Accepted Manuscript

Title: The Twin Deficit Hypothesis in Egypt

Author: Heba E. Helmy

PII: S0161-8938(18)30015-2
DOI: https://doi.org/10.1016/j.jpolmod.2018.01.009
Reference: JPO 6409

To appear in: Journal of Policy Modeling

Received date: 16-8-2017

Accepted date: 12-1-2018

Please cite this article as: & Helmy, Heba E., The Twin Deficit Hypothesis in
Egypt.Journal of Policy Modeling https://doi.org/10.1016/j.jpolmod.2018.01.009

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
The Twin Deficit Hypothesis in Egypt

Heba E. Helmy

Department of Economics, Faculty of Management Sciences

October University for Modern Sciences and Arts (MSA)

6th of October City, Egypt.

T
IP
hebaezzhelmy@yahoo.com

R
hezz@msa.eun.eg

SC
U
N
We employ a new approach to the twin deficit hypothesis aimed at enhancing policy
making in Egypt. In contrast to the conventional twin deficit hypothesis between the
current account, which comprises many items out of governments’ scope of
A
maneuvering, and the budget deficit, we track the causal link between Egypt’s
M

merchandise trade deficit and the budget deficit. We begin first by examining the
conventional twin deficit hypothesis using a VAR model, which implies short run
reverse causation running from the current account deficit to the budget deficit.
Second, as cointegration exists between the budget deficit and the merchandise trade
ED

deficit, we run a multivariate VECM model which refutes the twin deficit hypothesis
in favor of the current account targeting hypothesis. In policy terms, ameliorating
Egypt’s trade balance would ultimately improve its fiscal balance as well.
PT

Keywords: Egypt, twin deficits, merchandise trade deficit, cointegration, Granger causality,
E

Vector Autoregression (VAR), Vector Error Correction (VECM).

CC

JEL Classifications: C32, E62, F32, H62, O53

A

1. Introduction

Historically, the twin deficit hypothesis evolved from early studies tackling the so-

called dual-gap analysis. Such studies (which began by Chenery et al 1966) suggested
1
that if a country at the early stages of development is forced to import more than it

exports, resulting in an import-export gap exceeding its national saving capacity, then

this country will be forced to borrow from abroad, thereby increasing its saving

investment gap. The theory was a justification why most developing countries face a

foreign debt problem at their early stages of development. The above argument can be

expressed in the simple national accounting equation:

T
Income (Y) = Consumption (C) + Investment (I) + Exports (X) – Imports (M) (1)

IP
But since S = Y – C (where S = private saving), therefore

R
S = I + X – M, or by rearranging

SC
I – S = M – X. (2)

U
The above equation means that if a country has a balance of payments deficit then this
N
will necessarily imply that the resources it utilizes is more than those supplied to it;
A
alternatively, it may indicate that if a country saves more than it invests then this will
M

appear in the balance of payments account (Thirlwall 2003).

Dual gap analysis later developed into the twin deficit hypothesis. In the above
ED

model, splitting the private and public consumption and saving, and adding the

government sector as one of the components of effective demand would yield the
PT

following Keynesian model of an economy:

Y=C+G+I+X–M (3)
E
CC

where

Y= national income
A

C= private consumption

G=government expenditure

I= investment

(X-M) = current account surplus or deficit

2
Since Y = C + S + T (where T = taxes), therefore by rearranging:

(X – M) = (S – I) + (T – G) (4)

The above equation implies that the difference in the current account has to equal the

difference in national saving whether private or public. While the above equation is

true from an accounting perspective, it does not explain the independent and the

dependent variables, the direction of causation and the magnitude of changes that take

T
in the relevant variables place until equilibrium is ultimately attained.

IP
The twin deficit hypothesis supposes that if a country suffers a budget deficit

R
(G>T) then it has to suffer a trade deficit as well (M>X), assuming S and I are

SC
constant. The logic behind this is that a large budget deficit will lead to a rise in

U
interest rates and therefore to an increase in demand on the domestic currency; the
N
resulting appreciated exchange rate will ultimately lead to a fall in exports and a rise
A
in imports or in other words a current account deficit (The Mundell-Fleming
M

framework). The current account deficit can also result from a budget deficit if an

increase in government expenditure or a decrease in taxes (which led to the deficit)

ED

causes a rise in national income and therefore to an increase in imports. Hence, the

twin deficit hypothesis has its basis in the Keynesian model of an open economy.
PT

Opponents of the twin deficit hypothesis refute the rigidity of (S) and (I), and base

their theories on their flexibility implying that an increase in government deficit

E
CC

would force either (S) to rise or (I) to fall; the positive (S-I) balance would offset the

negative (T-G) balance thereby equating the right hand side of the equation and
A

preventing any effect on the external account (X-M). Advocates of this hypothesis –

or opponents of the twin deficit hypothesis – base their view on the Ricardian

Equivalence hypothesis. This hypothesis, which was developed by Robert Barro

(1974) assumes that, under certain conditions, an increase in government deficit

3
financed through issuance of bonds would not lead to an increase in income nor to an

increase in interest rate (and therefore to increased spending on imports) as people

would save the revenue generated from the expansion expecting that the government

would increase future taxes to finance the deficit. Within the same context rejecting

the twin deficit theory came the Feldstein-Horioka puzzle. The Feldstein-Horioka

puzzle that first appeared in a study in 1980 on 16 OECD countries for the period

T
1960-1974 suggested a strong correlation between national saving and investment

IP
despite the existence of worldwide capital mobility (and hence the puzzle). Changes

R
(or precisely decreases) in national savings are mirrored in similar changes (or

SC
decreases) in investment and thus produce no impact on the current account (Coakley

U
el al, 1998). A deficit in overall national saving (both private and public) would be
N
correlated by a one-to-one fall in investment (I), leaving the current account
A
unchanged.
M

Other possibilities, which were reflected in the opinions of other economists,

included a unidirectional reverse causality from the current account deficit to the
ED

budget deficit. The logic behind this relationship is that deterioration in the current

account would negatively affect economic growth and worsen the budget deficit, a
PT

phenomenon evident in small open economies depending on foreign loans to finance

investment and economic development. The accumulation of loans will inflate debt
E
CC

service payments in the current account, leading to current account deficit, and

ultimately to budget deficit to finance it. This reverse causation was referred to by
A

Summers (1988) as the ‘current account targeting’. A third possibility lies in the

existence of bi-directional causality from the two deficit to each other, a possibility

built upon the simultaneous existence of the factors that support mutual causation;

while the last possibility disapproved the existence of causality from any direction, a

4
possibility built in this case upon the simultaneous existence of the factors that refute

mutual causation.

