Earnings quality among private firms: evidence from the ELITE context

Abstract

Purpose

Recent accounting literature claims that private firms’ heterogeneity influences the quality of earnings. Along with certain drivers of heterogeneity, private firms get involved in specific programs aimed at fostering their access to capital, competencies and networks (CCN programs). Such programs can enhance private firms’ exposure to stakeholders that demand higher reporting quality, affecting their financial reporting choices. Therefore, this study investigated whether membership in CCN programs affects private firms’ earnings quality.

Design/methodology/approach

Focusing on the ELITE program, an international platform that since 2012 aims to support the growth of the most promising SMEs, and employing different econometric specifications facing endogeneity concerns, this paper carries out a quantitative empirical analysis to test the effect of CCN programs on private firms’ earnings quality.

Findings

Employing different earnings quality measures, empirical evidence reveals that firms belonging to CCN programs experienced an improvement in their earnings quality.

Research limitations/implications

Even though endogeneity concerns have been addressed, we are nevertheless aware that they might, at least partially, have affected our results.

Practical implications

Although the contributions of the study are mostly academic, the empirical evidence obtained also carries practical implications. CCN programs not only act, as one might assume, as catalysts for economic and dimensional growth but also contribute to better earnings quality, mitigating the information asymmetries between firms and their stakeholders.

Originality/value

By adding new evidence to the literature concerning the impact of private firms’ heterogeneity on earnings quality, this is the first study to analyze the impact of specific programs aimed at supporting the affiliated SMEs to foster their access to capital, competencies and networks.

Citation

Ricciardi, G., Fera, P., Moscariello, N. and De Nuccio, E. (2024), "Earnings quality among private firms: evidence from the ELITE context", Journal of Applied Accounting Research, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/JAAR-09-2023-0276

1. Introduction

Private firms play a key role in the global economy and are the main source of growth in most national economies. For instance, in the European market, private firms represent approximately 99% of the total number of registered companies (European Commission, 2022). Similar proportions apply to both the United States (Asker et al., 2015) and numerous other countries around the world (Hope and Vyas, 2017).

Despite their prominent role, the accounting literature only began to analyze private firms and their financial reporting quality in the last two decades (Beuselinck et al., 2023). This late focus on private firms likely relates to data availability: data on private companies are not readily available in most contexts (Pacter, 2007; Hope and Vyas, 2017). Additionally, compared to public firms, fewer stakeholders and investors are interested in the outcomes of private firms (Ball and Shivakumar, 2005).

However, an analysis of private firms’ accounting choices is relevant as financial reporting quality is as essential for private firms as it is for public firms (Bharath et al., 2008; Chen et al., 2011; Hope et al., 2017; De Meyere et al., 2018; Belot and Serve, 2018).

Literature on the earnings quality of private firms mainly consists of studies which compare private and public firms and, relying on the dispute between the demand hypothesis and the opportunistic behavior hypothesis, has provided mixed evidence (see Bar-Yosef et al., 2019 for a detailed literature review).

However, these studies set up private firms as a homogeneous sample, ignoring how the strong heterogeneity among such firms can influence their incentives for supplying high-quality earnings (Bigus et al., 2016; Liu and Skerratt, 2018; Bonacchi et al., 2019).

Studies on heterogeneity among private firms represent a relatively new strand of research, highlighting that differences between private firms in terms of stakeholders (Hope et al., 2017), legal form (Bigus et al., 2016), size (Liu and Skerratt, 2018), organizational structure (Bonacchi et al., 2019), disclosure regime (Bernard et al., 2018) and accounting regulatory settings (Cameran and Campa, 2020) have a significant influence on earnings quality. These results suggest that private firms’ accounting choices are driven not only by market demand and managerial opportunism but also by “nonmarket forces” that need to be controlled (Bonacchi et al., 2019).

Private firms’ heterogeneity, however, is also influenced by the presence of specific programs aimed at supporting and mentoring the affiliated companies to foster their access to capital, competencies and networks (hereinafter CCN programs). These programs help firms to boost their growth, facilitate active networking among its participants and provide links with domestic and international financiers and investors. As such, CCN programs can enhance private firms’ exposure to stakeholders that demand higher reporting quality, affecting their financial reporting choices.

Earnings quality within private firms reacts to the demand for monitoring made by three types of stakeholders: shareholders, debtholders and suppliers (Katz, 2009; Hope et al., 2017). In particular, skilled equity investors, more lending relationships and a greater number of suppliers create a higher demand for earnings quality (Katz, 2009; Hope et al., 2017; Bigus and Hillebrand, 2017; Martinez-Martinez et al., 2021). Since private firms enrolled in CCN programs are more likely to be involved with a broader pool of equity investors, lenders and suppliers, they are expected to face a higher demand for earnings quality.

Hence, the aim of this study is to examine whether the financial accounting choices made by private firms is influenced by their affiliation in CCN programs.

Among the various initiatives, this study focuses on the ELITE program, established in 2012 by the London Stock Exchange. ELITE is a network of private SMEs which support companies in their long-term growth by accelerating the process of accessing capital, expertise and networking. With over 1,600 companies around the world, 200 partners admitted since the start, and about €16 billion raised by ELITE firms through more than 1,500 corporate transactions, ELITE is one of the most important institutional programs committed to supporting SMEs.

The empirical analysis is carried out on a unique dataset provided by the ELITE network, consisting of the list of Italian companies which joined the program between 2012 and 2019. To mitigate the potential endogeneity problem arising from firms voluntarily enrolling in the ELITE program, this study employs a methodology that deals with self-selection bias: i.e. a comparable group of non-ELITE private firms (built using the propensity score matching technique, PSM) and a differences-in-differences approach (DID). Relying on the assumption according to which high-quality accounting numbers are associated with higher accrual quality and more timely loss recognition (Healy and Wahlen, 1999; Barth and Schipper, 2008), the empirical findings reveal that ELITE firms experienced a significant improvement in the quality of accounting numbers. Therefore, this study finds that firms belonging to CCN programs exhibit higher quality of financial reporting because they face a more stringent demand for earnings quality.

