A. Episodes of Tax Revenue Mobilization

In the first step, we identify episodes of rapid increases in tax revenue mobilization that might be driven by fiscal policy decisions and reforms. It may of course be that some observed tax increases are be driven by the business cycle or shifts in the composition across economic goods. We therefore adjust the tax revenue ratio to represent only its structural component, taking out the cyclical element by applying an output elasticity to tax revenue of one, as common in the literature.¹⁰ The initial identification thus takes the ratio of the cyclically adjusted tax revenue-to-trend GDP as a baseline.

View Full Size

Distribution of 172 Identified Episodes

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Sources: WoRLD and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Episodes are defined as 3-year compounded windows of an average increase of 0.5 percent per year or more. The first episode is triggered by the first 3-year window in which the tax-to-GDP ratio increases by an average of at least 0.5 percent per year (1.5 percent within three years). The threshold of 0.5 percent of GDP per year is often set as a target in IMF supported programs (e.g. Bangladesh 2004 and Madagascar 2016) or by governments.¹¹ It was also suggested that to build stronger institutions and support growth, countries should aim to achieve revenue gains of 0.5 percent of GDP per year.¹² Following the initial window, successive three-year windows (including the second and third years of the previous window) for which this criterion continues to be satisfied are compounded into the same episode. The end-point of an episode is the last overlapping window fitting this criterion. From the sample of 136 countries over 2000-2015, we identify 172 episodes that fit this criterion. Of those, 61/30 are observed in non-RR/RR EMs, and 65/16 in non-RR/RR LICs. Episodes are thus relatively evenly spread over the two income groups. The RR results are robust to using non-oil GDP as the denominator rather than nominal GDP. In Section V, when we analyze the underlying policy choices, we narrow the 172 episodes down to those with more tangible tax revenue mobilization results using two additional episode qualifiers.

B. Main Characteristics of Tax Revenue Mobilization Episodes

We begin with a general characterization of all episodes during which large tax revenue increases were observed.

Duration of episodes. Figure 6 plots the length of the identified episodes. The average episode duration is 3.5 years in RR countries and 3.4 years in non-RR countries, with a median duration of 3 years across all countries. Episodes can also be shorter than three years if the entire 1.5 percent increase (or more) occurred within one or two years only. Indeed, 16 percent of episodes lasted only one year. Notably, 23 percent of the identified episodes lasted for longer than 5 years, implying that some countries registered a sustained increase over a significantly longer period.

View Full Size

Episode Duration

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Note: Episodes are defined as 3-year compounded windows of an average increase of 0.5 percent per year or more. The first episode is triggered by the first 3-year window in which the tax-to-GDP ratio increases by an average of at least 0.5 percent per year (1.5 percent within three years).Sources: WoRLD, WEO, and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Initial tax ratios: Episodes of tax revenue mobilization occurred with both low and high initial tax-to-GDP ratios (Figure 7). Figure 7 presents the initial tax ratio also in relation to the “tipping point” of tax revenue, around 13 percent of GDP, after which countries tend to experience higher and sustained economic growth, as defined in Gaspar et al. (2016a) and Gaspar et al. (2016b). For LICs and RR countries, most large increases began while the tax-to-GDP/non-commodity revenue-to-GDP ratio was below 15 percent. For example, while not sustained, Afghanistan increased its tax-to-GDP ratio by 3.3 percent within two years, from an initial ratio of 5.6 percent. Liberia increased its tax-to-GDP ratio significantly from a level of 11.4 percent to above 16 percent within two years and succeeded in sustaining the increase. EMs frequently registered larger increases starting beyond a ratio of 15 percent, which reflects the higher average tax ratios in these countries. It thus seems that revenue mobilization can be achieved from a wide range of initial revenue ratios, although it is more frequently observed from a lower starting point.

