USING LINEAR REGRESSION IN THE ANALYSIS OF FINANCIAL-ECONOMIC PERFORMANCES

Prof. Lucian BUŞE

Assist. Mirela GANEA
Lect. Daniel CÎRCIUMARU
University of Craiova
Faculty of Economics and Business Administration
Craiova, Romania

Abstract: Seeing that the steady pace of development and evolution of

modern society has left its mark on the phenomena in all fields of activity,
including in the field of economics, the analysis of financial-economic
performances, as a major preoccupation of firms, has made a huge
qualitative leap, by shifting the focus to the exploitation of databases
(through adequate techniques) and the thorough interpretation of results.
Starting from the idea that in economics, as well as in other sciences,
anything has the tendency to depend on anything else, in this paper we
intended to develop an econometrical model capable of expressing the
relation between the economic rate of return- as a fundamental indicator of
expressing the firm’s financial-economic performance- and its determinant
factors. The multiple linear regression model has been developed through
the analysis of data from 30 Romanian companies in the processing
industry and by using the specific SPSS instruments, version 16.0.

JEL classification: C01, G30

Key words: multiple linear regression model, financial-economic performances,

economic rate of return

1. The analysis of performances and the multiple linear regression

Taking into consideration the rich informational society at the present time, the
analysis of financial-economic performances can be seen as a real challenge. The much
easier and more rapid access to different financial and nonfinancial information and
data of economic entities, determines the analysts to be less interested in the analysis
and description of only one economic variable, but rather more interested in the
analysis and description of relations between two or more economic variables.
In economic theory as well as in any other science, the idea that anything has a
tendency to depend on one or more causes is plausible. Our research is also centered
round this idea, and its main goal is to develop an econometrical model to examine the
relation existent between different indicators resulted from the firm’s financial
statement and its performance, expressed through the economic rate of return. In other
words, the analysis will be centered round the explanation of the evolution of rate of
return depending on the evolution of main financial indicators available at the level of
firm.
In order to carry out such an analysis we have used the multiple linear
regression method. Briefly speaking, the goal of the multiple linear regression is to
point out the relation between a dependent variable (explained, endogenous or
resultative) and a great deal of independent variables (explanatory, factorial,
exogenous). With the help of multiple linear regression we can determine to what
extent a part of the total variation of the dependent variable is influenced by the
variation of the independent variables.
The general form of the equation of multiple linear regression is:
Yi = β 0 + β1 ⋅ X i ,1 + β 2 ⋅ X i , 2 + ..... + β k ⋅ X i , k + ε
and:
i = 1,2,…..,n are the observations from the sample;
Yi = observation i of the dependent variable;
X1, X2,….,Xk = independent variables;
β0 = constant(free term of equation);
β1,…., β k = coefficients of independent variables;
ε = error term of equation.
Developing a multiple linear regression model involved the analysis of data
from 30 Romanian companies in the processing industry, companies which are
presented in the table in the annex no. 1.
The data analyzed refer to the financial statements of the fiscal year 2008 and
were obtained using the access to the online database www.amadeus2.bvdep.com .
We took into consideration the companies in the same industry because there
are factors specific to each filed of activity and we wanted to avoid adding dummy
variables for the sector.
As a dependent variable we used the economic rate of return, because we
consider that it synthesizes best a company’s financial-economic performance.
Independent variables were selected by applying the statistical tests
corresponding to the correlation analysis (dispersion diagram or Scatterplot and
correlation coefficient) to a number of 20 potential variables. In other words, following
the testing of the correlation between the economic rate of return and each of the 20
indicators of performance available at the level of firms (considered to be possible
independent variables), there were identified and retained 6 independent variables as
being linearly correlated with the economic rate of return.
The values registered by the dependent variable and the 6 independent
variables, at the level of each of the 30 firms from the sample, are presented in the
following table:

Table no. 1: The levels of the regression model variables

Economic Gross Operating

Good Cash Flow EBIT Total debt
rate of operating gross
No money /Turnover margin turnover
return surplus margin rate

