TESTING THE TWIN DEFICIT HYPOTHESIS FOR
KENYA 1970-2012
ERASTUS KAIBA NJOROGE
A research project presented to the Department of Applied Economics, in the School of Economics, in partial fulfillment of the requirements for the award of Degree of Master
of Economics (Finance) of Kenyatta University
DECLARATION:
This project is my original work and has not been presented for a degree in any other University or any other award.
Signature: ' --
Date:'J:;L,}OS}
k()14£
Erastus Kaiba Njoroge, BlEd Arts Reg. No.: KI02/CTY/PT/21S0S12010
This project has been submitted for examination with our approvalas University Supervisors
Signature
~J
.
.
Date:.~?
l
tJJ.f.[f?
IY
«")/
Name: Dr. George K. Kosimbei
Position: Lecturer, Department of Applied Economics
Signature: _.--."
.
,
Name: Dr. Julius Korir
DEDICATION
ACKNOWLEDGEMENT
I wish to express my sincere gratitude to my Supervisors. I owe the success and completion of this project to Dr. George Kosimbei and Dr. Julius Korir
Special thanks goes to Dr. George Kosimbei, my supervisor, for his undivided support during the writing of this project and for being more than generous with his expertise and precious time. His patience and understanding, his countless hours of reflecting, reading, encouraging, and most of all patience throughout the entire process went a long way in ensuring I came up with a quality project.
I would like to acknowledge Dr. Julius Korir for guiding me step by step and assisting me to ensure that I finalize with the project in time. His unlimited assistance is the reason I managed to come up with a final copy.
My sincere thanks go to the lecturers in the School of economics, Kenyatta University for their constructive criticism. Their willingness to provide feedback made the completion of this study an enjoyable experience.
Finally, I take this opportunity to thank my wife for reading through the project page by page to ensure errors and omissions are minimized;
ABSTRACT
Macro-economic theory suggests that there exists a causal relationship running from Budget Deficit to Current account deficit a concept normally referred to as the Twin Deficit Hypothesis (TDH). This claim has more often than not been the subject of debate in the scholarly and policy front. Majority of existing literature
on the twin deficits hypothesis focuses on its relevance in already developed economies. Further, existing literature has concentrated on analyzing the TDH on a bivariate approach using annual data. This study sought to investigate the relevance of the TDH nexus for Kenya using quarterly data in a multivariate approach which included budget deficits, current account balance, interest rates and exchange rates. The study analyzed quarterly data for a four decade period spanning from 1970Ql - 2012Ql. To estimate if a relationship exists between the variables, the proposal estimated the cointergration properties using Johansen &
Juselius model. The study also applied the VAR model to estimate IRF and Variance decomposition. Finally, the study investigated the causal relationship of the two deficits in the framework of Toda- Yamamoto's Granger causality test for causality. From the analysis, the study concluded that the twin deficit hypothesis does exist in Kenya when interest rates and exchange rates are included. In effect,
Ta
b
le of Contents
DECLARATION: i
DEDICATION ii
ACKNOWLEDGEMENT iii
ABSTRACT IV
TABLE OF CONTENTS v
LIST OF ACRONYMS AND ABBREVIATIONS vii
OPERATIONAL DEFINITION OF TERMS viii
CHAPTER ONE: INTRODUCTION 1
1.1 Background 1
1.2 Overview oftrends inTwin Deficit Hypothesis Variables in Kenya 3
1.3 Statement of the Problem 6
1.4 Research Questions 9
1.5 Objectives ofthe Study 9
1.5.1 General Objective 9
1.5.2 Specific Objectives 9
1.6 Significance of the Study : 10
1.7 Scope of the Study 12
1.8 Organization of the Study 12
CHAPTER TWO: LITERATURE REVIEW 13
2.1 Introduction : 13
2.2 Theoretical Literature 13
2.2.1: Keynesian! Conventional proposition 13
2.2.2: The Mundel Flemming Approach 14
2.3 Empirical Literature in other countries 19
2.5 Overview of Literature 26
CHAPTER THREE :METHODOLOGy 29
3.1 Introduction 29
3.2 Research Design 29
3.3 Theoretical Framework 29
3.4 Model Specification 36
3.5 Definition and Measurement of Variables 37
3.6 Data Collection 38
3.7 Data Analysis 39
CHAPTER FOUR: EMPIRICAL FINDINGS .43
4.1 Introduction 43
4.2 Diagnostics 43
4.2.1 Unit root test ~ 43
4.2.2 Lag Length Selection 45
4.2 Cointergration Test : : 46
4.3 Test for causality 49
4.4 Impulse Response Functions andVariance Decomposition 51
CHAPTER FIVE: SUMMARY, CONCLUSIONS AND POLICY
IMPLICATIONS 54
5.1 Summary 54
5.2 Conclusion 55
5.3 Policy Implications 55
5.3 Areas F or Further Research 55
REFERENCES 57
APPENDICES
Appendix I: List of Tables 62
Appendix II:List of Figures 63
Appendix III:Results for the Unit Root Tests 64 Appendix IV: Graphical Results for the Impulse Response Function Tests 67 Appendix V: Results forthe Variance Decomposition Tests 68 Appendix VI: Tabular Results for the Impulse Response Function Tests 69 Appendix VII: Summary Effect of Monetary and Fiscal Accommodation under
Fixed Exchange Rates 74
Appendix VII: Summary Impact of Monetary and Fiscal Accommodation under
TDH VAR KNBS BD CAD GDP CBK IMF OLS VECM ECM IFS
LIST OF ACRONYMS AND ABBREVIATIONS
Twin Deficit Hypothesis Vector Autoregressive
Kenya National Bureau of Statistics Budget Deficit
Current account deficit Gross Domestic Product Central Bank of Kenya International Monetary Fund Ordinary Least Squares
Vector Error Correction Model Error Correction Model
International Financial Statistics
OPERATIONAL DEFINITION OF TERMS
TERM MEANING
Twin deficit Hypothesis The concept states that there exists a positive relationship between budget and current account deficit, and that the causality runs from budget deficit to Current Account Deficit.
The relationship between cause and effect Causality
CHAPTER ONE
INTRODUCTION
1
.
1
Background
Mccoskey and Kao (1999) define the Twin Deficit Hypothesis (TDH) as the long
run positive relationship between the budget deficit and the current account
deficit. The TDH nexus postulates that an increase in budget deficits (BD) leads
to an increase in current account deficits (CAD). According to Ahmad, Lau and
Ahmed (2006), the link between the two deficits has been at the core of research
interest since the 'Reagan Fiscal Deficits' in the 1980s, a period that marked
expansive growth in the two deficits especially in the US rousing researchers and
scholars to explore their relationship. Over time research on the relationship
between the two deficits has spread to other developed and developing nations.