The remainder of the paper is organized as follows. The following section presents the background on the financial reporting quality of private firms and develops the research hypothesis. Section 3 details the research setting (i.e. the ELITE program). Section 4 describes the sampling process and the research methodology employed. Section 5 describes and discusses the empirical evidence, while Section 6 offers concluding comments.

2. Literature review and hypothesis development

This study lies at the intersection of two strands of research: literature concerning the earnings quality of private firms and literature examining the heterogeneity within private firms.

The accounting literature concerning private firms’ earnings quality mainly consists of studies that analyze mixed samples which include (and typically compare) private and public firms (Bar-Yosef et al., 2019). Empirical evidence on the relative earnings quality of public versus private firms is mixed.

A first strand of research embraces the demand hypothesis and suggests that public firms have (on average) higher accounting quality than private firms as they face a stronger demand for high-quality earnings (Ball and Shivakumar, 2005; Burgstahler et al., 2006; Hope et al., 2013). A second strand, instead, relies on the opportunistic behavior hypothesis, according to which managers suffer stronger pressure from public equity markets to meet earnings benchmarks or exceed stakeholders’ expectations, boosting managerial incentives to manipulate earnings (Beatty et al., 2002; Kim and Yi, 2006; Givoly et al., 2010).

In conducting their research, some authors became aware that these studies treat private firms as a homogeneous group (Bigus et al., 2016) and started questioning the reliability of failing to control for private firm heterogeneity (Bonacchi et al., 2019, p. 1070). Such evidence calls for additional research to address whether and how earnings quality varies among private firms.

Studies on heterogeneity among private firms constitute a relatively new research strand. Bigus et al. (2016), analyzing German private firms, found that corporations exhibit higher levels of income smoothing and conservatism than partnerships and one-person businesses. Bernard et al. (2018) reported that private firms which voluntarily disclose financial information exhibit higher earnings quality than both private firms which are subject to mandatory disclosure requirements and public firms. Focusing on the UK, Liu and Skerratt (2018) revealed that small and micro companies have much higher earnings quality than large private and medium-sized companies. Bonacchi et al. (2019), concentrating on the European Union and examining heterogeneity in organizational structure, found that private business groups have higher earnings quality than standalone firms. Finally, Cameran and Campa (2020) examined the accounting regulatory setting in the EU and reported that IFRS adoption has a positive impact on private firms’ financial reporting quality.

Overall, by outlining some endogenous drivers of heterogeneity among private firms, these studies provide an explanation for both the conflicting results in the comparative literature on earnings quality between public and private firms and the disparity in the quality of earnings among private firms.

Along with these drivers, this paper argues that the heterogeneity of private firms may also emerges as result of their affiliation with programs aimed at supporting the growth of affiliated companies, facilitating their access to capital, competencies and networks (CCN programs). These programs aim at boosting the capacity for growth and offer bridging services. As a boosting service, CCN programs offer tutoring session for developing growth strategy and strengthening the management competencies of the firm. As bridging service, CCN programs facilitate active networking among its participants and provide links with domestic and international investors. To this regard, CCN programs operate also as branding mechanisms that enhance the credibility of its participants.

In this vein, such programs may enhance private firms’ exposure to stakeholders that demand higher reporting quality, potentially affecting their financial reporting choices [1]. Specifically, companies within CCN programs have a higher likelihood of engaging with institutional investors, gaining more financing opportunities and establishing a broader network of suppliers. This increased exposure and connectivity with diverse stakeholders is likely to significantly influence their accounting choices. Research by Hope et al. (2017) and Lisowsky and Minnis (2020), indeed, highlight the pivotal role of various stakeholders, including shareholders, debtholders and suppliers, in driving the demand for earnings quality.

Regarding shareholders, demand-side matters suggest that separated ownership is associated with a higher demand for earnings quality due to the monitoring role of minority shareholders (Hope, 2013). However, minority shareholders are not the only ones able to determine higher earnings quality: Katz (2009), for instance, showed that the tighter monitoring and reputational considerations of a private equity sponsor can increase accrual quality.

Focusing on debt market characteristics, scholars have argued that lenders are likely to demand high-quality accounting reports from borrowers to enhance debt-contracting efficiency (Bharath et al., 2008; Hope et al., 2017). Specifically, firms with a greater number of bank relationships are less likely to engage in earnings management activities (Bigus and Hillebrand, 2017). In addition, Haw et al. (2014) found that private firms with publicly traded debt exhibit higher earnings quality due to the demand for monitoring by bondholders.

Finally, with respect to lenders, also suppliers face the counterparties’ credit risk and, therefore, assess the accounting information quality of the related firms (Hope et al., 2017).

As such, firms participating in CCN programs, due to their increased engagement with institutional investors, expanded financing opportunities and a broader supplier base, are subjected to heightened scrutiny, potentially influencing their accounting decisions. In essence, the multifaceted scrutiny, emerging from a diverse and broader network of stakeholders, may shapes and influences firms’ accounting choices.

This reasoning leads to the development of the following hypothesis:

Hp.

CCN private firms exhibit a significantly higher quality of earnings than other private firms.

3. Institutional background: CCN programs

Small and medium enterprises (SMEs) play a key role in the EU economy. They are the backbone of our economies, the industrial fabric of many regions and cities – they are the key to social cohesion and an engine of regional job creation and well-being (European Commission, 2022, p. 16). According to the Annual Report on European SMEs, in 2021, SMEs account for 99.8% of non-financial firms, employ most of the workers (64.4%), and contribute 51.8% of the value-added.

