Size is another variable that determines board effectiveness in its monitoring function. Even if managers have the right incentives, they can make mistakes (Sah & Stiglitz, 1986, 1991). Given that group members can disagree, group decisions imply differing opinions, requiring some sort of consensus on decisions. There should be less risk in decisions made by group consensus than in decisions made by individuals.

Sah and Stiglitz (1986) developed a model in which the larger the group of decision-makers, the less likely they were to accept a bad investment project. And Sah and Stiglitz (1991) found that highly variable returns, implying higher levels of risk, are less likely in firms with many managers. The larger the size of the decision group, and less power for individual decision-makers, reduce firm risk (Mínguez-Vera & Martín-Ugedo, 2010).

Large boards may benefit from a greater variety of opinions and experiences and increased monitoring capacity. In contrast, coordination/communication problems are also more severe (Hermalin & Weisbach, 2003). These problems of communication and decline of individual responsibility are more pronounced in larger boards because the cost to any director of not exerting due diligence in monitoring falls in proportion to the total number of board members.

Considering these arguments, Yermack (1996) and Core et al. (1999) suggest a curvilinear relationship between board size and monitoring effectiveness, according to agency theory. In small boards, as the number of directors grows, supervision of top executives increase, resulting in an alignment of interests and riskier decisions. However, when an optimum number of directors is reached, the trend reverses. Jensen (1993) argues that when a board has more than seven or eight directors, it becomes more difficult to exert effective control over the CEO. So, the final effect it is not clear, according to this theory.

In addition, as previously mentioned, according to the power perspective, the final decision of a group represents an average among individual positions. Thus, decisions adopted by a group are, on average, less risky than those adopted by individuals (Moscovici & Zavalloni, 1969; Sah & Stiglitz, 1986). As a consequence, increasing the size of the board should reduce firm risk (Halliday et al., 2021).

Previous evidence on board size focuses mainly on its effect on firm value, and not on risk. However, most evidence indicates that board size has a negative impact on a firm's risk (Cheng, 2008; Nakano & Nguyen, 2012).

According to the power perspective, we propose the following hypothesis:

Hypothesis 3

A larger board will lead to less risky decisions adopted and thus lower levels of debt.

As in the case with the board size, the effect of insider ownership is not clear in the context of agency theory. Morck et al. (1988) argue that low levels of director ownership lead to more effective control (convergence hypothesis). This alignment of interests between shareholders and insiders can lead to riskier decisions being made (Mínguez-Vera & Martín-Ugedo, 2010). However, high levels of director ownership may lead directors to entrench themselves (entrenchment hypothesis) with the opposite effect.

In general, a large concentration of shares in the hands of a small number of shareholders will give those shareholders more power, and this power is increased when those shareholders are insiders. As a consequence, it is expected that when insiders have large holdings of shares they will be able to decide that the firm will take more risk, and incur more debt. In addition, large shareholders may be reluctant to engage in equity financing, as that would tend to dilute their ownership. To maintain the control of the firm, they prefer debt financing.

Unlike the American market, where more dispersed ownership structures have been observed, Spanish SMEs have highly concentrated ownership structures (La Porta et al., 1999; Leech & Manjón, 2002). Spanish companies are mainly controlled by internal, dominant shareholders, who frequently act simultaneously as directors and managers (Baixauli-Soler & Sanchez-Marin, 2015). In this scenario, controlling shareholders exercise considerable power.

According to previous arguments, we consider that larger insider ownership will lead to more concentrated power and thus we present the following hypothesis:

Hypothesis 4

More insider ownership will lead to riskier decisions adopted and thus higher levels of firm debt.

The sample includes Spanish non-financial SMEs, between 2014 and 2019, obtained from the SABI database. We use the size criterion proposed by the European Commission (2003), according to which SMEs employ fewer than 250 people and have either an annual turnover not exceeding €50 million, or an annual total balance sheet not exceeding €43 million. After selecting SMEs from the database, removing companies without data and with errors, we have an unbalanced panel of 2385 companies and 9715 observations.

