ASSOCIAÇÃO DE POLITÉCNICOS DO NORTE (APNOR) INSTITUTO POLITÉCNICO DE BRAGANÇA Ratios and indicators that determine return on equity Davit Kharatyan Final Dissertation presented to Instituto Politécnico de Bragança To obtain the Master Degree in Management, Specialisation in Business Management Supervisors: Prof. Dr. Jose Carlos Lopes Prof. Dr. Alcina Maria de Almeida Rodrigues Nunes Prof. Dr. Lusine Aghababyan Bragança, May, 2016.
67
Embed
Ratios and indicators that determine return on equity Kharatyan Master thesis... · DuPont model is a widely used gauge of profitability which links several factors to ROE. Liesz
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ASSOCIAÇÃO DE POLITÉCNICOS DO NORTE (APNOR)
INSTITUTO POLITÉCNICO DE BRAGANÇA
Ratios and indicators that determine
return on equity
Davit Kharatyan
Final Dissertation presented to Instituto Politécnico de Bragança
To obtain the Master Degree in Management, Specialisation in Business
Management
Supervisors:
Prof. Dr. Jose Carlos Lopes
Prof. Dr. Alcina Maria de Almeida Rodrigues Nunes
Prof. Dr. Lusine Aghababyan
Bragança, May, 2016.
ASSOCIAÇÃO DE POLITÉCNICOS DO NORTE (APNOR)
INSTITUTO POLITÉCNICO DE BRAGANÇA
Ratios and indicators that determine
return on equity
Davit Kharatyan
Supervisors:
Prof. Dr. Jose Carlos Lopes
Prof. Dr. Alcina Maria de Almeida Rodrigues Nunes
Prof. Dr. Lusine Aghababyan
Bragança, May, 2016.
i
Abstract
This study aims to investigate factors that may affect return on equity (ROE). The ROE is a gauge of
profit generating efficiency and a strong measure of how well the management of a firm creates value
for its shareholders. Firms with higher ROE typically have competitive advantages over their
competitors which translates into superior returns for investors. Therefore, seems imperative to study
the drivers of ROE, particularly ratios and indicators that may have considerable impact. The analysis
is done on a sample of 90 largest non-financial companies which are components of NASDAQ-100
index and also on industry sector samples. The ordinary least squares method is used to find the most
impactful drivers of ROE. The extended DuPont model’s components are considered as the primary
factors affecting ROE. In addition, other ratios and indicators such as price to earnings, price to book
and current are also incorporated. Consequently, the study uses eight ratios that are believed to have
impact on ROE. According to our findings, the most relevant ratios that determine ROE are tax burden,
First and foremost, I would like to thank International Credit Mobility Programme (ICM ERASMUS+) for
providing me with a scholarship to carry out my studies in Instituto Politécnico de Bragança and giving
me an opportunity to get a double diploma degree.
I would like also to express my sincere gratitude to my research supervisors Prof. Dr. Jose Carlos
Lopes, Prof. Dr. Alcina Maria de Almeida Rodrigues Nunes and Prof. Dr. Lusine Aghababyan for the
useful comments, remarks and engagement through learning process of this master thesis. Without
their assistance and dedicated involvement in every step throughout the process, this paper would
have never been accomplished.