Y X1 X2 X3 X4 X5 X6
1 0,056 0,442 10383120 0,087 0,057 0,036 3,513
2 0,076 0,022 9869544 0,070 0,027 0,042 2,565
3 0,227 1,111 26685671 0,249 0,136 0,152 5,201
4 0,041 0,003 12138277 0,110 0,024 0,076 1,400
5 0,063 0,060 12032959 0,096 0,056 0,037 1,704
6 0,077 0,004 10798008 0,083 0,012 0,080 0,905
7 0,032 0,007 10696810 0,112 0,052 0,047 1,601
 8 0,281 2,452 23165418 0,212 0,194 0,165 9,502
9 0,047 0,006 5421437 0,058 0,027 0,025 1,807
10 0,174 0,043 13324530 0,132 0,063 0,117 1,937
11 0,127 0,062 14044440 0,154 0,039 0,105 1,355
12 0,110 0,006 13390931 0,149 0,058 0,096 1,344
13 0,060 0,040 1492526 0,024 0,007 0,012 3,051
14 0,245 0,025 20313159 0,227 0,046 0,179 1,358
15 0,066 0,030 29643236 0,542 0,246 0,287 1,357
16 0,047 0,044 3735283 0,053 0,023 0,017 3,479
17 0,122 0,087 13251207 0,153 0,083 0,094 3,600
18 0,004 0,095 3989975 0,046 0,057 0,004 2,723
19 0,135 0,177 14794181 0,222 0,034 0,136 0,703
20 0,163 0,165 11972850 0,144 0,087 0,115 2,385
21 0,060 0,044 3873616 0,082 0,023 0,030 1,400
22 0,066 0,009 9019110 0,108 0,059 0,062 1,338
23 0,116 0,164 3811847 0,056 0,003 0,043 1,878
24 0,416 0,347 14421802 0,191 0,086 0,143 6,435
25 0,090 0,004 14018015 0,207 0,110 0,073 1,116
26 0,150 0,074 11856849 0,159 0,061 0,099 1,243
27 0,136 0,261 8376274 0,194 0,064 0,078 1,706
28 0,137 0,025 15355836 0,279 0,092 0,132 0,795
29 0,229 0,058 19331352 0,287 0,180 0,139 2,343
30 0,120 0,346 4741264 0,073 0,039 0,046 3,295

The variables presented in table no. 1 were determined for each firm, on the
basis of the following calculation relations:
◊ Economic rate of return (Rec):
Earnings before int erest and tax ( EBIT )
Rec = , (1)
Average balance of total assets ( At )
and:
EBIT = Gross profit + int erest exp enses , (2)
The average balance of total assets ( At ), was determined as an average of the
sums reported at the beginning and at the end of the financial period 2008.
◊ Good money (Li):
Liquid assets
Li = (3)
Current debts
◊ Gross operating surplus (GOS):
GOS = VA + Se − It − Cp , (4)
and:
VA = Added value;
Se = Operating subsidies;
It = Value of rates and taxes owed (without tax profit and VAT);
Cp = Personnel expenditures (gross income and state budget contributions).
◊ Operating gross margin rate (Rmb):
 Gross operating surplus (GOS )
R mb = (5)
Turnover (T .O.)
◊ Ratio between the cash flow and the turnover (CF/CA):
Operational cash flow(CFO )
CF / CA = , (6)
Turnover (T .O.)
◊ EBIT Margin (MEBIT):
Earnings before int erest and tax ( EBIT )
M EBIT = (7)
Total income(Vt )
◊ Total debt turnover (RDt):
Turnover (T .O.)
R Dt = (8)
Total debt ( Dt )

We don’t pretend that the list of the above-mentioned variables is exhaustive,

because the economic rate of return indicator can be influenced also by other factors
beside the mentioned ones-factors that we will group into a stochastic variable called
error.
We established and tested the following six hypotheses on this stochastic
variable (error variable), but also on the form of the model and its other components:
1. The link between the dependent variable and the independent variables is
linear;
2. Independent variables are random. Also, there is no linear relation between
independent variables included in regression.
3. The expected value of the error term, ε, is zero, E(ε)= 0;
4. The variation of the error term , ε, is the same for all observations,
E (ε 2 ) = σ ε2 ;
5. The error term, ε, is not correlated between observations E(εt * εs)= 0, s ≠ t;
For the testing of the availability of hypotheses on which the regression model
is based, as well as for the estimation of the model’s parameters, the testing of these
parameters and the validation of the regression model, we used different statistical tests
offered by the instruments of the SPSS software (Statistical Package for the Social
Sciences), as we will present further on.

