1 Introduction

The shadow economy has attracted a great attention from policy-makers and scholars because its presence distorts the allocation of resources, alters income distribution and reduces governments’ tax revenue (Alm and Embaye 2013). Moreover, it is biased to evaluate the outcome of various economic policies without considering the shadow economy since national accounts fail to capture economic activities in the shadow economy. Thus, governments—especially in developing countries—are always trying to control the size of shadow economy. According to the Voluntarism school of thought on shadow economy, the rising burden of taxes is one of the main causes for the increase of shadow economy. The higher taxes make the after-tax earnings smaller, leading to stronger incentives for people to work underground to reduce the tax wedge (Feige 1989; Tanzi 1982, 1999; Giles 1999; Schneider 2007, 2010). Therefore, fiscal policies with the two tools of taxation and government expenditure can contribute not only to economic growth but also to the control of the shadow economy. Although taxation is examined as a determinant of the shadow economy, the impact of government expenditure on the shadow economy is still largely ignored in the literature. In addition, the effectiveness of fiscal policies is affected by the level of corruption (Gauthier and Goyette 2016) while corruption is also a determinant of shadow economy (Johnson et al. 1997; Hindriks et al. 1999). The implications are how fiscal policies affect the shadow economy and how this effect depends on the presence of corruption.

Developing Asian countries provide a fruitful context to study the above problems for their rising thorny features such as the large shadow economy size and the high corruption level. The estimated average size of shadow economy in 24 selected Asian countries over the period of 2002–2015 is 33.88% of official gross domestic products (GDP), and this period experiences an increase of 11.24% in the shadow economy size (Medina and Schneider 2018). The lowest and the highest sizes are 30.77% and 37.50% in 2006 and 2013, respectively. Meanwhile, the average index for control of corruption in 24 Asian countries over the period 2002–2015 fluctuates around − 0.7 and − 0.5 per year (Worldwide Governance Indicators, 2017). The control of corruption index is scaled from − 2.5 (the most corruption level) to + 2.5 (the least corruption level). As a result, Asian countries face challenges of various degrees of fiscal policy implementation to deal with the large shadow economy size amid the high corruption level.

This paper examines the impact of fiscal policies on the shadow economy through the two tools of taxation and government expenditure, taking corruption into consideration, in 24 developing Asian countries over the period of 2002–2015. For the first time in the literature, we investigate how government expenditure and the interaction between corruption and the government behavior influence the shadow economy. The results from estimation by Feasible Generalized Least Squares (FGLS) and two-step Generalized Method of Moments (SGMM) contribute to the literature in two aspects. First, we extend the Voluntarist school of thought on shadow economy by using the fiscal policy approach: the shadow economy is affected not only by taxation but also by government expenditure. We argue that government expenditure can reduce the shadow economy because of its contribution to the growth of the official economy, the poverty reduction and public goods improvement. In addition, we found a stronger impact of indirect taxes, compared to direct taxes, on the shadow economy in developing countries, contributing to the Voluntarism in terms of tax structure impact on the shadow economy. Second, the presence of corruption attenuates the impact of fiscal policy on shadow economy for two reasons: (i) tax collectors may get bribes from taxpayers in exchange for ignoring underground production activities and (ii) the effectiveness of government spending is dampened by corruption through public projects, diminishing the indirect effect of government expenditure on shadow economy through channels of economic growth and public services. These findings imply that government can utilize expansionary fiscal policies (to cut taxation or/and to increase government spending) to reduce the shadow economy as long as corruption is controlled.

The remainder of this study is organized as follows. Section 2 reviews the literature. Section 3 describes models, data and methodology. Empirical results and discussion are provided in Sect. 4. Section 5 concludes and offers main policy implications.

2 Literature review and hypotheses

2.1 Shadow economy

2.1.1 Definitions

The “shadow economy” named by Schneider and Dominik (2000) has been known with various labels such as the underground economy (Simon and Witte 1982; Feige 1989), the informal economy (Smith 1985) and the unofficial economy (Johnson et al. 1998).

