Journal of Sustainable Finance & Investment

ISSN: 2043-0795 (Print) 2043-0809 (Online) Journal homepage: https://www.tandfonline.com/loi/tsfi20

Does informal economy impede economic growth?

Evidence from an emerging economy

Nguyen Vinh Khuong, Malik Shahzad Shabbir, Muhammad Safdar Sial & Thai
Hong Thuy Khanh

To cite this article: Nguyen Vinh Khuong, Malik Shahzad Shabbir, Muhammad Safdar
Sial & Thai Hong Thuy Khanh (2020): Does informal economy impede economic growth?
Evidence from an emerging economy, Journal of Sustainable Finance & Investment, DOI:
10.1080/20430795.2020.1711501

To link to this article: https://doi.org/10.1080/20430795.2020.1711501

Published online: 09 Jan 2020.

Submit your article to this journal

Article views: 2011

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tsfi20
JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT
https://doi.org/10.1080/20430795.2020.1711501

Does informal economy impede economic growth? Evidence

from an emerging economy
Nguyen Vinh Khuonga,b, Malik Shahzad Shabbirc, Muhammad Safdar Siald and Thai
Hong Thuy Khanhe
a
Faculty of Accounting and Auditing, University of Economics and Law, Ho Chi Minh City, Vietnam; bVietnam
National University, Ho Chi Minh City, Vietnam; cLahore Business School, University of Lahore, Pakistan;
d
Department of Management Sciences, COMSATS University Islamabad, Pakistan; eFaculty of Finance and
Accounting, Nguyen Tat Thanh University, Ho Chi Minh City, Vietnam

ABSTRACT ARTICLE HISTORY

The objective of this study is to re-examine the impact of the Received 1 September 2019
informal economy on economic growth in Pakistan. This study Accepted 1 January 2020
first computed the informal economy through currency demand
KEYWORDS
equation and then the adopted auto-regressor distributed lags Informal economy; economic
(ARDL) technique for data analysis. The result indicates that 56% growth; ARDL; Pakistan
informal economy of gross domestic product (GDP) exists in
Pakistan. The Wald F-test shows that the overall model is
statistically significant because the value of this test (13.4) is more
than the upper and lower bounds values. Whereas Engle-Granger
causality test describes that the growth rate of real GDP causes
the Granger to GDP at 5%. This study tries to solve these issues
and give a new policy implication for policymakers to control the
informal economy and make sure that this sector will convert into
a recorded or reported form.

1. Introduction
The word informal economy is described through various ways in the existing literature,
such as black, shadow, hidden, irregular, unofficial, unobserved, underground and grey
economy (Gylys 2005). First of all, there is no difference between the informal economy
and the underground economy, because several countries called informal or underground
economy or some other names. However, the approach and objective of both (informal
and underground) are the same. The informal economy is renowned with diverse appella-
tions around the world. For instance, the Japanese recognize it as ‘Hidden Incomes’, the
French identify it as ‘Travail au noir’, Italy considers it as ‘The Lavorno Nero’, the
English call it ‘Fiddle’, the Swedish Russian term it as ‘Hidden Economy’, while in Pakistan
it has been analyzed as an ‘Hidden Economy’ or ‘Informal Economy’. Such kind of econ-
omic activities, which are not reported in National Income Accounts, come under the
definition of the informal economy. It is also called the subordinate zone of the overall
economy, which can play a positive role for the growth of the overall formal sector of
any economy (Gulzar, Junaid, and Haider 2010).

CONTACT Nguyen Vinh Khuong [email protected] Faculty of Accounting and Auditing, University of Econ-
omics and Law, VNU-HCM, Ho Chi Minh City, Vietnam
© 2020 Informa UK Limited, trading as Taylor & Francis Group
2 N. V. KHUONG ET AL.

The informal economy encompasses those activities that necessitate a cost but are
excluded from the benefits and rights of the formal deeds. They may transpire as informal
housing, informal transport, informal industrial sector, etc. However, these activities are
illegal but are not anti-social activities like the drug trade, smuggling, gambling and
pork (Ahmed and Ahmed 1995). The informal economy is expanding its growth with a
rapid pace in Pakistan from the last couple of decades. However, the position of the
formal sector is on the retreat stage. There are many factors, which play to the continuous
growth of the informal economy in Pakistan from the last several years. These factors
include a constant share of the formal manufacture sector in total employment and
output, a continuous increase of informal jobs in total employment level, the formal
economy became the victim of growth share of cash transactions in the overall
economy and an important one is a boost in tax gap estimation. Kemal (2007) concluded
the underground economy with these remarks that the formal and informal economies
have a positive relationship between them because when the formal economy upsurges,
then automatically proliferates the underground economy. For instance, when an organ-
ization appoints some contract and permanent employees, then it is liable to pay taxes for
hiring permanent employees. However, the organization has no panic of taxes for the con-
tract-based employees. This is one the reasons that most small-level businesses preferred
to hire contract-based employees to evade taxes.
To some extent, a collective role of some partial elements also became the cause for
declining the formal economy, for instance, firm-level behaviour has provided such
genius clues to identify the trends in the economy, tax-registered firms in terms of sale
and income tax has declined in the last couple of years according to the tax experts.
The anecdotal proof suggests that the underground economy production has rapidly
increased due to large-scale manufacturing firms that have either partially or completely
shifted their production units. But, the federal board of revenue (FBR) has caught many
significant pieces of evidence in these phenomena. Kemal and Qasim (2012) reported
that the size of the informal economy in 2008 lies between 74% and 91% of the reported
formal economy. It is noted that more than 73.3% of the informal sector people have got
their jobs outside of the agriculture sector in rural areas than urban areas. Most of them are
connected with the construction and transport sector, personal and social or community
services, the trade manufacturing sector and finally wholesale and retail businesses accord-
ing to the labour force survey 2008–2009.
However, several researchers conducted their studies to examine the size and role of
underground economies with the limited time period and adopted trivial methods.
However, these adopted methods have serious issues, which were emphasized by Tanzi
(1983) and some of the researchers. This study addresses following the area of the informal
economy, such as expansion of the informal economy, previously used trivial methods or
approaches to identify the informal economy, time period or data for a short time period
used in the past studies. This study also uses latest data set from 1973 to 2017, as we did
not find any comprehensive study on this topic in the perspective of Pakistan’s informal
economy; we expect that some reports have been published by the Pakistan Institute of
Development Economics (PIDE) and the State Bank of Pakistan (SBP) to highlight this
issue. Moreover, this study also identifies to what extent, the informal economy impacts
the formal economy of Pakistan. Furthermore, how does this informal economy
influence the policy design for policymakers?
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 3

