BEC 4340: Econometrics

Year IV – Semester I
2020

Mini Research Project on

The Furniture Industry in Moratumulla

By
Group No. 01

Department of Business Economics

Faculty of Management Studies and Commerce
University of Sri Jayewardenepura
 List of Group Members
Department:

Student Name MC CPM Signature

1. P.Y.D. Peiris 83873 16343

2. M.W.T. Uddamika 84155 16350

3. C.R. Munindradasa 83819 15861

4. M.A.N.P. Manasinghe 83801 16328

5. K.D.N.N. Gunasekara 83551 16255

6. E.A.K. Arachchige 83359 16030

7. W.D.N. Deshapriya 83450 16337

8. A.M.A.N. Sathsara 84055 16178

9. B.R.P. Sewwandi 83394 16205

10. M.A.D.S. Amarasiri 83352 16341

Group Leader’s Name: M.A.D.S. Amarasiri

Group Leader’s Contact No.: 071-4199507

Table of Content

Introduction ........................................................................................................................5
Methodology .......................................................................................................................7
Data ............................................................................................................................................. 7
Model Specification .................................................................................................................... 7
Variable Selection ....................................................................................................................... 7
Quantitative Data .................................................................................................................... 8
Qualitative Data ...................................................................................................................... 9
Hypothesis................................................................................................................................. 10
Based on Z Values ................................................................................................................ 10
Based on P-Values ................................................................................................................ 13
Confidence Intervals for True Coefficients .......................................................................... 15
Analysis & Findings .........................................................................................................17
Normality Assumption .............................................................................................................. 17
Graphical Tests ..................................................................................................................... 17
Statistical Tests ..................................................................................................................... 19
Interpretations of Coefficients .................................................................................................. 21
Statistical Significance .............................................................................................................. 23
t- Test ........................................................................................................................................ 23
R2 and adjusted R2 ..................................................................................................................... 24
Root-Mean-Square Error (RMSE) ............................................................................................ 25
Standard Errors ......................................................................................................................... 26
Overall Significance of the Model ............................................................................................ 26
Probability F Test .................................................................................................................. 26
Independent Variables Analysis ............................................................................................... 27
Checking for Errors................................................................................................................... 29
Multicollinearity ................................................................................................................... 29
Heteroscedasticity ................................................................................................................. 30
Specification Bias ................................................................................................................. 31
Testing Normality Assumption ............................................................................................. 32
Conclusion ........................................................................................................................34

iii
Recommendations ............................................................................................................36
Limitations ........................................................................................................................36
Bibliography .....................................................................................................................37
Appendix ...........................................................................................................................38

Table of Tables
Table 1: Data Set .......................................................................................................................... 10
Table 2 : Critical Values and t-Statistics ...................................................................................... 11
Table 3 : p-Values ......................................................................................................................... 14
Table 4 : Confidence Intervals for Coefficients ............................................................................ 15
Table 5 : Regression Results ......................................................................................................... 17
Table 6: Anderson-Darlin Test ..................................................................................................... 20
Table 7: Coefficients ..................................................................................................................... 21
Table 8 : Multicollinearity Test .................................................................................................... 29

Table of Charts
Chart 1: Histogram - Data Normality Test................................................................................... 18
Chart 2: Probability Plot- Data Normality Test ........................................................................... 19
Chart 3 : Residuals and Fitted Values .......................................................................................... 30

iv
 Introduction

Furniture industry is one of the major industries in Sri Lanka and according to (Padmasiri, 2012)

it has almost above 10000 furniture plants and wood working firms across the country. It has been

contributing to the Sri Lankan economy through creating job opportunities as well as to the growth

of gross domestic product (GDP). According to (Padmasiri, 2012) Since there is an increasing

demand for high quality wooden furniture in local market due growing population and the per

capita income the furniture industry plays a vital role in the manufacturing sector.

