Impact of corporate governance on banks evidence from yemen & gcc coun...
WriteUp.docx
1. Assignment 3: Empirical study: Boards
Characteristics: Have They Changed?
Seminar supervisor: Dr. Marc Gabarro
Lubomir Rashev 435593
Konstantinos Sfitsoris 431167
Vincent Donovan 425303
Nr. of characters: 12,106/14,399 (no spacing/spacing)
Nr. of words: 2,325
Erasmus School of Economics
FEM11001115
2. 1. Introduction
2. Literature review
3. Board size
3.1 Methodology
3.2 Results
4. Board size and financial performance
4.1 Methodology
4.2 Results
5. Conclusion
1 of 28
3. 1. Introduction
The existence of boards is one of the mandatory prerequisites for incorporations; however
the regulation on this matter is fairly lenient, leading to diversity in terms of board size,
structure, compensation etc. This endogenous governing body has the power to employ,
dismiss and compensate upper management along with the responsibility of handling
disagreements among upper management and shareholders. As with most governance
practises there are heated debates into which structure is optimal in terms of firm market
valuation, shareholder return and agency costs. This paper analyzes if there is a change in
board size after the crisis and attempts to determine if pre-crisis board size correlates with
post crisis financial success.
The 2007/08 financial crisis is the worst economic event to take place since the Great
Depression. The overall effect of the financial downturn was a decline in consumer wealth
and a decrease in economic activity, leading to a global recession. Consequently this
economic catastrophe exposed several flaws in firms corporate governance which were not
able to provide the necessary monitoring in order for the business to operate smoothly.
2. Literature review
The negative correlation between board size and financial performance (Yermack 1998),
would suggest that there will be a decrease in board members during the crisis because of
an extra incentive to cut costs and operate at a financial maximising level. Hermalin &
Weisbach (2001), further this view, in that larger boards are less responsive to poor
performance signals, again justifying why a decrease in board members after the crisis would
be expected. An increase in the amount of independent directors could have both positive
and negative effects for the firm. New skills, diversity and a clear separation of ownership
could be factors which are included along with the chance of the aggravating the free rider
problem (Harris and Raviv 2008).
Fama (1980) demonstrates that outsider directors have an easier time challenging CEO’s that
gray or internal directors. However these independent board members do not have access
to the same amount of information as insiders, this reduces the optimality of firm
2 of 28
4. monitoring. (Jensen 1993; Raheja 2005; Adams and Ferriera 2007; Harris and Raviv 2008;
Masulis and Mobbs 2011). Independent directors also typically hold small equity stakes in
firms in which they are board members, limiting their financial incentives to be vigilant
monitors (Perry 1999).
It should be mentioned that it is difficult to establish a clear relationship between
governance decisions and firm performance due to issues of reverse causality. Since boards
are fundamentally endogenous, their composition and characteristics might be based on the
observed firm performance by the shareholders and reflect their expectations. Therefore,
board structure decisions might be taken on the basis of unobserved factors, correlating
with the error terms of the performed regressions (Hermalin and Weisbach, 2001).
A generalizing conflict is that between the CEO and directors. As a results, the amount of
bargaining power between the two effects how the board evolves over time along with CEO
turnover, firm performance and changes in the ownership. Board size level has a negative
correlation with firm’s performance and the basic argument behind this conclusion, derived
from the fact that an increase in size causes a simultaneous augment in agency problems,
therefore board tend to be more symbolic than a really effective part of the management
process.
Yermack (1996) enhance this theory choosing Tobin’s q as indicator and examined its
correlation with the board size on a representative sample of large U.S. firms. Yermack’s
(1996) results suggest that there is a significant negative correlation between board size and
Tobin’s Q. Furthermore, Wu (2000) noticed a decrease of board size between 1991-1995
which was normally explained because of the willingness of active investors who suppose
that small boards are superior in terms of monitoring than their larger counterparts.
On the other hand, although board composition is not correlated to a firm’s performance, its
member’s actions are correlated with its characteristics. More accurately firms with bigger
proportion of outside directors tend to make better decisions which are of paramount
importance e.g. executive composition, poison pills, acquisitions and CEO replacement.
Besides, there is a both statistically and economically significant increase in stock prices
3 of 28
5. while outside directors added to the board, contrary to insiders, where no definite effect can
be implied.
Yermack (1996) observed that the correlation among pay-performance and the board size
inclined to decrease because small boards provide CEOs better incentives and force them
to bear more risk than large boards do. This evidence propose that CEOs’ effect over their
boards lead in higher pay.