The twin deficit hypothesis was ensued by a plethora of studies on both developed

and developing countries with no conclusive results. One of the main reasons for this

ambiguity we believe is that many of these studies equated the current account with

the external balance of a country comprising the current account, the capital account

T
and the financial account, assuming that the external balance can be simplified into

IP
the current account when in fact it cannot. The second reason is that even when the

R
current account itself is examined, conclusions may be ambiguous as that account

SC
itself encompasses many variables that are exogenous to any logical causation

U
between the two deficits. We thus attempt in this study to unravel such variables from
N
the current account by, first, examining the conventional causal relationship between
A
the ratio of the budget deficit to GDP and the ratio of the current account deficit to
M

GDP; and, second, reexamining the causal relationship between budget deficit to GDP

and a component of the current account, which is the merchandise trade deficit, taking
ED

Egypt over the period 1975 till 2014 as our case study. This novel approach to

understanding the twin deficit hypothesis is particularly suitable for Egypt. The reason
PT

for this is that in Egypt the main variables that account for the differences between the

current account and the merchandise trade balance are the services trade balance (that
E
CC

mainly includes the revenues from tourism and the Suez Canal), remittances or

transfers from Egyptians working abroad and debt service payments. If we examine
A

the main channels by which the budget deficit influences the current account deficit,

or vice versa, those three variables seem to be historically irrelevant or at least

immune to any changes in such channels. Contemporary incidents such as the Russian

passenger air flight crash over Sinai in 2015 which harshly inflicted the tourism sector

5
in Egypt, or changes in the value of remittances of Egyptians working in the Arab oil

producing countries in response to changes in oil prices, all prove that such 'primarily

exogenous' variables are unaffected by government balances. Encompassing such

variables may distort any possible links between the two accounts, implying a

relationship when there is not, or a disconnection when in fact there is. In addition to

being beneficial in clarifying the association of the two deficits, disentangling these

T
variables is advantageous in policy making, as governments usually have weak

IP
influence on such exogenous variables, and relatively stronger influence on other

R
endogenous ones through fiscal or monetary policies. Accordingly, elucidating the

SC
impact of government policy (and deficit) on the only affected variables in the current

U
account would help rectify government action and provide relevant policy
N
recommendations. Thus while earlier studies were concerned with splitting the right
A
hand side of the equation Eq. (4) (private and public saving), our study is concerned
M

with splitting the left hand side; making Eq. (4) in the form:

(X – M)m+ (X – M)s + TR + DBSER = (S – I) + (T – G) (5)

ED

where

TR = remittances from Egyptians working abroad

PT

DBSER = debt service payments

(X-M)s= balance of services account

E
CC

Assuming the above three terms to be exogenous, the above equation can be written

as:
A

(X – M)m= (S – I) + (T – G) (6)

Renaming the terms for simplification, the above equation could be expressed as:

TmD = (S – I) + BD (7)

Where

6
(X – M)m = TmD

(T – G) = BD

Our study thus attempts to examine the short run and long run relationship between

the merchandise trade account (TmD) and the government deficit (BD) by testing

whether cointegration exists between the two variables in Egypt using annual time

series data from 1975 until 2014. Our new approach is innovative in another respect,

T
as we incorporate in our multivariate Vector Error Correction model time series of

IP
both the exchange rate (ER) and the interest rate (INT), as both private saving and

R
investment are dependent on them. The aim is to shed light on how the transmission

SC
mechanism takes place (if any) between the two deficits (following Kalou and

Paleologou). Eq. (7) can thus be rewritten as

TmD = S(INT, ER) – I(INT, ER) + BD U

N
(8)
A
Our study will be structured as follows: Following this introduction and theoretical
M

background, section two briefly reviews the literature on the twin deficit hypothesis

over the last decade, section three expounds the empirical methodology used and
ED

highlights the findings of the paper, while the last section concludes by providing

relevant policy recommendations.

E PT

2. A review of the relevant literature after 2005

CC

While studies on twin deficit hypothesis in Egypt are rather scant, studies that
A

tackled this topic across countries ever since the eighties are innumerable. For reasons

of parsimony, we compile in the following table some selected studies on the twin

deficit hypothesis over the last ten years listed chronologically. The selection was

7
based on the studies that encompass diversity of findings, multiplicity of

methodologies and inclusion of our case study, Egypt.

Similar to the diversity of results on twin deficits across countries, the three studies

that tackled Egypt also produced conflicting evidence. While Hashemzadeh and

Wilson (2006) and Marinheiro’s (2008) studies both rejected the twin deficit

T
hypothesis for Egypt, the latter’s study supported a reverse causality from the current

IP
account deficit to the budget deficit. On the other hand, Ahmad, Aworinde, and

R
Martin’s (2015) study confirmed positive cointegration or the twin deficit hypothesis

SC
for Egypt.

U
N
3. Empirical Analysis
A

3.1 Egypt’s budget and current account deficits

M

To what extent are Egypt’s external balances correlated with her internal ones? In
ED

fact, a quick glance at Fig. 1 depicting the budget deficit and the current account
PT

deficit (both as ratios of GDP), and Fig. 2 depicting the budget deficit and the

merchandise trade deficit (also as shares of GDP) over the last decades emphasizes
E

how a strong correlation exists in the second case compared to a somewhat weaker
CC

correlation in the first. At start, we attempted to analyze whether cointegration existed

between Egypt’s budget deficit to GDP (or BD) and current account deficit to GDP
A

(or CAD). Data on Egypt’s CAD was obtained from UNCTAD online1 while BD was

1
UNCTAD defines the variable as follows: “Balance-of-payments current account
data cover all transactions between residents and non-residents of a reporting
economy, involving economic values and mainly concerning goods, services, income
and current transfers. In general, the current account balance describes the difference
8
calculated by the author depending on official national statistics from the Central

Bank of Egypt (CBE) and the Central Agency of Public Mobilization and Statistics

(CAPMAS). However, the current account of Egypt, which witnessed some years of

surplus and many years of deficit, was stationary, or I(0), as the series did not show

evidence of the existence of unit roots. On the other hand, the budget deficit to GDP

ratio was non-stationary and integrated of order I(1) according to Augmented Dickey

T
Fuller (ADF) and Philips Perron (PP) unit root tests (not reported here). Accordingly,

IP
cointegration between the two variables was not possible. We therefore began by

R
assessing the relationship between the two variables through an unrestricted Vector

SC
Autoregression (VAR) model. Inclusion of dummy variables to account for Egypt’s

structural adjustment program (ERSAP) in 1991, and for the 25th of January
U
N
revolution in 2011 (not reported here) proved insignificant and therefore we excluded
A
them from the model specification.
M

3.1.1 Vector Autoregression analysis

ED

The VAR model treats all variables as endogenous, and explains the behavior of
PT

one variable in terms of the lagged values of the other variables. To assess how the

transmission mechanism (if any) takes place between the two variables we
E

incorporated two important tools, the interest rate and the exchange rate into our
CC

system. Accordingly, our VAR system takes the following form:

A

between current receipts and expenditures for internationally traded goods, services
and income payments. At the same time, from a national perspective, the current
account balance would equal the gap between national savings and domestic
investment” (UNCTAD online).