In June 2008, the European Commission published the Small Business Act (SBA) with the dual aim of acknowledging the central role played by SMEs in the European setting and introducing an effective business policy framework to support small businesses. The main rationale for this is that SME and entrepreneurship support, regardless of the forms adopted (i.e. mentoring, training, enhanced access to finance), can boost the growth of SMEs (Rao et al., 2021; Idris et al., 2023) as well as the rest of society through positive spill-over benefits in terms of job and wealth creation, as well as economic growth (OECD, 2018).

Over the last 2 decades, several industry and policy initiatives have been established across the EU to foster the growth of SMEs and their listing. One of the most successful of these initiatives is the ELITE program launched in 2012 by the London Stock Exchange Group [2]. This program aims to accelerate the growth of the most promising SMEs through training programs, coaching sessions and access to different forms of funding. At the end of the first half of 2021, the program covered 43 countries, including more than 1,600 firms, yielded an aggregate turnover of € 111 billion, registered 665,000 employees and raised €16 billion through more than 1,500 corporate transactions (i.e. bond issues and IPO). Among the 43 countries involved, Italy is the most representative in terms of companies (970 out of 1,600 are based in Italy), contributing significantly to the success of the program (most of the raised funds refer to Italian companies). At the end of the first half of 2021, these 970 firms group more than 500,000 employees and reached an aggregate turnover of €94 billion, representing 5% of the aggregate revenues of the Italian SMEs [3].

Access to the ELITE program is subject to the fulfillment of specific economic requirements. Firms entitled to be included in the program are those with sales greater than €10 million [4], an operating income greater than 5% of the sales, and non-negative net profits.

Once enrolled, companies can benefit from training courses designed to help managers and entrepreneurs improve their competencies, receive assistance in making any necessary changes (i.e. corporate and financial communication practices) to consolidate their capabilities, and access the benefits and opportunities provided by the ELITE bridging services, such as a network of suppliers, advisors, institutions and investors.

In a context as heterogeneous and complex as that of private firms, the ELITE program, given its unique characteristics, affords researchers the opportunity to explore how the quality of accounting information change as private firms grow and become more similar to public firms in term of incentives (Bar-Yosef et al., 2019).

4. Research design

4.1 Sample composition

Companies belonging to the ELITE program in Italy represent the reference population. In particular, we focused on the list of the 830 Italian companies which joined the program between 2012 and 2019. According to the AIDA database published by Bureau van Dijk, the Italian private firm population consists of approximately 2,3 million firms. Of these firms, just over 11,000 firms meet ELITE’s selection criteria. Thus, the coverage of the ELITE program is quite notable relative to the target population. Moreover, the focus on a single country allowed us to rule out any issues linked to differences in accounting standards, ownership structure and investors protection which are inherent in cross-country studies (Leuz et al., 2003; Belot and Serve, 2018).

Using the AIDA database, we collected data from 2011 to 2020, focusing on annual accounting reports. Because empirical variables include lagged values and a pre-enrollment period, we focused on companies which joined the ELITE program from 2014 to 2018. Moreover, we omitted all financial and assurance firms because their peculiar financial reporting rules can bias earnings quality models (Fera et al., 2022). In addition, we excluded from the sample those firms with a lack of both financial and accounting data, and implausible values. Finally, to rule out any possible bias linked to public firms’ incentives, we excluded firms that were listed during the reference period. Table 1 summarizes the sample selection process for ELITE firms.

To assess the effect of the ELITE program on private firms’ earnings quality, we needed to address the potential endogeneity arising from firms’ self-selecting to join ELITE. Therefore, we first built a control sample of non-ELITE private firms. The companies included in this group were the result of a preliminary screening process on AIDA that employed the following criteria: (1) data available from 2011 to 2020; (2) active firms (i.e. not involved in a liquidation process); (3) firms are corporations; (4) total assets, equity, sales and net profit are 30% lower than the minimum and 30% higher than the maximum of the same variables for the ELITE firms in the year prior to joining ELITE (see Cameran and Campa, 2020, p. 10). This process resulted in a sample consisting of 34,142 unique entities.

Then, following the prior literature (Rosenbaum and Rubin, 1983; Heckman et al., 1997; Armstrong et al., 2010; Lawrence et al., 2011; Giovannetti et al., 2013), we associated to each treated firm joining into ELITE initiative in the calendar year t with a non-treated firm for the same year, industry and geographical location (northeast, northwest, center and south of Italy) that has the closest possibility of being selected (i.e. the closest propensity score). We estimated propensity scores by employing a logistic model that used the following variables as matching dimensions (see Appendix for the definition of all the variables): (1) we included three variables which capture the ELITE economic requirements, that are Sales, Net_Income and ROS in the year before joining the ELITE program [5]; (2) we also included Size and Leverage in the year before joining the ELITE program, since they are also part of the ELITE assessment. Next, we matched each ELITE firm to non-ELITE firms using a one-to-one nearest-neighbor matching without replacement [6]. In this way, we obtained a control sample consisting of companies that, having a mirror profile to ELITEs, would have had the same opportunities and incentives to join the program.

The matching procedure resulted in 680 unique entities (340 ELITE firms and 340 non-ELITE firms) from which we obtained a final sample of 3.446 firm-year observations [7].

Table 2 reports the means of the treatment and the control groups, along with the results of the t-test for group differences in means. None of the t-tests is significant, confirming the efficacy of the propensity-score matching (see Bonacchi et al., 2019, p. 1100).

