Get documents about "Composition of Board of Directors and Organizational Performance
Document Sample
Composition of Board of Directors and
Organizational Performance
-- An empirical study on Chinese
shareholding companies
2008/11/5 Jenny Tian 1
Overview
Theoretical Background
Hypotheses and Results
Key methodological issues
Sample and variables
Initial model
Diagnostics and decisions
3 Alternative models
2008/11/5 Jenny Tian 2
Theoretical Background
Research question
How does the board of directors influence
organizational performance?
Different theoretical lenses
Corporate governance theory: focused on where the
directors come from (insider vs. outsider)
Strategic leadership theory: focused on who the
directors are and how they work together (team
composition and process)
2008/11/5 Jenny Tian 3
Hypotheses and Results
(DV: Organizational Performance)
Corporate governance structure
H1: Proportion of independent directors (+) [Agency theory, not
supported]
H2: Proportion of affiliated directors (+) [Stewardship theory, supported]
Strategic leadership theory
H3: Board size (-) [Supported]
H4: Average age of the board (-) [Not supported]
H5: Age diversity of the board (-) [Note supported]
H6: Educational background diversity of the board (-) [Marginally
supported]
H7: Functional background diversity of the board (-) [Supported]
2008/11/5 Jenny Tian 4
Key Methodological Issues
Two methods used to deal with outliers
Robust regression (LTS) on the full sample
OLS on the reduced sample from which the outliers
were deleted
LTS increases R square when the outlier problem
is present, though the improvement is limited.
Outlier deletion leads to better estimation.
However, such a practice needs to be justified
carefully.
2008/11/5 Jenny Tian 5
Sample and Variables
Sample
Chinese shareholding companies that completed their initial
public offering (IPO) in 1996. N=203.
Variables
Variable Name Description
DV: ROA Return on asset
Control variables
INDUSTRY Dummy (0=non-manufacturing; 1=manufacturing)
FIRMAGE History: # of years since incorporation
ASSET Firm size
STATESHARE Shares owned by the state (%)
IVs
BSIZE Board size (total # of directors)
INDE_IPO Proportion of independent directors
AFF_IPO Proportion of affiliated directors
BAGE_MEAN Average age of the board
BAGE_DIV Board age diversity (coefficient of variation = s.d. / mean)
BEDU_DIV Board educational background diversity (Heterogeneity
index), ranging from 0 to 1, higher value indicating higher
diversity
BFUN_DIV Board functional background diversity (Heterogeneity index)
2008/11/5 Jenny Tian 6
Initial Model (n=203, OLS, no transformation)
Table 1. Initial Model (n=203, OLS, No Transformation)
Model 1 (Control Model) Model 2 (Full Model)
Variables Unstandardized Standard error Standardized Unstandardized Standard error Standardized
coefficients coefficients coefficients coefficients
(Constant) 5.700E-02 .007 7.635E-02 .045
INDUSTRY 2.052E-02 .007 .197** 1.831E-02 .007 .176*
FIRMAGE 1.232E-04 .000 .044 -4.8E-05 .000 -.017
ASSET -5.0E-08 .000 -.125† -3.9E-08 .000 -.096
STATESHARE -6.0E-04 .011 -.004 5.007E-03 .011 .034
BSIZE -3.2E-03 .001 -.201**
INDE_IPO -8.1E-03 .016 -.042
AFF_IPO 4.182E-02 .017 .196*
BAGE_MEAN 3.368E-04 .001 .032
BAGE_DIV 5.248E-02 .073 .057
BEDU_DIV -2.0E-02 .071 -.020
BFUN_DIV -3.4E-02 .027 -.094
F value 2.929 2.614
R square .06 .13
Adjusted R square .04 .08
R square change .075
† p < .10
* p < .05
** p < .01
*** p < .001
2008/11/5 Jenny Tian 7
Diagnostics and Decisions
Assumption Test Result Decision
Outlier (19 cases) Studentized 19 cases were ANOVA shows that the 19 cases were
residuals identified as outliers. significantly different from the others on
Cook’s distance several dimensions.
