The Data

The proof can be found in Appendix 1. Since team preferences (σ and μ) of both sexes are the same and gender preferences (v) are also the same, gender would not be considered as an important element here, which means gender is irrelevant in this proposition.

Proposition 2: Gender Neutral

.

f and v v then implies that and

If σ σ σ σ β β α α β β

different preferences (σ_f ≠σ_m ) of team formation of the two sexes are assumed in Proposition 2. It allows gender neutrality to be sustained even when team preferences of the two genders are different. For example, we might observe that female chairpersons have a higher propensity to cooperate with female CEOs than males (i.e. β_m^FC <β_f^FC), and the gender neutrality hypothesis (v_f =v_m =1) can still hold if men are more likely

to work alone than women (i.e.

σ

_m >

σ

_f ).

Based on these two propositions, we will first test whether there is difference of the partnership between chairman and chairwoman. And, if there is a difference, the single team type can then be tested in order to find support for the gender neutral hypothesis.

Ⅲ. The Data

Data used in this paper is from “Top5000: The Largest Corporations in Taiwan”, which is published by China Credit Information Service, Ltd., in June every year. The 2006 edition is used. China Credit Information Service, Ltd., sent out 16,780

more than 60 million NT dollars (about 2 million US dollars) in case of manufacturing companies, or had assets of more than 30 million NT dollars (about 1 million US dollars), in case of services companies. Of the total, 5,183 questionnaires were returned.

Besides the information in returned questionnaires, the source publication also links companies to their financial data from Taiwan Stock Exchange Corporation. There are 4,857 companies included in the composite ranking. Several companies were found to have missing values, or had unrecognized information. So finally the total number of companies we use is 4,485. In the analysis data set, the main variables are the composite rankings of companies, names of chairpersons and CEOs, established years, zip code, and industry code. Genders of chairpersons and CEOs are identified by their Chinese first names.

Chairpersons and CEOs of companies in the data are sorted by gender as shown in Table 1. Column 1 shows companies are sorted into even and single teams. A company with an even team is one which has different persons functioning as chairperson and CEO, while a company with a single team is one which has the same person holding both posts. Column 2 shows the number of female top executives corresponding to the team type, and Column 3 is the number of male top executives. Column (4) is the number of companies corresponding to the team types.

Row (A) presents the gender composition of chairpersons and CEOs in even teams.

There are 3,142 companies that have different persons as chairperson and CEO. Row (B) presents the gender composition of single teams in 1,343 companies covered in this data set. The sum of each column is shown in Row (C). It is found the total number of females observed is 460, and the total number of males is 7,167, in 4,485 companies covered by the data used for this paper.

We find that female top executives are relatively scarce in Taiwan. In Table 1, the percentage in the parenthesis is the share calculated by rows: females’ share in chairpersons in even teams is 7.45%, while the share of females in single teams is only

3.43%. It is found that in both even and single teams, males dominate. The proportion of female and male workers is perhaps fairly equal at the entry level of labor markets.

Then why at the top end, the ratio of females and males plunges to 1:16? Besides, there are fewer female CEOs than chairpersons. This makes one wonder whether there might be a gender preference among female chairpersons while hiring a CEO.

Next, Table 2 examines gender compositions of even teams only. There are 3,142 of them. In 1^st and 2^nd columns, four types of gender compositions of teams (chairperson + CEO) are shown:

1. A female chairperson and a female CEO, 2. A female chairperson and a male CEO, 3. A male chairperson and a female CEO, and 4. A male chairperson and a male CEO.

In 3^rd and 4^th columns, it shows the number and percentages of companies corresponding to different team types. In the 5^th column, conditional probabilities are calculated, i.e. P (Gender of CEOi | Gender of Chairmani). For example, the conditional probability that a given female chairperson chooses a female CEO is 19 divided by 234 (the total number of female chairpersons = 19 + 215), which equals 8.12%, i.e.

P(Female CEOi | Female Chairpersoni) = 8.12% 234

19 )

(

)

( = =

=

=

∩

=

f president P

f CEO f

president

P

.

