Chapter 1- Introduction
1.1 Introduction
Construction is one of the major contributors for National Economy in any country. In 
another words, it accounts for a sizable proportion in Gross Domestic Product (GDP) and 
Gross National Product (GNP). Therefore, this analysis was undertaken to determine the 
relationship between national economy and construction activities in all segments in Sri 
Lanka mainly categorized as residential buildings, non-residential buildings and others. These 
categories cover all types of construction activities in the country including hotels, housing 
complexes, hospitals, trade centers, towers, schools, other buildings, urban infrastructure 
including water supply, sewerage and drainage. Also it comprises construction of roads, 
railways, highways, ports, airports; power & oil storing systems; irrigation, recreation 
facilities and agriculture systems and telecommunications etc. that indicates input - output 
flux of national fiscals.
As per the definition of United Nations (UN) construction is “an economic activity directed 
to the creation, renovation, repair or extension of fixed assets in the form of buildings, land 
improvements of an engineering nature and other such engineering constructions as roads, 
bridges, dams and so forth”. It is a process that consists of the building or assembling of 
infrastructure in the fields of architecture and civil engineering. Construction industry 
involves a broad range of stakeholders and provides substantial employment opportunities to 
unskilled, semi-skilled and skilled labor markets. It supplements the foreign revenue derived 
from trade in construction material and engineering services. Further, it has strong linkages 
with other sectors which positively contribute towards the national income.
Construction activities in Sri Lanka has shown fairly upward trend year-on year since 1990. 
However, after the end of 30 year prolonged war in 2009, construction industry has shown a 
steep development with an average of 13.0% increase from 2009 to 2013. This has been 
accelerated specially due to reconstruction activities undertaken in North and East parts of the 
country and many other development projects carried out around the island. The trend for real 
GDP and construction sector GDP fluctuation is shown in figure 1 below.
§
£ UfcftAJW ^1
* >
Figure 1 - Construction Sector and overall GDP Growth Rate (Source : CBSL Annual Report 
2013)
As shown in Table 1.1 below, construction sector has contributed 8.1% to the overall GDP 
(real) in 2012 and 8.7% in 2013. Similarly, construction subsector has contributed 26.7% 
towards ‘Industry Sector’ in 2012 and subsequently 27.8% in 2013.
Table 1 - Comparison of construction sector in terms of GDP and Industry Sector (Source : 
CBSL Annual Report 2013)
ConstructionIndustry 
Sector (Real)
Construction 
as % of 
Industry Sector
GDP (Real) ConstructionGDP (Real)Year
as % of GDP(Rs. Mn)
(Rs. Mn)(Rs. Mn)
23.2%6.6%701,129 162,7902009 2,449,214
23.4%6.7%2010 177,9122,645,542 760,334
24.2%7.1%2011 2,863,715 838,932 203,204
26.7%8.1%2012 247,0913,045,288 925,152
27.8%8.7%282,7422013 3,266,099 1,016,886
2
However, by looking at just the row figures it may be difficult to get a clear understanding of 
if any money invested in construction industry may bring profits or not at the end. Hence, 
entrepreneurs and investors are interested to have a reliable and logical approach to be used 
in the decision making process to determine whether or not there is an obvious link between 
construction industry investments and constant economic growth of the country to get a 
feeling about the returns of their investments (ROI). Therefore, the intention of this study is 
to provide a mathematical approach to assist the decision making process in the context of Sri 
Lankan economy to identify the correct relationships between the investments in construction 
activities and the economic indicators of the country.
1.2 About Sri Lanka
Sri Lanka is an island located in the Indian Ocean with total land area of 65,610 squire kilo 
meters and 20.3 million of population with a distribution of 47.4% male and 52.6% female in 
gender (2012 census). Further, Sri Lanka has recorded 1% of population growth rate (2011). 
Sri Lanka is strategically located at the cross roads of major shipping routes connecting South 
Asia, Far East and the Pacific with Europe and the Americas next to the fast growing Indian 
sub-continent with close proximity to Southeast Asia and the Middle East.
As per the Socio Economic Indicators published by Central Bank of Sri Lanka (2013) Ethnic 
diversity of the country has been reported as Sinhalese - 74.9%, Sri Lankan Tamils — 11.2%, 
Indian Tamils - 4.2%, Moors - 9.2%, and Other - 0.5% (2012 census) with overall literacy 
rate (Aged 15 years and above, 2010) 91.9%. (Male - 93.2% and Female - 90.8%) Also the 
life expectancy (2011) has been reported as average of 74.9 years in total population.
Labor force as a percentage to the population has been recorded as 48.2% in the country 
(exclude Northern and Eastern provinces, 2011) whilst Unemployment percentage out of total 
labor force has been recorded as 4.2% with the same conditions. Per capita income has been 
recorded as USD 2,805 (LKR 310,124) in 2011 and this is an increase of 15.73% compared 
to 2010 and 14.65% compared to 2009.
Sri Lanka is rich with natural resources such as graphite, apatite, limestone, dolomite, mica, 
quartz, calcite, silica sand, gems, mineral sands, clay, hydropower and phosphates. Certain
3
minerals such as Graphite and Silica (mostly monazite) and few verities of gems are mainly 
exported as row materials with no purification or processing. The Government has emanated 
the State policy under ‘Mahinda Chinthana’ vision to develop Sri Lanka as one major hub in 
the South Asian region consist with five sub components such as Maritime, Aviation, 
Commercial & Tourism, Knowledge and Energy. This is called ‘five-hub concept’ in 
government strategy plans. However, still the main economic revenue of the country is 
generated through tourism, apparel & textile, tea, rubber & coconut exports and other 
agricultural products such as rice production. Further, Sri Lanka receives significant 
remittance through work force deployed in foreign countries for employment.
Most of third world countries are facing the problem of racing funds to improve their capital 
expenditures. As a developing country, Sri Lanka is also facing for the same issue in terms of 
inadequacy funds to improve infrastructure facilities in the country. Nevertheless, it’s a 
challenge to meet global standards and to par with global economic trends specially to attract 
investors from the outside the country. Therefore, developing of traveling and transportation 
facilities, aviation, ports, telecommunication, information technology and uninterrupted 
power solutions are essential and high priority.
1.3 Sri Lankan Economy and Outlook
As per Central Bank Statistics for past five years (2009 -2013), Sri Lanka has gained average 
of 6.7% growth in real GDP. Further, it was noted a sustainable growth trajectory throughout 
the period and has achieved 7.3% growth in 2013 compared to 6.3% in 2012. (Refer Table 
1.2). This has been supported by growth of all sectors including improved earnings from 
merchandises, service exports and workers’ remittances received in to the country. It was also 
noted the inflation (Colombo Consumers' Price Index, Department of Census and Statistics, 
Base 2006/07=100) remained at a single digit level for the fifth consecutive year, recording 
6.9 in December 2013 whereas 7.6 in December 2012. Latest reports indicate that this has 
been further reduced to 5.6 in April 2014. Year-on-year headline inflation too has moved on a 
decelerating path since March 2013 with the improvements in supply conditions. 
Continuation of this trend will help to reduce wage pressures in the economy and also to raise 
investor confidence in the long run.
4
Further, as per World Economic Outlook by IMF (April, 2014) it is noted that, Sri Lanka has 
made a remarkable progress in comparison to the emerging and developing markets in the 
Asian continent.
Table 2 - World GDP - Real (Source : World Economic Outlook by IMF (April, 2014))
Average
1996-2005 2009 2010 2011 2012 2013
1.0-2.7 1.8 1.4 1.22.9Advanced Economies 
United States 
United Kingdom 
Germany 
France
1.61.8 2.43.9 -3.0 1.5
1.6-0.6 1.43.9 -4.8 1.2
0.4 0.6-1.6 1.8 2.90.7
-0.1 0.31.6 1.0-1.42.2
2.2 2.1-2.3 2.0 0.70.8Japan
Emerging and Developing 
Asia
Bangladesh
Bhutan
China
6.57.9 6.77.7 9.77.1
6.1 5.85.9 6.4 6.55.4
6.5 55.7 9.3 10.16.9
7.79.2 10.4 9.3 7.79.2
4.48.5 10.3 6.6 4.7India 6.4
6.5 6.3 5.8Indonesia 2.6 4.6 6.2
6.3 7.3Sri Lanka 3.5 8 8.24.3
All sectors of the economy has contributed positively towards the steady growth of DGP and 
backed by favorable climate conditions prevailed during the year 2013. As a result, GDP in 
nominal terms has been grew by 14..5% to LKR 8,674 billion (USD 67 billion) in 2013 
compared to 15.8% of growth amounting LKR 7,582 billion (USD 59.4 billion) in 2012. This 
has helped to raise GDP per capita from USD 2,923 in 2012 to USD 3,280 in 2013 by 
recognizing Sri Lanka as the second highest country for GDP per capita among the SARRC 
members after Maldives.
| UBRAM |5
Figure 2 - GDP Current and Year-On-Year Change % (Source : CBSL)
In terms of expenditure, GDP growth supports increasing of consumption as well as 
investments in the country. Also it was noted interest rates have been reduced from double 
digit levels to single digit levels by the end of year 2013. This will lead to a positive move of 
the economy as relatively low interest rates will help to increase resources for investment 
activities in the future with the expansion in credit growth. Credit to both Industry and 
Services sectors have been increased by LKR 201.1 billion in 2013. The Central Bank policy 
introduced in 2013 has resulted to increase declarations of credit obtained from commercial 
banks in Sri Lanka. Further, it has helped to boost the access to the domestic and global 
capital markets by the private sector in 2013.
As per the preliminary findings of the Household Income and Expenditure Survey (HIES) for 
2012/13, population below the Poverty Line ratio has been declined to 6.5% from 8.9% in 
2009/2010. This is also a positive trend that indicates the development of Sri Lanka in terms 
of investor confidence. Further, it was noted that new reforms have been new introduced to 
simplify the tax system while reducing its inefficiencies and leakages. This will help reducing 
tax evasion and increase state revenue.
As per the CBSL economic indicators, it is observed an increase of labour force by 4.1% in 
2013 compared to the previous year. Further, it was noted that labour force participation rate 
(LFPR) also has increased to 53.8% in 2013 compared to 52.6% in 2012. Although, 
unemployment rate has recorded high in 2013 this would have been resulted by increased
6
entry from rural females into the labor market seeking local job opportunities which 
attributing 6.6% of female unemployment rate.
By considering all facts above, in overall, the economy of Sri Lankan is expected to continue 
its growth momentum in the medium term underpinned by increased investment, improved 
macroeconomic stability and improving global economic conditions. These are favorable 
indicators for entrepreneurs who look opportunities to invest in Sri Lanka in any sector, 
including the construction sector as well.
1.4 Construction Sector Contribution to Gross Domestic Product
The construction sector share to overall GDP has been improved steadily since 1990. CBSL 
statistics indicates 13% of average growth (in real terms) in Construction Sector during the 
past five years (2009 -2013). Further, it was noted the Construction Sector GDP in real terms 
has achieved 282.7 billion (USD 2.2 billion) in 2013 in comparison to LKR 247 billion (USD 
1.9 billion) in 2012. Whilst the highest growth rate of 21.6% recorded in 2012, still the sector 
has maintained 14.4% of growth in 2013. (Refer figure 1)
GDP Construction and Year-On-Year Change Percentage
^ 1,000,000 
g 900,000
I 800,000
5 700,000
^ 600,000 
•2 500,000
s 400,000 
* 300,000
g 200,000
- 45.00%
- 40.00%
- 35.00%
- 30.00%
- 25.00%
- 20.00%
- 15.00%
- 10.00%
- 5.00%
- 0.00%
u 100,000
ft 0 o«—'<Nm’^,w^vor^oooNO^<Ncn,^r»nvor^ooo'0*—‘CNm
ONONOsONOnOnONOnOnOnOOOOOOOOOO’—< — —• — OnOsONOnOnOnOsOnONOsOOOOOOOOOOOOOOo
—Construction GDP (Current) Change %
Figure 3 - GDP Construction (Current) and Year-On-Year Change % (Source : CBSL 
Annual Report 2013)
As a result of robust construction and manufacturing activities initiated in Sri Lanka, the 
Industry Sector recorded a growth rate of 9.9%, from LKR 925 million in 2012 to LKR 1 
billion in 2013. Industry Sector share also has been increased from 30.4% in 2012 to 31.1%
7
in 2013 from overall GDP contribution perspective. Similarly, Construction Sector has grown 
by 14.4% during 2013 as described above providing 8.7% share to the overall GDP. 
Construction Sector share to the overall GDP in 2012 was 8.1% in 2012 and 7.1% in 2011.
Figure 4 - GDP, Industry Sector & Construction Sector Performance in Real terms (Source : 
CBSL Annual Report 2013)
Heavy investments carried out by the current government of Sri Lanka on infrastructure 
development projects across the country have strongly contributed towards a sustainable 
growth momentum in the construction sub sector. The key infrastructure projects that are 
currently ongoing and recently completed are described under section 1.5 in the chapter 1. As 
a result gross domestic fix capital formation for Construction Sector as well as for Industry 
Sector has been increased. This trend and year-on-year change percentages are depicted in the 
figure 1.5 below.
National accounts for gross domestic fix capital formation for Construction Sector has been 
measured in three main areas comprising residential building, non-residential building and 
other activities such as construction of roads, railways, highways, ports, airports; power & 
oil storing systems; irrigation, recreation facilities and agriculture systems, 
telecommunications etc.
8
Construction Sector Gross Fix Capital Formation (Current) and 
Year-On-Year Change Percentage
1,800,000
1,600,000
40.00%
B
- 35.00%o
•43 ^
S a 1/400,000 
B o
o £ 1,200,000 
£— U 1,000,000 
£.2
Si 
^ £
£
fa o
Cfl U
30.00%
r 25.00% 
20.00%
800,000
600,000
400.000
200.000
■
- 15.00%
r 10.00%
i
5.00%
o 0.00%o o — tNcn-^-'ovor^oooNO — <NmCj\ O' CN O' O' O' O' O' O' O' O O O O 
O'O'O'O'O'O'O'O'O'C'OOOOOOOOOOOOOO
'^■mor^ooc'O — <Ncn 
oooooo — —
““Gross Fixed Capital Formation - Construction (Current) •—Change %
Figure 5 - GDP, Industry Sector & Construction Sector Performance in Real terms (Source : 
CBSL Annual Report 2013)
Building construction is one of the major components attributing over 52% of the total 
investments within the construction industry during last five years. (Refer Table 3 below). 
Therefore, this has a significant contribution to the sub sector especially in the areas such as 
condominiums, mega hotel projects and housing units, etc.
With the introduction of Public-Private Partnerships (PPPs) program for the development, 
financing, & operation of infrastructure, private sector got Immense potential to get involve 
with construction and development projects mainly in residential and non-residential 
buildings as well as other sectors such as transportation, energy, education, housing, health, 
etc. Also the figures indicate the private sector contribution in housing development projects 
attributed to 9.1% and 16.6% for the overall construction sector by commercial banks while 
the public sector was also involved in housing development projects, highways, rail roads and 
transport sector development projects such as phase two of the Southern Expressway, the 
Colombo - Katunayake Expressway, the Colombo Outer Circular Highway Project and the 
Northern Railway Project as well as port development projects. Those have been contributed 
towards sustaining growth momentum in this sub sector. Furthermore, Private sector 
participation remains significant in the areas such as water supply, drainage and other 
construction activities inclusive telecommunication, power and energy sectors. Meantime the 
public investment on economic and social infrastructure development activities have
9
attributed LKR 447 billion (5.2% of overall GDP) in 2013 in comparison to 415 billion (5.5% 
of overall GDP) in 2012.
The growth in the construction sector is also reflected by the increase in imports of 
investment goods and building materials. In 2012 it was recorded LKR 157.4 million spent 
on importing building materials to Sri Lanka and it has been increased to LKR 175 million in 
2013 reflecting 11.2% of growth.
Table 3 - Comparison of construction sector in terms of GDP and Industry Sector (Source : 
CBSL Annual Report 2013)
2009 2010 2011 2012 2013
Gross Domestic Fixed Capital 
Formation 1,147,440 1,452,002 1,772,515 2,189,805 2,536,648
Construction 802,445 996,190 1,118,634 1,400,370 1,631,404
Construction Sector as a % of
GFCF 69.9% 68.6% 63.1% 63.9% 64.3%
Residential Buildings 318,944 420,673 433,167 527,478 595,585
Non Residential Buildings 148,426 217,776 200,398 256,673 261,023
Building Sector as a % of 
Construction GFCF 58.2% 64.1% 56.6% 56.0% 52.5%
Other 335,075 485,069 616,219 774,796357,741
As per the all construction cost index formulated by the Institute of Construction Training 
and Development (ICTAD), the total cost of construction activities has been increased by 
12.2% in 2012 whereas it had decreased to 7.2% in 2013. This reflects the price increases of 
raw materials and labor in the industry. These price hikes are invariably link to the supply 
and demand in the industry. In another words, the rise in labor costs in the construction sub 
sector reflects the lack of labor to meet the demand arising from the expansion in the industry 
and the increase of cost links to an increase in demand. This issue has to overcome with the 
introduction of project management concepts to enhance efficiency and productivity and 
using the technology to automate the areas where need no human intervention.