It is an integral part of the official economy. There is not any adequate and coherent definition on the shadow economy in the literature. However, most scholars agree that the shadow economy consists of “all market-based goods and services that escape inclusion in official accounts” (Alm and Embaye 2013). Schneider (2007, 2010) defines the shadow economy as including “all market-based legal production of goods and services that are deliberately concealed from public authorities to avoid payment of income, value-added or other taxes; to avoid payment of social security contributions; to avoid having to meet certain legal labour market standards, such as minimum wages, maximum working hours, safety standards, etc.; and to avoid complying with certain administrative procedures, such as completing statistical questionnaires or other administrative forms”. This definition of the shadow economy is also used in this research due to its specific features.

2.1.2 Measurements

On measurement, there are three major approaches widely used to assess the size and development of the shadow economy: (i) direct methods that are based on microeconomic theories to employ well-designed surveys or tax auditing methods (Isachsen and Strom 1985; Mogensen and Pedersen 1995; Haigner et al. 2013); (ii) indirect methods that are based on macroeconomic indicators such as the currency demand (Tanzi 1983, 1999; Alm and Embaye 2013), the discrepancy between national income and expenditure statistics (Thomas 1999), the discrepancy between the official and actual labor force (Contini 1981); and (iii) the model methods (known as the MIMIC approach—Multiple Indicators Multiple Causes) in which the shadow economy can be measured through a set of causes and indicators of the shadow economy (Giles and Tedds 2002; Bajada and Schneider 2005; Buehn and Schneider 2012; Vo and Pham 2014; Schneider and Buehn 2017; Medina and Schneider 2018). Many causes of the shadow economy distinguished in these studies include tax and social contribution burdens, self-employment, institutional quality and tax morality. The size of the shadow economy is reflected in various indicators such as the official economic growth, currency/cash outside banks and labor force participation rate. In this research, we use data on shadow economies from Medina and Schneider (2018), in which the MIMIC approach was used for shadow economy estimation.

2.2 The impact of fiscal policy and corruption on shadow economy

2.2.1 Fiscal policy and shadow economy

According to Keynesian theory, the government can use fiscal policy by changing levels of taxation or/and government expenditure to stabilize the economy through affecting the aggregate demand and economic activity levels (O’Sullivan and Steven 2003). Meanwhile, Voluntarists contend that tax burden is one of the main causes of shadow economy. People and businesses have stronger incentives to work in the shadow economy as a choice to avoid paying high taxes (Feige 1989; Tanzi 1982, 1999; Giles 1999; Schneider 2007, 2010). Thus, an expansionary fiscal policy conducted by cutting taxes—the first tool of fiscal policy—will expectedly reduce the size of shadow economy. However, the impacts of different taxes on the shadow economy still receive little investigation. In this study, we categorize taxes into two groups: direct tax and indirect tax, and examine their impacts on the shadow economy separately. On average, the direct to indirect tax ratio is 0.745 in 24 Asian countries over the period 2002–2015 (our calculations based on data from WDI, WB). It seems easier for governments in developing countries to increase their revenues through indirect taxes than through direct taxes, since indirect taxes have much wider coverage as compared to direct taxes in which the rich and large corporations have tools to hide their incomes and profits. However, this taxation policy can possibly obtain equality but not equity for the so-called regressive nature of indirect taxes when the rich and the poor share the same burden for indirect taxes (such as VAT, customs duty…). Indirect taxes may also lead to market distortions since governments can impose on certain industries but not others. Moreover, the trend of basing on indirect taxes to generate public revenue may create conditions for the growth of the shadow economy through corruption and indirect tax evasion by the rich and multinational corporations.

The second tool of fiscal policy is government spending. Nevertheless, the linkage between government spending and shadow economy is largely ignored in the literature except some hints. We argue that government spending can reduce the size of shadow economy for three possible reasons. First, the government expenditure reallocates scarce resources to the official economy, reducing resources in the unofficial sector. Second, the government expenditure boosts economic growth through aggregate demand (Kolluri et al. 2000; Bose et al. 2007; Wu et al. 2010), and the growth in the official economy reduces the shadow economy since Dualists view the shadow economy as a residual or a by-product of the official economy and economic growth is the cure for the shadow economy (La Porta and Shleifer 2008, 2014; Williams 2008). Third, government spending may improve the quality of public services which in turn discourages people from working in the shadow economy because, according to the Legalist school of thought, the high quality of public services reduces transaction costs and increases benefits of being in the official economy (Johnson et al. 1998; Friedman et al. 2000; Dreher and Schneider 2010).