View Full Size

Initial Revenue Ratios at the Onset of Successful Revenue Mobilization

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Note: Vertical lines represent the “tipping point” of tax revenue to GDP in Gaspar et al. (2016a) and Gaspar et al. (2016b), which finds that countries which achieve a tax-to-GDP rate of at least 12.75 experience sustained economic growth. Even past this tipping point, the data displays room from improvement.Sources: WoRLD, WEO, and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Sustainability of end-of-episode tax ratios: Sustainability of the achieved revenue gain is a crucial element of any tax reform. Figure 8 shows that revenue increases were sustained/increased in slightly less than one third of the cases, while it fell by up to one third of the initial tax revenue increase in close to one third of the cases. In more than one third of the cases the gains were not sustained, which is especially the case in RR countries. In the majority of cases, the decline in tax revenue at the end of the episode was explained by macroeconomic shocks (internal or external, above all the global financial crisis) and/or revenue-reducing policy measures.¹³ For example, Moldova’s tax revenue fell by about 1.3 percent in 2008 following a deep recession caused by the global financial crisis (GFC). In Antigua and Barbuda, and Dominica, the effects from the GFC were emphasized by a granted temporary tax relief and a reduction in personal income tax rates.

View Full Size

Short-Term Sustainability of Revenue Increase

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Notes: Sustained indicates that revenue increased or stayed the same for at least three years after the revenue period ended (two years for episodes ending in 2015). Fell moderately indicates revenue fell by less than one third of the overall increase. Fell strongly indicates a fall of more than one third of the overall increase.Sources: WoRLD, WEO, and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

C. Robustness: “Top Revenue Performers” Based on Income Levels

Some of the above tax-to-GDP increases might have been driven by temporary economic or structural factors, rather than underlying revenue administration or tax policy changes. To complement the in-depth analysis of tax policy and revenue administration measures by controlling for GDP and output, we employ a second strategy for all countries, following von Haldenwang and Ivanyna (2012). This strategy relates countries’ tax-to-GDP ratios to their development levels, as measured by their GDP per capita. Since the relationship likely varies across income groups, we split the analysis in line with the income categories used in the previous sections.¹⁴

For each income group (LICs and EMs, excluding RR economies) we regress the 2015 tax-to-GDP ratio on logged GDP per capita. We then categorize the countries into four groups: (i) “High tax performers” – countries lie above the 85^th percentile of the trend line (fitted values); (ii) “Good tax performers” – countries are within the 50^th and 85^th percentiles; (iii) “Low tax performers” – countries fall within the 15^th and 50^th percentiles; and (iv) “Poor tax performers” – countries below the 15^th percentile. We then plot the 2009 tax-to-GDP ratio/logged GDP per capita combinations against the 2015 categories to examine if there have been changes over time.

Between 2009 and 2015, some countries have experienced significant changes in their tax performance, and moved categories in an upward direction (Figure 9). Overall, 18 out of 79 countries registered an improvement. The Maldives jumped two categories, from poor to good collection relative to peers. Of those 18 identified cases, all except for six (Albania, Bahamas, Jamaica, Philippines, Senegal, and Turkey) were also identified under the main identification strategy. The method is robust to varying the baseline years of the analysis. This strategy complements the first strategy as it takes a longer view and moves away from the across-the-board criterion regardless of the level of development. For instance, some countries may have fallen slightly short of the 0.5 percentage point improvement in their tax-to-GDP ratios in three consecutive years, but may still have consistently improved their ratios relative to their peers at the same level of development. These cases deserve consideration as “top tax performers” and may yield valuable lessons for countries struggling to improve their tax ratios. We add the six countries that were not in the original sample to the in-depth analysis to capture all countries that improved their tax performance.¹⁵

View Full Size

Comparison of Tax Revenue to GDP per Capita Performance Matrix, 2009 vs. 2015

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Notes: Comparisons made to line of best fit based on 2015 values by country group. The residual from the 2015 line of best fit is compared to the percentile. Below the 15^th percentile is classified as “Poor”, between the 15^th and 50^th percentile is “Low”, from 50^th to 85^th is “Good” and above 85^th is “High.”Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

A. Key Findings – Tax Policy Measures

Large tax revenue increases often coincide with higher collection of revenue from indirect taxation. Figure 10 shows the frequency of the different tax policy measures, presented as the ratio of the observations in a specific category to total identified tax policy measures. Changes to indirect taxation make up more than 56 percent of all discretionary tax policy changes. This result mirrors the overall increase in indirect tax ratios in both LICs and EMs. Changes to income taxes account for another 26 percent.

