THE MODERATING ROLE OF FAMILY OWNERSHIP ON THE RELATIONSHIP BETWEEN OWNERSHIP CONCENTRATION AND COMPLY-OR-EXPLAIN DISCLOSURE: AN ANALYSIS ON ITALIAN LISTED COMPANIES

The moderating role of family ownership on the relationship between ownership concentration and comply-or-explain disclosure: An analysis on Italian listed


INTRODUCTION
Corporate governance (CG) code represents a form of soft law useful to provide firms with a voluntary mean for improving CG practices. The need for good CG codes mainly arises from the separation between ownership and control, which generates agency conflicts (Berle & Means, 1932;Fama & Jensen, 1983).
Since the first code of good governance was issued in 1978 in the United States, many countries developed codes of good governance to remedy weaknesses in CG practices (Enrione, Mazza, & Zerboni, 2006), sharing best practices and increasing the legitimation of national firms in the international context. These CG codes have been intended as sets of best practice recommendations (Aguilera & Cuervo-Cazurra, 2004;Rose, 2016) and have been guided by the comply-or-explain principle (Haxhi & Van Ees, 2010;MacNeil & Li, 2006;Seidl, 2007) established by the Cadbury Committee in the UK and adopted by the European Union (EU) through latest CG code was issued in July 2015. Since 2007, listed companies have been required to draw up a CG statement, either as a part of the directors" report or as a separate document, to disclose all of the essential information about CG practices to investors (art. 123-bis Legislative Decree 58/1998, introduced by Legislative Decree 229/2007). This report, based on the flexibility of the comply-orexplain principle, states whether a company has complied with the provisions of the CG code or not, and it provides specific explanations in cases of noncompliance. With regard to the explanation to provide in cases of non-compliance, the code refers to the European Commission Recommendation 2014/208/EU on the quality of CG reporting.
This study has two aims. The first is to assess the degree of comply-or-explain disclosure regarding the composition and functioning of boards of directors provided by Italian listed companies. To achieve this aim, we developed a Comply-or-Explain Disclosure Index and content analysed the CG statements issued by firms. Differently, from most previous studies, our index considers both the comply and the explain aspects, examining in depth the underexplored issue of "explanation". The majority of the literature has mainly focused on the compliance aspect (see, e.g., Alves Talaulicar, & Kolot, 2005). The second aim of this study is to investigate the relationship between ownership structure and the level of the Comply-or-Explain Disclosure Index to understand whether ownership concentration affects the degree of comply-or-explain disclosure. According to previous research (i.e. Lepore, Paolone, Pisano, & Alvino, 2017), each CG mechanism provides its own effectiveness when interacting with other internal and institutional governance mechanisms. Therefore, to evaluate the effectiveness of code as good CG mechanism, it becomes essential to analyse its interaction with other governance variables, particularly ownership concentration. The latter, in fact, acquires increasing importance as CG control mechanism (Connelly, Hoskisson, Tihanyi, & Certo, 2010): interacting with other CG mechanisms, ownership concentration can influence CG effectiveness in protecting shareholder rights (Demsetz & Lehn, 1985;Shleifer & Vishny, 1997). We also investigate the moderating role played by family-firm, as the largest shareholder, in the mentioned relationship. Thus, we exceed the issue of ownership concentration, which has already been examined extensively, raising the issue of shareholders" identities, which remains underexplored. In this way, our results contribute to explaining some causes of the diverse findings that research has found about the relationship between ownership concentration and voluntary disclosure, demonstrating the importance of the largest shareholder"s identity.
We conducted an analysis on a sample of 227 companies listed on the Italian stock exchange at the end of 2016. To collect data on the degree of comply-or-explain disclosure, we investigated the CG statement issued for the year 2016 using content analysis. In particular, we exclusively analysed the section concerning the composition and functioning of the board of directors. To test the relationship between ownership structure and the level of comply-or-explain disclosure, we developed an OLS regression. To verify and quantify the moderating effect of family-firm as the largest shareholder in the mentioned relationship, we added an additional moderator term to the model, affecting both the direction and strength of the relationship. We found that the sampled companies tend to comply with the majority of the recommendations concerning the composition and functioning of the board of directors and disclose this information; however, when they decide to not comply, the sampled firms do not provide adequate explanations according to Recommendation 208/2014. The analysis also shows a negative relationship between ownership concentration and the degree of complyor-explain disclosure. Moreover, when there is a family firm as the largest dominant shareholder, the mentioned relationship is stronger. This means that, when self-regulating initiatives are designed and implemented, legislators, regulators and managers should not ignore the characteristics of the firms" ownership structure.
