Economic Modelling 74 (2018) 207–218

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.journals.elsevier.com/economic-modelling

The effect of corruption on labour market outcomes

Arusha Cooray a, b, *, Ratbek Dzhumashev c
a
School of Economics, University of New South Wales, Kensington, NSW, 2052, Australia
b
Centre for Poverty Analysis, Sri Lanka
c
Department of Economics, Monash University, Clayton, Victoria, 3800, Australia

A R T I C L E I N F O A B S T R A C T

JEL codes: We develop a theoretical model to investigate the relation between corruption and labour supply. The theoretical
O11 model shows that corruption affects labour supply in the formal sector by reducing productivity, changing the
Keywords: supply of labour in the shadow economy, altering the tax burden, and distorting the saving-consumption trade-off.
Corruption The predictions of the theoretical model are tested by using panel data methods for 132 countries. Using the
Labour market labour force participation rate (LFPR) and employment to population ratio as proxies for labour supply, the
estimated empirical results show that corruption has a statistically significant robust direct negative effect on the
LFPR and employment to population ratio. Corruption also has an indirect effect on the LFPR and employment to
population ratio through a higher tax burden and increase in size of the shadow economy. Higher wages, an
increase in consumption, and better regulatory quality are found to reduce the negative impact of corruption on
labour supply, however, the overall effect on labour supply is negative, suggesting that the negative effects of
corruption outweigh the positive effects of improved regulatory quality, wages, and higher consumption. Our
findings imply that in order to reduce the negative effect of corruption on labour supply, governments need to
develop a comprehensive approach to not only combatting corruption itself but also working on improving
regulation and promoting policies that decrease activities in the shadow economy.

1. Introduction LFPR and employment to population ratio through different channels. It

Studies show that corruption can be damaging for an economy.1 In
this study, we investigate how corruption affects labour supply. Labour
supply is measured by the labour force participation rate (LFPR) and the
employment to population ratio,2 both directly and indirectly.
Numerous papers have investigated the effect of corruption on labour
supply and have found that corruption can affect labour supply through
a number of channels. However, little has been done to establish a more
general theoretical framework to investigate how corruption alters the
is well known that low LFPRs and employment to population ratios
affect productivity at the firm and country levels, reduce tax revenues,
and retard economic growth and development. Thus, investigating this
issue is important from a policy perspective. It is crucial for govern-
ments to minimise the adverse effect of corruption on labour supply to
promote growth. In light of this, we aim to develop a more general
theoretical framework which links corruption to the labour market, and
then we test the predictions stemming from the model, using panel data
methods.

* Corresponding author. School of Economics, University of New South Wales, Kensington, NSW, 2052, Australia.
E-mail addresses: Arusha.Cooray@unsw.edu.au (A. Cooray), ratbek.dzhumashev@monash.edu (R. Dzhumashev).
1
The most commonly used definition of corruption is the abuse of public power for private gain (Mauro, 1998a, b). Corruption has been found to constrain growth
(Rose-Ackerman, 1999; Tanzi and Davoodi, 2002; Mauro, 1995; Mo, 2001), discourage investment (Mauro, 1996; Brunetti et al., 1998; Campos et al., 1999), reduce
foreign direct investment (Wei, 2000; Abed and Davoodi, 2002), and limit productivity (Lambsdorff, 2003) and reduce human capital accumulations through dis-
tortions in the education system (Patrinos and Kagia, 2007). Studies also show that more corrupt countries face higher inflation (Al-Marhubi, 2000), larger shadow
economies (Friedman et al., 2000; Johnson et al., 1997; Schneider et al., 2010), lower state bond ratings (Depken and Lafountain, 2006), and reduced expenditure on
education and health (Mauro, 1998a, b). Shleifer and Vishny (1993) and Bardhan (1997) provide an extended review of different aspects of corruption and its
economic effects.
2
The LFPR is defined as the proportion of the population aged 15 and older who is economically active, that is, all people who supply labour for the production of
goods and services during a specified period as defined by the World Bank (2015) and the International Labour Organisation (ILO, 2015). The employment to
population ratio is the proportion of a country's population who is employed (World Bank, 2015; ILO, 2015).

https://doi.org/10.1016/j.econmod.2018.05.015
Received 2 January 2018; Received in revised form 12 May 2018; Accepted 12 May 2018
Available online 22 May 2018
0264-9993/© 2018 Elsevier B.V. All rights reserved.
A. Cooray, R. Dzhumashev
To address the gap in the literature, we construct a simple theoretical
framework which captures the link between corruption and the labour
supply. The main features of the framework are based upon the studies of
Dzhumashev (2014a, 2014b), Aidt et al. (2008), and Delavallade (2006).
In this framework, corruption is captured as distortions in the burden of
taxes and externalities that arise due to government spending. We also
incorporate the effect of the shadow economy by assuming that some
labour is supplied to the underground sector.3
By analysing the links between corruption and the labour supply, the
present study extends upon the literature in a number of ways. Firstly, the
simple theoretical framework developed captures the main aspects of the
relationship between corruption and the labour market. Secondly, we
test empirically the theoretical conjectures and determine the overall
effect of corruption on the labour supply, as well as the effects of cor-
ruption through different channels. In the estimation, we measure labour
supply by using the labour force participation rate (LFPR), and the
employment to population ratio (EPR). We test the results for robustness
in a number of ways. We use additional control variables and interaction
terms to capture a range of possible influences on the labour market. To
account for country-level time-invariant unobservable influences, we use
different estimation methods, including fixed effects estimation. To
correct for any potential endogeneity bias, we use the system generalised
method of moments (GMM) and instrumental variable (IV) estimations.
Given the uncertainty and likely measurement errors in corruption, we
test the robustness of the results using three different data sets on cor-
ruption: the Transparency International (TI), Kaufmann et al. (2014), and
International Country Risk Guide (ICRG) corruption data sets. Estimation
is also carried out by splitting the sample into two groups – countries with
low levels of corruption (below the mean level of corruption) and those
with high levels of corruption (above the mean level of corruption) – to
investigate whether the adverse effects of corruption on the labour
market are more pronounced for high-corruption countries.
The primary findings of the study can be summarised as follows. The
theoretical framework highlights that corruption reduces labour supply
by encouraging activities in the shadow economy, by reducing the pro-
ductivity of the private sector, and altering the tax burden by distorting
the saving-consumption trade-off. The empirical results of the present
study suggest that corruption reduces the overall LFPR. The estimated
results indicate that corruption has a statistically significant robust direct
negative effect on the LFPR. Corruption is additionally found to have an
indirect adverse impact on the LFPR through the higher burden of taxes,
reducing private sector productivity and increase in size of the shadow
economy. However, higher wages, an increase in consumption, and
better regulatory quality are found to reduce the negative impact of
corruption on labour market outcomes. The findings have the following
policy implications. To reduce the negative effect of corruption on the
labour market, governments need to develop a comprehensive approach
not only to combatting corruption itself but also work towards improving
regulatory quality and promoting policies that reduce activities in the
shadow economy. Improvements in tax administration and encouraging
the private sector not only have a direct effect on output but may also be
viewed as policy measures for improving labour supply in the formal
economy.
The rest of this paper is structured as follows. Section 2 discusses the
literature. Section 3 presents the theoretical framework which links
corruption and the labour supply. Section 4 presents the data and
methodology. Section 5 evaluates the results, and Section 6 concludes.
2. Related literature
Only few studies have investigated the direct effect of corruption on
labour supply. In developing our model, we attempt to synthesise a more
3 4
The shadow economy is defined as ‘the market-based production of goods
and services that escapes detection in official estimates of GDP’ (Smith, 1994).
208
Economic Modelling 74 (2018) 207–218
general link between corruption and the labour market, by incorporating
the mechanisms presented in the literature. We briefly discuss this
literature below.
Mandal and Marjit (2010) develop a theoretical model which shows
the impact of corruption on wage inequality between skilled and un-
skilled labour. Pi and Zhou (2013) extend the wage-inequality analysis
accounting for the effect of institutional quality rather than corruption.
These studies show that the difference in capital intensity between sec-
tors influences the relation between corruption/institutional quality and
the wage-gap between skilled and unskilled labour.4 The cost of cor-
ruption (low institutional quality) in these studies is captured as a drain
on the final output produced, due to the shift of labour and capital to
unproductive use, or loss in output due to corrupt transaction costs. We
follow their modelling technique in spirit by expressing the cost of cor-
ruption as a direct change in final income through the tax burden and
diversion of resources from productive to unproductive use. The main
difference between our model and these models is that their focus is on
explaining the wage-gap between skilled and unskilled labour due to
corruption, whereas our focus is on the distribution of labour between
official and shadow economies and leisure.
Ferraz et al. (2012) in a study of corruption in Brazilian municipal-
ities, observes that students in municipalities with high corruption
receive test scores that are considerably lower and have higher dropout
and failure rates compared to students in municipalities without cor-
ruption. Higher corruption levels are found to lead to fewer teachers,
labs, and other educational facilities. Round et al. (2008) establish that in
Ukraine, individuals find jobs primarily through connections that require
bribe payments. Those who gain long-term employment also face a
number of problems, which involve the lack of job security, informal
payment of wages, and the absence of legal protection from employers.
Walsh (2010) documents similar findings for the Chiang Rai region in
Thailand. Corruption is found to lead to human trafficking,
under-payment of wages, and unsafe working conditions. These studies
are restricted to one country and are based on qualitative methods, as
opposed to our study, which is undertaken on many countries and em-
ploys quantitative methods.
Lacko (2006) examines how tax rates, corruption, and institutional
features of the labour market affect unemployment, employment,
self-employment, and activity in the hidden economy in a group of OECD
countries over the period 1995 to 2000. She finds that the subjective tax
rate is an important factor that explains cross-country differences in
unemployment, employment, and self-employment rates, as well as the
size of the shadow economy. Using arguments along the same lines,
Johnson et al. (1997) show that high regulatory policies and tax collec-
tion encourage individuals to move from the formal to the informal
sector. Ahn and De La Rica (1997) use data on the Spanish labour market
and find that individuals prefer working in the underground sector as
opposed to being unemployed. Field (2007) examines the effect of in-
creases in ownership security on household labour supply in Peru. The
results indicate that tenure rights can increase social welfare with labour
supply shifting from work at home to work in the outside market.
When corruption is ingrained in a society, the labour market is dis-
torted and tends to operate under a ‘shadow’ system (Cooney et al.,
2002). In other words, workers and employers lose faith in industrial and
labour laws, and settle disputes outside the formal system. Dreher and
Schneider (2010), Johnson et al. (1997), Ahn and De La Rica (1997),
Fugazza and Jacques (2003), and Frederiksen et al. (2005) examine the
relation between the shadow economy and the labour market. Dreher
and Schneider (2010) argue that the high levels of regulation that
accompany large shadow economies, lead to higher levels of unem-
ployment and a fall in the number of hours worked in the official sector.
Frederiksen et al. (2005) in a study of the male labour supply in the
See Marjit and Acharyya (2003, 2006), Marjit and Kar (2005), Pi and Zhou
(2012), and Pi et al. (2013) for wage inequality analyses.
A. Cooray, R. Dzhumashev Economic Modelling 74 (2018) 207–218

