authority agcom (chair-person cheli, 2000-2004).

Changing ideas and regulation in consumer protection policy: striking similarity, continuity and change
Close examination of the content of the annual reports permitted us to better focus on the micro-macro interactions of ideational entrepreneurship. The results revealed that the authority's presidents do in fact engage in institutional entrepreneurship, effectively shaping policy change. The empirical study of the content of annual Agcom reports 7 allowed us to identify three distinct stages of change in Agcom's consumer protection policy over time, and, during those periods, three equally different ways of promoting and including these issues in the authority's regulatory agenda, which also evolved markedly over time. While during the authority's first presidency (2000)(2001)(2002)(2003)(2004) the focus was on strengthening Agcom's constitutional framework and functional legitimacy, giving only a marginal role to the issue of consumer protection in the institutional agenda, during the second presidency (2005-2012) regulative attention to consumer issues began to increase. However, it has only been during the current presidency that consumer protection has become a priority in the institutional agenda (Figure 3).
Diverse semantic areas (lexical worlds) emerged 8 as a result of the analysis of the sub-corpus 'consumers' during the Cheli presidency (2000)(2001)(2002)(2003)(2004) (Figure 3). In particular, the semantic areas referred to included widely varied authority functions such as protection (tutela, protezione), transparency (trasparenza), reporting (segnalazione), legislation (normativa), supervision (vigilanza), control (controllo) and monitoring (monitoraggio). The topic of consumer protection (consumatore) was strictly correlated to the role of and the services delivered by providers: in fact, the word consumer is quite close to the centre of graph, and mixed with other issues such as broadband (banda larga), mobile cell (cellulare), and digital (digitale), but it is distant from the words authority (autorità) and protection (tutela). This distance indicates that the authority has not yet taken up the goal of consumer protection as one of its strategic regulative priorities. The Cheli presidency was in fact more focused on the authority's legitimacy and certification processes, and with the overall aim of strengthening Agcom's functions and activities in relation to other policy actors and the telecommunications market. Consumer protection, during Mr. Cheli's term, was not considered to be a specific or central issue in the regulation agenda, also because of the relative weakness of consumer associations throughout that period. It was not until 1998 that national Italian law for the first time transposed the principles contained in European legislation, recognizing individual and collective consumer rights and promoting collective and individual consumer protection. The structure of Agcom's annual reports during Mr. Cheli's presidency was fairly standardized, and basically focused on the following directions: presentation of the global communications system; developments and trends in the various communications sectors (telecommunications, telephony, broadcasting, publishing); the evolution of regulation at state and international levels; domestic inter-institutional 7 the results discussed in this paragraph are those of the analysis of sub-corpus 'consumer protection' , extracted from agcom annual reports (2000-2015) (Figure 2). 8 a semantic area is composed by groups of words which share a common meaning or which relate to the same subject in the corpus. to facilitate the interpretation of graphs, different circles indicate different semantic areas as identified by the software utilized (Reinert's method). Moreover, words placed in the same quadrant of the graph reflect the same semantic area, and the proximity of words indicates the level of association with the ecU, i.e. the same words which appear with a similar frequency in the same ecU are close to each other in the quadrant. correspondence analysis was applied to the original texts in italian. However, in order to facilitate the reading of this text, the italian utilized words are in brackets after the english translations.
relations, and the organizational development of the authority. Organizational issues were particularly relevant in those years, due to the institutionalization of the internal structure (sectors and services) in 2002. As regards consumers, in the annual public reports issued during the Cheli presidency, the identification of the consumer was quite generic. Consumer protection issues were not well differentiated from other important issues, despite the fact that in 2003 Agcom signed an important agreement with the twenty Italian Regions establishing decentralized offices (given the name Co.re.com) for dealing with alternative dispute resolution (ADR) in the telecommunications sector and to resolve the increasingly diffuse conflicts between operators and phone service consumers (fixed and mobile).