10
AH Construction Cost Index and Year-On-Year Change Percentage
700.0 t 20.00%
l 18.00%
x 600.0
-a
<5 500.0
u 400.0
s
•2 300.0
V
3
t 200.0
C
(3 100.0 
^ 0.0
j- 16.00% 
r 14.00%
r 12.00%
r 10.00%
- 8.00% 
-- 6.00% 
4.00%
V 2.00% 
1 0.00%
L
j
o-^r'im^-w-ivor-'Goavo — <Nm^-w->vor^oooo — n n 
OnOnOn^>0''0'iO\0'>C'iOnOOOOOOOOOO—<»— *— 
ONOnOnO'On05^0>CT'OnOOOOOOOOOOOOOO,_^_,_^_I_^^rt__^rv1(NrNcNcN(NtN<N<N<NfNrltNrv4
All Construction Cost Index ......Change %
Figure 6 - All Construction cost index and change Percentages (Source : ICTAD 
Publications)
1.5 Recent infrastructure development projects undertaken by the 
Government of Sri Lanka
The year 2013 was a booming year on all types of construction activities invested and 
initiated by both public and private sectors. It was observed the completion and continuation 
of many key projects in terms of improving physical infrastructure in the country. Whilst 
Table 1.4 below indicates a summary of massive public investments carried out by the 
current Government of Sri Lanka. Private sector got opportunity to join with government 
projects through Public-Private Partnerships Program. Also the private sector invested 
heavily specially in the areas like real estate projects.
11
Table 4 - Major Ongoing And Recently Completed Infrastructure Development Projects 
(Source : CBSL Annual Report 2013, Chapter 3)
Major Ongoing and Recently Completed Infrastructure Development Projects
Year
Project Name To be 
CompletedCompleted
Power Projects
Norochcholai Coal Power Plant
Phase I
Phase II: Unit 2 
Phase II : Unit 3 
Uma Oya Hydropower Plant 
Sampur Coal Power Plant
2011
2014
2014
2015
2017
Road Development Projects
Southern Expressway 
Phase I 2011
Phase II
Colombo-Katunayake Expressway 
Outer Circular Highway 
Phase I 
Phase II 
Phase III
2014
2013
2014
2015
2017
Railway Development Projects
Northern Railway Line Reconstruction Project 
Medawachchiya - Madu 
Madu - Thaleimannar 
Omanthai - Pallai 
Pallai - Kankasenthurai 
Signalling and Telecommunication System 
Matara-Kataragama Railway Line Project 
Phase I: Matara-Beliatta
2013
2014
2014
2014
2015
2016
Port Development Projects
Colombo South Harbour Project 
South Container Terminal 
East Container Terminal 
MagamRuhunupuraMahindaRajapaksa Port 
Phase I 
Phase II
Oluvil Port Development Project
Airport Development Projects
MattalaRajapaksa International Airport - Phase I 
_______ Bandaranaike International Airport Expansion Project
2013
2014
2010
2015
2013
2013
2017
12
The share of Foreign Direct Investments (FDIs) in infrastructure development activities 
attributed 56.5% from total FDIs to the country in 2013. Those have been mainly invested in 
the sectors such as telephone and telecommunication networks, housing and property 
development and ports and container terminals. In addition to the mega projects described 
below, government investments on infrastructure developments have been further expanded 
to several urban development initiatives among many of the major cities in the island with the 
aim of enhancing life quality of public. Countrywide township development projects and 
Colombo city beautification project aided the recreational and wellbeing of community. 
Further, those have helped to attract tourist attention and promoting Sri Lanka as an up 
market tourist destination. Meanwhile, many small scale projects also have been carried out 
with the scope of enhancing and facilitating rural livelihood. These projects included rural 
road developments and carpeting projects, electrification projects, irrigation projects and 
community based water supply projects, etc.
1.5.1 Power and Energy Projects
The government has invested in developing of power & energy projects as described below.
• Norochcholai Coal Power Project -The construction of first phase of this Power Plant 
(300 MW) was completed at the end of 2010 and with an investment of USD 450 million 
loan obtained from China. This was added to the national grid by end of March 2011. The 
second and third phases (600 MW) are due completing before the end of 2014 with 
another investment of USD 891 million.
• Uma Oya Hydropower Plant -The estimated cost of this project is USD 529 which 
mainly funded by the Government of Iran through Export and Development Bank as a 
loan. This is expected to connect to the national grid in 2015 with the output of 120 MW 
to the generation system with 230GWh of annual energy generation.
• Sampur Coal Power Plant - Phase I of this project had estimated USD 300 million and 
USD 200 million has been financed by the Indian government in 2010. It is expected to 
generate 500 MW of electricity to the national grid in 2017 from this project.
• Upper Kotmale Hydro Power Plant - Construction of the Upper Kotmale Hydro Power 
Plant incurred a cost of USD 475 million with the assistance from Japan International
13
Cooperation Agency (JICA).This power plan provides 150 MW output capacity using 
two turbines each can generate 75 MW.
1.5.2 Road Development Projects
The government strategies are in placed (National Road Master Plan) to construct highways 
and bridges at both national and provincial levels. Further, large investments can be seen on 
widening of roads, introduction of flyovers, reconstruction of bridges, etc. with the aim of 
increasing mobility in freight and passenger transportation through land network and 
reduction of traffic congestion and fuel costs.
• Southern Expressway -The Southern Highway (Southern Expressway) project has cost 
USD 741.1 million and it is a 126km-long express highway running from Colombo to 
Matara on the south coast. Further, it is the longest expressway out of the proposed 
expressway networks with 100 km/h constant speed until the end of the journey. This 
highway consists with 4-lane capacity. The first section (Phase I) from Kottawa to 
Pinnaduwa (Galle) was completed in November 2011 and the second section (Phase II) 
from Pinnaduwa (Galle) to Godagama (Matara) was opened to public in March 2014.
• Colombo-Katunayake Expressway - This is an investment of USD 292 million from 
Exim Bank in China. This Expressway is 25km-long with 4-lane capacity. It links the 
capital Colombo with Bandaranaike International Airportin Katunayake and further 
extended up to Negombo city. This was opened to the public since October 2013.
• Outer Circular Highway (OCH) -This is located in the Colombo Metropolitan Region 
and passes through two administrative districts, namely Colombo and Gampaha. This 
Highway runs around 20 km away from the City center of Colombo, connecting radial 
routes and has a total length of 29.2 km. The northern end of the highway is located at 
Kerawalapitiya on Colombo - Katunayake Expressway and the southern end in located at 
Kottawa.
The expressway is implemented in three phases as given below;
Phase 1 - Section from Kottawa to Kaduwela - 11km (opened for public since March 
2014 with the financial assistance from Japan.)
14
Phase 2 - Section from Kaduwela to Kadawatha - 8.9km (commenced in May 2014 
and due completion by 2015.)
Phase 3 - Section from Kadawatha to Kerawalapitiya - 9.3km (Due completion by 
2017.)
• Colombo - Kandy Highway - This is an alternative Highway between Colombo and 
central hill capital of Kandy with 98km-long. This is connected to the Outer Circular 
Highway at Kadawatha and run up to Katugastota. Field surveys have been completed 
and design and land acquisitions are in progress.
Phase 1 - Kadawatha to Ambepussa 48.2 km (will be commencing in 2015 with the 
Chinese assistance. This is expected to have 4 lanes track with provision for 6 lanes in 
future. Expected cost would be USD 1 billion.)
Phase 2 - Ambepussa to Katugastota 50.7 km (This is expected to have 2 lanes track 
with provision for 4 lanes in future.)
1.5.3 Railway Development Projects
Following railway projects are currently under way. These are extensions to the existing 
railway lines with the state mission to expand the national rail network.
• Northern Railway Line - The Northern Railway Line in Sri Lanka before terminating at 
the railway station call Kankasanthurai at the Northern Province and it stretch was 
339km-long. After the end of war in 2009, Sri Lankan government initiated 
reconstructing of Northern Railway Line from Vavuniya to Omanthai by Sri Lankan 
military and Omanthai to Pallai by IRCON International. (Total 96 km-long in 
length)Reconstructed Northern Railway Line from Kilinochchi to Pallai (30 km) was 
opened to the public in March 2014. The re-constructed railway line from Omanthai to 
Kilinochchi (62 km) was commissioned and opened to the public in September 2013. 
Reconstruction of Omantei to Palei Railway Line was cost USD 185.49 million which 
was supported by Indian government.
• Matara-Kataragama Railway Line Project - The extension of the new railway line 
from Matara to Kataragama will be carried out in three phases. Estimated length of this^
15
Railway Line is a 114.5 kilo-meters. Once the project is completed the trains will travel at 
a speed of 120 kilo-meters per hour.
Phase 1 - from Matara to Beliatta, (This stretch is 26.75 km-long, single line rail track 
and has been commenced construction on October 2013. Chinese government has 
granted a USD278 million loan for this phase)
Phase 2 - from Beliatta to Hambantota(48km-long)
Phase 3 -from Hambantota to Kataragama.(39.5km-long)
1.5.4 Ports Development Projects
The Sri Lankan government is capitalizing on the island's strategic location to promote it as 
an economic hub in South Asia region and to cater the increasing demand of services in the 
international shipping industry. Therefore, developing of ports and maritime also one of the 
key priorities in government agenda in terms of expand the capacity and improve the 
efficiency of existing ports through modernization and construction of new ports in strategic 
locations.
• Colombo Port Expansion Project (CPEP) - This Project was initiated in 2006 to cater 
for increasing demands of services in the international shipping industry. After 
completing of this project, Colombo port will gain the strength to handle large cargo ships 
that carry more than 8,000 containers. This project is jointly financed the USD 300 
million from Asian Development Bank (ADB) and USD 200 million from the Sri Lanka 
Ports Authority (SLPA). This entire infrastructural development project is called 
“Colombo Port Expansion Project (CPEP)”
i. South Port Expansion Project and Container Terminal- The Colombo 
South Harbour is situated west to the existing port of Colombo comprising an 
area of approximately 600 hectares. The harbour will be served by a new two- 
way channel with a depth of 20m and a width of 570m. The harbour has 3 
terminals each having 1,200m length and facilities to accommodate 3 berths 
alongside with capacity of 2.4 million TEUs per annum. The project was 
intended to increase the capacity of the Colombo Port by 160% upon
16
completion. The small boat harbour at the end of secondary breakwater has 
400m length of quay wall and the construction work of this breakwater has 
been completed. The breakwater cost over USD 400 million and USD 300 
million has been funded by Asian Development Bank (ADB).
ii. East Container Terminal - The Terminal is being developed as a twenty-first 
century facility to serve international shipping requirements. The construction 
work has been undertaken by Colombo International Container Terminals 
Ltd., (CICT) in December 2011 with an envisaged investment of USD 500 
million, including the installation of the latest state of the art terminal 
equipment. It is so far the single largest foreign direct investment in Sri Lanka 
by a private entity. The Port of Colombo with a current capacity of over 4.5 
million TEUs has embarked on a large infrastructure development programme 
to increase the total capacity of the Port of Colombo by another 7.2 million 
TEUs in three separate phases.
• Hambantota (MagamRuhunupura - MahindaRajapaksa) Port- This is located in the 
Southern Province and about 10 km away from the new Mattala Rajapakse International 
Airport with an initial investment of USD 450 million. The total project will occupy an 
area of 4,000 acres (16km2) and expected to accommodate 33 vessels at time. The 
constructions are underway for Phase I with room for three further terminals in the 
coming years.
Phase I - The total estimated project cost is USD 361 million and 85% from 
that to be funded by the government of China. The construction was carried 
out in 43 hectares of land. The main construction work in Phase I was 
commenced in 2008. The port is expected initially to function as a service and 
industrial port accommodating vessels up to 100,000 DWT and later be 
developed to handle transshipment cargo.
Phase II - This has planned to construct an additional quay approximately 
2500 meters long to accommodate four 100,000t and two 10,000t wharf 
berths. Phase II of the port is underway at an estimated cost of USD 800 
million. The port is ideally located to serve the main East-West shipping lane 
connecting Europe and the Middle-east with South East Asia.
17
Phase III - This phase has planned to construct of a container oil terminal of 
300m long and 17m depth, four 100,000 DWT container berths, one 100,000 
DWT oil wharf and two 30,000 DWT feeder berths. With the completion of 
the phase, the port will facilitate around 33 ships at a given time provide 
services such as bunkering facilities, ship repair, ship building and general 
shipping services while, ‘Raw Raw’ services for importing and re-exporting.
• Oluvil Port Development Project - Oluvil port is located in east coast of Sri Lanka and 
proposed project comprises the construction of a commercial port and a basin for fishing 
crafts. The port is to be built on an open shore and will be spread along 1.4 km of the 
coastline. The commercial port will comprise 330 meters of quay with a water depth of 8 
meters enabling 5000 ton ships while the fishing port will comprise 200 meters of quay 
with a water depth of 3 meters. Construction work of this project commenced in 2008 
with the funds granted through a loan by the Ministry of Foreign Affairs of Denmark. 
(DANIDA) Oluvil port is expected to provide more convenient and cost effective access 
to and from the southeastern region for goods and cargo originating on the west coast.
1.5.5 Airport Development Projects
The government of Sri Lanka expects two million tourist arrivals by 2016. Development of 
infrastructure at the airports is highly important to meet this objective as well as to increase
international passengers and international aircraft movements. Also it is a key to support the 
overall economic development of Sri Lanka. Therefore, the Government has decided to
Phase 2 and toimplement Bandaranaike International Airport Development Project 
construct of second International airport at Mattala.
• Expansion of Bandaranayake International Airport Development Project, Phase II - 
By keeping a mile stone, aircraft movements through Bandaranayake International 
Airport (BIA) has passed the mark of 52,000 in and out of Sri Lanka in 2013 showing a 
7.1%increase compared to the previous 2012, in the history' of the country’s civil aviation. 
Furthermore, the total number of passenger moments in and out of Sri Lanka in 2013 has 
been recorded as 7,328,798 with 3,621,822 passengers arriving and 3,690, 047 departing
of Civil and Aviation estimates that more than 8 millionusing the BIA. Ministry
18
" LANKA
JWA
international passengers will move through the country by 2015. Therefore, a project is 
under way to expand the Bandaranayake International Airport (BIA) with adding 
transit area, construction of new baggage-reclaim area, aircraft parking apron, taxiway, 
multi storied car park and widening of the existing runway. The agreements have been 
signed in 21012 with the Government of Japan (GOJ) to funded USD 346 million for this 
project as a way of long term loan and this project is expected to be completed in 2017.
a new
• Development of second International airport at Mattala - The growth in air service 
sector in South Asian Region has created more opportunities for Sri Lanka to capture 
overflying aircrafts and to provide them vast range of services. Therefore, the government 
of Sri Lanka decided to construct the country’s 2nd International Airport, Mattala 
Rajapaksa International Airport serving the city of Hambantota in southern province, Sri 
Lanka with the aim of enabling international trade, tourism, vocational training and 
employment in addition to above. Features of this new airport include a 4,000m-long, 
75m-wide runway and two taxi ways, as well as a 5,000m2 cargo facilities. The airport 
can also house the world’s largest commercial aircraft such as Airbus A380. Upon 
completion the full project, the new airport will gain the capability to handle one million 
passengers a year, 30,000 aircraft movements and 45,000tons of cargo.
Phase I - China's Export-Import Bank came forward to fund the phase I of this 
project by granting a USD 210 million loan. Construction work for 1st phase 
commenced in November 2009 and it was completed and opened on 18th March 
2013. That included one runway, aerodrome facilities. Passenger and Cargo 
Terminals and a taxi way along with basic infrastructure such as access roads, 
accommodation for officials, fuel farm, sewerage treatment plant, water supply 
facilities, meteorological building, fire building, catering facility and a car park.
Phase II - This stage will include a full length parallel Taxi way, a flying school, an 
airport hotel and recreational facilities along with 20 parking bays for Aircrafts and 15 
bridges. It will increase an additional five million passenger capacity once
implemented in 2015.
aero
19
1081)66
1.5.6 Water supply and irrigation projects
The government has initiated many water supply and irrigation projects around the country to 
augment urban water supply, boost agriculture activities by improving irrigation systems, 
reservoirs and water resources management.
As per the central bank annual report in 2013, National Water Supply and Drainage Board 
(NWS&DB) has managed to increase pipe borne water to 44.3% in 2013 in comparison to 
43.5 % in 2012 through several new projects carried out in various districts such as Colombo, 
Gampaha, Kandy, Galle, Ampara, Kurunegala and Jaffna. During the year, the government 
has spent 26.4 billion rupees for the implementation of water supply and sewerage projects. 
Meanwhile, two major projects launched namely Kelani River Right Bank and the towns 
North of Colombo-Stage II, are aiming to provide benefits over 1.25 million consumers in 
Colombo and Gampaha districts. Also it should be essentially stated that some of the large 
scale water supply projects have been carried out particularly in the North province in 
association with the Emergency North Recovery Project (ENRP) in 2013. These water supply 
projects included Nadunkemy, Vidathalathiv, Thevanpiddy, Adampan, Valvatithurai, 
Maruthankemy, Pandiyankulam, Mallavi and Oddusudan benefiting 51,220 people in 60 
Grama Niladhari divisions. Japan has granted USD 8.31 million for Kilinochchi Water 
Supply Scheme.
Work on several major irrigation projects has been commissioned and continuing. A large 
number of irrigation schemes including Rambakan Oya, abandoned Moragahakanda, Uma 
Oya and Deduru Oya are currently in progress.
In addition to above, following projects such as Yan Oya Project, Morana Reservoir,
Ellewewa Reservoir, Digili Oya Reservoir, Kalugal Oya Reservoir, Kubukkan Oya 
Reservoir, Talpitigala Reservoir, Lower Malwathuoya Multi sector Development Project, 
Redeemaliyadda Integrated Development, Wilakandiya, Kaudulla Stage 2 & NWP canal, 
Weli Oya Integrated Development Project, Weheragala, Kirindi Oya, Heda Oya and Uda 
Walawe left bank also have been commenced or completed in terms of new capacity creation.
11 facilitate cultivation activities specially in the abandon lands and domesticThese projects wi 
water supply.
