Building Integrated Economies in West Africa
Chapter

Chapter 12. Fiscal Space and Investment Scaling Up

Author(s):
Alexei Kireyev
Published Date:
April 2016
Share
  • ShareShare
Show Summary Details
Author(s)
Christine Dieterich, Karim Barhoumi, Qiang Cui, Sergio Sola and Alexander Raabe 

West African Economic and Monetary Union (WAEMU) countries face an important common challenge of creating sufficient fiscal space to finance ambitious growth, development, and poverty-reduction programs in individual countries. Such additional fiscal space can be created by either enhancing tax revenue or improving the efficiency of spending. While WAEMU countries are broadly in line with comparator countries in total tax collection, WAEMU’s tax revenue relies heavily on trade taxes, which will inevitably be reduced with impending trade liberalization. Also, high reliance on trade taxes makes the WAEMU’s revenue base vulnerable to the fluctuation of international prices. Panel regression and stochastic frontier analysis suggest substantial room to improve domestic tax collection in the WAEMU by 0.8 to 2 percent of GDP. The effort should be country-specific with each government focusing on its underperforming tax category. On the expenditure side, WAEMU countries have significant scope to improve efficiency of their spending on education and health. If all WAEMU countries achieve the spending efficiency of the top performer among them, the fiscal savings on average can add 1 to 3 percent of GDP annually to the available fiscal space of the region.

The Need for Scaling Up

WAEMU countries need to mobilize substantial financial resources to address the infrastructure gap, which has been widely identified as a growth bottleneck. Many studies find that inadequate infrastructure impedes growth (for example, Commission for Africa 2005 and Foster and Briceño-Garmendia 2009). Infrastructure development was estimated to have contributed about 1 percentage point to per-capita growth in West Africa in 2001–05 (for example, Calderon 2009, Figure 12.1). For Benin, Domínguez-Torres and Foster (2011) estimate that infrastructure contributed 1.6 percentage points to per-capita growth; while in Senegal, Torres, Briceño-Garmendia, and Dominguez (2011) find the contribution was about 1 percent point. Also, raising the two countries’ infrastructure endowment to that of Africa’s middle-income countries could boost annual growth by 3.2 and 2.7 percentage points, respectively. Recent reports also confirmed a continued infrastructure bottleneck in other WAEMU countries (for example, IMF 2013a, b, c).

Figure 12.1.Per-Capita Growth and Changes in Selected Growth Fundamentals 2001–02

Source: Calderón 2009.

Note: East Asian Tigers = Hong Kong, Singapore, South Korea, and Taiwan.

To generate the financing for the scaling up of public investment while preserving macroeconomic stability, WAEMU countries have to use their fiscal space efficiently. While WAEMU countries’ external debt levels declined owing to the heavily indebted poor countries/multilateral debt relief initiative, leaving some scope for external borrowing, the availability of financing at attractive terms is limited. Also, some countries’ total government debt has increased considerably since those countries received the debt relief, which suggests that caution is warranted in additional borrowing. Therefore, it is essential for the sustainable financing of scaling up infrastructure investment that the two major channels for creating fiscal space be used. These channels are increasing tax revenue and increasing the efficiency of spending.

Raising Tax Revenue

Improving tax collection remains the main channel for enlarging the fiscal space. This has been well recognized in the WAEMU, which has a convergence criterion of 20 percent for the tax-to-GDP ratio, even though several member countries have not been in compliance with this criterion for years.

The WAEMU’s relatively high indirect tax rates have not resulted in higher tax collection. The higher indirect tax rates in the WAEMU compared with those in sub-Saharan Africa and low-income country averages,1 especially for goods and services taxes and trade taxes, have not translated into higher revenues. Roughly, the tax-to-GDP ratio has been below the sub-Saharan Africa average throughout the observation period (2000–11), and just broadly in line with the low-income country averages (Figure 12.4). Looking at the trend over time, the WAEMU’s tax-to-GDP ratio improved from 11.7 percent of GDP in 2000 to 14.7 percent of GDP in 2011, driven by a broad trend in all member countries except Côte d’Ivoire, where results were affected by internal conflicts. However, the size of improvements varied considerably among the countries. For instance, Benin’s total tax revenue increased by 2.1 percentage points, while Togo’s total tax revenue rose by 6.6 percentage points.

Figure 12.2.Total and External Government Debt in the WAEMU

Source: IMF staff estimates.