3. Board size
3.1 Methodology:
In order to assess whether there is a change in the board size we have collected board-size
data from 2002-2006 and from 2010-2014, and performed a T-test to establish a significant
difference in the means of the board size for both periods. The period of the financial crisis
as an exogenous factor is assumed to take place during 2007, 2008 and 2009.
An independent group t-test was used to test if the mean values of number of directors
differ before and after the crisis. All firms which were included in the analysis have data for
board size for the years from 2002 to 2014. Financial institutions were omitted due to
regulation purposes, which affect the business's’ financial performance and governance
history. The crisis is assumed to run from 2007 – 2009 and an equal amount of observations
is taken pre/post.
3.2 Results
The t-test shows that the average amount of directors has increased marginally post-crisis.
The t- statistic is -2.19 with 4538 degrees of freedom, this relates to a two-tailed p value of
0.029 which in turn tells us that the variance in means is different from 0. As seen from the
diff = mean(0) – mean(1), the mean(1) is larger and shows that the difference is negatively
correlated. (<0)
Figure 1 demonstrates the information portrayed in the t-test, an increase in the mean
(green) amount of directors post crisis. This can be explained by shareholders attempting to
gain more control by increasing the influence on the board. One of the main deficiencies
4 of 28
6. that came to light during and post crisis was that the boards did not monitoring accordingly
(Cameira, 2003), therefore directors were not able to prevent the company from damaging
itself before the consequences were irreversible. In order to increase the monitoring role of
the board an increase in the amount of independent directors was enforced post crisis. The
positives involved with an addition of independent directors include increase in specialist
skills, diversity of board and a clear separation of ownership and control leading to a more
efficient governance and monitoring system.
Before the crisis (2002-2006) mean board size (reflected in the DIRECTOR variable) is 9.66
and the average board consist of 75% independent members. After the crisis (2010-2014)
the average board size has grown to 9.8 (2.04 standard deviation), while the composition
has remained the same.
The information acquired during this t-test is similar to that from a Moody’s survey. Which
found that there was a 14% increase in outside directors post crisis. Most of these outside
directors had financial background, they hoped that this would enable them to realize poor
internal and external environments before the consequences were irreversible. (Reuters :
Bank boards' financial expertise improves -Moody's, 2010)
4. Board size and financial performance
4.1 Methodology
In order to test the relationship between board size prior to the financial crisis and the firm
performance during the crisis, we calculate a mean board size per company for the period
from 2002 until 2006. We combine this data with performance indicators, namely Tobin’s Q
[TQ] and Return on Assets [ROA], for the period during the crisis. In regards to the time
frame of the crisis we select the period from 2007 to 2008. This differs from our assumptions
for the previous question due to the nature of the data we are using. Here we assume that
board size changes are slower and are reflected over a further period of time (2007-2009),
while performance data is taken into account faster. In relation to this the crisis period can
be delimited to two years.
5 of 28
7. Companies utilize key performance indicators in an attempt to accurately measure firm
performance. The proxies used to regress board size with performance are Tobin’s Q and
Return on Assets [ROA].
Return on Assets is an accounting calculation which analyses a firm's profitability compared
to the total assets. This gives an indication into the efficiency of the management team, and
their ability to generate earnings-returns to ordinary shareholders given the total assets of
the firm.
Tobin’s Q is a valuation ratio, its main benefit is the incorporation of current market
expectations. (Chan-Lee, 1986) In the last 50 years it has been the essential key in the
theory of investments. Recently, it has been applied in financial economics in an attempt to
capture the anticipated return, and therefore represents the measure of risk. (Richard Roll
and J. Fred Weston December 3, 2008).
Our main hypothesis is that companies with larger boards perform worse during periods of
sharp downturn. Our sample consists of publicly traded U.S. companies with available data
for the years between 2002 and 2014. We exclude financial services companies, due to the
disproportionate effect of the exogenous shock on their performance. We perform and
Ordinary Least Squares (OLS) regression to test for the relationship between the average TQ
and average ROA, and the average board size. In our first regression regression we use our
entire sample of companies with available board size and financial data. For our regression
we focus only on S&P 500 companies, due to their similarities in size and performance.
Tobin’s Q = (Total Market Value Fiscal year + Total Liabilities) / Total Assets
ROA = Net Income / Total Assets
Furthermore, we use board composition characteristics and board member characteristics to
attempt to explain the difference in TQ and ROA.