As highlighted by (Amarasekara, 2012) majority of furniture firms are in Moratuwa which is well-

known in the country for their carpentry. Also, most of wooden activities and products are taken

place within city of Moratuwa. Majority of these firms are small in scale and owned by private

sector. According to the same report this furniture industry is the main economic activity of the

people in Moratuwa area. Moratuwa has the concentrated furniture activities in the country

(Dasanayaka, 2011) because of the comparative advantage they have gained by the easy supply of

raw materials needed in producing furniture and also in marketing due to its popularity among the

people in the country for their products.

Most commonly used timber species by these firms is Teak and other timber species include

Mahogany and jack. They use mainly consumer preference, traditional designs by carpenters and

designs from foreign catalogues for designing of furniture (H.S.C.Perera, 2009) and accordingly

Teak Cabinets, Teak Elmira, and Teak Chairs are the main manufactured items.

Main issues and challenges faced by firms in furniture industry in Sri Lanka are financial and credit

relates issues, skill labor mismatching, technological problems, resource management issues and

irresponsibility of state sector supportive agencies (Dasanayaka, 2011) Therefore it is compelling

5
to concentrate on these problems and identify the factors that can improve and make a significant

growth in the furniture industry of Sri Lanka (Shantha, 2013)

6
 Methodology

Data

Using primary data sources such as surveys and direct observations, primary data was extracted to

conduct this research. Structured interviews were conducted using questionnaires provided by the

Department of Economics. These questionnaires were interviewer administrated questionnaires

which included both open and closed questions.

Our sample consists of 20 observations selected to cover the population of all the furniture firms

exist in Moratumulla area. All the interviews were conducted as one-to-one interviews in

collecting data.

Model Specification

The research depends on more than one independent variable to analyze the growth of the furniture

industry in Moratumulla. Since it follows a non-linear relationship (depending on several

independent variables) the research is analyzed using multiple regression model.

Y = B0 + (B1*x1) + (B2* x2) + … + (Bn*xn) + Ui

Variable Selection

We have taken both quantitative and qualitative data to analyze the growth of furniture industry in

Moratumulla. As for the dependent variable of the model we have taken the production value and

as for independent variables we have taken years of experience in the industry, number of laborers,

family inheritance, quantity of timber used, electricity payment and the training experience. All

these variables are measured on a weekly basis.

7
Quantitative Data

Quantitative data includes all the variables we can identify through a numerical measurement.

Including the dependent variable and four independent variables, the model is based on five

quantitative variables.

• Dependent variable

1. Production value

Value of the production is taken from multiplying the number of items produced by its unit

price.

• Independent variables

1. Years of experience

Owner of each firm had experience in the industry which we have measured on yearly

basis.

2. Number of laborers

We have taken the number of laborers who are currently working in the firm.

Quantity of timber used

The quantity of timber used is also taken on a weekly basis and measured by cubic feet.

3. Electricity payment

Electricity cost that firm incurs was given monthly and we have divided it by four to get

the payment for a week.

8
Qualitative Data

All the non-numerical data taken for the research represents the qualitative data included in the

model. We have selected two independent variables such as family inheritance and training

experience as our qualitative variables. These variables are analyzed through a dummy variable

regression.

• Independent variables

1. Family inheritance

This variable concern whether the business was newly acquired or inherited from

family by the current owner of the firm.

Dummy variables- The business is inherited from family,

Yes - 1
No - 0
2. Training experience

This variable state whether the current workers has any training experience regarding

the industry of furniture.

Dummy variables- Gained training experience,

Yes - 1
No - 0

9
Table 1: Data Set

Hypothesis

Based on Z Values

H0: β1= 0

(There is no significant relationship between independent variable and the dependent variable.)

H1: β1≠ 0

(There is a significant relationship between independent variable and the dependent variable.)

Decision rule

Reject H0 if -2.145>Z value >2.145

Do not reject H0 if -2.145 <Z value <2.145

10
Table 2 : Critical Values and t-Statistics

Critical value at
t-
Explanatory the 0.05 level of
Hypothesis statisti Decision
variables significance for 14
c
df (Two tailed test)

Intercept Economically
- 0.28 2.145
meaningless

Years of H0: There is no any statistically There is enough

experience significant relationship between sample evidence to

years of experience and the say that there is a

revenue per a week. statistically

significant

H1: There is a statistically 5.72 2.145 relationship

significant relationship between between years of

years of experience and the experience and the

revenue per a week. revenue per a week

at 95% confidence

level.