Figure 12.3.Tax Rate in the WAEMU

Sources: IMF Fiscal Affairs Department database; and IMF staff calculations.

Note: Data refer to 2010–13 averages. SSA = Sub-Saharan Africa; WAEMU = West African Economic and Monetary Union.

Figure 12.4.Total Tax Revenue

(Percent of GDP)

Sources: IMF Fiscal Affairs Department database; and IMF staff calculations.

Looking at the performance tax by tax, the improvement in the WAEMU’s tax ratio is driven by higher collection from income tax and goods and services taxes, while trade revenues are broadly flat due to limited trade liberalization. For example:

  • Trade taxes—In contrast with sub-Saharan African and low-income countries, where weighted average tariff rates declined, reflecting trade liberalization over the last decade, the WAEMU’s tariff rates dropped only marginally and the tax-to-GDP ratio has remained broadly stable over time (Figure 12.2). In the comparator groups, sub-Saharan Africa’s drop in trade tax revenues reflects the rate decline, while it seems that low-income countries were able to offset the rate decline by efficiency measures that allowed these countries to broadly preserve the trade tax-to-GDP ratio (Figure 12.3 and 12.5).
  • Income taxes—The WAEMU increased the tax-to-GDP ratio from close to 3 to close to 4 percent of GDP, but it remained below the ratios for low-income countries and sub-Saharan Africa (Figure 12.6).
  • Goods and services taxes—The francophone tradition of relying more on direct than on indirect taxation is reflected in the comparably higher rates. This translates into a higher level of tax revenues than is the case in comparator countries by around 0.6 to 0.8 percent of GDP. Also, the improvement in the WAEMU countries over the observation period was most pronounced in this tax category (Figure 12.4 and 12.7).

Figure 12.5.Trade Tax Revenue

(Percent of GDP)

Sources: IMF Fiscal Affairs Department database; and IMF staff calculations.

Figure 12.6.Income Tax Revenue

(Percent of GDP)

Sources: IMF Fiscal Affairs Department database; and IMF staff calculations.

Figure 12.7.Goods and Services Tax Revenue

(Percent of GDP)

Sources: IMF Fiscal Affairs Department database; and IMF staff calculations.

WAEMU countries show considerable variation in the drivers for revenue collection by tax categories. For example, in Togo, income tax revenue declined from 2.9 percent to 2.5 percent of GDP, but goods and service tax revenue rose sharply from 2 percent to 9.2 percent of GDP. In Benin, the revenue gain was driven by higher trade tax revenue, while goods and services tax revenue declined. In Côte d’Ivoire, however, the decline in tax revenue was mainly driven by falling trade tax revenues.

Panel regressions were used to analyze the tax potential of WAEMU countries based on determinants identified in the literature. Drawing on the existing literature on determining tax potential (for example, Gupta 2007; Davoodi and Grigorian 2007; and Pessino and Fenochietto 2010), the following variables were considered as the determinants to estimate the tax potential, which was defined as the maximum level of tax revenue that a country can achieve given its macroeconomic fundamentals: GDP per capita, consumption, gross fixed capital formation, inflation, import and export as a share of GDP, share of agriculture in GDP, share of the urban population, natural resource rents, and broad money as a share of GDP (Annex Table 12.1.1). Obtaining the tax potential allows for calculating the tax gap, which is the percentage deviation of actual revenue from potential revenue.2 The regression analysis is not only conducted for total tax revenue, it is also conducted for revenue in the subcategories of goods and services, trade, and income (see Annex Table 12.1.2 through Annex Table 12.1.4).

Our analysis suggests that WAEMU countries are ahead of comparator countries in their total tax collection, but have room to improve in the collection of income taxes. In 2011, total tax collection in the WAEMU exceeded the potential revenue by around 6 percent and 12 percent when compared with low-income countries and sub-Saharan African countries, respectively (Figure 12.8, panel 1). This compares favorably with 2000 when the WAEMU’s total tax collection was below potential by around 4 percent compared with both low-income and sub-Saharan African countries. The following factors explain this trend:

Figure 12.8.WAEMU Tax Revenue Gaps

Source: IMF staff estimates.

Note: All panels estimate the potential WAEMU tax in two groups: Sub-Saharan Africa and low-income countries. More precisely, each bar refers to the WAEMU tax gap in Sub-Saharan Africa as well as low-income countries.