2. Literature review
Some of the working papers from SBP and PIDE tried to throw light on the informal sector
of Pakistan, to identify the size and role of the informal economy in Pakistan. All these
studies were based on Guttman (1977, 1980) approaches. It is noted that researchers
such as Ahmed and Ahmed (1995), Shabsigh (1995); Aslam (1998); Iqbal and Qureshi
(1998); Khalid (2002); Kemal (2003, 2007); Arby, Malik, and Hanif (2010); adopted the
Tanzi monetary approach partially or completely, and they conducted their studies on
the underground economy in the context of Pakistan. It is impressed one for forthcoming
researchers and readers that when Tanzi concluded with these remarks that monetary
approach has given only the order of magnitude and wider symptom of trends instead
of a precise measure of the informal economy due to the sensitivity of assumptions. Fur-
thermore, this monetary approach has failed to provide the real size or exact numbers
regarding the informal economy, even some of the researchers, such as Kemal (2007),
also highlighted the problems of monetary approach in detail in both studies.
Now we are discussing those particular studies, which were trying to identify the under-
ground economy, its size and role in the perspective of Pakistan. Most of them adopted the
monetary approach completely or partially with some other methods, such as electrical
approach, multiple indicators and multiple causes (MIMIC) model, income and expendi-
ture approach and labour market approach.
Iqbal and Qureshi (1998) concluded that the underground economy continuously
increased and reached 20.2% in 1996. Khalid (2002) concluded that the value of the
underground economy (UE) as a percent of GDP suddenly enlarged and reached a
maximum level in 1998 and then rapidly declined. Yasmin and Rauf (2004) used the
data set from 1974 to 2002 to estimate the underground economy through the monetary
method. The result of their study revealed that in 1974 the value of the underground
economy lies in 12 billion Pakistani rupees, whereas in 2002 it increased and reached
1085 billion rupees. Kemal (2003) determined the underground economy increased in
1974 with 20% and rapidly upsurged this increase till 1998 with 54%, but declined in
2002 with 37%. Kemal (2007) estimated the underground economy for Pakistan and con-
cluded the result that UE and tax evasion increased in the 1980s till the 1999s, while it
became slow till 2003. However, tax evasion lies between 5.7% and 6.5% annually and UE
reached 54.6% in 2005. Arby, Malik, and Hanif (2010) used a modified monetary
approach to measure the underground economy from 1966 to 2008. The result of
their study indicates that UE lies 30% in the 1960s and then continuously increased
till 2008.
Aslam (1998) estimated the underground economy and defined those activities which
were directly or indirectly involved in generating income for the informal economy. This
study also recommended direct taxes on such hidden activities. However, this study has
taken large data from 1960 to 1998 to overcome this tax evasion problem. This study
also used Tanzi methodology with some minor changes to get relief from the dollarization
of the economy to control the model of currency circulation. This study concluded that
underground activities have significant negative impact on the formal economy. It
further described that underground elements due to the high cost of ambiguity and ineffi-
cacies and tax evasion closed on 4.7% of the GDP. It is noted that the foreign exchange
4 N. V. KHUONG ET AL.

(dollarization) in the underground economy has significant impact, creating a problem for
the formal economy.
Gulzar, Junaid, and Haider (2010) ascertained the size, causes, issues and implications
of the hidden economy in Pakistan. They collected the annual data frequencies from 1973
to 2010. In their study, they used multiple approaches such as, simple monetary approach
as Tanzi (1980), modified monetary approach using Dynamic Ordinary Least Square
(DOLS) technique of Co-integration, electricity consumption approach and labour
market approach, using statistical accounting, structural estimation approach using
multi-indicators multi-causes (MIMIC), to reach the objective of their study. The
growth of the informal economy has a significant impact on small-scale manufacturing
industry and cottage industry according to the electrical consumption approach. The
labour market approach determined that the informal economy has a substantial role in
poverty alleviation. However, the GDP of Pakistan economy gave the red signal and
also highlighted the corruption of government and hidden expansion of the informal
sector in the MIMIC approach.
Haque (2013) examined the estimated sized of the black economy in Bangladesh by
developing the regression model with time-series variables for 1973–2008. The author
also used Fully Modified Phillips-Hansen Estimate (FMOLS) to identify the long-run
co-integration relationship. The main assumptions of Tanzi’s approach is that mostly
hidden transaction activities are carried out in the form of cash payments and a rise in
the size of the black economy will also increase the demand for more currency. He esti-
mates the black economy under the following three assumptions: (a) velocity of illegal
money (IM) equals to the velocity of legal money, (b) velocity of IM equals to the velocity
of narrow money (M1) and (c) velocity of IM equals to the average velocity of legal money
and narrow money. This study used the data for multiple variables from various sources
that include annual reports of Bangladesh Bank and monthly economic trends on Bangla-
desh Economy by the Central Bank of Bangladesh, Bangladesh Bureau of Statistics, annual
reports of National Board of Revenue. The findings show that in 1973, the black economy
contributed to only 7% of nominal GDP. However, it increased to 62.75% of GDP in 2010.
It creates an enormous burden on the economy, which, in turn, results in heavy tax dis-
tortion and flawed measurement of macroeconomic variables. The study’s results regard-
ing the long-run co-integrating relationship (FMOLS method also support the Tanzi’s
original model.
Blackburn, Bose, and Capasso (2012) investigated to what extent the financial develop-
ment of a country is related to agents’ decision to indulge in the informal economy due to
their undeclared full incomes to avoid government taxes along with their business in the
formal sector. They used a model of tax evasion and financial intermediation to analyze
the relationship of informal market activity with credit market development. Their
study employed the analysis on three types of financial development regimes by highlight-
ing that an increase in economic growth from lower, to an intermediate, and then high
development growth would reduce the percentage of tax-evading agents. The results
showed a negative relationship between the practices of tax evasion and a financial devel-
opment level. A government may afford a low level of tax evasion if there are higher costs
to mitigate it and the country’s financial development and income are endogenously
growing. However, if a country in the initial process of financial development with the
stagger economy would experience difficulty to create tax revenues equally important
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 5

