1

Chapter # 4

Data Analysis and Discussion

4.1 Introduction

This chapter presents and discusses results on the impact of corporate governance on working

capital management. The chapter started by presenting the main diagnostic test of OLS model

before adopting various model for analysis. After the diagnostic test the study presents two types

of data analysis; namely descriptive statistics and inferential statistic. First the descriptive

statistic shows the relevant characteristic of variable such as maximization, minimization, mean

and stander deviation of variables used in the study. Second, inferential results highlight an in -

depth examination of the relationship between corporate governance and working capital

management.

4.2 Ordinary least Square:

Ordinary least square (OLS) is the most common method use to fit a line to the data. According

to Brooks (2008, p. 31) ordinary least square is the most comment method use to fit a line to the

data and to estimate slope and intercept in linear regression model. This study used ordinary least

square (OLS) regression to estimate the equation.

4.2.1 OLS Diagnostics Test:

The diagnostic tests carried out to verify the main assumption underlying the ordinary least

square regression and to remove possible problem associated to panel data. The detail of

diagnostic test of the study explains below.

 2

4.2.2 Multicollinearity:

Multicollinearity is the main assumption of ordinary least square model. It is occur when there is

linear relationship between two or more than two independent variables. According to Brooks

(2008, p 171) the multicollinearity problem occurs when the explanatory variables are highly

correlated with each other. In this study the researcher used variance inflation factor VIF and

tolerance to measure the multicollinearity problem.

According to rule of thumb if the VIF value is greater than 10, highly multicollinearity problem

exist in data.

Table 1: Collinearity Statistic

Tolerance VIF
Model BS .903 1.107
1 BM .993 1.007
BC .897 1.115

Dependent Variable: Receivable

Model
2 BS .903 1.107
BM .993 1.007
BC .897 1.115

Dependent Variable: Inventory

Model
3 BS .903 1.107
BM .993 1.007
BC .897 1.115

Dependent Variable: Payable

Note: BS stand for board size, BM stand for board meeting and BC stand for board board
committee. Model 1, 2, and 3 indicated that all three model are tested for collinearity problem.
All independent variables that is BS, BM and BC run with receivable, inventory, payable
respectively, which revealed no problem of collinearity because of VIF less than 10 in all model.
 3

4.2.3 Heteroscedasticity:

Heteroscedasticity problem exist when the error terms do not have a constant variance (Brooks,

2008, p. 132). If the heteroscedasticity occur in the OLS model then the result of hypothesis

become invalid and will no longer reliable.The heteroscedasticity can be deducted by

transferring variables into log. There are several test to check and deduct heteroscedasticity in

model, such as Harvay test, ARCH test, white test and Bruesh Pagan Godfrey test . In this study

Bruesh Pagan Godfrey test is used to test whether or not heteroscedasticity is present in the

models. The null hypothesis is that the variance of the residual is homogenous and an alternative

is the variance of residuals is heterogeneous. Thus, if the P- value significant less than 5 % or

0.05, then null hypothesis should be rejected and accept alternative hypothesis that the variance

is heteroscedastic and vice versa.

H0: The model is Homoscedastic

H1: The model is heteroscedastic

The below result show that there is no problem of heteroscedasticity in all model and show that

the variance of residual is homogenous.

Equation 1: ARD = Bo + B1 (BS) + B2 (BM) +b3 (BC) +r

Table 2: Breusch-Pagan Godfrey Test for Heterockedasticity

F-statistic 2.358527 Prob. F(3,231) 0.0724

Obs*R-squared 6.984176 Prob. Chi-Square(3) 0.0724

Scaled explained SS 16.59758 Prob. Chi-Square(3) 0.0009

Note:

Equation 2 : AID = Bo + B1 (BS) + B2 (BM) +B3 (BC) +r

 4

Table 3: Breusch-Pagan Godfrey Test for Heteroscedasticity

F-statistic 0.338495 Prob. F(3,231) 0.7975

Obs*R-squared 1.028548 Prob. Chi-Square(3) 0.7943

Scaled explained SS 3.428684 Prob. Chi-Square(3) 0.3301

Note:.