In other words, we hypothesize that:

Hypothesis 1

The shadow economy size is negatively affected by expansionary fiscal policy and positively affected by contractionary fiscal policy, ceteris paribus.

2.2.2 Corruption and shadow economy

Most researchers such as Johnson et al. (1997), Hindriks et al. (1999), Hibbs and Piculescu (2005) and Dreher and Schneider (2010) have a common conclusion from their empirical studies that corruption and the shadow economy are complements. To these scholars, corruption increases the size of shadow economy because: (i) corruption is viewed as a particular form of regulations and institutional quality which are drivers of shadow economy (Johnson et al. 1997), (ii) inspectors collude with taxpayers for getting bribes in exchange of underreporting the tax liability of the taxpayers (Hindriks et al. 1999), and (iii) businesses bribe dishonest bureaucrats for ignoring their unofficial production activities (Hibbs and Piculescu 2005). Thus, we predict that:

Hypothesis 2

Corruption has a positive impact on the shadow economy, ceteris paribus.

2.2.3 The effect of fiscal policy on shadow economy in the presence of corruption

Corruption can attenuate the impact of fiscal policy on shadow economy in two ways as follows. First, tax collectors get bribes in exchange for underreporting the tax obligations of taxpayers or ignoring their unofficial production activities, and this collusion unexpectedly weakens the effect of fiscal policy on the shadow economy. For example, the government carries out an expansionary fiscal policy by tax cuts to reduce the shadow economy. However, firms and people still pay bribes if the bribe amounts are smaller than their tax liability plus the red tape cost or if benefits of working underground are higher than bribe amounts. As a result, they continue operating unofficially and the shadow economy cannot be reduced. Second, corruption drains the government’s purse through public projects and reduces the effectiveness of government spending, diminishing the indirect effect of government expenditure on shadow economy through channels of economic growth and public services improvement. Therefore, we postulate that:

Hypothesis 3

Corruption intensifies the impact of direct and indirect taxes on increasing the shadow economy and attenuates the impact of government expenditure on reducing the shadow economy, ceteris paribus.

3 Models, data and methodology

3.1 Models

Based on the Voluntarism school of thought on shadow economy, previous studies and the above arguments, we examine the impact of fiscal policy on the shadow economy, taking the presence of corruption into consideration by proposing the model as follows:

$$ {\text{SHADOW}}_{it} = a_{0} + a_{1} X_{it} + a_{2} {\text{COR}}_{it} + a_{3} X_{it} *{\text{COR}}_{it} + \, Z_{it}^{\prime } a_{j} + \varepsilon_{it} $$
(1)

where t: time; i: country; a1, a2, a3, and aj: respective coefficients; and ε: error terms.

The dependent variable is the size of the shadow economy (SHADOW), measured by the percentage of gross domestic product (GDP). The regressors are fiscal policy (X) and corruption (COR). X*COR is the interaction of fiscal policy and corruption. Z is a vector of control variables. Since we measure fiscal policy by two tools that are taxation (both direct and indirect taxes) and government expenditure, model 1 is estimated in the form of two equations below:

$$ {\text{SHADOW}}_{it} = \alpha_{0} + \alpha_{1} {\text{TAX}}\_{\text{DI}}_{it} + \alpha_{2} {\text{TAX}}\_{\text{INDI}}_{it} + \alpha_{3} {\text{COR}}_{it} + \alpha_{4} {\text{TAX}}\_{\text{DI}}_{it} *{\text{COR}}_{it} + \alpha_{5} {\text{TAX}}\_{\text{INDI}}_{it} *{\text{COR}}_{it} + Z_{it}^{\prime } \alpha_{j} + u_{it} $$
(2)
$$ {\text{SHADOW}}_{it} = \beta_{0} + \beta_{1} {\text{EXP}}_{it} + \beta_{2} {\text{COR}}_{it} + \beta_{3} {\text{EXP}}_{it} *{\text{COR}}_{it} + Z_{it}^{\prime } \beta_{j} + v_{it} $$
(3)

where α, β: respective coefficients; u, v: error terms.