9
 n n
BDt  a 0   a1i (1  L) BDt 1   a 2i (1  L) INTt 1 
i 1 i 1
n n

a 3i (1  L) ERt 1   a 4i (1  L)CADt 1  u t
i 1 i 1 (9)

n n
INTt  b0   b1i (1  L) BDt 1   b 2i (1  L) INTt 1 
i 1 i 1
n n

b (1  L) ERt 1   b4i (1  L)CADt 1  et

T
3i
i 1 i 1 (10)

IP
n n
ERt  c0   c1i (1  L) BDt 1   c 2i (1  L) INTt 1 

R
i 1 i 1

SC
n n

c 3i (1  L) ERt 1   c 4i (1  L)CADt 1   t
i 1 i 1
(11)

CADt 
n n
U
d 0   d1i (1  L) BDt 1   d 2i (1  L) INTt 1 
N
i 1 i 1
n n
A
d
i 1
3i (1  L) ERt 1   d 4i (1  L)CADt 1   t
i 1 (12)
M

Where:
ED

BD: is the ratio (calculated by author) of government deficit (revenues minus

PT

expenditures) in year (t) over the GDP in year (t), all in Egyptian pounds, and data

being obtained from official sources. The budget deficit is obtained from Central
E

Bank of Egypt’s (CBE) annual reports (various issues of online and printed editions),
CC

while GDP is obtained from Arab Republic of Egypt Statistical Yearbook published

by the Central Agency from Public Mobilization and Statistics (CAPMAS) (various
A

issues of online and printed editions). It is worth noting that the values for BD are

somewhat different (usually higher) from the values of the ratio issued by government

10
sources. Yet we believe that this method of calculating the budget deficit is more

reliable than the readily available figures by official sources2.

CAD: is the ratio current account to the GDP in year (t), data being obtained from the

United Nations Conference on Trade and Development (UNCTAD) online.

INT: is average annual interest rate (the discount rates by CBE at end-of-year rates),

T
data being obtained from the International Financial Statistics (IFS) of the

IP
International Monetary Fund online.

R
ER: is the natural log of the nominal exchange rate or the value of LE 1 in terms of

SC
dollars, data being obtained from the International Financial Statistics (IFS) of the

International Monetary Fund. Raw data that was provided in dollars was converted to

U
Egyptian pounds (that is, the value of 1LE in terms of dollars rather than vice versa)
N
for consistency with other variables. Values were then transformed to natural logs.
A
M

(1-L) is the lag operator

ut, et, εt and υt are random error terms

ED

We start first by computing the appropriate lag length for our system (Table 2).
PT

The various tests gave different lag preferences, as the Schwartz information criterion

favored one lag, the LR test preferred two lags, while the Hannan-Quinn information,
E

Akaike information criterion and the Final prediction error tests favored four lags.
CC

Since our sample comprises only 35 annual observations and to prevent consuming
A

2
Marinheiro (2008) casts some doubts on the reliability of official government figures
on the percentage of budget deficit to GDP. For example, he states that two years
before Egypt’s adoption of the IMF stabilization program in 1991, government
sources put that deficit at 5.5% of GDP while some independent researchers put the
average of the last three years preceding the signing of the agreement at 15% of GDP.
If the official figure was true, the government would not have needed to sign the
agreement, according to Marinheiro (p.1043).

11
many degrees of freedom, we based our selection on the Schwartz test of only two

lags. The VAR satisfied the stability condition, as no root lied outside the unit circle.

We proceed by estimating the VAR with two lags, which appears in Table 3.

The first equation in our VAR model which has the budget deficit to GDP as the

T
dependent variable purports that the budget deficit is affected by the lagged values of

IP
the current account deficit. In fact all three variables explain 83% of the changes in

R
the budget deficit. On the other hand, the equation which had the current account

SC
deficit as the dependent variable did not provide any evidence of the impact of the

U
budget deficit to the current account. In fact, there was no impact from any of the
N
other three variables on the current account, which emphasizes the relative exogeneity
A
of this variable. The previous two equations pinpoint to the fact that causality runs
M

from the current account deficit to the budget deficit rather than vice versa.

It is interesting to note that the equation of interest rate in the VAR system
ED

illustrated a very significant impact of the exchange rate on the interest rate, reflecting

a conventional policy by successive Egyptian governments which respond to

PT

deteriorating exchange rate by raising the interest rate. The last equation of the

exchange rate signals the exogeneity of the exchange rate similar to the current
E
CC

account deficit.
A

3.1.2 Granger causality tests

To confirm the robustness of the results we performed Granger causality tests on

each pair of variables in our system (Table 4). As evident from the table the null

hypothesis that the current account deficit does not Granger cause the budget deficit

12
was rejected at the 5% significance level, whereas the null hypothesis that the budget

deficit does not Granger cause the current account deficit failed to be rejected

ascertaining the previous results of the VAR. Two important inferences from the

Granger tests is the causation that goes from the interest rate to the budget deficit and

from the exchange rate to the interest rate.

T
The previous VAR analysis and Granger causality tests both refute the twin deficit

IP
hypothesis in favor of the alternative hypothesis, the current account targeting. The

R
method of transmission may begin first with a deficit in the current account, which

SC
decreases the exchange rate, forces the government to raise interest rate to prevent

U
further deterioration of the pound (and dollarization on the part of the citizens),
N
ultimately leading to larger government deficit due to higher interest rate on new
A
debts.
M

3.2 Egypt’s budget and merchandise trade deficits

ED
PT

In this section, we have a closer look on the relationship between the budget deficit

and the merchandise trade deficit for reasons explained earlier.

E
CC

3.2.1 Cointegration between the budget deficit and the merchandise trade deficit
We first examine whether cointegration exists between the budget deficit and
A

merchandise trade deficit by removing other components of the current account such

as the balance of services, investment income, debt service payments and private

transfers or remittances. If the variables are non-stationary and cointegrated of the

same order, a VECM representation can examine their causal relationship as follows:

13
 n n
BD t  a 0  a1 ECTt 1   a 2i (1  L)BD t 1   a3i (1  L)INTt 1 
i 1 i 1
n n

a 4i (1  L)ER t 1   a5i (1  L)TmD t 1  u t

i 1 i 1 (9)

n n
INTt  b0  b1 ECTt 1   b2i (1  L)BD t 1   b3i (1  L)INTt 1 
i 1 i 1
n n

b (1  L)ER t 1   b5i (1  L)TmD t 1  et

T
4i
i 1 i 1 (10)

IP
n n
ER t  c0  c1 ECTt 1   c 2i (1  L)BD t 1   c3i (1  L)INTt 1 

R
i 1 i 1

SC
n n

c 4i (1  L)ER t 1   c5i (1  L)TmD t 1   t

i 1 i 1
(11)

TmDt 
n n
U
d 0  d1 ECTt 1   d 2i (1  L)BD t 1   d 3i (1  L)INTt 1 
N
i 1 i 1
n n
A
d
i 1
4i (1  L)ER t 1   d 5i (1  L)TmDt 1   t
i 1 (12)
M

Where:
ED

BD, INT, ER, (1-L), ut, et, εt and υt are the same as before.
PT

ECT is the error correction term

E

TmD is the ratio (calculated by author) of the merchandise trade deficit in year t over
CC

the GDP in year t, all in Egyptian pounds, both variables being obtained from the

Central Agency from Public Mobilization and Statistics (CAPMAS) publication of

A

Arab Republic of Egypt Statistical Yearbook (various issues of online and printed

copies).