  • Goods and services taxes—The relative improvement between 2000 and 2011 was mainly driven by a more positive goods and services tax gap (Figure 12.8, panel 2). However, it should be taken into consideration that higher tax rates in the WAEMU explain at least part of this positive tax potential.
  • Trade taxes—Despite the higher weighted average tariff, our tax potential analysis indicates only a moderately positive tax gap in 2011. However, improvements of the trade tax revenue compared 2000 range from below to slightly above potential (Figure 12.8, panel 3) for both benchmark groups.
  • Income taxes—Revenue performance as measured by the tax gap deteriorated from around −1½ percent to around −5½ percent compared with the gap in sub-Saharan African countries (Figure 12.8, panel 4), and closed only slightly from around −3½ percent to around −2 percent with respect to the low-income country benchmark.

This signals a need for the WAEMU to improve income tax collection compared with the peer groups.

Our analysis on the country level shows deviations from the overall WAEMU trend for the tax gap. During the observation period, six of the eight WAEMU members augmented tax collection efforts compared with what was done in sub-Saharan African and low-income countries (Figure. 12.8, panel 1). As for income tax revenue collection efforts, only Mali, Niger, and Senegal experienced an increase in comparison with the benchmarks (Figure 12.8, panel 3). The goods and services tax gap widened in the case of Benin, Burkina Faso, Côte d’Ivoire, Guinea-Bissau, Mali, and Senegal compared with at least one benchmark group (Figure 12.8, panel 3). Among these countries, the trade tax revenue gap closed significantly or turned positive for Benin, Burkina Faso, Niger, and Senegal, meaning that the overall improvement for these countries was mostly driven by an improved trade tax collection relative to the benchmarks.3 In sum, the diverse developments within the WAEMU group suggest the need for country-specific policies to improve revenue collection in the identified tax categories. These policies should leave room for amelioration as suggested by the empirical analysis.

Similar results were obtained with stochastic frontier analysis. Following Pessino and Fenochietto (2010), the stochastic frontier analysis estimation (Annex 12.2) applies a time-varying parameter of technical inefficiency in tax collection to the different tax categories already discussed in the panel regression analysis. The results confirm the findings using panel regressions, but with some modifications (Figure 12.9). Namely, regardless of the reference group of countries (sub-Saharan African and low-income countries) used in the estimation, the WAEMU performed relatively well in terms of tax collection. This result, however, seems to be mostly driven by trade taxes. When looking at goods and services and income taxes, WAEMU countries seem to be less efficient compared with sub-Saharan African countries and all low-income countries. The main division in this finding is the assessment of the Value Added Tax performance, which was positive in the panel regression approach and negative in the stochastic frontier analysis. In contrast, the findings were broadly consistent for income tax and trade taxes, even though the stochastic frontier analysis gave a more positive assessment of the trade tax performance than did the panel regression model.

Figure 12.9.Tax Collection Efficiency in the WAEMU by Category

Source: IMF staff estimates.

Note: LIC = low-income countries; SSA = Sub-Saharan Africa; WAEMU = West African Economic and Monetary Union.

Overall, the WAEMU’s good performance on tax revenues relies significantly on trade taxes. This situation reflects, in part, the region’s heritage, as import duties have traditionally constituted the main source of tax revenue there, and, in part, the slow pace of trade liberalization. WAEMU countries have recently initiated reforms toward trade liberalization, most notably, the introduction of a common external tariff for all Economic Community of West African States countries in January 2015. While the implementation will be gradual, it is important for the WAEMU to enhance its domestic tax revenue base, in particular, income tax revenues, where both analytical approaches indicate room for improvement by around 0.8 to 2 percent of GDP.

Improving Spending Efficiency

Is there scope for creating fiscal space by improving the efficiency of public spending? Our analysis focused on the technical efficiency of translating public spending into the corresponding results by comparing Benin’s input-output performance in public spending to those of other sub-Saharan African countries with similar levels of development. In addition, to reflect WAEMU countries’ aspirations to accelerate growth, specific comparisons with the fast-growing non-resource rich sub-Saharan African countries were provided.4 Quantitative assessments were conducted through a nonparametric data envelopment analysis (DEA).5 While public spending covers many sectors, only the education and health sectors were analyzed because these are sectors in which public spending plays a major role, and consistent cross-country data are readily available. Furthermore, based on data for 2008-12, we estimated the potential budgetary savings from higher efficiency in education and health to better inform policy discussions.