The conditional probability of a male chairperson choosing a female CEO is 161 divided by 2,908 (the total number of male chairpersons = 161 + 2747), which equals 5.54%.

The 6^th column is used for comparison, which has the proportions of CEOs by gender, in Table 1. It can be seen that team types that have female CEOs (in 2^nd column) are to be compared with 5.73%, which is the proportion of female CEOs in Table 1.

Also, team types with male CEOs are compared with the proportion of male CEOs in Table 1, which is 94.27%. It can be inferred that if a chairperson chooses a CEO

randomly from a pool of CEOs, then he/she has a 5.73% chance of choosing a female CEO, and there is a 94.27% chance of choosing a male CEO. Through the comparison mechanism, Table 2 shows that female chairpersons have a relatively higher tendency to have female CEOs (8.12% > 5.73%), and a lower propensity to have male CEOs (91.8% < 94.27%). In contrast, male chairpersons have a relatively higher tendency to name a male as CEO (94.46% > 94.27%), and a lower tendency to have a female CEO (5.54% < 5.73%). The comparison suggests that gender preferences might exist in composition of top executive teams, but the disparity is not very distinct, especially in case of male chairpersons.

Using the available information in the data set, we also sort the companies by their industry code, firm size, established years, and geographic locations. After controlling for these firm characteristics, we find similar results as in tables 1 and 2: female chairpersons and female CEOs are the minority among top executives, and female chairpersons show a relatively higher tendency to have same sex CEOs, than male chairpersons do, in most of the classifications. Details of the statistics are available on request.

. Empirical Results

Ⅳ

In this section, an empirical model is introduced to test whether the gender irrelevance and neutral hypotheses are sustained. The structure of the empirical model is based on that of Boschini and Sjögren (2007). The probit method is applied.

Y_ij^FC*= 'X _ij^FC β +ε_ij^FC (10) Y_ij^S*= X'_ij^S β +ε_ij^S (11)

Where Y_ij^FC* and Y_ij^S * are unobserved variables. Equation (10) denotes a chairperson’s tendency to cooperate with a female CEO while Equation (11) denotes a chairperson’s tendency to form a single team (to be the CEO as well). The observed

outcome in Equation (10) is a binary variable: if Y_ij^FC*>0 (i.e. the chairperson of i

company in j industry cooperates with a female CEO), then Y_ij^FC =1, otherwise

FC

Yij =0. The observed outcome variable in Equation (11) is also a binary variable: if 0

*

Y_ij^S > (i.e. the chairperson and the CEO of i company in j industry is the same

person), then Y_ij^S =1, otherwise Y_ij^S=0.

Both equations share the same explanatory variables. The 1^st explanatory variable is the sex of the chairperson,

f . If the chairperson of company i is female, then

_i

f =1,

_i otherwise

f =0. The 2

_i ^nd explanatory variable is the share of female CEOs in j industry,

φ

ij. There are three different industry classifications used in this paper: SCP, MCP and ACP. The first industry classification is SCP (Simple index of female CEO proportion).

All companies are divided into 5 different industries, which are manufacture, service, banking and finance, public enterprises and private universities. We then compute the female CEO proportion in each of the five industries.

The second industry classification is MCP (Main index of CEO proportion). The main difference between MCP and SCP is that the industries are divided into 41 sub groups, and the representative industry code is chosen by the main product of a company. Representative industry codes are used to calculate the proportion of female CEOs.

The third industry classification is ACP (Average index of CEO proportion), and it also uses the same 41 industry codes as MCP. But, since each company may not be listed for only one industry code, the number of corresponding female CEOs is calculated on a weighted basis. For example, if a company reports 3 different industry codes, it will be counted in all the three industries.

The 3^rd explanatory variable is the interaction term of the sex of the chairperson

and the share of female CEOs in the company’s industry,

f

_i

φ

_ij. The 4^th explanatory variable is a dummy variable of regions, i.e. the location of a company,

POST . If i

_i company is located in north Taiwan, then POSTi=1, if a company is located in non-north Taiwan, then POST_i=0. The 5^th explanatory variable is a dummy variable of established years of a company,

EST . They are divided by intervals of 10 years into four groups.