4.2 Measurement of earnings quality

Accounting literature provides several definitions of earnings quality (see Dechow et al., 2010, for a detailed review). In general, earnings are characterized by high quality when they reflect past performance and allow investors to forecast future income (Dechow and Schrand, 2004; Richardson and Tuna, 2012). As highlighted by Christensen et al. (2015), contemporary accounting literature usually examines three dimensions of earnings quality: earnings management, timely loss recognition and value relevance. Specifically, high-quality earnings are associated with less earnings management, more timely loss recognition and greater value relevance (Healy and Wahlen, 1999; Barth and Schipper, 2008; Barth et al., 2008). Since value relevance metrics rely on market data and, in turn, are not suitable for investigating private firms, this study employs earnings management (proxied by the magnitude of abnormal accruals) and timely loss recognition metrics. This choice follows previous studies that have investigated the earnings quality issue among private firms by employing two measures: earnings management and accounting conservatism (see Bar-Yosef et al., 2019, for a detailed literature review).

4.2.1 Proxy for earnings management

Prior literature suggests several methods for determining discretionary accruals (Jones, 1991; Dechow et al., 1995; Peasnell et al., 2000; DeFond and Park, 2001; Kothari et al., 2005). Building on Francis et al. (2005), we used the Dechow and Dichev (2002) model as modified by McNichols (2002):

(1)∆WCAit=β0+β1CFOi(t−1)+β2CFOit+β3CFOi(t+1)+β4∆Salesit+β5PPEit+εit

where ∆WCA represents the change in working capital accruals, CFO denotes the operating cash flows, ∆Sales represents the change in sales and PPE stands for property, plant and equipment [8].

We ran this model cross-sectionally for each industry-year from 2011 to 2020 and then used the regression coefficients obtained from Equation 1 to estimate the expected values for each firm-year observation of our sample. Subsequent to this, the abnormal component of working capital accruals for each firm-year observation (AA) was determined as the difference between the observed values in our sample and the expected values obtained, as presented above (Moscariello et al., 2020).

Once the abnormal accruals had been estimated, we scaled AA by end-of-the-year total assets and used the absolute value of AA because our purpose is to estimate the magnitude of abnormal accruals, irrespective of the intent to increase or decrease income.

4.2.2 Proxy for timely loss recognition

Following Cameran and Campa (2020), we tested timely loss recognition (TRL) using the following panel clustered-robust regression model developed by Basu (1997) as modified by Ball and Shivakumar (2005):

(2)ACCit=α+β1NCFOit+β2CFOit+β3(NCFO*CFO)it+β4ELITEit+β5(ELITE*NCFO)it+β6(ELITE*CFO)it+β7(ELITE*NCFO*CFO)it+∑(Year effects)it+∑(Industry effects)it+εit

where ACC is the total operating accruals scaled by the average total assets, NCFO is a dummy variable that equals 1 for negative operating cash flows, ELITE is a dummy variable that equals 1 for ELITE firms, and CFO stands for the operating cash flows scaled by the average total assets.

To examine the difference in TLR between ELITE firms and their control group, before and after the decision to join ELITE, we ran Model (2) first, for observations of the ELITE firms and their controls, referring to the period post-ELITE entry, and then for observations referring to the period pre-ELITE entry.

After estimating regression (2) and verified the fulfillment of the common-trends assumption trough the inspection of β3 trends in the pre-treatment period (Heckman et al., 1999; Wooldridge, 2010; Goodman-Bacon, 2021) (see Figure 1), we measured timely loss recognition using β7 coefficients from the two regression models. A significantly positive β7 coefficient implies that firms belonging to the ELITE program report bad news (losses) on a more timely basis than non-ELITE firms. Moreover, if the β7 coefficient estimated for the post-ELITE time-window was significantly higher than the same coefficient for pre-ELITE observations, it implies that firms joining the ELITE program experience an improvement in timely loss recognition.

4.3 Independent and control variables

To test whether the ELITE program affects private firms’ earnings quality, our independent variable consists of a dummy variable (ELITE) that takes the value 1 for firms belonging to the ELITE program and 0 otherwise, i.e. the firm belongs to the control group.

To better test our hypothesis, we included in the model several control variables that might affect the magnitude of abnormal accruals and that are widely used in accounting literature [9].

Size is defined as the annual sales’ natural logarithm. Prior studies suggest that large companies tend to have lower abnormal accruals because they are more closely examined than small companies and tend to have more stable and predictable operations (Klein, 2002; Dechow and Dichev, 2002). Because both variables have been related to earnings management (Francis et al., 2005; Klein, 2002), we also control for the volatility of operating cash flows (CFO_Vol), computed as the year-over-year relative changes, and the frequency of negative earnings (Neg_Earn), a dummy variable that takes the value of 1 for each firm-year observation with a loss. We include a measure for the firm’s profitability (as proxied by ROA) which may affect the magnitude of abnormal accruals (McNichols, 2002; Kim and Yi, 2006) and the annual leverage (Lev), computed as the ratio between debt and equity (Hope et al., 2017).

We also consider external features which consist of Inst_Inv, a dummy variable which assume the value of 1 if at least one institutional investor owns more than 2% of the share (Katz, 2009), and Own_Conc, a categorical variable that assumes values from 1 to 4 depending on the level of direct control [10] (Moscariello et al., 2020). Finally, we included Group, a dummy variable that takes the value of 1 if the firm directly owns subsidiaries at a stake higher than 50% (see Bonacchi et al., 2019, p. 1076), and we also control for the organizational form of each company, which has been related to the earnings quality of private firms (Bonacchi et al., 2019).

4.4 Model specification

To test our hypothesis, we set up the following differences-in-differences model which measures the impact of the ELITE program on the magnitude of abnormal accruals:

(3)AAit=β0+β1ELITEit+β2Post_ELITEit+β3ELITEit X Post_ELITEit+β4Sizeit+β5CFO_Volit+β6Neg_Earnit+β7ROAit+β8Levit+β9Inst_Invit+β10Own_Concit+β11Groupit+∑(Year effects)it+∑(Industry effects)it+εit

where Post_ELITE is a dummy variable that takes the value of 1 when the observation of the ELITE firm and its control refers to the period post-ELITE entry, i represents the specific firm, t indicates the reference year, and ε stands for the regression error.