The TDH nexus can be explained usmg ,the Keynesian income-expenditure
framework, (Brian, 2011), and the Mundell Flemming framework (Ahmad et al.,
2006). According to the former, an expansionary budget leads to increased
income ultimately resulting in increase in aggregate demand for domestic and
imported goods. The increase' in imports leads to a worsening of the current
account balance. According to Mundell Flemming, an increase in budget deficit
causes an upward rise in interest rates if government borrows domestically to
finance the deficit. This rise in interest rate leads to ,capital inflows and
consequently an appreciation of the exchange rates. This means exports become
less attractive while imports become attractive ending up worsening the current
account. This approach however depends on the openness of the economy and the
exchange rate regime.
In
a fixed exchange rate regime expansionary fiscal policywould lead to increased income a process that would still worsen the current
account (Mundel, 1968).
Government spending can be considered as an engine of growth in any country.
As an agent of the people, the government is tasked with the responsibility of
provision of social amenities such as education, health services and infrastructure.
As such, it is very difficult for a government to curtail its expenditure appreciably
at the expense of the aforesaid responsibilities. The dilemma is compounded
further by the fact that due to low levels of income and development mainly in
developing nations, there is a very narrow scope for increasing tax revenues
(Khalid and Guan, 1999). In the process of discharging the enormous
responsibility, expenditure outstrips revenue. The government may resort to
borrowing and the challenge with this option lies in ensuring that the debt is
prudently utilized to meet people's needs, aid, growth as well as ensure that
derived income is sufficient to meet the repayment needs.
Understanding the connection between the two deficits has a number of
advantages. First, large budget deficits caused by either internal or external
borrowing has consequences on future generations who are left with a repayment
adoption of the budget policy by the National Assembly. Chapter 12 of the
constitution on the other hand outlines principles shall guide aspects of public
finance in the country, revenue sharing and allocation guidelines as well as
appointment of various relevant state officers.
Despite stringent fiscal policies that seek to bring about fiscal discipline in the
country, the problem of expanding budget and current account deficits continue to
face the country. Figure 1.1shows the trends in the country's current account and
budget balances as a percentage of GDP
%ofGDP
4~---,
o
-4 -8 -12
-16
-204-~'-~'-~'-~'-~'-~-'~-'~-'~-''-~'-~'-~'-~'-~
1970 1975 1980 1985 1990 1995 2000 2005 2010
1
--
BD--CAB1
Source: IMF, World Economic Outlook Database, April20J3
Figure 1.1:Kenya's Current Account and Budget Deficits as a Percentage of
GDP
As shown Figure 1.1,a striking feature of the country's fiscal operations since the
1970's indicate that Kenya has been running budget deficit and current account
deficits for many years since independence. This rise in spending is due largely to
government initiatives to improve infrastructure and support the country's free
education system for an increasing population as well as due to poor budgetary
planning (Wawire, 2006). Government spending has been on a rapid increase that
has not been matched by a commensurate rise in government revenue. The worst
Budget deficit recorded was in 1992-1994 when donor funding was cut from
Kenya. In efforts to rein in spending, the Kenyan government made policy
changes in 2000 and 2001, including efforts to improve fiscal discipline and
transparency by strengthening the government's Office of the Controller and
Auditor General. In order to increase tax revenue, the Kenyan government has
also expanded its consumption tax policy to apply to more goods. These policy
changes have been successful at managing the recent budget shortfall, but budget
deficits continue to be a problem for Kenya.
The country's current account balance has also remained mostly negative over the
years. The current account deficit rose from 2.9 percent of GDP over 1964-73 to
6.9 percent over .1974-79 on account of the two oil shocks, widening trade
balance and overvalued domestic currency. Long term flows turned from a
position of 5 percent of GDP over 1964-73 to a -l.8 percent of GDP over 1996
-2000 prompting the country to rely increasingly on risky short term flows to
balance the accounts. The situation was aggravated in the late 90's due to afreeze
in donor aid and lending with the country recording lowest budget balance. The
trend has been worsening since 2008 partly due to the aftermath of the disputed
Figure 1.2 presents movements in volumes of imports and exports forKenya.
40
30
!I ...•
.
.
.
".
:
..
--
.
-10
, 0 ~I
'fit r0 "'" lf) CD r-,
.
s
'"" N r0 "" lf) r-. 00~ 0 en en
~
en en en en ·en
.
0 0 0 0 0 0 0 e-ten (;, en en en en
i
'
en 01 0 0 0 0 0 0 0 0 0 0.--j rl rl rl rl rl ,--l rl
N N N r-I N N N N N N
Year -20
Volume of Imports ofgoods % Change ••• Volume ofexports ofgoods %Change
Source: IMF, World Economic Outlook Database, April 2013
Figure 1.2: Percentage Change in Volume of Imports &Exports
As shown Figure 1.2,the country's import bill continues to expand at a rate that is
not commensurate to the expansion of the export bill. It was largely anticipated
that, with the implementation of the EAC Customs Union in 2005, Kenya would
dominate regional trade by diversifying its exports to the EAC market, given its
comparative advantage, especially in the manufacturing sector. However, this has
not been the case and as shown above, imports largely exceed the volume of
exports.
1
.3
Sta
te
ment
o
f the Problem
Kenya has continuously been running budget deficits and current account deficits
with the exception of 1993 and 2002 where a current account surplus of +2% of
GDP was recorded. Stephen (2010) noted that in order to achieve macroeconomic
stability and sustained economic growth, the two deficits have to be kept in
control. Keeping a balance between the two macro-economic variables is a
problem not only for developed countries but also for most developing countries.
Recurrent budget deficits tend to overburden future generations through principal
and interest repayments especially if the borrowed funds are allocated to recurrent
as opposed to development budgets (Egwaikhide et al., (2002).
Many studies on the twin deficit hypothesis have been carried out but there is
hardly any consensus on the direction of causality between budget ~Qleficitsand
current account deficits. In the case of Kenya, Kosimbei (2002) and Egwaikhide
et al. (2002) have carried out studies and found contradicting results. Kosimbei
(2002) found support for Ricardian equivalence while Egwaikhide et al. (2002)
found support for unilateral causality running from current account deficit to
budget deficits. These contradicting findings raised a need to investigate the
deficits using a different model and different data set.
These studies by Kosimbei (2002) and Egwaikhide et al., (2002), concentrated on
analysis using two variables that is, budget deficit and current account deficit.
However, the transmission mechanism between the two variables may not
necessarily be direct but through effects on variables such as interest rates and
exchange rates. This is because interest rates and exchange rates affects other
important economic variables such as investments and GDP growth. Therefore,
current account deficit through intermediating variables of interest rates and
exchange rates. No study had earlier included the two variables for a study on the
causality between the budget deficit and current account in Kenya.