taxable and non-taxable underground sectors in Denmark, find that wage shadow economy; hence, h ¼ 1  l  s is devoted to work for firms
rates and non-labour income have important effects on the labour supply operating in the formal economy. Note, to avoid cluttering, we omit time
in both sectors. They find that taxes tend to move labour away from the indices. The agents consume part of their incomec, and save the rest, a.
formal to the informal labour market. Analogously, Fugazza and Jacques Their savings are supplied as capital to firms in the formal economy ða ¼
(2003) in a theoretical model show that individuals move underground _ This also implies that the production in the shadow economy is based
kÞ.
in order to avoid taxation and government labour regulations. In on labour only (see, e.g. Foellmi and Oechslin, 2007). In the formal
contrast, higher wages and unemployment benefits are found to economy, labour income is taxed at the rate of τw , whereas capital income
encourage greater participation in the formal sector. is taxed at the rate of τk . The government exogenously determines the tax
Corruption can also affect the labour market indirectly through many rates. Given the governmental parameters and corruption outcomes, the
channels. Cooray and Schneider (2016) show that countries which are agents choose their leisure, consumption, and capital accumulation
corrupt find it difficult to retain and encourage skilled workers, which within their budget constraint. The distribution of their labour between
affects the formal labour market adversely. In addition, corruption has the formal and shadow economies is determined by the marginal pro-
been found to change the size and composition of public expenditure ductivities in the corresponding sectors.
away from important sectors, such as health and education (Mauro,
1998a, b; Wei, 2001), towards sectors such as defence, which involve 3.1.2. Production
greater secrecy. Similarly, Svensson (2003) observe that corrupt bu- The firms in the official economy produce a single good using the
reaucrats price discriminate to manipulate access to public services. If following production function,
corruption affects the provision of health and education services, by
increasing the cost and lowering the quality of these services (Gupta y ¼ Agα k1α ð1  l  sÞη ; (1)
et al., 2002; Kaufmann, 2010), corruption will lower the returns to a
country's workforce. Corruption can increase tax evasion, by weakening where g is productive governmental services, k is capital per capita, and A
the tax system. This reduces the resources available for social welfare is the exogenously given technology coefficient. In the absence of cor-
spending (Gupta et al., 2002) which will lower the returns to the labour ruption, the governmental input is determined endogenously as
force.
g ¼ τw wð1  l  sÞ þ τk rk; (2)
However, scholars have found that corruption might be beneficial for
an economy. Smarzynska and Wei (2000) observe that higher levels of
where wis the real gross wage rate in the formal economy, and r is the
corruption in Eastern Europe have increased the potential for joint ven-
rate of return to capital. The factors are paid their marginal products:
tures. Uhlenbruck et al. (2006) similarly note that firms adjust to cor-
ruption by entering into short-term contracts and engaging in joint ∂y ηy
w¼ ¼ ; (3)
ventures. Nevertheless, this is just one of many other possible channels ∂ð1  l  sÞ 1  l  s
through which corruption affects the labour market.
This study extends upon previous studies by developing a theoretical ∂y ð1  αÞy
model which links the effects of corruption to labour supply through the r¼ ¼ : (4)
∂k k
different channels that have been considered in the literature mentioned
above. Then, using a panel dataset for 132 countries, we empirically The shadow economy produces using only labour, and the production
investigate the impact of corruption on the LFPR and employment to function is given as follows:
population ratio through the channels established in the theoretical
ys ¼ Bs;
model.
where B is the productivity coefficient. Therefore, the wage rate in the
3. The theoretical framework shadow economy is given by:

To motivate the empirical analysis, we present a simple theoretical ws ¼ B:

framework which links corruption to labour supply in the formal and
The agents face the following budget constraint:
shadow economies. Although existing models of the labour supply
consider different aspects of the labour market, the effects of corruption k_ ¼ ð1  τw Þwð1  l  sÞ þ ws ð1  σ Þs þ ð1  τk Þrk  c; (5)
on the labour supply have not been investigated analytically. It is known
that in theoretical modelling, to focus on some important aspects of the _
where, kstands for a change in the capital stock owned by the agent. The
economic problem, other important aspects are ignored initially. Based
first term, on the right side of the equation, is the after-tax labour income
on this rationale, the standard labour supply model ignores the effect of
earned in the formal sector. The second term is the income earned in the
corruption. The conventional practice in taking into account the ignored
shadow economy after some costs are incurred that are given as a fraction
factor is that the assumption about that factor is relaxed and its effect is
of the wage rate in the shadow economy, σ ðFugazza and Jacques; 2003Þ:
incorporated into the model solutions. In our context, the effect of cor-
We assume that there are no other costs for agents operating in the
ruption would be incorporated into the outcomes of the labour supply
official economy, except taxes. In reality, there can be other costs, but we
model. Therefore, we follow this logic and commence with the bench-
normalise those possible costs to zero relative to the costs in the shadow
mark labour supply model, which ignores the existence of corruption.
economy. The reason is that due to the illicit nature, working in the
Then we incorporate the effects of corruption by linking it to different
shadow economy incurs higher costs than working in the official econ-
mechanisms through which corruption creates distortions.
omy. With higher corruption, this cost decreases by making shadow ac-
tivities more attractive (Dreher and Schneider, 2010).
3.1. The standard model of labour supply
3.1.3. Market clearing conditions
3.1.1. Agents The goods market clears according to the conditiony þ ys ¼ c þ a þ g:
We assume the following preliminaries. There is a continuum of That is, the total income is spent on consumption, c, saving, a, and
identical agents. They consume a single good, c. The agents are hetero- governmental expenditures,g. The labour market clears by each unit of
geneous in terms of wealth and productivity, but in terms of labour time is being used to work in official and shadow economies, and the rest
endowment, the agents are identical. Each agent is endowed with a unit is spent as leisure time: 1 ¼ h ¼ s þ l. The capital market clears by
of time, out of which l is chosen for leisure, and s is spent working in the

