Journal Article Critique
Professor Thomas Spencer, Ph.D.
RSCH-8250Z-1 – Advanced Quantitative Reasoning & Analysis
July 13, 2013
Hsu, Chen, and Cheng (2013) used a cross-sectional study to analyze 187 small and medium enterprises (SMEs) in Taiwan to determine if CEO attributes contributed to internationalization success as defined by firm performance. Based upon upper echelon and information processing theories, the researchers hypothesized that five CEO characteristics (age, tenure, education, international experience, and duality – that is, CEO also as board chairman) were significant moderating effects upon international firm performance. Using regression analyses, the researchers concluded that four of the predictors, excluding CEO tenure, had a significant moderating effect upon firm performance.
The rationale for the study was described by the researchers as a need to analyze CEO characteristic attributes because such CEO predictors had not been addressed in previous studies of internationalization and firm performance. Therefore, this study appeared to fill a gap in the literature. The three authors seemed to be a well-balanced team to tackle this particular academic issue. They individually had well-suited combinations of experience in finance, risk management, and international business.
The literature review was extensive and was broken down logically into two main sections: Internationalization/Firm Performance and Upper Echelon/Information Processing theories. Upper Echelon theory and Information Processing theory appeared to adequately address the study’s main thrust of analyzing CEO personal attributes in a complex business setting at the individual level (versus organizational structure). In addition, the hypotheses section, which contained discussions of the study’s six hypotheses, also included copious literature references. However, out of the study’s 88 referenced materials, only 8 references were written within 5 years of the article’s publication date. The 88 references for this relatively small study seemed to be excessive, so there is a concern that the authors may have been making up for a lack of recent relevant articles by citing an abundance of older articles and studies. This could have been driven by the perceived gap in the literature, which is frequently seen in this kind of situation. Breaking new ground in a well-documented topic sometimes has to be performed without the support of recent academic journal articles.
One area of general concern was the study’s focus upon SMEs and the off-shore internationalization of small and medium business dealings. It seemed more appropriate that the focus upon internationalization would be better applied to larger enterprises which would be better suited for this CEO study because larger organizations would logically have more abundant resources of all kinds and more experienced CEOs who could better internationalize their businesses versus the small and medium enterprises analyzed by this study. However, one must realize that Taiwan is a small island nation which must perforce rely upon many international partners for resources, commerce, and labor. Given the rather unique context of Taiwan, it seems plausible that even small and medium Taiwanese businesses would indeed be vitally concerned with international business success and their CEO’s part in that success.
The six hypotheses stated in the study appeared to be logically driven by the identified gap in the academic literature. The only possible bias in the approach of the topic seemed to be frequent references to the age of CEOs and the fact, revealed below, that an increase in CEO age appears to significantly relate negatively to international firm performance. Given that this is an Asian nation where age typically equates with increased knowledge, the study’s finding that CEO age does not actually equate to increased firm performance may have been a preconceived bias by the researchers and they may have been keen to use the study results to help move aging CEOs out of their hierarchical positions or may have a bias towards changing CEO selection criteria. At a minimum, the results of this study, if truly applicable to the larger population represented by the sample, have the potential to make many aging Asian CEOs, and their boards, uncomfortable.
Another bias in the study sample is that the companies are all publicly-held companies, therefore private companies were excluded from the sample – presumably because it was a convenience sample. Comparable data for private companies is difficult to obtain.
For the purposes of obtaining company data for this study, SMEs were defined by the researchers as Taiwanese public companies that had fewer than 200 employees. These companies were selected from the Taiwan Stock Exchange and GreTai Securities Market. Financial data for the companies was obtained from the Taiwan Economic Journal Databank and CEO demographic data was manually gathered from company reports. Any companies that lacked any of the required data were removed from the study sample. As stated earlier, n=187. Field (2013) suggested that for a regression study, the minimum sample size should be 104 cases plus the number of predictors, or in this case, 104 + 5 = 109. Therefore the minimum sample size for the regression was met.
Because the data was gathered from public sources, no IRB intervention was necessary in the data gathering process. There are no apparent ethical concerns about the study, particularly since the data was randomly selected and anonymous in nature. There is no individually identifiable CEO information in the dataset.
The authors addressed several concerns of a regression study. They used a generalized least squares (GLS) regression method to address the problem of cross-sectional heteroskedasticity and within-unit serial correlation. The GLS method was used to test the study’s hypotheses. The Hausman test was used to determine that there were no significant correlations between independent variables. However, elsewhere in the study, the researchers related that two variables did indeed have significant correlation and they stated that they simply “added” the two variables together. This seemed rather a simplified method to avoid too much variable correlation in an already complex model, but the researchers claimed it was a valid method to avoid multicollinearity and included a citation to affirm their “addition” action. The authors did mention that all variance inflation factors (VIF) values were well under the benchmark of 10.0, indicating that multicollinearity was not a serious factor in the regression model.
Also, it was not clear that all variables were continuous. For example, the education variable was encoded using a Likert scale between 1 (elementary) and 7 (PhD). These values can be seen as continuous when averaged across all cases, but are not a continuous variable, per se. There was also concern on the part of this reviewer that there were too many variables in the regression analysis, making it too complicated. The authors had to use seven regression models to determine the effects of their predictors. This seemed like a cumbersome and overly complex regression model, particularly when control and dummy variables were added into the regression equation, such as “Firm Size,” “R&D Intensity,” “Debt Ratio,” “Product Diversity,” and “Regulative Distance.” It wasn’t clear exactly how these variables were used to control the cross-sectional design in this complex regression model.
Descriptive statistics (mean, standard deviation), correlations, and significance (p) values were provided for all variables. The main results were reported as regression coefficients and significance values. Two very large tables of results with very small type were provided. All of the reporting of hypothesis confirmations were done by referencing only the first table of results. It was not clear why the second table was never referenced. If it was not referenced, why was it included in the study? The final size of the study sample was never compared to the population sample. That is, n=186 was never related to the proportion of the total population. Interaction plots for the four significant CEO attributes (age, experience, education, duality) were listed, but the X-axis of each plot was not described. This made the importance of the plots less impactful.
Each hypothesis was analyzed based on the regression results and confirmed or denied. Four of the predictors were significant. However, CEO tenure was not statistically significant, yet the authors chose to discuss its importance. If a variable, such as CEO tenure is not statistically significant, the researchers should not opine about its potential meaning for the study or how it may have affected a particular hypothesis.
Limitations of the study were discussed. Generizability for other non-Taiwanese businesses was discussed as a limitation. Asian cultural aspects inherent in the study, particularly around the attitudes of age and knowledge, were discussed as well. The authors also recognized that there may be other factors that affect successful internationalization, such as organizational structure, that were not addressed in the study. “Three-stage” theory was mentioned as a potential limitation, but what this meant was not clear as it was not clearly defined. It was suggested that further research could be done using a longitudinal study involving SME CEO attributes. Practical implications of the study revolved around CEO selection and grooming based on the significant predictors listed above.
This CEO attributes study appears to support several hypotheses about the significance of CEO characteristics that positively and negatively impact the successful internationalization of SMEs in Taiwan. The three author’s discussions about the significance of CEO age, CEO international experience, CEO level of education, and CEO duality (that is, a company has a better chance of being successful if the CEO is not also the chairman of the board) all could be justifiably and reliably cited in other studies about CEO characteristics and their affect upon firm performance.
Hsu, W., Chen, H., & Cheng, C. (2013). Internationalization and firm performance of SMEs: The moderating effects of CEO attributes. Journal of World Business, 48, 1–12.
Field, A. (2013). Discovering statistics using SPSS (4th ed.). London: Sage.