No of H0: There is no any statistically There is enough

Laborers significant relationship between sample evidence to

3.27 2.145
number of laborers and the say that there is a

revenue per a week. statistically

11
 significant

H1: There is a statistically relationship

significant relationship between between number

number of laborers and the of laborers and the

revenue per a week. revenue per a week

at 95% confidence

level.

Family H0: There is no any statistically There is enough

Inherent significant relationship between sample evidence to

family inherent and the revenue say that there is no

per a week. any statistically

H1: There is a statistically significant

significant relationship between -0.35 -2.145 relationship

family inherent and the revenue between family

per a week. inherent and the

revenue per a week

at 95% confidence

level.

Timber H0: There is no any statistically There is enough

Quantity significant relationship between sample evidence to

0.98 2.145
timber quantity used per a week say that there is no

and the revenue per a week. any statistically

12
 H1: There is a statistically significant

significant relationship between relationship

timber quantity used per a week between timber

and the revenue per a week. quantity used per a

week and the

revenue per a week

at 95% confidence

level.

Training H0: There is no any statistically There is enough

significant relationship between sample evidence to

training and the revenue per a say that there is no

week. any statistically

H1: There is a statistically significant

1.65 2.145
significant relationship between relationship

training and the revenue per a between training

week. and the revenue

per a week at 95%

confidence level.

Based on P-Values

H0: β1= 0

(There is no significant relationship between independent variable and the dependent variable.)

H1: β1≠ 0

13
(There is a significant relationship between independent variable and the dependent variable.)

Decision Rule

Reject H0 if P value <α

Do not reject H0 if p value >α

Table 3 : p-Values

Parameters are

Explanatory statistically different

P-Values Significance Level
Variables from zero at the 0.05

level of significance

Economically
Intercept 0.782 0.05
meaningless

Years of experience 0.000 0.05 Yes

No. of Laborers 0.006 0.05 Yes

Family Inherent 0.728 0.05 No

Timber Quantity 0.342 0.05 No

Training 0.121 0.05 No

14
Confidence Intervals for True Coefficients

Table 4 : Confidence Intervals for Coefficients

Explanatory
b±se*t.05/2 Lower limit Upper limit
Variables

Intercept 28,145.68± (99,657.62*2.145) -185,598.6 241,890

Years of experience 18,140.35± (31,68.967*2.145) 11,343.59 24,937.11

No. of Laborers 25,285.54± (7,738.076*2.145) 8,689.016 41,882.06

Family Inherent -24,562.76±(69,257.63*2.145) -173,105.6 123,980.1

Timber Quantity 1,459.454± (1,482.788*2.145) -1,720.811 4,639.719

Training 156,159.1± (94,653.97*2.145) -46,853.5 359,171.7

Here the confidence interval for the coefficient of years of experience at 95% confidence level is

from 11,343.59 to 24,937.11, which means the true β1 for years of experience lies in between that

range. Since both limits are positive figures it cannot be zero. So, we can say that there is a

significant relationship between revenue and experience.

Here the confidence interval for the coefficient of number of laborers at 95% confidence level is

from 8,689.016 to 41,882.06, which means the true β1 for number of laborers lies in between that

range. Since both limits are positive figures it cannot be zero. So, we can say that there is a

significant relationship between revenue and number of laborers.

However, confidence intervals for coefficients of family inherent, timber quantity, and training

have negative figures for lower limits, which means zero lies in between. Here we do not reject

15
the null hypothesis and thereby conclude that they do not have a significant relationship with the

dependent variable, revenue.

16
 Analysis & Findings

The following analysis and interpretations are based on the results obtained through Eviews.

Dependent Variable: REVENUE Table 5 : Regression Results

Method: Least Squares
Date: 02/15/20 Time: 14:34
Sample: 1 20
Included observations: 20

Variable Coefficient Std. Error t-Statistic Prob.

