CEO Education and Firm Performance: Evidence from Corporate Universities

Abstract

Corporate universities are increasingly being established in response to corporate dissatisfaction with the knowledge provided by traditional universities. While some argue that they may pose the greatest threat to traditional universities, others propose that corporate universities might provide better undergraduate-level education than traditional universities. To this end, this research tries to answer the following research question: can integrating corporation education into traditional universities provide better educational outcomes? Using a sample of 40,563 firm-year observations from 2001 to 2019 and exploiting the unique setting of Japanese corporate universities, the authors find that the CEOs who are graduates of corporate universities significantly underperform compared to their counterparts graduating from traditional universities. The results are robust under various alternative estimations, such as general method of moments (GMM) IV regressions and propensity score matching. Overall, the results indicate that integrating corporation needs into traditional higher education will lead to lower individual productivity.

1. Introduction

It has been well established that CEO characteristics matter for a wide range of corporate decisions (Bernile et al. 2017; Bertrand and Schoar 2003; Mackey 2008). Among the various personal attributes of a CEO, educational attainment has been confirmed to be an important factor of management style in extant studies (Hambrick and Mason 1984). However, the literature provides mixed evidence regarding the impact of CEO education on firm performance (Miller et al. 2015; King et al. 2016). One potential reason is that the relationship between CEO education and firm performance may differ by the type of educational institution. Our paper contributes to this literature stream by investigating a particular type of educational institution, the corporate university.

Corporate universities are increasingly being established in response to corporate dissatisfaction with the knowledge provided by traditional universities (Cappiello and Pedrini 2017).1 As of 2015, there were around 4000 corporate universities in the United States and 3700 in China (Lui Abel and Li 2012). More recently, some corporate universities have become accredited, offer degrees, and opt to outsource corporate education to traditional universities (Andresen and Lichtenberger 2007; Thompson 2000).2 Since traditional universities are also interested in partnering with corporation because of the potential economic gains, Narasimharao (2010) asserts there are advantages to integrating corporate education into traditional universities. This approach is assumed to narrow the boundaries between corporate and traditional universities and, thus, make traditional universities resemble businesses. Given the prosperity of corporate universities, some authors argue that they may pose the greatest threat to traditional universities (Ball and Butler 2000; Baporikar 2014; Nixon and Helms 2002; Thompson 2000). Others propose that corporate universities might provide better undergraduate-level education than traditional universities (Blass 2005). To this end, this research tries to answer the following research question: can integrating corporation education into traditional universities provide better educational outcomes?

To explore the above research question, the performance of Japanese firms led by CEOs who graduated from traditional universities is compared with those led by CEO who graduated from corporate universities. The Japanese setting is useful to examine our idea because the corporate universities in Japan are more comparable to traditional universities than their counterparts in other countries. For example, in the United States and European countries, the main target of corporate universities are the employees, most of whom have already completed their undergraduate education in traditional universities before they join companies, which makes it difficult to disentangle the effects of corporate universities from those of traditional universities. However, in Japan, both traditional and corporate universities provide undergraduate education.3

Our empirical results show that the firms led by CEOs whose last educational degree is a graduate from a corporate university significantly underperform their peers who graduated from traditional universities, which suggests that integrating corporation needs into traditional higher education will lead to lower individual productivity. One potential reason is that market-oriented education may prohibit creative and critical thinking abilities because subjects that do not fit well with the for-profit purpose (e.g., arts, humanities, critical social science) may be neglected (Elton 2000; Rutherford 2005; Anwaruddin 2013). The primary results hold for a battery of supplemental tests, as to mitigate endogeneity problems and alternative scenarios.

The contributions of this article to the literature are as follows. First, it investigates the heterogenous educational effects of CEOs. While the literature provides evidence on the effects of CEO education on firm performance, no studies consider the degree to which a university integrates corporation needs (King et al. 2016; Chevalier and Ellison 1999; Beber and Fabbri 2012; Bertrand and Schoar 2003). Some studies find that the CEOs who hold an MBA degree are more likely to pursue risky strategies and outperform their peers. In a similar vein, CEOs with stronger educational credentials (Ivy League schools) are more welcomed by the market (Elsaid et al. 2015; Miller et al. 2015; Gottesman and Morey 2010). This research offers a new insight: the type of university matters. Second, this paper extends the debate over what should universities be (Rutherford 2005; Lynch 2006; Thompson 2000; Blass 2001) by comparing CEOs who graduated from traditional universities with those who graduated from corporate universities. While some papers compare traditional universities with corporate universities (Nixon and Helms 2002; Blass 2001), there is little empirical evidence. This research takes advantage of the Japanese data and highlights that the CEOs who graduated from corporate universities show inferior firm performance. Finally, our research also contributes to the strand of literature examining the determinants of firm performance (Khuong et al. 2020).