Also majority of past studies including the studies for Kenya by Kosimbei (2002)
and Egwaikhide et al., (2002), had concentrated on testing for causality using
Granger causality test which have been blamed for concentrating on time
precedence rather that causality thus being weak in establishing the relation
between forward looking variables (Abdur and George, 2003). Since these
causality tests are grounded on asymptotic theories that are only valid for
stationary variables, non-stationary data has to be differenced until it's stationary.
Various models offer Error correction which are cumbersome and sensitive to the
values of nuisance parameters in finite samples and thus their results are
unreliable, (Toda and Yamamoto 1995) and (Zapata and Rainbaldi 1997). Toda
and Yamamoto (1995) proposed a better procedure that permits casual inference
to be performed in level VARs that may contain integrated process without
involving the rigorous attention and strict reliance upon intergration and
cointergration properties variables in the system. The purpose of this study was to
investigate the relation between budget deficits and current account Deficits
within the context of indirect transmission and testing causality using the recent
Toda Yamamoto methodology.
1.4
Research Questions
The study sought to answer the following questions
i) What is the relationship between budget deficits, current account balance,
interest rates and exchange rates?
ii) Are interest rates and exchange rates significant as intermediating variables in the transmission mechanism between budget deficit and current account
balance?
iii) What is the impact of shocks on budget deficits, exchange rates, interest rates and current account deficits on each other in Kenya
1.5
Objectives of the Study
1.5.1 General Objective
The general objective of this study was to test for the existence of the Twin deficit
Phenomena in Kenya while including interest rates and exchange rates as
intermediating variables.
1.5.2 Specific Objectives
The specific objectives were to:
i. Investigate the relationship between budget deficits, current account
ii. Establish the significance of interest rates and exchange rates as
intermediating variables in the transmission mechanism between budget
deficit and current account balance?
iii. Investigate the impact of shocks on Budget Deficits, exchange rates,
interest rates and Current Account deficits on each other in Kenya
1.6
Significance of the Study
The Vision 2030 aims at transforming Kenya into an "industrializing,
middle-income country by the year 2030". One key pillar is to maintain a sustained
economic growth of 10%. The vision acknowledges that a stable macroeconomic
environment is the only way 'in which confidence among investors can be
maintained. With a 10% GDP growth target, managing the deficits is key.
Kenya's worst economic performance in the early 1990's due to donors having
withdrawn their funding led Kenya to having the worst economic performance
since independence, (Kosimbei, 2009). This reliance on donor funding has been
characteristic of Kenya's economy despite the vital lessons learnt during the times
that donor aid had been withdrawn. Wawire, (2006) notes that poor budgetary
process coupled with limited resources is to blame for the budget deficit in Kenya.
The 'Jubilee government' is faced by a myriad of challenges including a largely
expanding budget deficit and current account deficit. To make matters worse, the
government also has an obligation to fulfill various national and international
commitments like increased pay for teacher, Millennium Development goals, and
war on terrorism among others. This proposed expenditure is expected to push the
Kenyan budget & current account deficits plummeting further. Sustainable
fulfillment of these obligations will require no less than proper management of deficits in order to among others, attract foreign investors and consequently
generate tax revenues.
This study was therefore deemed to be important to scholars as well as policy
makers in Central Bank and Ministry of Finance seeking to identify an appropriate solution to the deficits problem. Identifying the connection between
the two deficits would help to provide an appropriate policy recommendation on the specific deficit to target in order to reduce both deficits. Kenya provided an
.ideal scenario for studying the relationship between fiscal deficit and the current account deficit not only because it's a developing nation but also because it has
been running the two deficits for several decades.
Large and persistent current account deficits are among the most serious problems
of many developing countries since they result in economic crises like currency
crises, the burgeoning external debts and the reduction in international reserves. If it is the case that budget deficits do cause current account deficits, then policy
1.
7
Scope of the Study
The study used a four decade secondary quarterly data spanning 1970 Q1 to 2012
Q
1. Data for GDP, Exchange rate, Interest rates, Budget and Current accountDeficits were collected from the KNBS, CBK, IMF and World Bank's World
Economic Outlook Database.
1.8
Organization of the Study
The rest of the paper is structured as follows;
Chapter two presents the theoretical and empirical literature. The theoretical
literature begins by explaining the four possible directions of causality for the
budget and current account deficits. The empirical literature highlights a number
of past studies that have investigated the problem of the twin deficit not only in
Kenya but the rest of the world. The chapter ends with an overview of the
literature where a highlight of relevance of the literature as well the knowledge
gap is expounded. Chapter three presents the methodology of the study. The
research design, and theoretical framework is discussed. The chapter explains the
models specification, definition of variables. The chapter finalizes by explaining
the set of econometric tests that will be undertaken. Chapter four presents the
findings and chapter five concludes with policy implications and policy
recommendations.
CHAPTER TWO
LI
T
ERATURE REVIEW
2.1 Introduction
This section provides a discussion of existing theoretical literature and later
reviews a selection of past literary works in order to appreciate how far we have
come. At the end of this section, a critique of the empirical literature is provided
while identified research gaps that the study sought to address are'pointed out.
2.2 Theoretical Literature
Majority of past literature generally concentrated on two testable hypotheses
namely, the Keynesian! conventional proposition and the Ricardian Equivalence
proposition. However, these are not the only testable approaches by which the
interaction between the two deficits can be. studied. As a matter of fact, this
interaction can be studied through four testable approaches.
2.2.1: Keynesian/ Conventional proposition
The first approach is based on the Keynesian/ Conventional proposition. The
argument postulated by the Keynesian-income spending approach indicates that
open budget policies that occur as a result of a decrease in taxes or an increase in
public expenditures cause an increase in the national income. This increase in the
national income boosts the import sector and this consequently results in the
Considering the aforesaid, it would be true to say that there exists a relationship
between the two deficits and that this relationship runs from budget deficits to
current account deficit. Researchers like Perera and Liyanage (2010) have found
this true and in fact found support for the role of intermediating variables like
exchange rate and interest rates asexplained in the Mundell Flemming approach.
2.2.2: The Mundel Flemming Approach
As postulated by Mundel-Fleming model, a cut in public expenditures in a small
economy with a flexible exchange rate regime would lead to a reduction in
aggregate domestic demand and consequently a fall in GDP. This causes a
reduction in the transactional demand for money pilling a downward pressure on
domestic interest rates which causes a gap between international and domestic
interest rates. The difference in domestic and international interest rates causes
capital outflows and an increase in the demand far investment goods.
Depreciation of the domestic currency due to higher domestic demand and lower
interest rates brings about correction in both the trade balance (a proxy for the
current account balance) and the rest of the macro-economy, including the budget
balance, until the alignment of domestic with international interest rates is
restored.