C 28145.68 99657.62 0.282424 0.7818

YEARS_OF_EXPERIENCE 18140.35 3168.967 5.724373 0.0001
NO_OF_LABOURERS 25285.54 7738.076 3.267678 0.0056
FAMILY_INHERENT -24562.76 69257.63 -0.354658 0.7281
TIMBER_QTY 1459.454 1482.788 0.984263 0.3417
TRAINING 156159.1 94653.97 1.649789 0.1212

R-squared 0.836664 Mean dependent var 587100.0

Adjusted R-squared 0.778330 S.D. dependent var 289735.8
S.E. of regression 136413.1 Akaike info criterion 26.72809
Sum squared resid 2.61E+11 Schwarz criterion 27.02681
Log likelihood -261.2809 Hannan-Quinn criter. 26.78640
F-statistic 14.34258 Durbin-Watson stat 2.478363
Prob(F-statistic) 0.000044

Normality Assumption

Based on the sample, we make inferences to generalize it to the population. So that estimators

chosen (β0, β1, β2, β3, β4, β5) should be best to represent the population. If we assign probability

distribution for these estimators, it will give all the possible values with their probabilities. For that

we assume the chosen random variables are normally distributed. To test the normality, we have

used following normality tests.

Graphical Tests

1. Histogram

17
6
Series: Residuals
Sample 1 20
5
Observations 20

4 Mean -8.44e-11
Median 7578.646
3 Maximum 216873.6
Minimum -245151.6
Std. Dev. 117096.3
2
Skewness -0.267104
Kurtosis 2.703284
1
Jarque-Bera 0.311182
0 Probability 0.855909
-200000 -100000 1 100001 200001

Chart 1: Histogram - Data Normality Test

Here the vertical axis represents frequency of residuals at each interval. The horizontal axis

represents intervals of the residuals.

The distribution can be considered reasonably symmetrical. (Exhibits a reasonable bell shape)

Therefore the residuals follow a normal distribution.

2. Normal Probability Plot

Here the residuals are plotted on and closer to the line, therefore we can conclude that the residuals

are normally distributed.

The error terms (residuals) are considered because it is statistically proven that if we make an

assumption about residuals that will be valid for the parameters as well. So we can conclude that

the estimators follow a normal distribution.

18
Even though these graphical tests are used in determining the normality of residuals, these

interpretations can be highly subjective according to the analysts’ perception about the shape of

the curve. So further analysis is required by using statistical tests.

300,000

200,000
Quantiles of Normal

100,000

-100,000

-200,000

-300,000
-300,000 -100,000 0 100,000 300,000

Quantiles of RESID
Chart 2: Probability Plot- Data Normality Test

Statistical Tests

1. Jarque – Bera Test

Hypothesis Testing:

H0: Residuals are normally distributed

H1: Residuals are not normally distributed

Significance Level - 0.05

19
 Decision rule:

Reject H0 if prob < 0.05

0.8559> 0.05

Decision: We fail to reject H0

Conclusion:

Under 5% level of significance, we have enough evidence to prove that the residuals are

normally distributed under Jarque - Bera Test.

2. Anderson-– Darling
Table 6: Anderson-Darlin Test

Em pirical Dis tribution Tes t for RESID

Hypothes is : Norm al
Date: 02/15/20 Tim e: 14:39
Sam ple: 1 20
Included obs ervations : 20

Method Value Adj. Value Probability

Lilliefors (D) 0.101572 NA > 0.1

Cram er-von Mis es (W2) 0.031203 0.031983 0.8192
Wats on (U2) 0.029629 0.030369 0.8139
Anders on-Darling (A2) 0.185649 0.193655 0.8936

Method: Maxim um Likelihood - d.f. corrected (Exact Solution)

Param eter Value Std. Error z-Statis tic Prob.

MU -9.02E-11 26183.53 -3.45E-15 1.0000

SIGMA 117096.3 18995.53 6.164414 0.0000

Log likelihood -261.2938 Mean dependent var. -8.44E-11

No. of Coefficients 2 S.D. dependent var. 117096.3

20
Hypothesis Testing:

H0: Residuals are normally distributed

H1: Residuals are not normally distributed

Significance Level - 0.05

Decision rule:

Reject H0 if prob < 0.05

0.8936 > 0.05

Decision: We fail to reject H0

Conclusion:

Under 5% level of significance, we have enough evidence to prove that the residuals are normally

distributed under Anderson – Darling Test.

