Corruption, Taxes and Compliance1
Author:
Ms. Anja Baum https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Ms. Anja Baum in
Current site
Google Scholar
Close
,
Mr. Sanjeev Gupta
Search for other papers by Mr. Sanjeev Gupta in
Current site
Google Scholar
Close
,
Elijah Kimani
Search for other papers by Elijah Kimani in
Current site
Google Scholar
Close
, and
Mr. Sampawende J Tapsoba
Search for other papers by Mr. Sampawende J Tapsoba in
Current site
Google Scholar
Close

Contributor Notes

Author’s E-Mail Address: abaum@imf.org; sgupta@imf.org; stapsoba@imf.org

This paper revisits the effects of corruption on the state’s capacity to raise revenue, building on the existing empirical literature using new and more disaggregated data. We use a comprehensive dataset for 147 countries spanning 1995-2014, compiled by the IMF. It finds that—consistent with the existing literature—corruption is negatively associated with overall tax revenue, and most of its components. This relationship is predominantly influenced by the way corruption interacts with tax compliance. The establishment of large taxpayer offices improves tax compliance by dampening the perception of corruption, thereby boosting revenue.

Abstract

This paper revisits the effects of corruption on the state’s capacity to raise revenue, building on the existing empirical literature using new and more disaggregated data. We use a comprehensive dataset for 147 countries spanning 1995-2014, compiled by the IMF. It finds that—consistent with the existing literature—corruption is negatively associated with overall tax revenue, and most of its components. This relationship is predominantly influenced by the way corruption interacts with tax compliance. The establishment of large taxpayer offices improves tax compliance by dampening the perception of corruption, thereby boosting revenue.

I. Introduction

States that do not raise adequate revenues are unable to build institutions to support economic development (Besley and Persson, 2014). In a recent study, Gaspar, Jaramillo, and Wingender (2016) show that once the tax-to-GDP ratio reaches around 15 percent, GDP per capita increases sharply, a threshold that several developing countries have not yet crossed. Determinants for why revenue mobilization might be lagging are many, and the literature repeatedly and increasingly identifies corruption as one of the factors.

In this paper, we revisit the effects of corruption on the state’s capacity to raise revenue, building on the existing empirical literature with new and more disaggregated data. We use a comprehensive dataset for 147 countries spanning 1995–2014, compiled by the IMF. The data are disaggregated by revenue type: taxes on personal and corporate income, social security, property, goods and services, and international trade. The robustness of these results is then tested against another data source. In this regard, this paper is the first to make use of data generated by two IMF tools to benchmark countries’ tax performance, the Revenue Administration–Fiscal Information Tool (RA-FIT) and to assess actual tax collection vis-a-vis the potential, the IMF’s Revenue Administration Gap Program (RA-GAP).

Another novelty of this paper is the analysis of the impact of establishing public institutions designed to improve tax compliance. For the first time, a comprehensive dataset for the establishment of “specialized” taxpayer offices—both large and medium-sized—and anti-corruption agencies is compiled. It has been argued that establishing such institutions should be part of an anti-corruption strategy (IMF 2016).

We find that, consistent with the existing literature, corruption is negatively associated with overall tax revenue, and most of its components. This relationship is predominantly influenced by the way corruption interacts with tax compliance. Building on this relationship, we assess the possibility of improving tax compliance via strengthening institutions, specifically those focusing on taxpayer segmentation. When properly implemented, segmentation by taxpayers could lead to better focus on managing the unique compliance risks presented by the respective segments, thereby reducing the perception of corruption and increasing taxpayer compliance. Our results indicate that this is indeed the case.

The rest of the paper is structured as follows. Section II briefly reviews the existing literature. Section III describes a series of key stylized facts. The dataset is presented in the Section IV. The empirical methodology and the results are discussed in Sections V and VI. Section VII concludes.

II. Literature Review

During the previous decades, there has been a prolific literature—both theoretical and empirical—on the relationship between corruption and taxation. It finds: First, widespread corruption harms the culture of compliance, thereby increasing tax evasion (Aghion et al., 2016). Tax exemptions perceived to be the result of a bribe undermine the trust in government and the compliance with tax laws (Dreher and Herzfeld 2005); Second, corruption fosters the development of the informal sector, and therefore erodes the potential tax base (Schneider, 2005, Schneider and Denste, 2000).

Many papers have studied the direct influence of corruption on revenue performance (see Table 1), and most find a negative relationship between corruption and tax revenue. Sen Gupta (2007) finds that, among other factors (such as per capita GDP, the ratio of agriculture to GDP, trade openness and foreign aid), corruption is a significant determinant of a country’s revenue performance. More recently, Besley and Persson (2014) discuss why developing countries tax so little, examining considerations such as the economic structure of these economies, political factors (including strength of institutions, fragmented polities, and a lack of transparency due to weak news media), and sociological and cultural influences (such as a weak sense of national identity and a poor norm for compliance). The authors find a strong negative correlation between corruption and the tax revenue-to-GDP ratio. They attribute this correlation to corrupt systems of government that face resistance to increasing taxes.