The DEA methodology provides a parsimonious model at the aggregate level to assess the efficiency of public spending based on cross-country comparison of the input-output relationships. It uses a nonparametric approach to identify an “efficiency frontier” from the input-output relationships across the countries that share the same technology (see Herrera and Pang 2005 and Grigoli and Kapsoli 2013 for details). Each country’s efficiency is then compared with this frontier in the corresponding range of spending to obtain an efficiency score of between 0 and 100 percent, and variable returns to scales are taken into account in the estimation given the observed patterns of the data. In this analysis, data for 46 low-income and lower, middle-income sub-Saharan countries in 2003-12 were used in defining the efficiency frontier. However, data used in estimating potential savings were limited to those from WAEMU countries6 in 2008–12 to further strengthen the cross-country comparability to arrive at the most realistic estimates.

Education sector indicators in WAEMU countries have improved significantly in recent years, largely supported by higher spending. Between 2003–07 and 2008–12, average education spending increased from 4 percent to 4.4 percent of GDP in WAEMU countries. The increased spending also supported better result indicators, such as an increase of about 14 percentage points in primary school enrollment rates and an increase of about 11 percentage points in adult literacy. But the relative magnitude of the increase in result indicators was less than that of the spending increase. For example, education spending in peer countries remains stable, but achieved a similar increase (11 percentage points) in enrollments and a much higher increase (18 percentage points) in the adult literacy rate—which better reflects the quality of the education. This suggests lower efficiency in WAEMU countries in achieving quality education results, as compared with the fast-growing sub-Saharan African countries.

The quantitative DEA analysis confirms that WAEMU countries rank low in the efficiency of education spending compared with the peers. Despite progress in improving education indicators, WAEMU countries lag behind the peer countries in the technical efficiency of their education spending. This is illustrated by the fact that all WAEMU countries are below the efficiency frontier achieved by the most efficient countries in translating education spending into the result indicators in two periods: 2003–07 and 2008–12 (Figure 12.10). For example, at an average public education spending close to 5 percent of GDP in 2008–12, Benin stays well within the efficiency frontier (Figure 12.10, panel 1) and lags behind peer countries (for example, Liberia, Rwanda, and Togo) that are located to the northwest of Benin in the scatter plot chart. This means that these countries achieved a higher adult literacy rate at a lower per-capita spending7 than did Benin. Another example is Togo, which is closer to the frontier than is Benin, but achieved much lower adult literacy as compared with Uganda, which had the same level of spending. Based on the distance to the efficiency frontier, potential fiscal saving in achieving the same results can be estimated for WAEMU countries. To achieve realistic estimates, the calculation is limited to WAEMU countries that possess similar institutions and development status. Using this method, on average, WAEMU countries could save between 1 and 2 percent of GDP by improving their spending efficiency to the highest level in the Union (Table 12.2).

Figure 12.10.Efficiency of Education Spending: WAEMU and Peer Groups

Sources: World Development Indicators; IMF Fiscal Affairs Department database; and IMF staff estimates.

Table 12.1Education Spending and Result Indicators
Education Spending in Percent of GDPPrimary Enrollment Rates (Percent)Adult Literacy Rate (Percent)
2003–072008–122003–072008–122003–072008–12
Benin3.74.7106124.238.542.4
Burkina Faso4.53.655.476.424.7na
Mali4.24.370.879.825.131.1
Niger3.34.248.66628.7na
Senegal4.35.481.986.741.949.7
Togo3.74.1117.8135.356.957.1
Ethiopia5.54.979.2102.934.9na
Mozambique4.9na102.2na48.2na
Rwanda4.34.5133.8142.6na71.1
Tanzaniana6.5na103.3na73.2
Uganda53.1126.7120.471.473.2
WAEMU mean44.474.988.935.446
Peer country mean4.94.890.7101.541.559.2
Sources: World Development Indicators; IMF Fiscal Affairs Department database; and IMF staff calculations.Note: Data are not available for all countries.
Sources: World Development Indicators; IMF Fiscal Affairs Department database; and IMF staff calculations.Note: Data are not available for all countries.
Table 12.2Efficiency Scores in Education Spending and Estimated Potential Savings1, 2
Baseline ScorePotential Saving (Percent GDP)Scores in Alternative EstimatesPotential Saving (Percent GDP)
Benin35%2.242–90%0.4–2
Burkina Faso16%2nana
Mali15%2.637–57%1.3–1.9
Niger21%2.3nana
Senegal8%3.829–32%2.8–2.9
Togo100%0100%0%
WAEMU average2.21.1–1.7
Note: Given the pattern of fixed cost and decreasing return to scale exhibited in the cross-country data and discussed in the literature, this is estimated as (Xi-Xmin)*(1-Ei), where X refers to spending in percent of GDP and Ei refers to the relative efficiency score for country i. This estimate represents potential savings while achieving the same level of result indicators.

