I reported in a recent post about an article published recently by The Commentator, ‘Socialism’s Trojan Horse: “improved” gender diversity in the boardroom’:
I’m pleased to report the article has been well received, and the comments well-considered and interesting. In the course of replying to one person’s comments I invited her to provide me with evidence of the oft-claimed positive causal link between ‘improved’ gender diversity in the boardroom and enhanced corporate performance, which remains a ‘missing link’. Another lady replied to my request for evidence and cited two studies, one about American firms, the other about Danish firms. The reports are very technical in parts so I invited Michael Klein, the renowned commentator on the content of research papers, to review them. He kindly agreed to do so, and the remainder of this post is his commentary. My thanks to Michael for this.
STUDY ON AMERICAN FIRMS
‘The diversity of corporate board committees and financial performance’ (Carter et al.)
There are some oddities in table 3 which aren’t addressed in the text, so far as I can tell. Though Table 3 shows that the share of women on boards had a positive effect on performance (Tobin’s q) it doesn’t show an effect of Tobin’s q on share of women in boards, which one could interpret as evidence against the assumption that it is mostly well-performing firms which add women to their boards. However, the oddity is the variable ‘Additional directorships’ which shows in Equation 1 a coefficient of -.242 and in Equation 2 a coefficient of 1.387. This means that holding more than one directorship had a moderate and negative effect on firm performance, but it had a strong and positive effect on the share of women on boards.
I would interpret this result as indicating that the authors had a great number of women in their dataset who held not one but several directorships. Hence, they didn’t measure the effect of diversity, because to measure diversity you’d need different women, but of sameness, because the same women who held several directorships were measured time and time again… Given that boards include 11.4% women, I would consider this one a result-wrecker.
Furthermore, total assets made a negative contribution to Tobin’s q, but a positive one to women’s share in boardrooms, hence, it was asset wealth which explained board composition to a larger degree than financial performance and it was firms with lower assets which explained higher profits to a greater degree than women’s share on boards (equation 1). And, last but not least, it wasn’t women’s share on boards which had a positive effect in Tobin’s q (financial performance), but women’s share on boards of firms in service industries (Variable D8, Table 3), and particularly not on boards in mining industries (D2) and retail (D7). Again, the authors read more into their results than there is to be found, and again I would think that the conclusion they draw isn’t backed up by data. I’m not surprised that this paper didn’t mature into a contribution to a scientific journal. I guess it would have had a number of problems getting over the peer-review hurdle….
STUDY ON DANISH FIRMS
‘Do Women in top management affect firm performance? A panel study of 2,500 Danish Firms’ (Smith et al.)
Overall assessment: It’s awful!
(1) the 2,500 firms examined in the paper included 46 with a female CEO in 1993 and 99 in 2001.
(2) the authors label their variables (e.g. in Table III) consistently Board Top CEOs 1993-2001 implying panel data, but they have only a follow-up with two time points, hardly what the label ‘panel’ would lead us to expect.
(3) Again Table III shows a correlation-coefficient of 0.063 (5% significance level) between having a female CEO and gross profits. That is rather minuscule and given that only 46/99 firms had a female CEO I wouldn’t be surprised if this is a statistical artefact. I can, however, say nothing about it, because information about the quality of the model, usually included in tables (and as a matter of scientific rigour required) is not provided.
(4) Table VII displays coefficients of an Ordinary Least Square Regression (‘OLSR’) – as does Table III – and shows higher education of CEO to be highly correlated with Gross profit, beta= .283. Now, why have the authors not included higher education in Table III, as a control? Guess why? Because the effect between female CEO and Gross profit would vanish into thin air, leaving only higher education. I will eat my Tottenham Hotspur cap if this is not what happens when you calculate an OLSR with female CEO and higher education both independent variables to explain gross profit.
(5) By the way, gross profit is hardly a good measure of financial performance, it depends on too many exogenous variables beyond the control of CEOs. At least, it would have been necessary to adjust gross profits to industry means… Anyway, this study proves that education matters, it doesn’t prove that female CEOs or female share in boardrooms have a positive effect on firm performance.
(6) Just to repeat myself: Why is it that female composition is expected to influence financial performance? What is the theory behind the claim? Ask Smith, Smith and Vernier and you will draw a complete blank.