The remainder of the paper is organized as follows. The next section reviews the literature to form the basis for hypotheses development. Section 3 describes the sample selection process and research design. Section 4 reports and discusses the results of the study. The final section presents the conclusions, the main limitations and future perspectives of the study.  The degree of compliance has increased during the years. The explanations are of variable quality. Rose (2016) Denmark 2010 155 listed firms Dummy variable: 1 if a company complies with a recommendation, 0 otherwise. A company deciding not to comply with a recommendation receives a score of 1 if releases an explanation, 0 if it does not disclose the reason for not follow the recommendation.

LITERATURE REVIEW AND HYPOTHESES
The majority of firms comply or adequately explain the decision to not follow a recommendation. There is a positive relationship between the level of complying or explain disclosure and firm performance. Companies tend to comply with the CG code but, when they decide to not comply, they do not provide adequate explanations. There is a negative relationship between the ownership concentration and the level of comply-or-explain disclosure. The presence of a dominant financial shareholder at a high level of ownership concentration creates inefficiency of the degree of adherence to the comply-or-explain principle.
The main findings of previous studies show that overall adherence to CG code recommendations is relatively high. However, the level of compliance with CG codes varies significantly across countries. For example, in the UK, Pass (2006) showed that British listed firms to a large extent complied with the Cadbury Report"s recommendations. Both Arcot et al. (2010) and Shrives and Brennan (2015) noted that the scale of compliance with the CG code has increased over time. Additionally, Alves and Mendes (2004) in Portugal and Tariq and Abbas (2013) in Pakistan showed that the average degree of compliance has increased over time. In Germany, Werder et al. (2005) found that firms exhibited a high degree of acceptance of code recommendations. Similar results were found in Denmark, where Rose (2016) showed that most companies comply with CG code recommendations, mainly with respect to risk management and internal controls and the responsibilities of board recommendations. Additionally, in the Netherlands Akkermans, Van Ees, Hermes, Hooghiemstra, Van der Laan, Postma, and Van Witteloostuijn (2007) found that firms exhibit a high level of compliance with the code, and in Italy Lepore et al. (2018a) showed that Companies tend to comply with the CG code recommendations concerning the composition and functioning of the board of directors. In contrast, Bianchi, Ciavarella, Novembre, and Signoretti (2011) showed that the level of effective compliance with the Related Party Transactions principle of the CG code by Italian companies is low, and in Greece Nerantzidis (2015) found that the degree of compliance with code recommendations is low.
Most of the literature mainly focused on the compliance aspect (see, e.g., Sanderson  Similar to the comply aspect, the quality of the explanations provided varies significantly across countries. For example, in Denmark Rose (2016) found that most companies adequately explain why they have decided not to follow a specific CG code recommendation. In contrast, in the Netherlands both Akkermans et al. (2007) and Hooghiemstra and Van Ees (2011) reported that both the nature and the content of the explanations provided are remarkably similar across companies, indicating symbolic compliance with the code"s best practice provisions. In the UK, Arcot et al. (2010) showed that companies released no or few explanations and tended to use poor language when explanations were given. Additionally, in Italy Lepore et al. (2018a) found that non-compliant companies do not provide adequate explanations. Similar results were found by Nerantzidis (2015) in Greece.
However, it is important to emphasize that previous differences found across countries, both in the level of compliance and in the quality of the explanations, could be the consequence of the different manners in which scholars assessed the comply-or-explain disclosure. First, the authors used different CG codes to identify the recommendations with which companies should comply, depending on the country in which the study was conducted. Moreover, the methods used to investigate the level of comply-or-explain disclosure, i.e. content analysis, inevitably require some qualitative judgement and necessarily contain some subjectivity.
In summary, the findings of previous studies have suggested that soft law can be an efficient means of increasing the quality of CG statements among listed companies; however, code authorities should be more active in strengthening the explanations provided in cases of non-compliance, as well as move towards better monitoring of the information disclosed (Sergakis, 2013). In fact, until now companies have mainly emphasized the comply aspect, placing less importance on the explanations of non-compliance. Thus, the explain part of the code is largely ineffective. As mentioned in the European Commission Green Paper, "the large majority of respondents were in favour of requiring companies to provide better explanations for departing from codes" recommendations" (European Commission, 2012, p. 6). As a consequence of previous results and to increase the quality of the explanations provided, the European Commission issued Recommendation 2014/208/EU on the quality of CG reporting, requiring companies to release detailed information in cases of non-compliance.
Moreover, although the comply-or-explain principle should be applied to all companies and for all countries to be identified as a good governance practice (Nerantzidis, 2015), it is often differently applied in each country (Andres & Theissen, 2008;Hooghiemstra, 2012;Nerantzidis, 2015), and CG practices are not uniform across nations (Davies & Schlitzer, 2008, p. 532). This difference occurs because each country is characterized by its own legal and financial system, corporate ownership structure, cultural and economic situation; and these differences among countries could influence the manner in which the comply-or-explain principle is applied.