v
Acronyms
ROE - Return on equity
TB - Tax burden
IB - Interest burden
OM – Operating margin
AT – Asset turnover
FL – Financial leverage
PE – Price to earnings ratio
PB – Price to book ratio
CUR – Current ratio
OLS - Ordinary least squares
VIF- Variance inflation factor
vi
Table of Contents
List of Tables ........................................................................................................................................... vii
1. Ratios/indicators framework and literature review ............................................................................... 3
1.1 Financial ratios and indicators ............................................................................................................ 3
1.2 The DuPont model ............................................................................................................................. 7
1.3 Literature review ............................................................................................................................... 10
2. Research Methodology ...................................................................................................................... 14
2.1 Data and sample .............................................................................................................................. 14
2.2 Methodology and data treatment ..................................................................................................... 18
Table 3. Statistical distribution of variables’ values for the complete set of firms in the sample ........... 21
Table 4. Results of Pearson correlation coefficient between the independent variables and the return
on equity ................................................................................................................................................. 24
Table 5. Results of the OLS regression analysis for all companies, using original measurement units
and logarithmic values ............................................................................................................................ 27
Table 6. Statistical distribution of variables’ values for the Technology sector sample ......................... 28
Table 7. Results of Pearson correlation coefficient between the independent variables and the return
on equity ................................................................................................................................................. 31
Table 8. Results of the OLS regression analysis for the technology sector sample, using original
measurement units and logarithmic values ............................................................................................ 32
Table 9. Statistical distribution of variables’ values for the Consumer sector sample ........................... 33
Table 10. Results of Pearson correlation coefficient between the independent variables and the return
on equity ................................................................................................................................................. 35
Table 11. Results of the OLS regression analysis for the consumer sector sample, using original
measurement units and logarithmic values ............................................................................................ 37
Table 12. Statistical distribution of variables’ values for the other sectors sample ................................ 38
Table 13. Results of Pearson correlation coefficient between the independent variables and the return
on equity ................................................................................................................................................. 41
Table 14. Results of the OLS regression analysis for the other sector sample, using original
measurement units and logarithmic values ............................................................................................ 42
Table 15. Comparative analysis of OLS regression results ................................................................... 45
Table 16. Global sector companies ........................................................................................................ 52
price to earnings, price to book/ market to book and current2. It is noteworthy, that the analysis is not
only carried out for all index constituents, but also for two major industry sectors, namely: Technology
and Consumer sectors. The index constituents that do not belong to one of the two above mentioned
sectors are grouped in a residual sector called "other" sector. The reason behind this decision is that
industries have different characteristics resulting in discrepancies in many ratios. Mubin and Iqbal
(2014) agree that there is a sector impact on ROE. Therefore, industry analysis is crucial to make a
comparison between industries checking if all indicators provide the same important insights for
different industries and how those differences alter return on equity.
Tax burden, interest burden, operating profitability, asset turnover and financial leverage indicators are
the components of the extended DuPont model which explains the return on equity. In this respect, the
reason behind the inclusion of these variables into the analysis is very obvious - they all have a direct
impact on return on equity, as the decomposition explicitly links them to return on equity. The price to
earnings ratio and the price to book ratio were previously documented to have impact on profitability.
2 The formulas used to calculate these variables are presented in Table 2
16
According to Saleem et al., (2011) they were the first who attempted to link profitability and liquidity
measures. They found that liquidity measures are not related to return on equity. A liquidity measure
(current ratio) is incorporated in this study to further study the relationship between liquidity and
profitability measures. For this reason, and in conjunction with the extended DuPont components,
these variables were also included in the analysis as explanatory variables for the change in the return
on equity.
The unit of measure of the variables is either euro amounts or percentages. Formulas depicted in
Table 2 can differ from other sources as different databases use different formulas to calculate
indicators. The ratios from table 2 are acquired from the Bloomberg database and were used to
calculate the independent variables/indicators.
17
Table 2. Description of dependent and independent variables and the expected relation between them
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016
Variable Abbrevation Description RatioUnit of
measure
Type of
association
Operating margin OMMeasures how much is lefot of revenue cosidering cost of
goods sold and operating expenses% +
Note: The ratios are acquired from Bloomberg database and were used to calculate the variables in study. The notation n.a. means that is an expected relation is not applicable. ROE is the
dependent variable
-
+
+
n/a
+
+
+
(+) / (-)
PB Compares a stock's market value to its book value €
Return on equity
Tax burden
Interest burden
Price-to-earnings
Price-to-book
Measures the effect on interest on ROE %
PEMeasures a company's current share price relative to its
per-share earnings€
ROEAmount of income returned as a percentage of
shareholders equity%
TBThe proportion of the company's profits retained after
paying income taxes%
IB
Asset turnover ATMeasures the efficiency of a company's use of its assets
in generating sales revenue €
Financial leverage FLIs the use of borroewed capital to increase potential return
of an investment€
Current ratio CURMeasures a company's ability to cover its short-terrm
liabilities with its current assets€
( )
18
2.2 Methodology and data treatment
With respect to methodology of inferential data analysis, the Ordinary Least Squares (OLS) regression
method is used in this study to both identify the most relevant indicators that explain the changes on
return on equity and to quantify the relation between each indicator and the return on equity. In other
words, the OLS regression method is applied to find out which variables have the most explanatory
power or variations occurring in return on equity quantifying that explanatory power.