209
A. Cooray, R. Dzhumashev Economic Modelling 74 (2018) 207–218

savings being spent on capital investment according to the conditionk_ ¼ The condition w ¼ Bð1 σÞ ηy
ð1τw Þ ¼ ð1lsÞ implies that the equilibrium value
ð1  τw Þwð1  l  sÞ þ ws ð1  σ Þs þ ð1  τk Þrk  c. The public budget is of the official sector labour supply is given by
balanced as all tax revenue collected is spent on public goods production.
ηyð1  τw Þ
h¼1ls¼ : (15)
3.1.4. Equilibrium of the model Bð1  σ Þ
In the model, the variables c; a; l; s; w; ws ; r; k; g; y; ys ; h; and f are
We can also determine the fraction of the labour supply in the official
endogenous as they are determined within the model. However, the
sector by denoting the fraction as f , so that h ¼ ð1  lÞf . Thus,
variables τw ; τk ; σ ; bw ; bk ; μ; ε; and χ are exogenous, and the parameters. θ;
γ; α; η; A; B; and e are assumed as given: h ηyð1  τw Þ
The equilibrium is defined as a sequence of choice variables for an f ¼ ¼ : (16)
1  l Bð1  σ Þ  θc
agent:c;l;s, the state variable k, and factor prices w; ws ; and r such that: (i)
agents maximise the utility given their budget constraint; (ii) firms Using equation (9) and equation (12), one can solve for the growth
maximise profits given factor prices and the optimal response of private rate of consumption:
agents to corruption; and (iii) all markets clear as described above.
c_ y
¼ ð1  τk Þr  ρ ¼ ð1  τk Þð1  αÞ  ρ:
c k
3.1.5. Solution to the agents problem
By choosingc; l; s and k, the agents maximise their life-time utility On the balanced growth path, consumption and capital need to grow
_
given by at the same rate. Therefore, cc_ ¼ kk holds on the balanced growth path.
Z Using equation (7) and substituting for the solutions of l and s, we can
∞
1 θ γ ρt
U¼ cl e dt; (6) write:
0 γ
 
k_ y Bð1  σ Þ  θc  ηyð1  τw Þ y c
subject to: ¼ ð1  τw Þη þ þ ð1  τk Þð1  αÞ  :
k k k k k
k_ ¼ ð1  τw Þwð1  l  sÞ þ ws ð1  σ Þs þ ð1  τk Þrk  c; (7) _
Substituting the right sides of both growth ratescc_ ¼ kk , we can find:

and standard initial and transversality conditions. ð1  τk Þð1  αÞ yk  ρ ¼ ð1  τw Þη yk þ Bð1σÞθc
k
ηyð1τw Þ
þ ð1  τk Þ
The problem of the agent is solved by maximising the following 
current - value Hamiltonian: ð1  αÞ yk  kc ; solving we obtain the consumption function,

1 θ γ
H¼ cl þ λ½ð1  τw Þwð1  l  sÞ þ ws ð1  σ Þs þ ð1  τk Þrk  c: (8) 1
γ c¼ ½Bð1  σ Þ þ ρk: (17)
1þθ
Using this Hamiltonian, the following first-order conditions with
respect to c; l; sand k are obtained
3.2. How to incorporate the effects of corruption?
cγ1 lθγ  λ ¼ 0; (9)
Next, given the benchmark solution for the labour supply, we need
γ θγ1 to develop an argument about how the effects of corruption alter this
θc l  λ½ð1  τw Þw ¼ 0; (10)
solution. Before discussing the effects of corruption, we need to define
corruption. Corruption is defined as the misuse of a public position to
ð1  τw Þw þ ws ð1  σ Þ ¼ 0; (11)
create and capture private rents. In pursuit of private rents, corrupt
bureaucrats distort the functioning of the public sector in two broad
∂H _
 ¼ λ  ρλ: (12) aspects (see Aidt et al., 2008; Delavallade, 2006; Dzhumashev, 2014a;
∂k 2014b): the burden of taxes will be different from the statutory rates,
Substituting for λin equation (10) from equation (9) and simplifying and public production will be lower due to embezzlement and in-
yields the following: efficiency. Due to corruption-induced tax evasion, the taxpayers pay
only a fraction of the tax liability. This implies that the government
θc collects only effective tax revenue given by the μτw and ετk tax rates,
¼ ½ð1  τw Þw:
l where μ and ε are less than unity due to tax evasion. However, cor-
ð1τw Þw ruption can be also extortive, which generates an additional cost in the
From equation (11) we obtainws ¼ ð1σ Þ
. By recalling thatws ¼ B,
Bð1σ Þ form of bribe rates denoted as bk and bw . The costs of and gains from
we writew ¼ ð1τw Þ
.
corruption are proportional in the tax liability as the interaction of the
Substituting for w in the condition given byθcl ¼ ½ð1  τw Þw, we taxpayer and the tax inspectors is based on the statutory tax liability.
solve for l and obtain the following equilibrium condition for leisure: With corruption, instead of statutory tax rates, we use effective tax
rates:
θc
l¼ : (13)
Bð1  σ Þ ~τw ¼ ðμ þ bw Þτw and ~τk ¼ ðε þ bk Þτk : (18)
Bð1σ Þ ηy
Accounting for equation (13) and thatw ¼ ð1τw Þ
¼ ð1lsÞ
, we solve We capture the effect of embezzlement and inefficiency in public
fors: good production (Aidt et al., 2008; Delavallade, 2006; Dzhumashev,
2014a, 2014b) through the effective governmental productive input
θc þ ηyð1  τw Þ given as follows:
s¼1 : (14)
Bð1  σ Þ
~g ¼ eg;
This result imposes an additional feasibility conditionθcþηBð1
yð1τw Þ
σÞ < 1.
To rule out trivial cases with no shadow economy, we assume that this where e stands for the effective quotient that depends on the degree of
condition holds. corruption in the government. As corruption always leads to in-
efficiencies, we can safely assume that e < 1.

210
A. Cooray, R. Dzhumashev Economic Modelling 74 (2018) 207–218

Corruption increases government input requirements, and thereby additional burden stemming from bribesbw and bk , we may have ~τw  τw ;
decreases productive governmental services and output per capita. We ~τk  τk or have ~τw < τw and ~τk < τk . This implies that the effect of cor-
can establish this by considering the production function y ¼ ruption on the burden of taxes is ambiguous; and
Ag α k1α ð1  l  sÞη that if corruption reduces the value of productive thus,∂∂χ
~τw
≷ 0 and ∂~τk
≷ 0; in general: (iii) The consumption function
∂χ
public inputs ðgÞ, then the level of output ðyÞ should also fall. Another given by equation (16) implies that corruption may reduce or raise the
possible effect of corruption is transmitted through the cost differential burden of taxes, in addition to a negative impact of corruption on income.
between the shadow and formal economies. Corruption decreases the These changes jointly imply that the effect of corruption on consumption
cost of activities in the shadow economy by allowing more loopholes for ∂~c ≶ 0, depending on which effect dominates.
is not clear; thus∂χ
such activities (Fugazza and Jacques, 2003). In our setting, given that the
costs associated with shadow activities reduce ð~ σ < σ Þ with corruption,
the share of the labour supply in the shadow economys, increases. Cor- 3.3. Analysis of the effect of corruption on labour market outcomes
ruption may also affect consumption as illicit income may be spent
differently compared to legally obtained income (Blackburn et al., 2006). Given that output, consumption, the burden of the public sector, and
Therefore, this effect of corruption may alter the trade-off between the size of the shadow economy depend on the level of corruption in the
consumption and leisure, which then alters the labour supply. economy, we can consider a comparative static analysis of the equilib-
In light of this discussion, we can incorporate the effects of corruption rium share of the labour supply in the formal sector. Therefore, we
into the labour supply. Specifically, we modify the share of labour analyse equation (19) and using the conjectures from Proposition 1, state
engaged in the official sector. Recall that the share of the official labour the following proposition.
ηyð1τw Þ
supply in equilibrium is given by f ¼ Bð1σ Þθc. We re-write this expres- Proposition 2. An increase in the level of corruption affects the share of
sion as follows: labour supply in the formal sector through the following channels:

η~yð1  ~τw Þ
f ¼ ; (19)
Bð1  σ~Þ  θ~c
where y~ is the output,~c is the consumption per worker, σ~is the cost of
working in the shadow economy, and ~τw is the effective tax rate on labour
income, in the presence of corruption. In this expression, formally, there
is no difference from the benchmark case. However, the main difference
is that this formulation allows identifying the channels through which
corruption affects the labour market outcomes: through distortions in the
burden of taxes, in consumption, in the level of productive public input
provided, and in the cost of working in the shadow economy.
How corruption alters the burden of taxes or public regulation versus
how corruption distorts production depends on different interactions
between the private and public agents. However, the uniting aspect for
both types of corruption is the underlying institutional setting that drives
the corrupt behaviour of bureaucrats (Blackburn et al., 2006; Aidt et al., Proof.
2008). The multi-faceted nature of corruption makes it difficult to mea-
sure it objectively. The existing solution in the field is to construct an ∂f
(i) A fall in output per capita of the formal sector due to corruption,
has a negative effect on labour supply in the formal economy;
(ii) Corruption decreases the participation cost of the shadow econ-
omy, σ. This leads to an increase in the size of the shadow econ-
omy, and thus, reduces the share of the labour supply in the formal
sector;
(iii) If corruption reduces the burden of the labour income taxation,
then the labour supply in the formal sector decreases; otherwise, it
increases;
(iv) Corruption may affect consumption, and hence, a trade-off be-
tween consumption and leisure, which alters the share of labour
supply in the formal sector.
By taking the first-order derivative of equation (19), we obtain:
h i h i
∂~y ∂~τ
∂χ ηð1  ~τw Þ  ∂χ ηye Bð1  σ~Þ þ η~yð1  ~τw Þ B ∂∂χσ~  θ ∂χ
∂c