In Eq. (2), fiscal policy is represented by the tool of taxation in which we consider both direct taxes (TAX_DI) and indirect taxes (TAX_INDI). Both of them are measured by revenue as the percentage of GDP. In Eq. (3), government expenditure per GDP (EXP) is employed to measure fiscal policy.

The interaction terms between direct tax revenue and corruption (TAX_DI*COR) as well as between indirect tax revenue and corruption (TAX_INDI*COR) measure the extent to which direct and indirect tax revenues affect the shadow economy with the presence of corruption. By taking turns to make a partial derivative of TAX_DI or TAX_INDI in Eq. (2), we get the total effects of TAX_DI or TAX_INDI on shadow economy at the margin, respectively:

$$ \frac{{\partial ({\text{SHADOW}}_{it} )}}{{\partial \left( {{\text{TAX}}\_{\text{DI}}_{it} } \right)}} = \alpha_{1} + \alpha_{4 } {\text{COR}}_{it} $$
(4)
$$ \frac{{\partial ({\text{SHADOW}}_{it} )}}{{\partial \left( {{\text{TAX}}\_{\text{INDI}}_{it} } \right)}} = \alpha_{2} + \alpha_{5 } {\text{COR}}_{it} $$
(5)

Similarly, we get the total impact of EXP on shadow economy at the margin by making a partial derivative of EXP in Eq. (3):

$$ \frac{{\partial ({\text{SHADOW}}_{it} )}}{{\partial \left( {{\text{EXP}}_{it} } \right)}} = \beta_{1} + \beta_{3 } {\text{COR}}_{it} $$
(6)

The coefficient pairs of interest are α1 and α4, α2 and α5, as well as β1 and β3 in Eqs. (4), (5) and (6), respectively. Because we hypothesize that direct and indirect taxes can increase the shadow economy, and the increase in corruption intensifies this impact, we expect that α1, α2, α4 and α3 are all significantly positive. With the hypothesis that government expenditure reduces shadow economy and the increase in corruption attenuates this effect, we anticipate that β1 is significantly negative and β3 positive.

Control variables (Z) consist of the burden of government regulations (BURDEN), GDP growth (GDPG), retirement (RETIRE), unemployment (U_RATE) and urbanization (URBAN). The selection of these control variables is based on previous studies and arguments as follows.

The burden of government regulations Legalists debate that burdens of regulations and procedures induce the rise of shadow economy. For example, over-regulations on labor market (minimum wages, limited official working hours) or trade barriers (import quotas) increase labor costs and structural unemployment and reduce individuals’ choice in the official economy. This leads to higher motivation to work underground. The conclusion that heavily regulated burden causes larger shadow economy is found from various studies including those by Loayza (1997), Friedman et al. (2000) and Schneider et al. (2010).

GDP growth The effect of official economy on informal sector is ambiguous (Schneider and Bajada 2003; Schneider and Dominik 2013; Vo and Pham 2014). According to the Dualist approach, the economic growth negatively affects shadow economy because dualists view the shadow economy as a residual or a by-product of the official economy and economic growth is the cure for the shadow economy (La Porta and Shleifer 2008, 2014; Williams 2008). Nevertheless, Greenfield (1993) sees the development of the informal and formal sectors in a parallel way, with the notion that the direct and indirect demand for goods and services produced in the informal sector will increase its size as the formal economy expands. Elgin and Oztunali (2014) explore that higher GDP per capita is correlated with the larger informal sector in countries with low institutional quality.

Retirement: When people retire, they cannot work officially but they may participate in the shadow economy. Therefore, the shadow economy growth is attributed to the increase of retirement (Dell’Anno et al. 2007; Williams et al. 2016).

Unemployment: Most schools of thought on informal economy presume that unemployment has a positive impact on the shadow economy because high unemployment gives individuals more incentive to work unofficially to earn a living. This significant positive relationship is confirmed by Boeri and Garibaldi (2002) and Dell’Anno and Solomon (2008).