Equation 3 : APD = Bo + B1 (BS) + B2 (BM) +B3 (BC) +r

Table 4: Breusch-Pagan Godfrey Test for Heteroscedasticity

F-statistic 0.868778 Prob. F(3,231) 0.4580

Obs*R-squared 2.621884 Prob. Chi-Square(3) 0.4537

Scaled explained SS 15.17679 Prob. Chi-Square(3) 0.0017

Note:

4.2.4 Autocorrelation:

In the presence of autocorrelation phenomenon, ordinary least square are no longer blue (best

linear unbiased). Therefore there is need to test autocorrelation in the residuals.

According to Brooks (2008, p. 139) if the errors terms are correlated with one another, it would

be stated that autocorrelation or serial correlation problem exist in the model. Bruesh Godfrey

test was used to test and deduct autocorrelation problem of models. The null hypothesis is that

there is no autocorrelation problem and alternative hypothesis is that there is autocorrelation

problem. Thus , if the P- value significant less than 5 % or 0.05 ,then null hypothesis should be
 5

rejected and accept alternative hypothesis that lead to presence of autocorrelation problem and

vice versa.

H0: No autocorrelation

H1: Presence of autocorrelation

After Bruesh Godfrey test all model were the problem of autocorrelation and removed by one

period lag on dependent variable (Y) of each model. According to Brooks (2008, p. 140) to

remove the autocorrelation problem take one period lag of Y t, by shifting all of the observation

one period forward in the spreadsheet, written Yt – 1.

Equation 4 : AID = Bo + B1 (BS) + B2 (BM) +b3 (BC) +r

Table 5 : Breusch-Godfrey Test for Autocorrelation

F-statistic 0.537373 Prob. F(2,227) 0.5850

Obs*R-squared 1.102667 Prob. Chi-Square(2) 0.5762

Note:

Equation 5: AID = Bo + B1 (BS) + B2 (BM) +b3 (BC) +r

Table 6 : Breusch-Godfrey Test for Autocorrelation

F-statistic 0.546920 Prob. F(2,227) 0.5795

Obs*R-squared 1.122164 Prob. Chi-Square(2) 0.5706

Note:.
 6

Equation 6: APD = Bo + B1 (BS) + B2 (BM) +b3 (BC) +r

Table 7: Breusch-Godfrey Test for Autocorrelation

F-statistic 8.694215 Prob. F(2,227) 0.0002

Obs*R-squared 16.64929 Prob. Chi-Square(2) 0.0002

Note:

4.3 Model selection criteria (Random or Fixed effect model):

In this research Hausman test used to select either fixed or random effect. According to

Brooks (2008,p ) the easy method to select either fixed effect or random effect is Huasman

test .The null hypothesis is that that the random is an appropriate model and the alternative is that

the fixed effect is an appropriate.

Thus, if the P- value significant less than 5 % or 0.05, then null hypothesis should be rejected

and accept alternative hypothesis and vice versa. Further according to Brooks (2008, p. 506) that

pooled regression assumes that the intercept are the same for each firms (cross section) and for

each year (time series), which is inappropriate assumption, recommended that we could instead

estimate a model with fixed effect.

Table 8: Correlated Random Effects - Hausman Test

Test Summary ChiSq.Static Chi-Sq. d.f. Prob

Cross-section random 21.896067 4 0.0002

Note:
 7

Table 9 : Correlated Random Effects - Hausman Test

Test Summary Chi-Sq.Statistic Chi-Sq. d.f. Prob.

Cross-section random 79.085408 4 0.0000

Note:

Table 9: Correlated Random Effects - Hausman Test

Test Summary Chi-Sq.Statistic Chi-Sq. d.f. Prob.

Cross-section random 79.085408 4 0.0000

Note:

Table 10: Correlated Random Effects - Hausman Test

Test Summary Chi-Sq.Statistic Chi-Sq. d.f. Prob.

Cross-section random 92.971524 4 0.0000

Note:.
 8

4.4 Descriptive statistic:

Table 11: Descriptive Statistics

N Minimum Maximum Mean Std. Deviation

BS 235 5.00 13.00 8.5830 1.72603

BM 235 3.00 35.00 5.6213 2.95934

BC 235 1.00 5.00 2.1957 1.04390

Receivable 235 .00 140.00 19.5361 26.12731

Inventory 235 3.00 417.00 79.5191 65.64078

Payable 235 12.00 355.00 70.7191 44.44258

Valid N 235

(listwise)

Note: BS stand for board size, BM stand for board meeting, BC stand for board committee.