index that attempts to capture these diverse aspects of the phenomenon
of corruption. This has led to the construction of corruption measures,
such as the corruption perception index (CPI) provided by Transparency
International and Corruption Control by the World Bank. We employ
these measures of corruption as proxies for the levels of corruption across
countries in our empirical analysis. To relate to the empirical measure of
corruption given as CPI, we formalise the measure of corruption in our
analytical results by introducing a corruption index, χ. That is, we assume
that the variables which are affected by corruption, given in equation
(19), can be treated as functions of the corruption index χ. This allows us
to map the analytical results to the empirical model that uses CPI as a
measure of corruption.
Based on the discussion above, we state the following Proposition
about the effects of corruption on different variables that drive the share
of labour allocated in the formal sector.
Proposition 1. An increase in the level of corruption has the following ef-
∂y~
fects: per-capita output in the formal sector falls, ∂χ < 0; the changes in the
effective tax burden and consumption are ambiguous. Thus, ∂~τw
∂χ
and ∂~c ≶ 0:
∂χ
Proof. (i) Recall that corruption reduces the value of the public input
ð~ y ¼ A~g α k1α ð1  l  sÞη < y ¼ Ag α k1α ð1  l  sÞη Þ; thus,
g < gÞ. Then, ð~
the output level should fall with the corruption level. Hence, ∂∂χy < 0: (ii)
Depending on whether μ and ε reduce the tax burden more than the
¼ ; (20)
∂χ ðBð1  σ~Þ  θcÞ2
(i) From equation (20), it can be established that the direct effect of
the fall of the output per worker on the formal labour is negative,
b
as ∂∂χy < 0.
(ii) Recall that∂σ
∂χ < 0 . Based on equation (20), this implies that with a
higher level of corruption, the share of the labour supply in the
formal sector falls.
(iii) It can be verified from equation (20) that if ∂∂χ
~τw
< 0, then its effect
on the share of the labour supply in the formal sector is positive;
otherwise, the effect is negative.
∂c ≶ 0, equation (20) indicates that the effect of cor-
(iv) Given that∂χ
ruption on the share of the labour supply in the formal sector is
ambiguous.
≷ 0; ∂∂χ~τk ≷ 0; Therefore, the overall effect of corruption on the labour supply de-
pends on the dominating effect of corruption through these channels.
The analytical result above identifies several channels through which
corruption affects the share of the labour supply in the formal sector.
However, due to the analytical ambiguity, determination of the effective
~ impact of corruption on the labour supply becomes an empirical
question.

211
A. Cooray, R. Dzhumashev Economic Modelling 74 (2018) 207–218

4. Estimation methodology and data test indicated that a fixed effects model was more reliable than a random-
effects model. Therefore, results are reported only for the panel fixed-
4.1. Methodology effect estimation. The fixed-effects estimator allows controlling for any
unobserved country-specific time-invariant effects. The panel data model
We use equation (19) and Proposition 1 as the basis for our empirical is expressed, using the traditional notation, as follows:
model. One cannot employ a simple linearization using a log form in this
case. Taking logs of equation (19) does not yield a simple linear equation. yit ¼ Xit β þ ϖ i þ ηt þ μit ; (22)
Therefore, instead we use a heuristic approach to linearizing the share of
whereyit is the LFPR/employment to population ratio for country i in
labour in the formal sector given by equation (19), by focusing on the
period t. Xit is a vector which includes all independent variables,
variables that are affected by corruption.
including corruption and the control variables; ϖ i captures a country-
Recall the equilibrium expression for the official labour supply given
ηy~ð1~τw Þ
specific effect, ηt, takes into account relevant time effects, and uit is a
as f ¼ . In this expression, we can identify the following vari- random error term. Interaction terms are added to equation (22) to
Bð1~ σ Þθc
ables: consumption per capita~c, the tax rate~τw ; and the shadow economy investigate the desired differential effects, whilst disregarding other
cost parameter~ σ . The last parameter is not observed; thus, we need to differences not measured at the aggregate level.
proxy it with another available measure. Given that the size of the It can be argued that not all explanatory variables used in our
shadow economy, ~ζ, depends on the costs associated with shadow ac- empirical model are strictly exogenous. An approach that allows con-
tivities, σ~; we use ~ζinstead of~ σ . The share of the labour supply in the trolling for the joint endogeneity of explanatory variables with internal
formal sector also depends on consumption, ~c, and per-capita income, y~. instruments is the Arellano and Bover (1995) and Blundell and Bond
There are also some country-fixed effects (denoted as~ ω) to capture such (1998) system GMM estimator. In summary, equation (23), which in-
parameters asη, andθ. These parameters depend on the institutional set- volves variables in levels, is combined with equation (24), which in-
up of the economy. Thus, we assume that from the empirical perspective, volves variables in the first differences. Equation (23) is instrumented by
we can proxy these parameters using the measures of institutional the lagged first differences of the variables, whereas equation (24) is
quality,~q. In light of this discussion, we can write equation (19) in a more instrumented by lagged variables in levels:
general form that depends directly on the discussed variables,
yit ¼ γyit1 þ Xit β þ ϖ i þ ηt þ μit ; (23)
 
q; ~ζ; ω
h ¼ ϕ ~τw ; ~y; ~c; ~ ~ : (21)
yit  yit1 ¼ γðyit1  yit2 Þ þ βðXit  Xit1 Þ þ ηt þ ðμit  μit1 Þ: (24)
In general, equation (21) is non-linear, as is equation (19). However,
The variable definitions are the same as above for equation (22), with
for empirical analysis we need to simplify and represent the equation
the lagged values of the variables now entering the equation. The GMM
using some linear form. The simplest form to express this function should
estimator is based on the assumption that the error terms are not serially
be that each of the arguments on the right side can be expressed as a sum
correlated and that the explanatory variables are weakly exogenous or
of the direct effect of the variable and its effect conditional on corruption.
not statistically significantly correlated with future realisations of the
Therefore, we use the findings of Proposition 1 combined with our as-
error terms under which the following moment condition holds for the
sumptions to identify which variables are affected by the level of cor-
first difference estimator:
ruption. In practice, the level of corruption is usually represented using
E[yit-s (μit - μit1 )] ¼ 0; E[Xit-s (μit - μit1 )] ¼ 0; where i ¼ 1 …. .n, t ¼ 3
the corruption perceptions indices such as the CPI of Transparency In-
…. T and s  2.
ternational and the Corruption Control Index of the World Bank. To
As mentioned above, the levels equation is instrumented with the
relate the theoretical model to the empirical measure of corruption given
lagged first differences of the variables, which leads to an additional
as CPI, we formalise the measure of corruption in our empirical analysis
moments condition as follows:
by using a corruption index, χ. That is, we assume that the variables,
E[yit-s (ϖ i þμit )] ¼ 0; E[Xit -s (ϖ i þ μit )] ¼ 0 fors ¼ 1.
which are affected by corruption, given in equation (19), can be treated
Two diagnostic tests are carried out: the Hansen test for over-
as functions of the corruption index χ. Thus, given that χ is the level of
identifying restrictions under which the null hypothesis is that the in-
corruption measured by the corruption index, we can write the effect of
struments are not correlated with the residuals and the Arellano-Bond
corruption on the variables it influences; for example, ~τw ¼ τw ð1 þ χ Þand
test for second-order correlation in the first differenced residuals.
~τk ¼ τk ð1 þ χ Þ. This approach allows us to use the corruption perception
It is possible that the LFPR and corruption are both related to a third
indices available in the literature as a proxy for χ . Following this
variable, which is omitted from the estimation. In this case, the rela-
approach and taking into account that corruption affects labour and
tionship between the variables could be a correlation rather than a causal
capital incomes differently, we separate the effect of income per capita y~
relation. Therefore, as a further test for establishing causality, the IV
into labour income~ w ¼ wð1 þ χ Þ and capital income~r ¼ rð1 þ χ Þ based on
estimation could be useful in addition to the system GMM. In the IV
an agent's budget constraint, equation (7). We recall that Meon and Weill
estimation, a good instrument should be correlated to corruption and not
(2010), Aidt (2009), Aidt et al. (2008), and Mendez and Sepúlveda
influence the LFPR through other channels. If corruption is also influ-
(2006) find that the effect of corruption on economic performance de-
enced by a third factor which also influences the LFPR, then the esti-
pends on the quality of institutions and governance in a country. This
mated regression would change when different variables related to
finding implies that the effect of institutional quality on labour supply
corruption but not directly associated to the LFPR were used as in-
can depend on its interaction with the level of corruption; we write~ q ¼
struments for corruption (see Pritchett and Summers, 1996). Amongst
qð1 þ χ Þ. Analogously, the size of the shadow economy and consumption
the variables used as instruments for corruption are ethnolinguistic
are conditioned as follows: ~ζ ¼ ζð1 þ χ Þ and ~c ¼ cð1 þ χ Þ.
fractionalisation (Mauro, 1995), the settler mortality data of Acemoglu
In light of the discussion above, we formulate the equation that
et al. (2001) and Ahlin and Pang (2008), latitude (Gupta et al., 2002;
captures the relationship between the share of the labour supply in the
Cooray and Schneider, 2016), and the initial level of corruption (Gupta
official sector and other variables of interest as follows:
et al., 2002; Cooray and Schneider, 2016). Following Gupta et al. (2002)
f ¼ βX þ u; and Cooray and Schneider (2016), latitude and the initial level of cor-
ruption are used as instruments. Latitude is correlated to institutions
where X 2 ½~τ;~r ; w ~; ~ζ; χ .
~ ;~c; q (Acemoglu et al., 2005) but should not directly influence labour market
This model can be estimated using panel fixed effects. The Hausman outcomes. The initial level of corruption is used because it is
pre-determined compared to the current values of corruption (see Gupta