Table 1.

Literature Summary of Panel Estimation Techniques for the Impact of Corruption on Public Revenue

article image
Impact: (nI) no significant impact; (+) significantly positive; (-) significantly negative; (/) depends on the specification/ambiguous

However, due to data limitations, few empirical studies have considered the effect of corruption on different tax types. These studies suggest that indirect taxes requiring frequent interactions between tax authorities and individuals are more prone to corruption (Hwang 2002, Attila et al. 2008, Thornton 2008, Imam and Jacobs 2014). For example, taxes on international trade, which are an important source of revenue for developing countries, are negatively and statistically significantly affected by corruption. Furthermore, corruption tends to increase the share of international trade taxes to total tax revenues. Similarly, the impact of corruption on taxation of goods and services (such as VAT and excise) is found to be statistically significant (Attila et al., 2008, Thornton, 2008, and Imam and Jacobs, 2014). Estimates for direct taxes are mixed: both personal and corporate income taxes are found to be insignificantly impacted by corruption in Attila et al. (2008) and Thornton (2008), and only Imam and Jacobs (2014) find a negative impact on corporate taxes in the Middle East. Though these papers attempt to solve the endogeneity issue (see the empirical section), they remain focused on regions where tax collection has historically been weaker.

Much of the negative correlation between tax revenue and corruption comes from the effect corruption has on tax compliance, which constitutes the second focus of this paper. A revenue administration reform that aims at compliance improvements is the establishment of specialized taxpayer offices, focusing on taxpayers with different characteristics and different risks to revenue. Such segmentation, primarily but not only by size, enables the distinct compliance risks they pose to be best addressed. For example, IMF (2015) lists effective monitoring of large taxpayers as one way to address compliance deficits. This specialization may be reinforced by segmentation into economic sector to improve knowledge of personnel on sector-specific issues (e.g., for extractive industries, financial institutions, or telecommunications companies). By building specialized tax offices with the specialist skills to service distinct taxpayer segments, compliance and service can be enhanced, perceptions of corruption reduced and tax yield raised. Focusing on the largest taxpayers can be an important signal to the taxpaying public of the government’s commitment to enforcing the tax laws, thereby lowering the perception of corruption.

There is some anecdotal evidence that large taxpayer offices (LTOs) improve compliance, and thus increase tax revenue (Imam and Jacobs, 2014). Ebrill et al. (2001) show that countries with LTOs have generally performed better in terms of VAT revenue collection than those without one. Ebeke et al. (2016), on the other hand, find that LTOs have no significant direct impact on tax revenues, but are a pre-condition for other revenue increasing tax policy and administrative measures, such as VAT. They employ a propensity score matching (PSM) method on a group of sub-Sahara African countries. We build on the PSM methodology, extending the LTO dataset to more than 100 countries, while also adding data for medium- and small taxpayer offices (grouped with the acronym MTO here) as well as anti-corruption agencies (ACAs).

III. Dataset

We predominantly rely on the World Revenue Longitudinal Database (WoRLD),2 which has been put together at the IMF and comprises data on total revenue, taxes on individual income, corporate income, goods and services, and international trade, as well as social security contributions. The dataset is compiled from multiple sources and covers 147 countries (both advanced and developing) between 1995 and 2014.3 We conduct a robustness check with revenue data compiled by the International Centre for Tax and Development (ICTD), whose results are presented in Appendix III.4 All the revenue data are presented as a percentage of nominal GDP.

In addition to the WoRLD database, we use data from the IMF’s Revenue Administration Gap Program (RA-GAP) and the Revenue Administration–Fiscal Information Tool (RA-FIT).5 The RA-GAP program has been applied to 23 countries to evaluate the difference between potential and actual collection of VAT between 2003 and 2013. RA-FIT is a web-based data gathering tool used to establish a baseline for revenue administration performance, to facilitate comparative study and benchmarking. There are over 140 countries included in this tool with hundreds of variables on revenue collection services, revenue authority staffing, tax base registration, dispute resolution, auditing and verification, among others. The surveys have been conducted over the past five years and country specific data are currently not publicly available.

The dates of establishment of LTOs, MTOs and ACAs are identified for more than 100 countries, making it the largest information set on LTOs and MTOs currently available. The data draws on information from IMF teams working on these countries and their reports, as well as country officials, websites of revenue authorities or finance ministries, and other online sources. In addition, we cross-checked the results with dates presented for subsets of countries in Baer (2002), Fossat and Bua (2013), and Ebeke et al. (2016). Appendix IV summarizes the results. Overall, we collect dates (or verify their non-existence) for 109 LTOs, 62 MTOs and 77 ACAs.