Countries rely on a wide range of tax policy instruments in their revenue mobilization reforms. As can be seen in Figure 11, most of the country-episodes used multiple tax instruments (“0” tax policy instruments indicate revenue administration was the primary reform-driver). For example, Belize, 2013-2014 (GST and excise taxes), Kyrgyz Republic, 2011-12 (CIT, VAT, excises, and other GST), and Seychelles, 2008-11 (PIT, CIT, VAT, excises, other GST, trade taxes and subsidies) all used two or more tools in parallel.

Rate changes and reduction of exemptions are leading the tax policy mix. We split the tax policy measures into tax introductions (“new taxes”), rate changes (positive or negative), threshold changes (increases or reductions) and changes to exemptions, the latter two falling under base broadening. Appendix IV provides a detailed breakdown per country, summarized in Table 2 and Figure 12. Rate changes are most common (85 out of 187 observations, excluding subsidies), driven by those to excise taxation. Reductions in exemptions, above all to VAT, CIT, and other goods and services taxes, are the second most frequent tax policy tool to mobilize revenue. Threshold changes are not common during the identified episodes.

Rate increases are observed primarily for indirect taxes, while rate changes to PIT and CIT were primarily negative (Table 2). 7 out of 10 rate changes were negative for both PIT and CIT during the identified episodes (see also Appendix IV), partly aimed at stimulating economic activity, and primarily following international trends, especially for CIT rates. Often, these reductions were balanced by tax increases in indirect tax categories (for example in The Gambia, Liberia, and Seychelles). Similarly, half of the identified 8 threshold adjustments were increases (positive direction in Table 2). The impact from threshold reductions is not necessarily positive, and changes are often aimed at re-distributional and efficiency concerns rather than revenue mobilization.

Broadening the tax base, which has sometimes involved reducing the number of tax rates, was frequently part of revenue mobilization strategies. For example, Georgia simplified the tax system by reducing the number of taxes to be paid from 21 to 7. Targeting the tax base more directly, Morocco reduced the number of VAT special regimes, Jamaica introduced VAT to residential electricity consumption, and Seychelles converted the narrowly applied 7 percent GST into a broad-based 12 percent GST.

View Full Size

Frequency of Tax Policy Instruments

(in percent)

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Note: Ratio of observations in a specific category to total identified discretionary measures.Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Number of Tax Policy Instruments Used Per Episode

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Table 2.

Introductions, Rate and Base Changes

Note: General GST (sales taxes) excludes VAT and excise tax. Positive rate and threshold changes describe an increase in the tax rate or threshold value, respectively. Columns add up to more than one if positive and negative adjustments were made in the same episode. Source: IMF staff estimates.

Table 2.

Introductions, Rate and Base Changes

Tax	Measure	Observations	Positive Change	Negative Change
PIT	New	5	5	0
	Rate	10	4	7
	Threshold	2	2	0
	Exemptions	4	4	0
CIT	New	5	5	0
	Rate	10	3	8
	Threshold	1	0	1
	Exemptions	11	11	0
General GST	New	8	8	0
	Rate	9	8	1
	Threshold	1	0	1
	Exemptions	11	11	0
VAT	New	10	10	0
	Rate	6	3	3
	Threshold	4	2	2
	Exemptions	16	16	0
Trade	Rate	13	13	6
	Exemptions	7	7	0
Excise	New	6	6	0
	Rate	35	34	5
	Exemptions	7	7	0
Property	New	3	3	0
	Rate	2	1	2

Note: General GST (sales taxes) excludes VAT and excise tax. Positive rate and threshold changes describe an increase in the tax rate or threshold value, respectively. Columns add up to more than one if positive and negative adjustments were made in the same episode. Source: IMF staff estimates.

View Table

Table 2.

Introductions, Rate and Base Changes

Tax	Measure	Observations	Positive Change	Negative Change
PIT	New	5	5	0
	Rate	10	4	7
	Threshold	2	2	0
	Exemptions	4	4	0
CIT	New	5	5	0
	Rate	10	3	8
	Threshold	1	0	1
	Exemptions	11	11	0
General GST	New	8	8	0
	Rate	9	8	1
	Threshold	1	0	1
	Exemptions	11	11	0
VAT	New	10	10	0
	Rate	6	3	3
	Threshold	4	2	2
	Exemptions	16	16	0
Trade	Rate	13	13	6
	Exemptions	7	7	0
Excise	New	6	6	0
	Rate	35	34	5
	Exemptions	7	7	0
Property	New	3	3	0
	Rate	2	1	2

Note: General GST (sales taxes) excludes VAT and excise tax. Positive rate and threshold changes describe an increase in the tax rate or threshold value, respectively. Columns add up to more than one if positive and negative adjustments were made in the same episode. Source: IMF staff estimates.