212
A. Cooray, R. Dzhumashev
et al., 2002). The latitude data are sourced from the GeoDist dataset
compiled by CEPII (2014). The F-statistic, Cragg-Donald Wald F-statistic,
and Kleibergen-Paap Wald F-statistic are used for testing instrument
relevance. The Hansen test for over-identifying restrictions under which
the null hypothesis is that the instruments are not correlated with the
residuals is carried out under the IV estimation.
4.2. Data
The data cover the period 1996–2012, for 132 countries. The TI and
Kaufmann et al. indices are available only from 1995 to 1996 onwards,
respectively. For this reason, the panel dataset starts from 1996. The
sample constitutes a representative panel of regions covering Eastern
Europe and Central Asia, the Middle East and North Africa, Latin America
and the Caribbean, East Asia and the Pacific, South Asia, Sub-Saharan
Africa, and the high-income OECD countries. The grouping of regions
is based on the World Bank classification.
The dependent variables in the study are the labour market variables.
They include the LFPR and the employment to population ratio. The
employment to population ratio is used as an additional robustness check
(Nickell, 1997). The labour market variables are as defined in Section 1.
These data are obtained from the World Development Indicators (WDI).
The main independent variable of interest is corruption. Three measures
of corruption are used in the empirical study which follows. One is the
corruption measure from Transparency International (TI). Here, the es-
timate of corruption ranges from 0 (totally corrupt) to 10 (not corrupt).
The second is the estimate of corruption from Kaufmann et al. (2014)
which ranges from approximately 2.5 (totally corrupt) to 2.5 (not
corrupt). Due to problems associated with the TI and Kaufmann et al.
indices (changing methodology and inter-year variation of the former,
changing sources used in the construction of both indices (Treisman,
2007), and close correlation of the Kaufmann et al. index with other
indicators of governmental quality), the ICRG corruption index is used as
an additional robustness check. The ICRG index ranges from 0 (totally
corrupt) to 6 (not corrupt). All corruption indices have been reversed so
that 0 stands for not corrupt, and higher values on the indices stand for
higher levels of corruption.
A number of other control variables based on the theoretical model
are used in the empirical analysis that follows. In our theoretical model,
income per capita incorporates labour income and capital income based
on the agent's budget constraint. Wage and salaried workers as a per-
centage of the total employed are used to capture labour income. If the
substitution effect outweighs the income effect, then an increase in wages
will lead to an increase in the LFPR. If however, the income effect is
stronger than the substitution effect, then there will be an inverse rela-
tion between an increase in wages and the LFPR (Mincer, 1962). To
capture the full effect of income, we also account for a possible effect of
capital income on the labour supply by including the interest rate as a
measure of capital income. The lending rate is used for this purpose.
Similarly, an increase in consumption (fall in leisure) will lead to an in-
crease in the LFPR, and a fall in consumption (increase in leisure) will
lead to a fall in the LFPR. We use household final consumption expen-
diture as a percentage of the gross domestic product (GDP) to measure
consumption. A number of variables to capture different aspects of
institutional quality are included. The literature and our theoretical
5
In calculating the shadow economy, Schneider et al. (2010) assume that
with an increase in direct and indirect taxation, an increase in social security
contributions, and higher regulation, the lower the quality of state institutions,
the lower the tax morale, and the larger the shadow economy. They also assume
that the higher the unemployment, the more people engage in shadow economic
activities and the lower the GDP per capita of a country, and the higher the
incentive to work in the shadow economy. Because the shadow economy cannot
be measured directly, they use the Multiple Indicator Multiple Cause (MIMIC)
approach to estimate the shadow economy.
213
Economic Modelling 74 (2018) 207–218
model show that institutions affect the labour market (Gonzales et al.,
2015; Cooray et al., 2017). The polity index from the Marshall and
Jaggers (2014) database and regulatory quality from Kaufmann et al.
(2014) are used to capture institutions. Evidence shows that the larger
the shadow economy in a nation, the lower the employment in the formal
sector (Dreher and Schneider, 2010). Data for the shadow economy are
from Schneider et al. (2010). According to Schneider et al. (2010), the
main factors influencing the shadow economy are the overall tax and
social security contribution burden, intensity of regulations, public sector
services, and the official sector.5 According to the Laffer curve, as tax
rates increase from low to high levels, tax revenues and labour force
participation rates will increase, and beyond a certain point, an increase
in tax rates would discourage people from working thereby reducing tax
revenues. We use tax revenues as a percentage of GDP to control for the
effect of taxes on the LFPR.
In line with the discussion of our empirical model and extant results in
the literature, a number of interaction terms are also added to the
empirical analysis that follows to investigate whether these variables
magnify or reduce the effect of corruption on the LFPR. Evidence shows
that a high degree of governmental regulation can cause a larger shadow
economy as firms may need to bribe governmental officials to receive
approval for projects (Fugazza and Jacques, 2003). Therefore, an inter-
action term is incorporated for corruptionregulatory_quality.6 Our theo-
retical model and Dreher and Schneider (2010) find that the larger the
shadow economy in a nation, the higher the negative effect of corruption
on formal sector employment. We include an interaction term for cor-
ruptionshadow_economy to capture this effect. As corruption will distort
the consumption-leisure trade-off, by reducing a person's effort (Lambs-
dorff, 2005), an interaction term is added for corruptionconsumption. We
incorporate an interaction term, corruptiontaxes, as studies find that
high tax rates move workers away from the formal to the informal labour
market (Fugazza and Jacques, 2003; Frederiksen et al., 2005; Lacko,
2006). Interaction terms are also incorporated for corruptionwages as
higher returns to employment can reduce the negative effects of cor-
ruption on the LFPR. If corruption reduces a person's effort, then they
could devote more time to leisure than work (Lambsdorff, 2005). An
interaction term is incorporated for corruptionconsumption, as corrup-
tion can distort the consumption-leisure trade-off.7
5. Empirical results
5.1. Fixed effects estimation
The results are reported in Table 1. Columns (1) and (2) report pre-
liminary results for the LFPR using the TI corruption index and the
Kaufmann et al. corruption index, respectively. Columns (3) and (4)
report results using the two indices and incorporating a number of
interaction terms. All variables except for the corruption, regulatory
quality, and polity indices have been converted into logarithmic form for
the empirical estimation.
We first discuss the results in columns (1) and (2). The results suggest
that corruption does indeed have a negative effect on the LFPR using the
TI and Kaufmann et al. corruption indices. Column (1), for example,
suggests that a 1-unit increase in the TI corruption index leads to a 0.15%
fall in the LFPR, and column (2) that a 1-unit increase in the Kaufmann
et al. corruption index leads to a 0.13% fall in the LFPR. Wages and
consumption have a positive and statistically significant effect on the
LFPR. Higher wages and interest income encourage greater participation
6
Regulatory quality captures perceptions of the ability of the government to
formulate and implement sound policies and regulations that permit and pro-
mote private-sector development (Kaufmann et al., 2014).
7
We also trialled with an interaction term for corruptioninterest_rate, to see
how this income affected the corruption-labour supply relation. Since it was not
statistically significant, we have not reported this.
A. Cooray, R. Dzhumashev Economic Modelling 74 (2018) 207–218