Differences among countries are accompanied by differences among firms because the usefulness of each recommendation, and thus its adoption, depends on how the management and the board of directors "feel their own" recommendations (Mintz, 2005).
Several factors, at both the firm and country levels, can affect the manner in which the comply-orexplain principle is applied.
There are various studies focusing on the determinants of the level of comply-or-explain disclosure (e.g. Akkermans  In this paper, we focus on the relationship between ownership structure and the level of comply-or-explain disclosure. Moreover, considering that the effectiveness of a particular governance mechanism, such as ownership concentration, could depend on interactions with other governance mechanisms (Zattoni, Witt, Judge, Talaulicar, Chen, Lewellyn, & Shukla, 2017), we investigate whether previous relation is moderated by the presence of a family as a dominant shareholder.

The relationship between ownership concentration and comply-or-explain disclosure
Voluntary disclosure is a useful CG mechanism to mitigate the agency problem between competing shareholders and managers (e.g. Oliveira Research often used agency theory to investigate two specific hypotheses: the monitoring hypothesis and the expropriation hypothesis. According to the monitoring hypothesis, voluntary disclosure is a mechanism for monitoring the controlling shareholders or management activities. In fact, through voluntary communication the controlling shareholders or management can signal acting on a long-term perspective or in the best interests of all owners, and other shareholders can use this disclosure to efficiently control the activities of agents ( The importance of the agency problem is not the same in concentrated and widely held firms. As Fama and Jensen (1983) proposed, the potential for conflicts between principals and agents is greater for firms characterized by high ownership diffusion than for more closely held companies. Consequently, the amount of information disclosed by companies to mitigate such conflicts is likely to be greater in widely held firms. When the level of ownership concentration is low, more monitoring is required (Fama & Jensen, 1983). In contrast, firms with concentrated ownership are characterized by less information asymmetry between the management/dominant shareholder and other shareholders, because large shareholders typically have access to the information that they need and can provide an active governance system, which is difficult for smaller and less-informed investors (Cormier, Magnan, & Van Velthoven, 2005).
In such a context, the research usually suggested and found a negative relationship between ownership concentration and voluntary disclosure, supporting the expropriation hypothesis (e.g. Alsaeed (Shleifer & Vishny, 1986). Finally, there are also studies failing to find a statistically significant relationship (Mak, 1991;Craswell & Taylor, 1992;Raffournier, 1995;Eng & Mak, 2003;Donnelly & Mulcahy, 2008). In synthesis, there is no convergence between the sign and form of the relationship between ownership concentration and voluntary disclosure. As a consequence, corporate disclosures voluntarily provided by firms, for example, the comply-or-explain disclosure on CG code recommendations, provide an interesting field of analysis for researchers to test the competing expropriation and monitoring hypotheses.
Research into the relationship between complyor-explain disclosures and ownership structure, to the best of our knowledge, has to date been quite minimal. Warning (2011) analysed the relationship between shareholder concentration and compliance with the CG code in Germany and found an inversely U-shaped relation. In other words, firms with very low or very high ownership concentration had a small probability of compliance with the CG code, while a moderate concentration led to a higher degree of compliance. Samaha, Dahawy, Hussainey, and Stapleton (2012) found that the extent of CG voluntary disclosure in Egypt was lower for companies with higher ownership concentration. Vander Beauwhede and Willekens (2008), analysing the relationship between the level of CG disclosure and the percentage of shares closely held for the FTSE Eurotop 300, found that the ownership concentration negatively influenced the level of disclosure. In contrast, Barako (2007)  Although the majority of disclosure research sustains that shareholders prefer more disclosure of timely information, there are studies documenting the presence of different signs of the association between ownership concentration and voluntary disclosure, it depends on the dominant shareholder identity. Some categories of shareholders could generate an incentive to disclose more information (i.e. financial shareholders) or less information (i.e. families).
With specific regard to family-controlled companies, previous studies (Chau & Gray, 2002, 2010Ho & Wong, 2001) mainly found a negative relationship between family ownership and the level of disclosure. The underline rationale is that familycontrolled firms have little motivation to disclose information in excess of mandatory requirements because the demand for public disclosure is relatively weak in comparison with companies that have wider ownership (Chau & Gray, 2002). This is because dominant family shareholders, who frequently are also board members, have direct access to the firm"s financial and non-financial information and, as a result, have less need for disclosure.