By doing so, we are able to assess the difference between the ELITE firms and the control group in our first earnings quality measure, both before and after access to the program.

Since the effectiveness of a diff-in-diff model is contingent upon adhering to the common-trends assumption (Heckman et al., 1999; Wooldridge, 2010; Goodman-Bacon, 2021), a visual inspection of the common-trends assumption within our diff-in-diff model is displayed in Figure 2. Given the trends in the pre-treatment period, it is possible to assert that the common-trends assumption within Model (3) is not violated.

5. Results and discussion

5.1 Descriptive statistics and univariate analysis

Table 3 reports the main descriptive statistics for the variables included in the study. They are reported for the pooled sample and by distinguishing between ELITE firms and their controls.

The average value of the magnitude of abnormal accruals (AA) is approximately 4.9% of total assets (with a median value of about 3.5%). On average, the sample is characterized by a high ownership concentration (median value equals 4) and relatively high leverage (mean value equals 2.72). About 66% of the sample is organized as a business group and, on average, ELITE firms exhibit a higher percentage of institutional investors among their shareholders (12% vs 9%). However, as expected, ELITE firms differ slightly from the control group (see Panel B and Panel C). The only differences can be observed for ROA (ELITE firms perform better than their peers) and Group as there are more business groups among the ELITE firms.

The univariate correlation matrix, relative to the total sample, is reported in Table 4. Focusing on correlations significant at 5 and 1% levels, the matrix highlights a positive correlation between AA and firms that decided to join the ELITE program. A positive correlation is also documented between AA and both leverage and ownership concentration. The size, the performance (ROA) and the presence of institutional investors among shareholders are positively correlated with the group of ELITE firms. Finally, Table 4 shows that none of the variables included in the regression model suffers from collinearity as all coefficients are much lower than 0.8 (Fera et al., 2022).

5.2 Empirical results

5.2.1 Earnings management

After controlling for a set of variables that influence the quality of accounting numbers, Model (3) allows us to assess whether the membership in the ELITE program affects the magnitude of abnormal accruals. The relatively low R² of the model (0.0181) does not represent an issue, since the purpose of this models is to test an association in the field of earnings management studies (Breiman, 2001; Lukacs et al., 2010; Kliestik et al., 2021).

Table 5 reveals a negative and significant relationship between the iteration variable (ELITE X Post_ELITE) and AA (|t| = −2.01; pv = 0.045), showing that the difference in the level of abnormal accruals between ELITE and non-ELITE firms decreased significantly during the period after the enrollment of the treatment group (i.e. ELITE sub-sample). This evidence confirms the proposed hypothesis, as the ELITE firms show a significant decline in abnormal accruals compared to the control group and, therefore, exhibit significantly higher earnings quality.

Moreover, the findings from Model (3) also indicate that, all else being equal, the overall leverage of firms and the level of ownership concentration exhibit a positive and significant relationship with the magnitude of abnormal accruals. Conversely, the magnitude of abnormal accruals is lower when firms are business groups.

Finally, Table 5 reveals also a positive and significant relationship between our main independent variable (ELITE) and AA (|t| = 3.39; pv = 0.001), showing that the ELITE firms, before to join the program, exhibit a higher level of abnormal accruals than their peers. However, such results are not surprising. On the one hand, it is noteworthy that specific managerial policies often used by growing firms (e.g. ELITE firms) are likely to be identified as upward earnings management by widely used discretionary accrual methods, whereas those positive discretionary accruals are motivated by growth (Pfeiffer and Velthuis, 2009; Liu, 2019). On the other hand, as the IPO setting, the ELITE program may provide a motive and an opportunity for companies to manage earnings. As for the motive, the chance to enter in ELITE and extracting the benefits provided by the network. The opportunity to manage earnings, instead, results from the opacity in terms of financial data that surrounds a private firm (Premti and Smith, 2020).

5.2.2 Timely loss recognition

Table 6 reports the results for Model (2), the timely loss recognition model developed by Basu (1997) as modified by Ball and Shivakumar (2005).

Column A of Table 6 presents the estimation coefficients of both ELITE and non-ELITE firms during the pre-enrollment period (pre-ELITE), while Column B shows the estimation coefficients of both ELITE and non-ELITE firms during the post-enrollment period (post-ELITE).

As discussed previously, the analysis focuses on the coefficient β7, which is negative and significant in Column A (|t| = −2.19; pv = 0.030), indicating that ELITE firms accounted for losses on a less timely basis than non-ELITE firms before enrollment on the ELITE program, highlighting the same pattern observed within the earnings management analysis. The same result is evident in Column B (|t| = −3.79; pv = 0.000), indicating that after joining the ELITE program, ELITE firms still exhibited less timely loss recognition. However, the DID coefficient related to the difference in both coefficients β7 from Column A and Column B is positive and statistically significant.

Thus, despite ELITE firms still accounting for losses on a less timely basis after their enrollment in the ELITE program, empirical findings reveal that the effect of the ELITE program was beneficial as their timely loss recognition significantly improved relative to the control group. Therefore, this evidence also confirms the proposed hypothesis, as firms joining the ELITE program exhibit significantly higher earnings quality [11].

6. Concluding remarks and limitations

Recent accounting literature claims that private firms’ heterogeneity influences the quality of earnings. In this regard, the involvement of private firms in programs aimed at fostering their access to capital, competencies and networks (CCN programs) is still an unexplored driver of heterogeneity. Such programs could enhance private firms’ exposure to stakeholders that demand higher reporting quality, potentially affecting their financial reporting choices. Therefore, this study examined whether the affiliation in CCN programs influences private firms’ earnings quality.