Receivable, inventory and payable (days).

The table presents the descriptive statistic for 47 manufacturing firms of Pakistan which have

period range 2010 – 2014. The study used six variables for the analysis purpose which was

classified into three dependent and three independent variables. The dependent variables which

measured the working capital of the firms are account receivable, account inventory and account

payable. The independent variables which measured corporate governance of the firms are BS,

BM and BC. The section shows the standard deviation and mean of the study. In addition, shows

the minimum and maximum values of each variable which indicates the wide range of each

variable.
 9

As it reveal in table, the minimum value of BS 5 and maximum value is 13, which reflect

minimum and maximum numbers of board of director of Pakistani manufacturing firms. The

mean and standard deviation of BS 8.583 and 1.720 respectively, which show that on average

firms, contain almost 9 directors on aboard and there are fewer variation on the boar size.

Similarly the minimum and maximum values of BM are 3 and 25 respectively, which indicate

minimum and maximum number of board meeting held by board of directors in manufacturing

companies. The mean of BM is almost 6 which show on average meeting held by board of

directors in firms.

The minimum and maximum score of BC is 1 and 5 respectively, which show the minimum and

maximum range of board committee maintain by board of director in firms. The mean value is

almost 2 which reveal on average BC maintain in firms.

The account receivable minimum and maximum range is 0.00 to 140, which show period range

of collection from debtors’. On average manufacturing firm lend credit to debtors for 20 days.

The inventory period range 3 and 140 show period range in which firms sell their inventory .And

on average firms sell their inventory within 80 days.

The descriptive statistic of account payable show that minimum period range of firms to pay to

suppliers is 12 days and maximum 355 days. On average, Pakistani manufacturing pay their dues

to suppliers within 71 days.

The descriptive statistic shows that corporate governance variables that is BS,BM and BC

have stand deviation 1.7, 2.9 and 1.4 respectively, which show fewer variation. This variation

revealed that Pakistan manufacturing firms have stable policies of mentioned variable.

On the other side if we look at the working capital management variables that is receivable,

inventory and payable have stand deviation 26, 66 and 44 respectively, which show larger
 10

variation. This variation reveals that WCM are very sensitive in manufacturing firms of

Pakistan. Further indicated that management of firms continuously revises theirs WCM policies.

Similarly the descriptive statistic show that the mean of receivable less than payable that is 19

and 70 respectively. This combination of receivable and payable reveals good sign for Pakistani

firms because the best case for the firms is that when collect faster from debtors. If firm fail to

collect faster, then the firm have to finance more for operation of working capital . In such

situation the firm fall in bridge finance which goes to more cost that is not good sign for the firm.

4.5 Regression Results:

After describing the diagnostics test and descriptive statistic respectively, the regression analysis

is used to dig out more about impact of corporate governance on working capital management.

This study estimate determinants of working capital management using ordinary least square in

which 3 regression model have been run in order to investigate the impact of corporate

governance on working capital management.

 11

4.5.1 Regression for Model: 1

Equation 7: ARD = Bo + B1 (BS) + B2 (BM) +B3 (BC) +r

Dependent Variable: RECIEVABLE (Log)

Table 12 : Panel Least Squares Method

Variable Coefficient Std. Error T-Statistic Prob.

BS(Log) -0.846052 1.082513 0.781563 0.4358

BM(Log) -0.251339 0.359452 -0.699227 0.4856

BC(Log) 0.153647 0.223571 0.687240 0.4931

RECIEVABLE(Lag) 0.001430 0.007175 0.199258 0.8424

C 0.513787 2.355125 0.218157 0.8276

Effects Specification

Cross-section fixed (dummy variables)

R-squared 0.839756 Mean dependent var 2.044761

Adjusted R-squared 0.781272 S.D. dependent var 1.777206

S.E. of regression 0.831170 Akaike info criterion 2.694128

Sum squared resid 94.64558 Schwarz criterion 3.572098

Log likelihood -202.2480 Hannan-Quinn criter. 3.049848

F-statistic 14.35888 Durbin-Watson stat 2.656929

Prob(F-statistic) 0.000000

Note: Here log indicate that mentioned variables transfer to log. Model 1 indicates that

dependent variable is account receivable.