However, there are studies demonstrating that previous negative relationship is not always confirmed. Chen, Chen, and Cheng (2008), for example, found that family firms provide more or less voluntary disclosure; it depends on the type of information to release. More specifically, the authors reported that family firms exhibit a lower likelihood of providing management forecasts than non-family firms, but they are more likely to issue bad news earnings warnings. Chau and Gray (2010) found that the negative relationship between family ownership and the disclosure level is not linear in form when splitting the levels of family shareholding at less than 5%, between 5% and 25% and greater than 25%: at moderate and low levels of family shareholding (25% or less), the relationship is negative, meaning that companies provide low levels of voluntary disclosure; at higher levels of family shareholding (more than 25%), the association is positive, meaning that the extent of voluntary disclosure is higher. The author explained this different disclosure behaviour using the convergence of interest and management entrenchment hypotheses suggested by the finance literature (Morck, Shleifer, & Vishny, 1988). According to the convergence of interest hypothesis, when the ownership concentration increases, the need for disclosure decreases. This is the consequence of a convergence of interests between dominant shareholder and minority shareholders so that the former is more unlikely to act opportunistically. In these situations, the agency conflict between majority shareholder/management and minority shareholders decreases as ownermanagers" ownership of the firm increases. As a consequence, the relationship between ownership concentration and disclosure is negative. On the other hand, according to the management entrenchment hypothesis, when the ownermanagers" ownership of the firm increases, also the agency conflict between majority shareholder/management and minority shareholders enhances. This is because the dominant shareholder could become entrenched and pursue policies that are in its own best interests (Kumar & Tsetsekos, 1993;Wright, Ferris, Sarin, & Awasthi, 1996). In this situation, dominant shareholder/management decisions will likely be made to benefit the personal wealth of the dominant shareholder, and expropriation of minority shareholders by the controlling owner could occur (Fan & Wong, 2002; La Porta, Lopez-de-Silanes, & Shleifer, 1999). As a consequence, minority shareholders will increase their monitoring of dominant shareholder/management decisions. A possible way to reduce this monitoring activity is to disclose a greater amount of information. Thus, in the management entrenchment hypothesis, the relationship between ownership concentration and disclosure is positive.

Hypotheses development
This paper revisits the issues developed above in a specific country, i.e. Italy, with regard to disclosures on comply-or-explain. We attempted to overcome the aforementioned limitations of other studies considering an important shareholder type: the family.
The Italian setting is of interest because of its low level of investor protection, prevalence of family firms, high ownership concentration, and poorly developed capital market (Lepore et al., 2017 and 2018b), whereas prior research has often been conducted in countries with a relatively dispersed corporate ownership structure, a more efficient legal system and more developed capital markets.
In the Italian context, the risk that firm value will be expropriated by majority shareholders from minority shareholders is particularly significant, due to the high ownership concentration, the low counterweight power of minority shareholders and the low level of shareholder protection due to judicial system inefficiency 2018b).
We expected that the distribution of power among shareholders would play a pivotal role in influencing firms" disclosure about the principle of comply-or-explain. Therefore, we tested the following hypothesis: H1: Ownership concentration negatively affects the degree of comply-or-explain disclosure.
Moreover, we hypothesized a relevant role for a family owning a significant share of a firm"s control. We assumed that the presence of a dominant family shareholder should play a pivotal role in the relationship between ownership concentration and disclosure and we tested the following hypothesis: H2: The relation between ownership concentration and the degree of comply-or-explain disclosure is moderated by the presence of a dominant family shareholder.

Sample
The sample consisted of 227 companies listed on the Italian stock exchange on 31 December 2016. Table 2 summarizes the sample selection process. Our original sample was composed of 237 firms, i.e. all of the companies listed on the Italian stock exchange on 31 December 2016. We excluded 10 of these companies from further analysis because we could not find their CG statements and/or accounting and ownership structure data.  Table 3 presents the distribution of the 227 companies across the three sectors. In the final sample, services and industrial had the largest proportion of companies (34,8%), followed by financial (9,2%), the remaining part of the firms is in the others category (20,7%).
We gathered accounting and ownership structure data from the Amadeus database, which was created by Bureau Van Dijk, as well as information about the degree of effective adherence to CG code from the CG statements issued for 2016 by each company. In particular, we exclusively analysed the section concerning the composition and functioning of the board of directors. We decided to investigate the CG statements drawn up for 2016 because companies had the latest CG code, issued by Borsa Italiana in July 2015.

Dependent variable
To collect data on the dependent variable, i.e. Comply-or-Explain Disclosure Index (CoE_DI), we used content analysis (Krippendorff, 2013). The coding procedure was organized as follows. We first identified the recommendations referring to the composition and functioning of the board of directors discussed in the CG statement, using the CG code issued by Borsa Italiana in July 2015 and the "format for the CG statement" issued by the Italian stock exchange in January 2017. This format has been developed to help firms in drawing up their CG statement according to the code recommendations.