20
Telecommunication Infrastructure Development projects
Private sector telecommunication companies in Sri Lanka has been invested in spectrum for 
4G mobile broadband to support the expansion of modem mobile broadband technology and 
4G services in the commercial environment. Further, the Government has focused in 
developing of a high speed national fiber backbone network to facilitate the requirements of 
high speed connectivity throughout the country.
1.5.7
Sri Lanka is connected to the South East Asia-Middle East-West Europe 4 (SEA-ME-WE 4) 
project, the submarine cable system approximately 20,000 km long connecting 17 countries 
from Europe to the Middle East and South East Asia. The total project cost has been 
estimated to be over USD 500 million. The project is aimed to significantly increasing the 
bandwidth and global connectivity of users and to provide a bandwidth capacity of 1.28 
terabits per second, with a 25 year guaranteed lifespan for the technology.
1.6 Construction related bodies in Sri Lanka
There are two main accredited and apex bodies in Sri Lanka established to look after various 
aspects of Construction Industry and to assist Government Policies. Objectives of these 
bodies are:
• Recommend strategies for the development of the Construction Industry.
• Facilitating business partnerships linkages between Construction Industry 
stakeholders.
• Regulate registration and grading of Construction Contractors.
• Promote professionalism and coordinate activities of professional bodies.
• Promote /Facilitate export of construction industrial services.
• Provide advisory services.
• Provision of information& publications.
• Assist in the provision of training facilities.
• Conduction of Seminars and Construction Industry Exhibitions.
• Promote /Undertake research on matters related to the Construction Industry.
• Promote Quality Assurance and productivity solutions.
21
The Institute of Construction Training & Development (ICTAD) -This institute 
up by the Government of Sri Lanka to develop and promote the domestic Construction 
Industry, Contractors, Professionals, Work Force, etc to create a reliable and globally 
competitive construction industry for Sri Lanka.
was set
Chamber of Construction Industry Sri Lanka (CCI) -This is the apex body of the 
construction industry that servicing the SME (small and medium sized enterprises) sector by 
facilitation, training of all types of craftsmen associated to the industry. Further it has a 
primary objective to focus on national interest and facilitating the development of the 
construction industry.
There are few big main construction companies in Sri Lanka. Some of them are owned by the 
government and rests of majority are belonging to private sector. The main State-owned 
construction companies are:
• State Engineering Corporation of Sri Lanka (SEC) - Established in 1962 and 
functioning under the Ministry of Construction & Engineering Services of Sri Lanka. 
Only government construction organization certified with ISO 9001:2008. As the 
premier engineering organization in Sri Lanka, State Engineering Corporation is 
engaged in the following discipline of Engineering such as Engineering Design, 
Construction. Manufacturing, Fabrication, Project Management and Information 
Management.
• Central Engineering Consultancy Bureau (CECB) -Leading, Highly Diversified, 
Multidisciplinary Engineering Consultancy and Construction Organization established 
in 1973 under the Ministry of Irrigation and Water Management. Currently CECB is 
engaged in four core business segments including Water Resources Development, 
Buildings, Water Supply and Sanitation and Transportation (Roads, Bridges, 
Railways, Tunnels, Ports and Airports)
. State Development and Construction Corporation (SD&CC)-This was established 
State organization under the State Industrial Corporation to undertake
works. SD&CC is now under the purview of
in 1971 as a
heavy civil engineering construction
Ministry of Construction, Engineering Services, Housing and Common Amenities.
22
This Corporation plays a major role in the development of Sri Lanka, specially in the 
up-liftment of Infrastructure Construction. Further, it has undertaken various major 
engineering projects in the fields of Construction such as bridges, roads, highways, 
dams, irrigation determent schemes, hydro power tunnels and power houses, water 
supply and treatment works, multi storied buildings.
Below are few big private sector construction and engineering services companies currently 
operating in Sri Lanka.
• Access Engineering PLC.
• ICC -International Construction Consortium (Pvt) Ltd.
• Maga Engineering (Pvt) Ltd.
• Nawaloka Construction Company (Pvt) Ltd.
• Sanken Construction (Pvt) Ltd.
• Tudawe Brothers (Pvt) Ltd.
1.7 Objectives of the Study
The objective of this study is to identify mathematical relationships between construction 
industry outputs and the national economy or vice versa. It is expected to identify any short 
term and long term associations between the chosen variables. Such model can be used as a 
very helpful tool by entrepreneurs who are interested to invest money in construction industry 
in Sri Lanka.
1.8 Scope of the Study
This study may help in economic policy development standpoint to evaluate the structures of
as well as direct and indirect impact to the socialgovernment policies, their effectiveness 
wellbeing in the country. It is focused on causality relationships between the developments of 
construction activities and the GDP Growth in Sri Lanka. Empirical data of economic
indicators and construction index were used to determine the Granger Causality Test for the
It was checked the associations between national economic statistics 
activates taken place in the country to identify unidirectional and
period of 1990 to 2013 
and construction
23
bidirectional relationships among the variables as well as short term and long term 
relationships.
This statistical hypothesis test was used to find the correlation between them. Therefore, it 
was tested on following relationships to determine the association between each of them.
• National Economy and Construction Activates of Sri Lanka does not have a causal 
relationship.
• National Economy has a unidirectional relationship towards Construction Activates in 
Sri Lanka. (Unidirectional - X to Y)
• Construction Activates has a unidirectional relationship towards National Economy in 
Sri Lanka. (Unidirectional - Y to X)
• National Economy causes Construction Activates and the same time Construction 
Activates also causes National Economy in Sri Lanka. (Bidirectional- X causes Y 
whilst Y causes X)
1.9 Significance of the Study
Many researchers have attempted to assess the relationship between the construction outputs 
and national economy both in developing and developed countries. They have used many 
different variables to determine association between national economy and construction 
sector. However, majority of researchers have assessed the relationships merely based on 
Granger Causality Test whilst some of them have used simple models to elaborate causality 
relationships further.
Hence, this study attempts to construct a model using statistical approach specially to find 
accurate and specific relationships among national economy and construction sector
to determine the return of their investment in the
more
outputs. This will help for entrepreneurs 
construction industry both in short and long run.
There are two main statistical tests were used in this dissertation namely Ganger Causality 
test and Vector Auto Regression model and assumption tests to validate Cointegration issues.
24
1.10 Limitations of the Study
The following can be considered as the limitations of this study.
• Since some of the variables are only available on annual basis (i.e. - Gross Fixed 
Capital Formation for construction sector) the annual data was used for this study. 
However, if more frequent data (i.e. - daily, monthly, and quarterly) could be 
obtained, the output would have been more precise and qualitative.
• The base year used to calculate GDP has been changed time to time. Hence GDP 
current (nominal) was selected to eliminate the time value of money based on the 
index calculation. As a result current value has been selected for rest of all applicable 
variables.
• Data from 1990 to 2013 has been used in the study due to lack of historical data 
beyond 1990 in certain variables, (i.e. - Gross Fixed Capital Formation for 
construction sector)
1.11 Organization of the Dissertation
This dissertation is comprised with five main chapters. The first Chapter describes a 
comprehensive introduction into this research. Also it provides in detail view about the 
country level and construction sector economic indicators and how those have been moved 
period of time. Also this chapter describes recent key infrastructure projects initiated 
and undertaken by the government of Sri Lanka. The second chapter includes the review and 
discussion of the literature. Also it contains an overview of construction sector main 
accredited and apex bodies in Sri Lanka. The third chapter explains the methodology to the 
study, theoretical concepts and their applications that is implemented in this study. The fourth 
chapter presents a comprehensive description of the analysis using the theories described 
above. The derived models are explained in there. Finally, the conclusion is included in 
chapter five. Also it describes recommendation and future work in to the study.
over a
25
Chapter 2 - Literature Review
2.1 Introduction
Many researchers have studied the relationship between construction activities and 
fiscals in the history. Also it was noted that, many of them have used various multivariate 
theories and tested the causality relationship among the variables to build a connection 
among construction industry and its activities as a subsector which can be used to describe 
the trends of national economy. The studies have been carried out with the empirical data of 
developed countries as well as developing countries using various measures of construction 
output and national economic statistics. This chapter is mainly intended to study about 
previous researches which had been carried out in relation to this topic. Further, it will 
explain the key differences among various studies with respect to geographical locations and 
the identified patters of relationships among different variables.
economic
2.2 Literature Review
It was noted that the previous studies for the construction industry is driven back to 1940s 
(i.e. - Simon, 1944; Phillip Report, 1950; Emmerson, 1962; Banwell, 1964; Great Britain, 
EDC for Building, 1967; Wood, 1975). After 1960s, most of the studies have begun focusing 
contribution by construction and building industry towards economic development 
activities, (i.e. - Bowley, 1966; Higgin G. and Jessop N. K., 1963 and 1965; Hillebrandt, 
1974, 1984, 1985). Subsequently, it has been paid more focus on theory building in the
into theoretical framework to assist economic
on
construction industry emphasizing more 
concepts.(i.e. - Edmonds, 1979;Ball M„ 1988; Wells, 1986; Miles, D. & Neale R. H„ 1991; 
Ofori G, 1993, 1994). Some of the other researchers have introduced systems ideas. These
ideas included thought processes such;
• Interrelatedness of human 
organizations (Tavistock, 1966).
. Napier (1970) discussed about Swedish construction industry enhancements through 
correlations with power, status, learning, boundaries, goal evaluation, innovation, and
elements and technological imperatives within
group values, etc.
• Ofori (1980)
developing countries.
identified eight factors for construction industry development in 
factors included Economic growth and stability,These
26
Government recognition, Indigenous
Development planning and policies, Codes and procedures, Use of local materials,
Education and training, Appropriate technologies, and Incentives for local 
contractors.
• Fox (1989) developed a causal model of 50 factors to explain the development of 
Hong Kong’s construction industry.
• Al-Omari (1992) considered the statistical technique of factor analysis deriving six 
key factors from a case study in Abu Dhabi. These factors included External 
environment, Indigenous environment, Development planning & policies, Planning 
prerequisites & measurement tools, Implementation strategies and Working 
environment.
• Tassios (1993) expressed that the ratio between structural components cost and the 
finishing components cost, reflect the level of construction development comparing 
both developing and developed countries.
environment, Planning and resources,
Some of the scholars have discussed certain important ideas and have been presented detail 
description through research papers. Those were used to gain a better understanding about;
• Our prime objective. This means to determine whether construction industry 
supports national economy or vice versa.
• Where does Sri Lanka stand in comparison to other developing countries within the 
region or outside the region.
A significant role of the construction industry in the national economy has been highlighted 
by Duccio Turin (1969). He has based on cross section of data from developing countries as 
well as industrial countries. His focus was to build a relationship between the output of the 
construction industry and the Gross Domestic product (GDP) and particularly the per capita 
of each. He argued about a positive relationship between construction output andmeasures
economic growth. Furthermore, he showed as economies grow construction output grows at a 
faster rate, assuming a higher proportion of GDP particularly in developing countries. Turin
much of the construction work take place in the(1969) suggested that this may be due to 
developing countries and these countries need importing of construction materials and skills.
accounts for a significantly lower proportion of GDP inIn the same token, capital formation 
most developing countries in 
higher income countries are more
in comparison to developed countries. He explained this trend as
willing and able to invest in fix capital formation.
27
Following the same view point the World Bank report (1984) indicated in general, 
construction activities tend to increase with the increase of a country's resource base and level 
off only after a high degree of economic development has been achieved. Among countries
that are members of the Organisation for Economic Co-operation and Development, the 
average share of construction during the 1970s ranged between 7 percent and 8percent of the 
GDP.
Subsequently, this point has been endorsed by few other scholars. Hillebrandt (1985) 
expressed that construction in any country is a complex sector of the economy, which 
involves a broad range of stakeholders and has wide ranging linkages with other areas of 
activity such as manufacturing and the use of materials, energy, finance, labor and 
equipment.
Similarly, Wells (1986) analyzed the statistical data of a large number of countries and 
demonstrated a positive relationship between GDP per capita and the three separate measures 
of construction's contributions namely, value-added by construction as a percentage of GDP, 
gross construction output as a percentage of GDP and construction employment as a 
percentage of total Economically Active Population. He has noted that all three contributions 
of construction increase with the increase of income. Further he has observed that the annual 
of both GDP per capita and value-added in construction per capita of rich countries 
is substantially higher than that of poor countries. So, he has concluded in the countries 
where construction demand is small, rational planning and production of basic construction 
materials appropriate for the size of the construction market is essential for development of
the construction industry.
increase
Bon (1992) discussed the changing role of the construction sector at the various stages of
since World War II in Finland,economic development. He studied the construction activity 
Ireland, Italy, Japan, the UK, and the USA. The data underlying his analysis spans a 50-year 
period and appears to place special emphasis on Europe. He argued that construction follows 
the bell-shaped pattern of development or an inverted U-shaped relationship. The Bon curve
relationship between the share of construction in output and(Figure 2.1) claims that a 
economic development.
0*
t
0*1
y>c-
■t-28 O
Share of 
Constriction 
in GNP
NICs
LDSs AICs
GNP Per Capita
Figure 7 - The Bon curve (Source: Bon 1992)
The inverted U-shaped relationship presented by Bon (1992) is very different from the S- 
shaped relationship found by Turin (1978).Turin,s analysis was mainly focused on 
developing countries. Bon’s 1992 argument concerns the entire path from LDC (least 
developed countries) to NIC (newly industrialised countries) to AIC (advanced industrial 
countries) status. The share of construction in total output first increases and then decreases 
with economic development, this is called the inverted U-shaped relationship.
Ofori (1993) enhanced his previous views in 1980 and suggested that construction industry 
development as ‘the deliberate and managed process to improve the capacity and 
effectiveness of the construction industry to meet the national economic demand for building 
and civil engineering products, and to support sustained national economic and social 
development objectives.5 In another hand he introduced that the following components: 
human resource development; materials development; technology development; corporate 
development; development of documentation and procedures; institution building; and 
development of operating environment of the industry.
Jin et al. (2003) found a non-linear relationship between the shares of construction output in 
GDP with the GDP per capita. Jin used the statistics across 34 countries and regions to
analyses this.
construction economics (Wells 1986; Field and OforiIn addition to above many studies on 
1988; Bon and Pietroforte 1990; Green 1997; Hillebmndt 2000; Lean 2001; Rameezdeen 
2007; Anaman and Amponsah, 2007; Myers 2008; Dlamini 2011) emphasize the important
29
role of the construction sector in national economic growth. They all argue that construction 
makes a noticeable contribution to the economic output of a country. However, still there are 
some opposing views also can be noted.
Authors such as Wang and Zhou (2000), Tan (2002), Hassan (2002), Kim (2004), and 
Dlamini (2011) all argue that the construction sector and its related activities are not drivers 
of economic growth.
Also it was noted that Thanuja and Raufdeen (2006) and Thanuja, James and Raufdeen 
(2013) have conducted a causality test with no mathematical modeling in the same area and 
the conclusion justifies a strong linkage between national economic growth and the 
construction industry activities.
30
Chapter 3 - Methodology
3.1 Introduction
This chapter describes the data and the methodology used in this study and the impact of the 
growth of the construction sector to the national economy.
3.2 Selection of Data
In this empirical analysis it is used basic set of variables consists of both national economy 
fiscals and construction sector annual figures from 1990 to 2013. Whilst National Economy 
Fiscals included variables such as Nominal Gross Domestic Product (GDP) and Balance of 
Trade (BOT); Construction Sector variables included Construction Gross Domestic Products 
(CGDP), Gross Fixed Capital Formation for Construction Sector (CGFCF) and All 
Construction Cost Index (ACINDEX). All the variables are expressed in natural logarithms 
so that they may be considered elasticity of the relevant variables. Annual observations of 
GDP, BOT, CGDP, CGFCF were extracted from data published by the Central Bank of Sri 
Lanka on the annual reports and All Construction Cost Index (ACINDEX) was extracted 
from The Institute of Construction Training & Development (ICTAD) bulletins.
3.3 Granger Causality
Clive J Granger (1969) introduced the Granger causality tests to analyze the effect of one 
time series on another one. He thought out of the box and said that ‘regressions’ does not 
only show ‘correlations’ but if certain tests are performed on them they may reveal 
information about causality. The idea of Granger causality is that a variable X Granger- 
courses variable Y can be better predicted using the histories of both X and Y than it can be 
predicted using the history of Y alone. This is shown if the expectation of Y given the history
unconditional expectation of Y. It was then widely used in
of X is different from the 
economics.
31
3.3.1 Definition of Granger Causality
We say that xt is Granger causal for yt with respect to Fx if the 
predictor of yt+h based on Fx has smaller
based on zt, zx.x... for any h. In other word xt is Granger causal for yt if xt helps predict yt at 
some stage in the future.
variance of the optimal linear
variance than the optimal linear predictor of yt-*-h
Often you will have that xt Granger causes yt and yt Granger causes xt. In this case we talk 
about a feedback system. Most economists will interpret a feedback system as simply 
showing that the variables are related (or rather they do not interpret the feedback system).
Sometimes econometricians use the shorter terms “causes" as shorthand for “Granger 
causes". You should notice, however, that Granger causality is not causality in a deep sense 
of the word. It just talks about linear prediction, and it only has “teeth" if one thing happens 
before another. (In other words if we only fmd Granger causality in one direction). In 
economics you may often have that all variables in the economy react to some un-modeled 
factor and if the response of xt and yt is staggered in time you will see Granger causality even 
though the real causality is different. There is nothing we can do about that (unless you can 
experiment with the economy) Granger causality measures whether one thing happens before 
another thing and helps predict it and nothing else. Of course we all secretly hope that it 
partly catches some “real" causality in the process. In any event, you should try and use the 
full term Granger causality if is not obvious what you are referring to.