14
3.2.2. Unit root tests

To test for stationarity of the variables, standard unit tests were performed on the

variables. Both the Augmented Dickey Fuller and Phillips-Perron tests indicated the

existence of unit roots as the tests in levels failed to reject the null hypothesis of the

existence of unit roots implying that the variables are non-stationary. All series turned

stationary when first differenced indicating that all the four series are integrated of

T
order one or I(1) (Tables 5 and 6).

R IP
The VAR model was selected with two lags. The selection was based on the results

SC
of six lag selection tests. Four results out of the six supported the selection of two lags

(Table 7).

U
N
3.2.3 Cointegration results
A

We use the Johansen cointegration procedure to test whether the four variables are
M

simultaneously cointegrated. Two tests of cointegration were conducted, the trace and
ED

max-eigen33. Both tests rejected the null hypothesis of no cointegration, and suggested

one cointegrating relationship in the four variable model at the 5% significance level
PT

(Table 8).

3.2.4 VECM analysis

E

Since the four variables are cointegrated, implying that a long run relationship
CC

exists among them, then this relationship can be represented through an error
A

correction model (ECM). The Vector Error Correction Model (VECM) representing

the four variables appears in Table 9. Similar to the situation with the previously

3
The choice of whether the cointegrating vector and the VAR includes or does not
include a concept and/or a trend was decided by the Pantula principle.

15
estimated unrestricted VAR model comprising the current account deficit and the

other three variables (with the exception of the merchandise trade deficit), inclusion

of dummy variables to account for Egypt’s structural adjustment program (ERSAP) in

1991, and for the 25th of January revolution in 2011 (not reported here) proved also

insignificant; hence, we excluded them from the model specification.

T
R IP
SC
The long run error correction term ECT for Eq. (9) in our VECM where the budget

U
deficit is the dependent variable was equal to -0.032663 and its p-value is 0.0283
N
which is significant at 5% level, attesting to the long run causality that runs from the
A

interest rate, the exchange rate and the trade deficit to the budget deficit. In the short
M

run, the budget deficit was temporarily impacted by previous changes in the
ED

merchandise trade deficit (although this impact was significant only at the 10% level)

in addition to being influenced by its own previous deficits. The model was
PT

significant as the R2 was 0.517 and F-statistic at 3.223. The model also passed all

diagnostic tests displaying normality, no serial correlation and no heteroskedasticity,

E

in addition to passing a stability test44. The second model where interest rate is the
CC

dependent variable is also significant and has an R2 of 0.649; however the residuals
A

do not pass the normality test. The third model which has the exchange rate as the

dependent variable has a relatively low R2; in addition, the residuals suffer from non-

normality. The forth model which has the trade deficit as the dependent variable (Eq.

4
Cusum test of stability

16
12), and which practically tests the twin deficit hypothesis, does not exhibit long run

causality relationship from the other three variables to it, despite the fact that it passes

all the diagnostic tests. The merchandise trade deficit also displays some short run

influences from previous changes in the exchange rate in addition to responding to its

own previous changes. Thus models (9) and (12) are verified and imply that there is a

long run impact from the trade deficit to the budget deficit and not vice versa.

T
IP
To check the robustness of the results especially with respect to the short run

causality, we conducted the Wald test to examine whether the lagged values of each

R
SC
variable jointly influence the budget deficit. We conducted the same exercise when

the other variables were the dependent variables. The results for the Wald test

U
revealing the joint significance of the lagged values of each variable on the dependent
N
variable appears in Table 10.
A
M
ED
PT

As can be gleaned from the table, in the short run the budget deficit is influenced by
E

its own previous changes. The merchandise trade deficit does not appear to impact it
CC

probably because Table 9 revealed its influence only at the 10% significance level. On
A

the other hand, the merchandise trade deficit is not influenced by any short run

changes in the other three variables. It is interesting to note that the interest rate is

induced by previous changes in exchange rate in addition to previous changes of its

own.

17
 To ascertain the robustness of the results we conducted granger causality tests. As

can be inferred from Table 11, both budget deficit and merchandise trade deficit seem

to influence each other in the short run. However, the impact of the budget deficit on

the merchandise trade deficit is short lived lasting for only two periods, whereas the

merchandise trade deficit has a deeper impact on the budget deficit lasting for four

periods. The results of the Granger tests attest to some bi-directional relationship in

T
the short run between the two deficits, unlike that inferred from the Wald tests. The

IP
Granger tests confirmed the Wald tests on the strong short and medium run causality

R
running from the exchange rate to the interest rate. Lastly, Granger causality tests

SC
hinted to some sort of causality from the exchange rate to the merchandise trade

U
deficit (not reported here) at the 10% significance level, rather than vice versa.
N
A
M

4. Conclusion
ED

This paper employs a new approach to the twin deficit hypothesis aimed primarily

at enhancing policy making in Egypt. While the conventional twin deficit hypothesis
PT

attempts at discovering the causal link between the budget deficit and the current
E

account deficit, we believe that detecting such a link is futile in a country like Egypt
CC

where most of the current account variables are affected by factors out of the of

governments’ scope of maneuvering. Alternatively, we attempt to track the causal link

A

between Egypt’s merchandise trade deficit and its budget deficit. The merchandise

trade deficit, in contrast to other components of the current account, namely the

balance of services, transfers of Egyptians working abroad and debt service on

previously accumulated loans, is a sector which represents the real productivity of the

18
economy and the outcome of previous and current economic development policies. A

strong export sector, reflected in a balance of goods surplus, is a success record for

long and concerted efforts for development. If that sector was affected one way or

another by the budget deficit, governments should remodel their policies into reducing

the deficit, as the outcome of the deficit would mean hampering economic

development plans by inhibiting the growth of the exports and/or inflating the growth

T
of imports. To prove our point and explore this link we begin first by examining

IP
whether the budget deficit and the current account deficit are integrated in a long run

R
relationship. Such a link was not verified as expected (since the current account ratio

SC
to GDP proved stationary) and long run cointegration between the two deficits could

not be confirmed for the reasons mentioned earlier55. Following the non-existence of
U
N
cointegration, we estimated the influence of the two variables on each other by
A
employing an unrestricted VAR, which lay some evidence of a short run reverse
M

causation running from the current account deficit to the budget deficit. A second

route of causation ran from the exchange rate to the interest rate. Both routes of
ED

causation were supported by Granger causality tests, which added a third possible

route from the interest rate to the budget deficit.