We consider the firm's leverage (LEVER), computed as the ratio of total debt to total assets, as the dependent variable (Hernández-Cánovas et al., 2016; Hernández-Nicolás et al., 2019). Alternatively, and for robustness, the analyses have been repeated including the long-term and short-term debt ratio as dependent variables. The results do not differ significantly from those shown for total debt. The results for those two additional analyses have not been included for synthesis reasons. As independent variables, we measure the gender diversity or the proportion of women on the board, PWOMEN, as the number of female directors divided by the total number of directors (Valls Martínez et al., 2022). For duality, a dummy variable, DUAL, takes a value of 0 when the CEO and chairman of the board are the same person and 1 otherwise (Mishra & Mohanty, 2014). The board size, LNDIR, is computed as the logarithm of the number of directors on the board (Cashman et al., 2012). And insider ownership is calculated as the proportion of shares owned by the directors and CEO, DIROWN and CEOWN respectively (Fahlenbrach et al., 2010). We also include the squares of these two variables, DIROWN2 and CEOWN2, in order to account for a possible non-monotonic effect.

Finally, we consider the heterogeneity across firms by including several firm-specific control variables in the model. Variables generally used as determinants of leverage are (Brav, 2009; Gaud et al., 2005; Hall et al., 2004) the firm's size measured as the logarithm of total assets, FSIZE; the firm's age, approximated by the logarithm of the number of years the firm has been established, LFAGE; the ratio of intangible assets to total assets, INTAS; the debt cost, DEBCOS, measured as the ratio of financial expenses to total debt; the return on assets, ROA, calculated as the ratio of profits, before financial expenses and taxes, to total assets; and the interest coverage ratio, INCO, the number of times that net profit before interest and taxes related to a firm can cover interest payments.

Table 1 shows the descriptive statistics of all variables. The mean value of the proportion of women on the board is only 7.2 %. We also observe that only a third of the firms in our sample have a separate power structure relating to CEO and Chairman, although the Spanish Good Governance Code (2020) recommends the separation of those posts. The number of directors, NDIR, has a mean value of 5, just in the lower level of the recommendation of the Spanish Good Governance Code for listed firms: “The board of directors should have an optimal size to promote its efficient functioning and maximize participation. The recommended range is accordingly between five and fifteen members”. Finally, the ownership of directors has a value of 35.2% while the percentage of shares owned by the CEO is about 18.4 %. These values are higher than those reported by Mínguez-Vera and Martín-Ugedo (2010) for Spanish listed firm (3 % and 2 %, respectively).

According to the theoretical arguments, the following model is proposed to capture the relationship between the characteristics of the board of directors and the debt ratio of the firm:

Considering the hypotheses presented in Section 2, a negative sign is expected for the proportion of female directors, PWOMEN, the dummy that indicate the separation of the posts of CEO and Chairman, DUAL and for the board size, LNDIR, on debt ratio, LEVER. According to the power perspective, we expect a positive relationship between the proportion of shares owned by board members, DIROWN and the CEO, CEOWN and firm debt.

For the control variables, there are several arguments why firm size is related to capital structure (Kieschnick & Moussawi, 2018), including economies of scale in lowering information asymmetry, and scale in transaction costs, and market access. These reasons suggest that smaller firms should have lower debt levels. So a positive sign between firm size and debt ratio is expected.

A negative relationship between the age of the firm, LAGE and the leverage level is expected. Older firms have accumulated funds and, therefore, will probably require fewer external funds (Michaelas et al., 1999). With regard to the intangible assets to total assets, INTAS, a negative sign can be obtained as a consequence of the indirect correlation between the level of intangible assets and the debt structure of the firm (Parsons & Titman, 2009). The expected sign for the debt cost, DEBCOS is negative (firms with higher cost of debt will have a lower debt ratio) (Barry et al., 2008).

