The question of whether the educational and the professional background of Chief Executive Officers play an active role in determining corporate outcomes has been addressed quite extensively by the related literature. Although the studies on the matter are rather fragmented, it is possible to identify a common starting point in the “Upper Echelons Theory”, initially theorized by Hambrick and Mason (1984). The latter states that the executives’ background is a strong predictor of organizational results, and it has often been considered as an important tool for hiring purposes. This analysis follows that direction. More specifically, references such as Gottesman and Morey, who tie CEO educational background with firm performance by citing the higher degree of intelligence, the refined methodologies and the network that educated executives are supposed to have over non educated ones (Gottesman & Morey, 2006), are addressed. It is thus worth to deep dive into CEOs’ education and work experience by investigating their impact on three major dimensions: company performance, export intensity and research and development (R&D) intensity. These three are examined as dependent variables in a statistical framework and are represented by the following responses: return on assets, international revenues to total revenues, and R&D expenses to total revenues, respectively.
· By predictor (the specific type of background studied)
o Education
o Work experience
· By response (the specific corporate dimension analyzed)
o Company performance
o Export intensity
o R&D intensity
The sample considered comprises the 100 European companies with the highest market cap which have had a unique Chief Executive Officer over the five years from 2017 to 2021. Compared to the existing literature’s samples, the locational choice has been “rationalized”: the attention is not posed on a specific country but is rather spread over the whole Europe. This decision is due to the willingness of both exploring a geographical dimension that still has not been extensively tackled before and avoiding possible country-specific effects by broadening the overall point of view adopted. On the other hand, the market cap criterion has been implemented to ensure that efficient comparisons within large enough companies could be performed. Finally, the five years span has been defined to account for the initial phase in which an executive needs to develop a clear understanding of the new context before finally getting into full operation.
For each of the three dependent variables previously introduced, a specific model is built. The three all share the same independent variables, which are defined with the objective of measuring CEOs’ background as precisely as possible. In particular, the quantities explored are the level of degree held, the academic field tackled and the affinity with the previous corporate role. Coming to the overall analytical process, this can be broken down into four steps which have been repeated for each of the three models.
1. Univariate analysis: one sub-model is built per each independent variable with the aim of considering one explanatory variable at the time and checking for differences.
2. Multivariate analysis: the starting point is now a complete model that includes all the explanatory variables available. The latter is then compared with multiple reduced models, each having a different explanatory variable removed.
3. Logarithmic transformations: after logarithmically transforming the continuous, strictly positive predictors, the most relevant analyses are repeated. At the end of this step, a final conclusive model is identified.
4. Final checks: for the final model, three distinct situations are controlled (residuals’ behavior, outliers’ presence and heteroskedasticity).
. , * , ** , *** indicates significance at the 90%, 95%, 99% and 99.9% level, respectively
From the analysis of the three models emerges that there is a single dimension which consistently has more than one significant level: the one accounting for the academic field. Moreover, it is interesting to notice that its effect varies based on the dependent variable in object. CEOs with a background in social sciences are found to have a positive impact on return-on-assets, while at the same time they turn out to have a negative impact on research and development intensity. The opposite holds for CEOs with a technical background: the effect on return-on-assets is negative, while the one on R&D intensity is positive. For what concerns the educational level dimension, among the factors describing the different types of degrees that CEOs hold, the only one that is consistently significant is the one that measures observations without a degree. However, the latter is not reliable as the percentage of individuals within the sample who do not have started or completed their undergraduate studies is lower than 3%. Finally, professional background variables are much less statistically significant: the only model in which they have some relevance is the one investigating export intensity, which also happens to be the only one that meaningfully suffers from heteroskedasticity. Its results can therefore only be considered with some degree of reservation. From a general standpoint, overall all the models present some structure in the error, outliers and high leverage observations, as well as a rather low R-squared. These limitations are however in line with the ones from the models addressed in the literature and are regarded as hardly surmountable.
On another note, certain limitations are also identified, such as the fact that pairing different companies from different industries makes comparisons more difficult. These limitations are appreciated as starting points to determine new directions for future studies: all of them are thus analyzed and for each a specific recommendation on how to improve is proposed.
Comments