Moving from the Cheli to the Calabrò presidency (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012), a number of regulatory issues, such as Internet, mobile phones (cellulare), universal service (servizio), access (accesso), phone services (telefonia), television (televisione), users (utente), grew increasingly important in the Agcom agenda, and the annual reports issued during those years reflect this change (Figure 4). In this period, the salience of the consumer protection issue in the telecommunications sector emerged, resulting in increased attention by the authority to the final beneficiaries of regulation, as evidenced by similarly increased use of terms indicating both users (tutela dell'utente) and consumer associations (associazioni dei consumatori). Moreover, specific activities and functions in this field were developed and pursued: supervision (vigilanza), monitoring (monitoraggio) and inspection (controllo). The structure of the first two annual reports (2005 and 2006) perfectly mirrored those of the previous presidency (Cheli). The report issued in 2007, however, marked a radical change with regard to consumer issues. A greater salience of the issue can be identified, both from a change on the Figure 4 for this period and from the structure of the overall annual reports issued, which included a new section entitled 'The relationship with consumers and users' where, for the first time, the authority stated that its fundamental mission was to protect users and consumers (Annual Report 2007, p. 193). The authority thus publicly acknowledged that the relationship between providers and users was always unbalanced in favour of the former; consequently, control and litigation tools available to all consumers at low cost were decidedly in favour of operators, and Agcom additionally acknowledged that it was necessary for it to take charge of such activities. Thus, the authority's entrepreneurship with regard to consumer protection policy emerged, both in pursuit of the public interest and with the aim of empowering consumer problem-solving capacities. Agcom entrepreneurship can also be measured by its ability to communicate with and to respond to the needs of citizens, and by the rapidity and effectiveness of interventions in correcting market distortions. In October 2006, the authority established a new department for consumer protection and a permanent consultation with user associations, and has since then engaged in a continuous and constructive dialogue with consumers and the National Consumer Council (CNCU), aimed at facilitating and accelerating the authority's interventions in favour of consumer and user protection. The authority 'recognizes the unique role of consumer associations, as important and strategic allies of the authority's institutional activities' (Annual Report 2007, p. 193). In 2009, Agcom introduced another new chapter 'Instrumental and auxiliary bodies of the authority' into its annual reports. This marked an important turning point, following the delegation to the Regions and the autonomous Provinces of the ADR, in that it aimed to empower consumers and to give them more effective instruments for self-protection. Increased ADR activities, after 2007, were due to the approval of new legal rules and procedures facilitating access to ADR processes.
During the current Cardani presidency (2012-2017), consumer protection has become one of the authority's most salient issues, and a clear priority in the Agcom agenda. Consumer and user protection is now defined as public general interest and not only as individual interest to be protected through the operator contracts. The semantic lexical field of regulation reveals that the terms consumers (consumatori), consumer protection (tutela dei consumatori), and authority (autorità) have been extended to the postal sector as well, due to the privatization of postal services and the new role of Agcom in the field of communications ( Figure 5).
This increasing salience is reflected in the semantic area, where the topic 'consumers' assumes a central position in the very ambit of protection of the authority: in Figure 5, consumers (consumatori) are very close to protection (tutela), authority (autorità), guarantee (garanzia) and rights (diritti).
In 2013, with the transition to the Cardani presidency, the annual reports became much more complex, in continuity with previous reports, but at the same time introducing significant variations. In the annual reports for that year, visibility of the issue of consumer protection was strengthened through the inclusion of a more full-bodied and complex chapter entirely dedicated to the matter. Additionally, and related to the increasing attention to the issue, the authority re-organized its Department for Consumer Protection, creating three distinct offices (the Office of Disputes and Sanctions, the Office of Management Reporting and Supervision, and the Office of Quality, Universal Service and Relationships with Consumer Associations). This department has continued to implement the activities identified under the Memorandum of Understanding signed in 2008 between the authority and the consumer association members of the National Council of Consumers and Users. Since 2012, the department has also been working to create a new 'code of consumer rights for the electronic communications services market' , which is particularly important for the protection of user and consumer rights. At the end of the content definition process the Figure 5. the cardani Presidency. 'a policy paradigm: consumer protection perceived as public interest' . Source: sub-corpus 'consumer Protection' extracted from annual reports of the italian communication authority (chair-person cardani, 2013-2015). new code will be submitted to public consultation for comments and contributions from stakeholders (users, consumer groups, traders) (Annual Report 2012, p. 327). Consistently with the existing European legislation, the theme of consumers and their protection in TLC markets is thus expanding markedly, and affects regulators, regulated entities and end users (i.e. consumers), as well as a broader constituency of citizens and stakeholders through public consultation.