View Full Size

Frequency of Introductions, Rate and Base Changes

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

Curbing tax exemptions was a prominent measure. In more than half of the country episodes, various exemptions were eliminated, most frequently related to VAT, general GST, and CIT. Specific measures include reducing statutory and discretionary exemptions (for example in Guyana and Solomon Islands); ending tax holidays (Burkina Faso); ending CIT exemptions (Mauritania); and eliminating VAT exemptions (Uganda).

VAT and other general goods and services taxation (sales taxes) were a common measure for revenue mobilization. In some cases, it was observed that substantial increases in tax revenue resulted from the introduction of VAT (Bahamas, Bosnia and Herzegovina, Dominica, The Gambia, and Guyana), and after rate hikes (for example Barbados and Tuvalu). In 48 percent of episodes, changes to VAT (introduction, rate changes, thresholds or exemptions) were an element of the revenue mobilization strategy, and 29 percent of episodes made use of changes to other general goods and services taxation.

Increases in excise taxation were frequently applied. Excises are easy to implement and often used for immediate revenue needs or to balance revenue shortfalls from other reforms. In our sample, higher excise taxes were often levied on fuels, tobacco, alcoholic and non-alcoholic drinks, and cars. In other cases, increases in excise taxes were imposed on the telecom sector (for instance, Guyana and The Gambia). Changes to excises often accompanied other tax changes and did not generate large tax revenue changes by themselves.²⁰

Taxing sector-specific domestic economic rents (e.g., tourism, telecom, oil), while less frequent, helped some countries to boost tax revenue. This strategy targets sectors in which profits have an element of location-specificity that can imply high potential for revenue mobilization. Examples include the Maldives (tourism sector), and Malawi and Rwanda (telecom).

Property taxation, and the reduction of subsidies were least frequently observed. Property taxation has not been a common tool to mobilize revenue, being observed in only 3 percent of cases, likely because it has not been broadly implemented in many developing countries. Similarly, legal loopholes (for instance, the absence of cadaster to define property rights) and weak tax administration capacity could limit its application as a policy tool, especially at the local level to which it is often allocated. At 4 percent of observations, the reduction of subsidies was infrequent, but demonstrates revenue raising potential from targeting inefficient subsidies.

B. Key Findings – Revenue Administration Measures

The analysis identifies several RA measures that were implemented during the tax reform periods. Given the difficulty of measuring the impact of revenue administration measures, which often impact multiple revenue sources, this paper does equally not attempt to quantify their impact on tax collection, nor to empirically establish a causal link between such measures and the outcome.²¹ Based on international experiences, however, it is broadly accepted that tax administration measures that facilitate and enforce tax compliance will help reduce the tax gap and thereby increase the tax yield (Brondolo et al. 2008).²² With this understanding, we highlight key revenue administrative measures which were implemented during the reform periods (when tax collection increased significantly).

Revenue administration measures have been used widely across many episodes. As shown in Table 1, Figure 13, and Appendix V, revenue administration measures were common in almost all (90 percent) of the episodes, regardless of a country’s income group and the range of tax policy measures. In some cases (for example Sierra Leone and Republic of Congo), the authorities focused only on administrative measures during the tax reform period.

Steps to improve audits and other verification measures were undertaken in most countries that achieved an increase in tax revenue. 89 percent of the episodes that involved an improvement in revenue administration targeted changes in audits and verification programs (Figure 13), with a focus on administering audits, combatting tax evasion and reducing corruption. Evidence has shown that strengthening the auditing and enforcement capacity remains crucial for effective tax collection in developing countries (Carrillo et al. 2017). The findings in this study show that improvements in tax revenue collection in Georgia and Guinea coincided with intensified tax and customs audits.