Hat value Keep the “outliers” and try robust
DEFITS regression method.
Drop the outliers and use OLS.
Heteroscedasticity Graph No violation
Statistical test (Ui)
Non-normality Q-Q plot No violation
Auto-correlation Durbin-Watson test No violation
Collinearity VIF No violation
Tolerance
Non-linearity Bivariate plots Some variables appear Variable transformation
to be associated with ASSET LGASSET (Log 10)
ROA in an nonlinear STASHARE STA_SQRT (Square root)
manner. BSIZE BSIZE_LG (Log 10)
Variable selection None All the variables were kept for the theoretical
reason.
2008/11/5 Jenny Tian 8
Alternative Model 1 (n=203, transformation, OLS)
Model 1 (Control Model) Model 2 (Full Model)
Variables Unstandardi Standard Standardized Unstandardi Standard Standardized
zed error coefficients zed error coefficients
coefficients coefficients
(Constant) .142 .047 .198 .064
INDUSTRY 1.959E-02 .007 .188** 1.811E-02 .007 .174*
FIRMAGE 1.273E-04 .000 .046 -4.466E-05 .000 -.016
LGASSET -1.865E-02 .010 -.129† -1.814E-02 .010 -.125†
STA_SQRT -4.069E-03 .010 -.031 1.038E-03 .010 .008
BSIZE_LG -7.621E-02 .027 -.205**
INDE_IPO -9.174E-03 .016 -.048
AFF_IPO 4.006E-02 .016 .188*
BAGE_MEAN 4.689E-04 .001 .045
BAGE_DIV 4.645E-02 .073 .051
BEDU_DIV -1.538E-02 .071 -.015
BFUN_DIV -3.771E-02 .027 -.105
F value 3.040 2.686
R square .06 .13
Adjusted R square .04 .08
R square change .076
2008/11/5 Jenny Tian 9
Alternative Model 2 (n=203, transformation, LTS)
Variables Unstandardized coefficients
(Constant) 0.2946
INDUSTRY 0.0119
FIRMAGE -.0001
LGASSET -.0307
STA_SQRT .0070
BSIZE_LG -.1043
INDE_IPO -.0219
AFF_IPO .0435
BAGE_MEAN .0004
BAGE_DIV .0508
BEDU_DIV -.0542
BFUN_DIV -.0479
Robust R square .161
2008/11/5 Jenny Tian 10
Alternative Model 3 (n=184, transformation, OLS)
Model 1 (Control Model) Model 2 (Full Model)
Variables Unstandardi Standard Standardized Unstandardi Standard Standardized
zed error coefficients zed error coefficients
coefficients coefficients
(Constant) .171 .043 .262 .055
INDUSTRY 1.074E-02 .006 .130† 9.720E-03 .006 .118†
FIRMAGE 9.896E-05 .000 .044 -8.086E-05 .000 -.036
LGASSET -2.475E-02 .009 -.201** -2.525E-02 .008 -.205**
STA_SQRT 3.640E-03 .009 .033 1.111E-02 .008 .101
BSIZE_LG -8.421E-02 .022 -.276***
INDE_IPO -4.919E-03 .013 -.032
AFF_IPO 5.068E-02 .013 .295***
BAGE_MEAN 3.364E-04 .001 .038
BAGE_DIV 6.119E-02 .062 .078
BEDU_DIV -.116 .062 -.130†
BFUN_DIV -7.554E-02 .026 -.224**
F value 3.213 5.111
R square .07 .25
Adjusted R square .05 .20
R square change .18
2008/11/5 Jenny Tian 11
Validation (Alternative Model 3)
Calibration graph
ROA3_96 Actual
0.1
0
.0 .1
ROA3_96 Predicted P<.0001 RSq=0.25
RMSE=0.0337
Leave-out-one-group cross-validation
20 groups; Average R square = .27
Bootstrap: Unstandardized regression coefficients
The observed coefficients are not significantly different from
the bootstrap means.
2008/11/5 Jenny Tian 12
Related docs