We focused on the ELITE program (as a proxy for CCN programs), an international platform that since 2012 aims to support the growth of the most promising SMEs through training programs, coaching sessions, networking and facilitating the access to different forms of funding.

Given the potential endogeneity which arises from firms’ self-selecting to take part in ELITE, we employ a methodology that deals with self-selection bias, based on a treatment group of ELITE firms and a comparable group of non-ELITE firms ‒ built using propensity scores ‒ and the application of a DID approach.

Focusing on the earnings management dimension, the DID analysis revealed a beneficial effect of the ELITE program on firms’ accruals quality, suggesting that the ELITE program constitutes a stimulus for improving the quality of financial reporting.

Similarly, employing timely loss recognition as an earnings quality metric, the effect of the ELITE program has been beneficial ‒ despite ELITE firms continuing to account for losses on a less timely basis than their peers ‒ as the empirical evidence shows that ELITE firms accounted for losses on a more timely basis after their enrollment in the ELITE program.

This paper contributes to the previous literature in several ways. First, it is of particular interest to academics as it provides new evidence on the determinant role played by private firms’ heterogeneity in affecting their accounting choices. Second, this study enriches the literature on the earnings quality of private firms, examining the effect of an unexplored source of heterogeneity among private firms (i.e. being part of CCN programs). Furthermore, this paper also responds to the call for additional research on how accounting information changes as firms grow while remaining private (Bar-Yosef et al., 2019). Finally, this is the first study to analyze the accounting choices of private firms involved in specific policy initiatives (e.g. CCN programs), shedding light on a distinctive attribute of private firms that may have significant implications also in other areas (e.g. strategic, managerial, financial, etc.).

Although the contributions of the study are mostly academic, the empirical evidence obtained also carries practical implications. In particular, CCN programs not only act, as one might assume, as catalysts for growth but also contribute to better earnings quality, mitigating the information asymmetries between firms and their stakeholders. Moreover, even if the ELITE firms represent a small portion of the universe of private firms, it is noteworthy that the number of companies joining ELITE is steadily increasing, and the success of the program could catalyze the emergence of further analogous initiatives, thereby engaging a progressively expanding cohort of companies. In this vein, despite the current narrow setting, our results carry valuable insights for policymakers and regulators involved in the design of growth programs, in assessing the benefits achievable from such initiatives.

Despite the contributions made, we recognize that our results come with some potential caveats. For instance, because we focused on the Italian market, our findings might not necessarily generalizable as earnings quality determinants among private firms are likely to be affected by the country-specific regulatory environment as well as corporate governance features. Moreover, while the paper offers adequate theoretical foundations, it falls short in justifying why ELITE firms, before joining the program, demonstrate a lower earnings quality than the control group. However, such evidence does not undermine what emerges from the DID analysis (i.e. ELITE firms experienced an improvement in their earnings quality) but rather represents a controversial issue that can be addressed by future research. A further shortcoming of interest for future research is the empirical analysis of the mechanism through which CCN programs affect private firms’ accounting choices. Finally, even though endogeneity concerns have been addressed, we are nevertheless aware that they might, at least partially, have affected our results.

Figures

Figure 1

Model 2 outcomes visualization

Figure 2

Model 3 outcomes visualization

Table 2

Descriptive statistics of treated and matched firms

	Mean	Mean	t	p > |t |
Panel A – Sample 2014
Matching variables	ELITE (#18)	Matched (#18)
Sales	109.926.647	74.160.623	0.61	0.545
ROS	0.074	0.162	−0.30*	0.219
Net_Income	3.392.995	2.408.279	0.43	0.332
Size	18.059	18.244	−0.43*	0.668
Leverage	3.141	3.738	−0.59*	0.558
Panel B – Sample 2015
Matching variables	ELITE (#31)	Matched (#31)
Sales	76.395.669	81.916.264	−0.22	0.830
ROS	0.079	0.060	*0.71	0.479
Net_Income	1.527.021	2.236.611	−0.83	0.412
Size	17.813	17.953	−0.49	0.627
Leverage	2.434	2.240	*0.35	0.724
Panel C – Sample 2016
Matching variables	ELITE (#48)	Matched (#48)
Sales	69.882.452	79.841.392	−0.48	0.633
ROS	0.067	0.052	0.89	0.373
Net_Income	2.836.980	3.260.690	−0.28	0.783
Size	17.658	17.737	−0.36	0.723
Leverage	2.972	3.201	−0.34	0.731
Panel D – Sample 2017
Matching variables	ELITE (#102)	Matched (#102)
Sales	73.252.703	75.557.771	−0.14	0.890
ROS	0.071	0.097	−1.10	0.273
Net_Income	3.183.946	3.401.483	−0.22	0.829
Size	17.593	17.779	−1.21	0.227
Leverage	2.990	3.720	−1.18	0.241
Panel E – Sample 2018
Matching variables	ELITE (#141)	Matched (#141)
Sales	63.782.943	69.317.370	−0.57	0.568
ROS	0.070	0.065	0.57	0.569
Net_Income	2.813.744	2.543.942	0.61	0.542
Size	17.561	17.541	0.17	0.863
Leverage	2.906	2.971	−0.21	0.833