Interpretations of Coefficients

Table 7: Coefficients

Explanatory Variables Estimated Coefficients Interpretation

Intercept Coefficient +28,145.68 This is the expected mean value of
revenue when all independent
variables equal to zero. So, this value
also represents revenue of all furniture
Manufacturers/Sellers who started the
business by his own without a family
inherent and who did not engage with
any training program.

21
Years of Experience +18,140.35 There is positive relationship between
years of experience of furniture
Manufacturers/Sellers and revenue
earned by them per a week. So, when
the experience increases by one year,
revenue earned per week also will
increase approximately by Rs.18,140.

No of Laborers +25285.54 There is a positive relationship

between revenue earned by furniture
manufacturers/sellers per week and
no. of laborers employed by them.
Each one unit increase in laborers is
associated with a Rs.25,285.54
increase in expected average revenue
of furniture manufacturers/sellers.

Family Inherent -24,562.76 There is a negative relationship

between revenue earned by furniture
manufacturers/sellers per week and
the inherent nature of the business.
The revenue of those who acquired
the business from family is less Rs.
24,562.76 than who started the
business by his own. It was
investigated during the field visit also
that people who acquired the business
from their family have been just
limited to traditional operations and
do not use technical advancements for
the business improvements.

22
Timber Quantity +1459.454 There is a positive relationship
between timber quantity (cubic feet)
used by furniture
manufacturers/sellers during a week
and average revenue earned by them
during a week. Each one cubic foot
increase in timber usage is associated
with a Rs.1,459.454 increase in
revenue.

Training +156,159.1 There is a positive relationship

between revenue earned by furniture
manufacturers/sellers during a week
and their involvement for the training
programs. Revenue earned by a
person who participated for any
training program is Rs. 156,159.1
more than those who do not have any
training related with the industry.

Statistical Significance

t- Test

• Step 01

Building hypothesis:

H0: β1= 0

(There is no significant relationship between independent variable and the

dependent variable.)

H1: β1≠ 0

23
 (There is a significant relationship between independent variable and the

dependent variable.)

• Step 02

α (probability of rejecting H0) = 0.05

• Step 03

Calculating the t statistic:

t- statistic = Beta coefficient / Standard error

• Step 04

Decision rule:

Reject H0 if -2.145>Z value >2.145

Do not reject H0 if -2.145 <Z value <2.145

• Step 05

Decision:

State whether there is a significant relationship between the independent

variable and the dependent variable at a 95% confidence level.

R2 and adjusted R2

R-squared (R2) is a statistical measure that represents the proportion of the variance for a dependent

variable that's explained by an independent variable or variables in a regression model. Whereas

correlation explains the strength of the relationship between an independent and dependent

variable, R2 explains to what extent the variance of one variable explains the variance of the second

variable. So, if the R2 of a model is 0.50, then approximately half of the observed variation can be

explained by the model's inputs.

24
We use adjusted R2 to compare the goodness-of-fit for regression models that contain differing

numbers of independent variables. The adjusted R2 adjusts for the number of terms in the model.

Importantly, its value increases only when the new term improves the model fit more than expected

by chance alone. The adjusted R2 value decreases when the term doesn’t improve the model fit by

a sufficient amount. And also, it explains whether the included variables really reflect the

dependent variable, in our case the revenue. So, the adjusted R2 and R2 should be almost closer. If

it’s not closer, it means that we have not included necessary variables.

In our study the variation of revenue is explained by the variations of Independent variables. Our

R2 was 0.8367 while the value of adjusted R2 was 0.7783.

The R2 indicates that 83.67% of the variance in revenue of carpenters can be predicted from the

independent variables that above mentioned. And the gap between Adjusted R2 and R2 is 0.0584

which is comparatively low. So, we can conclude that our model has an overall significance

because even though we have used different variables they all are necessary for the model.