_i The benchmark of the established years is a company which was established less than 10 years ago. The 6^th explanatory variable is the size of a company,

SIZE . The firm

_i size is based on the net sales of the company, which means the higher is a company’s sales revenue, the bigger the company is. Firm sizes are divided into five levels.

Based on the results in the model section, the first step is to test the gender neutrality, i.e. to check whether female and male chairpersons have different attitudes towards teaming up with female CEOs. The key coefficient in this step is

β

₃^FC of Equation (10). Second, the single team tendency is examined, which can provide further support for the gender neutrality hypothesis.

β

₁^S and

β

₃^S of Equation (11) are two key coefficients that need to be estimated.

C F

β

3 is the coefficient of the interaction term of the chairperson’s sex (

f ) and the

_i share of female CEOs (

φ

_ij). If

β

₃^FCis statistically significantly different from zero, then it can be inferred that female and male chairpersons do have different attitudes towards the gender of CEOs, when forming a team. In other words, if the coefficient is insignificant, then it suggests that gender irrelevance might be true.

S

β

1 is the coefficient of the chairperson’s sex (

f ) in Equation (11). If it is

_i statistically significantly different from zero, then it can be concluded that the gender of chairpersons does influence the decision to have a single team.

β

₃^S is the coefficient of interaction term of chairperson’s sex and the proportion of female CEOs in Equation

(11), which is used to test whether there is a difference between genders in deciding to form a single team, when the share of female CEOs is taken into account. If these two coefficients are not consistent to the previous model’s expectations, then the gender neutral hypothesis will not be sustained.

Estimation results of equations (10) and (11) are in tables 3 and 4. Three sets of independent variables are used:

(1) Chairperson’s sex (

f ) for firm i and share of female CEOs (

_i

φ

_ij) in industry j are included as explanatory variables.

(2) In addition to the variables in (1), an interaction term of chairperson’s sex and share of female CEOs (

f

_i

φ

_ij) is added.

(3) In addition to (1) and (2), region (

POST ), established years (

_i

EST ) and firm

_i size (

SIZE ) are included.

_i

Table 3 shows the estimates of Equation (10), which are used to test the tendency of chairpersons of different sexes to opt for a female CEO. The total number of companies used in the estimation is 3,142, since single team companies are excluded.

The table has three parts: columns (1), (2) and (3) use the same index of female CEO

share, which is SCP, and columns (4), (5) and (6) are estimations using the MCP index

as the share of female CEOs, while columns (7), (8) and (9) use the ACP index instead.

Coefficients of the first explanatory variable, female chairperson (PSEX), is positive and statistically significantly different from zero at the 90% level in columns (5), (6), (8) and (9), which means female chairpersons tend to work with female CEOs under classifications of both MCP and ACP. The second explanatory variable, the

female CEO share, is positive and statistically significantly different from zero in all

estimations. It can be inferred that as the female CEO share increases, the number of chairpersons willing to team with female CEOs also increases.

The third explanatory variable is the interaction term of female chairperson and the

female CEO share. Coefficients under the indices of MCP and ACP are negative and

statistically significantly different from zero at 90% and 95% levels, respectively. This implies that when the female CEO share increases, a female chairperson has a lower tendency to cooperate with female CEOs, than male chairpersons.

Next, the results of estimations of Equation (11) are shown in Table 4. The layout of Table 4 is the same as that of Table 3, since explanatory variables of single team estimations are the same as those of female teams estimations. All observed companies are used for single team estimation in Table 4; there are 4,485 companies.

From the first row of Table 4, coefficients of female chairpersons are negative and statistically significantly different from zero at 95% level in seven out of nine columns, which means female chairpersons have lower possibilities of working alone than male chairpersons. Coefficients of the explanatory variable, female CEO share, are negative and statistically significant in columns (3), (6) and (9), which means that as the share of female CEOs increases, the number of companies that opt for a single team decreases.