Table 1 the findings of Fugazza and Jacques (2003), Frederiksen et al. (2005),
Fixed effects estimation. and Lacko (2006), the results indicate that the negative effect of cor-
VARIABLES (1) (2) (3) (4) ruption on the LFPR is increased due to taxes. Similarly, the interaction
terms on corruptionwages suggest that higher returns to labour reduce
LFPR LFPR LFPR LFPR
the negative effect of corruption on the LFPR. Thus, whilst the negative
Corruption TI - 0.153*** – 0.140*** – effects of corruption outweigh the positive effects of improved regulatory
(0.033) (0.022)
Corruption K – 0.127*** – 0.153***
quality, wages, and higher consumption, they have the potential to
(0.051) (0.032) reduce the effect of corruption on the LFPR.
Wages 0.119*** 0.176*** 0.172*** 0.148***
(0.041) (0.054) (0.052) (0.048)
Interest Rate 0.102 0.113** 0.110 0.111* 5.2. Correcting for endogeneity
(0.119) (0.055) (0.105) (0.064)
Consumption 0.114*** 0.215* 0.203* 0.160**
(0.034) (0.145) (0.106) (0.080) Next, the estimation is carried out by using the system GMM and IV
Taxes 0.214** - 0.361** 0.043** 0.050** methods to address potential endogeneity concerns. Table 2 replicates
(0.102) (0.174) (0.021) (0.022) the estimation in columns (3) and (4) of Table 1 using system GMM
Shadow Economy 0.030*** 0.030*** 0.045*** 0.156***
estimation.
(0.012) (0.010) (0.012) (0.041)
Polity Index - 0.003*** - 0.001 0.001 0.001 The system GMM results are, in general, consistent with those ob-
(0.001) (0.002) (0.005) (0.005) tained under the fixed estimation method. Corruption as measured by the
Regulatory Quality 0.004** 0.002* TI and Kaufmann et al. indices have a negative and statistically signifi-
(0.002) (0.001) cant effect on the LFPR. Consumption and wages have a positive and
Corruption TI#Reg. – – 0.060** –
Quality (0.005)
statistically significant effect on the LFPR, suggesting once again that the
Corruption K#Reg. – – – 0.009*
Quality (0.005) Table 2
Corruption TI#Shadow – – 0.165** System GMM estimation.
Economy (0.072)
Corruption K#Shadow – – – 0.180*** VARIABLES (1) (2)
Economy (0.048)
LFPR LFPR
Corruption TI#Taxes – – 0.233***
(0.101) Corruption TI 0.122*** –
Corruption K#Taxes – – – 0.191** (0.045)
(0.090) Corruption K
Corruption TI#Wages – – 0.014* – Wages – 0.129***
(0.007) (0.028)
Corruption K#Wages – – – 0.012* 0.179*** 0.208*** (0.060)
(0.006) (0.050)
Corruption – – 0.053** – Interest Rate 0.111** 0.045
TI#Consumption (0.026) (0.050) (0.052)
Corruption – – – 0.062** Consumption 0.135* 0.153*
K#Consumption (0.031) (0.076) (0.086)
R-squared 0.68 0.65 0.75 0.70 Taxes 0.030* 0.024**
Observations 185 198 185 198 (0.018) (0.012)
Shadow Economy 0.102*** 0.118***
Note: Robust standard errors in parentheses. Significant at the 1%, 5% and 10% (0.020) (0.024)
levels respectively. Polity Index 0.002 0.008
(0.002) (0.009)
Regulatory Quality 0.004** 0.002
in the formal sector. The results appear to suggest that the substitution
(0.002) (0.002)
effect outweighs the income effect in this case. The interest rate is not Corruption TI#Reg. Quality 0.014** –
consistently statistically significant. (0.007)
The shadow economy and tax revenues have a negative and statisti- Corruption K#Reg. Quality – 0.014*
cally significant effect on the LFPR. The results suggest that a larger (0.007)
Corruption TI#Shadow Economy 0.129** –
shadow economy discourages employment in the formal sector (Dreher (0.055)
and Schneider, 2010). Higher tax revenues similarly discourage people Corruption K#Shadow Economy – 0.145* (0.076)
from participating in the labour force. The coefficient on the polity index Corruption TI#Taxes 0.265*** –
is statistically significant only in column (1) and the coefficient on the (0.112)
Corruption K#Taxes – 0.317***
regulatory quality is significant in columns (1) and (2). Perhaps, this is
(0.118)
because corruption and the quality of institutions are closely related. Corruption TI#Wages 0.113** –
Next, the model is estimated in columns (3) and (4), by incorporating (0.052)
a number of interaction terms. The coefficients on the main variable of Corruption K#Wages – 0.120***
interest, corruption, continue to be negative and statistically significant. (0.010)
Corruption TI#Consumption 0.017* –
The results are in general consistent with those obtained above. A look at (0.010)
the interaction term on corruptionregulatory_quality suggests that better Corruption K* Consumption – 0.015**
regulatory quality reduces the impact of corruption on the LFPR. The (0.007)
coefficient on the interaction term however, is still negative suggesting Lag Dependent Variable 0.389*** 0.421***
(0.044) (0.052)
that the negative effects of corruption outweigh the positive effects of
Hansen Test for Over-identifying Restrictions: p 0.17 0.20
regulatory quality. More sound policies that encourage private-sector value
participation will reduce the adverse effects of corruption on the LFPR. Arellano-Bond Test for 2nd Order 0.26 0.33
The interaction term on corruptionshadow_economy has a negative and Autocorrelation: p value
statistically significant effect on the LFPR consistent with the views of Observations 185 198

Dreher and Schneider (2010) in that a larger shadow economy intensify Note: Standard errors reported in parentheses. *, **, *** Significant at the 10%,
the negative effects of corruption on the labour market. Consistent with 5% and 1% levels respectively.

214
A. Cooray, R. Dzhumashev Economic Modelling 74 (2018) 207–218

substitution effect outweighs the income effect. Taxes and the shadow F-statistic and the Kleibergen-Paap Wald F-statistic become more reliable
economy have a negative and statistically significant effect on the LFPR. as statistical inferences. An F-statistic above the critical value (10%
The polity index and the interest rate are not consistently statistically maximal test size) indicates the rejection of weak instruments. Addi-
significant. The interaction terms have a similar effect as above with tionally, the instrumental variables should not vary systematically with
corruptionregulatory quality reducing the effect of corruption on the the disturbance term in the second-stage equation. The Hansen J-test
LFPR, corruptionthe_shadow_economy, and corruptiontaxes intensifying (Hansen, 1982) is employed to check whether the selected instruments
the adverse effects of corruption on the LFPR, and corruptionwages and satisfy the exclusion restriction. The joint F-statistics from the first stage
corruptionconsumption, reducing the negative effect of corruption on the reject the null that the instruments selected are not relevant. The
LFPR. The Hansen test and the serial correlation test in the system GMM Cragg-Donald Wald F-statistic and the Kleibergen-Paap Wald F-statistic
estimation confirm that the moment's conditions cannot be rejected. test the identification of the equation as a whole. These tests reject the
The correlation between corruption and LFPR appears to be robust to hypothesis that one of the endogenous regressors is under- or weakly
the inclusion of a number of control variables. It is necessary, however, to identified.
ensure that the direction of causality is from corruption to the LFPR. For The main variable of interest, corruption, is once again negative and
this, using the IV estimation could be a useful addition to fixed effects and statistically significant. The results are qualitatively similar to those ob-
system GMM. Table 3 reports IV results. Latitude and the initial level of tained in Tables 1 and 2
corruption are used as instruments for corruption in the IV estimation
(Gupta et al., 2002; Cooray and Schneider, 2016). As a rule of thumb, if 5.3. Further robustness checks
the F-statistics in the models exceed a value of 10, then it suggests in-
strument relevance (Staiger and Stock, 1997). However, this has been Next the estimation in Tables 1–3 are replicated using the ICRG index
found to be an inadequate measure of instrument relevance (Stock et al., due to the shortcomings mentioned in the TI and Kaufmann et al. indices.
2002). In a weak instrument setting, the tests of Cragg-Donald Wald The results are reported in Table 4.
Corruption as measured by the ICRG index has a negative and sta-
tistically significant effect on the LFPR consistent with the results above.
Table 3 The results are in general similar to the results obtained by using the TI
IV estimation. and Kaufmann et al. indices. Thus, it can be concluded that the results are
VARIABLES (1) (2) consistent to the use of the measure of corruption.
LFPR LFPR
The results could be sensitive to the labour market outcome measure
used. Therefore, next, the LFPR is replaced with the EPR to account for
Corruption TI 0.178*** –
this. The results are reported in Table 5. Once again, the results are
(0.065)
Corruption K – 0.160*** consistent with those obtained above, suggesting that the results are
Wages (0.046) robust to the labour outcome measure.
0.328*** 0.311***
(0.129) (0.127)
Interest Rate 0.110 0.021
(0.101) (0.032)
Consumption 0.252** 0.215** Table 4
(0.125) (0.104)
Using the ICRG index.
Taxes 0.370** 0.401*
(0.137) (0.234) VARIABLES (1) (2)
Shadow Economy 0.114** 0.196***
LFPR LFPR
(0.038) (0.030)
Polity Index 0.008* 0.004*** Corruption ICRG 0.110** 0.148**
(0.005) (0.001) (0.046) (0.060)
Regulatory Quality 0.004** 0.002* Wages 0.081* 0.136***
(0.002) (0.001) (0.040) (0.034)
Corruption TI#Reg. Quality 0.160** (0.071) – Interest Rate 0.055 0.146*
Corruption K#Reg. Quality – 0.136* (0.065) (0.082)
(0.083) Consumption 0.166** 0.152***
Corruption TI#Shadow Economy 0.191*** – (0.075) (0.066)
(0.063) Taxes 0.021* 0.018*
Corruption K#Shadow Economy – 0.240*** (0.012) (0.009)
(0.084) Shadow Economy 0.053** 0.124*
Corruption TI#Taxes 0.255*** – (0.026) (0.064)
(0.112) Polity Index 0.004* 0.004
Corruption K#Taxes – 0.227*** (0.002) (0.006)
(0.111) Regulatory Quality 0.002* 0.002*
Corruption TI#Wages 0.018*** – (0.001) (0.001)
(0.006) Corruption ICRG # Reg. Quality – 0.121*
Corruption K#Wages – 0.010* (0.079)
(0.006) Corruption ICRG#Shadow Economy – 0.153*
Corruption TI#Consumption 0.044** – (0.082)
(0.021) Corruption # Taxes – 0.269***
Corruption K#Consumption – 0.034** (0.112)
(0.015) Corruption ICRG#Consumption – 0.065*
F Statistic (first stage) 30.0 34.3 (0.035)
Kleibergen-Paap rk F statistic (first stage) 42.1 455.3 Lag Dependent Variable 0.455*** 0.685***
Cragg-Donald Wald F statistic (first stage) 35.6 32.1 (0.065) (0.188)
Hansen Test for Over-Identifying Restrictions: p Hansen Test for Over-identifying Restrictions: p vale 0.31 0.40
value Arellano-Bond Test For 2nd Order Autocorrelation: p 0.32 0.25
R2 0.59 0.64 value
Observations 185 198 Observations 385 385

Note: Standard errors reported in parentheses. *, **, *** Significant at the 10%, Note: Standard errors reported in parentheses. *, **, *** Significant at the 10%,
5% and 1% levels respectively. 5% and 1% levels respectively.