In this study, it is intended to determine which among eight selected variables influence return on
equity of companies operating in Nasdaq-100 NDX index. The existence of more than an explanatory
variable puts the present analysis in the framework of a multiple linear regression analysis. In this
case, the dependent variable (return of equity) in approximately linearly related to the independent
OM 90 18,65 -95,58 68,00 163,57 17,67 0,95 -2,84 21,54
AT 90 0,79 0,09 3,55 3,46 0,61 0,78 2,23 8,38
FL 90 2,57 1,11 11,97 10,86 1,59 0,62 3,44 18,48
PE 90 37,21 4,58 453,04 448,46 59,84 1,61 5,44 34,60
PB 90 5,61 1,03 40,30 39,28 5,18 0,92 3,91 24,35
CUR 90 2,41 0,14 11,25 11,10 1,77 0,74 2,20 10,02
Note: All the values are presented in the same unit of measurment of the variables with the exception of the coefficient of variation
that is presented in %
22
To sum up, return on equity, interest burden, operating margin and price to earnings variables are
characterized by a significant degree of dispersion around their respective means compared to tax
burden, asset turnover, financial leverage, price to book and current ratios as shown above by
coefficient of variation, skewness and kurtosis values.
3.1.2 OLS regression analysis results for global sample
The OLS method is applied to identify and quantify which of the selected variables determine changes
in the return on equity of the 90 companies of Nasdaq-100 NDX index selected for analysis. It allows
also to verify the possible relation between each independent variable and the dependent variable –
ROE.
As explained in the previous subsection 3.1, some variables are presented in percentage terms while
others are presented in monetary units (€) which makes the comparison of each variable’s impact on
ROE difficult. For an obvious reason it is necessary to present them in a same unit of measure to
simplify the comparison of results. Additionally, the descriptive statistical analysis showed that some
variables exhibit high values of range (the distance between their minimum and maximum values were
big). Therefore, the linear functional form of the model is transformed into a logarithmic functional form
– all the variables will be used in their logarithmic format. Logarithmic values are known to decrease
the degree of dispersion of a variable’s values. Second, the transformation allows to analyze all the
coefficients in percentage values. Thus, a second model using the same variables is estimated – the
only difference between the first and the second model is that the former uses the values with original
units of measure, whereas the latter uses logarithmic values. The second model is presented in
equation 7:
(7)
Due to presence of negative values in the dataset some observations are excluded from the second
model. For each model the number of effective observations (companies) used is presented in OLS
regression analysis.
Another important statistical indicator to present is the Pearson correlation coefficient between each
explanatory/independent variable and the dependent variable the study wants to explain. The
presentation of such an indicator allows to explore which independent variable may be positively or
negatively related with the return on equity and the strength of such a relation. The results of the
23
Pearson correlation coefficient with statistical significance for both original values and logarithmic
values are presented in Table 4. Results with no statistical significance are not present. The number of
observations available for each variable is presented in brackets. According to table 4 tax burden,
interest burden and price to book indicators are strongly correlated with return on equity. Whereas,
Price to earnings ratio has a negative and strong association with return on equity and current ratio is
not correlated with return on equity.