We primarily use two corruption measures throughout this paper, the Corruption Perception Index by Transparency International (TI) and the World Bank’s Control of Corruption Index. Both indices measure the perception of corruption, defined as the misuse of public power for private benefit, and cover about 180 countries. The two indicators differ in the sources used and specific questions asked to survey participants. Both are primarily perception-based, which has been criticized across the literature since actual and perceived levels of corruption might differ significantly. In the absence of measures for actual levels of corruption, and given that tax evasion might be equally impacted by the perception of corruption, the indices here are taken as proxies for actual corruption. In addition, while highly correlated, the indices have been subject to methodological changes over time (for example, the TI’s Corruption Perception Index underwent a significant change in 2012). For this reason, both indicators are used to verify the robustness of the results. The corruption variables have been rescaled to indicate that higher values are associated with greater corruption.

IV. Stylized Facts

At the outset, we look at the correlation between corruption, tax revenue, and compliance indicators. Total tax revenue is negatively correlated with the Transparency International Corruption index,6 which implies that increased corruption is associated with lower levels of the tax-to-GDP ratio.

A disaggregation of data into individual income and corporate income taxes shows that the former is more strongly negatively correlated with corruption (Figure 2). Large corporate enterprises often fall under the supervision of LTOs and are part of the formal sector, in contrast to smaller enterprises, which could partly explain this result. Furthermore, the scope for discretion tends to be more pronounced in the case of personal income tax (Besley and Persson, 2014). For individual income tax, the perception of corruption lowers the willingness of individual taxpayers to comply fully with tax laws because of the perception that their taxes will be misused by the authorities.

Figure 1.
Figure 1.

Total Tax Revenue

(1995–2014, period average)

Citation: IMF Working Papers 2017, 255; 10.5089/9781484326039.001.A001

Sources: Transparency International, IMF WoRLD Database.
Figure 2.
Figure 2.

Tax Revenue Components and Corruption

(1995–2014, period average)

Citation: IMF Working Papers 2017, 255; 10.5089/9781484326039.001.A001

Sources: Transparency International, IMF Staff Estimates.

Panel C in Figure 2 shows a negative correlation between corruption and social security contributions. This is likely attributable to a characteristic of the informal sector in many countries, leading individuals and small and medium-sized enterprises to not meet their social security obligations. For international trade, the negative correlation with corruption reflects the delays that traders face at the border and bribes they are asked pay to clear their goods.

Ideally, one would like to go beyond the correlation between tax revenue and corruption by analyzing compliance for different taxes. Data on the VAT’s C-efficiency ratio offers such an analysis. The C-efficiency ratio compares the actual VAT revenue with its potential when a uniform tax rate is applied to all consumption. A shortfall in the former (i.e., a ratio less than 1) is influenced by both VAT design and compliance. Data for 108 countries shows that there is a strong correlation between a lower C-efficiency and a higher perception of corruption (Figure 3, Panel E).

Figure 3.
Figure 3.

Tax Compliance and Corruption

Citation: IMF Working Papers 2017, 255; 10.5089/9781484326039.001.A001

*: Data spans 1995–2014, and points represent the period average for each country.**: Data represents between 4 and 8 observations in the years 2004–2013 for each country.Sources: Transparency International, IMF RA-GAP, IMF RA-FIT, IMF Staff Estimates.

As noted above, the RA-GAP program has generated data on the VAT compliance gap for 23 countries. When plotted against the level of perceived corruption (Panel F), we find a positive relationship, that is, the higher the level of perceived corruption, the larger is the VAT compliance gap. In addition, Panel G shows that for countries with higher perceived corruption, more days are required to clear imports through customs, as importers go through unofficial channels to get their items released. Finally, we find the relationship between corruption and VAT refunds to be negative— a result that confirms problems with getting VAT refunds in more corrupt countries.

V. Revisiting the Impact of Corruption on Tax Revenue

A. Empirical Methodology – Panel Techniques

The empirical section assesses the effect of corruption on tax revenue. The following equation is estimated:

y i t = α i + λ t + β * C o r r u p t i o n i t + n k ( θ k * X i t k ) + i t ( 1 )

Where yit is the tax revenue-to-GDP ratio for a country i at time t, and Corruptionit represents a measure of corruption. The regressor Xitk represents a set of k control variables, which, following the literature (Baunsgaard and Keen 2010), include proxies of the development level, the size of the tax base, the degree of openness, and the macroeconomic environment. αi and λt represent a full set of country and time fixed effects, respectively. it is the error term. For our controls we use trade openness, which is the value of the total imports and exports as a percentage of GDP, as well as inflation, the share of agriculture in the economy, and GDP per capita in purchasing power parity terms.7