2.2.3: Ricardian Equivalence Hypothesis
The second- and largely controversial approach- based on the Ricardian
Equivalence Hypothesis postulated by Ricardo states that there exists no
relationship between the two deficits. The approach explains that an
inter-temporal shift between taxes and budget deficit do not matter for real interest
rates, investment or current account balance. Ricardian equivalence hypothesis
makes a strong presumption people that understand that any increase in
government expenditure that is financed by increased budget deficit will
ultimately be compensated by taxation in the future. Critics of the Ricardian
equivalence hypothesis have objected the five major assumptions which are; i)
Capital markets are perfect, ii) People do not live forever and as such, they do not
care about future taxes, iii) Economic agents and mainly consumers are rational
and farsighted meaning that they satisfy the infinite horizon condition, iv) Taxes
are not distortionary lump-sum per capita and iii) Full employment exists (See
Barro, (1989) for more explanation). Despite the presumptions, some researchers
have found evidence to support this view (See for instance, Kosimbei, 2002;
Ratha, 2009; Brian, 2011).
The two are not the only testable approaches as far as the interaction between
budget and current account deficit are concerned. The third approach- known as
current account targeting- is based on a unidirectional causality that runs from
current account deficit to Budget deficit. Such situation may happen when
worsening of the current account leads to slowing of the economic growth thus
leading to budget deficit. Egwaikhide et al., (2006) notes that this argument is
particularly more likely for countries that highly depend on foreign direct
importance this approach in analyzing the TDH for Kenya. Scholars like
Egwakhide, et al.(2002; Lau (2006) and Halil (2012) have found support for this
approach.
The final approach is based on a bi -directional causality that runs from Budget
deficit to Current account deficit and vice-versa. Existence of a feedback causality
may cause the causality to run in both directions. Jayaraman and Lau (2008)
-having found supporting evidence for this proposition for 6 Pacific Island
Countries indicate that Bi directional causality is not an unusual phenomenon for
a country that relies on export revenues. This therefore makes this approach
, important for the Kenya situation. Since Kenya operates in a flexible exchange
rate (Ndungu & Ngugi, 1999), the theoretical literature specified on the impact of
fiscal policy in a floating exchange rate regime. Stephen, Benjamin, Francis,
Njuguna and Ngugi (2010) observed that Kenya's capital mobility "is substantial
but far from perfect". For this reason, the theoretical literature was restricted to
the impact of a fiscal accommodation in the case of flexible exchange with low &
relatively high capital mobility.
An expansionary fiscal accommodation raises income causmg an mcrease m
demand shifting the IS curve to the right leading to an increase in interest rates as
shown in Figure 2.1.
v
Figure 2.1: Fiscal Policy under Flexible Exchange Rates and Low Capital
Mobility
The rise in income causes an increase in the demand for imports & the current
account deteriorates. An expansionary fiscal accommodation causes the IS Curve
to shift from IS to IS1. A new equilibrium at B represents a deficit in BOP since it
is below the original BOP curve. This increase in income raises the transactions
demand for money, driving up the interest rate & causing higher imports. The
resultant capital inflow is however not sufficient to finance the deficit. As such,
domestic currency depreciates shifting the BP to the right &further causes the IS1
to shift out further to IS2•A new equilibrium is reached where IS-LM-BP intersect
at C. Note that the LM curve stays put while the IS curve shifts to the right twice.
The first shift is due to the expansionary fiscal policy, and the second shift is due
to the improvement in the trade deficit brought about by the depreciation.
If capital flows are more elastic to domestic interest rate movements, the capital
movement's resultant from a fiscal accommodation could be huge. Figure 2.2
shows Fiscal Policy under Flexible Exchange Rates and High Capital Mobility
LM
-, -,
, ,
, . BP
"B
-,
'.
--~
_-~-
y
-
\
P-
,
.
IS
v
-
OutputFigure 2.2: Fiscal Policy under Flexible Exchange Rates and High Capital
Mobility
An expansionary fiscal policy accommodation shifts the IScurve to IS' and the
economy moves from old equilibrium A to a new equilibrium B. Now, point B is
above the BP line and corresponds to a Balance of payment surplus - at B, the
capital inflows induced by the increase in the interest rate more than compensate
the deterioration in the balance of trade. The external balance is in disequilibrium
causing an appreciation that shifts the BP curve to the left. The appreciation of the
domestic currency has negative effects on trade: the IS shifts back wiping away
part of the fiscal expansion. A new IS-LM-BP equilibrium is reached in C. The,
IS also shifts twice, but in opposite direction. Higher degree of capital mobility
has undesirable effects on the effectiveness of fiscal policy accommodation with
flexible exchange rates. As noted, the resultant appreciation of domestic currency
has a negative impact on the current account leading to a reduction in output.
Although the expansion is not as large as with low capital mobility, some
expansion still takes place.
2.3 Empirical Literature in other countries
Lau and Baharumshah (2004) used annual data for 1976- 2000 to investigate the
causal link between the two deficits for Malaysia. By employing the Toda
Yamamoto granger causality test, the authors found support for bi-directional
causality. They conclude that the scenario is not an unusual phenomenon for an
emerging country and that any decision to reduce the current account deficit
problem cannot be achieved by taking a fiscal stance only but also use of
monetary policies.
Lau et al. (2006) examined the TDH link for 4 countries which are India,
Malaysia, Thailand and Philippines. While still making use of the Tod
a-Yamamoto granger causality test, the authors found support for bi-directional
The study also found strong support for TDH for Thailand while for Indonesia,
reverse causality was confirmed. The study confirms that exchange rate and
interest rates are important channels for causality.
Marinheiro (2006) analyzed data for Egypt for the period 1974 - 1989. Using
VECM model, the author tried to determine whether the budget deficit leads to an
external deficit in Egypt. The findings indicated that there exists a causal
relationship from the current account deficit to the budget deficit. This study will
help us compare the findings with those of Kenya, both countries being
developing nations.
Lau, Mansor and Puah (2007) used quarterly data to examine the TDH nexus in
the Five Asian Countries; Korea, Malaysia, Philippines, Indonesia, Thailand. The
researchers used Johansen and Juselius (1990) cointergration procedure and
VECM for Granger causality. Their study was split into two sub periods of 1976
Q 1 to 1997 Q2 and 1997Q3 to the end for each country's sample. In the pre-crisis
period, the authors found support for the TDH for Malaysia, Philippines, and
Thailand. Results indicate causality runs in the opposite direction Korea and
Indonesia. In the Post Crisis period, their research returned similar results with the
exception of Philippines which had a bi directional causality. The researchers
concluded that while as managing the deficits should be an important national
agenda, TDH is not a universal phenomenon, and rather, it is country specific.
Comparing this study with that of Lau et, al (2006) one gets the notion that model
specified and data characteristics play an important role in determining the results.
Jayaraman and Lau (2008) studied six Pacific Island Countries (6 PIC's) for
1988-2004 using VECM and fully modified OLS method. The study found a
bi-directional relationship in the short run which they indicate is not an unusual
phenomenon for countries that rely on export revenues. In the long run, the study
could not find any causality.