2. Developing a multiple linear regression model

Taking into account the above-mentioned information, we developed the
multiple linear regression model by using the specific SPSS instruments, version 16.0.
The stages of the multiple linear regression model and the results registered by
using SPSS, are:

A. Determining and testing the correlation ratio

In order to determine and test the correlation ratio between the dependent
variable and each independent variable we calculated the Pearson Coefficient and the
Statistic-t test and the probability associated to it, for each combination of variables- the
obtained results being presented in the following table:
 Table no. 2: Partial correlation matrix

Table no. 2 is structured around three parts, in accordance with the significance
of data, as it follows:
a) the first part encompasses the values of the Pearson correlation coefficients;
b) the second part encompasses the values of the significance threshold (Sig.)
corresponding to the testing of the significance of values registered by Pearson
coefficients;
c) the third part points out the number of observations considered (n=30 in our
case).
The level of Pearson coefficient offers information on the meaning and
intensity of the correlation between the analyzed variables. This coefficient can take
value within the interval [-1, 1].
When appreciating the intensity of the correlations between variables we took
into consideration also the thresholds of significance (Sig.), considering a minimal
significance threshold of 0.05, below which coefficients are considered significant from
a statistical point of view. In other words, Sig. values below 0.05 for each calculated
coefficient suggest that there is a significant correlation between the analyzed variables.

B. Selecting independent variables in the model

In order to come up with the best combination of independent variables which
explain the variation of the economic yield, we used the backward method, the obtained
results being exposed in the following table:
 Table no. 3: Selecting independent variables

As we can notice in table no. 3, in the first stage, we introduced all considered
variables in the model, while in the following stages we eliminated one after the other,
along the line of the lowest influence on the economic yield the following independent
variables:
● Gross operating surplus;
● Operating gross margin rate;
● Good money.
C. Estimating the model’s parameters. Testing the significance of the model’s
parameters
We will carry the analysis of the model’s parameters on the basis of the results
in the following tables:
 Table no. 4: Correlation coefficient (R)

Table no.4 contains the values of the R correlation coefficient at the level of the
whole group of variables which form the regression models, calculated distinctly in
each stage of the backward method of optimal assessment of linear regression.
As it concerns our study, due to the value calculated for the R correlation
coefficient R=0.796, we can state that the independent variables detected within model
no. 4 (Total debt turnover, ratio between the Cash flow and the turnover, and EBIT
margin) are those which explain best the evolution of the dependent variable.
The same conclusion can be obtained by analyzing also the ANOVA table:

Table no. 5: ANOVA Table

Hence, through the ANOVA test the threshold of significance is calculated for
each model, noticing that the registered values are below the significance threshold
(0.05), which means that the independent variables explain the variation of the
dependent variable.
Estimating the parameters of the regression model and testing their significance
involves analyzing the results in table no. 6: Regression model parameters. In this table,
in the first part we can find the coefficients of the regression model, standard errors, t-
test statistic value for each coefficient, as well as the value of the threshold of
significance (Sig.).
As it is about a multiple regression, in the second part of the table, the colinearity
statistics, tolerance and variation inflation factor (VIF) are specified, as it can be
noticed:
Table no.6: Regression model parameters:

The analysis of the results in columns t and Sig, for the 4 different models
confirms us the conclusion according to which the Ratio between the cash flow and the
turnover, the EBIT margin and the Total debt turnover are variables which estimate
best the evolution of economic efficiency. This conclusion is fostered by the values
below 0.05 of the significance threshold that corresponds to each of these independent
variables (Sig. 0.037 for the ratio between the cash flow and the turnover, and 0.00 for
the other two independent variables), but also by the tolerance values for these three
independent variables and VIF values.
Taking into consideration these aspects we will retain only the values estimated
for the coefficients of no. 4 model in the previous table. Thus, the estimated values for
the three parameters of the model and their significances are:
 a) value of -0.588 for the ratio between the cash flow and the turnover, which
means that when the indicator of the ratio between the cash flow and the turnover
increases by one unit, while the other independent variables remain constant, the
economic rate of return decreases on average by 0.588 units;
b) value of 1.085 for EBIT margin, which means that when EBIT margin
increases by one unit, while the other characteristics remain constant, the economic rate
of return increases by 1.085 units;
c) value of 0.029 for the Total debt turnover, that is, when the Total debt
turnover increases by one unit, while the other independent variables remain constant,
the economic rate of return increases by 0.029 units.
d) The free term has the value of -0.006 and does not have an economic
interpretation.