Urbanization According to the Dualists, urbanization is positively related to the shadow economy as the shadow economy appears in the process of industrialization and urbanization when migrants move from rural sector to urban sector (Safa 1986). However, Elgin and Oyvat (2013) argue that there is an inverted-U relationship between the level of urbanization and the share of shadow economy. In their interpretation, the share of the shadow economy rises in the early phase of urbanization due to several pull and push factors, but it tends to fall in the latter phase when the impact of these pull and push factors is reduced as a natural result of rural dwellers getting wealthier.

3.2 Data

All data in Eqs. (1) and (2) are extracted from various sources, including Medina & Schneider (2018); World Development Indicators (WDI) from World Bank (2017a) and Worldwide Governance Indicators (WGI) from World Bank (2017b); and Global Competitiveness Index (GCI) from World Economic Forum (WEF). Especially, we use the component “Control of Corruption” in WGI as a proxy for corruption because this is viewed as a particular factor of institutional quality which is the crucial driver of shadow economy (Johnson et al. 1997). “Control of Corruption” is defined as “the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as capture of the state by elites and private interests” (World Bank 2017b). Besides, the component “Burden of government regulations” in GCI is employed to measure how burdensome it is “to comply with governmental administrative requirements” which reflect the possibility that people and businesses can work underground because of higher costs to be official (Schneider et al. 2010).

We collect the data for 24 Asian developing countries—Bangladesh, Bhutan, Cambodia, China, India, Indonesia, Iran, Jordan, Kazakhstan, Kyrgyzstan, Laos, Lebanon, Malaysia, Maldives, Mongolia, Myanmar, Nepal, Pakistan, Philippines, Sri Lanka, Tajikistan, Thailand, Vietnam and Yemen—over the period of 2002–2015. Measurements, expected signs and sources of all variables are described in Table 1.

Table 1 Measurements, expected signs and sources of all variables
Full size table

Table 2 provides the summary statistics for the variables.

Table 2 Summary statistics
Full size table

The selection of 24 Asian countries is owed to two reasons. First, all of these countries are middle income countries (lower middle income with GNI per capita between US$1006 and US$3955; or upper middle income with GNI per capita between US$3956 and US$12,235), defined by the World Bank (2018). Second, the selection is also due to data constraints. The data for shadow economy, taxes and corruption must be sufficient for econometric analysis. Studies based on cross-country analysis do not eliminate differences in many aspects such as culture, population and correspondingly render the results dubious in nature because of data comparability and heterogeneity problems (Atkinson and Brandolini 2001). However, the selection of 24 Asian countries is our best choice with data availability and possible minimization of heterogeneity problems.

3.3 Econometric methodology

To estimate the two models (1) and (2), the Feasible Generalized Least Squares (FGLS) is conducted for correcting the presence of autocorrelation within panels and heteroskedasticity across panels (Greene 2012). Some authors argue that lagged levels of shadow economy can affect its current value (Berdiev et al. 2015). Thus, the dynamic forms of models (1) and (2) are also examined, being estimated by two-step Generalized Method of Moments (SGMM) with robustness check, in which lagged differences and levels of the dependent variable are used as instrumental variables. SGMM estimates linear dynamic panel-data models and copes with the issues of endogeneity, heteroskedasticity and autocorrelation (Blundell and Bond 1998). Two types of tests will be used for the empirical model (Arellano and Bond 1991). First, the Sargan test checks the validity of instruments and specifications. Second, the Arellano and Bond test examines the hypothesis that the residuals from the estimations are first-order correlated (AR1) but not second-order correlated (AR2).

4 Results and discussion

The empirical results and associated test statistics for Eq. (2) are provided in Table 3 for FGLS and SGMM, with three specifications. In the specification (1.1), the impacts of direct tax, indirect tax and corruption on shadow economy are estimated. The interaction terms between taxation (both direct and indirect taxes) and corruption are added in the specification (1.2). The specification (1.3) includes control variables.

Table 3 Regression results for Eq. (2)
Full size table

Similarly, the empirical results for Eq. (3) are presented in Table 4 with three specifications. The impact of government expenditure and corruption on shadow economy is examined in the specification (2.1). The specification (2.2) includes the interaction term between government expenditure and corruption. Then, control variables are added in the specification (2.3).