We identified 54 recommendations and we grouped these recommendations into 7 categories (see Table 4). To identify the 7 categories, we specifically referred to the "format for the CG statement". Once the recommendations were identified, the CG statement of each sampled company was analysed, and data were collected for each recommendation. The analysis was performed by two associate professors. Sentences were selected as the recording units because they are considered the most reliable unit of analysis (Milne & Adler, 1999), despite the potential difficulty scholars may encounter in allocating information to only one category (Beattie & Thomson, 2007). All of the companies were assessed based on how they communicate their governance practices. More specifically, we assigned a score of 1 to a specific recommendation if the company complied with it and disclosed this information and a score of 0 if the firm did not comply with this recommendation. Thus, the "comply indicator" could range from 0 (no compliance with any of the provisions identified exante) to 54 (in cases of absolute compliance with all of the provisions identified).
In cases of non-compliance with a specific provision, we investigated the quality of the explanation released. We assigned a score ranging from 0 to 1 on the basis of the explanation provided.
To assess the quality of the explanation, we referred to art. 8 of the European Commission Recommendation 2014/208/EU on the quality of CG reporting. According to this article, if a company decides not to comply with a specific recommendation, it must release the following information: 1) the manner in which the company has departed from a recommendation; 2) the reasons for the departure; 3) how the decision to depart from the recommendation was made within the company; 4) if the departure is limited in time, when the company envisages complying with a particular recommendation; and 5) where applicable, the measures taken instead of compliance and how these measures achieve the underlying objective of the specific recommendation or of the code as a whole or how it contributes to good CG of the company.
We assigned a score to each explanation provided. More specifically, we assigned a score of 1 if a company disclosed all of the information required by Recommendation 208/2014 (points 1-5). Otherwise, a score of less than 1 and ranging between 0 and 1 was assigned. More specifically, we assigned a score of 0 when no explanation was provided, a score of 0.20 to companies that released explanations for one of the five points identified by Recommendation 208/2014, a score of 0.40 to firms that provided explanations for two of the five points, a score of 0.60 to companies that released explanations for three of the five points, a score of 0.80 to firms that provided explanations for four of the five points, and a score of 1 to companies that released explanations for all of the five points. The "explain indicator" could also range from 0 (no explanation for any of the provisions identified exante) to 54 (in cases of explanation adherent to Recommendation 208/2014 for all of the provisions identified).
Our final dependent variable was the sum of the "comply indicator" and the "explain indicator". Before adding together these two indicators, we converted these scores into indices, dividing the score assigned to each firm by 54 (the maximum value that a company could obtain if it is fully compliant with all of the recommendations identified or if it provides all of the explanations required by Recommendation 208/2014). As a consequence, we computed our dependent variable CoE_DI as follows: The CoE_DI awarded to firm i is equal to the sum of the "comply indicator" obtained by company i divided by 54 and the "explain indicator" assigned to firm i divided by 54.
To verify the inter-coder reliability, the two coders first defined a set of coding rules. Next, each researcher independently coded the CG statements of two companies to identify the differences between coders. Finally, the differences identified were discussed and, on the basis of this discussion, the final set of coding rules was defined. Krippendorff"s α was calculated to quantify the level of inter-coder reliability, yielding an acceptable result of 87 percent.

Independent and moderating variables
We measured the ownership concentration using the natural logarithm of the percentage of share held by the first shareholder (OwnConc). Higher values of OwnConc correspond to higher concentrations of power in the hands of the largest shareholder and, as a consequence, lower contestability of his power. We expected a negative relationship with CoE_DI. We computed the natural logarithms of OwnConc to control for skewness.
We computed the moderating variable, the presence of a dominant family shareholder, using a dummy variable (FamFirm). More specifically, we assigned the score of 1 to firms presenting a family as the largest shareholder with almost 10% of shares and 0 otherwise. We also included the interaction term computed as the multiplicative effect between FamFirm and OwnConc.

Control variables
We also included control variables demonstrated to impact the dependent variable. In particular, we included firm size (Size), which was calculated as the natural logarithm of total assets and was predicted to have a positive association with CoE_DI because larger firms are expected to provide more information to satisfy investor demand for information, considering that they support lower average costs for collecting and disseminating information than smaller firms (Cerbioni & Parbonetti, 2007). We included leverage (Lev), which was calculated as the total long-term debt over total assets and was predicted to have a positive association with CoE_DI because, according to Jensen and Meckling (1976), firms with higher leverage have more incentive to disclose information voluntarily because they hope to reduce agency costs with long-term and short-term creditors. We included both profitability (Profit), calculated as the return on assets (ROA), and growth sales (GrowthSales), computed as the growth rate of total sales, and we predicted finding a positive relationship with our dependent variable because companies characterized by high profitability and growth sales could have incentives to make more corporate disclosures (Raffournier, 1995) to underscore their good performance to investors. Finally, we included in our model the industry variable (services, industrial, financial and others) and coded this variable using a dummy to account for possible unobserved heterogeneity among industries. We included in the industry category firms that NaceRev2 codification identify as construction and manufacturing firms; in the services sector, we included firms that supply electricity, gas, steam, water, sewerage, waste, repair of motor vehicles, transportation, food services and accommodation, ICT and administrative services; we include in financial sector firms that make financial and insurance activities; other firms are included in a residual category, that is other. Table 5 summarizes all of the model variables and provides more information about their descriptions, measurements and data sources.