The definition of Granger causality did not mention anything about possible instantaneous 
correlation between xt and yt. If the innovation to yt and the innovation to xt are correlated we 
say there is instantaneous causality. You will usually (or at least often) find instantaneous 
correlation between two time series, but since the causality (in the “real" sense) can go either
usually does not test for instantaneous correlation. However, if you do find Granger
feel that the case for “real" causality is stronger if 
innovations to each series can be thought
way, one
causality in only one direction you may
there is no instantaneous causality, because then the
enerated from this particular series rather than part of some vector
Of course, if your data is sampled with a long sampling
of as actually being g 
innovations to the vector system 
period, for example annually, then you would have to 
the other after such a long lag (you may 
depending on your application).
explain why one variable would only
have a story for that or you may not,
cause
32
Granger causality is 
be described by the model,
p rticularly easy to deal with in VAR models. Assume that our data can
Miyt AuAnA'n
A21A22A23
^A3\AnAi3j
yt-1 AUAV yt-k 
+ AUWL
ak ak ak ^A.3]A.32 A33 j
Hit
Mizt + 1Zt-1 U2tZt-k
Xt, M 3V / \ J \xt-i) U3t[Xt-k J
Also assume that,
S11Z12Z13
Z„ = Zl2Z22Z23
Z13 Z23 Z33
This model is a totally general VAR-model - only the data vectors has been partitioned in 
3 sub-vectors - the yt and the xt vectors between which can be tested for causality and the zt 
vector(That could be empty) which we condition on.
In this model it is clear that xt does not Granger cause yt with respect to the information set
,k or AU = 0; and A[2 = 0; i =generated by zt if either^{3-0; and^23~0; i - 1,
1,.. .k. Note that this is the way you will test for Granger causality. Usually you will use the 
VAR approach if you have an econometric hypothesis of interest that states that xt Granger
causes yt but yt does not Granger cause xt.
3.3.2 Limitations of Granger Causality
. If both X and Fare driven by a commonGranger causality is not necessarily true causality
with different lags, one might still accept the alternative hypothesis of Granger
causality. Yet, manipulation of one of the variables would not change the other. Indeed, the
of variables, and may produce misleading results
variables. A similar test involving more
third process
Granger test is designed to handle pairs 
when the true relationship involves three 
variables can be applied with vector auto regression.
or more
33
Vector Auto Regression (VAR) Analysis3.4
The vector autoregression (VAR) model is one of the 
use models for the analysis of multivariate time 
univariate autoregressive model to dynamic multivariate time series. A way to summarize the 
dynamics of macroeconomic data is to make use of Vector Auto Regressions. The VAR 
model has proven to be especially useful for describing the dynamic behavior of economic 
and financial time series and for forecasting. Therefore, VAR models have become 
increasingly popular in recent decades. They are estimated to provide empirical evidence on 
the response of macroeconomic variables to various exogenous impulses in order to 
discriminate between alternative theoretical models of the economy. It often provides 
superior forecasts to those from univariate time series models and elaborate theory-based 
simultaneous equations models. Forecasts from VAR models are quite flexible because they 
can be made conditional on the potential future paths of specified variables in the model.
most successful, flexible, and easy to 
series. It is a natural extension of the
In addition to data description and forecasting, the VAR model is also used for structural 
inference and policy analysis. In structural analysis, certain assumptions about the causal 
structure of the data under investigation are imposed, and the resulting causal impacts of 
unexpected shocks or innovations to specified variables on the variables in the model are 
summarized. These causal impacts are usually summarized with impulse response functions 
and forecast error variance decompositions. This simple framework provides a systematic 
way to capture rich dynamics in multiple time series, and the statistical toolkit that came with 
VARs was easy to use and to interpret.
In addition to measuring the broad correlation in the variables of a system, VAR helps us to 
the lead-lag relationships. VAR is commonly used for forecasting systems of 
interrelated time series and for analyzing the dynamic impact of random disturbances on the
VAR approach side steps the need for structural modeling by 
variable in the system as a function of the lagged values of all of 
The estimated VARs are used to calculate the
measure
system of variables. The 
modeling every endogenous 
the endogenous variables in the system 
percentages of each endogenous 
explanatory variables and pro 
innovation to the variable in the
++ ApYt-p + pXt +
variable that can be explained by innovations in each of the
vides information about the relative importance of each random 
VAR. The mathematical form of a VAR is
Yt = A1Yt-1
34
Where Yt is a k vector of endogenous variables, X, i 
Ap and (3 are matrices of coefficients to be estimated 
may vary contemporaneously.
is a d vector of exogenous variables, Aj, 
, and at is a vector of innovations that
Multiple Regression and Assumpti
Multiple regression is most effect at identifying relationship between a dependent variable 
and a combination of independent variables when its underlying assumptions are satisfied. In 
another words, it relies upon certain assumptions about the variables used in the analysis. 
When these assumptions are not met the results may not be trustworthy.
3.5 ons
These assumptions include
• The errors are normally distributed
• The mean of the errors is zero
• Errors have a constant variance
• The model errors are independent
Each metrics variables are normally distributed, the relationships between metric variables 
are linear, and the relationship between metric and dichotomous variables is homoscedastic. 
Failing to satisfy the assumptions does not mean that our answer is wrong. It means that our 
solution may under-report the strength of the relationships.
Outliers can distort the regression results. When an outlier is included in the analysis, it pulls
result in a solution that is more accurate for the 
cases in the data set. It shall be checked for
the regression line towards itself. This can
outlier, but less accurate for all of the other 
univariate outliers on the dependent variable and multivariate outliers on the independent
of satisfying assumptions and detecting outliers are intertwined. For 
value on the dependent variable that is an outlier, it will affect the
variables. The problems
example, if a case has a 
skew, and hence, the normality of the distribution. Removing an outlier may improve the
distribution of a variable. Transforming 
for a case will be characterized as an outlier.
variable may reduce the likelihood that the value
35
The order in which we check assumptions and detect outliers will affect our results because
we may get a different subset of eases in the final analysis. In otder to maximize the number
of cases available to the analysis, it shall be
evaluated assumptions first. It shall be substituted
any transformations of variable that enable 
transformed variables that are required in our analysis to detect
us to satisfy the assumptions. It shall be used any 
outliers.
Least Squares Method (LSM)3.6
The Least Squares Methods (LSM) is one of the very popular techniques in statistics due to 
following reasons.
• Most common estimators can be casted within this framework, (i.e — the mean of a 
distribution is the value that minimizes the sum of squared deviations of the scores.
• This method has recognized as very tractable as when the error is independent of an 
estimated quantity, it can add the squared error and the squared estimated quantity.
• Mathematical tools and algorithms are involved in LSM (derivatives, Eigen 
composition, singular value decomposition) have been well studied for a relatively 
long time.
The Method of Least Squares is a procedure to determine the best fit line to data; the proof 
simple calculus and linear algebra. The basic problem is to find the best fit straight line
y = ax + b given that, for n E {1,........ N}, the pairs (xn; yn) are observed. The method
easily generalizes to finding the best fit of the form;
uses
+ ckfk(x);y = a^Cx) +
it is not necessary for the functions fk to be linearly in x - all that is needed is that y is to be a 
linear combination of these functions.
define the error associated to saying,(xN;yN)}' we mayGiven data {(x1;* yi), 
y = ax + bby
N
E(a,b) = ^On - (axn + b))2
n=l
36
This is just N times the variance of the data set {yx - (aXl + b),
,yn - (axN + b)}. It
makes no difference whether
or not we study the variance or N times the variance as our
error, and note that the error is a function of two variables.
The goal is to find values of a and b that minimize the error. In multivariable calculus 
learn that this requires us to find the values of (a; b) such that,
dE
we
dE
db °
Therefore the final solution can be drawn as,
S
b =-^SJXX
a = y - bx
It is not necessary to worry about boundary points: as |a| and |b| become large, the fit will 
clearly get worse and worse. Thus it was not necessary to check on the boundary.
Preliminary Analysis3.7
Both graphical representations as well as descriptive statistics were comprehensively 
employed in examining the data properties.
Variables are depicted using scatter plots and line graphs which is useful in identifying 
characteristics of the series, detecting possible outliers, observations that could be used in the
determining the probable transformation to make the seriesmodeling process such as 
stationary etc.
is carried out under Chapter 4 for all the variables 
and special characteristics of the series. The 
of central tendency such as arithmetic mean, median
and mode and measures of dispersion such as variance, standard deviation and range which 
are widely known. Further, measures of distribution shape including skewness and kunosis 
which is normally a concern in most of the economic data series is also considered.
A comprehensive descriptive level analysis
selected, in order to identify the patterns
summary measures included measures
37
3.8 Skewness
Skewness is a measure of asymmetry of the distribution of. series around its arithmetic mean 
and computed by the formula,
HW)3 
1 = 1
Where d is an estimator for the standard deviation and n is the sample size.
The skewness of a symmetric distribution, such as the normal distribution, is zero. A positive 
value implies positive skewness of the distribution and a negative value implies negative 
skewness of the distribution.
3.9 Kurtosis
Kurtosis measures how sharply peaked the distribution is (peaked ness or flatness of the 
distribution) of a series and calculated as,
i=l
Where d an estimator for the standard deviation and n is is the sample size. The kurtosis of 
the normal distribution is 3. If the kurtosis exceeds 3, the distribution is peaked (leptokurtic) 
relative to the normal; if the kurtosis is less than 3, the distribution is flat (platykurtic) relative 
to Normal distribution.
3.10 Cointegration
If two or more series are themselves non-stationary, but a linear combination of them is
said to be cointegrated. That is if they share a commonstationary, then the series are 
stochastic drift.
spurious or nonsense regressions in timecauseNot concerning about co-integiation may 
series in modeling. Ms is where ihe usual procedure for resting hypotheses concerning the
relationship between non-sta,ionary variables was to run ordinary leas, scares,egress,ons on
38
data which had initially been differenced is i 
cointegrated.
is incorrect if the non-stationary variables are
Let Yt - (yi t> • ■ •, yn t)' denote an (
Yt is cointegrated if there exists an (n x ]) vector p =
PoYt = PiYi t + • • ■ + pnyn t ~ 1(0)
The intuition is that 1(1) time series with a long-run equilibrium relationship cannot drift too 
far apart from the equilibrium because economic forces will act to restore the equilibrium 
relationship. Here the individual series are first-order integrated that is 1(1) but some 
cointegrating vector of coefficients exists to form a stationary linear combination of them. 
Therefore I such cases Error Correction Models recommended when it comes to the modeling 
process.
1) vector of 1(1) time series.n x
(Pi/---,Pny such that
3.11 Detecting Cointegration - Johansen Cointegration Test
Johansen Co-integration test is a technique for testing cointegration of several time series. 
This test permits more than one cointegrating relationship.
It was selected a VAP (p)
+ Apy( p + e,yt-\i+Axyt ,+
Where y, is an n x 1 vector of variables that are integrated of order one commonly denoted 
1(1) and et is an n x 1 vector of innovations. This vector auto regression can be re-written as,
pA
2 r,Av +&Ay,=p+n y +t l
i=i
Where,
II = ^Ai = / And 1/ - Aj •
j
then there exist nxr matrices a and P each 
stationary. R is the number of co-integrating
If the coefficient matrix II has reduced rank r 
with rank r such that n=aP' and P y* IS
39
relationships, the elements of a known as the adjustment parameters
a co~integrating vector .It can be shown that for a
are in the vector error
correction model and each column of (3 is 
given r, the maximum likelihood estimator of (3 defines the combination of yt-i that yields the 
of Ayt with yM after correcting for lagged differences andr largest canonical correlations
deterministic variables when present. Johansen theo
ry proposes two different likelihood ratio 
of these canonical correlations and thereby the reduced rank of the ITtests of the significance
matrix.
3.12 Trace Test
H0 : Cointegration rank is less than or equal to r
H0 * There is m cointegrating relations” (i.e., the series are stationary)
Where r = 0,1,....., m - 1.
T 2ln(l- jj,,)Where J
i/ trace
/=r+l
Where T = sample size and ^ is the largest canonical correlation.
Maximum Eigen value statistic test
H0: Number of cointegration vectors equal to r 
H0: Number of cointegration vectors equal to r + 1
where
^max ~ Tl7l(l ^r+l)^ ^
likelihood estimates of the parameters in a vector error-The tests also produce maximum 
correction (VEC) model of the cointegrated series.
3.13 Selection of the lag length
nsidering the VAR lag order selection criteria of the
The optimum lag length is arrived by co 
lag structure. The available criterias are
40
• Sequential modified LR
• Final prediction error
• Akaike information criterion
The Akaike Information 
smaller values are Preferred. It is computed as,
test statistic
Criterion (AIC) is used in model selection processes and
AIC = -2-+ — 
n n
Where l is the Log likelihood and k is the number of parameters in the model and n i; 
the sample size.
• Schwarz information criterion
This is an alternative to the Akaike Information Criterion that imposes a larger 
penalty for additional coefficients. It is calculated as,
l _ k log (n)
SC = -2 —+n n
Smaller values are preferred in this criterion also.
• Hannan-Quinn information criterion
l 2k log (log (n))
HQ= -2- +
7171
This criterion also imposes another penalty function for unimportant inclusions to the model. 
Commonly selected lag by the available techniques 
the model while taking more emphasis on Akaike
criterion and Hannan-Quinn information criterion which are widely employed.
is considered as the best possible lag of
information criterion, Schwarz information
41
Chapter 4 - Analysis of Data
4.1 Introduction
In this chapter the methodology elaborated in chapter 3 is implemented to the selected 
variables described under notation section. Gross Domestic Product (GDP) and Balance of 
Trade (BOT) have been selected to reflect the national economic statistics whilst 
Construction Sector of Gross Domestic Product (CGDP), Gross Fixed Capital Formation for 
Construction Sector (CGFCF) and All Construction Cost Index (ACINDEX) have been 
selected to reflect construction outputs. In order to understand the characteristics of the
variables the graphical as well as summary measures were studied at first. Then the series 
was tested to determine whether the assumptions are met and the given regressions are 
explained our objective reflecting any unidirectional bidirectional relationship between 
construction outputs and national economic statistics. A keen attention was paid to safeguard 
the characteristics and real behavior of row data. Hence, transformations deemed unnecessary 
as it results to compromise the details and smoothness of variables particularly used in this 
analysis. Then the residual tests were tested for unit root, serial correlation, heteroscedasticity 
and multivariate normality to validate the model. Finally, the conclusion was arrived based 
on the test results.
consist of 24 data elements from year 1990 to 2013. TheIt was obtained a sample
information such as CGFCF is only published annually. Hence, for higher accuracy and
reliability, the stud, was based on published annual dam b, the respective national authorities
, , . , . observations of GDP, BOT, CGDP, CGFCF wereduring the mentioned period above. Observ
tral Bank of Sri Lanka on the annual reports fromextracted from data published by the Cen 
2009 - 2013 and All Construction Cost Index (ACINDEX) 
of Construction Training & Development (ICTAD) bulletins.
extracted from The Institutewas
4.2 Notation
Let,
Abbr: GDPic Product (Nominal)• Gross Domestic
• Construction Sector of Gross Dome
• Balance of Trade - Abbr: BOT
Stic Product (Nominal) - Abbr: CGDP
42
. Gross Fixed Capital Formation for Const 
. All Construction Cost Index
ruction Sector - Abbr: CGFCF
- Abbr: ACINDEX
43
Graphical Representation of the Vari4.3
ables
plot of each series was checked 
plot the graphs.
parately to identify the trend. E-views have been used to
GDP
CGDP
10,000.000
1.000,000
8,000.000-
800,000-
6,000,000-
600.000-
4,000,000- 400.000-
2,000,000- 200,000-
® i i i i ■ i i i i i i i i i i i i i—i—i—i—i—r~
90 92 94 96 98 00 02 04 06 08 10 12 P. I ( i i I I i i i I I I i i fill I I i T90 92 94 96 98 00 02 04 06 08 10 12
BOT CGFCF
0 2,000,000
1,600,000--2,000-
1,200,000--4,000-
800,000--6,000-
400,000--8,000-
°90T^^4 ‘ 96'98'00 02 ‘ 04'06 ‘ 08 ‘ 10'12-10,000 90'92'94'96'98 ‘ 00'02'04 06'08'10 12
ACINDEX
120" j j |j j i i i i ^
90 92 94 96 98 00
variablesFigure 8 - Time series line plots of the
44
When observing the scattered plots of the vari 
with their trends indicating possible consist
anables, all of them are fairly concentrated along 
ence in variance except BOT.
Testing for Ganger Causality4.4
The Granger Causality was «ed for the Mowi„g ^ ^ N„u Hypothcs|s M ^
ppeared m different lag levels (1 to 7) depicted below. Howe 
combinations where national
versa.
a
ver, it was considered only the 
economy can be compared with construction sector and vice 
It has been selected Gross Domestic Product (GDP) and Balance of Trade (BOT) to 
express the national economic statistics whilst Construction Sector of Gross Domestic 
Product (CGDP), Gross Fixed Capital Formation for Construction Sector (CGFCF) and All 
Construction Cost Index (ACINDEX) have been chosen to reflect construction sector outputs. 
Then the associations between the variables were tested
In this study, only the combinations of variables were considered from the opposing sector 
and the combinations of variables from the same sector were omitted. Hence following 
relationships have been used to analyse and express the causality relationship between GDP 
& CGDP, GDP & CGFCF, GDP & ACINDEX, BOT & CGDP, CGFCF & BOT and 
ACINDEX & BOT. It has been disregarded the combinations that indicate relationships 
between the same sector variables such as GDP & BOT, CGFCF & CGDP, ACINDEX & 
CGDP and CGFCF & ACINDEX.