Sadullah and Deniz (2008) used quarterly data for 1996 Q
1-
2006 Q4 to study sixemerging countries. Using panel cointergration and fully modified OLS Method,
the authors in similarity with Akbostanci and Tunc (2002) found supports the twin
deficit hypothesis for Czech, Brazil, Mexico, Colombia, South Africa and Turkey.
Their study points out on the important role of intermediary variables i.e. interest
rates and exchange rates.
Pahlavani and Saleh (2009), used annual data for Philippines spanning from
1970-2005 to test the the direction of causality between the two deficits. Using MWald
procedure for causality analysis, they confirmed a bi-directional causality. The
study advised that policy measures to reduce the budget deficit could play an
important role in reducing the current account imbalances and vice-versa. The
study is consistent with Lau, Mansor and Puah (2007) for the case of Philiphines.
Ganco (2010) undertook to establish the relationship between the two deficits for
error correction models found out that Budget deficit has impact of Current
account deficits. The study concluded that TDH is not valid for Bulgaria and that
fiscal policies should not be used as a substitute for monetary policy.
Using VECM, Akbar and Mehrara (2011) examine the relationship between no
n-oil deficits in Iran for the period spanning 1959 to 2007. The study established a
positive long run relationship between the current budget deficit and non-oil
current account deficit & found a bidirectional relationship between the two
deficits. They also found that taxes and government expenditure do affect
economic variables especially current account deficit. They thus recommended
that government should control budget deficit in order to manage current account
deficit.
In an effort to analyze Twin deficit hypothesis and Feldstein-Horioka Hypothesis
(FHH) in Turkey, Halil Sami (2011) analyzed data for 1974 to 2010. The study
used ARDL bound testing approach and found evidence in support of TDH &
FHH.
Oladipo and Akinbobola (2011) examined annual data for Nigeria for the period
between 1970 and 2005. The study included inflation and exchange rate as
intermediating variables. The authors used Johansen maximum likelihood
procedure for long run cointergration and granger causality test. They found a
unidirectional causality running from Budget deficit to inflation. Though the test
is not on the budget deficit versus current account deficit, it pinpoints the
important role played by intermediating variables like inflation and exchange rate
in the economy. The study concludes that relevant measures to enhance policy
coordination among various arms of the government should be put in place. Also,
monetary policy should be made to complement fiscal policy.
Perera and Liyanage (2011) undertook to investigate the relationship between
Budget deficit and Current Account deficit. Using annual data from 1960 to 2009
and quarterly data from 1990 Q1 to 2009 Q4, the authors used Engel Granger
residual based test for cointergration, Johansen maximum likelihood procedure
for long run cointergration and Granger causality to test for direction of causality
& TDH was confirmed. They conclude that there is a possibility of relying on
curtailing budget deficits in an attempt to trim down current account deficits.
In another study for Nigeria, Oladipo et al. (2012) conducted an empirical
analysis of TDH in Nigeria using pair wise granger causality tests. Their study
used annual data for 1970-2008. They found out that budget deficit and current
account deficit are cointergrated with causality running from running from Budget
deficit to current account deficit. They concluded that it's important to
complement budget cuts with a coherent package focusing on policies for export
promotion, productivity improvement and exchange rate.
In another study aiming to analyze The TDH in Iran, Farzane, (2012) studied
annual data for 1971-2007. The study employed Johansen test for cointergration
causality runs from budget deficit to current account deficits. The authors
concluded that government should curtail nonproductive expenditure in order to
solve the problem of the current account deficits.
Saeed and Khan (2012) used annual data for 1972-2008 to investigate the
dynamics of the two deficits in Pakistan. Using Johansen maximum likelihood
procedure for long run cointergration and error correction model for granger
causality, the researchers found a long run relationship between the two deficits.
This causality runs from budget deficit to current account deficit thus conclude
that Pakistan is a non Ricardian economy.
Ucal and Mehmet (2012) analyze quarterly data for 1996 Q1- 2011 Q4 to test
whether budget deficit has correlation with current account deficit and to check
direction of correlation for Turkey. Using Johansen Cointergration Analysis and
granger causality tests, they found that current account deficit has significant
negative effect on budget deficits. This study contradicts the findings of Sarni
(2011). Saeed and Khan (2012) explain that most researches have reached
contrasting conclusions due to the pre specifications of the models they use.
Brian (2012) employed the use of VAR model to empirically examine the causal
relationship between BD and CAD for Argentina. Using quarterly data from 1976
Ql-2010 Q3, the study couldn't find causality in any direction. As such, the study
concludes that Argentina is a Ricardian nation. The authors are however quick to
warn on the imperfect nature of the data used. They also indicated that there is
evidence that a better fitting model can be made
Merza et al. (2012) used Ganger causality test, Johansen cointergration test,
Vector Auto regression and Impulse response functions to study the Twin Deficit
Hypothesis for Kuwait. Results for the data spanning from 1993 Q 1- 2010 Q4 indicates that though budget deficits and current account deficits are cointergrated
in the long run, budget deficits respond negatively to shocks in current account
deficits. The findings that twin deficit hypothesis is not existent for the Kuwaiti
economy is consistent with study by Alaksami (2000) for Saudi Arabia which is a
similar economy i.e, oil rich economy.
2.4 Empirical Literature in Kenya
Egwaikhide et al., (2002) studied a number of African countries using data for
1970-1999. Their study employed the use of OLS to analyze for correlation
between the two deficits. The authors found that there exists a positive
relationship between the two deficits for all nations under study except for
Cameroon, Cote d'Ivoire, Gambia, Guinea-Bisau and Mali. Granger causality test
confirmed twin deficit hypothesis for Benin, Burkina Faso, Ghana, Nigeria and
South Africa and bi-lateral causality for Togo. The study found a unilateral
causality running from current account deficits to budget deficits for the case of
Kenya. Their fmdings were particularly important for us as they helped us to
Kosimbei (2002) used annual data for 1964 to 2000 to analyze the relationship
between fiscal and current account deficit. The author carried out Granger
causality tests which revealed that there was no causality between fiscal and
current account deficits. The study concludes that the Ricardian equivalence is
valid in the Kenyan case with short-run dynamics being the same as long run
equilibrium relationships. In the same year and in contrast to findings by
Kosimbei (2002), Egwaikhide et al., (2002) found a unilateral causality that runs
from current account deficits to budget deficits for the case of Kenya. This
contradicting finding raised the need to investigate the deficits using a different
model, different data set and a longer period data.
2.5 Overview of Literature
It was noted that the past studies suffered from various shortcomings. The first
shortcoming involved the use of cross-sectional data which may not address
country specific issues (See for example Lau, Mansor and Puah (2007) who
examined TDH for Five Asean Countries, Jayaraman and Lau (2008) who studied
six Pacific Island Countries & Festus et al (2002) who studied a collection of
African countries including Kenya).
