**4. Results and Discussion**

In the case of using BINARYDELAY as a factor, Welch robust ANOVA indicates (see the descriptive statistics in Table 2) that the means are different for all six independent variables at *p* < 0.001. Thus, all independent variables could potentially exhibit significance in discriminating between (non-)violators in BLR.

The conducted BLR analysis (see Table 3) testing the model specified in Section 3.4 indicates that at *p* < 0.05 level all six independent variables discriminate between (non-)violators, while the control variable BOARDSIZE is significant only at the *p* < 0.1 level. When the presence of a majority owner (MAJORITY) and board memberships in other firms (TIES) lead to a higher likelihood of violation, then in turn older managers (MANAGERAGE), women on the board (WOMAN), longer tenure (TENURE) and a larger amount of shares owned by the board members (BOARDOWNER) all reduce the likelihood of violation. Thus, H1a, H2a, H2b and H3a are supported in BLR, while H1b and H3b are rejected. Although larger boards could to a certain extent exhibit a lower likelihood of delay, the significance level of that variable does not enable to draw any ultimate conclusions, especially when considering the population size used in this study.


**Table 2.** Descriptive statistics of corporate governance variables.

Source: Own elaboration.

According to our expectation, P1 assumes a positive relationship between both variables of the board's power concentration dimension and TADV. However, our results are inconclusive. The ownership concentration variable enables the support of P1, as high levels of ownership concentration can foster risk-taking, in line with Nguyen (2011). Moreover, minority shareholders might not make much pressure as outside investors who demand more transparency (Carney 2005). Conversely, when managers hold a larger proportion of the shares, they are less likely to be engaged in TADV. As the manager-owners of the firm, they are more engaged/committed to decision-making processes, and in this case, they also have a direct responsibility to face law violations. It can be assumed, that although manager-owners have much information "in the head" (Uhlaner et al. 2007), and thus, are not in need to publish annual reports quickly, they are still more worried about the personal reputation loss and legal consequences of violations.


**Table 3.** Logistic regression model for BINARYDELAY (0—non-violator, 1—violator).

Source: Own elaboration. Notes: Average variance inflation factor (VIF) 1.30. See the model's general form in Section 3.4.

Related to the proposition P2, certain demographic characteristics should have a negative relationship with TADV, which found proof with the two variables employed. When members of the board are less risk-prone as women, have more life-experience measured as being biologically older, then the probability of TADV is lower. According to prior studies, age and gender are two relevant conditions against risk, that is, older managers and women are more risk-averse than young ones and men (Jianakoplos and Bernasek 1998; Troy et al. 2011; Ho et al. 2015). In addition, female directors are more likely to be objective and independent (Fondas 2000), therefore decreasing risk-taking (Elsaid and Ursel 2011), and thus, also following rules and official requirements to disclose financial information on time. Older managers with experience are less involved in dishonest and unethical behaviours than young ones (Troy et al. 2011; Ortiz-de-Mandojana et al. 2018). This could be due to the fact that old managers have experienced other law violations in their business life, which could have had negative consequences, for instance in the form of fees, penalties, reputation reduction, or decreases of credit ratings. Thus, they do not want to conduct more misbehaviours.

Regarding the third proposition P3 reflecting board experience, firms are supposedly less risk-taking when their managers have more experience, but proof for this was found only by using the TENURE variable. Being engaged in a firm for a longer period makes the managers more capable of consolidating financial information quicker, but also, they might have witnessed the negative consequences of TADV already before. In turn, being a board member in other firms acts in the opposite way. While multiple directorships are related to uncommon skills and strong abilities in both monitoring and advising (Falato et al. 2014; Harris and Shimizu 2004), such individuals could be busy directors who may lack the time needed to execute their monitoring well (Johnson et al. 2013; Jiraporn et al. 2009). However, some empirical research has concluded that "criticisms levelled against these directors may be unfounded" (Harris and Shimizu 2004, p. 791), and perhaps, there are other potential explanations related to this variable.

Our results show that board size is not associated with TADV. This might be because the board size in private firms is very small and many times is made up of the unique owner who is also the unique manager. In addition, when there are more members in private firms' boards, they could also be from the same family, therefore making the same decisions as they are defending the same interests (Zona 2015).

Table 4 extends the base BLR analysis by introducing different types of violators. When violators are broken into two types, that is, mild violators (up to 365 days delay) and severe violators (more than 365 days delay), an interesting feature is that the significances and effect directions of independent variables are not altered, although the magnitude of the effect of specific variables can (largely) vary. It is possible to generalize that when comparing non-violators with a specific type of violator (either mild or severe), in case of all independent variables, the effect is always stronger in the case of severe violators. Many independent variables are not significant when distinguishing between mild and severe violators, namely only two variables (i.e., MANAGERAGE and TENURE) are significant at *p* < 0.01. Thus, violators differ more from non-violators than different violators differ between themselves.

As the effects in the case of mild violators are not as strong, we can suggest that perhaps the decision to follow or not the disclosure regulation in the case of mild violators could be the case of "carelessness". Such managers do not really want to violate the regulation, but for instance, when the composition of the annual report is left "to the last minute", it cannot be prepared on time and perhaps not all board members can accept and sign the report enough quickly. The latter "carelessness" logic is corroborated by prior studies such as Cheng (2008) or Arosa et al. (2013).


**Table 4.** Additional logistic regression models for the subpopulations of BINARYDELAY in comparison with the base model.

Source: Own elaboration. Note: All firms, 54,081 non-violators and 23,131 violators, SP1 54,081 non-violators and 15,917 mild violators, SP2 54,081 non-violators and 7214 severe violators, SP3 15,917 mild violators and 7214 severe violators. See the model's general form in Section 3.4.