215
A. Cooray, R. Dzhumashev Economic Modelling 74 (2018) 207–218

Table 5 Table 6
System GMM estimation - using the employment to population ratio. Splitting the sample by corruption level.
VARIABLES (1) (2) (1) (2)

EPR EPR Low Corruption High Corruption

(below mean) (above mean0
Corruption TI 0.152*** –
(0.030) LFPR LFPR
Corruption K – 0.172***
Wages 0.265 0.121
(0.042)
(0.079)*** (0.042)***
Wages 0.189*** 0.163***
Interest Rate 0.140 0.053
(0.060) (0.059)
(0.041)*** (0.059)
Interest Rate 0.064 0.173**
Consumption 0.110 0.218
(0.051) (0.079)
(0.051)** (0.069)***
Consumption 0.124** 0.186**
Taxes 0.135 0.139
(0.055) (0.083)
(0.060)** (0.030)***
Taxes 0.047** 0.060***
Shadow Economy 0.108 0.135
(0.021) (0.020)
(0.067)* (0.032)***
Shadow Economy 0.084** 0.148***
Polity Index 0.008 0.004
(0.043) (0.042)
(0.004)* (0.006)
Polity Index 0.001 0.001
Regulatory Quality 0.008 0.002
(0.003) (0.002)
(0.004)** (0.001)*
Regulatory Quality 0.003 0.002
R-squared 0.75 0.72
(0.003) (0.002)
Observations 148 198
Corruption TI#Reg. Quality 0.012** –
(0.005) Note: Robust standard errors reported in parentheses. ***, **, *.
Corruption K#Reg. Quality – 0.005* Significant at the 1%, 5% and 10% levels respectively.
(0.003)
Corruption TI#Shadow Economy 0.166** –
(0.078)
Corruption K#Shadow Economy – 0.192* on the LFPR in the high-corruption group, consistent with arguments of
(0.094) higher rent seeking and tax burdens in high-corruption countries. Better
Corruption TI#Taxes 0.266** – regulatory quality has a greater statistically significant effect on the LFPR
(0.116)
Corruption K#Taxes – 0.315***
of the low-corruption group.
(0.121)
Corruption TI#Wages 0.015* – 6. Conclusions
(0.007)
Corruption K#Wages – 0.012*
This study considers a theoretical framework that links corruption to
(0.006)
Corruption TI#Consumption 0.030** – labour supply and empirically tests the hypotheses stemming from the
(0.015) theoretical model. The results indicate that corruption has a direct
Corruption K#Consumption – 0.038* negative effect on the LFPR and the employment to population ratio.
(0.021) Corruption also affects labour supply negatively through taxes and the
Lag Dependent Variable 0.628*** 0.662***
shadow economy. The estimated effects are robust to a number of control
(0.114) (0.150)
Hansen Test for Over- Identifying Restrictions: p value 0.35 0.29 variables, alternative measures of corruption and employment, and
Arellano-Bond Test for 2nd Order Autocorrelation: 0.37 0.45 estimation methods. Higher wages and better regulatory quality are
p value found to reduce the negative impacts of corruption on the LFPR and
Observations 207 207
employment to population ratio. However, the negative effect of cor-
Note: Standard errors reported in parentheses. *, **, *** Significant at the 10%, ruption outweighs the positive effects of these variables. These findings
5% and 1% levels respectively. have important policy implications. The results suggest that policy-
makers should be concerned with reducing corruption in order to
enhance labour force participation rates. Reducing corruption will in-
5.4. Splitting the sample crease access to jobs and permit people to take advantage of labour
market opportunities that arise. The results also suggest that improving
Corruption can have a greater adverse effect on the LFPR of countries regulatory quality that reduces the size of the shadow economy and
which are more corrupt than those which are less corrupt. Therefore, the increasing the returns to labour will increase the supply of labour in the
sample is split into two – countries falling below the mean level of cor- formal market. Our findings also highlight the importance of the effi-
ruption of 6.4 on the TI index (low-corruption countries) and above the ciency of the fiscal policy in terms of tax collection and provision of
mean level of corruption (high-corruption countries) – to investigate public goods, as these policies increase the labour supply in the official
whether greater adverse effects on the labour market are witnessed in economy by increasing the productivity of the official economy and
high-corruption countries. Note, the results reported in Table 6 are based indirectly reducing the relative attractiveness of activities in the shadow
on the TI index. Estimation was also carried out using the Kaufmann et al. economy.
and ICRG indices using system GMM. Similar results were obtained using
these two indices. Column (1) of Table 6 reports results for the low- Appendix. Countries used in study:
corruption group, and column (2) reports regression results for the
high-corruption group. Afghanistan, Albania, Algeria, Argentina, Armenia, Australia, Austria,
The results indicate that wages and consumption have a positive Azerbaijan, Bangladesh, Belarus, Belgium, Benin, Bhutan, Bolivia,
impact in both groups of countries, high- and low-corruption countries; Botswana, Brazil, Burkina Faso, Burundi, Cambodia, Cameroon, Canada,
however, a higher impact is observed for the low-corruption group. Central African Republic, Chad, Chile, China, Colombia, Comoros, Congo
Similarly, taxes and the shadow economy have a greater negative effect Dem Rep, Congo Rep, Costa Rica, Cote d’Ivoire, Cuba, Cyprus, Denmark,