Many countries focused on strengthening management, governance, and human resources (HR) to enhance their tax performance. Hiring more qualified staff, strategic planning and monitoring performance, focusing on training and development to strengthen technical skills of staff at tax and customs offices, and strengthening tax legislation to empower revenue collection agencies were commonly used measures (77 percent of the episodes). In some cases, their tax-to-GDP ratios increased following the setting up and empowering of a semi-autonomous revenue authority (e.g., Jamaica). This result is supported by Ebeke et al. (2016), who find that semi-autonomous revenue agencies have on average had a large and positive effect on non-resource taxes. However, establishing semi-autonomous agencies alone will not guarantee a revenue impact, especially if capacity and staffing are insufficient (see also Kidd and Crandall, 2006).

View Full Size

Revenue Administration Measures

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

In more than half of the episodes improvements in information technology (IT) systems provided support for core revenue administration processes. In these cases (for example in Burkina Faso, Tonga, Guinea-Bissau, and Uruguay), tax collection increased following the implementation of an IT strategic plan, the enhancement of the IT system, the computerization of tax and customs administration to support tax payments, and filing and use of third-party information. Russell (2010) noted that IT is important to build capacity in core tax administration functions, such as registration, filing and payment enforcement, and it supports compliance management. Also, the introduction of an electronic system could support tax collections by accelerating the processing of tax payments. However, as noted in IMF (2015), despite significant investment into IT, many developing countries still struggle with successfully automating basic compliance operations, and IT improvements will need to go hand in hand with advances in other administrative procedures to be fully beneficial.

Several countries employed measures to strengthen the registration process and improve filing. Improvements in tax revenue were supported by an increase in the number of registered taxpayers (Central African Republic), and in some cases re-registering taxpayers (Georgia, Guinea-Bissau). Several countries focused on ensuring timely filing and payment of taxes, and enforcing sanctions for failure to file.

Several countries improved voluntary compliance and strengthened the customs clearance process during the tax reform period. For instance, in Tonga, the increase in revenue coincided with strengthening tax compliance and reorganizing the one-stop shop for customs tariff payments. Guinea-Bissau strengthened customs administration by improving cross-border controls.

Some countries improved the management of payment obligations and strengthened segmentation (often focusing on large taxpayers). Developing a policy to limit, manage, and recover tax arrears contributed positively to tax revenue performance in several countries (Burundi, Comoros, Dominica). Some countries implemented a large taxpayer unit (LTU) (Cabo Verde, Uruguay) and strengthened management of large taxpayers to ensure timely filing and payment of taxes. Large taxpayers can account for as much as 80 percent of all tax revenue, therefore strengthening procedures targeted at them is seen as a critical reform step.²³

Countries often relied on an array of revenue administration instruments in their mobilization efforts. Most of the revenue administration measures are interlinked and closely associated, therefore a more comprehensive approach with multiple revenue administration instruments could enhance the results if they are successfully implemented. In more than 30 percent of the country-episodes, 4-5 elements of revenue administration were addressed in parallel (Figure 14). For example, in Paraguay the authorities streamlined the auditing and collection process, improved the efficiency of the customs clearance process, and adopted measures to reduce corruption. The anti-corruption driven revenue mobilization strategy in Georgia (2003-2008) also benefitted from a range of administrative measures, including the mass-replacement and training of staff, strengthening of legal procedures for tax disputes, upgrading of IT, simplifying of tax returns and improvement of audits (Akitoby 2018 and Akitoby et al. 2018a).²⁴

View Full Size

Number of Revenue Administration Instruments Used

Citation: IMF Working Papers 2018, 234; 10.5089/9781484361535.001.A001

Source: IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

In some cases, the initial success following tax policy measures was eroded in the absence of sustained reform efforts for revenue administration. For example, Antigua and Barbuda increased its tax-to-GDP ratio temporarily, helped by the introduction of a sales tax. However, most of the other tax categories declined during this time, amid limited improvements in tax and customs administration. Similarly, in The Gambia, the introduction of VAT was a positive contribution to revenue performance, but the limited capacity to enforce compliance caused VAT and other tax revenue growth to slow down, especially between 2015 and 2016. In general, almost all of the countries that sustained the increase have undertaken reforms to revenue administration, often in combination with other measures (see Table 1 and Appendix V).