Estimating equation (1) is challenging because of potential reverse causality between corruption and taxation. High taxation could encourage tax evasion and low tax capacity could favor corrupt behavior. To address this endogeneity issue, the literature has applied different instrumental variable approaches (Ghura 2002, Hwang 2002, Attila et al. 2008, and Thornton 2008). In its IV-GLS procedure, Ghura (2002) uses the contemporaneous, squared, and lagged values of population, population growth, and the urbanization rate as instruments. In Hwang (2002), corruption is instrumented with three variables: the percent of population professing protestant faith, a dummy for a former British colony, and an index of ethnolinguistic fragmentation. Attila et al. (2008) favors ethnolinguistic fragmentation, legal origins (French and British), distance from the equator, and the urbanization rate as instruments. Thornton (2008) uses openness, the black-market exchange rate premium, ethnic fragmentation, and a dummy for oil producers instead. In this paper, we follow Hwang (2002), Attila et al. (2008), and Thornton (2008) by using the ethnolinguistic fragmentation index (Taylor and Hudson, 1972), and adding an index capturing religious fragmentation (Mauro, 1995), both employing the argument that more fractionalized countries tend to have more dishonest bureaucracies.

An additional challenge is the possibility that some omitted variables may be correlated with both corruption and tax collection. To mitigate the omitted variable bias, the literature has used the within fixed effect estimator to control for country-specific factors. However, this technique expunges the heterogeneity in the corruption variable and the instrument variables, which exhibit significant country-specific components (see Appendix I). Accounting for both endogeneity and fixed effects would have been preferable, but the lack of idiosyncratic heterogeneity in the corruption index led to weak results.

For the reasons mentioned above, we apply pooled IV estimations. Pooled methods do not control for fixed effects and assume that the unobserved heterogeneity is uncorrelated with the covariates. This approach helps prevent the collinearity bias between the corruption indexes, the IV variables, and the country fixed effects. All estimates correct for autocorrelation and heteroskedasticity.

B. Results

Baseline specifications. The pooled IV results are reported in Table 2, columns (1) and (2), and the corresponding first-stage results are discussed in Appendix II. In line with the existing literature, the relationship between corruption and tax revenue in percent of GDP is negative and significant. The estimated coefficients are significant at least at the 5 percent level, indicating that an increase of one standard deviation (SD) in corruption perception―which corresponds to a differential between relatively highly corrupt and relatively less corrupt countries8―would correspond to an on average 12 percentage points lower tax revenue-to-GDP ratio.

Table 2.

The Effect of Corruption on Tax Revenue

article image
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Most control variables have the expected signs, comparable to the findings in Sen Gupta (2007). The coefficient associated with openness is positive, but only significant when the advanced economies dummy is added in the baseline specifications. Inflation has a positive impact on the tax-to-GDP ratio (inflation, when unanticipated, increases tax revenue). The increase in nominal income pushes people into higher tax brackets. There is a negative and significant relationship between the share of agriculture and revenue performance in the baseline specification. This confirms that the agricultural sector may be difficult to tax, especially if it is dominated by subsistence farmers. The coefficient associated with the log of per capita GDP is positive and significant in the baseline specification. This is in line with other studies that found that the capacity to collect and pay taxes increases with the level of development. The effect of the log of per capita GDP on tax revenue is nonlinear because the coefficient associated with the squared term is negative and significant.

We use the generalized method of moments (GMM) estimator to test for robustness of our results. Unfortunately, GMM does not lead to consistent estimates with our dataset. The number of instruments outweigh the cross-sectional dimension, and the model is over-identified, with the probability for the Sargan-Hansen test being well above the accepted thresholds in the literature.9 In summary, our baseline estimates confirm the evidence that corruption undermines tax collection.

Advanced economies versus developing economies. We further check whether the negative relationship between corruption and tax revenue varies between advanced and developing economies (the latter including both low income and emerging markets). We introduce additively and multiplicatively (with the corruption index) a dummy variable taking the value of 1 if the country is an advanced economy, and 0 otherwise. The results are shown in Table 2, columns (3) and (4). The adverse effect of corruption on tax collection is larger in advanced economies relative to the average impact for all countries in the sample. The coefficients associated with corruption are negative, and statistically significant at the 1 percent level. The coefficients of the interactive terms are significantly negative. This result is a novelty to the existing literature and implies that any given change in the level of corruption is more detrimental in advanced economies, likely because of the lower overall level of corruption in these countries.

Tax components: Next, we turn to the relationship between corruption and the components of tax revenue. We split overall tax revenue into personal and corporate income tax, social security, property tax, taxes on domestic goods and services, and international trade taxes. This approach should help understand how the adverse effects of corruption impact specific taxes and avoid misleading conclusions that can arise from broad aggregates, especially if tax components move in offsetting directions. We re-estimate equation (1) for each tax item using the pooled IV methodology described above.

The results are shown in Table 3, where all taxes are specified as a ratio to GDP. PIT, VAT, social security contributions, and trade taxes all confirm the negative relationship with corruption. The associated coefficients are negative and statistically significant at the 1 percent level.

Table 3.