 12

The table revealed the summary statistic of regression model 1 st. The coefficient of determination

(r2) is 83 %, which indicate that 83 % variation is explained by model and the rest of 17 %

attribute to errors .Further in other words the r 2 indicate that the sufficient enough data points are

very close to fitted regression line. The f- statistics indicate that model is fit with F- statistic14.35

at p- value of 0.00000.

The study found that there is no significant effect of BS on account receivable in days (ARD).

Further indicate that 1 percent increase in BS can lead to decrease in ARD by .846 percent. This

negative relationship reveals that increase in BS can fast the receivable collection from debtors.

The finding is in line of with finding of Dolatabadi & Faradonbeh (2015), Gill and Biger (2013),

Chaudhry & Ahmed ( 2015) and Kajananthan and Achchuthan (2013) show insignificant

negative relationship with account receivable (days). This negative relationship reveals that BS

can fast the collection from debtors.

The regression results of BM with ARD implies that 1 percent increase in BM can decrease ARD

by value of 0.251, but statistically insignificant. This negative relationship indicates that

increase in BS can increase receivables collection from debtors. The combination of these

variables consistent with the finding of Kamau & Basweti (2013) reveals negative relationship of

the variables.

Similarly, the result of BC indicate that 1 percent increase in BC can bring increase in ARD by

value of 0.251, but statistically insignificant. This positive relationship indicates that increase in

BC can slow the collection efficiency. Finding consistent with finding of Kamau & Basweti

(2013) and Kajananthan and Achchuthan (2013) revealed positive BC with working capital

management. On the other side Chaudhry & Ahmed (2015) find negative relationship.
 13

The last, the coefficient of C indicate that an absence of independent variables (BS, BM, BC) can

increase ARD by 0.514 unit with significant of 0.00000. Further it reveals that an absence of

these variables the collection efficiency from debtors can slow of manufacturing firms.

The results from regression model 1 are used to determine hypothesis stated in chapter 1 as

shown in 1.3 sections. The first research hypothesis was that there is no significant influence of

board size, board meeting and board committee on account receivable. So above result reveals

that there is no influence of mentioned variables on receivable .Therefore, it can be conducted

that hypothesis 1 is true.

Thus the overall estimated model can mathematically express as follow.

ARD = 0.514 + 0.846 (BS) – 0.251(BM) +0.154 (BC) 4.5.2

 14

4.5.2 Regression for Model: 2

Equation 8: AID = Bo + B1 (BS) + B2 (BM) +B3 (BC) +r

Dependent Variable: INVENTORY

Table 13: Panel Least Squares Method

Variable Coefficient Std. Error T-Statistic Prob.

BS 2.032888 4.548549 0.446931 0.6556

BM -0.437587 1.213176 -0.360696 0.7189

BC -0.994753 4.617018 -0.215454 0.8297

Inventory (lag) 0.102406 0.089622 1.142648 0.2552

C 58.90555 41.52455 1.418572 0.1583

Effects Specification

Cross-section fixed (dummy variables)

R-squared 0.847311 Mean dependent var 79.82447

Adjusted R-squared 0.791584 S.D. dependent var 64.68158

S.E. of regression 29.52879 Akaike info criterion 9.834701

Sum squared resid 119457.1 Schwarz criterion 10.71267

Log likelihood -873.4619 Hannan-Quinn criter. 10.19042

F-statistic 15.20492 Durbin-Watson stat 1.653491

Prob(F-statistic) 0.000000

Note: BS stand for board size, BM stand for board meeting and BC stand for board committee.

Model 2 indicate that here dependent variable is inventory. Lag indicates that take one period lag

of dependent variable.
 15

The table revealed the summary statistic of regression model 2. The coefficient of

determination (r2) is 84 %, which indicate that 84 % variation is explained by model and the rest

of just 16 % attribute to errors .Further in other words the r 2 indicate that the sufficient enough

data points are very close to fitted regression line. The f- statistics indicate that model is fit with

F- statistic 15.20492at p- value of 0.00000.

The study found that there is no significant effect of BS on AID (account inventory in days).