24
Table 4. Results of Pearson correlation coefficient between the independent variables and the return on equity
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016
TB IB OM AT FL PE PB CUR
Normal ROE0.411*
(90)-
0.346*
(90)-
0.554*
(90)
- 0.212*
(90)
0.177*
(90)-
Logarithmic ROE0.604*
(88)
0.434*
(87)
0.302*
(87)
0.320*
(88)
0.197
(88)
- 0.506*
(88)
0.4061*
(88)-
Note: (*) means that the coefficient presents 5% level of significance. Values with no stars indicate 10% level of signficance. (-) indicates no relationship between the dependent
and independent variable in question. Number of obervations available for analysis arepresented in brackets.
Variables
25
Table 5 presents the results of the regression analysis for all index constituents, using both the original
measures (model 1) and the logarithmic values (model 2).
In the table, the first column presents the independent variables that may have influence on return on
equity. The second column illustrates the estimated coefficients which reflect both the strength and
type of relationship an independent variable has with a dependent variable, that is, if the changes in
the independent variable make the return on equity change in the same direction5. The coefficients are
given in the same measurement units as their associated independent variables and denote the
expected change in dependent variable for every 1-unit change in the independent variable holding all
other independent variables constant. The third column presents the robust standard error (that
guarantees the hypothesis of homoscedasticity is not violated and therefore the results of the
estimation are robust and trustworthy). The fourth column presents the p-values and the associated
levels of statistical significance for each coefficient. Finally, the last column presents the VIF values
that allow to conclude about collinearity between independent variables.
Indicators of the estimation quality and accuracy are also presented in the table which are the
coefficient of determination (R-squared), the test of joint statistical significance (the F-test) and the root
mean squared error (Root MSE). The results of the Ramsey test for omitted variables are also
presented. N indicates the observations available to perform the estimation in each model.
The R-squared indicates how much of the variation that occurred in the return on equity are explained
by the variation that happened in the independent variables. A value near to 1 indicates that the model
explains all the variability of the response data around its mean. The F-test statistical significant
indicates that the variables jointly create a good model. The smaller the Root MSE the more accurate
is the estimation. The Ramsey test checks the existence of omitted variables. It indicates if the model
includes the most important variables that explain the changes in the return on equity or, in other
words, no important variable is omitted from the model.
As shown in Table 5, logarithmic values present better results as indicated by, for example, a higher R-
squared value. Moreover, the regression analysis with original values presents a Ramsey values
statistically significant which indicates the existence of omitted variables, that is, more variables should
be added to the model to make the analysis more accurate.
The model presents a R-squared equal to 0,6786 for original values which means that almost 68% of
the variation in the return on equity are explained by the variations that happen in the eight variables
presented in the model. However, the results of regression analysis for logarithmic values indicate a
much higher R-squared value - 93% of the variation in the return on equity is explained by changes in
independent variables. For variables presented with their original measures and in logarithmic values,
the remaining 38% and 7%, respectively, of the ROE variations are explained by the error term, that is,
5 When the sign associated with coefficient is negative, the relationship is negative. Otherwise, the relationship is
positive.
26
by factors like omitted variables, measurement errors or others that could not be included in the model.
The F-test results for both normal and logarithmic values are statistical significant for a significance
level of 1% which indicates that the independent variables jointly justify the variation on the return to
equity. However, as explained before the Ramsey test indicates the existence of omitted variables if
the original values are used. The R-squared and Root MSE values indicate that the results of
logarithmic model (model 2) are better.
According to the results of regression analysis with normal values only CUR and IB (current, interest
burden) are not statistically significant. The results of regression analysis with logarithmic values
indicate that only CUR is not statistically significant. Therefore, a conclusion cannot be withdrawn
regarding the influence of these variables on return on equity. All the other six variables for the first
model and seven for the second model are statistically significant and present the expected sign
between them and the return on equity.
The results of first model point out that asset turnover has a coefficient of 11.23 which means that 1€
change in asset turnover translates into 11.23% change in return on equity. Financial leverage has a
value of 9.02 which signifies that 1€ change in financial leverage translates into 9% change return on
equity. Nevertheless, the second model presents different results.