However, the interaction term of the female chairperson and the female CEO share is insignificant in all estimations. Thus, there is no conclusive information about how the female CEOs share can influence the different genders of chairpersons who opt for a single team.

Combining the estimation results and the two propositions derived in the model section, the gender irrelevant hypothesis is first examined. It is found that coefficients of the interaction term

β

₃^FC<0, which implies β <^FC_f

β

_m^FC. Thus, the gender irrelevant hypothesis is failed. Second, coefficients of single team are examined with coefficients of female chairpersons

β

₁^S< 0, which shows that female chairpersons have a lower tendency to form a single team than male chairpersons. However, coefficient of the interaction term of female chairpersons and female CEOs share,

β

₃^S, is insignificant.

Since the gender neutral hypothesis is sustained only when

β

₁^S>0 and

β

₃^S<0 are

satisfied, the gender neutral hypothesis is also failed.

. Conclusions

Ⅴ

Wage differential and occupation segregation are often considered as the main issues of gender discrimination in labor markets. Since women now receive higher education and have more choices, i.e. other than being housewives only, seriousness of wage gap and occupation segregation is decreasing. However, the promotion process and standards are still not the same and fair for female and male workers.

In this paper, data from the 2006 edition of “Top5000: The Largest Corporations in Taiwan”, published by China Credit Information Service, Ltd. is used to investigate whether there are gender preferences when a chairperson names a CEO. The total number of companies is 4,485. The team formation process is assumed as random matching, which is similar to Boschini and Sjögren (2007).

First, based on the descriptive statistics in the data section, there are only a few female chairpersons and CEOs in these top companies, i.e. about 6%. We also found that chairpersons have a higher tendency to work with same sex CEOs. This means there is gender gap in teamship choices between male and female chairpersons. Second, based on the results of the estimations, both the gender irrelevant hypothesis and gender neutral hypothesis in the random matching model are not sustained by the estimated coefficients of equations (10) and (11).

Notice that the empirical test suggests that a female chairperson has a lower tendency to cooperate with a female CEO than a male chairperson, when the female

CEO share increases in some industry segments. Promoting a candidate as CEO may be

a complex decision, especially in a big company. A chairperson needs to consider many aspects, such as opinions of company’s senior managers and the relationship between the competitors and future CEOs. Therefore, female chairpersons may face more pressure to name a same sex CEO in male dominated working environments. On the

different perspectives, especially in female dominated industries.

For further study, there are a few issues that could be considered. First, more characteristics of companies could be taken into account, such as family-controlled firms, i.e. whether the standard of promotion is based on employees’ performance or blood relationship. Second, board of directors’ characteristics might also help explain the choice of CEOs. For example, the gender ratio and the age structure of the boards might affect the CEO choice.

Appendix (1)

Proof of Propositions

(1) Proposition 1--Gender Irrelevance

.

Using the assumption of proposition 1, the relevant coefficients are derived:

( ) ^{( )} ( ) ^{( )}

(2) Proposition 2—Gender Neutral

.

Using the assumption of proposition 2, the relevant coefficients are derived:

( ) ^{( )} ( ) ^{( )} ( )

References

1. Albrecht, James, Björklund, Anders and Susan Vroman, 2003. Is There a Glass Ceiling in Sweden? Journal of Labor Economics. 21: 145-77.

2. Blau, Francine D. and Lawrence M. Kahn. 2000. Gender Differences in Pay. The

Journal of Economic Perspectives. 14:75-99.

3. Boschini , Anne and Anna Sjögren. 2007. Is Team Formation Gender Neutral?

Evidence from Coauthorship Patterns. Journal of Labor Economics. 25: 325-65.

4. Cannings, Kathy. 1988. Managerial Promotion: the effects of socialization, specialization, and gender. Industrial and Labor Relation Review. 42:77-88.