Discussion and conclusions
This article explored institutional entrepreneurship within IRAs as a variable for explaining change in policy mix over time and offered a new approach to identify agency in institutionalization processes. The case of consumer protection policy in the TLC sector indicates that changes in Agcom leadership are linked to changes in regulatory ideas, strategies and instruments used to pursue policy goals and to stabilize their implementation. In fact, our research indicated that different presidencies within Agcom have had different degrees of commitment over time that not only modified the salience of the issue of consumer protection but that also corresponded to a completely different range of tools and resources. During the first term, Agcom developed a symbolic discourse on consumer protection, mainly referred to operator policies and with an initial commitment to regional strategies but with no reference to specific instruments. The second term displayed far more instrumental discourse and entrepreneurship leveraged on authority decentralization, seeking to enforce individual consumer protection rights, and finally, under the most recent presidency, discourse and entrepreneurship has become increasingly paradigmatic. The current president now pursues a new regulatory paradigm, based on the implementation of ad hoc legal resources (Communications Code of 2003, adjourned in 2013) and a higher degree of visibility of consumer protection, meant as general public interest to be defined and reached through public consultations. Over the three periods examined, the agency's public discourse under the three different agency presidents has demonstrated a marked shift in policy paradigm, passing from a self-protection model, based on ADR and price cap tools, to a public interest protection model in which public debate co-defines regulatory strategies and specific codes of rules are universally applied. In fact, the current Agcom presidency has drawn attention to the European regulatory framework, promoting a new normative discourse on how relationships between regulatory authority and its context should be shaped in the field of consumer protection.
The empirical evidence obtained from this study indicates that over time, in a stable constitutional framework, individual entrepreneurs -IRA Presidents -exert different degrees of influence and abilities to seize opportunities for change, acting as internal drivers of regulative change, modifying the salience of diverse issues, and drawing increased attention to different problem definition strategies, resources and instruments for turning the problem of consumer protection into a solution. The results we obtained clearly indicate that, on the one hand, different types of entrepreneurship contributed to selecting ideas and values that inspired regulation according to different stages of Agcom institutionalization. On the other hand, the institutional entrepreneurs provided other actors and IRA stakeholders with periodic accounting of the work performed, activities completed and goals which still need to be implemented, recounting changes on the ground regarding regulatory goals, strategies and tools adopted. The annual reports analysed revealed the evolution over time of regulatory salience and content in the field of consumer and user protection.
The method used deserves special consideration. The analysis of textual data permitted us to notice certain aspects that would otherwise in all likelihood have gone undetected. In particular, it has been very useful to apply statistical methods to this type of corpus and sub-corpora of texts to quickly investigate trends, continuities and changes over time in policy goals, strategies and tools in a specific regulatory field. The method used for this study can in fact reinforce and support the traditional qualitative policy process analysis, with little or no interference of analyst bias, given that all 'data crunching' is performed by computer. The quantitative/qualitative content analysis of consumer protection policy reflects the evolution of regulatory entrepreneurship, and marks when and where the major changes occur. By focusing on consumer and user protection, a topic in which political delegation is fairly stable over time and not shared with other institutions, we were able to fully appreciate the role that the institutional entrepreneur can play in the institutionalization of consumer protection policy in the TLC sector (Righettini & Nesti, 2014). The authority's annual reports indicated a variable definition of the content of the consumer protection issue over time, and a variable degree of attention and relevance given to this issue not only by the authority's presidents but also by the authority as a whole. As we expected, the chronological evolution of topics relating to consumers over time clearly demonstrated variations between the change in Agcom's presidency and changes in consumer protection policy. Moreover, this study has allowed us to confirm the validity of the institutional entrepreneurship approach in highlighting the relationship between changes in institutional entrepreneurship and policy-mix change. The method used and the confirmation of the entrepreneurial role played by some of the IRA's presidents requires further verification by comparing different policy sectors or by comparing the Italian case with other similar case studies within the same policy area. We believe that this path could lead to interesting empirical and theoretical developments in the institutional entrepreneurship approach.