39
of the innovation activities of the firm. R&D investments can be measured by variables such as sales, total assets and market capitalization. But, because these variables also indicate the size of the firm, the relationship may not be portrayed as it is. Therefore, as shown in the Table total assets and market capitalization have strong significance for the dependent variable, and both logarithms have negative signs. A paper by Lee (2005) shows that the market capitalisation indicates the R&D elasticity and consequently this relationship is negative in this research. The variable used also is scaled to total assets, and as a denominator has a negative value. The CEO duality has a -0.0386 billion $ effect on the R&D of the firm, and this is possibly because when the CEO is also the Chairman, it deteriorates the expenses the company can make. The t-statistics are also relatively high for columns 3 and 5. For Panel A the percentage of independent directors variable is strongly significant, while in Table 8 there was significance only for Panel B. Furthermore, when compared to Table 8 the variable risk is now significant for the fixed-effects regressions. In Panel A the effect is positive, which means that more risk raised the innovation of the firm, while for Panel B the effect is negative, and since Panel B is the period that more women were added to the board this can potentially be interpreted as if women that take more risk according to the literature result in reducing the intensity of the innovation activity.
Shareholder’s equity is significant for almost every column. Shareholder’s equity is one of the main sources of funding for research and development and thus the relationship is positive and extremely significant. Lastly, both gender diversity indexes show higher values than in Table 8, yet the absence of significancy remains the same.
40
years and not consecutively for all the years, are still retained on the sample.
Moreover, in order to control for possible outliers in the sample, the variables are winsorized at the 1% level. Additionally, the initial sample had two R&D variables, R&D filed and R&D actual, but since the first variable had fewer missing observations, it was selected as the basis for the construction of the dependent variable. More specifically, the natural logarithm of 1 plus the value of R&D is constructed. The 21 duplicate observations that were found were then dropped.
Furthermore, the variables Blau Index, Shannon Index, Ln (Total Assets), Ln (Market Cap), Risk had to be constructed. The sample has now 5,551 observations instead of 5,544 initially. There are two questions that are examined for this research; the first one is what is the effect of the percentage of females on the board on the firm’s performance? The second one is, what is the effect of the percentage of females on the board on the firm’s innovation? The common assumption behind these research questions is that following the ISS voting policy, the percentage of women in the corporate board is going to rise and consequently this will affect the company’s performance and innovation.
The methodology of this paper begins with OLS regressions using return on assets and research and development expenses as dependent variables, and following the instrumental variable (IV) is manipulated for the TSLS regressions. The IV is the lagged value of the percentage of females on the board; the reason an IV is needed is to deal with potential endogeneity. More specifically it is possible that a firm with large size and therefore more power can get the best female candidates first and therefore, other companies that are smaller in size will get the lower qualified candidates. To test the gender diversity, the percentage of females on the board is employed as an independent variable, along with the Blau Index and the Shannon index, that also measure the gender diversity. Furthermore, the dependent variable for the firm performance is the return on assets and similarly for innovation the R&D expenses of the firm. Erhardt, Werbel, and Shrader (2003), examine the effect of diversity in the board on firm performance, using the return on assets and the return on equity; They found that board diversity was positively and significantly correlated with ROE and ROA in 1998, the relationship was not significant for 1993. From the results presented in the descriptive statistics it is visible that the percentage of women in the board did increase by a 5%. The results did not show a significance for the
41
percentage of females on the board for neither firm performance, nor innovation. Blau Index was significant and negative for Panel B in the ROA regression, but since there was not any other significance, robustness checks were realised in order to find significant results for the independent variables. Blau Index was significant and negative for Panel B with fixed effects in the ROE regressions. Ellis and Keys (2003) in contrast find a positive relation of the Blau Index. Shannon Index was also significant for Panel B but the coefficient had a positive sign. It could be the case that if the sample was extended beyond 2020, then the significance of the two indexes would be stronger. The general effect of gender diversity on firm performance was negative, as measured by the percentage of females on the board, consistent with the research of Ahern and Dittmar (2012), but not significant. And the effect of the percentage of females on the board on innovation was positive, consistent with the research of An et al (2021), but again not significant. Consequently, we cannot conclude that the hypotheses that the gender diversity has an effect on firm performance and corporate innovation are valid or not, since the results did not show a strong significancy across the tables.
Concerning the limitations of this paper, the employment of patents as proxies for corporate innovation was initially intended, but after thorough research the sample of patents gathered was insufficient. More specifically, patents for US firms were available in the USPTO database, but the companies did not belong to the Russell 3000 or the S&P 1500, and the NBER database had a sample available only until 2005. After downloading data from USPTO containing the names of the firms and the number of patents and matching them by the CUSIP codes of the firms and the years, only two firms matched from the whole sample. After that, I tried to combine the company names from both USPTO and NBER databases, and merge them to one dataset, but again the company names were different, therefore no match was made.
As a result, the R&D expenses of the firms were used solely as a dependent variable, and that perhaps signifies that the results do not portray the innovation of the firm spherically. Another limitation was the omitted variable bias; more specifically the dependent variables for firm performance and innovation were correlated with the variable Officer Age, which was an omitted variable and as a result it was removed.
Furthermore, there is a limitation when concerning the data points for each firm;
Since there were a number of missing observations for some of the firms during the
42
years, there is a possibility that the results are not as accurate as they could be with a full point sample. The reason that I did not drop all firms that had missing observations for certain years was that the final sample would consist of very few firms, and therefore would not be representative. Also, the sample would possibly provide improved results when the time period will be extended; More specifically since the ISS voting policy was introduced in 2019 to 2020, the results would be more accurate and less biased if the sample included more years after the policy.
Unfortunately, this paper had a limited sample since the data were available until 2020. This research can be re-examined as more data years become available.
Furthermore, ROA that was used as the dependent variable for firm performance focuses on short term results and also on operational excellence. The effect of gender diversity is better observed under a long-termed horizon and therefore the results might be more significant if long-term financial performance measures were used.
Even though the corporate governance literature has examined in great extend the gender diversity on the board, the gender quotas have not received this attention, especially now that worldwide efforts are realized to increase the representation of women in the boardroom it should be extremely interesting to measure the consequences that the addition of women in top management position has on the firms. The US do not have a quota yet, with the exception of the State of California, but when compared cross-sectionally by future studies it should provide strong insights. Therefore, there is room for further research for a topic that is constantly developing through quotas and policies.
43