6 of 28
8. 4.2 Results
The mean board size for companies for the the entire period 2002-2014 is 9.36 with a
standard deviation of 2.46. The smallest and biggest boards consist of 3 and 34 members
respectively. Mean age is 61 years with a standard deviation of 8.55.
Multicollinearity is the adverse situation in which the variables used in the model have high
correlations, which misleadingly affects the standard errors. In turn this will affect the
significance values of independent values because of the collinearity between dependent
variables. The Variance Inflation Test measures the extent that the variance of the estimated
components increases over the case of no correlation among mean amount of directors and
the percent of which are independent. Ideally the VIF Factors should be under 10, in this
case (Table 4, 6, 8, 9) the values are 1 indicating that there is negligible collinearity between
the variables.
Figures 2 and 4 show the correlation between Mean Tobin's Q and the Mean amount of
directors in both the standard and the S&P 500 delimited samples. There is a slight negative
gradient with a fairly weak correlation of -0.043 which is significant at a 5% level (Table 5).
The negative correlation could be attributed to some of the companies with smaller boards
having Tobin's q values of higher than 4, which is not present for companies with > 11
directors. Similar effects can be observed in the S&P 500 sample (Table 8), where the
coefficient is -0.059. Regarding the independent variable for independent board member
percentage we find no significant relationship to the dependent variable in both samples.
The R-squared value of both regressions is low (1.3% and 2.3% respectively). The R-squared
indicates how close the observations are to the regression line. This value in the regression
run shows that the model hardly explains any of the variability of the data around its mean.
Tables 6 and 9 show the results for the same model in both samples, where the dependent
variable is Return on Assets. The relationship between board size and ROA is statistically
significant in the full sample of firms, but economically negligible in both samples (coefficient
of 0.004 and 0.001 respectively). The R-squared is low in both samples (0.15% and 0.017%).
7 of 28
9. 5. Conclusion
The relationship between the size and structure of the board of directors and the financial
performance of firms is one which could be argued either way. This paper attempted to
address the empirical evidence which for change in board size from pre-post crisis, as well as
if the pre-crisis board size affected how well the firm coped during the economic downturn.
In the first part of the paper we find that the average amount of directors/firm increases
after the crisis. An additional number of board members post-crisis could be explained by
the Harris and Raviv (2008) paper in conjunction with the Reuters article. Both sources
mention an increase in the amount of independent directors. Firms could choose to do this
in the hope that new skills, structure and expert knowledge might help the company
monitor future downturns and react quicker to the problem.
In the second part of the paper we explore the effect of board size on the firm's’ abilities to
cope with exogenous shocks. The OLS regression demonstrates a fairly negligible
relationship between pre crisis board size and how well the firm fared during the crisis, in
terms of both Tobin’s Q and Return on Assets (ROA). All regressions that were run had a
considerably low R-squared values, this indicates that the models hardly explain any of the
variability of the data around the mean. This does not necessarily mean that the models are
wrong, it could indicate that no two firms are identical and that the real world problems
have no clear correlations. There is also the issue of endogeneity due to the nature of the
board structure. This means that board structure decisions might be taken on the basis of
unobserved factors, which correlate with the error terms of the performed regressions
(Hermalin and Weisbach, 2001). There is no reason to assume that there is no reverse
causality for the choice of board composition and size in relation to the firm’s performance.
As seen in the data, boards will continually evolve in an attempt to optimally deal with the
internal and external environments.
Bibliography
Adams, Renee B., Heitor Almeida, and Daniel Ferreira. “Powerful CEOs And Their Impact on
8 of 28
10. Corporate Performance.” SSRN Electronic Journal SSRN Journaln. pag. Web.
Chan-Lee, James, and Helen Sutch. “Profits And Rates of Return in OECD Countries.” OECD
Economics Department Working Papers(1985): n. pag. Web.
Chan-Lee, James H. Pure Profit Rates and Tobin's q in Nine OECD Countries. Paris:
Organisation for Economic Co-operation and Development, 1986. Print.
Fama, Eugene F. The Disciplining of Corporate Managers. Chicago: Graduate School of
Business, University of Chicago, 1980. Print.
Harris, M., and A. Raviv. “A Theory Of Board Control and Size.” Review of Financial Studies
21.4 (2006): 1797–1832. Web.