216
A. Cooray, R. Dzhumashev
Djbouti, Dominica, Dominican Republic, Ecuador, Egypt, El Salvador,
Eritrea, Ethiopia, France, Finland, Fiji, Gabon, Gambia, Germany, Geor-
gia, Ghana, Greece, Guatemala, Guinea-Bissau, Haiti, Honduras, India,
Indonesia, Iran, Iraq, Ireland, Japan, Hungary, Jordan, Kazakhstan,
Kenya, Kyrgyzstan, Laos, Latvia, Lebanon, Lesotho, Liberia, Libya,
Lithuania, Luxemburg, Macedonia, Madagascar, Malawi, Malaysia,
Maldives, Mauritania, Mauritius, Mexico, Mongolia, Morocco,
Mozambique, Myanmar, Namibia, Nepal, Netherlands, Nicaragua, Niger,
Nigeria, Norway, Pakistan, Panama, Papua New Guinea, Paraguay, Peru,
Philippines, Poland, Russia, Rwanda, Samoa, Senegal, Serbia, Sierra
Leone, Solomon Islands, Somalia, South Africa, Sri Lanka, Sweden,
Switzerland, Sudan, Suriname, Swaziland, Syria, Tajikistan, Turkey,
Turkmenistan, Tuvalu, Uganda, Ukraine, United Kingdom, United States,
Uruguay, Uzbekistan, Vanuatu, Venezuela, Vietnam, Yemen, Zambia,
Zimbabwe.
Appendix A. Supplementary data
Supplementary data related to this article can be found at https://doi.
org/10.1016/j.econmod.2018.05.015.
References
Abed, G., Davoodi, H., 2002. Corruption, structural reforms, and economic performance
in the transition economies. In: Abed, George T., Gupta, Sanjeev (Eds.), Governance,
Corruption, & Economic Performance. International Monetary Fund, Publication
Services, Washington, D.C, pp. 489–537.
Acemoglu, D., Johnson, S., Robinson, J., 2001. The colonial origins of comparative
development: an empirical investigation. Am. Econ. Rev. 91, 1369–1401.
Acemoglu, D., Johnson, S., Robinson, J., 2005. Part a. In: Institutions as a Fundamental
Cause of Economic Growth, Handbook of Economic Growth, vol. 1. Elsevier,
pp. 385–472.
Ahlin, C., Pang, J., 2008. Are financial development and corruption control substitutes in
promoting growth? J. Dev. Econ. 86, 414–433.
Ahn, N., De La Rica, S., 1997. The underground economy in Spain: an alternative to
unemployment? Appl. Econ. 29, 733–743.
Aidt, T.S., Dutta, J., Sena, V., 2008. Governance regimes, corruption and growth: theory
and evidence. J. Comp. Econ. 36 (2), 195–220.
Aidt, T.S., 2009. Corruption, institutions, and economic development. Oxf. Rev. Econ. Pol.
25 (2), 271–291.
Al-Marhubi, F., 2000. Corruption and inflation. Econ. Lett. 66, 199–202.
Arellano, M., Bover, O., 1995. Another look at the instrumental variables estimation of
error components models. J. Econom. 68, 29–51.
Bardhan, P., 1997. Corruption and development: a review of issues. J. Econ. Lit. 35 (3),
1320–1346.
Blackburn, K., Bose, N., Haque, M.E., 2006. The level and persistence of corruption in
economic development. J. Econ. Dynam. Contr. 30, 2447–2467.
Blundell, R., Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel
data models. J. Econom. 87, 115–143.
Brunetti, A., Kisunko, G., Weder, B., 1998. Credibility of rules and economic growth:
evidence from a worldwide survey of the private sector. World Bank Econ. Rev. 12,
353–384.
Campos, E., Lien, D., Pradhan, S., 1999. The impact of corruption on investment:
predictability matters. World Dev. 27, 1059–1067.
Cooney, S., Lindsey, T., Mitchell, R., Zhu, Y., 2002. Labour law and labour market
regulation in East asian states: problems and issues for comparative inquiry. In:
Cooney, S., Lindsey, T., Mitchell, R., Shu, Y. (Eds.), Law and Labour Market
Regulation in East Asia. Routledge, London.
Cooray, A., Schneider, F., 2016. Does corruption promote Emigration? An empirical
examination. J. Popul. Econ. 29, 293–310.
Cooray, A., Dutta, N., Mallick, S., 2017. Trade openness and labor force participation in
Africa: The role of political institutions. Ind. Relat. J. Econ. Soc. 56, 319–350.
Delavallade, C., 2006. Corruption and distribution of public spending in developing
countries. J. Econ. Finance 30 (2), 222–239.
Depken, C., Lafountain, C., 2006. Fiscal consequences of public corruption: empirical
evidence from state bond ratings. Publ. Choice 126, 75–85.
Dreher, A., Schneider, F., 2010. Corruption and the shadow economy: an empirical
analysis. Publ. Choice 144, 215–238.
Dzhumashev, R., 2014a. Corruption and growth: the role of governance, public spending,
and economic development. Econ. Modell. 37, 202–215.
Dzhumashev, R., 2014b. The two-way relationship between government spending and
corruption and its effects on economic growth. Contemp. Econ. Pol. 32, 403–419.
Ferraz, C., Finan, F., Moreira, D.B., 2012. Corrupting learning: evidence from missing
federal education funds in Brazil. J. Publ. Econ. 96, 712–726.
217
Economic Modelling 74 (2018) 207–218
Field, E., 2007. Entitled to work: urban property rights and labour supply in Peru. Q. J.
Econ. 122, 1561–1602.
Foellmi, R., Oechslin, M., 2007. Who gains from coercive corruption? J. Dev. Econ. 82,
95–119.
Frederiksen, A., Graverson, E., Smith, N., 2005. Tax evasion and work in the underground
sector. Lab. Econ. 12, 613–628.
Friedman, E., Johnson, S., Kaufmann, D., Zoido-Lobaton, P., 2000. Dodging the grabbing
hand: the determinants of unofficial activity in 69 countries. J. Publ. Econ. 76,
459–493.
Fugazza, M., Jacques, J., 2003. Labour market institutions, taxation and the underground
economy. J. Publ. Econ. 88, 395–418.
Gonzales, C., Jain-Chandra, S., Kochhar, K., Newiak, M., 2015. Fair Play: More Equal
Laws Boost Female Labour Force Participation. IMF Staff Discussion Note, 02,
February.
Gupta, S., Davoodi, H., Alonso-Terme, R., 2002. Does corruption affect income inequality
and poverty? Econ. Govern. 3, 23–45.
ICRG Report, 2014. ICRG Report on Methodology. ICRG, London. https://www.prsgroup.
com/?pdf_file¼http://www.prsgroup.com/wpcontent/uploads/2012/
icrgmethodology.pdf (retrieved January 2014).
Johnson, S., Kaufmann, D., Shleifer, A., 1997. The unofficial economy in transition.
Brookings Pap. Econ. Activ. 27, 159–239.
Kaufmann, D., Kraay, A., Mastruzzi, M., 2014. World Governance Indicators Project.
World Bank.
Kaufmann, D., 2010. Can Corruption Adversely Affect Public Finances in Industrialized
Countries. Brookings. April 19.
Lacko, M., 2006. Tax Rates with Corruption: Labour Market Effects Empirical Cross
Country Comparisons on OECD Countries. Institute of Economics, Hungarian
Academy of Sciences, 2006/4.
Lambsdorff, J.G., 2003. How corruption affects productivity. Kyklos 56, 457–474.
Lambsdorff, J.G., 2005. How Corruption Affects Economic Development. Transparency
International: Global Corruption Report 2005. Pluto Press, London, pp. 310–312.
Mandal, B., Marjit, S., 2010. Corruption and wage inequality? Int. Rev. Econ. Finance 19
(1), 166–172.
Marjit, S., Acharyya, R., 2006. Trade liberalization, skill-linked intermediate production
and the two-sided wage gap. Journal of Economic Policy Reform 9 (3), 203–217.
Marjit, S., Kar, S., 2005. Emigration and wage inequality. Econ. Lett. 88 (1), 141–145.
Marshall M. and Jaggers K. (2014) http://www.systemicpeace.org/polity/polity14.
htm#nam (retrieved January 2014).
Mauro, P., 1995. Corruption and growth. Q. J. Econ. 110, 681–712.
Mauro, P., 1996. The Effects of Corruption on Growth, Investment, and Government
Expenditure. International Monetary Fund, Washington, D.C. IMF Working Paper
WP/96/98.
Mauro, P., 1998a. Corruption and the composition of government expenditure. J. Publ.
Econ. 69, 263–279.
Mauro, P., 1998b. Corruption: causes, consequences and agenda for future research.
Finance Dev. 35, 11–14.
Mendez, F., Sepulveda, F., 2006. Corruption, growth and political regimes: cross country
evidence. Eur. J. Polit. Econ. 22, 82–98.
Meon, P., Weill, L., 2010. Is corruption an efficient grease? World Dev. 38, 244–259.
Mincer, J., 1962. Labour force participation of married women: a study of labour supply.
In: Aspects of Labour Economics. Princeton University Press. NBER.
Mo, P., 2001. Corruption and economic growth. J. Comp. Econ. 29, 66–79.
Nickell, S., 1997. Unemployment and labour market rigidities: Europe versus North
America. J. Econ. Perspect. 11, 55–74.
Patrinos, H.A., Kagia, R., 2007. Maximizing the performance of education systems: the
case of teacher absenteeism. In: Campos, J.E., Pradhan, S. (Eds.), The Many Faces of
Corruption: Tracking Vulnerabilities at the Sector Level. World Bank, Washington.
Pi, J., Zhou, Y., 2012. Public infrastructure provision and skilled–unskilled wage
inequality in developing countries. Lab. Econ. 19 (6), 881–887.
Pi, J., Zhou, Y., 2013. Institutional quality and skilled-unskilled wage inequality.
Economic Modeling 35, 356–363.
Pi, J., Zhou, Y., Yin, J., 2013. International factor mobility, monopolistic competition, and
wage inequality. Economic Modeling 33, 326–332.
Pritchett, L., Summers, L.H., 1996. Wealthier is healthier. J. Hum. Resour. 31, 841–868.
Rose-Ackerman, S., 1999. Corruption and Government: Causes, Consequences and
Reform. Cambridge University Press, Cambridge.
Round, J., Williams, C., Rodgers, P., 2008. Corruption in the post-soviet workplace: the
experiences of recent graduates in contemporary Ukraine. Work. Employ. Soc. 22,
149–166.
Schneider, F., Buehn, A., Montenegro, C., 2010. Shadow Economies All over the World:
New Estimates for 162 Countries from 1999–2007. World Bank Policy Research
Working Paper No. 5356.
Shleifer, A., Vishny, R., 1993. Corruption. Q. J. Econ. 108 (3), 599–617.
Smarzynska, B., Wei, S.-J., 2000. Corruption and the Composition of Foreign Direct
Investment: Firmlevel Evidence. National Bureau of Economic Research, Cambridge,
MA. NBER Working paper No. 7969.
Smith, P., 1994. Assessing the size of the underground economy: the statistics, Canada
perspectives,” canadian economic observer, catalogue 11-010, 3.16-33, at 3.18.Spiro,
P. S. (1993): "evidence of a post-GST increase in the underground economy. Can. Tax
J./Revue Fiscale Canadienne 41/2, 247–258.
Svensson, J., 2003. Who must pay bribes and how Much? Evidence from A Cross section
of firms. Q. J. Econ. 118 (1), 207–230.
A. Cooray, R. Dzhumashev
Tanzi, V., Davoodi, H., 2002. Corruption, public investment, and growth. In:
Abed, George T., Gupta, Sanjeev (Eds.), Governance, Corruption, and Economic
Performance. International Monetary Fund, Publication Services, Washington, D.C,
pp. 280–299.
Transparency International, 2014. Corruption Perceptions Index. http://www.
transparency.org/cpi2014/results (retrieved January 2014).
Treisman, D., 2007. What have we learnt about the causes of corruption from ten years of
cross-national empirical Research?”. Annu. Rev. Polit. Sci. 10, 211–244.
218
Economic Modelling 74 (2018) 207–218
Uhlenbruck, K., Rodriguez, P., Doh, J., Eden, L., 2006. The impact of corruption on entry
strategy: evidence from telecommunication projects in emerging economies. Organ.
Sci. 17, 402–414.
Walsh, J., 2010. Labour market and corruption issues in Chiang Rai, Thailand. Review of
Economic and Business Studies 3, 253–268.
Wei, S., 2000. How taxing is corruption on international investors? Rev. Econ. Stat. 82,
1–11.
Wei, S., 2001. Corruption in Economic Transition and Development. UNECE Spring
Seminar, Geneva.