Regression model
To test the hypothesis developed, we used the following regression model: To verify and quantify the effect of the moderating variable on our OLS regression, we added an additional moderator term (FamFirm) to the model. This term could affect the direction and strength of the relationship between the prediction variable and the dependent variable.
Moreover, we inserted into our regression the interactive effect of the moderator term.
(2) 4. FINDINGS Table 6 shows the descriptive statistics for the variables. The average value of our CoE_DI variable is 56.07%. This value shows that the sampled companies either complied or adequately explained why they decided to not follow a specific recommendation. However, it is necessary to better analyse both the "comply indicator" and the "explain indicator". Table 7 reports their values in absolute terms. On average, the sampled firms complied with 29.57 recommendations. The minimum value of the "comply indicator" was 2, and the maximum value was 51. Thus, there were no companies complying with all 54 recommendations identified ex-ante. The situation was completely different when analysing the "explain indicator". Companies did not provide adequate explanations for all of the recommendations: the average value, in absolute terms, was 0.71. The maximum value was 9 and the minimum 0. Definitively, while the sampled companies tended to comply with most of the recommendations identified and to disclose this information when they decided to not comply, they did not provide adequate explanations according to Recommendation 208/2014. It is likely that this choice also occurred because, in Italy, independent monitoring and enforcement mechanisms do not exist. The aforementioned results regarding complyor-explain principle applications could be read in the light of the findings of Zattoni and Cuomo (2008): in civil law countries, such as Italy, the issuance of codes and, in the same way, the introduction of related innovations are prompted more by legitimation reasons than by a determination to improve governance practices.

Descriptive statistics
Breaking down both the "comply indicator" and the "explain indicator" for all of the 7 categories into which the 54 recommendations were grouped, we found that the vast majority of the information provided about the compliance with the CG Code regards the category "Functioning of the board", followed by the "Composition of the board" and "Independent directors" categories. Passing to the "explain indicator", the vast majority of explanation belonged to the "Composition of the board" category, followed by the "Functioning of the board" category. Moreover, with specific regard to the "explain indicator", the majority of the explanations provided by the sampled companies concerned descriptions of the reasons for the departure from a specific recommendation and descriptions of the measures undertaken instead of compliance. This last result was in line with the findings of Hooghiemstra and Van Ees (2011), who documented that companies frequently justify deviation from a recommendation by revealing the internal practices adopted.
OwnConc exhibited a relatively high value on average, showing that, in the sampled companies, power is concentrated in the hands of the largest shareholder. Table 8 reports the percentage of shares held by the top three shareholders. The data demonstrate the peculiar situation of Italian non-financial listed companies, which are characterized by a high ownership concentration. The largest shareholder possesses, on average, 49.13% of the shares, whereas the second and the third shareholders have 13.86% and 7.10%, respectively. The first two shareholders possess, on average, 62.50% of the shares, and the first three shareholders possess 68.98%.
Moving to the analysis of the largest shareholder identity, 15.4% of the sampled companies are represented by family firms that are firms with a family as the largest shareholder that possesses more than 10% of the shares.
Finally, by analysing the other control variables, the sampled companies present the following levels: size (19.44), growth in sales (0.259), profit (1.73) and leverage (0.164).
Before conducting a regression analysis, we investigated the correlations between the model variables. Table 9 provides the correlation coefficients between CoE_DI and the independent and control variables.

Regression results
Regression (1) in the following table reports the effect of ownership concentration (OwnConc) on CoE_DI. In regression (2), we inserted both the moderating variable (FamFirm) and the interaction term (OwnConc*FamFirm). The explanatory power of the regressions varied from 33.49% to 37.6%.