For Return on Assets, ROA, Modigliani and Miller (1963), argue that tax shielding would induce profitable companies to use more debt. In contrast, the pecking order theory (Myers & Majluf, 1984) concludes that profitable firms would primarily finance with retained earnings rather than external financing. Finally, the interest coverage ratio, INCO (Robinson et al., 2015), shows the ability of a firm to repay its debt (Zarebski & Dimovski, 2012) and it is a measure of the riskiness of a firm (Andrade & Kaplan, 1998). Risk of insolvency is higher when earnings are relatively low with respect to interest expenses. Consequently, a positive relationship is expected.

Equations are estimated using a panel data methodology, applying the System GMM technique (Arellano & Bover, 1995; Blundell and Bond, 1998). This methodology makes it possible to control for individual heterogeneity (Himmelberg et al., 1999). These are unobserved variables that affect each of the firms in the sample unequally, but which are constant over time and which directly affect the decisions made by these units. Descriptions of the kind of things usually considered under this heading are entrepreneurial capacity, operational efficiency, capitalization of experience, and access to technology. The methodology also controls for macroeconomic effects on the dependent variable (those factors that equally affect all the companies in the study). In addition, System GMM estimation solves the endogeneity problem that may appear when the independent variables and the dependent variable are determined simultaneously, as is the case in the present study (for example, women can choose to serve on board of firms with less debt because of their risk aversion). In order to remove any possible bias arising from simultaneous estimation, this methodology estimates a system of two simultaneous equations. The first equation uses variables in levels (first differences instruments) and the other uses variables in first differences (lagged with respect to instruments).

Specification tests proposed by Arellano and Bover (1995) are applied to examine the validity of the instruments in System GMM estimation. First, the Arellano–Bond test for the serial correlation is adopted to test whether there is a second-order serial correlation in the first-differenced residuals. The null hypothesis is that the residuals are serially uncorrelated. If the null hypothesis cannot be rejected, it provides evidence that there is no second-order serial correlation and the System GMM estimator is consistent. Second, the Sargan test is applied to test the null hypothesis of instrument validity and the validity of the additional moment restriction necessary for system GMM, respectively. Failure to reject this null hypothesis means that the instruments are valid.

This methodology has important advantages over others. The OLS (Ordinary Least Squares) estimations do not consider the unobservable heterogeneity of firms, and considers the variables to be exogenous. Fixed Effects estimation addresses the problem of unobserved heterogeneity, but not the endogeneity. The 2SLS (Two Stage Least Squares) considers the endogeneity, but it is not efficient in samples with a large number of firms and small number of years and it also has the problem of the choice of instruments (Arellano & Bond, 1991).

The results are shown in Tables 2 to 5. In Model 1 of Table 2, we find a negative and significant effect of the proportion of female directors (PWOMEN) on firm debt ratio (LEVER). This finding supports Hypothesis 1 (A larger proportion of women on the board will lead to lower levels of debt) and is consistent with arguments from the power perspective and social identity theory (Tajfel & Turner, 2004; Roberson, 2019; and Watson et al., 1993). According to these authors, heterogeneous groups make less risky decisions. According to these arguments, decision making will be more complicated in a heterogeneous group. To decide, a consensus must be reached and the power of the male directors (with a riskier profile) can be diluted by the preferences of the female directors. Olsen and Cox (2001) found that women investors weight risk attributes, such as the possibility of loss, more than men. The World Bank (2012) report shows that women tend to be more cautious and have a lower level of ambition. Carter and Shaw (2006) also found that the presence of women in management is associated with lower levels of debt.

This result is important since it shows how female directors influence financial decisions and the survival of the company, and it suggests an important contribution to the empirical findings.

Model 2 shows a negative relationship between the separation of responsibilities (DUAL) of CEO and Chairman and firm debt. This result supports Hypothesis 2 (If a single person holds both positions (CEO and chairman), this will lead to higher levels of debt). So, according to the power perspective, when the post of CEO and Chairman coincide in the same person the decision maker can accumulate more power, and so the firm will have a higher proportion of debt. This evidence contradicts the conclusion of Adams et al. (2005) who found no significant relationship between firm risk and duality. However, Daily and Dalton (1994a) found that duality increased the likelihood of bankruptcy. The findings of this research contribute to reducing the controversy surrounding CEO duality in leadership by furnishing empirical evidence of how CEO duality in corporate governance structures affects managerial behavior in corporate strategic management.