Table 5 provides additional BLR models in case the applied population of firms is broken in two based on either median size or age of firms. Likewise, with the violation context, the BLRs focusing on different size or age groups indicate that the variables are significant and the effects are in the same direction, but the magnitudes of the effects vary. Still, unlike with the violation context, there is more variation with respect to whether smaller/larger size or younger/older age of firms leads to the independent variable having a weaker/stronger effect in distinguishing between (non-)violators.

**Table 5.** Additional logistic regression models of BINARYDELAY for smaller/larger and younger/older firms in comparison with the base model.


Source: Own elaboration. Note: For the distinction of smaller/larger and younger/older firms, the population is broken in two based on median size (natural logarithm of total assets) 9.98 or median age (firm age in years at 30 June 2015) 7.34. See the model's general form in Section 3.4.

When the BLR is run with another fiscal year (i.e., 2015), the results are not altered (see Table A1). Namely, the only variable clearly not significant, likewise with the base model calculated by using the fiscal year 2014, is the control variable BOARDSIZE. In turn, in the case of independent variables, the signs of the coefficients remain the same and absolute values of the coefficients are very similar, like for the base model documented in Table 3. Thus, the results are robust with respect to the year chosen for analysis. Table A1 also shows the bootstrapping results for the year 2014. In a 100-sample bootstrapping, the signs of independent variables' coefficients do not change for the lower and upper 95% confidence intervals, thus the subpopulations of firms are quite similar to the findings obtained

with the base regression model on the whole population documented in Table 3. The bootstrapping result is an expected scenario based on the age and size contexts in Table 5, which also do not indicate the change in variables' signs.

The results of the study are consolidated into Table 6, which in future research can be used as a benchmark for the association of timely accounting disclosure violation and corporate governance attributes in SMEs. As a contribution to the literature, we found that certain demographic attributes in the board make them less likely to be violators of the accounting regulation, while the power concentration and experience on the board can lead to varying violation behaviour, depending on what variable of the specific dimension is considered. In addition, corporate governance characteristics have more pronounced effects on the violation probability when the violation becomes more severe.


**Table 6.** Summary of the associations found in this study.

Source: Own elaboration. Note: The first column includes the result for the three research propositions (either true, inconclusive or false; inconclusive means one true and one false evidence), while the second column includes the result for the acceptance/rejection of postulated six hypotheses.

#### **5. Conclusions and Future Research**

The objective of this research was to analyse the association between corporate governance characteristics and timely accounting information disclosure violations in private SMEs. Relying on an SME population in a developed European economy, namely Estonia, a set of theoretically motivated corporate governance (independent) variables was studied with annual report submission delays (as the dependent variable) in different logistic regression analyses. Evidence was found that certain demographic diversity in the board (as portrayed by women on the board and managers' older age) reduces the likelihood of violation, while variables portraying power concentration (managerial ownership and ownership concentration) and board experience (tenure length and business ties) provided mixed results.

Varying stakeholders can benefit from the results of this study. First, as non-timely disclosure has been proven to be associated with either financial distress or bankruptcy (Altman et al. 2010; Lukason 2013; Luypaert et al. 2016; Lukason and Camacho-Miñano 2019), creditors can account specific corporate governance characteristics in case of lengthy delays. In the latter circumstance, financial information from the past can already be obsolete, and thus, non-financial variables could be of remarkable value to predict distress or bankruptcy. Second, based on the results, state institutions monitoring timely submission have a better understanding, which corporate governance characteristics in association with firm size and age can lead to a law violation with a higher likelihood. The latter enables, for instance, the targeting of likely lengthy violators earlier to guarantee better transparency in the business environment. Last but not least, as the general foundation of this study was risk behaviour more broadly, the findings can provide valuable hints, which corporate governance characteristics could potentially be triggers for other risk behaviour types.

Finally, this paper is not free from limitations, being fully related to future research proposals. First, our paper is focused on one country, Estonia, and thus, our findings could be altered by the peculiarities of this country, for example, the accounting disclosure (violation) legal framework and its implementation. Future research could be conducted in other countries in order to check whether cultural or legal settings have an impact on how corporate governance is linked to accounting disclosure violations. Second, our approach to corporate governance is limited to a certain set of dimensions and variables portraying them, and thus, future research could be enhanced to account more for psychological or personal characteristics such as ethical level, past violation behaviour or past training/education of managers. Third, although the results were validated with another fiscal year, the violations could be studied in a longer time frame, to either detect certain disclosure pattern changes or even consider corporate governance changes, should these occur.

**Author Contributions:** Both authors contributed to all parts. Both authors have read and agreed to the published version of the manuscript.

**Funding:** The first author acknowledges financial support from the University of Tartu Foundation's Ernst Jaakson Commemorative Scholarship, the Estonian Research Council's grant PRG791 "Innovation Complementarities and Productivity Growth" and the Estonian Research Infrastructures Roadmap project "Infotechnological Mobility Observatory (IMO)".

**Acknowledgments:** Authors thank the Estonian Centre of Registers and Information Systems for the data.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

**Table A1.** Model composed with another fiscal year 2015 and bootstrapping results for the year 2014.


Source: Own elaboration. Notes: BS—bootstrapping, CI—confidence interval. BS results were obtained with 100 bootstrap samples for the year 2014 population. B and Sig.—coefficient and *p*-value either for the whole populations from 2014 or 2015.

#### **References**


Adams, Renee B., and Daniel Ferreira. 2007. A theory of friendly boards. *Journal of Finance* 62: 217–50. [CrossRef]


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