Regarding Model 1, the results indicate a negative relationship between ownership concentration and CoE_DI. More specifically, the coefficient of OwnConc is statistically significantly better than a 1 percent level for explaining variations in CoE_DI (β = -0.182, p<0.01). The negative coefficient is consistent with our expectation, emphasizing that companies presenting more concentrated ownership tend to disclose less information about compliance with the CG code and to explain less about non-compliance. In other words, these companies tend to be less adherent to the comply-or-explain principle. This result is coherent with the considerations of Sergakis (2013), who affirmed that, in national contexts where the ownership structure is concentrated, soft law measures might not be the most suitable system to exert sufficient pressure for the improvement of CG practices. In these contexts, a new supervisory mechanism of the information disclosed becomes necessary. Our result is in line with Warning (2011), who found that German listed firms with very high ownership concentration showed a small probability of compliance with CG code, while a moderate concentration led to a higher degree of compliance. Our result is also in line with Samaha et al. (2012), who found that the extent of CG disclosure is lower for listed Egyptian companies with higher ownership concentration. Vander Beauwhede and Willekens (2008) also found that ownership concentration negatively influences the level of CG disclosure. Instead, our finding is in contrast with the results found by Barako (2007), who reported positive relationships between shareholder concentration and social and board disclosures.
In regression (2) the moderating variable is significant. With respect to the interaction term, OwnConc*FamFirm is significant and negative. In other words, the significant negative interaction term indicates that the negative relation between ownership concentration and CoE_DI is stronger for family firms. Our result is in line with the findings of Gray (2002 and2010) and Ho and Wong (2001), who documented that companies characterized by family-controlled ownership structure make significantly fewer disclosures at high concentration levels. In other words, the presence of large family block holders induces companies to disclose less information because the demand for public disclosure is relatively weak in comparison with firms that have wider ownership (Chau & Gray, 2002). According to the convergence of interest hypothesis (Morck et al., 1988), in these family-controlled firms, the convergence of interests between dominant shareholder and minority shareholders brings to a reduction of type II of agency conflict and, as a consequence, a decrease in the level of disclosure.
In our analysis, the simple slope of the independent variable on the dependent variable (CoE_DI) changes depending on the value of the moderator (FamFirm). More specifically, the results reveal that when there is a Family as a dominant shareholder of the company the negative relationship between OwnConc and CoE_DI becomes stronger. To better understand the effect of the moderating variable, we grouped companies into two categories: family firms and non-family firms.
The moderation can be interpreted as follows: ceteris paribus, following a certain increase in the ownership concentration, the comply-or-explain disclosure decreases more for family firms than for non-family firms. The graph below illustrates our interpretation.

Source: Authors' elaboration
With respect to the control variables, size positively contributes to CoE_DI and Leverage negatively influences CoE_DI. These results could be explained by considering that larger companies could have incentives to provide more information to satisfy investor demand for information, considering that they support lower average costs of collecting and disseminating information than smaller firms (Cerbioni & Parbonetti, 2007).

CONCLUSION, LIMITATIONS AND FUTURE PERSPECTIVES
In this paper, we considered the principle of complyor-explain as a whole, measuring the degree of disclosure under this principle by companies and testing the relationship between this degree of disclosure and ownership concentration.
The vast majority of the research exclusively analysed compliance with best practices, while this study considers both compliance and explanations provided in cases of non-compliance.
The sampled companies tend to comply with most of the recommendations identified; however, when they decide to not comply, the sampled firms do not provide adequate explanations according to Recommendation 208/2014. This phenomenon is likely aggravated by the absence of independent monitoring or enforcement mechanisms and, in Italy similar to in other civil law countries, by the issuing of codes and the introduction of related innovations being prompted more by legitimation reasons than by the determination to improve governance practices (Zattoni et al., 2017). Therefore, in civil law countries such as Italy, where companies can be expected to be least willing to comply with recent CG disclosure requirements, additional monitoring will be needed (Vander Beauwhede & Willekens, 2008; Sergakis, 2013).
Research in the past decades into the relationship between ownership concentration and voluntary disclosure has been inconclusive, and there is no convergence between the sign and form of the relationship. The main reason of these different findings depends on the institutional context where empirical studies were conducted. Moreover, the exclusive consideration of the ownership concentration for the study of the relation between ownership structure and disclosure is not enough for this aim. These divergent findings led our study to test the aforementioned relation, introducing an interaction term that considers the identity of the largest shareholder in the regression model.
The inclusion of an interaction term that refers to shareholders" identity is important for better understanding the "power games" of the dominant shareholders and the influences on those who control the company, which derive from the interactions between these shareholders.
Our study shows that there is a negative relationship between ownership concentration and the variable of comply-or-explain disclosure, revealing that companies presenting more concentrated ownership tend to disclose less information about compliance with CG code and to explain less about non-compliance. This outcome occurs because the largest shareholder, potentially having the power to influence CG practices and thus the reporting practices, does not use his/her power in the direction of incrementing disclosure about the comply-or-explain principle because he/she has access to the information required through private channels.