Secondly, the studies had a shortcoming of having used bivariate causality test
that may likely result in omission of important variables and hence give biased
results. In fact, many studies ignored the role of exchange rates and interest rates
in their studies. The fact that the discussion on the role of interest rates and
exchange rates and their impact on vanous economic variables has been
controversial warranted their inclusion in the research. No past study regarding
the causality between the budget deficit and current account nexus in Kenya had
included the two mediating variables. The study also took into consideration the
fact that a good econometric model ought to include all relevant variables since
omitting a relevant variable in a model specification would lead to biased results.
Thirdly, unlike past literature which suffered from the use of annual data, the
study used quarterly data to help capture the dynamics that could be omitted
erroneously by use annual data.
Lastly, the study noted that past literature has concentrated on testing for causality
using Granger causality tests which can be modeled by a straight forward
regression (Utku, 1995). Econometric tests have shown that such tests tend to
concentrate on time precedence rather that causality. They are thus weak in
establishing the relation between forward looking variables (Abdur and George,
2003). Also, since the tests are grounded on asymptotic theories that are only
valid for stationary variables, non-stationary data has to be differenced until it's
stationary. Various models offer Error correction which are cumbersome and
sensitive to the values of nuisance parameters in finite samples and thus their
results are unreliable, (Toda and Yamamoto 1995) and (Zapata and Rambaldi
1997). Toda and Yamamoto (1995) proposed a better procedure that permits
process without involving the ngorous attention and strict reliance upon
intergration and cointergration properties variables ill the system. The study
utilized this latest causality testing technique.
The latest attempt investigating the twin deficit hypothesis for Kenya was by
Kosimbei (2002) and by Egwaikhide et al., (2002). This study therefore sought to
improve on their studies by including more recent data bringing on board recent
developments on the macroeconomic front. The study therefore sought to build on
existing literature by bridging the four Gaps mentioned herein above.
CHAPTER THREE
METHODOLOGY
3.1
Introduction
This chapter begins by presenting the theoretical framework for the TDH nexus.
This is followed by the model specification highlighting the justification for the
model used. Later, the chapter presents a number of econometric tests that were
undertaken.
3.2
Research Design
The study used non experimental research design. Specifically, the design used
was non-descriptive time series design. Quarterly time series data for the period
1970 to 2011 were used.
3.3
Theoretical Framework
The theoretical foundations of the connection between the budget and current
account deficits can be traced from both the national income identity as well as
the Mundel Flemming (M-F) framework. Virtually, all analyses of the twin deficit
hypothesis begin with a review ofa basic national accounting identity as
highlighted in equation 3.1.
Y=C + G + I + (X -
ML
(3.1)Where;
C- Consumption
I- Investment
G- Government expenditure
ex
-
M)- Net ExportsThe current account is given as
CA
=
X- M+
NtL {3.2)Where Ntr is net transfer. This component was assumed to be very small or
negligible for case of Kenya.
The national investment equation in an open economy is given as;
S
=
Y- C- G+
CA {3.3)National Investment is given as
I=Y-C-G (3.4)
Considering an open economy, the savings equation can be given as
S
=
I+
CA (3.5)Equation (3.5) depicts that an open economy can source domestically and
internationally for the funds needed for investment activities, that is, borrowing
can allow domestic investments to exceed domestic savings. Savings can also be
separated into private savings (S'") and government savings (sg)to get
S=Spr
+
sg
:
..(3.6)Spr= yd - C {3.7)
Private savings is given by the part of disposable income (yd) that is saved after
consumption. The government savings on the other hand is given as government
revenue in terms of taxes less expenditure in terms of expenditure (G) and
government transfers (Tr) as given in equation (3.8) below.
sg= T- G -Tr _ .__.(3.8)
•
Rising from the above identities, and having separated private savings from
government savings, equation (3.9) holds.
S=Spr
+
sg= I+
CA . ._. ...(3.9)Equation (3.9) provides a possibility for re-writing previous equations to a form
that is useful for analyzing the impact of government savings on an open
economy.
Spr=I
+
CA - sg= I+
CA - (T- G -TrL . .._. 0.10)Re-arranging equation (3.10)
CA
=
S'"- 1- (G+
T, -T) ..(3.11)Where, the expression (G
+
T, - T) in equation 3.11 represents consolidatedpublic sector budget deficit. It represents the extent to which the government is
borrowing to finance its expenses.
The equation in 3.11 above can be expressed in a simplified form as,
(X -M) =(S- I)
+
(T-GL
0.12)Where (X-M) is the current account balance, -(S-I) is the saving and investment
balance, (T-G) is the budget balance. Considering the assumption that S = I, the
current account imbalance may be attributable to a fiscal imbalance.
Equation 3.12 presents the framework under which changes in the budget deficit
•
would result in an increase in the latter if and only if the rise in government deficit
decreases total private saving. For instance, if (S - I) remains the same and tax
revenues (T) are constant, an increase in government spending (G) will positively
affect the current account balance. This way, the government deficits that result
from increased purchases tend to reduce the country's current account surplus,
worsening of the external balance.
Critics dismiss equation (3.12) as a simple identity equation and its estimation as
an irrelevant exercise. Further, others consider its estimation to be misspecified to
the scope that financial variables such as exchange rate and interest rates are
omitted and their role as transmission variables ignored. Proponents of the
Mundel Flemming approach believe that a deterioration of the budget deficit
causes the domestic interest rate to rise which results in net capital inflow leading
to an appreciation of the domestic currency and ultimately worsening of the
current account balance via a decline in net exports (Salvatore (2006),
This study considered that the transmission mechanism is important and should be
explicitly taken into account. The study therefore adopted the Mundel Flemming
approach in the analysis of effects of government policies to budget and current
account balances.
The primary paradigm for analyzing the effect of government policies in the short
run is the Mundel Flemming Model. It is an economic model set forth by Robert
Mundel and Marcus Fleming being an extension of the IS-LM Model. They
extended the Keynesian closed economy IS/LM model in order to predict the
impacts of monetary and fiscal expansions/ contractions under fixed and floating
exchange rate regimes. The model depicts the short-run association between an
economy's exchange rate, interest rate, and output. In addition to the variables
explained in the Keynesian Income equation above, the Mundell Flemming model
also uses the following variables:
M- Nominal money supply
P- Price level
r- Nominal interest rate
e- Real effective exchange rate
L- Liquidity preference/ real money demand
Y" - foreign income
The Mundell-Fleming model is based onthe following equations;
1)The goods market components (IS equation):
Y=C(Y-T)
+
I(r)+
G+
X(Y*, e)- M(Y, eL .... O.13)Equation 3.13 states that output is the sum of consumption, investment,
government expenditure and net exports.