According to the results, tax burden, interest burden, operating margin, asset turnover, financial
leverage ratios (extended DuPont components) describe changes occurring in return on equity. The
coefficients of the second model for TB, IB, OM, AT and FL are 0.94, 0.95, 0.89, 0.90 and 0.89
respectively, which means that 1% change in TB, IB, OM, AT and FL translates into 0.94 %, 0.95 %,
0.89 %, 0.90 % and 0.89 % change in return on equity, respectively. The model asserts that TB, IB,
OM, AT and FL (extended DuPont components) are the most powerful drivers of ROE.
27
Table 5. Results of the OLS regression analysis for all companies, using original measurement units and logarithmic values
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016
Ramsey test: F (3, 75) = 51.19 *** Ramsey test: F (3, 75) = 0.44
Notes: * means that the coefficient presents a 10% level of significance; ** means that the coefficient presents a 5% level of significance; *** means that the coefficient presents a 1%
Note: All the values are presented in the same unit of measurment of the variables with the exception of the coefficient of variation that is
presented in %
29
skewness values indicating that most of the companies again present values near to the minimum
value. On the other hand, interest burden has a negative skewness meaning most of the companies
present values near to the maximum value. Kurtosis values of the second group are relatively lower
compared to the first group of variables meaning that the distribution of variables of the former are less
peaked (more dispersed) than the distributions of variables of the latter.
Since Nasdaq-100 NDX presents largest companies in the world, companies operating in the same
sector (Technology) have similar size and characteristics. It can be observed that variables are
characterized by significantly less dispersion compared to the values of table 3. As the results of
descriptive statistics in table 3 are for all companies from various industries, the variables exhibit
notable dispersion around their respective means. This can be seen by comparing the coefficient of
variations of table 6 and table 3.
3.2.2 OLS regression analysis results
The results of the Pearson correlation coefficient with statistical significance, for both original values
and logarithmic values, are presented in Table 7. Results with no statistical significance are not
presented. The number of observations available for each variable is presented in brackets. According
to table 7 operating margin and price to book indicators are strongly correlated with return on equity.
The asset turnover is also correlated with return on equity with 10% level of significance. The
remaining indicators are not correlated with return on equity.
The results of OLS regression analysis for technology sample are presented in Table 8. As shown in
Table 8, both models show high R-squared values indicating that variations occurring in the
independent variables effectively explain variations occurring in the dependent variable. The first model
presents a R-squared equal to 0,8621 for original values and the second model presents higher R-
squared value of 0.9847. The results with logarithmic values are better due to higher R-squared value.
The F-test results for both normal and logarithmic values are statistically significant for a significance
level of 1%. The Root MSE is much lower for regression model using logarithmic values, indicating
much higher accuracy compared to the model with normal values.
According to the results of regression analysis with normal values only PE and CUR (price to earnings,
current ratio) are not statistically significant. The results of regression analysis with logarithmic values
indicate that only CUR is not statistically significant. Therefore, a conclusion cannot be withdrawn
regarding the influence of these variables on return on equity. All the other six variables for the first
model and seven for the second model are statistically significant and present the expected sign
between them and the return on equity.
The results of first model point out that asset turnover has a coefficient of 22.22 which means that 1€
change in asset turnover translates into 22.22% change in return on equity. Whereas, financial
30
leverage has a value of 6.95 which signifies that 1€ change in financial leverage translates into 6.95%
change in return on equity. Nevertheless, the second model presents different results.
According to the results, tax burden, interest burden, operating margin, asset turnover, financial
leverage ratios (extended DuPont components) have the most impact on return on equity. The
coefficients of the second model for TB, IB, OM, AT and FL are 0.98, 0.54, 0.87, 0.87 and 0.71
respectively, which means that 1% change in TB, IB, OM, AT and FL translates into 0.98 %, 0.54 %,
0.87 %, 0.87 % and 0.71 % change in return on equity, respectively. The model asserts that TB, IB,
OM, AT and FL (extended DuPont components) are the most powerful drivers of ROE.