D. Multicolinearity
The diagnosis of colinearity involves the analysis of results in the following
table:

Table no. 7: Diagnosis of collinearity

The most important information transmitted by this table is represented by the

values of the condition indexes. Theoretically, an index higher than 15 shows that there
is a colinearity problem, while a value higher than 30 indicates serious colinearity
problems. In our study, we come across values of the condition index, above 15, for
models no. 1 and no. 2. Also from this point of view, model no. 4 represents the linear
combination of independent variables which explain best the evolution of economic
efficiency.
 Taking into account the stages so far, the model of multiple linear regression is:
Y = − 0 , 006 − 0 , 588 ⋅ X 1 + 1, 085 ⋅ X 2 + 0 , 029 ⋅ X 3 + ε
and:
Y = Economic rate of return
X1 = Ratio between the cash flow and the turnover
X2 = EBIT margin
X3 = Total debt turnover

Or:

Economic rate ofreturn = −0,006 − 0,588 ⋅ (ratio between cashflow and turnover ) +
+ 1,085 ⋅ ( EBIT m arg in) + 0,029 ⋅ (Total debt turnover ) + ε

The observance of hypotheses required by the regression analysis (errors are

distributed normally, at average 0; errors have a constant variation; errors are
independent of each other) was verified graphically using P-P plot and Scatterplot
diagrams.
Figures no. 1 and no. 2 show how these hypotheses are respected:

Figure 1: Normal P-P Plot Diagram

 Figure 2: Scatterplot Diagram

ACKNOWLEDGMENTS

This work was supported by CNCSIS –UEFISCSU, project number 299/2007, PNII – IDEI.

REFERENCES :
1. Andrei, T. Statistică şi econometrie, Ed. Economică, Bucureşti, 2003
2. Helfert, E. Tehnici de analiză financiară, Ed. BMT, Bucureşti, 2006
3. Jaba, E., Analiza statistică cu SPSS sub Windows, Editura Polirom, Iaşi, 2004
Grama, A.
4. Spircu, L., Econometrie, Ed. Pro Universitaria, Bucureşti, 2007
Ciumara, R.
ANNEX NO. 1

Table no. 1: Analyzed companies

Processing industry
No. Name of firm
Code Name
1 TABCO-CAMPOFRIO SA 10 Food industry
INDUSTRIALIZAREA LAPTELUI Food industry
2 MURES SA 10
3 CARNIPROD SRL 10 Food industry
4 GALMOPAN SA 10 Food industry
5 AGRO COMPANY SRL 10 Food industry
6 ZAHARUL LIESTI SA 10 Food industry
7 PAN GROUP SA 10 Food industry
Clothing manufacturing
8 PRODUCTIE ZARAH MODEN SRL 14
9 PRINCIPAL COMPANY SA 10 Food industry
Clothing manufacturing
10 OZTASAR SRL 14
11 DORNA SA 10 Food industry
12 CARREMAN ROMANIA SRL 13 Textile manufacturing
13 MARIBO PRODCARN SRL 10 Food industry
14 SUPREME CHOCOLAT SRL 10 Food industry
15 IASITEX SA 13 Textile manufacturing
16 LUCA SRL 10 Food industry
17 MEDA PROD 98 SA 10 Food industry
18 YARNEA SRL 13 Textile manufacturing
19 ZAHARUL SA 10 Food industry
20 NEGRO 2000 SRL 10 Food industry
21 AVI INSTANT SRL 10 Food industry
22 C+C SA 10 Food industry
Clothing manufacturing
23 STEILMANN ROMANIA SRL 14
24 MARCEL SRL 10 Food industry
25 CARMOLIMP SRL 10 Food industry
Clothing manufacturing
26 CORSSA SRL 14
27 TIP TOP FOOD INDUSTRY SRL 10 Food industry
ROULEAU-GUICHARD ROUMANIE Clothing manufacturing
28 SRL 14
29 BETTY ICE SRL 10 Food industry
30 LEFRUMARIN SRL 10 Food industry