PT

We proceeded to examine multivariate cointegration between the budget deficit

and the merchandise trade deficit both as ratios of GDP, together with the
E
CC

5
In fact, Egypt’s very recent history proves that inflows of billions of US dollars
from the Gulf and Western countries in the second half of 2013 was mainly due to
political factors rather than economic factors, let alone factors that deal with
A

government balances. Similarly, the main foreign exchange sector in the current
account, tourism, was and remains very vulnerable to any political factors that
enhance or deteriorate that sector; the 1997 Luxor attacks or the 2015 Russian plane
crash are exemplary of such vulnerability making the sector immune to any economic
stimulants on the part of governments. Finally, remittances of Egyptians working in
Gulf countries are primarily dependent on whether such countries face revenues or
deficits in their own budgets as a result of changes in oil prices.

19
transmission tools of the interest rate and the exchange rate, The study proved that the

four variables are cointegrated in a long term relationship through one cointegrating

equation. However, as cointegration does not show the direction of causation we ran a

VECM representation which refuted the twin deficit hypothesis in favor of the current

account targeting hypothesis. In other words, the worsening of budget deficit does not

lead to worsening the trade deficit. On the contrary, it is the trade deficit, together

T
with the other variables, that leads to worsening the budget deficit in the long run.

IP
Equally important is what the VECM proved about the long run relationship from the

R
other three variables to the interest rate. In the interest rate equation, the coefficient of

SC
the exchange rate was both statistically and economically significant. Both routes of

U
causation were supported by Granger causality tests, which added a third possible
N
route from the interest rate to the exchange rate. In other words, in the short run the
A
interest rate and the exchange rate might mutually affect each other. The tests also
M

hinted to some bidirectional causality between the budget deficit and the merchandise

trade deficit in the very short run; however, when longer lags are included it is only
ED

the trade deficit which influences the budget deficit. Such tests however were not

supported by Wald tests which pinpointed to only short run causality running from the
PT

exchange rate to the interest rate. Our study is thus in line with Hashemzadeh et al and

Marinheiro on the refutation of the twin deficit hypothesis in Egypt. It is also in line
E
CC

with Marinheiro on the reverse causality running from current account (in our case the

merchandise trade deficit) to the budget deficit. However, our study partially
A

contradicts with Marinheiro’s on the existence of cointegration between the current

account and the budget deficit in support for cointegration between the balance of

goods deficit and the budget deficit. Naturally our findings contradict with Ahmad et

al’s study in supporting the twin deficit hypothesis. Finally, it is worth noting that the

20
study proved the inapplicability of the Ricardian equivalence hypothesis to the

Egyptian case simply because interest rate does not respond to changes in the budget

deficit but merely to changes in the exchange rate6.

Summing up the above discussion implies that the current account targeting

hypothesis is materialized in Egypt through two possible transmission routes. The first

T
seems to begin with trade deficit (being the most exogenous of the four variables)

IP
which worsens through time as a result of rising food and energy imports resulting

from a rising population, and a weak export sector. The rising trade deficit pressures

R
SC
the exchange rate to deteriorate. To prevent huge depreciations of the pound and

dollarization, governments resort to raising interest rate. Higher interest rate inflates

U
future debts and results in higher budget deficits. The second route of transmission
N
starts with the exchange rate rather the trade deficit as hinted by Granger causality
A
tests. Since there is some indication that in short run, the exchange rate affects trade
M

deficit, thus a shock in exchange rate, maybe as a result in the fall in tourism or a rise

in food or oil prices, leads to a deterioration in the trade balance, forcing the
ED

government to borrow to finance the desperately needed imports. In all cases, it is

always the budget deficit which adjusts to the external deficit, whether it is the current
PT

account deficit or the merchandise trade deficit.

E

The main policy recommendation which we conclude from this study is that
CC

Egypt’s current account, but more importantly Egypt’s merchandise trade account,

6
This conclusion should not be surprising given the fact that the Central Bank of
A

Egypt is not totally independent. The chairman of the Central Bank and the CEOs of
the public sector banks which finance the government budget deficit are all appointed
by the government. There is no pressure on the part of successive governments to
raise interest rate to finance budget deficits as banks will abide by the government
demands anyway. However, the government is forced to raise interest rate on the
Egyptian pound whenever demand on the dollar rises to prevent further depreciation
of the pound.

21
must be ameliorated. Lost decades of economic development left Egypt with a weak

export sector with little to produce and sell, and a proliferating import sector due to

limited agricultural, energy and water resources. Egypt should go into manufacturing

and capital deepening to enhance its export sector, replicating models of small

productive economies with limited resources such as Singapore and South Korea.

Restructuring Egypt’s trade balances would ultimately improve its fiscal balances as

T
well.

R IP
SC
U
N
A
M
ED
E PT
CC
A

22
References

Ahmad, A. H., Aworinde, O. and Martin, C. (2015) Threshold cointegration and the
short-run dynamics of twin deficit hypothesis in African countries. The Journal
of Economic Asymmetries, 12, 80-91.
Baharumshah, A., Lau, E. and Khalid, A. (2006) Testing twin deficit hypothesis using
VARs and Variance Decomposition. Journal of Asia Pacific Economy. 11 (3)
331 – 354. August.
Barro, R. (1974) Are government bonds net wealth? Journal of Political Economy.
82(6) 1095-1117.

T
Bartolini, L. and Lahiri, A. (2006) Twin deficits, twenty years later. Current Issues in

IP
Economics and Finance. 12 (7) October.
Bouakez, H., Chihi, F. and Normandin, M. (2014) Fiscal policy and external
adjustment: new Evidence. Journal of International Money and Finance, 40, 1-

R
20
Central Agency for Public Mobilization and Statistics (CAPMAS) ARE Statistical

SC
Yearbook, various issues
Chen D. (2007). Effects of monetary policy on the twin deficits. The Quarterly
Review of Economics and Finance, 47, 279-292.

U
Chihi, F. and Normandin, M. (2013) External and budget deficits in some developing
countries. Journal of International Money and Finance, 32, 77-98
N
Chenery, H. and Strout, A. (1966) Foreign assistance and economic growth. The
American Economic Review, 56 (4) 679-733
A
Central Bank of Egypt (CBE) online, Annual Report, various issues. Available at
<http://www.cbe.org.eg/ar/Pages/default.aspx.
M

Coakley, J., Kulasi, F. and Smith, R. (1998) The Feldstein-Horioka puzzle and capital
mobility: A review. International Journal of Finance and Economics, 3, 169-188
Feldstein, M. and Horioka, C. (1980) Domestic saving and international capital flow.
ED

Economic Journal, 90, 314-329

Hashemzadeh, N. and Wilson, L. (2006) The dynamics of the current account and
budget deficits in selected countries of the Middle East and North Africa.
International Research Journal of Finance and Economics, 5, 1-20
PT

International Monetary Fund (IMF) online, International Financial Statistics (IFS).