31
Table 7. Results of Pearson correlation coefficient between the independent variables and the return on equity
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016
TB IB OM AT FL PE PB CUR
Normal ROE - -0.542*
(32)
0.334
(32)
0.334
(32)-
0.425*
(32)-
Logarithmic ROE - -0.542*
(32)
0.326
(32)- -
0.471*
(32)-
Variables
Note: (*) means that the coefficient presents 5% level of significance. Values with no stars indicate 10% level of signficance. (-) indicates no relationship between the dependent
and independent variable in question. Number of obervations available for analysis arepresented in brackets.
32
Table 8. Results of the OLS regression analysis for the technology sector sample, using original measurement units and logarithmic values
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016
Model 1: Normal values Model 2: Logarithmic values
p-value p-value
n = 32 n=32
Ramsey test: F (3, 20) = 0.28 Ramsey test: F (3, 20) = 0.19
Notes: * means that the coefficient presents a 10% level of significance; ** means that the coefficient presents a 5% level of significance; *** means that the coefficient presents a 1%
OM 34 18.39 -28.56 68.00 96.55 15.90 0.86 0.34 5.89
AT 34 1.11 0.22 3.55 3.34 0.85 0.76 1.10 3.48
FL 34 2.79 1.15 11.97 10.82 1.91 0.68 3.49 17.12
PE 34 28.69 4.58 72.15 67.58 14.69 0.51 0.89 3.83
PB 34 6.01 1.61 14.31 12.70 3.74 0.62 0.88 2.61
CUR 34 2.15 0.14 6.97 6.82 1.54 0.71 1.21 4.10
Note: All the values are presented in the same unit of measurment of the variables with the exception of the coefficient of variation
that is presented in %
34
minimum. On the other hand, interest burden has a negative skewness meaning that most of the
companies in the sample present values nearest to the maximum. Kurtosis values of the second group
are relatively lower compared to the first group of variables meaning that the distribution of variables of
the former are less peaked (more dispersed) than the distributions of variables of the latter.
Since all the companies operate in Consumer sector, it can be observed that variables are
characterized by significantly less dispersion compared to the values of table 3 as was the case for
Technology sector. This can be seen by comparing the coefficient of variations of table 9 and table 3.
3.3.2 OLS regression analysis results
The results of the Pearson correlation coefficient with statistical significance, for both original values
and logarithmic values, are presented in Table 10. Results with no statistical significance are not
presented. The number of observations available for each variable is presented in brackets.
According to table 10 tax burden and price to book and price to earnings indicators have the highest
impact on return on equity. Whereas, interest burden, operating margin and financial leverage have
relatively low impact on return on equity. It is noteworthy that price to earnings ratio has a negative
association with return on equity. According to table 10, asset turnover and current ratios have no
impact on return on equity.
35
Table 10. Results of Pearson correlation coefficient between the independent variables and the return on equity
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016
TB IB OM AT FL PE PB CUR
Normal ROE0.578*
(34)-
0.410*
(34)-
0.718*
(34)
-0.406*
(34)- -
Logarithmic ROE0.672*
(33)
0.327
(33)
0.293
(33)-
0.404*
(33)
-0.548*
(33)
0.525*
(33)-
Variables
Note: (*) means that the coefficient presents 5% level of significance. Values with no stars indicate 10% level of signficance. (-) indicates no relationship between the dependent and
independent variable in question. Number of obervations available for analysis arepresented in brackets.
36
The results of OLS regression analysis for technology sample are presented in Table 11. As shown in
Table 11, both models show high R-squared values indicating that variations occurring in the
independent variables effectively explain variations occurring in the dependent variable. The first model
presents a R-squared equal to 0,9022 for original values and the second model presents higher R-
squared value of 0.9934. Obviously, second model with logarithmic values is better due to higher R-
squared value. The F-test results for both normal and logarithmic values are statistically significant for
a significance level of 1%. The Root MSE is much lower for regression model using logarithmic values,
indicating much higher accuracy compared to the model with normal values.