The rest of this paper is organized as follows. Section 2 reviews the related literature. Section 3 presents the background of Japanese corporate universities. Section 4 describes our sample and methodology. Section 5 show our main empirical results and Section 6 offers a summary and the conclusions of this study.

2. Literature Review

Nearly 170 years ago, in his book The Idea of a University, Newman (2008) stated that a university is a place of teaching universal knowledge. Over the centuries, universities have served as guardians and creators of knowledge, as well as watchdogs of free exchange of ideas. However, there is an increasing tendence, both in developed and emerging countries, to adopt market-based reforms by privatizing universities (Gebreiter 2021; Parker 2011; Kallio et al. 2016; Parker 2012), which translates into for-profit higher education and a market-oriented curriculum (Rutherford 2005).

At the end of this spectrum of market-oriented higher education is the emergence of corporate universities. Different from training departments that provide orientation and on-the-job related skill training, a corporate university can be defined as an educational institution that can be used as a strategic tool designed to pursue long-term goals, as well as provide sustainable competitive advantage to parent companies (Andresen and Lichtenberger 2007).4 As opposed to a traditional university, where general-purpose education is provided, corporate universities are thought to have a hybrid nature, in that they provide both specific and general training.5 As an increasing number of corporate universities have become accredited and offer degrees, some authors argue that it is advantageous to integrate corporate education into traditional universities and that corporate universities may pose the greatest threat to traditional universities (Ball and Butler 2000; Baporikar 2014; Nixon and Helms 2002; Thompson 2000).

With respect to the educational outcomes of corporate universities, two competing hypotheses can be proposed. On the one hand, corporate universities can provide better educational outcomes than traditional universities because integrating corporate needs into higher education can close the gap between workforce demand and the educational supply of traditional universities. On the other hand, Elton (2000) contends that this increasing commercialization will lead to the unintended consequences, in that students go to university solely to gain career skills and credentials. Additionally, market-oriented education may prohibit creative and critical thinking abilities because subjects that do not fit well with the for-profit purpose (e.g., arts, humanities, critical social science) may be neglected (Rutherford 2005; Anwaruddin 2013). Consequently, the second idea asserts that a corporate university is associated with lower educational outcome. From the above argument, this paper proposes the following two competing hypotheses:

H1a. 

CEOs whose last educational degree is a graduate from a corporate university significantly outperform their peers who graduated from traditional universities.

H1b. 

CEOs whose last educational degree is a graduate from a corporate university significantly underperform their peers who graduated from traditional universities.

3. Background of Japanese Corporate Universities

As one of the best-known corporate universities in Japan, the Toyota Technological Institute (TTI) was founded in 1981 by the Toyota Motor Corporation. According to its history and mission, TTI was opened in 1981 and began offering a unique educational program based on its founding philosophy and aimed to maintain technological competitiveness.6 Initially, TTI only accepted students with industrial work experience. However, over 80% of the students admitted since 1993 have been new high-school graduates. More recently, TTI further opened its PhD program and established a sister institution, TTI Chicago, in 2003. Toyota Motor Corporation has been involved in TTI by appointing the chairman of the board of trustees, who is usually either the executive or ceremonial head of the university.7

Similarly, Seikei University was established by Koyata Iwasaki, the fourth head of the Mitsubishi zaibatsu.8 It used to be managed by Mitsubishi zaibatsu before World War II and, even after the dissolution of Mitsubishi zaibatsu, it has remained close to the Mitsubishi Group. The current chairman of the board of trustees is Ken Kobayashi, who is also currently serves as the chairman of Mitsubishi Corporation.9

Given the fact that most corporate universities were founded by corporations and the founding corporations can exert power over them through the appointment of the chairman, we assume that corporate universities are implicitly or explicitly used as a strategic tool to pursue the long-term goals, as well as the sustainable competitive advantages of founding companies. Indeed, Shigenobu Nagamori, the founder (CEO) of Nidec corporation, indicated that the Kyoto University of Advanced Science was created in 1969, aiming to develop human resources.10

As previously argued, the unique setting of Japanese universities enables us to compare the educational effects of corporate and traditional universities. Any Japanese high school student can either choose a corporate university or a traditional one, depending on his/her score of the National Centre Test for University Admissions, geographical considerations, as well as personal preferences. By contrast, the main target of corporate universities in the United States and European countries is the employee.