Hermalin, Benjamin E., and Michael S. Weisbach. Boards Of Directors as an Endogenously
Determined Institution: a Survey of the Economic Literature. Cambridge, MA.: National
Bureau of Economic Research, 2001. Print.
Jensen, Michael C. “The Modern Industrial Revolution, Exit, And the Failure of Internal
Control Systems.” The Journal of Finance48.3 (1993): 831–880. Web.
Masulis, Ronald W., and Shawn Mobbs. “Are All Inside Directors The Same? Evidence from
the External Directorship Market.” The Journal of Finance66.3 (2011): 823–872. Web.
Perry, Tod Tod. “Incentive Compensation For Outside Directors and CEO Turnover.” SSRN
Electronic Journal SSRN Journaln. pag. Web.
Raheja, Charu G. “Determinants Of Board Size and Composition: A Theory of Corporate
Boards.” SSRN Electronic Journal SSRN Journaln. pag. Web.
Yermack, David. “Higher Market Valuation of Companies with a Small Board of Directors.”
Journal of Financial Economics40.2 (1996): 185–211. Web.
Wu, Y. 2000. “Honey, I Shrunk the Board.”
University of Chicago working paper
9 of 28
11. APPENDIX A - Tables and figures
Table 1 - Descriptive statistics 2002-2014; source: h3.dta
reg1a.dta
Table 2 - Descriptive statistics by period: pre crisis (2002-2006) and post crisis (2010-2014);
source: reg1a.dta
10 of 28
12. Table 3 - T-test of board size grouped by periods: pre crisis (2002-2006) and post crisis
(2010-2014); source: reg1a.dta
Figure 1 - Line plot of average board size; source: reg1a.dta
11 of 28
13. Table 4 - Descriptive statistics for mean board size and mean board composition for
2002-2006, and mean Tobin’s Q and ROA for 2007-2008; source: LAST5.dta
Table 5 - Linear regression: Dependent variable - Mean Tobin’s Q (2007-2008); Independent
variables - Mean Board size (2002-2006), Mean Percentage of independent board members
(2002-2006); source: LAST5.dta
12 of 28
14. Figure 2 - Scatter plot of Mean Board size (2002-2006) and Mean Tobin’s Q (2007-2008);
source: LAST5.dta
13 of 28
15. ROA REGRESSIONS
Table 6 - Linear regression: Dependent variable - Mean ROA (2007-2008); Independent
variables - Mean Board size (2002-2006), Mean Percentage of independent board members
(2002-2006); source: LAST5.dta
14 of 28
16. Figure 3 - Scatter plot of Mean Board size (2002-2006) and Mean ROA (2007-2008); source:
LAST5.dta
Table 7 - Descriptive statistics for mean board size and mean board composition for
2002-2006, and mean Tobin’s Q and ROA for 2007-2008, S&P 500 companies; source:
LAST6.dta
15 of 28
17. Table 8 - Linear regression for S&P 500 Companies: Dependent variable - Mean Tobin’s Q
(2007-2008); Independent variables - Mean Board size (2002-2006), Mean Percentage of
independent board members (2002-2006); source: LAST6.dta
16 of 28
18. Figure 4 - Scatter plot of Mean Board size (2002-2006) and Mean Tobin’s Q (2007-2008), S&P
500 Companies; source: LAST6.dta
17 of 28
19. Table 9 - Linear regression for S&P 500 Companies: Dependent variable - Mean ROA
(2007-2008); Independent variables - Mean Board size (2002-2006), Mean Percentage of
independent board members (2002-2006); source: LAST6.dta
18 of 28
20. Figure 3 - Scatter plot of Mean Board size (2002-2006) and Mean ROA (2007-2008), S&P 500
Companies; source: LAST6.dta
19 of 28
21. APPENDIX B - Stata Commands
cd "C:UsersWildHostageGoogle DriveESESeminar ACFGassignment 3stata"
use b
format %ty year
format %3.1g age
format %20s name
format %20s fullname
format %2.1g female
format %2.1g attend_less75_pct
format %2.1g employment_ceo
format %2.1g employment_cfo
format %2.1g employment_chairman
format %ty year_term_ends
format %ty dirsince
gen cusip2=substr(cusip,1,6)
replace cusip=cusip2
drop cusip2
save e
use d
append using e, force
save g
20 of 28
22. use c
format %20s county
format %20s conm
ren fyear year
gen cusip2=substr(cusip,1,6)
replace cusip=cusip2
drop cusip2
save f
merge m:m cusip year using g, force
save h1
use h1
so cusip year at
keep if _merge==3
encode cusip, gen(cusip2)
drop cusip
ren cusip2 cusip
keep if cusip>0
so cusip year
by cusip year: gen id=[_n]
egen CUSIPYEAR = concat(cusip year), decode p("")
encode CUSIPYEAR, gen(CUSIPYEAR2)
drop CUSIPYEAR
21 of 28
23. ren CUSIPYEAR2 CUSIPYEAR
by cusip year: gen DIRECTOR=[_N]
keep if id==1
drop id
keep if cusip != .