Root-Mean-Square Error (RMSE)

The RMSE is the square root of the variance of the residuals. It indicates the absolute fit of the

model to the data–how close the observed data points are to the model’s predicted values. Whereas

R-squared is a relative measure of fit, RMSE is an absolute measure of fit. As the square root of a

variance, RMSE can be interpreted as the standard deviation of the unexplained variance and has

the useful property of being in the same units as the response variable. Lower values of RMSE

indicate better fit. RMSE is a good measure of how accurately the model predicts the response,

and it is the most important criterion for fit if the main purpose of the model is prediction.

25
In our study the relevant RMSE is 140,000 and the production mean is 587,100. So, compared to

mean value this is a lower value and we can conclude that how close the observed data points are

to the model's predicted values.

Standard Errors

The standard errors of the coefficients of the model explains the deviation of population mean

from the estimated mean value of the coefficients. Accordingly following are the standard errors

of the coefficients.

Standard error of;

̂ = 3168.967
𝛽1

̂ = 7738.076
𝛽2

̂ = 69257.63
𝛽3

̂ = 1482.788
𝛽4

̂ = 94653.97
𝛽5

Overall Significance of the Model

Probability F Test

H0: β1=β2=β3=β4= β5=0

H1: At least one βi≠0

α=0.05

Prob > F =0.0000

26
Decision Criteria:

If Prob > F < 0.05: Reject H0

If Prob > F > 0.05: Fail to reject H0

Decision: Reject H0

Conclusion:

There is enough evidence to say that at least one independent variable included in the model

significantly affects the dependent variable, weekly revenue.

Independent Variables Analysis

1. Number of laborers

The number of workers employed at these firms have a significant relationship with the output

they produce. Firms with higher number of workers have the opportunity to gain advantages of

division of labor and thereby increased productivity resulting in comparatively higher levels of

revenue. Therefore, the analysis has included the number of workers as a key determinant of the

revenue in these firms. Impact of labor in terms of cost was not considered due to the changes in

salaries paid to those workers, rather the number of laborers has been taken to show the relationship

between labor and production.

2. Years of experience

According to the years of experience in these firms there are changes in revenue since the p value

of experience variable is less than the significance level 0.05. Level of experience was measured

by the number of years since it gives a reasonable idea about the exposure of the firm in the

furniture industry. The reason behind this relationship is when the owner has a higher level of

27
experience it has a positive impact to the production. Firms with lower level of experience are new

to the industry and therefore have a lack of awareness and exposure to the industry. So, a firm with

higher level of experience can earn a higher revenue compared to a firm with lower level of

experience in the furniture industry.

3. Timber quantity

The analysis shows that the amount of timber used by each firm does not have a significant

relationship with the revenue. Therefore, even though a relationship existed, according to the

analysis the timber quantity is not a significant determinant of the revenue in these furniture firms.

4. family inherent

Since the p value of this variable is greater than the Significance level 0.05 it doesn’t have a

statistically significant relationship with revenue. So even though there might possibly be a

relationship it is statistically insignificant to the dependent variable. It doesn’t have any strong

impact on the revenue. The main reason behind this is that firms which have been inherited from

their owners' families are more reluctant to implement changes in their business or to adopt new

strategies. They tend to continue the same traditional production methods and technology. They

don’t feel the need to make changes in their established systems and therefore become less efficient

than the firms that use modern technology and new features in their furniture. On the other hand,

even though the firms that are family inherent avoid making changes in some instances it could be

beneficial for them. Since they have been repeating the same processes in manufacturing and

selling, they have more knowledge and skills to perfect their output than newly established firms.

And they have better brand recognition and customer loyalty. So in this case family inherent

variable shows an insignificant relationship with the dependent variable.

28
 5. Labor training

This variable was included to test the impact of the training program the workers have received on

the production of furniture in the firm. The results show that it is statistically insignificant to the

dependent variable. Therefore, output produced by a worker who has received training program

does not differ from output produced by a worker who hadn’t received any labor training programs.

The training hours or the training program itself do not affect to the variations in the revenue

among these firms.