Available at <http://www.imf.org/en/Data>
Kalou, S. and Paleologou, S. (2012) The twin deficits hypothesis: Revisiting an EMU
E

country. Journal of Policy Modeling, 34, 230-241.

Kim, C. and Kim, D. (2006) Does Korea have twin deficits? Applied Economic
CC

Letters, 13, 675 – 680

Kim, S. and Roubini, N., (2008) Twin deficits or twin divergence: fiscal policy,
current account and the real exchange rate in the US. Journal of International
A

Economics, 74, 362-383.

Marinheiro, C. (2008) Ricardian equivalence, twin deficits, and the Feldstein-Horioka
puzzle in Egypt. Journal of Policy Modeling, 30, 1041-1056
Neaime, S. (2015) Twin deficits and the sustainability of public debt and exchange
rate policies in Lebanon. Research in International Business and Finance, 33,
127-133.
Payne, J. and Kumazawa, R. (2006) Capital mobility and the Feldstein-Horioka
puzzle: re-examination of less developed countries. The Manchester School, 74
(5), 61-616

23
Summers, L. H. (1988) Tax Policy and International Competiveness in J. Frenkel (Ed)
International Aspects of Fiscal Policies, pp. 349-375. Chicago: Chicago
University Press.
Thirlwall, A. (2003) Growth and Development with Special Reference to Developing
Countries .New York: Palgrave Macmillan
Tranchanas, E. and Katrakilidis, C. (2013) The dynamic linkages of fiscal and current
account deficits: new evidence from five highly indebted countries accounting
for regime shifts and asymmetries. Economic Modeling, 32, 502-510
United Nations Conference on Trade and Development (UNCTAD) online. Available
at <http://unctad.org/en/Pages/statistics.aspx>
Xie, Z. and Chen, S. (2014) Untangling the relationship between government deficit

T
and current account deficits in OECD countries: evidence from bootstrap panel

IP
Granger causality. International Review of Economics and Finance, 31, 95-104

R
SC
U
N
A
M
ED
E PT
CC
A

24
Fig. 1: Current account deficit to GDP ratio (CAD) and
budget deficit to GDP ratio (BD) in Egypt, 1980-2014

T
IP
0.1

R
0.05

SC
0

U
N
-0.05
A
-0.1
M

-0.15
ED

-0.2

CAD BD
E PT

Fig. 2: Merchandise trade deficit to GDP ratio (TmD)

CC

and Budget deficit to GDP ratio (BD) in Egypt, 1975-

2014
A

25
 0.05

-0.05

-0.1

T
IP
-0.15

R
-0.2

SC
-0.25

TmD BD

U
N
A
M
ED
E PT
CC
A

26
Table 1
Review of Selected Studies on the Twin Deficit Hypothesis after 2005
Study Country Perio Variables estimated Methodology Main finding
or d used
countries estim
examined ated
Baharum Indonesia, 1976: Budget deficit to GDP Cointegration, The long run relationship between the
shah, Malaysia, 1- ratio, Current account Granger causality budget deficit and the current account
Lau and Thailand 2000: to GDP ratio, nominal tests, Vector deficit is confirmed in some ASEAN
Khalid and 4 exchange rate, short Autoregression
countries. Reverse causality and bi-
(2006) Philippine term interest rate (VAR),
s Variance directional causality also takes place in
Decomposition some countries
Bartolini Two Two Private consumption, Panel regression Although a link exists between fiscal and
, and samples: period GDP, fiscal deficit, with fixed effects current account deficits, yet that link is
Lahiri, 26 s government

T
very weak; thus, reducing the large US
(2006) developed 1972- consumption, public
budget deficit will not have a major
countries; 1998; debt, GDP growth,
effect on correcting the imbalances in the

IP
and 16 1972- population growth,
OECD 2003 current account current account
countries balance.
Hashem Middle Mid Budget deficit and Vector Twin deficit is confirmed only for Oman.

R
zadeh Eastern sevent current account Autoregression Reverse causation is confirmed for Syria
and countries ies till deficits (VAR), Impulse and Yemen. For other oil exporting
Wilson including late Response countries, current account explains a

SC
(2006) Egypt nineti Functions, considerable percentage of the changes
es. Variance in the budget balance. Twin deficit
Decomposition hypothesis is not confirmed for Egypt
Kim and Korea 1970- Budget deficit to GDP Modified Wald A causal relationship from the Korean
Kim 2003 ratio, current account test (MWALD) current account deficit to the budget
(2006)

Chen United 1975:

deficit to GDP ratio,
exchange rate
Budget deficit, trade 3-equation U deficit and not vice versa.

A 1% increase in the bond rate explains

N
(2007) States 01- deficit, real interest dynamic linear 22% of variations in the budget deficit. A
2004: rate, nominal exchange system to show one index point variation of the exchange
A
12 rate, real exchange rate the positive rate correlates with a 78% variation in
relationship the budget deficit. When real exchange
between the 5- rate is incorporated in the system the
M

year treasury rates become 14% and 86% respectively

bond rate and the
twin deficits
Kim and United 1973- Log of real GDP, Vector Twin divergence rather than twin deficit
ED

Roubini States 2004 Government deficit as Autoregression exists as an expansionary fiscal shock; or
(2008) Quart percentage of GDP, (VAR) a government deficit shock improves
erly current account as rather than deteriorates the current
data percentage of GDP, 3 account and depreciates the exchange
month real interest rate. Improvements in the current
PT

rate, log of real account result from increases in private

exchange rate saving and declines in investment.
Marinhei Egypt 1974- External debt, budget Cointegration, Rejects the twin deficit hypothesis for
ro, 2003 deficit, consumption, Vector Error Egypt, in favor of a reverse causality
(2008) public consumption, Correction running from the external deficit to the
E

budget deficit. It also refutes the full

GDP, Current account Model (VECM),
Ricardian equivalence for Egypt in favor
deficit/GDP, trade Granger causality of high capital mobility
CC

deficit/GDP, budget tests

deficit/GDP,
Investment/GDP
Kalou Greece 1960- Budget deficit, short Cointegration, The two deficits are positively linked
and 2007 term interest rate, Vector Error with causality running from the current
A

Paleolog nominal effective Correction account deficit to the budget deficit

ou exchange rate, current Model (VECM)
(2011) account balance
Chihi, Selected Post External variables such Covariance and The covariance between the two deficits
and developin 1960 as world interest rate, Generalized is positive and significant. The
Norman g period real exchange rate and Methods of covariance is stimulated by shocks
din, terms of trade; and related to internal conditions more than
countries . Moments
(2013) internal variables such external conditions.
as domestic resources (GMM)
and fiscal policies.
Trancha Portugal, 1971- Government budget Cointegration Supports the twin deficit hypothesis.
nas and Ireland, 2009 deficit and the current allowing for Decreases in fiscal deficit have stronger