save h3
import excel "C:UsersWildHostageGoogle DriveESESeminar ACFGassignment
3stataregression1.xlsx", sheet("Sheet1") firstrow clear
save extra
use extra
encode CUSIPYEAR, gen(CUSIPYEAR2)
drop CUSIPYEAR
ren CUSIPYEAR2 CUSIPYEAR
save extra, replace
use h3
gen CRISIS = (year>=2007)
ttest DIRECTOR, by(CRISIS)
drop if year==2007
drop if year==2008
drop if year==2009
ttest DIRECTOR, by(CRISIS)
22 of 28
24. drop _merge
merge m:m CUSIPYEAR using extra, force
keep if _merge==3
so cusip
by cusip: gen id=[_N]
drop if id<10
ttest DIRECTOR, by(CRISIS)
save reg1
use reg1
encode sic, gen(sic2)
drop sic
ren sic2 sic
so sic
drop if sic in 4001/4690
ttest DIRECTOR, by(CRISIS)
save reg1a
use h3
gen CRISIS = (year>=2007)
drop _merge
merge m:m CUSIPYEAR using extra2, force
keep if _merge==3
23 of 28
25. drop if year>2009
gen ROA= ni/ at
gen TQ=( mkvalt+ lt)/ at
keep if TQ>=0
keep if at!=.
keep if TQ!=.
keep if ROA!=.
save reg3, replace
encode sic, gen(sic2)
drop sic
ren sic2 sic
so sic
drop if sic in 6559/7063
ttest DIRECTOR, by(CRISIS)
save reg3a
so cusip
so cusip year
so cusip
by cusip: gen id=[_N]
drop if id<8
drop id
save reg3b
24 of 28
26. egen meanteq = mean(teq), by (cusip CRISIS)
egen meanava = mean(ava), by (cusip CRISIS)
egen meanpci = mean(pci), by (cusip CRISIS)
egen meanpce = mean(pce), by (cusip CRISIS)
egen meanpcl = mean(pcl), by (cusip CRISIS)
egen meantq = mean(TQ), by (cusip CRISIS)
egen meand = mean(DIRECTOR), by (cusip CRISIS)
egen meanroa = mean(ROA), by (cusip CRISIS)
egen meanmvl = mean(mkvalt), by (cusip CRISIS)
egen meanat = mean(at), by (cusip CRISIS)
gen logmeanat=log(meanat)
gen logmeanmvl=log(meanmvl)
gen meanPCI=meanpci*100
gen meanPCE=meanpce*100
gen meanPCL=meanpcl*100
so cusip CRISIS
by cusip CRISIS: gen id=[_n]
keep if id==1
drop id
save reg3d
use reg3d
drop if CRISIS==1
25 of 28
27. save reg3e, replace
use reg3d
drop if CRISIS==0
save reg3f, replace
use reg3e
drop _merge
save reg3e, replace
use reg3f
drop _merge
save reg3f, replace
use reg3e
gen meanroa2 = meanroa
gen meantq2 = meantq
drop TQ meantq meanteq meanroa meanmvl meanat logmeanat logmeanmvl
save reg3e, replace
use reg3f
drop meand DIRECTOR meanpci meanava meanpce meanpcl meanPCI meanPCE meanPCL
save reg3f, replace
use reg3e
merge m:m cusip using reg3f
save reg3g
gen meanTQ = meantq-meantq2
26 of 28
28. gen meanROA = meanroa-meanroa2
save LAST2, replace
drop _merge
merge m:m cusip using jesus, force
keep if _merge==3
ren meantq MEANTQ
ren meanroa MEANROA
ren meand MEANDIRECTOR
ren meanPCI PERCINDEPENDENT
ren meanava MEANAGE
gen MEANAGESQ=MEANAGE*MEANAGE
save LAST5
drop if SNP==0
save LAST6
27 of 28