4. Sample Selection, Data, and Methodology

This research selected corporate universities in Japan from two websites,11 namely the university name, the founding company, as well as its history. In total, we identified 14 corporate universities. Then, the authors double checked the list of corporate universities by referring to the homepages of these universities, as well as Japan Student Services Organization (JASSO). Information (e.g., last educational background, major in university, birthplace) for CEOs was taken from the Toyokeizai Director database. Then, the authors merged the list of corporate universities with the Toyokeizai Director database using the names of corporate universities. The financial information is available from Nikkei NEEDS Financial Quest.

Table 1. Sample distribution.

Year	Special CEOs	Others
2001	25	1759
2002	27	1779
2003	29	1807
2004	29	1829
2005	27	1905
2006	29	1968
2007	29	2021
2008	29	2050
2009	27	2057
2010	21	2073
2011	20	2090
2012	21	2113
2013	20	2158
2014	22	2189
2015	22	2257
2016	22	2318
2017	26	2369
2018	24	2418
2019	24	2457
Total	473	40,090

reports the presence of Japanese corporate universities. Column 1 indicates that the number of companies led by CEO who are graduates from universities founded by corporations is approximately 30 every year and accounts for 1.2% of all firm-year observations.

Table A2. The ranking for CEO producing universities in Japan.

Name	Freq.	Cum.
Keio University	3652	11.1%
Waseda University	2199	6.7%
University of Tokyo	2184	6.6%
Nihon University	1050	3.2%
Kyoto University	974	3.0%
Chuo University	865	2.6%
Meiji University	732	2.2%
Doshisha University	700	2.1%
Hitotsubashi University	530	1.6%
Osaka University	504	1.5%

(in the Appendix A) further shows the ranking for CEO producing universities in Japan.12 An important feature of the CEO labour market in Japan is that the top 10 universities (e.g., Keio University, Waseda University, University of Tokyo), all of which are traditional universities, account for around 40% of all observations in our dataset. Meanwhile, the authors found that three corporate universities stand out: 293 firm-year observations are from Seikei University, 59 from Asia University, and 50 from Musashi University.

Furthermore, as shown in

Table A3. Major of CEOs who graduated from universities founded by corporations.

Major	Freq.
Economics	215
Law	85
Engineering	62
Political Science and Economics	48
Commerce	28
Electrical Engineering	10
Pharmacy	9
Chemistry	6
Business	5
Humanities	5

(in the Appendix A), approximately 70% of the CEOs who graduated from universities founded by corporations majored in economics, law, or engineering. With respect to whether the appointment of a CEO from a corporate university is related to geographical characteristics.

Table A4. Prefecture Distribution of headquarters.

Prefecture	CEOs from Universities Founded by Corporations	Others
Tokyo	208	15,565
Osaka	40	4253
Aichi	17	2078
Chiba	14	456
Hiroshima	14	506
Kanagawa	13	1611
Shizuoka	13	490
Okayama	10	187
Ibaraki	7	87
Saitama	6	711
Niigata	6	369
Gunma	6	202
Kochi	5	42
Kagawa	3	159
Kyoto	1	651

(in the Appendix A) presents the prefecture distribution of the headquarters of Japanese firms, and a similar pattern was observed in that most of the Japanese companies are headquartered in Tokyo and Osaka regardless of their CEOs’ educational backgrounds.13

To test the hypotheses, the authors first employed a linear regression of firm performance to directly compare the educational impacts between corporate and traditional universities as a benchmark. As our key independent variable, Special CEOs is a dummy variable equal to one for firms with CEOs coming from universities founded by corporations, and zero otherwise. ROA (operating income scaled by total assets), SGR (sales growth ratio), and Tobin’s Q (total capitalization and book value of liabilities scaled by book value of assets) are used to measure firm performance.