To estimate potential savings in spending, input efficiency scores were calculated first and then converted to relative scores based on highest scores among WAEMU countries.

To enhance robustness, three input-output specifications were used: the baseline score was based on adult literacy and per-capita spending; alternative one was based on primary enrollment rates and per-capita spending; and alternative two was based on joint literacy and primary enrollment rates and per-capita spending. These specifications produced consistent rankings. The baseline specification has the best country coverage, while the other two cover less than half of the WAEMU countries.

Note: Given the pattern of fixed cost and decreasing return to scale exhibited in the cross-country data and discussed in the literature, this is estimated as (Xi-Xmin)*(1-Ei), where X refers to spending in percent of GDP and Ei refers to the relative efficiency score for country i. This estimate represents potential savings while achieving the same level of result indicators.

To estimate potential savings in spending, input efficiency scores were calculated first and then converted to relative scores based on highest scores among WAEMU countries.

To enhance robustness, three input-output specifications were used: the baseline score was based on adult literacy and per-capita spending; alternative one was based on primary enrollment rates and per-capita spending; and alternative two was based on joint literacy and primary enrollment rates and per-capita spending. These specifications produced consistent rankings. The baseline specification has the best country coverage, while the other two cover less than half of the WAEMU countries.

Health spending in WAEMU countries has increased significantly while the result indicators improved. Between 2003–07 and 2008–10,8 average health spending in the WAEMU increased from about 2.3 to 2.7 percent of GDP. The results indicators also improved, including an increase in life expectancy of two years and a reduction of the child mortality rate by about 3 percentage points. Figure 12.11 shows that while WAEMU countries are also below the efficiency frontier in general, a few countries moved closer to the frontier in 2008–10 (for example, Burkina Faso, Mali, Senegal), where result indicators improved faster, relative to the change in public health spending (Table 12.3).

Figure 12.11.Efficiency of Health Spending: The WAEMU and Peer Groups

Sources: World Development Indicators; IMF Fiscal Affairs Department database; and IMF staff estimates.

Note: Three-letter International Organization for Standarization abbreviations used for country names.

Table 12.3Health Spending and Result Indicators
Public Health SpendingPrivate Health SpendingLife ExpectancyChild Survival Rates
(Percent GDP)(Percent GDP)(in years)(per 1,000)1
2003–072008–102003–072008–102003–072008–102003–072008–10
Benin2.32.32.42.15759880903
Burkina Faso3.7433.45355841884
Côte d’Ivoire11.53.54.74749869886
Guinea-Bissau1.21.84.54.75253844863
Mali3.133.43.75254826862
Niger2.52.72.52.55457826872
Senegal2.73.32.92.56163900932
Togo1.93.34.345455888899
Ethiopia2.62.51.92.15761890923
Mozambique3.92.92.43.24849869899
Rwanda45.43.74.75562892935
Tanzania2.63.22.13.25459909938
Uganda2.12.16.67.15357891921
WAEMU mean2.32.73.33.553.855.6859.3887.6
Peer country mean33.23.34.153.457.6890.2923.2
Sources: World Development Indicators; IMF Fiscal Affairs Department database; and IMF staff estimates.

This indicator is derived by 1,000 minus the under-5 child mortality rate so that the rate is expected to be positively associated with health spending.

Sources: World Development Indicators; IMF Fiscal Affairs Department database; and IMF staff estimates.

This indicator is derived by 1,000 minus the under-5 child mortality rate so that the rate is expected to be positively associated with health spending.

The DEA analysis confirmed the efficiency of health spending was higher than in education, but there is still a scope for potential savings. In applying the DEA to the health sector, both public and private spending needs to be included. While basic education is generally considered a public good that should be fully supported by public spending, the health sector requires significant private spending beyond public spending to achieve the corresponding results, and thus, the DEA analysis included both sources of spending. Following the same method applied to education spending, WAEMU countries’ efficiency score ranges from 43 percent (Guinea-Bissau) to 100 percent (Senegal) in the baseline estimate. As a result, the estimated potential savings of increased efficiency of spending on health are about 0.4 to 0.8 percent of GDP on average (Table 12.4).