Finally, in Model 3, we observe a negative influence of board size (LNDIR) on firm leverage (LEVER) as anticipated in Hypothesis 3 (A larger board will lead to lower levels of debt). So, as the power perspective predicts, increasing the size of the decision-making group reduces the power of a specific decision-maker and, as a consequence, the firm risk.

In relation to the control variables, Table 2 shows a positive effect of firm size (FSIZE) on firm debt (LEVERAGE), in line with our expectations and the arguments of Kieschnick and Moussawi (2018) and Titman and Wessels (1988). The signs of the firm age, LFAGE, proportion of intangible assets, INTAS and debt cost, DEBCOS, are all as expected, and consistent with the arguments of Michaelas et al. (1999), Parsons and Titman (2009) and Barry et al. (2008), respectively. Supporting the pecking order theory (Myers & Majluf, 1984), we find that profitable firms tend to use less debt. Finally, firms with less risk, as indicated by the interest coverage ratio, INCO, have more debt in proportion (Andrade & Kaplan 1998).

Table 3 shows the effect of directors and CEO ownership, DIROWN and CEOWN, on firm leverage (LEVER).

In Model 1, we find a positive and significant influence of the directors’ ownership on firm debt, in agreement Hypothesis 4 (More insider ownership will lead to higher levels of firm debt). It is possible that high share ownership of directors gives them more power, and thus makes the firm more likely to adopt risky policies. A similar result is observed in Model 3, focusing on CEOs’ ownership, supporting Hypothesis 4. We do not find a quadratic relationship (Models 2 and 4).

This result is in the line with that of Palmer and Wiseman (1999). They showed that the stockholdings of senior management have a positive impact on managerial risk taking. Bajaj et al. (1998) suggest that in firms with low executive ownership, directors are incentivized to under-lever the firm to reduce bankruptcy risk. May (1995) also argues that CEOs with high ownership will display riskier behavior, such as selecting investments with more risk. We contribute by adding new evidence in a country with a high concentration of business ownership in small and medium-sized enterprises.

Table 4 presents all the variables together. This table confirms the previous results, namely a negative effect of the proportion of female directors (PWOMEN), duality (DUAL), and board size (LNDIR), on firm debt (LEVER), and a positive effect of director and CEO ownership (DIROWN and CEOWN). The sign of the control variables is the same as above.

Considering the results obtained in this study of the effect of the share ownership of directors and CEO on indebtedness, we set out to analyze whether the presence of female directors has a moderating effect (Baron & Kenny, 1986).

In this sense, the greater risk aversion that women can provide and the greater heterogeneity that they bring to the board of directors can cause even companies with the highest shareholding concentration to reduce their exposure to financial risk, limiting the power of those who own shares. In other words, we analyze whether the presence of women on boards of directors serves to offset the power of shareholder concentration and thus reduce the debt ratio.

Following the previous arguments, we study the moderating effect of adding women to the board on reducing the leverage of SMEs where more shares are in the hands of directors and the CEO. In Table 5, we observe that more female directors reduce the indebtedness of firms where directors (Model 1) and the CEO (Model 2) own more of the shares. So, women on the board help to reduce the level of firm debt.

The interactive term PWOMENxDIREOWN has significant and negative impact on debt ratio (β=−0.0007, p<0.01), meaning that if the percentage of female directors and directors’ ownership increase by 1 %, the firm financial risk in terms of debt ratio will decrease by 0.07 %. This can be interpreted as meaning that firms with high director ownership tend to have heterogeneous gender diverse boards which decrease the firms’ financial risk. Also, PWOMENxDIREOWN produces similar results.