The model now recognizes that the real exchange rate as the relative price of
demand is influenced by the real exchange rate, as well as by the interest rate,
income, and fiscal policy. Consumption is a function of disposable income while
investment is a function of interest rates. Exports depend on exchange rate and
foreign income. Imports on the other hand depend on exchange rates and
domestic income. The lower the exchange rate the less expensive domestic goods
are relative to foreign goods and thus the greater net exports are. The higher the
exchange rate the more expensive domestic goods are relative to foreign goods
and thus the smaller net exports are. Since the exchange rate is an endogenous
variable, the real exchange rate now adjusts to achieve goods market equilibrium.
2) The money market components (LM equation):
M/P
=
L(r, Y) {3.14)This equation defines the money market. It states that the supply of real money
balance (MIP) equals the demand L(r, Y). The demand for real money balances
depends negatively on interest rate and positively on income. Money supply (M)
is an exogenous variable controlled by the Central Bank. A higher interest rate or
a lower income (GDP) level leads to lower money demand. The domestic interest
rate is a function of the world interest rate and country risk. Real effective
exchange rate is used to provide the trade-weighted measure.
3) The balance of payment components:
BoP = CA
+
KA {3 .15)Where, BoP is the balance of payments, CA is the current account surplus, and
KA is the capital account surplus.
NX =NX
(e,
Y,Y*L
{3.16)Where; NX is net exports, e is the exchange rate, Y is GDP/ Output, and Y" is the
GDP of countries that are foreign trading partners. NX depends negatively on the
exchange rate. Higher domestic income (GDP) leads to more spending on imports
and hence lower net exports; higher foreign income leads to higher spending by
foreigners on the country's exports and thus higher net exports. A higher eleads to
higher net exports.
CA=NX {3.17)
Where, CA is the current account and NX is net exports. That is, the current
account is viewed as consisting solely of imports and exports.
K
.
A=
z (r-r*)+
k ~ {3.18)Where r* is the foreign interest rate, k is the exogenous component of financial
capital flows, z is the interest-sensitive component of capital flows, and the
derivative of the function z is the degree of capital mobility (the effect of
differences between domestic and foreign interest rates upon capital flows KA).
Considering the extra variables that come into playas far as the interaction
between the two deficits is concerned, Mundel Flemming provides an explanation
as follows. A cut in public expenditures in a small economy with a flexible
a) Reduction in aggregate domestic demand and consequently a fall in GDP,
b) Reduction in the transactional demand for money,
c) A downward pressure ondomestic interest rates,
d) Gap between international and domestic interest rates,
e) The difference in domestic and international interest rates causes capital
outflows and increase inthe demand for investment goods,
f) Depreciation of the domestic currency due to higher domestic demand and
lower interest rates,
g) Correction in both the trade balance (a proxy for the current account
balance) and the rest ofthe macro-economy, including the budget balance,
until the alignment ofdomestic with international interest rates is restored.
3
.
4
Model Specification
The study utilized a structural Vector Auto Regressive (VAR) model which
unlike other models doesn't- than necessary- impose restrictions to identify the
system. Sims (1980) described the other models as "incredible." The VAR' s
model is especially important because the variables are treated symmetrically in a
structural sense with each variable having an equation explaining its evolution
based on its own lags and the lags of the other variables in the model. Also, no
priori knowledge about the variables is required. VAR's modeling is also
advocated for because is it atheory-free technique (Sims, 1980). Furthermore; the
model has characteristic of simplicity and accuracy (Gordon, 1985).
VAR model took the form,
n· n n n ('" 19)
CAD
="
aCAD .+"/
3
ED .+""'EXRATE +" mINR +Jl J.l ~ 1 /-1 ~ J (-J ~ Y' /-k ~ -r I-I 1
;=1 .)=1 k=1 .[=1
Where;
CAD- Current Account Deficit
BD- Budget Deficit
EXRATE- Exchange Rate
INR - Interest Rate
3.5
Definition and Measurement
of Variables
The study utilized time series data spanning from 1960 Q 1 to 2011 Q4 which
included budget deficits, current account deficits, exchange rates and interest
rates.
Table 3.1: Variables definitions, measurements and data sources
S. Variable Definition Measurement Data Source
No
1 Budget Deficit- The difference between Data IFS, Statistical
(ED) national government expressed as a Abstracts, Economic
revenues and expenditures GDP Surveys
2 Current Account The difference between a Data -IFS, Statistical
Balance- (CAB) country's value of total expressed as a Abstracts, Economic
imports and value of total percent of Surveys
exports GDP
3 Exchange Rate- This is the value of one Quarterly real Central- Bank of
(EXRATE) currency for the purposes of exchange rate Kenya conversion to another between
Kenya and USA
4 Interest Rates- A rate which is charged or The 90 day Central Bank of
(INR) paid for the use of money CBK T-bill Kenya website
and is often expressed as an rate was used
annual percentage of the as a proxy of
principal. interest rates
5 Gross domestic The monetary value of all Quarterly IFS, Statisti cal
Product- (GDP) the fmished goods and value of GDP Abstracts, Economic
services produced within a growth in Surveys
country's borders in a shillings was
specific time period used
normally ayear
3.6
Da
t
a Collection
Quarterly time series data spanning from 1960
Q1
to 2011 Q4 were used. Wherequarterly data were not available, annual data were interpolated using cubic spline
interpolation method available in E-views statistical package. Data for exchange
rates and interest rates were collected from the Central Bank of Kenya. Current
account, GDP and budget deficit data were obtained from the IMF's International
Financial Statistics, statistical abstracts and economic surveys from the Kenya
National Bureau of Statistics.
3.7
Data Anal
y
si
s
The study pursued a number of empirical tests based on time series econometric
techniques. Diagnostics test such as unit roots were estimated. To establish
relationship between variables, VAR analysis was estimated before undertaking
cointegration analysis. To establish the direction of causality between the two
variables of study, causality analysis was done. The study utilized Toda and
Yamamoto (1995), MW ALD testing procedure to test for the causal inferences of
the variables. Lastly, impulse response analysis was carried.
Additionally, It was noted earlier that majority of past literature on the TDH have
concentrated on granger causality test to test for the causality on the two deficits.
These tests have been blamed for concentrating on time rather than causality itself
(Abdur and George, 2003). In most cases, time series data are non-stationary
which means that parameters estimated from the non-stationary variables do not
follow the standard statistical distributions for testing significance & have to be
differenced. Granger causality is done in
an
environment of Error Correction,(Granger, 1988). The Error Correction Models (ECM's) like Engel Granger ECM
and the Johansen & Juselius Vector Error Correction Models help omit
misrepresentation and omission of important constraints. The ECM's are sensitive
to the values of nuisance parameters in finite samples making the results are
unreliable, (Toda & Yamamoto, 1995) & involve cumbersome process which
Toda & Yamamoto (1995) proposed a procedure that allows for causal inferences
to be made at level VARs that may not be stationary without use of the rigorous
pretests and strict reliance upon cointergration and integration properties. They
propose a modified WALD -MW ALD- for testing granger non causality which
impose nonlinear restrictions in the properties ofthe VAR models needless to test
for unit roots and cointergration ranks. Not only is the procedure simple, it has
also been found to be superior to both the LR test, (Moscani & Gianni, 1992) and
the Wald test of Toda & Phillips (1993, 1994), as verified by Zapata & Rambaldi,
(1997).
i) Establish the maximum order of integration (dm(L'()
The Augmented Dicky-Fuller (ADF) test, the Phillips-Peron (PP) test and the
Kwiatkokowski, Phillips, Schmidt & Shin test (KPPS) were used. Both ADF and
PP tests are based on a null hypothesis of a unit root 1(1)while the KPSS test is
based on a null hypothesis of Stationarity 1(0).
ii) Determine the optimal lag length. (k).