Table 12.4Efficiency Scores In Health Spending and Estimated Potential Savings1, 2, 3
Baseline ScorePotential Savings (Percent GDP)Scores in Alternative EstimatesPotential Saving (Percent GDP)
Benin68%0.454–70%0.4–0.6
Burkina Faso84%0.536–89%0.3–1.9
Côte d’Ivoire65%0.284–100%0–0.1
Guinea-Bissau43%0.583–100%0–0.1
Mali89%0.241–98%0.1–1.2
Niger46%0.990–99%0–0.2
Senegal100%049–50%1.2
Togo73%0.658–73%0.6–1
WAEMU average0.40.8

To estimate potential savings in spending, input efficiency scores were calculated first and then converted to relative scores based on highest scored among WAEMU countries.

To enhance robustness, three input-output specifications were used: the baseline score was based on per-capita private and public spending and child survival rate; alternative one was based on per-capita public spending and child survival rate; and alternative two was based on public and private spending in percent of GDP and child survival rate. Data coverage across three specifications is similar.

Similar to education spending, this is estimated as (Xi-Xmin)*(1-Ei), where X refers to spending in percent of GDP and Ei refers to the relative efficiency score for country i. This refers to potential savings while achieving the same level of result indicators. This estimate represents potential savings while achieving the same level of result indicators.

To estimate potential savings in spending, input efficiency scores were calculated first and then converted to relative scores based on highest scored among WAEMU countries.

To enhance robustness, three input-output specifications were used: the baseline score was based on per-capita private and public spending and child survival rate; alternative one was based on per-capita public spending and child survival rate; and alternative two was based on public and private spending in percent of GDP and child survival rate. Data coverage across three specifications is similar.

Similar to education spending, this is estimated as (Xi-Xmin)*(1-Ei), where X refers to spending in percent of GDP and Ei refers to the relative efficiency score for country i. This refers to potential savings while achieving the same level of result indicators. This estimate represents potential savings while achieving the same level of result indicators.

Improving the efficiency of public education and health spending can contribute not only to fiscal space, but also to more inclusive growth in WAEMU countries. Our analysis found that WAEMU countries have significant scope to improve the efficiency of their spending in education and health. If all WAEMU countries could achieve the highest efficiency already reached by the top country in the union, the fiscal savings are estimated to be about 1 to 3 percent of GDP. Therefore, complementary to improved tax efforts, enhancing the efficiency of spending could provide a significant contribution to WAEMU countries’ fiscal space to support the scaling up of infrastructure investments. Furthermore, education and health services are essential to enhance the well-being of WAEMU citizens and to enhance human capital and build a more productive labor force, and thus improving spending efficiency could also support more inclusive growth.