As control variables, the authors include CEO age, CEO ties (natural logarithm of the total number of CEOs from the same university in a given year), foreign institutional investor (foreign institutional investors’ total ownership), proportion of outside director, Ln (total assets), and leverage (total debts scaled by total assets). One-year lagged data are used for these control variables. All variables are winsorized at the top and bottom 1% levels (except for the dummy variables).

To alleviate the endogeneity issue, we employ the following methodological approaches: an instrumental variables (IV) estimation strategy and propensity score matching (PSM). A battery of robustness checks has been further conducted to rule out some potential alternative scenarios.

Table 2. Summary statistics.

	Special CEOs	Others	p-Value
ROA	2.52% [2.56%]	3.79% [3.75%]	0.000 ***
	N = 473	N = 39,511	[0.000 ***]
SGR	2.19% [1.15%]	4.19% [2.77%]	0.007 ***
	N = 464	N = 38,753	[0.000 ***]
Tobin’s Q	1.14 [0.94]	1.17 [0.98]	0.518
	N = 473	N = 39,617	[0.001 ***]
CEO Age	59 [60]	60 [61]	0.66
	N = 473	N = 39,617	[0.080 *]
CEO Ties	2.24 [2.83]	3.59 [3.71]	0.000 ***
	N = 473	N = 39,617	[0.000 ***]
Foreign institutional investor	5.40% [1.60%]	9.36% [4.74%]	0.000 ***
	N = 469	N = 37,739	[0.000 ***]
Proportion of outside director	10.99% [0.00%]	13.82% [11.11%]	0.000 ***
	N = 473	N = 39,617	[0.000 ***]
Ln(Total assets)	10.36 [10.33]	10.72 [10.52]	0.000 ***
	N = 473	N = 39,617	[0.004 ***]
Leverage	22.63% [21.00%]	22.52% [19.62%]	0.890
	N = 473	N = 39,617	0.514

Notes: this table presents descriptive statistics. See Appendix A Table A1 for the variable definitions. p-values are for the mean [median] difference test. Significance at the 10%, and 1% levels is indicated by *, and ***, respectively.

presents the summary statistics of the variables separately for the subsample (Special CEOs versus Others). This research finds that Special CEOs significantly underperform their peers who graduated from traditional universities in terms of accounting-based, as well as market-base performance measures. ROA is 2.52% for Special CEOs, which is lower than for other CEOs (3.79%). Similarly, firms led by special CEOs tend to have a lower Tobin’s Q (0.94 versus 0.98, p < 0.01). These results are consistent with the second hypothesis that the educational effect of corporate universities is inferior to that of traditional universities. As argued by Elton (2000), increasing commercialization will lead to the unintended consequences, in that students go to university solely to gain career skills and credentials. Additionally, market-oriented education may prohibit creative and critical thinking abilities (Rutherford 2005; Anwaruddin 2013).

This research also considered various control variables that potentially affect firm performances. Table 2 shows that Special CEOs have significantly lower foreign institutional ownership and a lower proportion of outside directors than other CEOs. Additionally, CEO ties has a significant smaller value for Special CEOs than for other CEOs. Meanwhile, the authors do not find a significant difference in CEO age and Leverage.

5. Empirical Results

5.1. Baseline: Ordinary Least Squares Estimation

This section presents the results of ordinary least squares (OLS) regressions, where the dependent variables are ROA, SGR, and Tobin’s Q. Industry and year-fixed effects are included in all estimations to account for industry characteristics and macro-economic conditions. Models (1) and (2) in

Table 3. OLS regressions.

Dependent Variables	ROA	SGR	Tobin’s Q
	(1)	(3)	(4)
Special CEOs	−0.010 ***	−0.018 **	0.046
	(−3.38)	(−2.56)	(1.37)
Control variables:
CEO Age	−0.000	−0.002 ***	−0.011 ***
	(−0.55)	(−13.35)	(−20.48)
CEO Ties	−0.000	−0.000	0.011 ***
	(−0.97)	(−0.27)	(6.39)
Foreign institutional investor	0.085 ***	0.116 ***	1.800 ***
	(17.87)	(10.10)	(32.46)
Proportion of outside director	−0.032 ***	0.003	0.308 ***
	(−9.78)	(0.37)	(8.85)
Ln (Total assets)	−0.000	−0.003 ***	−0.106 ***
	(−0.61)	(−4.14)	(−29.06)
Leverage	−0.113 ***	−0.033 ***	0.016
	(−44.37)	(−5.36)	(0.72)
Constant	0.063 ***	0.195 ***	2.816 ***
	(9.08)	(13.02)	(42.67)
N	38,127	37,384	38,208
R2	0.152	0.106	0.239
Industry dummy	Y	Y	Y
Year dummy	Y	Y	Y