It is noteworthy that the regression analysis with original values presents a Ramsey values statistically
significant which indicates the existence of omitted variables, that is, that more variables should be
added to the model to make the analysis more accurate.
According to the results of regression analysis with normal values TB, IB, PE, PB, CUR (tax burden,
interest burden, price to earnings, price to book and current) are not statistically significant. The results
of regression analysis with logarithmic values indicate that PE, PB, CUR are not statistically significant.
Therefore, a conclusion cannot be withdrawn regarding the influence of these variables on return on
equity. OM, AT and FL (DuPont components) variables for the first model and TB, IB, OM, AT, FL
(extended DuPont components) for the second model are statistically significant and present the
expected sign between them and the return on equity.
The results of first model point out that operating margin has a coefficient of 1.47 which means that 1%
change in operating margin results in 1.47% change in return on equity. Asset turnover has a
coefficient of 18.48 which means that 1€ change in asset turnover translates into 18.48% change in
return on equity. Financial leverage has a value of 14.96 which signifies that 1€ change in financial
leverage translates into 14.96% change in return on equity. Nevertheless, the second model presents
different results.
According to the results, tax burden, interest burden, operating margin, asset turnover, financial
leverage ratios (extended DuPont components) significantly affect return on equity. The coefficients of
the second model for TB, IB, OM, AT and FL are 0.91, 1.05, 0.87, 0.88 and 0.93 respectively, which
means that 1% change in TB, IB, OM and 1€ change in AT and FL translates into 0.91 %, 1.05 %, 0.87
%, 0.88 % and 0.93 % change in return on equity, respectively. The model asserts that TB, IB, OM, AT
and FL (extended DuPont components) are the most powerful drivers of ROE which was the case both
in global and technology samples.
37
Table 11. Results of the OLS regression analysis for the consumer sector sample, using original measurement units and logarithmic values
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016
Model 1: Normal values Model 2: Logarithmic values
p-value p-value
n = 34 n=33
Ramsey test: F (3, 22) = 98.97*** Ramsey test: F (3, 21) = 1.82
Notes: * means that the coefficient presents a 10% level of significance; ** means that the coefficient presents a 5% level of significance; *** means that the coefficient presents a 1%
(extended DuPont components) significantly affect return on equity.
The coefficients of the second model for TB, IB, OM, AT and FL are 1, 1.01, 1.02, 1 and 1.02
respectively, which means that 1% change in TB, IB, OM and 1% change in AT and FL translates into
1 %, 1.01 %, 1.02 %, 1 % and 1.02 % change in return on equity, respectively. The model asserts that
TB, IB, OM, AT and FL (extended DuPont components) are the most powerful drivers of ROE.
As shown in table 5, 8, 11 and 14 TB, IM, OM, AT and FL are statistically significant in every sample
which is one of the most important findings of this study. The coefficients of these variables are
relatively similar in each sample which highlights the importance of extended DuPont model as a
determinant of return on equity. Those variables almost equally affect return on equity in each sample.
40
41
Table 13. Results of Pearson correlation coefficient between the independent variables and the return on equity
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016
TB IB OM AT FL PE PB CUR
Normal ROE - -0.433*
(24)- -
-0.416*
(24)- -
Logarithmic ROE0.749*
(23)
0.635*
(22)- - -
-0.590*
(23)- -
Variables
Note: (*) means that the coefficient presents 5% level of significance. Values with no stars indicate 10% level of signficance. (-) indicates no relationship between the dependent and
independent variable in question. Number of obervations available for analysis arepresented in brackets.
42
Table 14. Results of the OLS regression analysis for the other sector sample, using original measurement units and logarithmic values
Source: Author’s calculations using Bloomberg data retrieved on 23.02.2016