Annex 12.1. Regression Results by Tax Category
Annex Table 12.1.1Determinants of Total Tax Potential
Total Tax Revenue (Percent of GDP)LICsSSA
GDP per capita5.133***2.637
[1.571][1.879]
Inflation, consumer prices (annual percent)0.0010.002
[0.013][0.013]
Imports (percent of GDP)0.126**−0.09
[0.060][0.089]
Exports (percent of GDP)−0.0190.128
[0.058][0.102]
Agriculture (percent of GDP)−0.087*−0.166**
[0.046][0.062]
Consumption (percent of GDP)−0.0090.105
[0.052][0.093]
Gross fixed capital formation (percent of GDP)−0.0280.077
[0.058][0.098]
Urban population (percent of total)0.1260.109
[0.081][0.138]
Total natural resources rents (percent of GDP)0.02−0.015
[0.021][0.043]
M2 (percent of GDP)0.006−0.001
[0.039][0.027]
Constant−27.967*−14.466
[13.758][17.255]
Observations571707
Number of id3338
R-squared0.4450.113
R20.4350.101
Source: IMF staff estimations.Note: M = broad money as a share of GDP; id = number of countries used on the regression; R-squared = coefficient of determination; R2 = adjusted coefficient of determination.Robust standard errors in brackets.*** p < 0.01, ** p < 0.05, * p < 0.1
Source: IMF staff estimations.Note: M = broad money as a share of GDP; id = number of countries used on the regression; R-squared = coefficient of determination; R2 = adjusted coefficient of determination.Robust standard errors in brackets.*** p < 0.01, ** p < 0.05, * p < 0.1
Annex Table 12.1.2Determinants of Trade Tax Potential
Trade Tax Revenue (Percent of GDP)LICsSSA
GDP per capita−0.25−1.545
[0.739][1.375]
Inflation, consumer prices (annual percent)−0.005***−0.004
[0.001][0.003]
Imports (percent of GDP)0.050***−0.034
[0.012][0.045]
Exports (percent of GDP)0.0090.021
[0.018][0.040]
Urban population (percent of total)−0.036−0.012
[0.088][0.204]
Total natural resources rents (percent of GDP)0.019−0.008
[0.014][0.034]
Trend−0.04−0.009
[0.049][0.115]
Constant4.87817.929*
[5.736][9.827]
Observations590716
Number of id3338
R-squared0.2040.036
R20.1940.0261
Source: IMF staff estimations.Note: M = broad money as a share of GDP; id = number of countries used on the regression; R-squared = coefficient of determination; R2 = adjusted coefficient of determination.Robust standard errors in brackets.*** p < 0.01, ** p < 0.05, * p < 0.1
Source: IMF staff estimations.Note: M = broad money as a share of GDP; id = number of countries used on the regression; R-squared = coefficient of determination; R2 = adjusted coefficient of determination.Robust standard errors in brackets.*** p < 0.01, ** p < 0.05, * p < 0.1
Annex Table 12.1.3Determinants of Income Tax
Income Tax Revenue (Percent of GDP)LICsSSA
GDP per capita1.964***2.370***
[0.592][0.626]
Agriculture (percent of GDP)−0.015−0.035*
[0.020][0.020]
Consumption (percent of GDP)0.027***0.025*
[0.007][0.013]
Gross fixed capital formation (percent of GDP)0.024**0.01
[0.010][0.013]
Urban population (percent of total)0.0450.091**
[0.026][0.035]
Total natural resources rents (percent of GDP)0.023**0.006
[0.010][0.014]
M2 (percent of GDP)0.019***0.023**
[0.007][0.009]
Public wage bill (percent of GDP)0.196***−0.000***
[0.042][0.000]
Constant−15.962***−19.051***
[4.419][5.101]
Observations461629
Number of id2535
R-squared0.4620.201
R20.4520.191
Source: IMF staff estimations.Note: M = broad money as a share of GDP; id = number of countries used on the regression; R-squared = coefficient of determination; R2 = adjusted coefficient of determination.Robust standard errors in brackets.*** p < 0.01, ** p < 0.05, * p < 0.1
Source: IMF staff estimations.Note: M = broad money as a share of GDP; id = number of countries used on the regression; R-squared = coefficient of determination; R2 = adjusted coefficient of determination.Robust standard errors in brackets.*** p < 0.01, ** p < 0.05, * p < 0.1
Annex Table 12.1.4Determinants of Goods and Services Tax
Goods and Services Tax Revenue (Percent of GDP)LICsSSA
GDP per capita3.956***4.415***
[1.370][1.270]
Inflation, consumer prices (annual percent)0.0060.006
[0.008][0.006]
Agriculture (percent of GDP)0.004−0.026
[0.037][0.036]
Gov. consumption (percent of GDP)0.0140.023
[0.069][0.046]
Household consumption (percent of GDP)−0.0380.042
[0.054][0.037]
Gross fixed capital formation (percent of GDP)−0.0420.003
[0.054][0.034]
Urban population (percent of total)0.157**0.129*
[0.072][0.064]
M2 (percent of GDP)0.0130.023
[0.016][0.019]
Imports (percent of GDP)0.0810.021
[0.060][0.036]
Exports (percent of GDP)−0.0430.021
[0.063][0.035]
Constant−26.563**−37.550***
[11.500][9.178]
Observations571698
Number of id3338
R-squared0.3490.397
R20.3380.388
Source: IMF staff estimations.Note: M = broad money as a share of GDP; id = number of countries used on the regression; R-squared = coefficient of determination; R2 = adjusted coefficient of determination.Robust standard errors in brackets.*** p < 0.01, ** p < 0.05, * p < 0.1
Source: IMF staff estimations.Note: M = broad money as a share of GDP; id = number of countries used on the regression; R-squared = coefficient of determination; R2 = adjusted coefficient of determination.Robust standard errors in brackets.*** p < 0.01, ** p < 0.05, * p < 0.1
Annex 12.2. Stochastic Frontier Analysis

The stochastic frontier model of Pessino and Fenochietto (2010) can be represented as follows:

Where a represents a set of country specific intercepts, X is the vector that represents variables affecting tax revenue. The error term ε is a composite error term made of the standard component v and of a component u, which is distributed following a probability density function that is positively definite. The element u is the time varying element, which represents the degree of inefficiency: higher values correspond to higher inefficiencies. Similarly to the panel regression, four different models are estimates, each for one type of tax revenue. Starting from a general model for the estimation of the determinants of tax revenue to GDP, the model is then modified slightly to exclude variables that are not supposed to determine the behavior of some of the sub-categories of revenue.