The Effect of Corruption on Tax Revenues Components

article image
Notes: Robust standard errors in parentheses, *** significance at the 1 percent level, ** significance at the 5 percent level, * significance at the 10 percent level.

The negative impact on PIT suggests that corruption works through informality, which favors tax evasion. An increase of one SD in corruption perception or control of corruption would decrease the PIT ratio by 0.2 percentage points per year. The impact of corruption on goods and services is felt most strongly via its impact on the VAT—the tax with most revenue potential. An increase of one SD in corruption perception would decrease the VAT ratio by 0.2 percentage points per year. Excise taxes tend to be unaffected. Social security contributions (SSCs) are strongly affected. An increase of one SD in corruption would decrease SSCs by 0.5 percentage points per year. An increase of one SD in corruption perception would decrease the international trade tax revenue by 0.2 percentage points per year. While the importance of international trade taxes is declining for many developing countries, they remain a crucial source of revenue in fragile states.

Surprisingly, the effect of corruption on CIT is positive. An increase of one SD in corruption perception is associated with an increase in the CIT ratio by 0.1 percentage points per year. Given the deterrent effect of corruption on taxes prone to informality, there seems to be a substitution in favor of the tax burden on the formal sector. Finally, we do not find any impact on property taxes, which typically has relatively small share in overall revenue. Furthermore, property taxes are usually levied at the sub-national level where compliance may be weak, particularly in developing countries. The results remain broadly robust to the use of an alternative ICTD database (see Appendix III).

VI. Improving Compliance—The Impact of Institutions

In this section, we analyze the impact of specific public institutions—such as LTOs—on compliance. These institutions are expected to have a dampening effect on corruption and a positive influence on tax revenue. We collect dates of establishment on LTOs, MTOs and ACAs, which are detailed in Appendix IV. The resulting database on LTOs and MTOs is the largest of its kind, and will be useful for future impact analysis of fiscal and public institutions on economic performance.

The first wave of establishing LTOs can be observed during the mid- to late 90s, and currently only a minority of countries have not opted to establish them.

Figure 5.
Figure 5.

Effectiveness of Compliance Offices

Citation: IMF Working Papers 2017, 255; 10.5089/9781484326039.001.A001

Notes: Chart only shows observations for which date of establishment could be identified, or where it is certain that institutions do not exist.

A. Propensity Score Matching Technique

The dates of establishment of taxpayer offices or ACAs are the foundation of the propensity-score matching (PSM) methodology developed by Rosenbaum and Rubin (1983), which is applied here to analyze the impact of these institutions on corruption and tax revenues. PSM is a statistical matching technique that attempts to estimate the effect of a treatment, policy, or other intervention by accounting for the covariates that predict receiving the treatment. In our case, treatment is the establishment of a specific institution (LTO, MTO, or ACA) to counter corruption and improve revenue performance. Ebeke et al. (2016) applied this methodology to study the impact of LTOs, VAT and semi-autonomous revenue authorities on tax revenue in 41 sub-Saharan Africa, an approach that is applied here in a difference-in-difference specification.

The reason for using difference-in-difference methodology vis-à-vis panel estimation is that it is difficult to control for endogeneity and the resulting bias in the latter approach. The bias arises because the apparent difference in outcomes between countries with and without units may depend on characteristics that determine whether a unit was established in the first place, and therefore the outcome could be dependent on these characteristics rather than on the establishment (treatment) per se. Given the absence of suitable instruments to capture the institution’s establishment, we refrain from using standard panel estimation. PSM avoids the common endogeneity issue by using explanatory variables to find observations across countries when they were equally likely to establish the respective institution, by means of propensity scores. The impact of an institution on the dependent variable is then compared across countries with similar propensity scores.

Propensity scores are estimated with a Probit model, using a set of independent variables that are close to the previous Section V – a dummy variable for the state of development, the share of agriculture, real per capita GDP, openness, and population – as well as a variable for time. Both the general income category (AE, EM or LIDC), as well as the income level within these groups impacts the likelihood of establishment of institutions, specifically LTOs and ACAs. For example, many of the poorest LIDCs share the characteristic of below-average LIDC tax ratios and higher levels of perceived corruption. All controls, except for population, are significant in explaining the likelihood of establishment of these institutions. In an expanded baseline, we add CPI inflation, latitude, a dummy variable for oil exporters, total tax revenue as a share to GDP, and the ratio of external debt to GDP. Especially the latter two are likely to influence decisions to increase efforts to mobilize revenue or reduce corruption for economic and fiscal stabilization purposes.