After SGMM estimation, two types of tests are used (Arellano and Bond 1991). First, the Sargan tests with high P values for all specifications fail to reject the null hypothesis of the validity of the over-identifying restriction, indicating that instruments are appropriate and specifications are well determined. Second, the results from Arellano-Bond tests show that the null hypothesis of no first-order serial correlation in the residuals (AR1) is rejected and that the null hypothesis of no second-order serial correlation in the residuals (AR2) is not rejected at 1% level of significance for all specifications.

The results in Tables 3 and 4 show interesting findings as follows:

Table 4 Regression results for Eq. (3)
Full size table

First, the shadow economy is positively affected by direct and indirect taxes and negatively affected by government expenditure, confirming the first hypothesis that an expansionary fiscal policy can reduce the shadow economy but a contractionary fiscal policy increases it. Direct and indirect taxes have positive impact on the shadow economy in all specifications of Eq. (2) by both estimations of FGLS and SGMM, and there is a stronger impact of indirect taxes, compared to direct taxes, on the shadow economy in developing countries. When the ratio of direct taxes in GDP increases by 1%, the shadow economy measured as the percentage of GDP increases by 0.158–0.377%, while 1% increase in indirect taxes to GDP ratio leads to the rise in the shadow economy by 0.287–0.517%. This finding supports the conventional view that tax burden is one of the main causes of shadow economy (Feige 1989; Tanzi 1982, 1999; Giles 1999; Schneider 2007, 2010), and it is in line with the prediction that the trend of basing on indirect taxes to generate public revenue may create conditions for the shadow economy growth through corruptions and through indirect tax evasion by the rich and multi-corporations in developing countries.

On the other hand, government expenditure negatively affects the shadow economy in all specifications of Eq. (3), by both FGLS and SGMM. Apparently, 1% rise in government expenditure leads to a fall in shadow economy by around 0.232–0.496%. This finding fills the gap on the linkage between government spending and shadow economy in the literature. The finding that government spending is a tool to reduce the shadow economy can be supported by three possible reasons: (i) the government expenditure reallocates scarce resources to the official economy, reducing resources in the unofficial sector, (ii) the government spending reduces the shadow economy through boosting economic growth; and (iii) the government expenditure may improve the quality of public services and this improvement in turn discourages people from working in the shadow economy.

Second, corruption has a positive effect on shadow economy in all specifications of Eqs. (2) and (3), failing to reject the second hypothesis. In comparison with direct and indirect taxes and government expenditure, corruption has the strongest impact on the shadow economy. In particular, one unit rise in corruption increases the shadow economy by 0.470–3.701%. This confirms the postulation that corruption and the shadow economy are complements in the literature (Johnson et al. 1997; Hindriks et al. 1999; Hibbs and Piculescu 2005; Dreher and Schneider 2010).

Third, the third hypothesis that corruption intensifies the positive impact of direct and indirect taxes on the shadow economy and attenuates the negative impact of government expenditure on the shadow economy is affirmed.

On the one hand, the interaction effects of direct taxes and corruption (α4) as well as of indirect taxes and corruption (α5) on the shadow economy are significantly positive, specifying that the tax burden increases the shadow economy and the increase in corruption intensifies this effect. The total effect of direct taxes on shadow economy, as shown in Eq. (4), is derived from: (i) the direct effect on shadow economy (α1) and (ii) the indirect effect on shadow economy in the presence of corruption (α4*COR). Similarly, the total effect of indirect taxes on shadow economy, as displayed in Eq. (5), is the sum of α2 and (α5*COR). Thus, when corruption is taken into account, the impacts of direct and indirect taxes on shadow economy are strengthened.

On the other hand, the positive interaction effect of government expenditure and corruption on shadow economy (β3) shows that government expenditure reduces the shadow economy size and the increase in corruption attenuates this effect. The total impact of government expenditure on the shadow economy, as illustrated in Eq. (6), is the sum of its direct impact on the shadow economy (β1) and its indirect impact in the presence of corruption (β3*COR). By this way, in the presence of corruption, when the ratio of government expenditure in GDP rises by 1%, the shadow economy just declines by 0.104–0.192%, compared with 0.232–0.496% of its direct impact on the shadow economy without the interaction with corruption. This finding demonstrates that corruption reduces the negative effect of expenditure on shadow economy.