References

    BelhocineNazim. 2013. “Assessing Efficiency of Public Spending on Health and Education in Iceland. In Iceland: Selected Issues. International Monetary FundWashington.

    • Search Google Scholar
    • Export Citation

    CalderónCésar. 2009. “Infrastructure and Growth in Africa.Policy Research Working Paper 4914World BankWashington.

    Commission for Africa. 2005. Our Common Interest: Report of the Commission for Africa. London: Commission for Africa.

    DavoodiH. and GrigorianD.2007. “Tax Potential vs. Tax Effort: A Cross-Country Analysis of Armenia’s Stubbornly Low Tax Collection.Working Paper 07/106International Monetary FundWashington.

    • Search Google Scholar
    • Export Citation

    Domínguez-TorresCarolina and VivienFoster. 2011. Benin’s Infrastructure: A Continental Perspective. Policy Research Working Paper 5689World BankWashington

    • Search Google Scholar
    • Export Citation

    FosterVivien and CeciliaBriceño-Garmendia eds. 2009. Africa’s Infrastructure: A Time for Transformation. Paris and Washington: Agence Française de Développement and World Bank. http://www.infrastructureafrica.org.

    • Search Google Scholar
    • Export Citation

    GrigoliFrancesco and JavierKapsoli. 2013. “Waste Not, Want Not: The Efficiency of Health Expenditure in Emerging and Developing Economies.Working Paper 13/187International Monetary FundWashington.

    • Search Google Scholar
    • Export Citation

    GuptaAbhijit Sen. 2007. “Determinants of Tax Revenue Efforts in Developing Countries.Working Paper 07/184International Monetary FundWashington.

    • Search Google Scholar
    • Export Citation

    HerreraSantiago and GaoboPang. 2005. “Efficiency of Public Spending in Developing Countries: An Efficiency Frontier Approach.Policy Research Working Paper No. 3645.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2013a. “Benin: 2012 Article IV Consultation and Fourth Review under the Extended Credit Facility Arrangement.Staff Report Côte d’IvoireWashington.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2013b. “Côte d’Ivoire: 2013 Article IV Consultation and Fourth Review Under The Extended Credit Facility Arrangement.Washington.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2013c. “Mali: 2012 Article IV Consultation, Request for Disbursement Under the Rapid Credit Facility and Cancellation of the Extended Credit Facility Arrangement.Staff ReportWashington.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2013d. Regional Economic Outlook: Sub-Saharan Africa. Washington.

    PessinoCarola and RicardoFenochietto. 2010. “Determining Countries’ Tax Effort.Hacienda Pública Española /Revista de Economía Pública195 (4/2010): 6587.

    • Search Google Scholar
    • Export Citation

    TorresClemenciaCecilia M.Briceño-Garmendia and CarolinaDominguez. 2011. Senegal’s Infrastructure: A Continental Perspective. Policy Research Working Paper 5817World BankWashington.

    • Search Google Scholar
    • Export Citation
1

Excludes upper middle-income countries of South Africa, Botswana, Nigeria, Angola, and Namibia. The low-income countries refer to the World Bank definition of countries with a per-capita GNI of $1,025 or less in 2012.

2

A positive gap signals that tax revenue collection is above potential. A negative gap implies that tax revenue collection falls short of the potential.

3

Regarding the determinants of trade tax potential, we used different models, such as incorporating a proxy for the openness or using import/GDP as well as export/GDP separately.

4

These are countries with top growth performance in sub-Saharan Africa, as discussed in IMF (2013d): Ethiopia, Mozambique, Rwanda, Tanzania, and Uganda.

5

The DEA method has been used in a recent analysis on the efficiency of public spending in Iceland and in cross-country studies, such as Belhocine (2013) and Grigoli and Kapsoli (2013).

6

Data may not be available for all WAEMU countries, and this limits the coverage of the efficiency and saving estimates.

7

Similar analysis using education spending as a share of GDP and the corresponding result indicators yields consistent results.

8

Due to data constraints, the latest period is limited to 2008–10.

    Other Resources Citing This Publication