Once the propensity scores are estimated, the observations are matched with three different matching estimators – Stratification, Kernel, and local linear regression matching. We enforce common support for the estimation, i.e., treatment observations with propensity scores higher r than the highest t propensity score in the non-treated pool are left unmatched.10

The average treatment effect (ATE) – the impact from the establishment of the units on variables of interest, compared to countries without the units – is analyzed based on the identified matches. We use a difference-in-difference approach specified as follows:

A T E = E ( Δ t + x Δ t x | T = 1 ) = E ( ( Y t + x T = 1 Y t + x T = 0 ) ( Y t + x T = 1 Y t + x T = 0 ) ) | T = 1 = E ( Y t + x T = 1 Y t + x T = 1 ) | T = 1 ( Y t + x T = 0 Y t + x T = 0 ) | T = 1 ( 2 )

where T=1 if an institution is established, and T=0 otherwise. The first term on the lower right hand side refers to the differences in outcomes before (or at) and after the treatment for the treated group. The second term uses the differences in outcomes for the control group to eliminate the bias. This specification eliminates the bias arising from common time trends in the dependent variables. It further eliminates issues resulting from the likely case that countries without the respective institutions have a higher tax revenue ratio or higher corruption to begin with. For example, if we used the revenue ratio in absolute terms and not in difference levels as a dependent variable, the results would simply compare the revenue-to-GDP ratios between countries with and without institutions. If countries that establish the institutions have lower tax ratios on average, the ATE will suggest that LTOs lead to lower tax revenue mobilization. Instead, taking the difference-in-difference approach changes the interpretation of the ATE to ‘how much did the tax revenue ratio increase after the establishment of an LTO compared to countries that have not done so’.

B. Results

The PSM baseline results are shown in Table 4, with the dependent variables specified as follows: (i) for corruption we use the change in the Corruption Perception Index three years after and one year before establishment of an institution; (ii) for revenue the difference between tax revenue two years after and the year of establishment is used.11 In the baseline, treatment (establishment) is noted as one occurrence only, and all years after the establishment year are dropped. All years before establishment are kept, except for three (for corruption) or two (for revenue) years before the year of establishment to avoid an overlap in the dependent variable.

Table 4.

Impact of Institutions on Corruption and Tax Revenue – Baseline

article image
Notes: IMF estimates, *** significance at the 1 percent level, ** significance at the 5 percent level, * significance at the 10 percent level.

Column 1 of Table 4 shows that LTOs have a statistically significant impact on the perception of corruption (significance levels are based on the bootstrap T-statistics).12 This impact is negative, with an average treatment effect (ATE) of about −3, i.e., the perception of corruption significantly declines (or increases significantly less) during the three years after the establishment of the LTO compared to other countries. On the other hand, the impact of both anti-corruption units and MTOs on the Corruption Perception Index is insignificant. For ACAs, the result is in line with the notion that the establishment of anti-corruption units alone is likely to be insufficient to lower corruption, unless they are given wide ranging independence and power to pursue relevant cases. It further provides empirical support to those who have argued that ACAs are a mere additional bureaucratic step in a corrupt environment that is used to support corrupt activities (for example bribes to avoid prosecution).

Table 4 further shows a statistically significant and positive impact of the establishment of LTOs on tax revenue, with an average treatment effect of about 0.6 percent of GDP: following the establishment of an LTO, tax revenue increases on average by 0.6 percent of GDP more (or declines less) over two years compared to countries that have not established an LTO. At the same time, the impact of the establishment of MTOs and ACAs, while positive, is insignificant. These results suggest that the establishment of LTOs has led to improvements in how compliance of this key segment is managed – via targeted taxpayer service and improved enforcement programs. This may not be the case with respect to MTOs. The establishment of MTOs is a relatively new concept, and while offices have been set up in several countries, tailored compliance programs for this segment have, on average, not yet been implemented to the same degree as for LTOs. In addition, MTOs are often regionally organized and managed, with decentralization making targeted program implementation and supervision more challenging.13

The impact of LTOs on revenue and corruption might seem contrary to the notion that increased perception of corruption has a positive impact on CIT revenues. However, the cross-country regression describes a structural comparison of how revenue levels and its structure compare at different levels of perceived corruption. On average, countries with higher perceived corruption have higher CIT revenues. The establishment of LTOs could lead to a shorter-term reduction in the level of perceived corruption, as well as to an increase in CIT and other tax revenue, without altering this structural comparison.

Since the treatment effects of MTOs and ACAs on corruption and tax revenue remain largely insignificant for the remainder of specifications below, results for them are not shown.

We conduct several robustness tests in Table 5. First, we expand the set of independent variables to include CPI inflation, a control for latitude, a dummy variable for oil exporters, total tax revenue as a share to GDP, and the ratio of external debt to GDP. The second specification deletes all fragile states from the baseline estimation on the assumption that their institutions are weaker than other countries. Finally, and with fragile states included, the last test refrains from dropping the observations after an institution is established, increasing the number of observations with an LTO in place significantly. The results are consistent across all specifications. The perception of corruption declines and revenue increases significantly when fragile states are excluded, and the results seem to hold for the entire sample, suggesting that after LTOs are established, revenue continues to increase faster than in countries without LTOs.14

Table 5.