5. Chang, Chin-fen. 1993. Determinants and Expectation of Workers’ Promotion: The Applicability of Propositions of Internal Labor Markets. Journal of Social Sciences

and Philosophy. 6: 205-30.

6. Hsu, Mei and Been-Lon Chen and Jiun-Der Fang. 2006. The Effects of Industrial Structure and Sex Discrimination on Changes in Female and Male Wage Differential in Taiwan 1978-2003. Academia Economic Papers. 34: 505-39

7. Hung, Yu-Shen. 1996. A Study of the Transformation of Female Employment and the Differentials between Female and Male in Taiwan’s Service Industry.

MasterThesis of the department of Labor Relations of National Chung Cheng University.

8. Kao, Charng. 1991. Male-Female Wage Differentials in Taiwan--A Human Capital Approach. The National Chengchi University Journal. 62: 71-108.

9. Konrad, Alison M. and Kathy Cannings. 1997. The Effects of Gender Role Congruence and Statistical Discrimination on Managerial Advancement. Human

Relations, 50: 1305-28.

10. Landau, Jacqueline. 1995. The relationship of Race and Gender to Manager’s Ratings of Promotion Potential. Journal of Organizational Behavior. 16: 391-400.

11. Liu, Ying-chuan. 1989. Wage Discrimination by Sex in Taiwan. Taiwan

EconomicReview. 17: 357-88.

12. McCue, Kristin. 1996. Promotions and Wage Growth. Journal of Labor

Economics.14: 175-209.

13. Mincer, Jacob and Solomon Polachek. 1974. Family Investments in Human Capital:

Earnings of Women. The Journal of Political Economy. 82, no. 2, Part 2: Marriage, Family Human Capital, and Fertility: S76-S108.

14. Neumark, David and Michele Mclennan. 1995. Sex Discrimination and Women’s Labor Market Outcomes. The Journal of Human Resources 30: 713-40.

15. Olson, Craig A. and Brian E. Becker. 1983. Sex Discrimination in the Promotion Process. Industrial and Labor Relations Review. 36: 624-41.

16. O'Neill, June and Solomon Polachek. 1993. Why the gender gap in wages narrowed in the 1980s? Journal of Labor Economics. 11: 205-28.

17. Reskin , Barbara. 1993. Sex Segregation in the Workplace. Annual Review of

Socialogy. 19: 241-70.

18. Tseng, Min-Chieh. 2001. The Changes of Gender Differences in Earnings in Taiwan:

1982, 1992 and 2000. Journal of Population Studies. 147-209.

19. Ivanova-Stenzel, Radosveta and Dorothea Kübler. 2005. Courtesy and Idleness:

Gender Differences in Team Work and Team Competition. SFB 649 Discussion

Paper 049.

20. Wellington, Alison J. 1993. Changes in the Male/Female Wage Gap, 1976-85. The

Journal of Human Resources. 28: 383-411.

Table 1: Gender of Chairperson and CEO

(1) (2) (3) (4)

Female Male Companies

Chairperson 234 (7.45%) 2,908 (92.55%) (A) Even Team

CEO 180 (5.73%) 2,962 (94.27%)

3,142 (100%)

(B) Single Team 46 (3.43%) 1,297 (96.57%) 1,343 (100%)

(C) Total Observations 460 7,167 4,485

Table 2: Team Compositions of Chairperson and CEO

(1) (2) (3) (4) (5) (6)

Chairperson CEO Obs % Conditional probability (%)

Comparison with the proportion of CEOs (%)

-by gender-

Female Female 19 0.61 8.12

>

5.73

Female Male 215 6.84 91.88

<

94.27

Male Female 161 5.12 5.54

<

5.73

Male Male 2747 87.43 94.46

>

94.27

3142 100.00

Table 3: Probit Estimation of Team Composition with Female CEOs (Marginal Effects)

SCP MCP ACP

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Female Chairperson 0.0214 -0.00797 -0.00486 0.0161 0.0852* 0.0836* 0.0167 0.100* 0.0991*