In the presence of interaction with corruption (COR), the total impacts of direct taxes (α1 + α4*COR), indirect taxes (α2 + α5*COR) and government expenditure (β1 + β3*COR) on shadow economy in 24 Asian countries are presented in Table 5 (see in Appendix, as additional online material). It can be seen that in countries with low levels of corruption (such as Bhutan, Jordan and Malaysia), the total impact of taxes in increasing the shadow economy is weak, but the total impact of government expenditure in reducing the shadow economy is strong. Conversely, countries with high corruption levels (such as Cambodia, Kyrgyz Republic, Myanmar, Tajikistan and Yemen) endure the strong total impact of taxes in increasing the shadow economy, but the weak total impact of government expenditure in reducing the shadow economy.

Fourth, determinants that have positive impacts on the shadow economy size consist of the burden of government regulations, retirement and unemployment as asserted in previous studies. However, the official economy and urbanization are negatively associated with the shadow economy in Asian developing countries. These are advocated by Dualism with the postulation that official economy and shadow economy are substitutes, and by Elgin and Oyvat (2013) in the sense that there is an inverted-U relationship between the level of urbanization and the share of shadow economy, and Asia is currently in the latter phase of urbanization when the critical value of urbanization is exceeded.

4.1 Robustness checks

To overcome identification problems of using tax, government burden and unemployment as regressors for the shadow economy estimated by MIMIC approach (Medina and Schneider 2018), we use the shadow economy estimated by a two-sector (official and the shadow economies) dynamic general equilibrium model (Elgin and Oztunali 2012) for regression in Eqs. 2 and 3, and we obtain consistent results. The robust estimation results are provided in Appendix as additional online material (Tables 6 and 7). In addition, to minimize the heterogeneity in the type of countries, if we drop some countries in civil war (such as Lebanon and Yemen) and some centrally planned economies (China, Laos and Mongolia), the estimation results remain stable.Footnote 1

5 Conclusion

Controlling the size of shadow economy is one of the main concerns for policy-makers, especially those in developing Asian countries, which experience high levels of shadow economy and corruption. This paper empirically examines how fiscal policy affects the shadow economy through the two tools of taxation and government expenditure, and how this effect depends on the presence of corruption.

By using a panel data for 24 developing Asian countries over the period of 2002–2015, we find that the shadow economy is negatively affected by expansionary fiscal policies and positively affected by contractionary fiscal policies; and corruption and shadow economy are complements. In particular, the tax burden from both direct and indirect taxes increases the shadow economy and the increase in corruption intensifies this effect. Furthermore, the government expenditure reduces the shadow economy size and the increase in corruption attenuates this effect. Besides, there is evidence for a stronger impact of indirect taxes, compared to direct taxes; a stronger effect of government expenditure, in comparison with direct and indirect taxes; and a dominating impact of corruption on the shadow economy in developing countries. Other determinants of shadow economy in developing Asian countries are also found such as the burden of government regulations, economic growth, retirement, unemployment and urbanization.

The main findings from the study indicate that governments can utilize expansionary fiscal policies to reduce the shadow economy, provided that corruption is controlled. Since the impact of government expenditure on the shadow economy is stronger than that of taxes on the shadow economy, as found in this paper, governments should give priority on public expenditure rather than taxes when they aim to reduce the shadow economy by implementing fiscal policies. Moreover, the finding of a stronger impact of indirect taxes—compared to direct taxes—on the shadow economy suggests that governments in developing countries should consider appropriate weight of direct and indirect taxes in their tax revenue to help reduce the shadow economy more effectively. However, corruption control is definitely not to be missed because of its direct dominating impact on the shadow economy and indirect impact on the shadow economy through fiscal policy tools including direct taxes, indirect taxes and government expenditure. This also implies that corruption control is one of the key conditions that international financial organizations such as the World Bank, the International Monetary Fund and the Asian Development Bank should consider for their funding agreements. Besides, other supporting policies to reduce the shadow economy should focus on lessening the burden of government regulations, and unemployment as well as enhancing economic growth and urbanization.