Impact of LTOs on Tax Revenue—Robustness

article image
Notes: IMF estimates, *** significance at the 1 percent level, ** significance at the 5 percent level, * significance at the 10 percent level.

VII. Policy Conclusions

This paper revisits the relationship between perceived corruption and taxation, by using a recently available revenue dataset. It finds that corruption has considerable impact on aggregate revenue performance—as much as 0.6 percent of GDP annually. The impact is felt even on taxes with most revenue potential, such as the VAT. Taxes on income and international trade as well as social security contributions are adversely affected. It seems that the effect of corruption on revenue performance is mediated through poor compliance as suggested by a strong correlation between VAT compliance and perceived corruption. Corruption takes a toll on the functioning of the VAT; the higher the perceived corruption, the slower the speed at which VAT refunds are given.

The paper further examined the effect of institutions designed to reduce corruption and improve tax compliance. In this regard, it assessed the impact of over 100 LTOs, MTOs and ACAs across different country groups and regions. It found that LTOs do indeed reduce perceived corruption. They further seem to enhance revenue performance with countries with LTOs outperforming revenue generation of countries without LTOs by as much as 0.6 percent of GDP. The effect of MTOs and ACAs on corruption and revenue performance is not statistically significant.

Appendix I. Variance Decomposition of Corruption, Instruments, and Tax Revenues-to-GDP

article image
Notes: Values show the percentage of the variable that is explained by country fixed effects.

Appendix II. First Stage

All instruments are highly significant in explaining the corruption index. Ethnolinguistic fragmentation has a positive and significant impact on corruption, whereas religious fragmentation has a negative and significant effect. The partial R2 of the excluded instruments is between 0.26 and 0.28. Moreover, the F-statistics of the first stage are significant at 1 percent.

article image

Controls included

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Appendix III. Results with ICTD

article image

Appendix IV. LTO, MTO, and ACA Establishment Dates

article image
article image
article image
article image
Notes: (a) Georgia – The LTO was dismantled in 2010. (b) Iraq – An LTO does exist, but it is staffed with 4 people only. We therefore classify it as not-functional. (c) Kenya – Two dates are presented, because LTO was formed as an operations unit in 1998 to provide one-stop shop services in the administration of income tax and VAT matters affecting large taxpayers. The LTO was developed into a fully-fledged department in the year 2006 with the sole purpose of administering domestic tax matters affecting large taxpayers. Both dates are kept due to a likely impact from both reforms. (d) Uganda – The LTO was closed in 2001 and re-opened in 2004.

Appendix V. Summary Statistics

article image

References

  • Aghion, P., Akcigit, U., Cagé, J. and Kerr, W. (2016). Taxation, Corruption, and Growth. NBER Working Paper 21928, Cambridge, Massachusetts: National Bureau of Economic Research.

    • Search Google Scholar
    • Export Citation
  • Attila, G., Chambsa, G. and Combes, J. (2008). Corruption et Mobilisation des Recettes Publiques: Une Analyse Économétrique. Recherche Economique de Louvain.

    • Search Google Scholar
    • Export Citation
  • Baer, K. (2002). Improving Large Taxpayers’ compliance. A Review of country experience. Washington, D.C.: International Monetary Fund.

    • Search Google Scholar
    • Export Citation
  • Besley, T. and Persson, T. (2014). Why Do Developing Countries Tax So Little? Journal of Economic Perspectives, 28(4), pp. 99120.

  • Baunsgaard, T. and Keen, M. (2010). Tax revenue and (or?) trade liberalization. Journal of Public Economics, 94, (9–10), pp. 563577.

    • Search Google Scholar
    • Export Citation
  • Dreher, A. and Herzfeld, T. (2005). The Economic Costs of Corruption: A Survey and New Evidence. SSRN Electronic Journal, doi:10.2139/ssrn.734184, June.

    • Search Google Scholar
    • Export Citation
  • Ebeke, C., Mansour, M. and Rota Graziosi, G. (2016). The Power to Tax in Sub-Saharan Africa: LTUs, VATs and SARAs. Fondation Pour Les Etudes et Recherches Sur Le Developpement International Working Paper 154.

    • Search Google Scholar
    • Export Citation
  • Ebrill, L., Keen, M., Bodin, J. and Summers, V. (2001). The modern VAT. Washington, D.C.: International Monetary Fund.

  • Fossat, P. and Bua, M. (2013). Tax Administration Reform in the Francophone Countries of Sub-Saharan Africa. IMF Working Paper 13/173, Washington, D.C.: International Monetary Fund.

    • Search Google Scholar
    • Export Citation
  • Gaspar, V., Jaramillo, L. and Wingender, P. (2016). Tax Capacity and Growth: Is there a Tipping Point? IMF Working Paper 16/234, Washington, D.C.: International Monetary Fund.

    • Search Google Scholar
    • Export Citation
  • Ghura, D. (2002). Tax Revenue in Sub-Saharan Africa: Effects of Economic Policies and Corruption. In: G. Abed and San. Gupta, ed., Governance, Corruption, and Economic Performance, 1st ed. Washington, D.C.: International Monetary Fund.