Source(s): Table created by the authors

Table 3

Descriptive statistics

	Obs.	Min	Max	Mean	Median	StdDev
Panel A – total sample
AA	3,446	0.001	0.221	0.049	0.035	0.045
ACC	3,446	−0.441*	0.201	−0.014*	−0.011*	0.039
ELITE	3,446	0	1	0.500	0.500	0.500
Post_ELITE	3,446	0	1	0.605	1	0.489
Size	3,446	11.974	21.523	17.520	17.517	1.173
CFO_Vol	3,446	−56.844*	313.420	0.370	0.080	6.414
Neg_Earn	3,446	0	1	0.078	0	0.268
ROA	3,446	−0.492	0.638	0.056	0.043	0.065
Lev	3,446	0.010	43.230	2.797	1.984	2.645
Inst_Inv	3,446	0	1	0.105	0	0.307
Own_Conc	3,446	1	4	3.335	4	1.041
Group	3,446	0	1	0.663	1	0.473
Panel B – ELITE sub-sample
AA	1,723	0.001	0.221	0.051	0.037	0.047
ACC	1,723	−0.341*	0.172	−0.139*	−0.011*	0.041
ELITE	1,723	1	1	1	1	0
Post_ELITE	1,723	0	1	0.605	1	0.489
Size	1,723	12.259	21.523	17.630	17.615	1.063
CFO_Vol	1,723	−18.956*	313.420	0.537	0.104	8.473
Neg_Earn	1,723	0	1	0.059	0	0.235
ROA	1,723	−0.146	0.549	0.066	0.053	0.063
Lev	1,723	0.010	43.230	2.870	2.073	2.622
Inst_Inv	1,723	0	1	0.118	0	0.323
Own_Conc	1,723	1	4	3.383	4	1.016
Group	1,723	0	1	0.766	1	0.423
Panel C – control group
AA	1,723	0.001	0.207	0.046	0.037	0.044
ACC	1,723	−0.441*	0.201	−0.014*	−0.012*	0.037
ELITE	1,723	0	0	0	0	0
Post_ELITE	1,723	0	1	0.605	1	0.489
Size	1,723	11.974	20.780	17.410	17.412	1.263
CFO_Vol	1,723	−56.844*	89.956	0.202	0.055	3.234
Neg_Earn	1,723	0	1	0.097	0	0.297
ROA	1,723	−0.492	0.638	0.046	0.032	0.066
Lev	1,723	0.129	26.488	2.724	1.902	2.670
Inst_Inv	1,723	0	1	0.092	0	0.289
Own_Conc	1,723	1	4	3.287	4	1.063
Group	1,723	0	1	0.560	1	0.496

Source(s): Table created by the authors

Table 4

Univariate correlations matrix

		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
(1)	AA	1.000
(2)	ACC	0.135	1.000
(3)	ELITE	0.050***	0.006	1.000
(4)	Post_ELITE	−0.004	−0.043**	0.659***	1.000
(5)	Size	−0.002	−0.019	0.093***	0.066***	1.000
(6)	CFO_Vol	−0.008**	−0.009	0.026	−0.011	0.002	1.000
(7)	Neg_Earn	−0.013***	−0.064***	−0.072***	−0.006	−0.053***	−0.073***	1.000
(8)	ROA	0.024	0.513***	0.148***	−0.057***	0.065***	0.051***	−0.359***	1.000
(9)	Lev	0.075***	0.034**	0.027	−0.042**	0.043**	0.004	0.107***	−0.187***	1.000
(10)	Inst_Inv	−0.024	0.053**	0.042**	0.006	0.034**	0.009	0.072***	0.043**	−0.032*	1.000
(11)	Own_Conc	0.034**	0.034**	0.046***	0.007	0.126***	0.020	−0.002	0.061***	0.061	0.034**	1.000***
(12)	Group	−0.022	−0.079***	0.219***	0.254	−0.223***	−0.004	0.008	−0.085***	−0.054***	0.110***	0.138***	1.000

Note(s): This table shows the estimation results from the Pearson correlation test. Significance at *0.10; **0.05; ***0.01

Source(s): Table created by the authors

Table 5

Abnormal accruals and ELITE membership

Dependent variable: AA	McNichols (2002)
Intercept	0.032**
	2.23
ELITE	0.008***
	3.39
Post_ELITE	0.002
	0.78
ELITE × Post_ELITE	−0.006**
	−2.01
Size	0.001
	0.36
CFO_Vol	−0.001
	−0.86
Neg_Earn	−0.002
	−0.59
ROA	0.016
	1.19
Lev	0.001***
	4.06
Inst_Inv	−0.003
	−1.41
Own_Conc	0.002**
	2.09
Group	−0.003**
	−2.01
Year effects	Included
Industry effects	Included
R2	0.0181
F-value	3.02
Prob. >F	0.0000
Root MSE	0.0446
No. of observations	3.446

Note(s): This table shows the estimation results from Model (3). All variables are defined in Appendix and discussed in Section 4. Parameters are estimated using robust standard errors. Significance at *0.10; **0.05; ***0.01

Source(s): Table created by the authors

Table 6

Timely loss recognition and ELITE membership

Dependent variable: ACC	(A) Pre-ELITE entry	(B) Post-ELITE entry
Intercept	−0.014	−0.006
	−0.96	−0.47
NCFO	0.020**	−0.001
	2.02	−0.08
CFO	0.017	−0.066
	0.20	−0.99
NCFOxCFO	0.162	0.070
	0.41	0.72
ELITE	0.006	0.002
	0.87	0.33
ELITExNCFO	−0.015	−0.006
	−1.09	−0.48
ELITExCFO	−0.084	−0.009
	−0.68	−0.10
ELITExNCFOxCFO	−0.970**	−0.805***
	−2.19	−3.79
Diff-in-Diff
(β7 Post-ELITE Entry)	0.165*
(β7 Pre-ELITE Entry)	1.92
Year effects	Included	Included
Industry effects	Included	Included
R2	0.0665	0.0695
F-value	4.77	7.71
Prob. >F	0.0000	0.0000
Root MSE	0.0363	0.0383
No. of observations	1.360	2.086