for development in other sectors. The result highlights the financial development is helpful
to reduce the incidence of tax evasion only above a threshold level. If it is below a threshold
level, the informal economy will exist and cause serious financial repression. Shabbir,
Saleem, and Khan (2019) suggest that an imperfect financial market provides more
chances for the exchange of asymmetric information between lenders and borrowers
that affect the terms and conditions regarding the disclosure of wealth by individuals
and loan contract.

3. Description and sources of data

This study comprises different policy variables and data have been taken from different
sources for these variables. However, these variables include gross domestic product
(GDP), gross net product (GNP), inflation rate, growth rate of real GDP, foreign currency
accounts (FCA), demand deposits (DD), currency in circulation (CC), money supply (M1
and M2), banking services (BS) and total tax revenues. The data have been collected from
1973 to 2017, from various latest published annual reports of SBP, handbook of statistics
on Pakistan economy (2015) published by Pakistan Bureau of Statistics (PBS),1 various
economic surveys reports and data for total tax revenues are collected from several pub-
lished annual reports by federal board of revenues (FBR).2 Moreover, BS variable is com-
puted through the total amount of bank deposits over the total number of bank accounts
from the website of SBP.3
The existing literature indicates that several researchers, such as Ahmed and Ahmed
(1995); Shabsigh (1995); Aslam (1998); Iqbal and Qureshi (1998); Khalid (2002); Kemal
(2003, 2007); Kemal and Qasim (2012); Arby, Malik, and Hanif (2010) used numerous
trivial methods to measure the trends and percentage of the informal economy as com-
pared to the formal economy. But, the latest study (Khan and Khalil 2017) on Pakistan
underground economy used only data till 2010. The currency demand equation is used
to measure the informal or underground economy for Pakistan.

  
CC + DD + FCA T
= b0 + b1 + b2 BS + b3 GDP G + b4 INF
M2 Y

  
CC + DD + FCA
+ b5 t − l ... (1)
M2
whereas t shows the time period for subscript; CC currency in circulation; DD demand
deposits; FCA foreign currency accounts; Y real GDP; T total taxes; M2 money supply;
BS banking services; INF inflation; GDP_G growth rate of real GDP; ε error term.
This study gets the currency ratio values after estimating the regression equations with
taxes and without tax variables for each predicted year. The difference in both estimated
terms indicates us currency holding with reference to the persuasion of tax. Furthermore,
after getting the difference in both values, then multiply with money supply (M2) to get
IM. Moreover, legal money can be calculated after taking the difference between money
supply (M1) and IM. But, velocity (V ) plays a novelty role in finding out the informal
or underground economy (UE). Moreover, national income (GNP) dividing by legal
money portrayed the values of velocity, which was further used to get the estimated
value of the underground economy through legal money multiplied by velocity. Finally,
6 N. V. KHUONG ET AL.

total taxes divided by national income are then multiplied by the estimated underground
economy revealed the tax evasion figures.

    
CC + DD + FCA CC + DD + FCA
Illegal Money = t− wt ∗M2 . . ... (2)
M2 M2

Legal Money = M1–IM (3)

GNP
Velocity = (4)
LM
Underground Economy = IM∗V (5)
Total Taxes
Tax Evasion = ∗UE (6)
GNP
Appendix 1 gives complete detail of all above-mentioned equations with facts and figures.
But, Appendix 2 describes a relatively history for Pakistan underground economy from
this study as compared to other related studies.

4. Methodology
4.1. Empirical model and data description
The main objective of this study is to measure the informal or underground economy
using evasion of tax in Pakistan. This study uses annual observations of time-series data
over from 1973 to 2017. The significant regression model of the underground economy
in Pakistan is structured as follows:
LnGDPt = g0 + g1 INF + g2 ln UE + g3 lnTXt + g4 lnGGDt t + g5 lnEMt + et
(7)
where GDPt denotes real gross domestic product; INF inflation rate;
UET underground economy; TX tax evasion; GGD GDP growth rate; EM employment
rate; Ln natural logarithm; et error term.
The underground economy is an independent variable and with these policy variables
such as tax, employment, GDP growth, inflation and GDP as a proxy of economic growth
and treated as a dependent variable. The data have been taken from various issues of the
economic survey of Pakistan, handbook of statistics of Pakistan economy and data for
sales tax on imports, customs duties and total tax revenues are taken from various
issues of the FBR annual reports. Whereas real interest rate is computed by subtracting
the inflation rate from the nominal interest rate. Finally, the growth rate of GDP is esti-
mated by the following formula applied on GDP at constant factor cost:
 
GDPt − GDPt−1
GDP growth rate =
GDPt−1
The direct and indirect methods are used to estimate the size of the underground
economy or tax auditing (Kemal 2003; Schneider and Enste 2000; Pyle 1989). Whereas
samples and survey methods based on voluntary replies are related to the direct
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 7

method. Indirect estimation depends on determining the traces, while the black economy
leaves in evoke. Moreover, these indirect approaches include different methods, such as
monetary approach, labour market approach, physical input method and monetary
approach. The monetary approach is the best and easily applied and most common
approach among these to estimate the size of the underground economy and tax evasion.
The indirect approach (monetary approach) is used as Tanzi (1980) and explained that
the estimates of the underground economy from this approach should not be taken as
precise measures due to the sensitivity of the assumptions. These methods were used as
broad indications of trends and orders of magnitude. Kemal (2003) adapted the Tanzi
model specification for the macroeconomic condition in Pakistan and verified the robust-
ness of the estimates. But later both Tanzi (1983) and Kemal (2007) studies identified
several issues in monetary approach and did not recommend this method as the best
approach to measure the underground economy.
This study contributes by applying a comprehensive approach to find out the impor-
tance of the underground economy in economic growth with the help of a theoretical
and empirical analysis. The analysis of Monte Carlo exhibits that the ARDL method is
superior and produces consistent results even for small samples (Pesaran and Shin
1999; Saleem et al. 2019). The ARDL method reduces the problem of endogeneity
because it is free of residual correlation and it takes suitable lags which are accustomed
to serial correlation and issues of endogeneity.
The ARDL technique is a modern technique. This method of co-integration is a step-
wise procedure. The ARDL framework can be written as follows:

D LnGDPt = a1 + aINF LnINFt−1 + aUNE LnUNEt−1 + aTX LnTXt−1 + aGGD LnGGDt−1 +


V 
m 
n
aEM LnEMt−1 + ac D LnINFt−C + ak D LnUNEt−k + al D LnTXt−l +
U=0 k=0 l=1


d 
f
am D LnGGDt−m + am D LnEMt−m + D1i (8)
c=0 e=0

H0 :d1 = d2 = d3 = d4 = d5 = d6 = 0 (there is no co-integration).

H1 :d1 = d2 = d3 = d4 = d5 = d6 = 0 (there is co-integration).

The intercept is a drift component and D1 is a residual (error term) and supposed to be
white noise. This study applies the Akaike Information Criterion (AIC) to detect the
absence of serial correlation and will take the optimal lag length for the selected model.
When the null hypothesis is rejected, then the alternative hypothesis will show the exist-
ence of long-run relationship among variables.

4.2. Unit root analysis

The Augmented Dickey-Fuller (ADF) test is employed to check the stationary or non-
stationary state of the data; if variables are not stationary, the OLS results will be spurious.
The statistical results of this test are given in Table 1.
8 N. V. KHUONG ET AL.

The ADF is used to intercept as well as intercept and trend (simultaneously). The
results of unit roots have confirmed that some incorporated variables are stationary (at
all levels) and some of them at last become stationary at I (1).
The Wald F-statistics values gathered from Narayan (2005) are presented in Table 2.
The estimated Wald F-test value is 5.06 at lag 2 (which is greater than the upper and
lower bond critical values at all significance levels). Thus, the results indicate that all
selected variables are co-integrated.

4.3. ARDL estimation

Table 3 indicates that all three ARDL equations are examined by taking each variable as a
dependent variable. Moreover, Table 3 also shows each equation and their respective lag
length of dependent and all independent variables. The results propose that the null
hypothesis (H0 ) of no longer association between the variables is rejected at 5% and
10%. The estimated value of F-statistics is 13.4, while the upper bound value is 5.06,
4.01 and 3.52 at 1%, 5% and 10% level of significance, respectively (designed by
Pesaran, Shin, and Smith 2001 and Narayan 2005).
This section is related to the model of ARDL which identifies the short-run and long-
run association among dependent and exogenous variables. Thus there is the possibility of
various lags of the variables that can be applied (Duasa 2007). After concluding the exist-
ence of co-integration (long-run) among these variables, equation (8) has been calculated
applying the ARDL co-integration approach to find the long-run estimates. These results
are shown in Table 4.
In Table 5, a normal test relates to the Jarque–Bera statistic of the test for normality of
data, x2 the Breusch–Godfrey LM test is related to serial correlation, x2 ARCH test is
Engle’s test for ‘autoregressive conditional heteroskedasticity’, and x2 Ramsey reset test
is used to test for model specification test, whereas (*) represents the 10% level of signifi-
cance. Durbin–Watson (DW) test that is used to find the problem of autocorrelation

Table 1. Unit root testing.

Augmented Dickey–Fuller (ADF) Test
At level First Difference
Variables Constant Constant plus trend Constant Constant plus trend Conclusion
In GDPt 2.282 −3.271 −2.622** −4.142** I(1)
In INFt 3.6491** −1.973 −3.570** −3.519 I(1)
In UNEt −1.5532 −3.507*** −3.930* −5.079* I(1)
InTXt −4.082** −5.096 −3.6186 −3.5590* I(1)
InGGDt −4.14*** −3.39*** −3.773** −3.671** I(1)
Note: *, ** and *** symbolizes 10%, 5% and 1% significance level, respectively.

Table 2. ARDL Bounds testing to co-integration analysis.

Level of Significance Critical Values Wald Test (F-Value)
Lower Limit Upper Limit
1% 3.74 5.06
5% 2.86 4.01 13.4
10% 2.45 3.52
Note: Critical values (lower and upper).
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 9

Table 3. Results of error-correction (ECM) model.

Results of Error Correction (ECM) Model
Dependent Variable:
LGDP(−1)
Selected Model: ARDL(2, 0, 2, 2, 0, 0)
Variable Coefficient T-Statistic
D(LGDP(–2)) 0.014** 2.020
D(LINF) –0.000 −1.339
D(LUE) 0.007 0.737
D(LUE(–1)) –0.310* −1.807
D(LEM) –0.004 −1.033
D(LEM(−1)) 0.085* 1.744
D(LTX) 0.002** 2.296
D(GGD) –0.000* −1.357
Note: *, ** and *** symbolizes 10%, 5% and 1% significance level, respectively.

Table 4. Long-run results (GDPt is a dependent variable).

Variables Coefficient T-Statistic
LINF −0.000 −1.3623
LUE 0.994* 4.4627
LEM −0.282 −119.05
LTX 0.002** 2.24281
GGD −0.004 −1.30474
C 0.270* 8.7952
ECM(−1) −0.619*** −7.7031
Diagnostic tests
R2 0.92
Adjusted R2 0.91
F-statistics 1199.397
(x2 )JB normality test 0.38(0.82)
(x2 )Breusch–Godfrey lm test 8.18(0.20)
(x2 )ARCH test 1.03(0.30)
Durbin–Watson (DW) 2.18
(x2 ) Ramsey reset test 0.41(0.76)
Note: *, ** and *** symbolizes 10%, 5% and 1% significance level, respectively.

results shows that the value of DW is 2.18, which means that the model does not suffer
from autocorrelation.