Table 5 - Causality Test in different Lag Levels between GDP and CGDP
GDP Does Not Cause CGDPrr.np Does Not Cause GDPLag ProbabilityF-StatisticProbabilityF-Statistic 0.8337
0.9493
0.04527
0.05222
0.2796
0.4791
1.23513
0.76871
1
2 0.97740.065220.00855.820953 0.06493.037790.01335.208694 0.02284.98107
19.9189
8.3356
0,0229
0.05244.97484
4.83161
5 0.0024
6 0.11130.09310.10357
st and 2nd lags of CGDP
45
an unidirectional relationship from GDP 
relationship observed that t0 in lag 6. However, there ivambte GDP o, CGDP depend 
Therefore this was not examined this father as i
relationship between these two variables to fb
is no recent 
on resent past data (lag 1 or 2) 
it would not able to identify any short term 
recast any trend.
Table 6 - Causality Test in different Lag Le,e„ between GDP turd CGFCF
CGFCF Does Not Caiisp QDPLag GDP Does Not Cause CGFCFF-Statistic Probability F-Statistic Probability1 1.32122
0.61141
0.2639 6.71212 0.01752 0.5541 4.20032 0.03293 0.92367 0.4549 1.68143 0.21654 1.72011 0.2154 1.88107 0.184
5 5.43141
7.09323
17.4166
0.0179 2.38259 0.1319
6 0.0242 3.07376 0.1192
7 0.0554 5.06661 0.1747
It was noted no causal relationship exists among first four lag levels from CGFCF to GDP. 
Although, there is a unidirectional relationship from CGDP to GDP in lag 5 and 6, still this 
was not examined further as it would not able to identify any long or short term relationship 
from CGFCF to GDP in causality. However there was a unidirectional relationship from GDP 
to CGFCF both in lag 1 and 2. Hence, we examined this relationship further.
Table 7 — Causality Test in different Lag Levels between GDP and CGFCF
GDP Does Not Cause ACINDEXACINDEX Does Not Cause GDPLag ProbabilityF-StatisticProbabilityF-Statistic
0.93090.007710.19741.777961
0.77880.253750.43382 0.87748
0.6650.536320.03143.939063 0.02444.308230.01784.764584 0.069
0.3587
3.22285
1.42001
0,0717
0.1033
3.16916
3.34224
5
6 0.055717.30990.05467 17.6715
lationship from ACCINDEX to GDP in lag 3 & 4 and 
in lag 4, still this was not examined 
relationship among these two
Although, there is a unidirectional re
unidirectional relationship from 
fata, as it would not able to identify any k>"8 or short tern,
GDP to ACCINDEX
variables.
46
Table 8 - Causality Test in different Lag LeveIs between BOT and CGDP
—^IDowNolCaus^CDP
r-Statistic
Lag
_ CGDP Does Not Cause BOT
F-Statistic____  Probabilil
7,99639 
19.158 
3.07867 
1.25566 
~~ 1.86712
1.2015
Probabilil
1 14,5153
6.20009
6.80489
3.57876
4.32161
3.42797
0.0011
0.0095
0.0046
0.0104
0.00004
0.0621
0.3442
0.2061
2
3
4 0.0425 0.0334
0.09886 0.42957 7.59064 0.1213 2.37194 0.3284
Table 4.4 indicates productive and meaningful bi-directional 
CGDP and CGDP to BOT. At the outset this indicates that there
among variables BOT and CGDP in short run. Hence, it was tested further to understand long 
term relationship.
causal relationship from BOT to 
would be a relationship
Table 9 - Causality Test in different Lag Levels between CGFCF and BOT
CGFCF Does ^ot Cause BOT BOT Does Not Cause CGFCFLag
F-Statistic Probability F-Statistic Probability
1 16.6491 0.0006 0.88755 0.3574
0.5615 0.58060.000012 24.5656
0.02174.437290.01273 5.20794
0.09142.637590.20914 1.75006
0.05743.483460.09785 2.75479
0.06514.312710.08786 3.6645
0.09919.442050.2183.92563
There is a positive short term unidirectional relationship from CGFCF to BOT in lag 1 to 3. 
investigated further to understand any
Although there is a unidirectional relationship from BOT to CGFCF in lag 3, this 
examined further as i. would not able to identify any long or short tenn teladonship here.
n ACINDEX and BOT
mathematical long term relationship.
was not
This was
Table 10 — Causality Test in different Lag Levels betv
ACINDEX Does Not Cause_BQT 
Probabili
BOT Does Not Cause ACINDEX 
F-Statistic ProbabililLag
F-Statistic 0.42320.668540,0067
0,0006
0,1987
0,8584
0.6097
9.132461 0.2441.53437
0.0518
0.1424
0.1949
11.75642 3.29993
2.14952
1.9292
3 1.7717
0,32062
0.74817
4
5
47
6 0.43009
4J190T 0.8332
0.2092
7 0.64395
0.98016
0.6981
0.5915
A short term unidirectional 
and 2. No causal relationship can be noted vi
causal relationshi•P exists between ACINDEX and BOT in lag 1
ice versa from BOT to ACINDEX.
4.5 Summary Measures of the variables
Table 11 - Summary Measures
GDP CGDP BOT CGFCF ACINDEXMean 2559349. 201638.3 -2930.529 401720.9 268.9344
Median 1494642. 97730.50 -1548.650 170657.0 202.4125
Maximum 8673870. 894683.0 -702.5000 1631404. 590.4250
Minimum 321784.0 21541.00 -9710.000 35239.00 100.0000
Std. Dev. 2456862. 231775.7 2666.122 464442.0 154.9524
Skewness 1.164400 1.667581 -1.531855 1.382934 0.748164
Kurtosis 3.186333 5.006492 4.114979 3.740262 2.116075
Jarque-Bera 5.458027 15.14931 10.62950 8.198015 3.020321
Probability 0.000513 0.004919 0.016589 0.2208750.065284
As the values of the variable are in different scales it is less meaningful in comparing 
summary measures of the variables.
When considering the shape of the distributions, according to the Jarque-Bera statistics all the 
variables do not look normally distributed.
Ho: Distribution is Normally Distributed 
Hi: Distribution is not Normally Distributed
bability is less than 0.05 at 5% significance.
almost close to zero indicating their all
Reject the null hypothesis when pro
Further, skewnesses of the variable distributions are 
the variables are normally distributed.
48
4.6 Vector Autoregression Estim
We tested Vector Autoregression Estimates for diffe 
relationship equation explaining its 
model variables.
Since at 5% significance level, the value of the test statistics 
reject the null hypothesis.
ates
rent lag levels in each variable to build 
its own lags and the lags of the other
a
evolution based on
are greater than 0.05 we do not
4.6.1 VAR Estimate - Lag 1
Table 12 - VAR Estimate (Lag 1)
Vector Autoregression Estimates 
Date: 05/16/14 Time: 20:04 
Sample (adjusted): 1991 2013 
Included observations: 23 after adjustments 
Standard errors in () & t-statistics in [ ]
GDP CGDP BOT CGFCF ACINDEX
GDP(-l) 1.203189 
(0.19856) 
[ 6.05965]
0.067967 
(0.02606) 
[ 2.60788]
1.002809 
(0.17733) 
[ 5.65501]
-11.67808
(3.64348)
[-3.20520]
-0.131513
(0.11735)
[-1.12065]
-515.4916
(157.638)
[-3.27009]
26978.38 
(14641.3) 
[ 1.84262]
0.997348
0.996568
-0.002142
(0.00176)
[-1.21774]
0.073310 
(0.05711) 
[ 1.28365]
4.85E-05 
(2.6E-05) 
[ 1.86845]
0.004218 
(0.38859) 
[ 0.01085]
5.96E-05
(0.00018)
[0.33741]
0.037204
(0.01197)
[3.10777]
-0.106635
(1.35102)
[-0.07893]
CGDP(-l)
-0.004714
(0.00363)
[-1.29996]
-7.584921
(7.98397)
[-0.95002]
-0.029640
(0.24596)
[-0.12051]
-0.016547
(0.00792)
[-2.08870]
12.50492 
(10.6418) 
[ 1.17507]
-2000.654
(988.401)
[-2.02413]
0.908264
0.881283
21.71853 
(27.7583) 
[ 0.78242]
-0.236662
(0.89407)
[-0.26470]
446.6763
(1200.99)
[0.37192]
-56824.98
(111546.)
[-0.50943]
0.998616
0.998208
BOT(-l)
-0.000268
(0.00012)
[-2.29825]
0.748119 
(0.25716) 
[ 2.90919]
-3.093559
(345.434)
[-0.00896]
-31637.81
(32083.6)
[-0.98611]
CGFCF(-l)
0.932789 
(0.15690) 
[ 5.94524]
ACINDEX(-l)
-2.283689
(14.5724)
[-0.15671]
C
0.993958
0.992181
0.996826
0.995893^■squared 
Mi- R-squared
49
Sum sq. resids 
S.E. equation 
F-statistic 
Log likelihood 
Akaike AIC 
Schwarz SC 
Mean dependent 
S.D. dependent
1.85E+U
104308.7
2452.551
-294.9269
26.16756
26.46377
2656634.
2464360.
3.19E+09
13691.30
1278.758
-248.2232
22.10637
22.40258
209468.7
233716.4
14522620 
924.2682 
33.66288 
-186.2264 -266.2666
16.71534 
17.01156 
-3027.400 
2682.509
1.53E+10 3156.763
30001.79 13.62688
1067.827 559.3125
-89.23638 
23.67536 8.281425
23.97158 8.577640
417654.9 276.2793
468124.8 154.1039
Determinant resid covariance (dof 
adj.)
Determinant resid covariance 
Log likelihood 
Akaike information criterion 
Schwarz criterion
7.54E+32
1.66E+32
-1016.375
90.98915
92.47023
It is noted AIC (Akaike information criterion) in lag 1 is 90.99 and SIC (Schwarz information 
criterion) is 92.47. It shall be chosen the lag length that minimizes AIC and SIC for the VAR 
model. Therefore it was tested the second lag length of the VAR (Vector Autoregression) 
model and then test the correlations of residuals.
4.6.2 VAR Estimate - Lag 2
Table 13 - VAR Estimate (Lag 2)
Vector Autoregression Estimates 
Date: 05/16/14 Time: 20:10 
Sample (adjusted): 1992 2013 
Included observations: 22 after adjustments 
Standard errors in () & t-statistics in [ ]
CGFCF ACINDEXBOTCGDPGDP
0.254942 0.000107-0.0081170.103286
10 12471) (0.00851) (0.24380) (0.00013)
[0.82819] [-0.95398] [ 1.04570] [0.80656]
0.008920
(0.00745) (0.21342) (0.00012)
[ 1.19761] [-0.94598] [-0.22098]
-0.060623 3.84E-05
1.376091 
(1.09535) 
[ 1.25630]
GDP(-l)
-0.201886 -2.57E-05-0.029637-0.453766 
(0.95884) (0-109H)
[-0.47325] [-0.27148]
1.516740
GDP(-2)
-0.027805
tO 04342) (1.24410) (0.00068)
[-0.64036] [-0.04873] [ 0.05657]
-1.429155 -0.000952/ '
<58S
[0.87328] [ 2.38329J
-1.829311 
(1.02932)
CGDP(-l)
(01)7023) (2.01221) (0.00110) '-6.036992
(9.04052)
CGDP(-2)
Vi
\V50 V*
[-0.66777]
108.7569
[-1.77720] [0.97076] 
-1.090161
[-0.71024] [-0.86669]
BOT(-l)
U.36710] [-0 094001 n'Si (17'7067) (0.00967)
L y4°°] [-1-76406] [0.78757] [0.68532]
76.57792 1.632175
(84.0353) (9.56799)
[0.91126] [0.17059]
13.94529 0.006624
BOT(-2)
-0.856162 -16.75054
(0.65280) (18.7043) (0.01021)
[-1.31151] [-0.89554] [0.18064]
-2.529798 -0.356966 0.007776
r 131ml r(?'?i<56) (°-01498) (0-42921) (0.00023)
[ -31187] [-1.62583] [0.51908] [0.02554] [-1.00227]
,4'5.49544 0.731724 -0.041732 1.545865 0.000329
(2.60046) (0.29608) (0.02020) (0.57880) (0.00032)
[ 1.74952] [2.47137] [-2.06587] [2.67080] [ 1.04251]
3791.441 -242.0887
(5031.06) (572.820) (39.0824) (1119.80) (0.61128)
[0.75361] [-0.42263] [0.16487] [0.61315] [2.25625]
0.001844
CGFCF(-l)
0.010963 -0.000235
CGFCF(-2)
ACINDEX(-l) 6.443437 686.6023 1.379196
ACINDEX(-2) -2090.701 -87.12927 -6.415008 -691.1034 -0.704977
(6771.39) (770.969) (52.6016) (1507.16) (0.82273)
[-0.30876] [-0.11301] [-0.12195] [-0.45855] [-0.85687]
-16759.30 26474.00 -2716.995 -20001.55 32.10987
(323555.) (36838.9) (2513.45) (72015.9) (39.3122)
[-0.05180] [ 0.71864] [-1.08098] [-0.27774] [0.81679]
0.998573 0.995976
0.997275 0.992318
6.67E+09 1986.678
24618.86 13.43901
769.5173 272.2499
-80.75159 
23.36727 8.341054
23.91279 8.886575
434899.7 283.8193
471604.3 153.3265
C
0.947266 
0.899326 
8121025.
859.2293 
19.75935 
-172.2248 -246.0399
16.65680 
17.20232 
-3119.686 
2708.007
0.998504
0.997144
1.74E+09
12593.50
734.0740
-231.2926
22.02660
22.57212
217874.7
235630.7
0.998950
0.997995
1.35E+11
110608.3
1046.393
-279.0945
26.37223
26.91775
2760466.
2470322.
R-squared 
Adj. R-squared 
Sum sq. resids 
S.E. equation 
F-statistic 
Log likelihood 
Akaike AIC 
Schwarz SC 
Mean dependent 
S.D. dependent
Determinant resid covariance (dof
adj-)
Determinant resid covariance 
Log likelihood 
Akaike information criterion 
Schwarz criterion
1.98E+31
6.18E+29
-910.6413
87.78557
90.51318
51
Subsequent lag length indicates 
lag 2 is a better modification. However, 
the optimal level.
a minimizati n of A1C to 87.78 and SIC to 90.51. Therefore
11 Was checked the third lag length also to determine
4.6.3 VAR Estimate - Lag 3
Table 14 - VAR Estimate (Lag 3)
Vector Autoregression Estimates 
Date: 05/16/14 Time: 20:11 
Sample (adjusted): 1993 2013 
Included observations: 21 after adjustments 
Standard errors in () & t-statistics in [ ]
GDP CGDP BOT CGFCF ACINDEX
GDP(-l) 3.479916 0.481880 -0.020749 0.568011 0.000313
(1.04949) (0.13531) (0.01162) (0.17954) (0.00012)
[3.31582] [3,56140] [-1.78615] [3.16378] [2.58861]
-2.795751 -0.296971 0.026988 -0.463770 -0.000346
(0.72258) (0.09316) (0.00800) (0.12361) (8.3E-05)
[-3.86910] [-3.18776] [3.37431] [-3.75181] [-4.16299]
GDP(-2)
2.220802 0.203470 -0.016464 0.212530 0.000328
(0.65113) (0.08395) (0.00721) (0.11139) (7.5E-05)
[ 3.41070] [2.42377] [-2.28429] [ 1.90801] [4.37848]
-2.752454 -0.001998
GDP(-3)
-1.172407 0.108677-14.71995
(5.63501) (0.72650) (0.06237) (0.96398) (0.00065) 
[-261223] [-1.61377] [ 1.74236] [-2.85530] [-3.07882]
0.002025
CGDP(-l)
-0.136078 1.9054851.890045
Cl 20954) (0.10384) (1.60492) (0.00108) 
[ 1.56261] [-1.31039] [ 1.18728] [ 1.87482]
-0.003503
18.40258 
(9.38165) 
[ 1.96155]
CGDP(-2)
;31„87847 'altm (°02S (iS (0-00108)
[-3,39136] [-3.99H2] [2.18849] [-3.49654] [-3.23658]
0.004236
CGDP(-3)
22.51751-1.223029 
(0 65508) (10.1242) (0.00681)
[-1.86700] [2.22413] [0.62171]
0 979886 -39.11709
(093648) (14.4732) (0.00974)
[ L04635] [-2.70272] [-3.18353]
0.014354
4.69837194 46554
-31.60142
BOT(-l)
-0.031011
(8496043) (^077)
[-2.35953] [-2.89716]
BOT(-2)
41.10457-0.928631-4.72664558.58385BOT(-3)
52
[0.63880] [-0399761 /n!'513) 05.6888) (0.01056)
0.39976] [-0.91479] [2.62000] [ 1.35941]
1.637064 0.114167
0-38149) (0.17811)
[ 1.18500] [0.64099]
0.691693 0.403583
(1-54640) (0.19937)
[0.44729] [2.02427]
2.529342
(3.01340) (0.38851) (0.03336)
[ 0.83937] [ 0.29998] [-0.37603]
-5538.957 -1627.047
(5148.39) (663.763) (56.9872) (880.734) (0.59277)
[-1.07586] [-2.45125] [0.85188] [-1.51256] [0.17884]
44.00800 -715.5223
(8190.13) (1055.92) (90.6560) (1401.08) (0.94299)
[0.00537] [-0.67763] [0.18571] [0.56920] [-0.24916]
-2793.643 38.12013 -11.34743 -587.3804 -0.172266
(4485.93) (578.354) (49.6544) (767.407) (0.51650)
[-0.62276] [0.06591] [-0.22853] [-0.76541] [-0.33353]
179380.2 -7302.263 115130.0 105.7062
CGFCF(-l)
-0.027313 
(0.01529) (0.23633) (0.00016)
[-1 -78615] [ 4.63084] [ 2.61845]
-0.015273 
(0.01712) (0.26454) (0.00018)
[-0.89228] [ 1.84481] [-0.74670]
-0.012543
1.094410 0.000416
CGFCF(-2)
0.488031 -0.000133
CGFCF(-3) 0.116545 0.993829 -7.81E-05
(0.51550) (0.00035)
[ 1.92789] [-0.22509]
ACINDEX(-l) 48.54597 -1332.166 0.106009
ACINDEX(-2) 16.83539 797.4952 -0.234959
ACINDEX(-3)
702294.7
(462760.) (59662.0) (5122.26) (79164.3) (53.2809) 
[ 1.51762] [3.00661] [-1.42559] [ 1.45432] [ 1.98394]
0.988952 
0.955808 
1651538.