    • Search Google Scholar
    • Export Citation
  • Hwang, J. (2002). A Note on the Relationship between Corruption and Government Revenue. Journal of Economic Development, 27 (2), pp.161178.

    • Search Google Scholar
    • Export Citation
  • Imam, P. and Jacobs, D. (2014). Effect of Corruption on Tax Revenues in the Middle East. Review of Middle East Economics and Finance, 10 (1), pp.124.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund (IMF). (2015). Current Challenges in Revenue Mobilization: Improving Tax Compliance. IMF Policy Paper, Washington, D.C.

    • Search Google Scholar
    • Export Citation
  • Mauro, P. (1995). Corruption and Growth. Quarterly Journal of Economics, 110 (3), pp.681712.

  • Roodman, D. (2009). A Note on the Theme of Too Many Instruments. Oxford Bulletin of Economics and Statistics, 71, pp. 135158.

  • Rosenbaum, P. and Rubin, D. (1983). The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika, 70(1), pp. 4155.

    • Search Google Scholar
    • Export Citation
  • Schneider, F. (2006). Shadow Economies and Corruption All Over the World: What Do We Really Know? IZA Discussion Paper No. 2315, Bonn, Germany: Institute for the Study of Labor (IZA).

    • Search Google Scholar
    • Export Citation
  • Schneider, F. and Enste, D. (2000). Shadow Economies: Size, Causes, and Consequences. Journal of Economic Literature, 38(1), pp. 77114.

    • Search Google Scholar
    • Export Citation
  • Sen Gupta, A. (2007). Determinants of Tax Revenue Efforts in Developing Countries. IMF Working Paper 07/184, Washington, D.C.: International Monetary Fund.

    • Search Google Scholar
    • Export Citation
  • Taylor, C. and Hudson, M. (1972). World Handbook of Political and Social Indicators. 2nd ed. New Haven: Yale University Press.

  • Thornton, J. (2008). Corruption and the Composition of Tax Revenue in Middle East and African Economies. The South African Journal of Economics, 76 (2), pp.316320.

    • Search Google Scholar
    • Export Citation
1

We would like to thank Vitor Gaspar, Michael Keen, Paolo Mauro, Kevin Greenidge, Ashvin Ahuja, Davide Furceri, Claudia Berg, Suhaib Kebhaj, Carlos Goncalves, Antonio Spilimbergo, Santiago Acosta Ormaechea, Ricardo Llaudes, and John Ralyea for their helpful comments, as well as participants of IMF’s Fiscal Affairs Department seminar for their insightful comments. We are grateful to Mark Albertson for his research assistance. The paper is forthcoming in the Special Issue on Tax and Corruption of e-Journal of Tax Research.

3

The sources are IMF country reports, Government Finance Statistics (GFS), World Economic Outlook (WEO) and the Organization for Economic Co-operation and Development (OECD).

4

The ICTD Government Revenue Dataset (GRD) combines data from several international sources to compile a comprehensive database. The data includes 50 revenue indicators and global coverage. There are some methodological and source differences between the ICTD data and WoRLD, making it a strong candidate for robustness checks.

6

Similar results are found when the World Bank Control of Corruption index is used.

7

This data is obtained from the World Bank Development Indicators covering the years 1995 to 2014.

8

For example, the average Corruption Perception Index is 26.3 for advanced, 62.5 for emerging, and 74 for low income economies.

9

After following Roodman’s (2009) strategy, we found that the probability of the Hansen J-statistic test equals 1 in most cases. The issue is even more pronounced when examining various tax categories.

10

The distributions of the propensity scores for treated and control groups are visually inspected with kernel density plots for each of the presented estimates, which are almost identical to one another after matching for all presented results below.

11

The impact on corruption perception is likely to be realized with delay. However, where signal effects are strong, perception might change in the year of establishment already. In contrast, the effects of LTOs on revenue are unlikely realized in the first year of establishment, given a natural learning process. However, the results hold for the LTO impact on tax revenue also if tax revenue is specified like the Corruption Perception Index.

12

The Kernel and log-linear regression methods estimate the propensity scores after excluding all observations for which the dependent variable is not available (using the psmatch2 command in Stata). The stratification method first estimates the propensity scores, and in a separate step estimates the ATE (using the pscore command in Stata). Consequently, the treatment observations are different across these methods.

13

A third argument is that shareholders of larger corporates have an interest in a no-corruption policy in the companies they are invested in, while smaller taxpayers and firms often directly interact with the tax authorities.

14

Ideally, we would like to disaggregate the revenue data into its components, and analyze the institutions’ impact by level development. Unfortunately, these analyses lead to a significant loss of observations in the treatment group, rendering this analysis infeasible.

  • Collapse
  • Expand
Corruption, Taxes and Compliance