4.4. Stability test

The cumulative sum (CUSUM) and the square of cumulative sum tests are used to check
the stability of the model in the short-run. Brown, Durbin, and Evans (1975) elaborated
that regression coefficients and constancy of regression coefficients belong to CUSUM
and CUSUM square, respectively. The statistical result of this test shows that the model
is stable at (5%) interval bonds. Figures 1 and 2 show the stability of the model through
CUSUM and CUSUM square, respectively.

5. Empirical findings and discussion

Table 2 reveals the computed value of Wald Test (F-Value) statistics is five, while the
upper and lower bound values are less than Wald Test (F-value) statistics value
Narayan (2005). The critical values described by Narayan (2005) are suitable for this
10 N. V. KHUONG ET AL.

Table 5. Results of Granger causality by applying the MWALD test.

Variables Observations F-statistics P (values)
LINF does not Granger Cause LGDP 44 1.28408 0.2637
LGDP does not Granger Cause LINF 0.18994 0.6652
LTX does not Granger Cause LGDP 44 0.03344 0.8558
LGDP does not Granger Cause LTX 1.79897 0.1872
GGD does not Granger Cause LGDP 44 4.84460 0.0334**
LGDP does not Granger Cause GGD 0.26620 0.6087
LEM does not Granger Cause LGDP 44 0.29949 0.5872
LGDP does not Granger Cause LEM 1.82980 0.1836
LUE does not Granger Cause LGDP 44 0.28045 0.5993
LGDP does not Granger Cause LUE 0.85310 0.3611
LTX does not Granger Cause LINF 44 0.37893 0.5416
LINF does not Granger Cause LTX 4.42809 0.0415**
GGD does not Granger Cause LINF 44 0.83156 0.3672
LINF does not Granger Cause GGD 2.65660 0.1008*
LEM does not Granger Cause LINF 44 0.13930 0.7109
LINF does not Granger Cause LEM 0.01802 0.8939
LUE does not Granger Cause LINF 44 0.19077 0.6646
LINF does not Granger Cause LUE 0.78195 0.3817
GGD does not Granger Cause LTX 44 0.47194 0.4960
LTX does not Granger Cause GGD 0.14192 0.7083
LEM does not Granger Cause LTX 44 0.40227 0.5294
LTX does not Granger Cause LEM 1.11919 0.2963
LUE does not Granger Cause LTX 44 0.76630 0.3865
LTX does not Granger Cause LUE 0.06798 0.7956
LEM does not Granger Cause GGD 44 0.22405 0.6385
GGD does not Granger Cause LEM 0.05030 0.8237
LUE does not Granger Cause GGD 44 0.26531 0.0609*
GGD does not Granger Cause LUE 3.54315 0.0669*
LUE does not Granger Cause LEM 44 0.26531 0.6093
LUE does not Granger Cause LEM 1.82071 0.1846
LEM does not Granger Cause LUE 44 0.74241 0.3939

study with a small sample. The null hypothesis (H0 ) (no co-integration between the
selected variables) is rejected when all variables are treated as response variables. There-
fore, results indicate that there is a long-run relationship between dependent and indepen-
dent variables.
The ECM results of coefficients are shown in Table 3. The LGDP with lag 2 is significant
at 1%. The LUE (–1) is significant at 10%. Similarly LEM (–1) is significant at 10%. LTX
and GGD aresignificant at 10%. This study identifies that the one period lagged error-cor-
rection terms (ECM (–1)) are statistically highly significant at the 1% level or better when
GDP, LINF, LUE, LEM, LTX and LGGD are the variables for the equation of ECM. LUE
and LTX are significant at 10% and 5%, respectively.
Moreover, Table 4 shows that the coefficient of the ECM (–1) term relates to the speed
of adjustment (all given variables) to move toward their long-run equilibrium position
when a short-run shock occurs. The results of normalized long-run coefficients are
shown in Table 4. All the variables are significant and co-integration (long run) is
found among these variables. The values of R-squared (92) and adjusted R-squared (91)
statistics define that the ARDL technique (Table 4) for bounds test is well fitted for the
selected data.
The statistical results of the diagnostic test indicate that the values of ARDL technique
are free from severe econometric issues. The H0 (null hypothesis) showed that the model
of this study suffers neither in heteroskedasticity (applied ARCH test) nor in serial
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 11

Figure 1. CUSUM.

Figure 2. CUSUM Square.

correlation (Breusch–Godfrey LM test is used). The results found that the model has no
severe econometrics issues and this study accepted the null hypothesis (which means
long-run relationship exists). The result of H0 of Ramsey reset test shows that model is
correctly specified, furthermore, we also accept H0 in the case of Jarque Bera (JB test)
which shows that data are normally distributed where the null hypothesis is stated that
the values are greater than 5% level of significance.

5.1. Granger causality analysis

In this section, techniques of causal analysis as Granger test causality analysis and variance
decomposition (VDM) technique to investigate the robustness of the causal relationship
12 N. V. KHUONG ET AL.

between the dependent and independent variables in Table 5. A relationship describes the
dependency or connectivity between configuration items. A relationship can be either uni-
directional or bidirectional. The results of a search are affected when relationships are
bidirectional or unidirectional. In this section, the technique of causal analysis, Engle-
Granger causality analysis and VDM technique investigate the robustness of the causal
relationship between GDP and other independent variables. Nolte et al. (2010) estimate
the causal direction of interactions robustly with respect to instantaneous mixtures of
independent sources with arbitrary spectral content. The findings of GGDP Granger
cause to GDP at 5% level of significance. The LINF granger causes LTX and GGD at
5% and 10%, respectively. The bidirectional causality is running from LUE to GGD at
10% level of significance. The bidirectional relationship describes the two side association
between both variables. It means if a change occurs in either increase or decrease, then it
equally affects other associated variables with the same percentage of value. Whereas the
rest of all variables have a unidirectional relationship among the variables due to low F-
statistics value and high P-value. It further indicates that there is only one side relationship
between these variables. The unidirectional relationship elaborates that all the variables in
Table 5 do not have an effect on their associate variables either any change in increase or
decrease in percentage, value or number.