Further indicate that 1 unit increase in BS can lead to increase the AID by 2.032888 units. This

positive relationship reveals that increase in BS can lead to slow the inventory turnover

efficiency of manufacturing firms. The study consistent with Gill and Biger (2013) that BS have

insignificant negative relationship with inventory(days). On other side Chaudhry & Ahmed,

(2015) found positive relationship.

The regression results of BM with AID implies that 1 unit increase in BM can decrease AID by

value of 0.437587, but statistically insignificant This negative relationship reveals that increase

in BM can lead to fast the inventory turnover efficiency of manufacturing firms.

Similarly, the result of BC indicate that 1 unit increase in BC can bring decrease in AID by value

of 0.994753, but statistically insignificant. This negative relationship indicates that increase in

BC can fast the inventory turnover efficiency. Chaudhry & Ahmed (2015) investigated that BC

has positive relationship with inventory (days). The study consistent with Kamau & Basweti

(2013) shows that BC positive relationship with WCM.

The last, the coefficient of C indicate that an absence of independent variables (BS, BM, BC) can

increase AID by 58.90555units with significant of 0.00000. Further it reveals that an absence of

these variables the inventory turnover can slow of manufacturing firms.

 16

The results from regression model 2 are used to determine hypothesis stated in chapter 1 as

shown in 1.3 sections. The first research hypothesis was that there is no significant influence of

board size, board meeting and board committee on account inventory. So above result reveals

that there is no influence of mentioned variables on inventory .Therefore, it can be conducted

that hypothesis 2 is true.

Thus the overall estimated model can mathematically express as follow.

AID = 58.90555+ 2.032888 (BS) - 0.437587 (BM) - 0.994753 (BC)

 17

4.5.3 Regression for Model: 3

Equation 8: AID = Bo + B1 (BS) + B2 (BM) +B3 (BC) +r

Dependent Variable :PAYABLE

Table 14 : Panel Least Squares Method

 18

Variable Coefficient Std. Error T-Statistic Prob.

BS -3.857466 3.668242 -1.051584 0.2948

BM -1.439870 0.977839 -1.472502 0.1432

BC 0.751273 3.734526 0.201170 0.8409

PAYABLE (Lag) 0.007398 0.089714 0.082459 0.9344

C 109.4926 33.93173 3.226848 0.0016

Effects Specification

Cross-section fixed (dummy variables)

R-squared 0.798741 Mean dependent var 70.65957

Adjusted R-squared 0.725289 S.D. dependent var 45.40597

S.E. of regression 23.79856 Akaike info criterion 9.403220

Sum squared resid 77592.91 Schwarz criterion 10.28119

Log likelihood -832.9026 Hannan-Quinn criter. 9.758940

F-statistic 10.87432 Durbin-Watson stat 1.950001

Prob(F-statistic) 0.000000

Note: BS stand for board size, BM stand for board meeting and BC stand for board committee.

Model indicate that here is payable (days) use as dependent variable. Lag indicated that take one

period lag on dependent variable.

The table revealed the summary statistic of regression model 3 rd.The coefficient of determination

(r2) is almost 80 %, which indicate that 80 % variation is explained by model and the rest of just
 19

20 % attribute to errors .Further in other words the r 2 indicate that the sufficient enough data

points are very close to fitted regression line. The f- statistics indicate that model is fit with F-

statistic 10.87432at p- value of 0.00000.

The study found that there is no significant effect of BS on account payable (days). Further

indicate that 1 unit increase in BS can lead to decrease in the APD by 3.857466 units. This

negative relationship reveals that increase in BS can lead to fast the efficiency of payment to

suppliers of manufacturing firms. Chaudhry and Ahmed(2015) found that BS have negative

relationship with payable (days). On the other hand Dolatabadi and Faradonbeh (2015), Gill and

Biger (2013) found positive relationship with account payable.

The regression results of BM with APD implies that 1 unit increase in BM can decrease APD by

value of 1.439870, but statistically insignificant. This negative relationship reveals that increase

in BM can lead to fast payment to suppliers of manufacturing firms. Chaudhry and Ahmed(2015)

found negative relationship with BM and account payable (days). Kamau & Basweti (2013) also

found insignificant negative relationship .