574.7239 
29.83821 
-148.1610 
15.63438 
16.43021 
-3218.500 
2733.938
C
0.999913 0.999616
0.999650 0.998463
3.94E+08 178.6931
8882.331 5.978179
3813.817 867.2633
-205.6575 -52.27976
21.11024 6.502834
21.90607 
453323.3 
475068.8
0.999801
0.999206
2.24E+08
6694.144
1678.373
-199.7181
20.54458
21.34041
226893.3
237527.1
0.999890
0.999560
1.35E+10
51922.24
3027.443
-242.7369
24.64161
25.43743
2871665.
2474264.
R-squared 
Adj. R-squared 
Sum sq. resids 
S.E. equation 
F-statistic 
Log likelihood 
Akaike AIC 
Schwarz SC 
Mean dependent 
S.D. dependent
7.298661
291.6857
152.4960
determinant resid covariance (dof
adj-)
Determinant resid covariance 
Log likelihood 
Akaike information criterion 
Schwarz criterion
1.20E+26
9.21E+22
-704.1970
53
It was noted that the relative quality of Akaik 
values in the third lag level. Those h 
lag values.
« (AIC -
ave been minimized
74.68) and Schwarz (SIC - 78.66) 
and optimized compared to first two
4.7 Proposed Models
Based o. .he values chained tough the above VAR ^ reMom
were identified. However these relationships to be tested further
against model assumptions.
Model 1 -
GDP — 3.31582 GDP( 1) — 3.86910 GDP(-2) + 3.41070 GDP(—3)
- 2.61223 CGDP(-l) + 1.96155 CGDP(-2) - 3.39136 CGDP(-3)
- 2.35953 B0T(—2)
Model 2 -
CGDP = 3.56140 GDP(-l) - 3.18776 GDP(-2) - 2.42377 GDP(-3)
- 3.99112 CGDP(-3) - 2.89716 B0T(-2) + 2.02427 CGFCF(-2)
- 2.45125 ACINDEX(-l) + C
Here, C = 3.00661
Model 3 -
BOT = 3.37431 GDP(-2) - 2.28429 GDP(-3) + 2.18849 CGDP( 3)
Model 4 -
- 3 .75181 GDP(—2) - 2.85530 CGDP(-l)
2.22413 B0T(-1) - 2.70272 BOT(-2)
4.63084 CGFCF(-l)
CGFCF = 3.16378 GDP(-l)
- 3.49654 CGDP(-3) + 
+ 2.62000 B0T(-3) +
Model 5 -
„ .,-QQ rnp(-2) + 4.37848 GDP(-3)
4 W 923658 CGDP(-3) - 3,8353 BOT(-2,ACINDEX = 2.58861 GDP(-l) "
+ 3.07882 CGDP(-l) - 
+ 2.61845 CGFCF(-l) + C
Here, c = 1.98394
54
4.8 Testing of assumptions
Series of tests were undertaken to determine 
violate one or more of multiple linear 
not met the results may not be trust
whether above mentioned multiple regressions 
regression assumptions. When these assumptions
Y' ^ome the key assumption were tested as below.
are
• Variables are Normally distributed.
• Assumption of a Linear Relationshi 
Variable(s).
• Variables are measured without error (Reliably)
• Assumption of Homoscedasticity
p between the Independent and Dependent
4.8.1 Model 1
In model I, we noted the GDP has the following relationship.
GDP = 3.31582 GDP(-l) - 3.86910 GDP(-2) + 3.41070 GDP(-3)
- 2.61223 CGDP(-l) + 1.96155 CGDP(-2) - 3.39136 CGDP(-3)
- 2.35953 B0T(—2)
It shows the current year GDP is linked to the previous year GDP, CGDP and BOT.
Further analysis was undertaken. Since at 5% significance level, the value of the test statistics 
greater than 0.05, we do not reject the null hypothesis.are
Model 1.1 -
the significance of the lagged variables
It was used a lagged regression model to test 
identified in the VAR model 1.
for Model 1 (1st Iteration)
Table 15 - Least Squares estimate
Dependent Variable: GDP 
Method: Least Squares 
Date: 06/12/14 Time: 16:21
Sample (adjusted): 1993 2013
Included observations: 21 after adjus^m^^
Std. Error
86033.50
Prob.t-Statistic
.1,441611
CoefficientVariable 0.1731
-124026.8C
55
GDP(-l)
GDP(-2)
GDP(-3)
CGDP(-l)
CGDP(-2)
CGDP(-3)
B0T(-2)
1-314008
-1-232275
1-677201
-3.598568
11.49616
-15.26034
-68.99373
0.364276
0.748251
0.556039
2.497096
4.543061
4.067904
78.24270
3.607174
-1.646874
3.016337
-1.441101
2.530487
-3.751400
-0.881791
0.0032
0.1235
0.0099
0.1732
0.0251
0.0024
0.3939
^squared 0.999292 Mean dependent ™
S.E. of regression 816«iS Akait^ZZn
Sum squared resid 8.67E+10 Schwarz criterion
Log likelihood -262.2762 Hannan-Quinn criter.
F-statistic 2621.857 Durbin-Watson stat
Prob(F-statistic) 0.000000
2871665.
2474264.
25.74059
26.13850
25.82695
2.281746
The following variables were noted 
GDP(-l), GDP(-3), CGDP(-2) and CGDP(-3).
as significant during the lagged regression model 1.1.
Model 1.2 -
Based on the output, of model 1.1, GDP was regressed again with GDP(-l) GDP(-3) CGDP(- 
2) and CGDP(-3)
Table 16 - Least Squares estimate for Model 1.1 (2nd Iteration)
Dependent Variable: GDP 
Method: Least Squares 
Date: 06/12/14 Time: 16:25 
Sample (adjusted): 1993 2013
Included observations: 21 after adjustments _____ _
Prob.Std. Error t-StatisticCoefficientVariable
4.569668 0.0003
2.569405 0.0199
1.697494 0.1078
-2.679911 0.0158
2871665. 
2474264. 
25.82582 
26.02477 
25.86899
0.847234 0.185404
0.716986 0.279047
4.955498 2.919302
-9.960046 3.716558
GDP(-l)
GDP(-3)
CGDP(-2)
CGDP(-3)
Mean dependent var 
S D. dependent var
Akaike info criterion
^■-squared
Adjusted R-squared 0.998673
S-E. of regression 90142.58 rriterion
S«m squared resid 1.38E+11 Schwarz cr ^
J;°8 likelihood -267.1711, Hannan-Q 
urbin-Watson stat pL.88492^ *
0.998872
56
The following variables were noted as si 
CGDP(-3).
S1gnificant from model 1.2. GDP(-l), GDP(-3), and
lyTndel 1.3 -
Based on the output, of model 1.2, GDP 
CGDP(-3)
was regressed again with GDP(-l) GDP(-3) and
Table 17 - Least Squares estimate for Model 1.2 (3rd I
Dependent Variable: GDP
Method: Least Squares
Date: 06/12/14 Time: 16:30
Sample (adjusted): 1993 2013
Included observations: 21 after adjustments
teration)
Variable Coefficient Std. Error t-Statistic Prob.
GDP(-l)
GDP(-3)
CGDP(-3)
1.065655 0.140287 7.596239
0.453002 0.243504 1.860347
-4.515170 1.972940 -2.288548 0.0344
0.0000
0.0793
R-squared
Adjusted R-squared 
S.E. of regression 
Sum squared resid 
Log likelihood 
Durbin-Watson stat
0.998681 Mean dependent var 2871665. 
0.998534 S.D. dependent var 
94736.68 Akaike info criterion 25.88715 
1.62E+11 Schwarz criterion 
-268.8151 Hannan-Quinn criter. 25.91954 
2.017691
2474264.
26.03637
The following variables were noted as significant from model 1.3. GDP(-l) and CGDP(-3).
Model 1.4-
Based on the output, of model 1.3, GDP was regressed again with GDP(-l) and CGDP(-3).
Table 18 - Least Squares estimate for Model 1.3 (4th Iteration)
Dependent Variable: GDP 
Method: Least Squares 
Date: 06/29/14 Time: 15:55 
Sample (adjusted): 1993 2013 
Included observations: 21 after adjustments
Std. Error
0.088068 
1.632085
Meandependent var
Prob.t-Statistic
14.49149
-1.354626
CoefficientVariable 0.0000
0.19141.276242
-2.210864
GDP(-l)
^CGDP(-3)
Squared
2871665.
0.998427
57
Adjusted R-squared 0.998344
S.E. of regression 100685.1
Sum squared resid 1.93E+11
Log likelihood -270.6616
Durbin-Watson stat 2.296031;
S-P-dependent var 
Akatkemfo criterion 25 96778
Schwarz criterion 26 06 25 
Hannan-Quinn criter.
2474264.
25.98937
At the end it has been received a rela<i„„shlp betwee„ QDp >nd GDp
previous ,ear. This concluded «ha. ,be c„M GDP depend „„ly « „ p„viou5 year GOP 
and does not have a relationship with any other construction 
relationship was disregarded.
1 which is the
output parameters. Hence, this
4.8.2 Model 2
In model 2, it was noted that the CGDP has the following relationship.
CGDP = 3.56140 GDP(-l) - 3.18776 GDP(-2) - 2.42377 GDP(-3)
- 3.99112 CGDP(-3) - 2.89716 B0T(-2) + 2.02427 CGFCF(-2)
- 2.45125 ACINDEX(-l) + C
Further analysis was undertaken. Since at 5% significance level, the value of the test statistics 
are greater than 0.05, we do not reject the null hypothesis.
Model 2.1 -
Based on the output, of VAR model 2, CGDP was regressed again with GDP(-l), GDP(-2), 
GDP(-3), CGDP(-3), BOT(-2), CGFCF(-2) and ACINDEX(-l).
Table 19 - Least Squares estimate for Model 2 (1st Iteration)
Dependent Variable: CGDP 
Method: Least Squares 
Date: 06/12/14 Time: 16:13 
Sample (adjusted): 1993 2013 
Included observations: 21 after adjustmen s
Std. Error
9925.878 
0.017682 
0.031069 
0.029970 
0.285563 
2.114318
Prob.t-Statistic
10.34744 
18.32022 
-6.992929 
5 928900
.10.81191 
.10.36373 !
CoefficientVariable
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
102707.5
0.323930
-0.217260
0.177691
-3.087483
-21.91222
C
GDP(-l)
GDP(-2)
GDP(-3)
CGDP(-3)
BOT(-2)
58
CGFCF(-2)
ACINDEX(-l)
0-542623
-1360.750
0.069937
101.7535 7.758750 -13.37301
Adjusted R-squared 0.999387 var
S.E. of regression 5880.578 Akaike^f^ V3r 
Sum squared resid 4.50E+08 Schwarz
Miwihood " hJsss*
0.000000 Durb,"-Wa's»”® 
All the variables concludSini^S3^"
0.0000
0.0000jAsquared
226893.3 
237527.1 
n°n 20.47903 
20.87694 
20.56539
3.016022
F-statistic
Prob(F-statistic)
assumption test was undertaken.
Testing of Stationarity of the model residuals with A
Ho: Residuals are non-stationary (There is a unit root) 
Hi: Residuals are stationary (There is no unit root)
ugmented Dickey-Fuller test
Table 20 - Augmented Dickey-Fuller test for model 2.1
Null Hypothesis: RES2 has a unit root 
Exogenous: None
Lag Length: 1 (Automatic based on SIC, MAXLAG=4)
t-Statistic Prob.*
-6.387454Augmented Dickey-Fuller test statistic
Test critical values: 1% level
5% level 
10% level
-2.692358
-1.960171
-1.607051
^MacKinnon (1996) one-sided p-values.
Warning: Probabilities and critical values calculated for
observations
and may not be accurate for a sample size of
Augmented Dickey-Fuller Test Equation 
Dependent Variable: D(RES2)
Method: Least Squares 
Date: 06/29/14 Time: 16:35
Sample (adjusted): 1995 2013
included observations: 19 after adjustments^
Std. Error
Prob.t-Statistic
-6.387454
2.512285
Variable Coefficient
RES2(-1) 
aaaJXRES2(-l))
Squared
0.0000
0.02240.359566 
0.206786
^Td^ndent var
-2.296711
0.519505
0.821985
-15.34085
59
Adjusted R-squared 0.811514 
S.E. of regression 
Sum squared resid 
Log likelihood 
Durbin-Watson stat 1.967183
S.D. dependent*... . . 8676.206
AKaike info criterion 19.40513 
Schwarz criterion 
Hannan-Quinn criter.
var3766.777 
2.41 E+08 
-182.3487 19.5045419.42195
P value < 0.05 we reject H0
Since at 5% significance level, the p-value of the Augmented Dickey-Fuller test statistic 
(0.0000) is less than 0.05
stationary.
reject the null hypothesis. Therefore, the residuals arewe
Serial Correlation - Correlogram of Residuals (Q statistics)
Ho: No serial correlation exists in residuals 
Hi : There exists serial correlation in residuals
1Date: 06/12/14 Time: 16:18 
Sample: 1993 2013 
Included observations: 21
AC PAC Q-Stat ProbAutocorrelation Partial Correlation
1 -0.510 -0.510 6.2703 0.012
2 -0.123 -0.516 6.6534 0.036
3 0.333 -0.047 9.6245 0.022
4 -0.392 -0.404 13.997 0.007
5 0.333 0.045 17.348 0.004
6 -0.180 -0.272 18.390 0.005
7 0.057 0.159 18.503 0.010
8 0.060 -0.148 18.635 0.017
9 -0.273 -0.154 21.642 0.010
10 0.355 -0.086 27.179 0.002
11 -0.166 0.022 28.512 0.003
12 -0.121 -0.223 29.292 0.004
C
=3
I
CE= □l
Cl
cc=
[
c cc
Figure 9 - Correlogram of Residuals for model 2.1
concluded that there exists serialless than 0.05 it wasSince some of the p-values 
correlation between residuals.
are
Breusch-Godfrey Serial Correlation LM Test
Ho: No serial correlation exists in residuals
in residualsHi : There exists serial correlation
60
Tabk 21 - Breusch-Godfrey Serial Correlation LM T 
Breusch-Godfrey Serial Correlation LM Test:
estfor model 2.1
F-statistic
Obs*R-squared
11.20109
14.08428
Prob. F(2,l 1)
Prob. Chi-Square(2)
0.0022
0.0009
Test Equation:
Dependent Variable: RESID 
Method: Least Squares 
Date: 06/12/14 Time: 16:21 
Sample: 1993 2013 
Included observations: 21
Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
C 2869.090 6252.025 0.458906
0.013343 0.013172 1.012991
0.016753 0.024809 0.675262
-0.046913 0.021945 -2.137717
0.253073 0.190368 1.329385
0.433999 1.338486 0.324246
-0.037874 0.044632 -0.848583
-19.70464 63.85451 -0.308586
-1.121400 0.243080 -4.613288
-0.808731 0.283000 -2.857701
0.6552
0.3328
0.5135
0.0558
0.2106
0.7518
0.4142
0.7634
0.0007
0.0156
GDP(-l)
GDP(-2)
GDP(-3)
CGDP(-3)
BOT(-2)
CGFCF(-2)
ACINDEX(-l)
RESID(-l)
RESID(-2)
0.670680 Mean dependent var -1.46E-10
4741.073
R-squared
Adjusted R-squared 0.401237 S.D. dependent var
3668.633 Akaike info criterion 19.55878
20.05617S.E. of regression Sum squared resid 
Log likelihood 
F-statistic
1.48E+08 Schwarz criterion 
-195.3672 Hannan-Quinn criter. 19.66673 
2.489130 Durbin-Watson stat 
0.078051
1.693209
Prob(F-statistic)
Breusch-Godfrey Serial Correlation LMSince at 5% significance level, the p-value of the
is lesser than 0.05 the null hypothesis was rejected.Test: statistic (0.0022) is
, this was notserial correlation among residuals. Therefore
That means this relationship has 
investigated further.
4-8.3 Model 3
In model 3, it was noted that the BOT has
the following relationship.
61
BOT = 3.37431 GDP(—2) - 2-28429 GDP(—3) + 2.18849 CGDP(-3)
Further analysis was undertaken. Si 
are greater than 0.05, we do not reject the
nee at 5 ^ significance level, the value of the test statistics 
null hypothesis.
Model 3.1 -
Based on the output, of VAR model 3, BOT 
CGDP(-3).
was regressed with GDP(-2), GDP(-3) and
Table 22 - Least Squares estimate for Model 3 (1st Iteration)
Dependent Variable: BOT
Method: Least Squares
Date: 05/16/14 Time: 20:22
Sample (adjusted): 1993 2013
Included observations: 21 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
C -125.8388 428.0259 -0.293998 0.7723
0.005463 0.002124 2.572653 0.0198
-0.008555 0.003181 -2.689744 0.0155
0.007538 0.026119 0.288601 0.7764
GDP(-2)
GDP(-3)
CGDP(-3)
0.898872 Mean dependent var -3218.500
2733.938
R-squared
Adjusted R-squared 0.881026 S.D. dependent var
943.0050 Akaike info criterion 16.70566 
15117393 Schwarz criterion 
-171.4095 Hannan-Quinn criter. 16J4884 
50.36816 Durbin-Watson stat PB8BBI 
0.000000
S.E. of regression 
Sum squared resid 
Log likelihood 
F-statistic
16.90462
Prob(F-statistic)
significant during the lagged regression model 3.1.The following variables were noted as 
GDP(-2) and GDP(-3).