5.2. VDM analysis

The VAR analysis includes broad forecast error VDM; this system examined the causal
correlation among GDP and other independent variables. The method of VDM
depends on the magnitude of the expected error variance which deals with a series
(from each of the exogenous variable over several time periods) for innovations. The
VDM indicates the amount of information each variable contributes to the other variables
in the autoregression. It determines how much of the forecast error variance of each of the
variables can be explained by exogenous shocks to the other variables. The method of
impulse response functions and VDM is the part of the approach of innovation accounting
approach (Wolde-Rufael 2009). Table 6 presents the results of the VDM method; the
direction of causality indicates the statistical results. Table 6 shows different blocks with
the statistical values of GDP and other factors and their effects on GDP. The results of
Table 6 in the first block (VDM of GDP) indicate the shock or impulse to LEM, LINF,
LUE, LTX and GGD in start (at period 2) or short-run contributes 96.7%, 86.5%,
0.09%, 71.60% and 35.01% fluctuation in GDP. However, in the long-run if we tested at
10th periods of time, then we found that the shock to LEM, LINF, LUE, LTX and GGD
contributes 79.0%, 65.9%, 1.57%, 48.5% and 21.8% variation in GDP, respectively.
Based on VDM statistical results, it is found that all other determinants show fluctuation
in each other.

5.3. Impulse response function (with multiple graphs)

Figure 3 indicates that one standard deviation (SD) bands demonstrate the upper and
lower bounds. Furthermore, Figure 3 provided the Impulse response regarding generalized
indicates one SD innovations with ±2 standard error (SE). In signal processing, the
impulse response, or impulse response function (IRF), of a dynamic system is its output
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 13

Table 6. Results of Variance Decompositions Method.

Variance Decompositions
Period LGDP LEM LINF LUE LTX GGD
1 100.0000 99.45061 95.33906 0.187519 76.51052 35.20424
2 93.59154 96.76058 86.55857 0.093600 71.60352 35.01325
3 91.32342 94.15269 81.91397 0.164649 66.79508 29.23270
4 88.13866 91.20935 77.76444 0.154090 63.02393 28.33108
5 84.60677 88.48212 73.42296 0.130330 59.21485 26.92643
6 80.87381 86.17781 69.70051 0.126343 56.00471 25.80054
7 77.30861 84.11398 67.71945 0.217605 53.34299 24.68987
8 73.83454 82.19707 66.67800 0.470842 51.25818 23.56780
9 70.33241 80.48327 66.18977 0.923954 49.68018 22.62613
10 66.77874 79.00890 65.93551 1.572424 48.50818 21.84886

when presented with a brief input signal, called an impulse. More generally, an impulse
response is the reaction of any dynamic system in response to some external change.
The results of figures specify that when one standard deviation shock is given to these vari-
ables, how it has effects on LGDP in the beginning and LGDP response to LEM increasing
constantly and then became flat or constant. The responses of LGDP to LINF and LUE are
decreasing. But the response of LGDP to LTX and GGD shows an increasing trend, but
after certain 8th period, the variable GGD shows a decreasing trend. Finally, the rest of
all variables also explain that one standard deviation shock is given to these variables
than how they react, as shown graphically in Figure 3. To determine an output directly

Figure 3. Impulse Response Function (Multiple graphs). Note: Author own estimation in the E-Views
software when considering 10 periods.
14 N. V. KHUONG ET AL.

in the time domain requires the convolution of the input with the impulse response. When
the transfer function and the Laplace transform of the input are known, this convolution
may be more complicated than the alternative of multiplying two functions in the fre-
quency domain.

6. Conclusion
The purpose of this study is to re-investigate the several effects of the informal economy on
the level of economic growth in Pakistan. It is observed that from the last couple of
decades, Pakistani economy continuously becomes the victim of the informal economy
due to several reasons such as political instability, war against terrorism, unremittingly
decline in exports revenue and law and order situations. But, several plans launched by
the different provincials and federal ministries are unable to overcome this whole situation
completely. Nowadays Pakistan is going through the economic tumult and desperately
increases in the size of the underground economy as compared to the formal sector,
which is disgusting. The PIDE and SBP have published several reports to highlight this
issue, but we did not find any comprehensive study on this topic, which has properly
addressed this issue and given suitable implication for policymakers.
The results indicate that the informal economy has a significant role in declining the
formal sector of Pakistan economy. Moreover, 56% of the informal economy is measured
of total GDP. This study adopted the ARDL approach as all the variables are stationary on
the level and the first difference with the trend and constant. The Wald F-test revealed that
overall model of this study is statistically significant as lower and upper bounds values are
lesser than the F-test value. This study used Engle-Granger causality and VDM tests to
check the causality robustness between dependent variable as GDP and rest of indepen-
dent/policy variables. But, bidirectional causality is measured at 10% level of significance
from LUE to GGD. Furthermore, stability and diagnostic tests show that model is a best fit
and it suffers neither autocorrelation and nor heteroskedasticity.
This study also suggests some policy recommendations for policymakers to reduce the
size of the informal economy from Pakistan. First, the government should implement
taxes on the agriculture sector, as approximately 70% Pakistan population belong to the
agriculture sector. But, Pakistan agriculture sector has not involved in any big direct
tax. Second, the FBR takes serious action against illegal transportation system and puts
them under documented system with a serial number as per provincial and federal
level. This action will give a proper income in the shape of revenue tax through a
proper channel. Finally, provincial and federal governments join hands to eradicate
non-documented housing schemes. These housing schemes are working at a small,
medium and large level such as, rent the house, small businesses inside the house and
big housing schemes, which are neither reported nor documented in complete form.
Moreover, the ministry of land and cooperative society will make sure that all resident,
industrial and agriculture sector lands to escape from landlords and fake owners of
lands. This activity gives a lot of revenue in the sense of taxes.