Checking for Errors

Multicollinearity

In a Classical Linear Regression Model, we assume that the model is free from the multicollinearity

along with some other assumptions. Multicollinearity problem exists only in multiple regression

model. Multicollinearity occurs when the independent variables which determine the dependent

variable are very highly correlated with each other. When multicollinearity occurs, standard error

may be inflated.

We can identify Multicollinearity through Stata by using “vif” command.

Table 8 : Multicollinearity Test

29
 If mean vif value or independent vif value is less than 10 there is no multicollinearity problem.

Since the mean vif (1.26) and each invidual variables’ vif values (1.45,1.31,1.24,1.23,1.10) are

less than 10 multicollinearity problem does not exist.

Heteroscedasticity

Heteroscedasticity means when the error variance (ui) team is not constant. If there is no

heteroscedasticity, we call it homoscedasticity. important assumptions of the classical linear

regression model are that the variance of each disturbance term ui, conditional on the chosen values

of the explanatory variables, is some constant number equal to σ2.

To identify Heteroscedasticity problem, we can use “rvfplot” and it graphical shows the pattern of

variance.
200000
100000
Residuals

0
-300000 -200000 -100000

200000 400000 600000 800000 1000000 1200000

Fitted values

Chart 3 : Residuals and Fitted Values

30
As you can see, there is no pattern that we can observe in this graph. Therefore, there is no

heteroscedasticity problem.

Since the diagram is subjective, the heteroscedasticity can be shown through Breusch Pegan Test.

For that we use “estate hottest” command.

. estat hettest

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity

Ho: Constant variance
Variables: fitted values of datasetrevenue

chi2(1) = 1.59
Prob > chi2 = 0.2075

H0: The model is free from Heteroscedasticity

H1: The model has Heteroscedasticity

If P value is less than 0.05, then we can reject null hypothesis. That means the model has

heteroscedasticity problem. In our model P value (0.2075) which is greater than 0.05. Therefore,

we fail to reject the null hypothesis. That means the model is free from heteroscedasticity.

Specification Bias

Specification bias means the specified model is not correct. Among many, omitted variable bias is

one form of specification bias. Testing for omitted variable bias is important for our model since

it is related to the assumption that the error term and the independent variables in the model are

not correlated.

In order to test this, we use “ramsey reset(ovtest)”

31
. ovtest

Ramsey RESET test using powers of the fitted values of datasetrevenue

Ho: model has no omitted variables
F(3, 11) = 1.23
Prob > F = 0.3443

H0: This model has no omitted variable.

H1: This model has omitted variable.

If P value is less than 0.05, we reject null hypothesis. That means the model has omitted variables.

In our model the P value (0.3443) is greater than 0.05, which means we do not reject H0. Therefore,

the model has no omitted variables. We can conclude that the model is correctly specified.

Testing Normality Assumption

From normality assumption we check whether the behavior of residuals is normally distributed,

for that we use “Shapiro-wilk test”.

. swilk e

Shapiro-Wilk W test for normal data

Variable Obs W V z Prob>z

e 20 0.98322 0.397 -1.861 0.96861

H0: Error terms are normally distributed

H1: error terms are not normally distributed

32
If p value is less than 0.05, then we can reject null hypothesis. That means the error terms are not

normally distributed. In our model key value (0.96) is greater than 0.05. That means we fail to

reject H0. Therefore, the error terms are normally distributed.

33
 Conclusion

Furniture industry is the one of major industries in Sri Lanka which plays a significant role in the

Sri Lankan economy. It has been making significant contribution towards the economy through

creating employment opportunities and increasing Gross Domestic Production (GDP) mainly in

terms of manufacturing sector. It is obvious that majority of furniture manufacturing firms are

located in Moratuwa which is dominated in manufacturing furniture. This mini research mainly

focused in deriving a most fitted multiple regression line for dependent variable, production value

based on independent variables, years of experience in the industry, number of laborers, family

inheritance, quantity of timber used, electricity payment and the training experience.

The research is mainly based on primary data and survey method was used to collect data from the

Moratuwa area. Survey was based on interviewer administrative questionnaires given by the

Department of Business Economics. Our sample consists of 20 observations selected to cover the

population of all the furniture firms exist in Moratumulla area. All the interviews were conducted

as one-to-one interviews in collecting data.