The optimal lag length was chosen using information criteria like the Akaike
information Criterion (AIC), the Swartz Bayesian Criterion (SBC) and the Sim's
modified log-likelihood test (LR) test.
iii) Set up a VAR-model in the levels. The VAR model will be of Var(k+
dmax/h order. This is the augmented VAR-model.
iv) Carry out a Wald test for thefirst k variables only with k degrees of
freedom
A modified WALD (MWALD) tests for restrictions on the parameters of a Var
(k) -where k is the optimal lag length as determined in step 1 above- is done by
using the Toda & Yamamoto Procedure. A four equation VAR model of (k+
dma.ih order was set up using the level values of the data. This was regardless of
the order of integration of the various time series data. This test has been proven
to display asymptotic chi square distribution following the usual degrees of
freedom, (Chang & Hsu, 2009).
To estimate the role of interest rates and exchange rates, the study tested for
impact response functions and variance decomposition.
An impulse response refers to the feedback of any dynamic system in response to
an external change. As such, Impulse response functions (IRF) describe how the
economic variables responds over time to exogenous shocks. The analysis is
commonly used in the empirical literature to discover the dynamic association
amongst various variables within vector-autoregressive (VAR) models.
Response ofYi,lts to one-time impulse in Yjt with all other variables dated tor
Since our VAR
Model
has four variables, we shall analyze the response of ashock in any of the variables to the others. We conducted IRF tests in order to
map out the time path of the dependent variables in the model, to shocks from the
other independent variables.
Variance decomposition specifies the amount! proportion of information each
variable contributes to the other variables in the model. As an alternative method
to the IRF, it helps to see the impact of shocks in independent variables to the
dependent variables. While as the IRF is used to evaluate the effects of a shock to
one and thus transmitted to other endogenous variables, it cannot tell us the
magnitude of shocks in the system. The Variance decomposition procedure is
applied to define how much of the forecast error variance for any variable in a
system, is explained by innovations to each explanatory variable, over aseries of
time spans. Commonly own series shocks explain most of the error variance,
though the shock will also impact other variables in the system.
CHAP
T
ER FOUR
EM
P
I
R
ICAL FINDINGS
4.1 Introduction
This chapter presents results of the study. The diagnostics are presented first
followed by cointegration results, causality, and impulse response. The
diagnostics consisted of econometric tests of unit root and lag length
determination. Co integration used Johansen test while for causality analysis, Toda
Yamamoto technique was used.
The empirical studies by Egwaikhide et al., (2002) and Kosimbei, (2002)
employed VAR model and error correction models. Borrowing from those
studies, this study utilized the VAR model which contains information on the
variables themselves, their interactions and the dynamic short-run and long-run
relationships. However, this study used a different model from those of
Egwaikhide et al., (2002) and Kosimbei, (2002) to test for causality between
budget deficit and current account deficit.
4.2 Diagnostics
4.2.1 Unit root test
The study first investigated the integration properties of the data since most of
economic time series data are non-stationary. Augmented Dicky-Fuller test
statistics were used. The lag lengths for the ADF equations were chosen by the
Modified Akaike Information Criterion (AlC) and Schwartz Bayesian Criterion
(SBC). The lag lengths for PP and KPPS tests were chosen using Newey-West
Bandwidth and Andrews Bandwidth.
Table 4.2 presents the results of the ADF unit root test. As shown in Table 4.2, the
variables under study were found to be integrated of order one, I (1). (See detailed
unit root test results in Appendix III).
Table 4.2: Augmented Dickey-Fuller results
Test critical values: 5% level
LAG LENGTH
CRITERIA SBC MODIFIED AIC
\
PROBABILITY AT 1ST AT AT 1ST
VARIABLE INCLUDED AT LEVELS DIFFERENCE LEVELS DIFFERENCE
BD Intercept 0.0828 0.0004 0.4267 0.0000
Trend &
Intercept 0.0757 0.0024 0.6391 0.0000
None 0.2524 0.0000 0.3345 0.0000
CAB Intercept 0.0049 0.0005 0.142 0.0005
Trend &
Intercept 0.0250 0.0037 0.3768 0.0037
None 0.0556 0.0000 0.0806 0.0000
EXRATE Intercept 0.918 0.0038 0.9045 0.0337
Trend &
Intercept 0.3763 0.0195 0.3752 0.0265
None 0.932 0.0010 0.9116 0.0124
INRATE Intercept 0.1821 0.0103 0.1821 0.0000
Trend &
Intercept 0.2354 0.0404 0.2354 0.0000
None 0.2749 0.0006 0.2749 0.0000
The four variables namely, budget deficits, current account balance, interest rates
and exchange rates were found to have a unit root at levels. They were however
found to have no unit root at first difference. Hence the variables became
stationary on the first difference.
4.2.2 Lag Length Selection
The maximum lag length was chosen using Akaike information Criterion (AlC),
the Swartz Bayesian Criterion (SBC) and the Sim's modified log-likelihood test
(LR) test. Table 4.3 illustrates the lag selection results.
Table 4.3: Optimal Lag Length Selection
Hannan
-Modified Final Akaike Schwarz Quinn
LRtest prediction information information information
Lag LogL statistic error criterion criterion criterion
0 -2101.812 NA 3176036. 26.32265 26.39953 26.35387
1 -1068.928 200l.213 9.584755 13.61160 13.99600 13.76769
2 -808.3828 491.7792 0.451018 10.55478 11.24670 10.83575
3 -802.7706 10.31237 0.514048 10.68463 11.68406 11.09047
4 -789.2295 24.20470 0.531040 10.71537 12.02232 11.24608
5 -736.3815 91.82347 0.335999 10.25477 11.86923 10.91035
6 -601.897l 226.9424 0.076729* 8.773714* 10.69570* 9.554165*
7 -597.3614 7.427207* 0.089074 8.917018 11.14652 9.822341
8 -588.5393 14.00502 0.098200 9.006742 11.54376 10.03694
9 -567.4779 32.38193 0.093112 8.943474 11.78801 10.09854
* Indicates lag order selected bythe criterion
Four Information criteria suggested that a maximum lag length of 6 for each
variable. However, additional test was carried to determine the number of lag