Model 3.2 -
, BOT was regressed with GDP(-2) and GDP(-3)
Based on the output, of model 3.1
Table 23 - Least Squares estimate for Model 3.1 (2nd Iteration)
Dependent Variable: BOT 
Method: Least Squares 
Date: 06/12/14 Time: 15:44 
Sample (adjusted): 1993 2013 
Included observations: 21 after adjustmen
62
Variable Coefficient
0.005543 0.002034
-0.008140 0.002368
Std. Error t-Statistic Prob.
GDP(-2)
GDP(-3) 2.724796 0.0134
-3.438068 0.0028
R-squared 0.895906 Mean dependent var -3218 500
Adjusted R-squared <3890428 S.D. dependent vat 2733 938
S.E. of regress,on 904.9805 Akaike info criterion 16.54410
Sum squared resid 15560803 Schwarz criterion
Log likelihood -171,7130 Hannan-Quinn criter
Durbin-Watson stat fl.752762
16.64357
16.56569
It was identified that both the variables were significant.
Testing of Stationarity of the model residuals with Augmented Dickey-Fuller test
H0: Residuals are non-stationary (There is a unit root) 
Hi : Residuals are stationary (There is no unit root)
Table 24 - Augmented Dickey-Fuller test for model 3.2
Null Hypothesis: RES has a unit root 
Exogenous: None
Lag Length: 0 (Automatic based on SIC, MAXLAG=4)
t-Statistic Prob.*
mm-3.883453Augmented Dickey-Fuller test statistic
-2.685718
-1.959071
-1.607456
1% levelTest critical values:
5% level 
10% level
♦MacKinnon (1996) one-sided p-values.
Augmented Dickey-Fuller Test Equation 
Dependent Variable: D(RES)
Method: Least Squares 
Date: 06/12/14 Time: 15:49
Sample (adjusted): 1994 2013 
Included observations: 20 after adjustments
Coefficient
-0.980827 0.252564
Prob.Std. Error t-Statistic
Variable
-3.883453 0.0010
RES(-l)
116.4395
1192.150Mean dependent var S.D. dependent var
0.436910R-squared 
Adjusted R-squared 0.436910
63
S.E. of regression 
Sum squared resid 
Log likelihood 
Durbin-Watson stat 1.817203
15205251i « Schwarz criterion 
163.7930 Hannan-Quinn criter.
16.47930
16.52908
16.48902
P value < 0.05 we reject H0
Since at 5% significance level, 
(0.0005) is less than 0.05 we 
stationary.
the p-value of the Augmented Dickey-Fuller test statistic 
reject the null hypothesis. Therefore, the residuals are
Serial Correlation - Correlogram of Residuals (Q statistics)
H0: No serial correlation exists in residuals 
Hi : There exists serial correlation in residuals
Date: 06/12/14 Time: 15:46 
Sample: 1993 2013 
Included observations: 21
AC PAC Q-Stat ProbAutocorrelation Partial Correlation
1 -0,001 -0.001 2.E-05 0.996
2 0.063 0.063 0.1017 0.950
3 0.010 0.011 0.1046 0.991
4 0.232 0.229 1.6330 0.803
5 -0.394 -0.418 6.3300 0.275
6 -0.141 -0.170 6.9710 0.324
7 -0.269 -0.285 9.4725 0.220
8 -0.079 -0.132 9.7029 0.287
9 -0.154 0.074 10.656 0.300
10 -0.064 -0.168 10.838 0.370
11 0.015 0.049 10.849 0.456
12 0,181 -0.001 12.603 0.399
l]
Zl=]
c=
cc
dcz
c[
]c
c[
I
□
Figure 10 - Correlogram of Residuals for model 3.2
spike could be observed lay beyond the confidence intervals. This 
greater than the significance level of 0.05. Hence, then; is
As per the correlogram no 
was confirmed by all p values 
no serial correlation in the residuals of the model.
are
Breusch-Godfrey Serial Correlation LM Test 
Ho: No serial correlation exists in residuals
64
H, : There exists serial correlation in residuals
Table 25 - Breusch-Godfrey Serial Correlation LM T 
Breusch-Godfrey Serial Correlation LM Test:
est for model 3.2
F-statistic
Obs*R-squared
0.110073
0.065147
Prob. F(2,17)
Prob. Chi-Square(2) 0.9680
Test Equation:
Dependent Variable: RESID 
Method: Least Squares 
Date: 06/12/14 Time: 16:05 
Sample: 1993 2013 
Included observations: 21
Presample missing value lagged residuals set to zero.
Variable Coefficient Std. Error t-Statistic Prob.
GDP(-2)
GDP(-3)
RESID(-l)
RESID(-2)
0.000956 0.002952 0.323686 0.7501
-0.001101 0.003419 -0.321946 0.7514
0.014643 0.284464 0.051476 0.9595
0.183232 0.395700 0.463059 0.6492
0.003102 Mean dependent var -84.83256
877.7720
950.5995 Akaike info criterion 16.72171 
15361870 Schwarz criterion 
-171.5779 Hannan-Quinn criter. 16.76488 
1.765551
R-squared
Adjusted R-squared -0.172821 S.D. dependent var 
S.E. of regression 
Sum squared resid 
Log likelihood 
Durbin-Watson stat
16.92066
5% significance level, the p-value of the Breusch-Godfrey Serial Correlation LM
do not reject the null hypothesis. Therefore,
Since at
Test: statistic (0.8964) is greater than 0.05 
there is no serial correlation among residuals.
we
65
Correlogram Squared Residuals
Date: 06/12/14 Time: 1 SiST" 
Sample: 1993 2013 
Included observations: 21
Autocorrelation Partial Correlation AC PAC Q-Stat Prob
1 0.057 0.057 0.0780 0.780
2 -0.174 -0.177 0.8440 0.656
3 -0.140 -0.122 1.3690 0.713
4 0.066 0.052 1.4928 0.828
5 0.391 0.359 6.1022 0.296
6 -0.041 -0.085 6.1566 0.406
7 0.045 0.205 6.2277 0.513
8 -0.098 -0.072 6.5878 0.582
9 -0.103 -0.132 7.0119 0.636
10 -0.072 -0.250 7.2399 0.703
11 -0.083 -0.105 7.5706 0.751
12 -0.036 -0.256 7.6398 0.813
C c
c c
] J
[
] □
c [
c c
c n=
[ :
i=
Figure 11 - Correlogram Squared Residuals for model 3.2
As per the correlogram of squad residuals, no spike could be observed lay beyond the 
confidence intervals. This was confirmed by the all the p values are greater than the 
significance level of 0.05. Hence, there is no squared serial correlation in the residuals of the 
model.
White’s Heteroskedasticity
H0: No Heteroskedasticity exists in residuals 
H, : There exists Heteroskedasticity in residuals
Table 26 - White’s Heteroskedasticity test for model 3.2
Heteroskedasticity Test: White
6.2039
0.1826
0.1431
1.704572 Prob. F(3,17)
Prob. Chi-Square(3) 
Prob. Chi-Square(3)
F-statistic 
Obs*R-squared 
Scaled explained SS 5.426489
4.856172
Test Equation:
Dependent Variable: RES1DA2 
Method: Least Squares 
Date: 06/12/14 Time: 15:59
66
Sample: 1993 2013 
Included observations: 21
Variable Coefficient Std. Error
t-Statistic Prob.
C 453473.9 317528.1
-1.37E-05 1-428138 0.17140.3197 
0.3113 
0.3046
GDP(-2)A2 
GDP(-2)*GDP(-3) 3.23E-05
GDP(-3)A2
1.34E-05 -1.025139 
3.10E-05 1.043662
1.79E-05 -1.058576-1.89E-05
R-squared 0-231246 Mean dependent var 
Adjusted R-squared 0.095584 S.D. dependent var 
S.E. of regression 1193121. Akaike info criterion 
Sum squared resid 2.42E+13 Schwarz criterion 
Log likelihood -321.4127 Hannan-Quinn criter. 31 03487
F-statistic 1.704572 Durbin-Watson stat
0.203866
740990.6
1254586.
30.99169
31.19064
2.202519
Prob(F-statistic)
Since at 5% significance level, the p-value of the White’s Heteroskedasticity Test (0.2039) is 
greater than 0.05 we do not reject the null hypothesis. Hence, there is no squared serial 
correlation in the residuals of the model.
Normality Test (Jarque-Bera Test)
8 Series: Residuals 
Sample 1993 2013 
Observations 217-
6- -84.83256
-145.4721
1736.390
-2305.723
877.7720
-0.255947
3.643344
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
5-
4-
3-
2- Jarque-Bera 0.591437 
Probability 0.7439971-
0 1000
_________ -2000 __ ________
Figure 12 - Jarque-Bera test for model 3.2
Ho: Residuals are normal 
Hj: Residuals are not normal
67
Since at 5% significance level, the p-value of the J 
0.05 we do not reject the null hypothesis. Th arque-Bera Test (0.591437) is greater than 
erefore, residuals are normal.
4.8.4 Model 4
In model 4, iit was noted that the CGFCF has the following relationship.
CGFCF= 3.16378 GDP(-l) - 3 .75181 GDP(-2) - 2.85530 CGDP(-l)
3.49654 CGDP(—3) + 2.22413 BOT(-l) - 2.70272 BOT(-2) 
+ 2.62000 BOT(-3) + 4.63084 CGFCF(-l)
Since at 5% significance level, the value of the test statistics are greater than 0.05 we do not 
reject the null hypothesis.
Model 4.1 -
Based on the output, of VAR model 4, CGFCF was regressed with GDP(-l), GDP(-2), 
CGDP(-l), CGDP(-3), BOT(-l), BOT(-2), BOT(-3) and CGFCF(-l).
Table 27 - Least Squares estimate for Model 4 (1st Iteration)
Dependent Variable: CGFCF
Method: Least Squares
Date: 06/12/14 Time: 16:48
Sample (adjusted): 1993 2013
Included observations: 21 after adjustments
Prob.Coefficient Std. Error t-StatisticVariable
0.233340264.34 1.255171
0 100733 0.455100 0.6572
0.140642 0.241130 0.8135
0.480564 -1.750981 
0.882093 -1.185661
50538.63 
0.045844 
0.033913 
-0.841458 
-1.045863 
12.38537 13.99309
-0.563484 12.81848
55.19842 19.37538
1.740107 0.355327
C
GDP(-l)
GDP(-2)
CGDP(-l)
CGDP(-3)
BOT(-l)
BOT(-2)
BOT(-3)
CGFCF(-l)
0.1054 
0.2587 
0.885106 0.3935
-0.043959 0.9657
2.848895 0.0147
4.897199 0.0004
453323.3
475068.8
23.06836
23.51602
23.16551
0.998794 Mean dependent var 
0^997990 S.D. dependent var 
21300.36 Akaike info criterion 
5.44E+09 Schwarz criterion 
-733 2178 Hannan-Quinn criter.
Durbin-Watson stat
R-squared 
Adjusted R-squared 
S.E. of regression 
Sum squared resid 
Log likelihood 
F-statistic 1242.095
68
Prob(F-statistic) 0.000000
Only BOT(-3) and CGFCF(-l) was significant
among the considered variables.
Model 4.2 -
Based on the output, of model 4.1 CGFCF was regressed again with BOT(-3) and CGFCF(-
1).
Table 28 - Least Squares estimate for Model 4.1 (2nd Iteration)
Dependent Variable: CGFCF
Method: Least Squares
Date: 06/12/14 Time: 16:49
Sample (adjusted): 1993 2013
Included observations: 21 after adjustments
Variable Coefficient Std. Error t-Statistic Prob.
BOT(-3)
CGFCF(-l)
13.42734 8.118632 1.653892 0.1146 
1.252350 0.037088 33.76698 0.0000
0.995774 Mean dependent var 453323.3
475068.8
R-squared
Adjusted R-squared 0.995551 S.D. dependent var 
S.E. of regression 
Sum squared resid
Log likelihood ______
Durbin-Watson stat 2.330218
31686.54 Akaike info criterion 23.65556 
1.91 E+l 0 Schwarz criterion 
-246.3834 Hannan-Quinn criter. 23.67715
23.75504
Only CGFCF(-l) was significant out of the considered variables.
Model 4.3 -
Based on the output, of 4.2, CGFCF was regressed again with CGFCF(-1).
estimate for Model 4.2 (3rd Iteration)Table 29 - Least Squares
Dependent Variable: CGFCF 
Method: Least Squares 
Date: 06/12/14 Time: 16:50
Sample (adjusted): 1991 2013 
Included observations: 23 after adjustments
Coefficient 
1.194746 0.012682
Prob.Std. Error t-Statistic
Variable
94.20515 0.0000
CGFCF(-l)
417654.9Mean dependent var0.995469R-squared
69
Adjusted R-squared 0.995469 
S.E. of regression 
Sum squared resid 
Log likelihood
23K sss-270 3596 ^chwarz criterion 23.64585
Durbin-Watson stat 2.814910 nan'Qumn criter- 23.60890
This results also pretty much similar to the 
means Construction Gross Fixed Capital Format!
year CGFCF value. This concluded that current CGFCF depend only on the previous year 
CGFCF and does not have a relationship with an, other national economic parameter,.
results which were received in the 1st model. That
(CGFCF) depends only on the previousion
4.8.5 Model 5
In model 5, it was noted that the ACINDEX has the following relationship.
ACINDEX = 2.58861 GDP(-l) - 4.16299 GDP(-2) + 4.37848 GDP(-3)
+ 3.07882 CGDP(-l) - 3.23658 CGDP(-3) - 3.18353 BOT(-2)
+ 2.61845 CGFCF(-l) + C
Since at 5% significance level, the value of the test statistics are greater than 0.05 we do not 
reject the null hypothesis.
Model 5.1 -
Based on the output, of VAR model 5, ACINDEX was regressed with GDP(-l), GDP(-2), 
GDP(-3), CGDP(-l), CGDP(-3), BOT(-2) and CGFCF(-l).
Table 30 - Least Squares estimate for Model 5 (1st Iteration)
Dependent Variable: ACINDEX 
Method: Least Squares 
Date: 06/12/14 Time: 16:51 
Sample (adjusted): 1993 2013 
Included observations: 21 after adjustments
Prob.Std. Error t-StatisticCoefficientVariable
7.188275 0.0000
7.396763 0.0000
-2.664242 0.0195
4117436 0.0012
-4.968231 0.0003
-8.110277 0.0000
-3.198533 0.0070
1.971038 0.0704
66.05422 9.189162
0.000275 3.71E-05
-0.000176 6.59E-05
0.000203 4.92E-05
-0.001251 0.000252
-0.002752 0.000339
-0.025065 
0.000193
C
GDP(-l)
GDP(-2)
GDP(-3)
CGDP(-l)
CGDP(-3)
BOT(-2)
CGFCF(-l)
0.007836
9.81E-05
70
R-squared 
Adjusted R-squared 0.997073
8.250676 
884.9575 
-69.07839 
974.1878 
0.000000
0.998097 Mean dependent var 291.6857 
T?’.dependent var 152.4960 
Akaike info criterion 7.340799 
Schwarz criterion 
Hannan-Quinn criter.
Durbin-Watson stat
S.E. of regression 
Sum squared resid 
Log likelihood 
F-statistic
7.738712
7.427156
2.058199Prob(F-statistic)
Out of the regression the lagged variables 
CGDP(-3), and BOT(-2) were significant.
of GDP(-l), GDP(-2), GDP(-3), CGDP(-l),
Model 5.2 -
Based on the output, of model 5.1, ACINDEX 
CGDP(-l), CGDP(-3) and BOT(-2).
was regressed with GDP(-l), GDP(-2),
Table 31 - Least Squares estimate for Model 5.1 (2nd Iteration)
Dependent Variable: ACINDEX 
Method: Least Squares 
Date: 06/12/14 Time: 16:52 
Sample (adjusted): 1993 2013 
Included observations: 21 after adjustments
Coefficient Std. Error t-Statistic Prob.Variable
0.0000
4.01E-05 7.199364 0.0000
7.24E-05 -2.464380 0.0273
5.29E-05 3.453221 0.0039
_ 0.0010 
0.000288 -8.079840 0.0000
0.008580 -2.781380 0.0147
60.17156 9.544545 6.304288 
0.000288 
-0.000178 
0.000183 
-0.001059 0.000255 -4.152937 
-0.002327 
-0.023864
C
GDP(-l)
GDP(-2)
GDP(-3)
CGDP(-l)
CGDP(-3)
BOT(-2)
291.6857
152.4960
7.507037
7.855211
7.582599mm
R-squared 0.997529 Mean dependent var
Adjusted R-squared 0.996470 S.D. dependent var
9.060996 Akaike info criterion 
1149.423 Schwarz criterion 
-71.82389 Hannan-Quinn criter. 
941.8227 Durbin-Watson stat 
0.000000 ____ _
S.E. of regression 
Sum squared resid 
Log likelihood 
F-statistic
Prob(F-statistic)
All the variables were significant in the regression. 
Therefore the model was reconstructed as below.