Availability of data and material

The data are available on request.
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 15

Notes
1. http://www.pbs.gov.pk/
2. https://www.fbr.gov.pk/
3. http://www.sbp.org.pk/

Acknowledgement
The authors are grateful to the anonymous referees and the editorial team of the journal for their
extremely useful suggestions to improve the quality of the article. All authors contribute equally.

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This research is funded by University of Economics and Law, Vietnam National University Ho Chi
Minh City, Vietnam.

References
Ahmed, M., and Q. M. Ahmed. 1995. “Estimation of the Black Economy of Pakistan Through the
Monetary Approach.” The Pakistan Development Review 34 (4): 791–807.
Arby, M. F., M. J. Malik, and M. N. Hanif. 2010. The Size of Informal Economy in Pakistan. (SBP
Working Paper Series, No. 33).
Aslam, S. 1998. “The Underground Economy and Tax Evasion in Pakistan: Annual Estimates
(1960–1998), and the Impact of Dollarisation of the Economy.” The Pakistan Development
Review 37 (4): 621–631.
Blackburn, K., N. Bose, and S. Capasso. 2012. “Tax Evasion, the Underground Economy and
Financial Development.” Journal of Economic Behavior & Organization 83 (2): 243–253.
Brown, R. L., J. Durbin, and J. M. Evans. 1975. “Techniques for Testing the Constancy of Regression
Relationships Over Time.” Journal of the Royal Statistical Society, Series B 37 (2): 149–162.
Duasa, J. 2007. “Determinants of Malaysian Trade Balance: An ARDL Bounding Approach.”
Journal of Economic Cooperation 28 (3): 21–40.
Gulzar, A., N. Junaid, and A. Haider. 2010. “What is Hidden, in the Hidden Economy of Pakistan?
Size, Causes, Issues, and Implications.” The Pakistan Development Review 49 (4): 665–704.
Guttman, Peter M. 1977. “Subterranean Economy.” Financial Analysts Journal 33: 26–27.
Gylys, Povilas. 2005. “Economy, Anti-Economy, Underground Economy: Conceptual and
Terminological Problems.” Ekonomika 72: 1–11.
Haque, S. T. 2013. Underground Economy of Bangladesh: An Econometric Analysis. Research Study
Series No–FDRS 01.
Iqbal, Z., and S. K. Qureshi. 1998. Underground Economy and Tax Evasion in Pakistan: A Fresh
Assessment. Pakistan Institute of Development Economics, Islamabad. (Research Report
No. 158).
Kemal, M. A. 2003. Underground Economy and Tax Evasion Evaluation. Pakistan Institute of
Development Economics. (Research Report No. 184).
Kemal, M. A. 2007. "Fresh Assessment of the Underground Economy and Tax Evasion in Pakistan:
Causes, Consequences, and Linkages with the Formal Economy." PIDE-Working Papers (13).
Kemal, M Ali, and Ahmed Waqar Qasim. 2012. “Precise Estimates of the Unrecorded Economy.”
The Pakistan Development Review 505–516.
16 N. V. KHUONG ET AL.

Khalid, M. 2002. “Estimation of Underground Economy, Analysis of Pakistan.” M.Phil Thesis,

Department of University, Islamabad.
Khan, A., and S. Khalil. 2017. “The Real Size of Underground Economy: A Case of Pakistan.”
Pakistan Journal of Applied Economics 27 (1): 89–100.
Narayan, P. N. 2005. “The Saving and Investment Nexus for China: Evidence from Co Integration
Tests.” Applied Economics 37 (17): 1979–1990.
Nolte, G., A. Ziehe, N. Krämer, F. Popescu, and K. R. Müller. 2010. “Comparison of granger
Causality and Phase Slope Index.” In Causality: Objectives and Assessment, Volume 6, edited
by Isabelle Guyon, 267–276. Berlin: JMLR Workshop and Conference Proceedings.
Pesaran, M. H., and Y. Shin. 1999. “Anautoregressive Distributed lag Modeling Approach to Co
Integration Analysis. Chapter 11.” In Econometrics and Economic Theory in the 20th Century:
the Ragnar Frisch Centennial Symposium, edited by S. Strom, 289–326. Cambridge:
Cambridge University Press.
Pesaran, M. H., Y. Shin, and R. Smith. 2001. “Bounds Testing Approaches to the Analysis of Level
Relationships.” Journal of Applied Econometrics 16 (3): 289–326.
Pyle, D. J. 1989. Tax Evasion and the Black Economy. New York: Springer.
Qasim, A. W. 2011. “Relationship between Inflation and Tax Evasion: A Case Study of Pakistan.”
MPhil Thesis. Department of Economics, Quaid-i-Azam University, Islamabad.
Saleem, H., M. Shahzad, M. B. Khan, and B. A. Khilji. 2019. “Innovation, Total Factor Productivity
and Economic Growth in Pakistan: A Policy Perspective.” Journal of Economic Structures 8 (1):
1–7.
Schneider, F., and D. H. Enste. 2000. “Shadow Economies: Size, Causes, and Consequences.”
Journal of Economic Literature 38 (1): 77–114.
Shabbir, M. S., H. Saleem, and M. B. Khan. 2019. “Impact of Internet Adoption and Mobile Phone
Penetration on Corruption: Evidence from Selected Asia-Pacific Countries.” Global Business
Review 20. doi:10.1177/0972150919832107.
Shabsigh, G. 1995. The Underground Economy: Estimation, Implications - the Case of Pakistan.
International Monetary Paper Series No 95/101.
Tanzi, V. 1980. Underground Economy and Tax Evasion and Implications. Banca Nazionale del
Lavoro quartly review 32-135 427-453.
Tanzi, V. 1983. The Underground Economy in the United States: Annual Estimates, 1930–1980. IMF
Staff Papers 30, 283–305.
Wolde-Rufael, Y. 2009. “Energy Consumption and Economic Growth: The African Experience
Revisited.” Energy Economics 31: 217–224.
Yasmin, В, and H. Rauf. 2004. “Measuring the Underground the Economy of Pakistan.” The Lahore
Journal of Economics 9 (2): 93–103.

Appendix 1. Trend Estimates of Underground Economy and Tax Evasion.