Notes: see Appendix A Table A1 for the variable definitions. Heteroskedasticity-consistent standard errors are used to compute Z-statistics. Significance at the 5% and 1% levels is indicated by **, and ***, respectively.

provide negative and significant coefficients on Special CEOs, implying that, all else being equal, Special CEOs underperform their peers in terms of ROA (SGR) by 1% (1.8%). However, Model (3) shows an insignificant sign for Special CEOs and, thus, this research does not find a difference in Tobin’s Q. With respect to the control variables, Table 3 implies that higher foreign institutional ownership and the proportion of outside directors are associated with better firm performance. CEO Ties has a significant coefficient on Tobin’s Q, highlighting the value of connections (social capital).

5.2. Robustness Checks

One can criticize our results, showing that Special CEOs’ firms differ from others in their various characteristics. If endogeneity is a concern, our main results may reflect correlation rather than causation. To mitigate this endogeneity concern, the authors use two additional approaches to establish the causal inference between Special CEOs and firm performance.

5.2.1. Instrumental Variable Regression

Firstly, as in Nam and Uchida (2019), the authors implement general method of moments (GMM) IV regressions with industry- and year-fixed effects of firm performance to deal with endogeneity concerns Since the Pagan–Hall test is always significant, GMM IV is used instead of simple two-stage least squares regression. To identify the impact of Special CEOs on firm performance, we need a variable that reflects the probability of choosing a corporate university that does not directly affect firm performance. While it is extremely difficult to find appropriate IVs, the authors employ the following two instrumental variables: (1) Tokyo, a dummy variable equalling one for firms with a CEO who was born in Tokyo, and zero otherwise, and (2) Female, a dummy variable equalling one for firms with a female CEO, and zero otherwise. Our rationale is as follows. First, unreported results show that around 60% of the corporate universities are located in Tokyo. This implies that, ceteris paribus, a high school graduate born in Tokyo is more likely to choose a corporate university due to its geographical proximity. Meanwhile, the birthplace of CEOs is more likely to be exogenously determined and, thus, is less likely to directly affect firm performance. Second, after verifying our director database, the authors found that none of the directors who are graduates of corporate universities are female. This result is not surprising given the high levels of gender inequality in Japan. In the Global Gender Gap Report 2020, Japan ranked 121 out of 153 countries according to gender equality. Traditionally, a women’s role has been considered to prepare them to be better wives and mothers, which shaped the persistent view that educating daughters is less important than educating sons (Fujimura-Fanselow 1985). Therefore, the authors expect that a female high-school student is less likely to go to Tokyo and choose a corporate university. Unreported regression results also suggest that none of the performance measures are related to the female dummy.

The first-stage results in

Table 4. Instrumental variable regression (GMM).

	1st Stage	2nd Stage
		ROA	SGR	Tobin’s Q
		(1)	(2)	(3)
Special CEOs		−0.123 **	−0.488 ***	−0.629
		(−2.09)	(−2.98)	(−0.99)
CEO AGE	0.000	−0.000	−0.002 ***	−0.011 ***
	(0.35)	(−0.52)	(−12.99)	(−20.47)
CEO Ties	−0.004 ***	−0.001 **	−0.002 **	0.009 ***
	(−16.91)	(−2.12)	(−2.51)	(3.03)
Foreign institutional investor	−0.027 ***	0.082 ***	0.103 ***	1.781 ***
	(−3.79)	(16.33)	(8.05)	(30.78)
Proportion of outside director	−0.001	−0.032 ***	0.002	0.307 ***
	(−0.23)	(−9.73)	(0.27)	(8.81)
Ln(Total assets)	−0.000	−0.000	−0.003 ***	−0.106 ***
	(−0.86)	(−0.69)	(−4.16)	(−28.95)
Leverage	−0.008 **	−0.114 ***	−0.036 ***	0.012
	(−1.98)	(−43.56)	(−5.56)	(0.53)
Constant	0.015 **	0.065 ***	0.203 ***	2.828 ***
	(2.29)	(9.18)	(12.95)	(42.08)
Tokyo	0.014 ***
	(7.02)
Female	−0.014 ***
	(−11.48)
Industry dummy	Y	Y	Y	Y
Year dummy	Y	Y	Y	Y
Partial R2	0.002
F-test of excluded instruments	88.75 ***
Hansen J test of overidentification (p-value)		0.001	3.430 *	0.252
		(0.978)	(0.064)	(0.616)
N	38,127	38,127	37,384	38,208
Centered R2	0.012	0.118	0.003	0.230