So, women on the board can help to reduce the level of firm debt. This finding represents a contribution to the study of corporate governance. We are not aware of any similar analysis before. In contexts of high share ownership, such as the Spanish case, adding women to the board of directors can serve to reduce the financial risk, and therefore the bankruptcy, of companies. The presence of women directors is not only important for ethical reasons. There are also the economic reasons, that it can mitigate the devastating effects of financial crises.

In addition, in Tables 6 and 7, we analyze the interrelation of the proportion of women on boards (PDIR) with the separation of positions of CEO and Chairman, DUAL (Table 6) and the board size (Table 7).

Table 6 shows a non-significant interaction effect, which means that the presence of women on the boards of Spanish SMEs does not produce a positive effect of separation of positions on corporate debt. This occurs whether we include the CEO ownership (Model 1) or the percentage of shares of the members of the board of directors (Model 2) as the dependent variable. There are no changes in the significance or sign of the other variables.

The result is similar when the interaction between the percentage of women and the size of the board is included (Table 7). That is, there is no increase in debt as a consequence of the joint effect.

Finally, in Tables 8 and 9, we study whether the other characteristics of the board, the duality of positions (Table 8) and the board size (Table 9) serve to mitigate the positive effect of both directors’ ownership and CEO ownership.

In Table 8 it can be seen that the interaction between the DUAL variable and the percentage of shares owned by members of the board of directors (Model 1) and the CEO ownership (Model 2) is positive and significant. Therefore, unlike the percentage of women on the board, the separation of positions of Chairman and CEO does not serve to offset the positive effect that ownership has on corporate debt.

The results are similar in Table 9. A positive interaction relationship is observed between the size of the board and the directors’ ownership (Model 1) and the percentage of shares owned by the CEO (Model 2). This result highlights the importance and effectiveness of female directors in reducing debt and, therefore, business risk.

For all the models, the z1 Wald Test shows the combined significance of the coefficients of the independent variables, while z2 reports the joint significance of the time dummies. The m2 statistic shows no second-order serial correlation in the first-difference residuals in all models. This means that the models are not mis-specified. Sargan's Test shows a lack of correlation between the error term and the instrument used.

Small and Medium Enterprises are extremely important in terms of number, employment and sales. As observed in periods of financial crisis, one of the factors positively influencing their survival is avoiding very high leverage. In this manuscript we examine, from the power perspective, how several board characteristics (gender diversity, duality, board size and insider ownership) influence the firm debt ratio for a sample of Spanish SMEs. Previous studies that have addressed this topic mainly focus on listed firms, and use agency theory in their analysis. However, in the Spanish context and SMEs, the power perspective is probably a more appropriate framework.

All the results obtained support our hypotheses. We find that more female directors, a lower proportion of shares owned by directors and by the CEO, larger boards and the separation of the roles of CEO and Chairman all lead to a decrease in firm debt. We also find that women directors reduce the level of debt in firms where more shares are in the hands of the CEO and directors. Focusing on this last finding, previous evidence has already indicated that women are more risk averse than men and they are more prone to share the power. However, to our knowledge, there is no previous evidence that shows so clearly how the presence of more women on the board can offset the negative effects of high levels of insider ownership. Furthermore, we find that the other board characteristics examined, separation of positions and size, do not exert a moderating effect on the share ownership of directors and CEOs to reduce corporate debt. Finally, the combination of the proportion of women with the separation of positions and size does not cause an increase in financial risk.

These results are important as they show that the characteristics of the boards of directors are important when it comes to reducing corporate debt and preventing financial crises. The proportion of women on boards of directors is especially decisive, since they not only reduce financial risk but have a moderating effect on insider ownership. Therefore, it is important that legislation is used to increase the proportion of female directors, not only in large companies, but also in SMEs, which are the vast majority of firms in the world economy.

The main limitation of this work is that it is focuses on a single country. Other relevant variables, such as the type of director, have not been included (due to lack of data). For future research, we will try to extend the sample to different countries and legal systems.