Years Underground Economy (U.E) Tax Evasion (T.E) GDP as U.E GDP as T.E
1973 8751.748 838.15 13.542 1.2
1974 10072.496 1023.95 16.257 1.154
1975 4708.252 458.42 21.254 0.408
1976 14475.725 1432.09 23.456 1.074
1977 52936.986 5341.85 19.354 1.562
1978 36654.138 3731.67 32.546 1.980
1979 60991.331 6522.60 31.874 2.14
1980 32172.280 3825.05 29.354 1.515
1981 40964.962 4733.01 35.214 1.573
1982 23248.804 2564.36 38.452 0.734
1983 27391.339 2937.88 39.541 0.728
1984 31297.440 3428.91 36.538 0.746
1985 84695.145 8695.69 34.846 1.703

(Continued )
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 17

Continued.
Years Underground Economy (U.E) Tax Evasion (T.E) GDP as U.E GDP as T.E
1986 102234.069 10887.86 41.584 1.959
1987 1238618.526 132844.05 43.531 1.254
1988 807510.399 86455.44 47.512 2.321
1989 1130612.24 128092.59 42.651 1.32
1990 1039340.194 121335.49 42.145 2.14
1991 1096213.211 115962.62 45.564 2.91
1992 1347258.634 153861.46 48.651 2.825
1993 1598798.492 181265.67 43.895 2.651
1994 1790598.026 195957.16 47.512 3.214
1995 2325442.072 280267.99 52.421 3.652
1996 3171750.903 402333.99 50.247 3.521
1997 3784756.897 443191.18 50.952 3.518
1998 4470832.839 494772.20 51.521 3.632
1999 3991664.468 422772.21 48.354 3.458
2000 3151595.649 289541.18 47.652 3.541
2001 3721319.173 351299.77 49.351 3.845
2002 4796967.339 433014.40 53.265 3.875
2003 12784273.73 1171323.42 54.265 4.012
2004 12886518.54 1164229.91 56.512 4.21
2005 18878631.43 1680025.68 58.716 4.341
2006 26162994.96 2231131.17 54.351 2.841
2007 23039066.93 2077116.18 53.421 2.951
2008 41702742.27 3875852.16 57.524 2.865
2009 45107277.96 3866555.60 59.562 2.178
2010 51140494.39 4398490.44 53.284 3.214
2011 35467768.3 2893661.25 54.653 3.652
2012 34616800.04 3091365.49 52.985 3.214
2013 64518398.14 5334444.04 56.321 3.526
2014 92691591.24 7886790.98 61.526 3.541
2015 100081457.9 8919346.72 62.541 3.894
2016 118358616.4 11252549.69 60.342 4.13
2017 185475314.4 17319062.79 56.251 4.029
18 N. V. KHUONG ET AL.

Figure A1. Historical Movement of Underground Economy in Pakistan.

Figure A2. Trends of Underground Economy and Tax Evasion as a percentage of GDP.
 JOURNAL OF SUSTAINABLE FINANCE & INVESTMENT 19

Appendix 2. Comparison of the Current Study with Other Studies.

Current Khan and Khalil Kemal Qasim Arby, Malik, and Hanif Gulzar, Junaid, and
Years study (2017) (2007) (2011) (2010) Haider (2010)
1973 13.542 _ 38.0 _ _ _
1974 16.257 56.67 33.1 24.01 27.1 30.7
1975 21.254 74.36 31.6 22.18 25.9 38.3
1976 23.456 58.22 30.9 24.03 28.4 43.4
1977 19.354 65.52 34.9 23.69 27.9 46.3
1978 32.546 51.54 39.2 28.11 29.2 54.8
1979 31.874 68.94 45.6 30.95 31.1 56.5
1980 29.354 78.46 43.0 33.47 33.3 50.1
1981 35.214 53.74 47.8 31.60 33.1 47.8
1982 38.452 78.84 42.0 38.95 31.6 51.5
1983 39.541 53.01 49.3 38.71 32.8 56.9
1984 36.538 71.14 39.3 38.22 32.1 53.0
1985 34.846 55.86 44.7 35.77 29.6 57.1
1986 41.584 51.92 50.5 36.85 35.2 62.2
1987 43.531 67.54 45.5 36.22 35.4 57.7
1988 47.512 51.42 42.7 35.47 32.7 52.5
1989 42.651 60.53 39.2 37.26 32.5 51.4
1990 42.145 36.34 36.1 39.15 30.0 55.5
1991 45.564 32.28 44.4 33.73 26.1 46.7
1992 48.651 37.59 45.5 37.35 27.7 46.5
1993 43.895 72.02 56.6 34.93 30.1 56.7
1994 47.512 49.16 60.6 33.97 33.3 44.1
1995 52.421 42.93 68.7 38.65 34.8 43.4
1996 50.247 56.73 74.9 41.64 36.8 51.0
1997 50.952 34.89 69.0 35.24 36.4 47.6
1998 51.521 79.81 46.1 33.23 36.4 54.1
1999 48.354 45.69 56.5 32.01 35.2 49.7
2000 47.652 50.43 65.7 33.78 26.0 58.4
2001 49.351 42.85 64.3 34.07 26.3 56.6
2002 53.265 49.98 68.2 33.23 27.0 61.0
2003 54.265 61.34 66.6 35.65 29.0 55.3
2004 56.512 44.38 64.8 35.45 24.9 50.8
2005 58.716 45.67 – 35.17 18.7 49.6
2006 54.351 41.82 – 35.56 18.3 50.1
2007 53.421 23.73 74.3 38.03 18.9 51.0
2008 57.524 57.84 – 37.27 19.6 36.1
2009 59.562 68.63 91.0 32.81 – 37.2
2010 53.284 70.79 38.0 31.80 – 47.6
2011 54.653 _ _ _ _ _
2012 52.985 _ _ _ _ _
2013 56.321 _ _ _ _ _
2014 61.526 _ _ _ _ _
2015 62.541 _ _ _ _ _
2016 60.342 _ _ _ _ _
2017 56.251 _ _ _ _ _
20 N. V. KHUONG ET AL.

Figure A3. Underground Economy Comparison of the Current Study with Other Studies.