Notes: see Appendix A Table A1 for the variable definitions. Heteroskedasticity-consistent standard errors are used to compute Z-statistics. Significance at the 10%, 5%, and 1% levels is indicated by *, **, and ***, respectively.

confirm these assumptions. Tokyo (Female) carries a significant positive (negative) coefficient of 0.014 (−0.014), suggesting the validity of the IV. Moreover, the partial F-statistic is 88.75 and the Stock–Yogo weak ID test critical value at 10% IV size is 19.93, thus rejecting the null hypothesis of a weak instrumental variable. Additionally, the Hansen J test statistic is not statistically significant at the 5% level, which indicates that the IVs are valid in all three estimations.

The second-stage regression uses the fitted value of Special CEOs obtained in the first-stage regression, and the results are qualitatively similar to those in our baseline regressions. The coefficients in Models (1) and (2) in Table 4 are −0.123 and −0.488 for ROA and SGR, respectively, thus highlighting the significantly negative impact of corporate universities on CEO productivity. While Model (3) also carries a negative sign for Special CEOs in the regressions of Tobin’s Q, it is not statistically significant.

5.2.2. Propensity Score Matching

Comparing the performance of firms led by special CEOs to that of companies led by other CEOs with similar characteristics is another effective way to mitigate the endogeneity concern. Following Fan and Uchida (2019), the authors first run a logit regression, where the dependent variable is Special CEOs, and all variables (including industry and year dummies) in Table 3 are included to estimate the probability of appointing a special CEO. For every Special CEOs, we selected the three (five) firms with the nearest propensity of being a Special CEOs as the matched subsample. The results are reported in Panel A of

Table 5. Propensity score matching estimates.

Panel A: Full Sample	1 vs. 3	1 vs. 5
ROA
Special CEOs	2.55%	2.55%
Others	3.47%	3.65%
Difference	−0.92% **	−1.09% ***
	(−2.49)	(−3.12)
SGR
Special CEOs	2.14%	2.14%
Others	4.65%	4.65%
Difference	−2.50% ***	−2.51% ***
	(−2.99)	(−3.13)
Tobin’s Q
Special CEOs	1.15	1.15
Others	1.09	1.09
Difference	0.05	0.06
	(1.27)	(1.41)
Panel B: Post-turnover Period	1 vs. 3	1 vs. 5
Ave ROA (1,3)
Special CEOs	2.33%	2.33%
Others	4.24%	3.88%
Difference	−1.92% *	−1.35%
	(−1.75)	(−1.22)
Ave SGR (1,3)
Special CEOs	0.98%	0.98%
Others	5.56%	4.83%
Difference	−4.57% **	−3.85% **
	(−2.42)	(−2.17)
Ave Tobin’s Q (1,3)
Special CEOs	1.02	1.02
Others	1.03	1.06
Difference	−0.01	−0.04
	(−1.03)	(−0.72)

Notes: this table reports the mean values of the firm performance measures of Special CEOs and the matching non-Special CEOs. See Appendix A Table A1 for the variable definitions. T-statistics are for the mean difference test. Significance at the 10%, 5%, and 1% levels is indicated by *, **, and ***, respectively.

. Given the difference between Special CEOs and its matched firms narrows for this matched subsample, this research still finds a similar pattern in that Special CEOs significantly underperform their counterparts. ROA is 2.55% for Special CEOs, whereas it is 3.47% for their matched peers (difference: −0.92%, p < 0.05).