Our results also reveal that, when there is a dominant family shareholder in the ownership of the company, the negative relationship between ownership concentration and comply-or-explain disclosure is stronger. This result indicates that, at high levels of ownership concentration, the presence of a dominant family shareholder can encourage sub-optimal decisions that are harmful to firms, such as to decrease the degree of disclosure about the comply-or-explain principle, because of the presence of managerial entrenchment, as suggested by both the conflict-of-interest and strategicalignment hypotheses. In other words, it seems that, when the ownership is more concentrated, the presence of a dominant family shareholder discourages managers from achieving higher levels of adherence to the comply-or-explain principle and disclosing this result. Family firms at high levels of ownership concentration seem to create inefficiency in the degree of adherence to the comply-or-explain principle and to generate a potential opportunity for management"s and/or large shareholders" entrenchment.
Coherently, these findings confirm previous literature showing (Hutton, 2007;Chen et al., 2008) some critical factors that limit the voluntary disclosure in family firms. Firstly, family owners have longer investment horizons than other shareholders and this implies that they are less interested in accelerating timely information, such as trading profits. Moreover, family firms tend to not disclose voluntary information in order to avoid the potential costs, such as proprietary costs, deriving from managers" emphasis on short-term rather than long-term performance. Secondly, family owners" active involvement in firms" management results in lower information asymmetry between themselves and managers. Moreover, because of the better monitoring of managers by family owners, the demand for information from non-family owners to monitor managers is lower due to the substitutive relation between direct monitoring and disclosure (Bushman, Chen, Engel, & Smith, 2004;Chen et al., 2008).
The study contributes to the academic literature in two ways, supplying some reasons useful to explain why previous research did not obtain convergent results. First, we develop a novel index of disclosure, the comply-or-explain disclosure index, which considers comply and explain aspects together. To the best of our knowledge, few studies investigated the explain aspect or developed specific indices. Different from previous studies, we assess the explanation provided by companies in cases of non-compliance using the indications contained in Recommendation 208/2014. Moreover, when analysing the ownership structure, we do not exclusively consider the level of ownership concentration, but we also introduce in our regression model a moderating variable that considers the shareholder identity, documenting that the presence of a dominant family shareholder influences in an interesting manner the relationship between ownership structure and the degree of disclosure regarding the comply-or-explain principle of Italian listed companies. This result is very interesting because it shows the importance of the largest shareholder"s identity, emphasizing that different types of shareholders in whose hands shares are concentrated could determine different types of relations (positive or negative) and/or relations characterized by different levels of strength. This result opens an interesting field of analysis for CG scholars and has implications for various actors. It is interesting for policy makers and practitioners because it suggests that, in countries such as Italy, where often dominant family shareholders play a pivotal role in firm control, the issuance of a good governance code could be ineffective in improving disclosure, accountability and CG in general. In other words, our result seems to confirm that, in contexts characterized by high levels of ownership concentration and weak institutional settings, the issuance of code could be prompted more by legitimation reasons than by the purpose of improving the governance practices of national companies. Moreover, this result is exacerbated when there is a dominant family shareholder. As a consequence, when legislative or self-regulating initiatives are designed, the legislators and other actors involved should not ignore the characteristics of the context in which the innovation is introduced. Considered together, our results suggest that, in national contexts characterized by high levels of firm ownership concentration, soft law measures might not be the most suitable mechanism to exert pressure for the improvement of CG practices. These contexts require a new supervisory mechanism for the information disclosed by firms.
Our findings also have implications for companies and their boards because they show that, in contexts characterized by the highest ownership concentration, there is a natural tendency to minimize disclosures about compliance and explanations for non-compliance, limiting the market disciplinary mechanism and preventing investors from understanding the governance practices. In such situations, in which there is little separation between ownership and control, it becomes important that the independent board members exert their independence to contrast the power of controlling owners and thus the tendency to limit the disclosure.
Our study also has implications for academics, suggesting the need for researches that analyse the effectiveness of CG mechanisms considering the interaction of different mechanisms that operate at both the firm and country levels.
This study has some limitations, so further research is necessary. First, the sample includes only companies listed in Italy, so the results might not be generalized. Second, the study was conducted in a period of financial crises, which could have influenced the behaviour of firms. Moreover, our comply-or-explain variable exclusively refers to the composition and functioning of the board of directors and has been developed using the CG code and the "format for the CG statement" issued by Borsa Italiana. Therefore, our comply-or-explain disclosure index could be utilized only to conduct analyses on Italian companies and with respect to the board of directors" structures. Thus, in future research, we must expand the sample to consider a different period without financial crises and develop a new version of the comply-or-explain disclosure index, useful to investigate also the degree of effective adherence to the CG codes of companies listed in other countries.