(PSEX) (0.0175) (0.0617) (0.0637) (0.0166) (0.0493) (0.0497) (0.0167) (0.0541) (0.0548)

Female CEO Share

(SCP) 1.108** 1.065** 0.972** (MCP) 0.891** 0.954** 0.932** (ACP) 0.919** 0.991** 0.970**

(0.325) (0.340) (0.346) (0.111) (0.116) (0.115) (0.118) (0.123) (0.122) PSEX*SCP 0.496 0.394 PSEX*MCP -0.682* -0.695** PSEX*ACP -0.807** -0.825**

(β₃^FC) (1.204) (1.190) (0.351) (0.349) (0.371) (0.370) North Taiwan 0.00136 0.00693 0.00758

(0.00926) (0.00841) (0.00840) 0.00881 0.00388 0.00388 Established Years

11~20 (0.0113) (0.0105) (0.0105) 0.00617 0.00131 0.00171 Established Years

21~30 (0.0128) (0.0116) (0.0117) -0.00849 -0.0154 -0.0155 Established Years

>30 (0.0113) (0.0101) (0.0101)

0.0247* 0.0202 0.0203

Firm Size Level

A2 (0.0145) (0.0135) (0.0136)

0.0167 0.0108 0.0109

Firm Size Level

A3 (0.0143) (0.0131) (0.0131) 0.0139 0.00911 0.00932 Firm Size Level

A4 (0.0142) (0.0130) (0.0131)

0.0329* 0.0273 0.0278

Firm Size Level

A5 (0.0191) (0.0178) (0.0179)

N=3,142. The robust standard errors are listed in the parentheses, and constant is not reported. *significant at the 90% level; **significant at the 95% level. SCP: 5 industry classifications.

MCP: 41 industry classifications. ACP: 41 industry classifications and each company may have more than one industry code.

Table 4: Probit Estimation of Single Team Composition (Marginal Effects)

SCP MCP ACP

(1) (2) (3) (4) (5) (6) (7) (8) (9) Female Chairperson -0.143** -0.0869 -0.100 -0.142** -0.161** -0.170** -0.142** -0.154** -0.163**

(PSEX,

β

₁^S) (0.0234) (0.122) (0.118) (0.0235) (0.0438) (0.0432) (0.0235) (0.0464) (0.0459)

Female CEO Share

(SCP) -0.557 -0.494 -1.244** (MCP) -0.320 -0.347 -0.486** (ACP) -0.361 -0.378 -0.497*

(0.554) (0.567) (0.590) (0.236) (0.245) (0.247) (0.249) (0.257) (0.260) PSEX*SCP -1.335 -1.337 PSEX*MCP 0.462 0.433 PSEX*ACP 0.296 0.239

β

3S

(2.628) (2.633) (0.934) (0.963) (0.983) (1.013)

0.0424** 0.0356** 0.0354**

North Taiwan

(0.0154) (0.0153) (0.0153)

0.0643** 0.0679** 0.0679**

Established Years

11~20 (0.0194) (0.0194) (0.0194) 0.0726** 0.0786** 0.0786**

Established Years

21~30 (0.0218) (0.0218) (0.0218) -0.00449 0.00442 0.00467 Established Years

>30 (0.0206) (0.0206) (0.0206) 0.0686** 0.0703** 0.0703**

Firm Size Level

A2 (0.0225) (0.0225) (0.0225) 0.124** 0.125** 0.125**

Firm Size Level

A3 (0.0228) (0.0229) (0.0229) 0.161** 0.160** 0.160**

Firm Size Level

A4 (0.0230) (0.0230) (0.0230) 0.138** 0.136** 0.136**

Firm Size Level

A5 (0.0278) (0.0277) (0.0277)

N=4,485. The robust standard errors are listed in the parentheses, and constant is not reported. *significant at the 90% level; **significant at the 95% level. SCP: 5 industry classifications.

MCP: 41 industry classifications. ACP: 41 industry classifications and each company may have more than one industry code.

在文檔中 台灣中高齡就業者漸進式退休行為之研究 (頁 46-62)