We have taken both quantitative and qualitative data to analyze the growth of furniture industry in

Moratumulla. As for the dependent variable of the model we have taken the production value and

as for independent variables we have taken years of experience in the industry, number of laborers,

family inheritance, quantity of timber used, electricity payment and the training experience. All

these variables are measured on a weekly basis and in order to quantify qualitative variables

dummy variables were used. We have selected two independent variables such as family

inheritance and training experience as our qualitative variables. These variables are analyzed

34
through a dummy variable regression. In order to analyze data collected through the survey, STATA

and EVIEWS software were used and we were able to derive a best fitted multiple regression line

based on selected independent variables.

According to the results of the regression analysis only years of experience and number of laborers

have a statistically significant relationship with the dependent variable, revenue. Therefore, even

though a relationship existed, according to the analysis all other independent variables are not

significant determinants of the revenue in these furniture firms. When it comes to explanatory

power of the model, the R2 indicates that 83.67% of the variance in revenue of carpenters can be

predicted from the independent variables that above mentioned. And the gap between Adjusted R2

and R2 is 0.0584 which is comparatively low. So, we can conclude that our model has an overall

significance because even though we have used different variables they all are necessary for the

model. On the other hand, F value of the model is less than 0.05 probability which concludes that

there is enough evidence to say that at least one independent variable included in the model

significantly affects the dependent variable, weekly revenue. Jarque – Bera Test and Anderson –

Darling Test were used to check whether residuals of the model have distributed normally or not.

So results indicate that, under 5% level of significance there are enough evidence to prove that the

residuals are normally distributed. When it comes to multicollinearity problem, as our mean “vif”

value and each individual variable’s “vif” values are less than 10, our model is free from the

multicollinearity problem. At the same time model does not subject to the Heteroscedasticity

problem as the p value derived from Breusch Pegan Test is greater than 0.05, where we failed to

reject the null hypothesis. Therefore, as a whole it is concluded that the regression model is best

fitted to derive weekly revenue earned by furniture manufacturers or sellers in Moratumulla.

35
 Recommendations

• Through the survey we carried out we identified that most of businesses do not use

technological advancements in their operations. So, we recommend using facilities like E-

marketing, Business website, Online buying and selling platform in order to increase the

market share and their revenue.

• Most of manufacturers are facing a problem of not having experienced workers. So, it is

recommended to facilitate their workers to going through a training program.

• It is identified that firms which have been inherited from their owners' families are more

reluctant to implement changes in their business or to adopt new strategies. They tend to

continue the same traditional production methods and technology. So, in order to attract

new customers while maintaining their long-term loyal customers we recommend them to

adopt with new changes shared by their competitors.

Limitations

When carrying out the mini research project the team had to face the following limitations,

• No consistency in the data we collected was recognized

• The respondents to the questionnaire sometimes objected to answer certain questions in the

questionnaire or some tried to boycott some questions

• The respondents were trying to be average in answering the questionnaire rather answering

with the real data

• Inadequate sample size

• Lack of previous research studies

36
 Bibliography

Amarasekara, H., 2012. A study on the status of furniture manufacturing industry in moratuwa

area. s.l., s.n.

Dasanayaka, S. W. S. B., 2011. Identification of barriers for development of the Sri Lankan small

and medium scale furniture and wooden products manufacturing enterprises, a study based on the

Moratuwa area.. Euro Asia journal of management, 21(40), pp. 61-101.

H.S.C.Perera, 2009. Manufacturing Strategy and Improvement Activities of Sri Lankan Furniture

Manufacturers. ENGINEER, Volume XXXXII, pp. 11-20.

Padmasiri, H. N., 2012. The role of human and social capital on the development of wooden

furniture clusters in Sri Lanka. International Journal of Development , Volume 11, pp. 19-36.

Shantha, A., 2013. Resource use efficiency of small scale furniture industry in Sri Lanka. s.l., s.n.

37
 Appendix

Appendix 1: Images of Moratumulla visit

Appendix 2: Questionnaire

38
Appendix 1: Images of Moratumulla visit

39
40
Appendix 2: Questionnaire

41