27
Katrakili Italy, account deficits as structural breaks impact on the current account rather than
dis Greece percentages of GDP and asymmetries the opposite
(2013) and Spain
Bouakez US, UK, 1973: Tax and government Structural Vector Weak support for the twin deficit
, , Chihi, Canada 1- spending shocks on the Autoregression hypothesis, and effects of an unpredicted
and and 2008: current account and (SVAR) tax are inconsistent with standard
Norman Australia 4 real exchange rate economic models; while real
din, depreciation is larger than predicted by
(2014) conventional identification measures

Xie and Eleven 1980- Current account to Panel Granger Barro-Ricardian equivalence hypothesis
Chen OECD 2010 GDP ratio, causality tests applies to the UK and France. The
(2014) countries government balance to with bootstrap Keynesian hypothesis applies to Norway
GDP ratio, private critical values and Switzerland. The current account
investment to GDP targeting hypothesis applies to Ireland,
ratio Spain and Sweden. Bi-directional

T
causality applies to Belgium, Finland,
Greece and Iceland.
Ahmad, Nine 1980: Budget deficit to GDP, Threshold Long run positive cointegration

IP
Aworind African 1- Current account deficit cointegration (supporting twin deficts) for six of the
e, and countries 2009: to GDP approach of nine African countries and negative
Martin, including 4 Hansen and Seo cointegration (supporting twin

R
(2015) Egypt divergence) for the other three. Diversity
exists in the speed of adjustment from
the current account compared to the

SC
budget balance.
Neaime Lebanon 1970- Government Cointegration, A long run relationship between the two
(2015) 2013 expenditures, tax Granger causality deficits does not exist in Lebanon, but in
revenues, budget tests the short run recurrent budgetary deficits
deficit, total public continues to widen the current account
debt, current account,
external debt, exports,
imports, ratio of
U deficit
N
external debt to GDP
Source: Compiled by the author
A
M
ED
E PT
CC
A

28
Table 2
Lag selection test for VAR model
Lag HQ SC AIC FPE LR LogL
0 -0.610340 -0.485625 -0.670655 6.01e-06 NA 14.39516
1 -6.635254 -6.011678* -6.936831 1.15e-08 189.7593 127.5209
2 -6.592653 -5.470216 -7.135491 1.00e-08 27.08020* 146.6001
3 -6.746998 -5.125700 -7.531098 7.85e-09 25.70156 168.7320
4 -7.345905* -5.225746 -8.371266* 4.67e-09* 26.21397 197.7546
Source: Author's calculations
*Indicates lag order selected by criterion
LR: sequential modified LR test statistic (each test at 5% level)
FPE: Final prediction error
AIC: Akaike information criterion

T
SC: Schwartz information criterion
HQ: Hannan-Quinn information criterion

R IP
SC
U
N
A
M
ED
E PT
CC
A

29
Table 3
Vector Autoregression (VAR) model
Variables Budget Deficit Interest rate Exchange rate Current account
(BD) (INT) (ER) deficit (CAD)
BD(-1) 0.236545 4.532096 -0.520089 -0.212478
[1.13012] [0.47990] [-0.47236] [-0.85192]
BD(-2) 0.513706** -0.712963 -0.074149 0.366334
[2.74830] [-0.08454 [-0.07541] [1.64474]
INT(-1) 0.002103 0.178801 -0.008420 0.00914
[0.48242] [0.90906] [-0.36718 [0.17596]
INT(-2) 0.005391 0.220207 0.028562 -0.001976
[1.54893] [1.40231] [1.56010] [0.47643]
ER(-1) 0.031856 -4.595374* 1.312138*** -0.052296

T
[0.63256] [-2.02240] [4.95306] [-0.87145]
D(ER(-2)) -0.031350 6.146802*** -0.391262 0.053003

IP
[-0.65457] [2.84454] [-1.55302] [0.92874]
CAD(-1) 0.420171** 6.426769 -1.501546 0.709570***
[2.28227] [0.77370] [-1.55049] [3.23450]

R
CAD(-2) -0.301253* -1.091616 0.520010 -0.192026
[-1.99144] [-0.155994] [0.65348] [-1.06529]

SC
C -0.108700** 8.536987*** -0.409154 0.016862
[-2.08575] [3.63059] [-1.49247] [0.27153]
R2 0.830160 0.867949 0.981596 0.643225
F statistic 14.66364 5.408659 160.0107 1.148180
Tests of residuals
Jarque-BeraNormality 0.681459 0.011426
U
0.019483 0.000000
N
test (p-value)
Breusch-Godfrey serial 0.3021 0.2197 0.3819 0.1550
correlation LM
A
(p-value of Obs. R2)
Breusch-Pagan-Godfrey 0.4432 0.0861 0.3055 0.4424
M

heterosdedasticity
Test (p-value of Obs.
R2)
Source: Calculated by author
ED

***Indicates significance at the 1% level

**Indicates significance at the 5% level
*Indicates significance at the 10% level
E PT
CC
A

30
Table 4
Pairwise Granger causality tests at 2 lags for VAR variables
Null Hypothesis F-statistic p-ratios
CAD does not Granger cause BD 3.81779** 0.0342
BD does not Granger cause CAD 2.49931 0.1003
ER does not Granger cause BD 1.70137 0.2008
BD does not Granger cause ER 0.78285 0.4669
INT does not Granger cause BD 4.63130** 0.0183
BD does not Granger cause INT 0.28360 0.7552
ER does not Granger cause CAD 1.67622 0.2053
CAD does not Granger cause ER 1.84490 0.1767
INT does not Granger cause CAD 0.20310 0.8174

T
CAD does not Granger cause INT 2.03184 0.1500
INT does not Granger cause ER 0.84935 0.4384

IP
ER does not Granger cause INT 14.3335*** 5.E-05
Source: Calculated by author
***Indicates significance at the 1% level

R
**Indicates significance at the 5% level
*Indicates significance at the 10% level

SC
U
N
A
M
ED
E PT
CC
A

31
Table 5
Augmented Dickey Fuller (ADF) Test
Variable At Levels Test critical values At 1st difference Test critical values
BD -2.182507 -4.219126*** -10.27203 -4.219126***
(Intercept, trend, -3.533083** -3.533083**
max. lag length -3.198312* -3.198312*
=9)

TmD -0.589977 -2.627238*** -9.286173 -2.627238***

(None, max. lag -1.949856** -1.949856**
length =9) -1.611469* -1.611469*
INT -2.306218 -3.610453*** -6.423961 -3.615588***
(Intercept, trend, -2.938987** -2.941145**

T
max. lag length -2.607932* -2.609066*
=9)

IP
ER -2.247439 -4.219126*** -4.161653 -4.219126***
(Intercept, trend, -3.533083** -3.533083**
max. lag length -3.1988312 -3.198312*

R
=9)
Sources and notes: Calculated by the author. The null hypothesis is that the series contain a unit root.