Similarly, the result of BC indicate that 1 unit increase in BC can bring increase in APD by value

of 0.751273, but statistically insignificant. This positive relationship indicates that increase in

BC can slow payments to suppliers. Chaudhry and Ahmed (2015) found positive relationship of

BC with account payable (days). Kajananthan and Achchuthan (2013) also found a positive

relationship with CCC.

The last, the coefficient of C indicate that an absence of independent variables (BS, BM, and

BC) can increase APD by 109.4926 units with p value is 0.0016. Further this indicates that an

absence of these variables can slow payment to suppliers of manufacturing firms.

 20

The results from regression model 3 are used to determine hypothesis stated in chapter 1 as

shown in 1.3 sections. The third research hypothesis was that there is no significant influence of

board size, board meeting and board committee on account payable. So above result reveals that

there is no influence of mentioned variables on payable .Therefore, it can be conducted that

hypothesis 2 is true.

Thus the overall estimated model can mathematically express as follow.

APD = 109.4926 - 3.857466 (BS) - 1.439870 (BM) + 0.751273 (BC)

In short, Study found from above empirical results that corporate governance have not

significant impact on working capital management .The study consist with Kamau & Basweti

(2013) CG does not improve WCM and also study by Kajananthan, R. Achchuthan, (2013) CG

have not significant influence on working capital management.

On against side Gil and Biger (2013) argue that corporate governance improve working capital

management. Similarly Chaudhry and Ahmad (2015) also reveals that corporate governance

have significant influence on working capital management

Chapter # 5 Conclusion and recommendation

5.1: Conclusion

This research studied the impact of corporate governance on working capital management of

manufacturing firms in Pakistan. The study used quantitative research approach.

 21

In this research the data of 47 manufacturing firms of Pakistan was analyzed using descriptive

statistics and regression analysis for period of 2010 – 2014. The OLS regression model has been

used to analyzed the impact of corporate governance on working capital management .The study

used account receivable in days (ARD) ,account inventory in days (AID) and account payable in

days (APD) as a dependent working capital management variables. And board size (BS) ,board

meeting ( BM) and board committee (BC) used as an independent corporate governance

variables .

Descriptive statistics were used to examine the characteristics of the chosen variables. The mean

value of account receivable was 20 which reveal that on average firms collected receivable from

debtors for 20 days. Average firms sell their inventory within 80 days. The average account

payable period reveals that firms pay their dues to suppliers within 71 days .Similarly on average

9 board of directors, 6 board meeting and 2 board committee maintain by firms.

The regression analysis of the account receivable in days (ARD) indicate that there is no

significant influence of board size ( BS) ,board meeting (BM) and board committee (BC) on

account receivable . Further BS and BM have negative relationship which reveals that increase in

BS and BC can lead to fast receivable collection of firms .Similarly BC have positive

relationship which reveals that increase in BC can lead slow receivable of firms .

The regression analysis of the account inventory in days (AID) indicates that there is no

significant influence of BS ,BM, BC on account inventory in days (AID). Further BS have

positive relationship which reveals that increase in BS can lead to slow inventory turnover

efficiency. Similarly BM and BC have negative relationship which reveals that increase in BC

can lead to fast the inventory turnover efficiency.

 22

The regression analysis of account payable in days (APD) indicates that there is no significant

influence of BS, BM and BC on account payable in days (APD). Further BS and BM have

negative relationship which reveals that increase in BS and BM can lead to fast payment

efficiency to suppliers. Similarly BC have positive relationship which reveals that increase in BC

can lead to slow the payment efficiency of suppliers.

Final in all models the BS and BM have negative relationship with dependent variables ARD,

AID and APD which lead to increase the efficiency of working capital management that is seem

to good sign for firm. And on flip side the BC seem not good exercise for working capital

management because BC has positive relationship in all models which lead to slow working

capital efficiency. So above result reveals that increase in BS and BM can be bit well for

Pakistani firms to improve working capital.

5.2: Recommendation for future research

There is need for further study to carry out the impact of corporate governance on working

capital management of firms by incorporating more corporate governance variables that effect

working capital management. Further study can be conducted to investigate impact of corporate

governance on working capital management of homogenous and heterogeneous firms of

Pakistan.
23