The impact of CEO education on convertible bond issuance

Abstract

We examine the impact of managerial education on a firm’s decision to issue convertible bonds instead of standard, non-hybrid securities. Upper echelons theory argues that better managerial education attainment fosters a higher ability to process complex information and tolerate ambiguity. Exploiting convertible bonds’ higher degree of complexity, relative to seasoned equity and straight bonds, we hypothesise that Chief Executive Officers (CEOs) with higher education levels are more likely to issue convertible bonds instead of non-hybrid security types. A multinomial probit model analysing firms’ choice between convertibles, seasoned equity, and straight bonds provides evidence consistent with this hypothesis. A one-level increase in a CEO’s highest academic degree raises the likelihood of substituting convertibles for non-hybrid securities by 2.86%, after controlling for standard corporate security choice determinants. Chief Financial Officer (CFO) education levels, by contrast, have no significant impact on convertible bond issuance. Consistent with higher managerial education attainment being associated with a higher cognitive ability, we also find a positive association between CEO education levels and the level of complexity in convertible bond design. Our findings, which hold under a range of alternative specifications, illustrate the influence of CEOs’ personal characteristics on securities issuance and design.

KEYWORDS:

• CEO education
• complexity
• convertible debt
• security choice
• security design

JEL:

• G32

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Plöckingera et al. (2016) and Wang et al. (2016) provide reviews of papers examining the upper echelons theory.

2 For brevity, we have limited this overview to studies on the relation between managerial characteristics and corporate finance decisions. A number of studies examine the role of managerial characteristics in an asset pricing context. Examples incluKaplan, Klebanov, and Sorensen (2012), who use CEO assessment scores to document an association between managers’ general ability and private equity and buyout fund performance, and Beber and Fabbri (2012), who document an association between holding an MBA and risk-taking behaviour in foreign currency derivatives markets.

3 Borgia and Newman (2012) use a mixed-method approach on a sample of Chinese small and medium sized Enterprises and do not find an effect of managerial education on leverage. They do not consider convertible bonds. Guo (2013), in an unpublished working paper, examines the effect of firm performance (i.e., returns and value) attributable to the CEO on convertible bond issuance. The paper does not examine CEO education or other CEO demographic characteristics.

4 Theory does not yield clear predictions on the role of managers’ general ability (as, e.g., evaluated through assessment scores, as in Kaplan, Klebanov, and Sorensen 2012) and convertible bond issuance – it really only predicts a role for cognitive ability, as this ability in particular helps managers to deal with and tolerate the higher complexity and ambiguity of convertibles.

5 We thank an anonymous referee for pointing out this alternative prediction. Yet another prediction arises from the “free lunch theory” on convertible bonds, which argues that managers choose these securities because they mistakenly believe that convertibles are better for the firm than both equity or straight bonds in all possible contingencies (Brennan and Schwartz 1988). If lower-educated managers are more prone to this fallacy, this would also lead to a hypothesis of a negative impact of CEO education on convertible bond issuance. However, Dong, Dutordoir, and Veld (2018) find no evidence that managers are deluded about the true cost of convertibles their interview study on convertible bonds.

6 Anecdotal evidence that we gathered from business school colleagues teaching at various education levels suggests that convertible debt is not systematically covered in finance courses, although it does appear to be mentioned in the textbooks typically used by business school professors. Thus, even for those managers who have a business school education, it is unsure whether they have effectively learnt about convertible debt as part of their curriculum.

7 We acknowledge it is difficult to empirically disentangle a complexity viewpoint from an innovation viewpoint, since those security design features that make convertibles potentially more innovative are also likely to make them more complex than other securities. Some studies argue that a higher cognitive ability can manifest itself as a greater ability to absorb new ideas, and therefore an increased tendency toward accepting innovations (Barker and Mueller 2002), suggesting that an innovation viewpoint could be narrowly associated with, and even embedded in a complexity viewpoint on the role of managerial education in convertible bond issuance and design.

8 Consistent with Lewis and Verwijmeren (2011), we only have a very small number of convertible preferred stocks in the raw sample, i.e. 45 issues in total. Furthermore, there are only three mandatory convertibles in our sample, suggesting that plain vanilla, traditional convertibles are by far the most important convertible bond class (albeit with variations in convertible bond design, as outlined further). Consistent with Henderson and Zhao (2014), a SEC EDGAR search reveals that a fraction of convertible bond issues (13.37% of the final sample) are issued together with seasoned equity. Our baseline security choice model results remain unaltered if we remove these combined offerings from the data set.

9 In an unreported analysis, we find that the characteristics of the subsample of issuers for which we are able to retrieve CEO educational data are not materially different from those of the subsample of issuers for which we are not able to retrieve these data. Thus, the requirement of availability of CEO characteristics does not seem to introduce a selection bias in our sample. Detailed results of all unreported tests referred to in the paper are available from the corresponding author.

10 Only 4% of the CEOs in our sample are female. When we control for CEO gender, the results remain unaltered, with a Female CEO dummy having no significant impact.

11 In untabulated tests, we find that the mean values of the following variables are statistically significantly different when comparing treated with non-treated observations in our un-matched, baseline dataset: Leverage, Tax, Issue size, ROA, and Market to book. In contrast, in the reduced dataset of matched observations, no difference remains statistically significant, indicating that this dataset is well-balanced. Results of the first-stage matching equation are available upon request.

12 In an unreported additional test, we estimate an inverse-probability-weighted regression consistent with the approach in King, Srivastav, and Williams (2016). We use the teffects ipwra command in Stata for this purpose. The probabilities account for the fact that better-educated CEOs may have a higher likelihood of choosing larger and more profitable firms in certain industries (King, Srivastav, and Williams 2016). We obtain a positive and statistically significant average treatment effect (ATE) for the effect of Highly educated CEO on the likelihood of a convertible bond issue rather than an equity issue, further corroborating our baseline result.

13 We use the heckprobit command in Stata, which is based on a maximum-likelihood estimation of both a selection equation and the probit model we are primarily interested in. We have also evaluated the suitability of other potential approaches to deal with endogeneity but deemed these unsuitable for our empirical design. More particularly, instrumental variable and General Method of Moments (GMM) estimations are problematic in our research context, given that both our outcome and test variables are non-continuous in nature. An analysis of changes in convertible bond issuance around CEO turnovers is not possible either. We only have 121 CEO turnovers in our dataset, among which a mere 19 are closely preceded or followed by convertible bond issues. This makes the resulting subsample too small for meaningful empirical tests.

14 This is based on issue proceeds of the 2,289 offerings in the final sample. For multinomial probit models we only include observations underwritten by the top 15 investment banks, as the security choice model does not converge otherwise.

15 Examples include Grimm and Smith (1991), Chevalier and Ellison (1999), Graham and Harvey (2001), Barker and Mueller (2002), Bertrand and Schoar (2003), Bhagat, Bolton, and Subramanian (2010), King, Srivastav, and Williams (2016), and Miller and Xu (2019).

16 In total, 12.5% of the convertibles in our dataset have a hard call feature, and 8.64% have a put feature.