Note(s): This table shows the estimation results from Model (2). All variables are defined in Appendix and discussed in Section 4. Parameters are estimated using robust standard errors. Significance at *0.10; **0.05; ***0.01

Source(s): Table created by the authors

Table A1

Variables summary

Variable	Description
Matching dimensions
Sales	Total annual sales
Net_Income	The difference between revenues and expenses
Size	Total assets’ natural logarithm
ROS	The ratio between operating income and sales
Leverage	Proxy for the overall indebtedness, computed as total debts over equity
Dependent variable
AA	Magnitude of abnormal working capital accruals which represent a proxy for earnings management, obtained employing the McNichols (2002) model, as defined in Section 4.2.1. AA are scaled by the end-of-the-year total assets and are considered in absolute value
ACC	The total operating accruals which represent a proxy for timely loss recognition, obtained the model developed by Basu (1997) as modified by Ball and Shivakumar (2005). ACC are scaled by the end-of-the-year total assets
Main variable
ELITE	Dummy variable that takes the value 1 for firms belonging to the ELITE program. Otherwise, the firm belongs to the control group
Control variables
Post_ELITE	Dummy variable that takes the value of 1 when the observation of the ELITE firm and its control refers to the period post-ELITE entry
Size	The annual sales’ natural logarithm
CFO_Vol	The volatility of operating cash flows, computed as the year-over-year relative changes
Neg_Earn	Dummy variable that takes the value of 1 for each firm-year observation with a reported loss
ROA	Proxy for profitability, defined as the ratio between operating income and total assets
Lev	Proxy for the overall indebtedness, computed as total debts over equity
Inst_Inv	Dummy variable that takes the value of 1 if at least one institutional investor owns more than 2% of the share
Own_Conc	Categorical variable that assumes the following values: (1) 1 in case of an average ownership concentration lower than 0.25; (2) 2 in case of an average ownership concentration between 0.5 and 0.25; (3) 3 in case of an average indirect control higher than 0.5; (4) 4 in case of an average direct control higher than 0.5
Group	Dummy variable that takes the value of 1 if the firm directly own subsidiaries at a stake higher than 50%
CFO	The operating cash flows scaled by the average total assets
NCFO	Dummy variable that equals to 1 for negative operating cash flows

Note(s): All non-dichotomous variables have been winsorized at the upper and lowest 1% of the distribution

Source(s): Appendix created by the authors

Notes

1. This study does not examine tax incentives as tax-motivated earnings management needs to be considered when analyzing different types of financial statements (consolidated vs individual) (Leuz and Wüstemann, 2003; Watrin et al., 2014). Hence, this study focuses on individual financial statements only.

2. Specifically, the ELITE program was born from the collaboration between Borsa Italiana (a company that was part of the LSEG) and the most important Italian institutions and organizations (i.e. Confindustria).

3. The “Rapporto Regionale PMI” drawn up by Confindustria in collaboration with Cerved, indicates and aggregate turnover of Italian SMEs of approximately € 2.000 billion.

4. Or more than 5 million and with a rate of growth exceeding 15% in the previous year.

5. As mentioned, access to the ELITE program is subject to the fulfillment of specific economic requirements. This assessment was carried out on the latest available accounting data. For instance, if a firm joined ELITE in 2012, the fulfillment of requirements was assessed according to the 2011 annual report.

6. This means that once firm i was matched with firm j, both firms were removed and not considered for further matching (Guo and Fraser, 2014).

7. Since the length of the post-ELITE period changes in accordance with the year of firms’ joining ELITE, the number of firm-year observations is not equal to 680 x 9 (6,120).

8. All variables were scaled by the average total assets.

WCA is defined as [(current assets – cash – short_term investments) – (current liabilities – short_term debt)].

9. We did not control for IFRS effects (Cameran and Campa, 2020) as only 7% of the firms prepare their financial statements under IFRS standards. We also did not control for companies with audited annual reports as 97.8% of the observations present an audited annual report. Finally, few companies are linked to listed firms and, therefore, we did not control whether firms included in our sample are subsidiaries of listed firms. Nevertheless, results are confirmed also controlling for these aspects.

10. Ownership concentration (%Own) is defined as a categorical variable that assumes the following values: (1) 1 in case of an average ownership concentration lower than 0.25; (2) 2 in case of an average ownership concentration between 0.5 and 0.25; (3) 3 in case of an average indirect control higher than 0.5; (4) 4 in case of an average direct control higher than 0.5.

11. As robustness test, we relaunch Model 2 including certain control variables. The selection of the variables is based on the previous literature (e.g. Watts, 2003; Qiang, 2007; LaFond and Watts, 2008; Garcia Lara et al., 2020). Specifically, they are: (1) Size, defined as the annual sales’ natural logarithm; (2) Leverage, computed as the ratio between debt and equity; (3) Big4, which is a dummy that takes the value of 1 if a company is audited by one of the Big Four audit firms (Deloitte, EY, KPMG or PWC). In particular, following a common approach in the conservatism literature, we add the control variables to the timely loss recognition model by interacting with them the interaction variable “CFOxNCFO”. For each control variable K, we add the vector “A, AxCFO, AxNCFO, AxCFOxNCFO”. Empirical findings reveal that the effect of the ELITE program was beneficial as ELITE firms’ timely loss recognition significantly improved relative to the control group, confirming what emerges from the main regression model.

Appendix

Table 1

Sample composition

Panel A – Sample selection process
Population of Italian ELITE firms (from 2012 to 2019)	830
ELITE firms which joined the program in 2012, 2013 and 2019	−304
Financial companies	−4
Financial and accounting data not available	−133
Companies with implausible values	−43
Companies that were listed during the reference period (2011–2020)	−6
Basic sample	340

Source(s): Table created by the authors