SC
The number of lags was automatically selected based on the Schwartz information criterion given a
maximum number of 9 lags. Graphical representations of the variables determined whether the test was
undergone with a constant, a constant and a time trend or none.
*** Critical values for 1% level
** Critical values for 5% level
* Critical values for 10% level
U
N
Table 6
The Phillips Perron (PP) Test
A
Variable At Levels Test critical At 1st difference Test critical values
values
BD -1.888312 -4.211868*** -10.27203 -4.219126***
M

(Intercept, trend) -3.529758** -3.533083**

-3.196411* -3.198312*
TmD -0.918472 -2.625606*** -12.25580 -2.627238***
ED

(None) -1.949609** -1.949856**

-1.611469* -1.611469*
INT -2.310646 -3.610453*** -6.425117 -3.615588***
(Intercept, trend) -2.938987** -2.941145**
-2.607932* -2.609066*
PT

ER -1.633707 -4.211868*** -4.052238 -4.219126***

(Intercept, trend) -3.529758** -3.533083**
-3.196411* -3.198312*
Sources and notes: Calculated by the author. The null hypothesis is that the series contain a unit root.
E

The number of bandwidths was automatically selected by the Newey-West automatic selection of
bandwidths using Bartlett Kernel.
CC

*** Critical values for 1% level

** Critical values for 5% level
* Critical values for 10% level
A

32
Table 7
Lag selection tests for VECM
Lag HQ SC AIC FPE LR LogL
0 0.061295 0.174051 -0.000103 1.18e-05 NA 4.001898
1 -4.480319 -3.916539* -4.787305 9.86e-08 180.8662 108.5651
2 -4.521105* -3.506300 -5.073680* 7.65e-08* 32.23471* 129.8631
3 -4.134413 -2.668584 -4.932577 9.60e-08 17.37028 143.2527
Source: Author's calculations
*Indicates lag order selected by criterion
LR: sequential modified LR test statistic (each test at 5% level)
FPE: Final prediction error
AIC: Akaike information criterion
SC: Schwartz information criterion

T
HQ: Hannan-Quinn information criterion

IP
Table 8

R
Cointegration analysis
Rank Trace statistic Critical value Max-eigen Critical value

SC
statistic
r=0 52.4668* 47.85613 36.78836* 27.58434
r>1 15.67804 29.79707 7.998022 21.13162
Source: Author’s calculations
Denotes significance at the 5% significance level
U
N
A
M
ED
E PT
CC
A

33
Table 9
Vector error correction model with budget deficit as the dependent variable
Variables Budget Deficit Interest rate Exchange rate Trade deficit

(ΔBD) (ΔINT) (ΔER) (ΔTmD)

ECT (-1) -0.032663** -2.402431*** 0.118383* -0.010867

[-2.31742] [-5.55189] [1.77877] [-0.74396]
D(BD(-1)) -0.684949*** -2.802584 -0.951970 -0.307132
[-3.88942] [-0.51835] [-1.14479] [-1.68281]
D(BD(-2)) -0.326787* -1.472182 -0.816767 0.108153
[-2.00793] [-0.29463] [-1.06282] [0.64122]
D(INT(-1)) -0.001545 -0.545501*** 0.041509* -1.65E-05

T
[-0.31370] [-3.60868] [1.78542] [-0.00323]
D(INT(-2)) -0.000808 -0.353640** 0.017206 -0.002260

IP
[-0.16965] [-2.41718] [0.76468] [-0.45760]
D(ER(-1)) -0.065775 -4.013491*** 0.496764** -0.073501*
[-1.62555] [-3.23070] [2.59996] [-1.75272]

R
D(ER(-2)) -0.023938 -5.306051*** 0.293301 -0.004990
[-0.44216] [-3.19230] [1.14732] [-0.08893]

SC
D(TmD(-1)) 0.363178* -4.704046 1.441447 -0.406362*
[1.79552] [-0.75749] [1.50920] [-1.93851]
D(TmD(-2)) 0.327800* 0.723093 1.493674 -0.173174
[1.70469] [0.12248] [1.64502] [-0.86897]
C

R2
-0.005809
[-0.84839]
0.517968
-0.541142*
[-2.57429]
0.649293 U -0.020714
[-0.64071]
0.370390
-0.005323
[-0.75015]
0.469058
N
F statistic 3.223649 5.554153 1.764857 2.650339
Tests of residuals
A
Jarque-BeraNormality test 0.252743 0.00018 0.010326 0.685233
(p-value)
M

Breusch-Godfrey serial 0.3308 0.8758 0.8021 0.2085

correlation LM
(p-value of Obs. R2)
Breusch-Pagan-Godfrey 0.1627 0.8109 0.2379 0.9652
ED

heterosdedasticity
Test (p-value of Obs. R2)
Source: Calculated by author
***Indicates significance at the 1% level
PT

**Indicates significance at the 5% level

*Indicates significance at the 10% level
E
CC
A

34
Table 10
Wald test for significance of short run effects (P-values of Chi-square values χ2)
Dependent variables
Independent (ΔBD) (ΔINT) (ΔER) (ΔTmD)
variables
(ΔBD) 0.0004*** 0.8679 0.4042 0.1028
(ΔINT) 0.9486 0.0005*** 0.1959 0.8929
(ΔER) 0.1825 0.0000*** 0.0062*** 0.1861
(ΔTmD) 0.1240 0.6481 0.1825 0.1526
ECT effect (t-statistic)
-2.31742** -5.55189*** 1.77877* -0.74396
Source: Calculated by author
***Indicates significance at 1% level

T
**Indicates significance at 5% level
*Indicates significance at 10% level

R IP
SC
U
N
A
M
ED
E PT
CC
A

35
Table 11
Granger causality tests for VECM variables
No. of lags D(BD) does not Granger p-ratios D(TmD) does not Granger p-ratios
cause D(TmD) (F- cause D(BD) (F-statistic)
statistic)
1 9.38176 ***0.0042 8.03001 ***0.0076
2 3.16259 *0.0558 2.98515 *0.0648
3 1.76055 0.1767 3.51555 **0.0274
4 1.60214 0.2036 2.44449 *0.0718
5 1.10868 0.3830 1.11645 0.3792
No. of lags D(ER) does not Granger p-ratios D(INT) does not Granger p-ratios
cause D(INT) (F-statistic) cause D(ER) (F-statistic)
1 5.97005 **0.0197 0.33491 0.5665

T
2 3.13492 *0.0571 0.20336 0.8170
3 5.49668 ***0.0041 0.52093 0.6713

IP
4 5.10766 ***0.0036 0.51488 0.7254
5 3.27399 **0.0223 0.62387 0.6831
Source: Calculated by author

R
***Indicates significance at 1% level
**Indicates significance at 5% level

SC
*Indicates significance at 10% level

U
N
A
M
ED
E PT
CC
A

36