While the main results remain unchanged, the authors further conducted an event study. Specifically, this research first identified CEOs turnover year (event = 0). To be included, the authors require that the incoming CEOs need to stay for at least three years and that all information should be available for one year prior to the turnover (event = −1). In total, the authors successfully identified 21 CEO turnovers for Special CEOs and 1614 turnovers for CEOs graduating from traditional universities. Then, the authors selected three (five) firms with the nearest propensity of being a Special CEO as the matched subsample based on information prior to the CEO turnover (event = −1). The results for the post-turnover period are shown in Panel B of Table 5. For instance, the three-year average ROA is 2.33% for newly appointed Special CEOs and 4.24% for the matched peers.

5.3. Additional Analyses

One can argue that our main results may be driven by the fact that none of the Japanese corporate universities rank in the top 10 of the CEO producing universities in Japan. Given that a top 10 university in Japan is equivalent to an Ivy League school in the United States, it is possible that the CEOs with stronger educational credentials are more likely to demonstrate superior long-term performance (Elsaid et al. 2015; Miller et al. 2015; Gottesman and Morey 2010). To address this concern, the authors first reran all our analyses while excluding the top 10 (top 20) universities, and the results for the GMM IV are reported in

Table 6. Instrumental variable regression (GMM) (sub-samples).

Panel A: Excluding top 10 universities
Dependent Variables	ROA	SGR	Tobin’s Q
	(1)	(2)	(3)
Special CEOs	−0.057 *	−0.198 **	−1.208 ***
	(−1.66)	(−2.22)	(−3.17)
Control variables:	Yes	Yes	Yes
Industry dummy	Yes	Yes	Yes
Year dummy	Yes	Yes	Yes
N	25,266	24,674	25,307
R2	0.139	0.081	0.213
Panel B: Excluding top 20 universities
	ROA	ROE	SGR
	(1)	(2)	(3)
Special CEOs	−0.126 ***	−0.281 ***	−1.261 ***
	(−3.34)	(−2.91)	(−3.16)
Control variables	Y	Y	Y
Industry dummy	Y	Y	Y
Year dummy	Y	Y	Y
N	24,785	24,185	24,826
R2	0.096	0.057	0.197

Notes: see Appendix A Table A1 for the variable definitions. Heteroskedasticity-consistent standard errors are used to compute Z-statistics. Significance at the 10%, 5%, and 1% levels is indicated by *, **, and ***, respectively.

. Surprisingly, the results are even stronger for this subsample, and it finds that Special CEOs significantly underperform for all three measures. The estimated coefficient in Model (3) indicates that Tobin’s Q is 1.208 lower for Special CEOs than for other CEOs.

To further differentiate the educational effects from CEO ability, the authors reran all our analyses by adding more control variables related to CEOs’ inherent abilities. Following Custódio and Metzger (2014), the following four variables are added to capture CEOs’ innate talents, as well as experience: (1) Fast Track: the age of the first CEO job; (2) Dual Directorship: a dummy variable equal to one for a CEO who also takes the director position of other firm, and zero otherwise; (3) Experience: a dummy variable equal to one for CEO who worked for other company before joining the current company, and zero otherwise; and (4) Sd Score: indicates how far from the statistical mean a typical student admitted to a university scores on university entrance examination. The authors expect that the younger an individual became CEO for the first time, the more talented he/she is. Similarly, high quality board members are more likely to serve on multiple boards (Brickley et al. 1999; Bugeja et al. 2009; Coles and Hoi 2003; Gilson 1989, 1990). Experience is also expected to positively affect firm performance (Crossland et al. 2014). One might argue that corporate universities are different from traditional top-ranking universities in terms of education credentials. To mitigate this concern, this research further adds Sd Score and assume a higher Sd Score is associated with higher education credentials, as well as intellectual ability.14 However, it is also worth noting that our sample is substantially reduced due to missing data.

The results are presented in

Table 7. Controlling for more CEO characteristics.