71
ACINDEX = 60.17156 + 0-000288 GDP(-i) _ 
0-000183 GDP(—3) _
0.000178 GDP(—2)
0-001059 CGDP(-l) _ 0.002327 CGDP(-3)
+
~ 0.023864 BOT(-2)
Testing of Stationary of the model residuals with A
Ho: Residuals are non-stationary (There is a unit root) 
H, : Residuals are stationary (There is no unit root)
Table 32 - Augmented Dickey-Fuller test for model 5.2
Null Hypothesis: RESS has a unit root 
Exogenous: None
Lag Length: 0 (Automatic based on SIC, MAXLAG=4)
ugmented Dickey-Fuller test
t-Statistic Prob.*
Augmented Dickey-Fuller test statistic -3.983851 &.0004
Test critical values: 1% level
5% level 
10% level
-2.685718
-1.959071
-1.607456
* MacKinnon (1996) one-sided p-values.
Augmented Dickey-Fuller Test Equation 
Dependent Variable: D(RESS)
Method: Least Squares 
Date: 06/12/14 Time: 16:54 
Sample (adjusted): 1994 2013 
Included observations: 20 after adjustments
Prob.Coefficient Std. Error t-StatisticVariable
0.00080.221402 -3.983851-0.882031RESS(-l)
-0.628209
10.08396
6.905333
6.955120
6.915052
R-squared 0.452910 Mean dependent var
Adjusted R-squared 0.452910 S.D. dependent var
S.E. of regression 7.458653 Akaike info criterion
Sum squared resid 1056.999 Schwarz criterion
Log likelihood -68.05333 Hannan-Quinn enter.
Durbin-Watson stat 2.058722
P value < 0.05 we reject Ho
1 nf the Augmented Dickey-Fuller test statistic 
significance Therefore the residuals aie stationary.Since at 5%
(0.0004) is less than 0.05 we
72
Serial Correlation - Correlo
gram of Residuals (Q statistics)
Date: 06/12/14 Time: 16:52 
Sample: 1993 2013 
Included observations: 21
Autocorrelation Partial Correlation AC PAC Q-Stat Prob
1 0.116 0.116 0.3277
2 0.022 0.009 0.3400
3 -0.008 -0.012 0.3418
4 -0.283 -0.285 2.6175 0.624
5 -0.039 0.029 2.6623 0.752
6 -0.034 -0.025 2.6986 0.846
7 0.064 0.081 2.8379 0.900
8 -0.009 -0.117 2.8407 0.944
9 -0.285 -0.308 6.1038 0.729
10 -0.105 -0.074 6.5852 0.764
11 -0.215 -0.178 8.8252 0.638
12 -0.098 -0.098 9.3369 0.674
3 3
0.567
0.844
0.952d d
] ]
C
d d
c [
c c
c [
Figure 13 - Correlogram of Residuals for model 5.2
Ho: No serial correlation exists in residuals 
Hj : There exists serial correlation in residuals
As per the correlogram no spike could be observed lay beyond the confidence intervals. This 
confirmed by the all the p values are greater than the significance level of 0.05.
Hence, there is no serial correlation in the residuals of the model.
was
Breusch-Godfrey Serial Correlation LM Test
Ho: No serial correlation exists in residuals 
Hi : There exists serial correlation in residuals
Table 33 - Breusch-Godfrey Serial Correlation LM Test for model 5.2
Breusch-Godfrey Serial Correlation LM Test:
0,7454
0.60540.301211 Prob. F(2,12)Prob. Chi-Square(2)F-statisticObs*R-squared 1.003845
Test Equation:
Dependent Variable: RESID 
Method: Least Squares 
Date: 06/12/14 Time: 16:53
73
Sample: 1993 2013
Included observations: 21
Presample missing value lagged residual
Variable
s set to zero.
Coefficient Std. Error
t-Statistic
-0.014021 
-0.531136 
0.424461 
0.092083 
0.228398 
-0.000230 0.000456 -0.504876
0.001660 0.009399 0.176573
0.419591 0.550101
-0.048538 0.540794 -0.089753
Prob.
C -0.141107 10.06374
-3.18E-05 5.99E-05
4.21E-05 9.91E-05
6.02E-06 6.54E-05
6.47E-05 0.000283
0.9890
0.6050
0.6787
0.9282
0.8232
0.6228
0.8628
0.4603
0.9300
GDP(-l)
GDP(-2)
GDP(-3)
CGDP(-l)
CGDP(-3)
BOT(-2)
RESID(-l)
RESID(-2) 0.762753
R-squared 0.047802 Mean dependent 
Adjusted R-squared -0.586996 S.D. dependent var 
S.E. of regression 
Sum squared resid 
Log likelihood 
F-statistic
var 1.44E-13 
7.580973
9.550211 Akaike info criterion 7.648531 
1094.478 Schwarz criterion 
-71.30957 Hannan-Quinn criter. 7.745683 
0.075303 Durbin-Watson stat 2.067602 
0.999458
8.096183
Prob(F-statistic)
Since at 5% significance level, the p-value of the Breusch-Godfrey Serial Correlation LM 
Test: statistic (0.7454) is greater than 0.05 we do not reject the null hypothesis. Therefore, 
there is no serial correlation among residuals.
Correlogram Squared Residuals
Date: 06/12/14 Time: 16:56
Sample: 1993 2013 
Included observations: 21
AC PAC Q-Stat ProbPartial CorrelationAutocorrelation
1 0 024 0.024 0.0141 0.905
2 -0 200 -0.201 1.0326 0.597
3 -0 180 -0.177 1.9017 0.593
4 0.091 0.059 2.1377 0.710
5 -0.022 -0.100 2.1524 0.828
6 -0.103 -0.115 2.4971 0.869
7 -0.136 -0.145 3.1371 0.872
8 -0.206 -0.314 4.7177 0.787
ss as as s
„ .(WW -0.126 6.4® 0.836
c□
cc
l]
:
cc cc c[=
]□ c
c[ 12 -0.120 -0.131cc
d Residuals for model 5.2
Figure 14 - Correlogram Square
74
As per the correlogram of squared 
confidence intervals. This 
significance level of 0.05. 
model.
residuals, spike could be observed lay beyond the 
p values are greater than the 
correlation among squared residuals of the
no
was confirmed by the all the
Hence, there is no
White’s Heteroskedasticity
Ho. No Heteroskedasticity exists in residuals 
Hi : There exists Heteroskedasticity in residuals
Table 34 - White’s Heteroskedasticity test for model 5.2 
Heteroskedasticity Test: Breusch-Pagan-Godfrey
F-statistic 
Obs*R-squared 
Scaled explained SS 3.254095 Prob. Chi-Square(6)
1.097443 Prob. F(6,14) 
6.717518 Prob. Chi-Square(6)
0.4109
0.3478
0.7763
Test Equation:
Dependent Variable: RESIDA2 
Method: Least Squares 
Date: 06/12/14 Time: 16:58 
Sample: 1993 2013 
Included observations: 21
Coefficient Std. Error t-Statistic Prob.Variable
85.97946 0.059370 0.9535
0.0950 
0.3165 
0.9659 
0.4199 
0.6554 
0.5001
5.104587 
0.000646 0.000361 1.790631
-0.000678 0.000652 -1.038910
-2.08E-05 0.000476 -0.043554
-0.001908 0.002296 -0.831054
0.001183 0.002595 0.455999
-0.053505 0.077291 -0.692259
C
GDP(-l)
GDP(-2)
GDP(-3)
CGDP(-l)
CGDP(-3)
BOT(-2)
54.73444
82.80799
R-squared 0.319882 Mean dependent var
Adjusted R-squared 0.028403 S.D. dependent var
S.E. of regression 81.62354 Akaike info criterion 1.90331
Sum squared resid 93273.62 Schwarz criterion 2.-5149
Log likelihood -117.9848 Hannan-Quinn enter. 1L97888
F-statistic 1.097443 Durbin-Watson stat -.659743
Prob(F-statistic) 0.410888 ___I
75
Since at 5% significance level, 
greater than 0.05 we do 
squared residuals of the model.
the p-value of the White’s Heteroskedasticity Test (0.4109) is 
not reject the null hypothesis. Hence, there is
no correlation among
Normality Test (Jarque-Bera Test)
'
Series: Residuals 
Sample 1993 2013 
Observations 21
i
6-
5-
Mean
Median
Maximum
Minimum
Std. Dev.
Skewness
Kurtosis
1.44e-13
-1.464088
18.96103
-11.21831
7.580973
0.620250
3.179886
4-
3-
2-
1 - Jarque-Bera 1.374799 
Probability 0.502882
0
-10 -5 0 5 10 15 20
Figure 15 - Jarque-Bera test for model 5.2
Since at 5% significance level, the p-value of the Jarque-Bera Test (0.502882) is greater than 
0.05 we do not reject the null hypothesis. Therefore, residuals are normal.
76
Chapter 5 - Conclusion
In general terms, the construction sector plays vital role in Sri Lankan 
industry sector has contributed positively towards the GDP 
noted the construction sector GDP growth 
country.
economy and the entire 
growth year-on-year. Also it was 
always higher than overall GDP growth in thewas
However, during our exercise to derive mathematical approach to express a relationship 
between construction sector output and national economic fiscals, It was noted two long term 
relationships as below.
BOT = 0.005543 GDP(-2) - 0.008140 GDP(-3) and;
ACINDEX = 60.17156 + 0.000288 GDP(-l) - 0.000178 GDP(-2)
+ 0.000183 GDP(-3) - 0.001059 CGDP(-l) - 0.002327 CGDP(-3) 
- 0.023864 BOTC-2)
Therefore, it could be concluded that Balance of Trade (BOT) has a relationship between 
previous year Gross Domestic Products (GDP) and a year before. In another words last two 
year GDP figures have influenced this year BOT figures.
Further, it explains that all construction cost index (ACINDEX) has an impact from last three 
GDP figures plus CGDP figures of last year and two years before plus BOT figures of
two years before.
year
Therefore it is evident that an existence of a strong relationship between construction activity
with the decisionand economic growth in Sri Lanka. Further, this conclusion is par 
statements provided by the previous researchers through their studies.
77
5.1 Recommendation and further studies
In this study it was determined that there exists a long term relationship between BOT and 
ACINDEX using a VAR model for annual data since 1990 to 2013. 
recommended to use the fitted model in determining relationships among those
Therefore it is 
variables.
However, no relationships were identified GDP, CGDP and for CGFCF 
methodology followed in this study. Therefore it is recommended to follow another 
methodology to determine such relationships between the said variables.
using the
78
References
• Al-Omari, J. A. 
industry in 
of Reading.
. Ball, M. (1988) Rebuildi
(1992) Critique of 
a capital-surplus developing
aspects of development theory using construction 
economy as an exemplar, PhD thesis, University
ng construction: economic change and the British constructionindustry. London .'Routledge. 
• Banwell, H. (1964) The Placing and Management of Contracts for Building and Civil 
Engineering work, A report of the Committe 
Banwell, HMSO, London 
• Bent E.
e under the chairmanship of Sir Harold
Sorensen (2005) ECONOMICS 7395 Spring 2005 March 1 2005.
The future of international construction: Secular patterns of growth and 
decline.Habitat International 16 (3): 119-128.
• Bowley M. (1966) The British Building Industry, Cambridge University Press: London.
• Dlamini, S. (2011).Relationship of construction sector to economic growth. School of 
Construction Management and Engineering, University of Reading, UK.
• Drewer, S (1997) Construction and development: Further reflections on the work of 
Duccio Turin. Proceedings of the First International Conference on Construction 
Industry Development, Singapore 9-11 December.
• Edmonds, G. A. & Miles, D.W.J. (1984). Foundations for Change: Aspect of the 
construction industry in developing countries. ITG publications, UK.
• Bon, R. 1992.
• Edmonds, G. A., (1979) The construction industry in developing countries, International 
Labour Review, 118, 355-369
• Emmerson, Sir Harold (1962). Survey of Problems Before the Construction Industries, A 
report prepared for the Ministry of Works. HMSO, London.
• Fox, P.W. (1989) A Study of the Hong Kong Construction Industry Using a Systems 
Approach, Unpublished MSc Thesis, University of Salford.
. Great Britain, EDC for Building (1967). Action on the Banwell Report, A report by a 
king party of the Economic Development Committee for Building, NEDO, HMSO:
London.
• Hassan, S.A. (2002)
2002.
• Higgin G. and Jessop N. K.
London. Tavistock Publications.
wor
. Construction Industry (Pakistan)?ub\ished by Economic Review
(1963 and 1965) Communications in the Building Industry.
%%
| *
79
& 6sfttv
. Hillebrandt P.M. and Mcikl 
. Hillebrandt P.M., (1974) E 
Publishers Ltd.: London.
• Hillebrandt P.M.,
Publishers Ltd.: London.
• Hillebrandt, P.M.
Macmillan, London.
Jin, W., X. Lu, and Y. He. 2003. A regression model of the growth path of construction 
industry.Chinese Civil Engineering Joumal36 (3): 105-109.
(2004). Housing and the Korean Economy. Journal of Housing Economics, 13,
u’ J-L (1985) Resource planning for construction, 
eonomic Theory and The Construction Industry
British construction industry, MacMillan 
0985) Economic Theory and The Construction Industry
MacMillan
(>984), Analysis of the
2nd Edn.
• Kim, K.H. 
321-41.
• Miles, D. & Neale R.H., (1991J Building for Tomorrow: International Experience in 
Construction Industry Development, Geneva: International Labour Office.
• Napier, G. (1970) A Systems Approach to the Swedish Building Industry. 
Stockholm:National Swedish Institute for Building Research, Document No. D9:1970
• Ofori G. (1980) "The Construction Industries of Developing Countries: the Applicability 
of Existing Theories and Strategies for their Improvement and Lessons for the Future - 
the Case of Ghana", Ph.D. Thesis University College, London.
• Ofori, G. (1990). The Construction Industiy: Aspects of its Economics and Management. 
Singapore: Singapore University Press.
• Ofori, G. (1993) Research on 
Construction Management & Economics, 11, 175-185.
Ofori, G. (1994) Practice of construction industry development at the crossroads.
International 18,2,41-56.
• Phillips Report (1950)Report of a 
London.
construction industry development at the crossroads.
Habitat
Working Party to the Minister of Works. HMSO:
and management of building contracts: Report of the Central 
an: Lord Simon) HMSO: London.
• Simon (1944) The placing
Council for Works and Buildings, (Chairm
Construction Sector In Economic Development
, Volume 17, Issue 3, pages 391-409, November 1970
• Tan, W. (2002). Construction and economic developm
ion Management & Economics, 20,59,-6,,.
construction industry with emphasis on
, Scottish
• Strassmann P.W., (1970) The
Journal of Political Economy in selected LDCs: past, present
andfuture. Construction 
• Tassios, S. (1992) “<Struclure
” Distr
& function of the 
limited. Paper ID/WG.528/L S September 1992.
developing countries,
80
Preparedfor meeting of UNIDO 
Construction industry, Tunis, Tunisia, 3-7 May 1993.
ThanujaRamachandra, James OlabodeBamideleRotimi and RaufdeenRameezdeen (2013): 
Diiection of the Causal Relationship between Construction and the National Economy of 
Sri Lanka. Journal of Construction in Developing Countries, 18(2), 49-63, 2013 
ThanujaRamachandra, RaufdeenRameezdeen (2006). Study of the relationship between 
construction sector and the Sri Lankan economy. Built-Environment-Sri Lanka - Vol. 06, 
Issue 02: 2006
• The Tavistock Institute of Human Relations (1966) Interdependence and Uncertainty: a 
Study of the Building Industry, London: Tavistock Publications Ltd. [see Crichton, C.]
• Turin, D.A. (1969) Construction Industry, based on the Proceedings of the International 
Symposium on Indusm'al Development held in Athens in Nov-Dec 1967, New York, 
Monograph no. 2.
• Wang, K. and Zhou, Y. (2000).Overbuilding: A Game-theoretic Approach. Real Estate 
Economics, 28, 493-522.
• Wells, J. (1985). The Role of Construction in Economic Growth and Development, 
Habitat international, 9(1), 55-70.
• Wells, J. (1986). The Construction Industry in Developing Countries: Alternative 
Strategies for Development, Croom Helm Ltd, London.
• Wood (1975) The Public Client and the Consfruction Industries (Chairman: Sir Kenneth 
Wood), Great Britain, EDCs for Building and Civil Engineering, London: NEDO, 
HMSO.
. World Bank, The Construction Industry: Issues and Strategies in Developing Countries, 
International Bank for Reconstruction and Development, The World Bank, Washington
DC, 1984.
and UNCHS (Habitat), First Consultation on the
81
Appendix
GDP CGDP
(Current)
(Rs. Mn)
(Current) BOT CGFCF 
(Rs. Mn)Period (Rs. Mn) (Rs. Mn) ACINDEX
1990 321784 21541 -702.50 35239 100.00
1991 372345 24535 -997.10 38270 110.40
1992 425283 28485 -1044.60 48005 118.63
1993 499565 32615 -1147.60 59223 127.43
1994 579084 38323 -1558.70 69297 137.58
1995 667772 44455 -1504.50 78385 147.08
1996 768128 48234 -1343.70 85977 153.45
1997 890272 56434 -1224.80 99389 159.18
1998 1017986 116357 165.9369301 -1091.70
1999 127569 167.751105963 -1369.2075538
140784 175.502000 1257636 82684 -1797.50
196.2316185195057 -1157.5014073982001
179463 208.60-1406.6010040415818852002
220.58197074-1538.6011011118224682003
263096 245.15-2242.6012769220908412004
290.63329611-2516.5016799924527822005
343.60419662-3370.3021683329386802006
387.63537633-3656.5026410435786882007
444.78705374-5980.6032713844106822008
456.25802445-3122.1036624848352932009
465.93996190-4825.1042341456041042010
490.901118634-9710.0051122065433132011
550.851400370-9416.7071227275785542012
590.431631404-7607.7089468386738702013
-
82