Panel A: OLS
Dependent variables:	ROA	SGR	Tobin’s Q
	(1)	(3)	(4)
Special CEOs	−0.009 ***	−0.017 **	0.047
	(−2.91)	(−2.38)	(1.41)
Fast Track	−0.001 ***	−0.001 ***	−0.005 ***
	(−10.96)	(−9.83)	(−9.51)
Dual Directorship	−0.002	0.013 *	0.175 ***
	(−0.61)	(1.78)	(4.70)
Experience	0.001	0.018 ***	0.084 ***
	(1.43)	(7.99)	(9.07)
Sd Score	−0.010	−0.024	0.386 ***
	(−1.47)	(−1.45)	(5.43)
Other control variables:	Y	Y	Y
N	26,238	25,896	26,305
R2	0.171	0.121	0.235
Panel B: IV GMM
Dependent variables:	ROA	SGR	Tobin’s Q
	(1)	(3)	(4)
Special CEOs	−0.086 *	−0.307 **	−0.035
	(−1.95)	(−2.57)	(−0.07)
Other control variables:	Y	Y	Y
N	26,238	25,896	26,305
R2	0.146	0.060	0.236
Panel C: PSM
ROA
	1 vs. 3		1 vs. 5
Special CEOs	2.64%		2.64%
Others	3.79%		3.72%
Difference	−1.16% ***		−1.08% ***
	(−3.17)		(−3.08)
SGR
Special CEOs	2.10%		2.10%
Others	4.10%		3.88%
Difference	−2.01% **		−1.78% **
	(−2.37)		(−2.22)
Tobin’s Q
Special CEOs	1.15		1.15
Others	1.07		1.08
Difference	0.07 *		0.07
	−1.75		(1.64)
Number of treated	465
Number of untreated	24,851

Notes: see Appendix A Table A1 for the variable definitions. Significance at the 10%, 5%, and 1% levels is indicated by *, **, and ***, respectively.

. With respect to the control variables, the estimated results are consistent with our conjecture and all four variables significantly affect firm performance, indicating that the additional variables successfully capture CEOs’ inherent abilities and experience. More importantly, qualitatively similar results are obtained even after controlling for more CEO-specific characteristics.

6. Conclusions

It is well established that the education background of a CEO significantly affects firm performance (Hambrick and Mason 1984; King et al. 2016; Chevalier and Ellison 1999; Beber and Fabbri 2012) because education shapes the cognitive abilities of the CEO, which in turn influences firm investment and general decision-making. Furthermore, the effects of education are likely to vary according to the type and quality of education. Beber and Fabbri (2012) found that CEOs with MBA backgrounds tend to aggressively engage in forex markets. Chevalier and Ellison (1999) show that the CEOs graduating from university (undergraduate) with high SAT scores substantially outperforms their peers. Miller et al. (2015) further argue that firms run by CEO having graduated from Ivy League schools show superior performance, especially for undergraduates program. This research adds to this literature stream by investigating whether the effects of CEO education on firm performance differ by university type. Comparing corporate universities with traditional universities also extends the debate over what universities should be (Blass 2001; Lynch 2006; Rutherford 2005; Thompson 2000).

This research developed two competing predictions regarding the educational outcomes of corporate universities. On the one hand, corporate universities can provide better educational outcomes than traditional universities because integrating corporate needs into higher education can close the gap between workforce demand and the educational supply of traditional universities (Ball and Butler 2000; Baporikar 2014). On the other hand, market-oriented education may prohibit creative and critical thinking abilities (Rutherford 2005; Anwaruddin 2013).

The empirical findings can be summarized as follows. First, the OLS regression reveals that CEOs that have graduated from corporate universities significantly underperform their counterparts graduating from traditional universities. Second, to mitigate the endogeneity concern, the authors use two additional approaches to establish the causal inference between Special CEOs and firm performance: GMM IV regressions with industry- and year-fixed effects, and propensity score matching. Finally, to further differentiate the educational effects from CEO ability, the authors reran all the analyses by adding more control variables related to CEOs’ inherent abilities. Overall, qualitatively similar results are obtained even after controlling for more CEO-specific characteristics.

The results indicate that the integration of corporation education into traditional universities is associated with relatively lower productivity, at least at the CEO level. While an increasing number of corporate universities have become accredited and offer degrees, (Ball and Butler 2000; Baporikar 2014; Nixon and Helms 2002; Thompson 2000), the findings indicate that integrating corporation education into traditional universities cannot provide better educational outcomes.

However, this research is limited by a lack of adequate controls over characteristics of universities (e.g., courses, modules, mission, and vision). Future research should consider whether our findings can be generalized to other institutional settings. Finally, there are many other factors that might affect our main results including main bank relationship, family businesses and political connections.