Chapter

Appendix 2. Assessing Potential Revenue: Two Approaches

Author(s):
International Monetary Fund. Fiscal Affairs Dept.
Published Date:
October 2013
Share
  • ShareShare
Show Summary Details

The main text reports on two rather different ways of assessing revenue potential, giving complementary perspectives on the scope to raise more.

Peer analysis

Peer analysis, the most traditional approach, models revenue ri in country i (in percent of GDP) as a function68

of observable characteristics xi (such as income per capita, with a very wide range of other variables explored in the literature). The “potential” for additional revenue is then the fitted residual, εi, which, by construction, averages to zero over the sample.

Torres (2013) extends this method by applying it to subcategories of revenue. For a cross-section of 164 countries, using data constructed from IMF reports (World Economic Outlook, Article IV staff reports, and revisions to ongoing programs), revenues are divided into those from income taxes, payroll taxes, other taxes, taxes on goods and services, taxes on international trade, grants, and nontax revenues. To calculate the revenue gaps, taxes on international trade, grants, and nontax revenues are excluded, as these are somewhat less under the government’s direct control. Control variables include per capita income, the old-age dependency ratio, and political participation, with revenues increasing in all three.

Table A.2.1 reports the estimated potential for additional revenue for selected advanced and emerging market economies and low-income countries; negative values indicate that observed revenues exceed predicted ones. There is quite a wide variation within each income group, with substantial implied scope to increase total revenue in some countries but little in others. The breakdown by tax category provides useful pointers as to where the most evident potential lies—generally consistent with the views in IMF (2010a). For example, in Germany and Mexico, VAT revenues could be enhanced by eliminating reduced VAT rates, and in Japan by increasing (as planned) the consumption tax rate. Along with Korea, Japan also raises less from the personal income tax than do its peers.

Table A.2.1.Revenue Gaps(Percent of GDP)
TotalConsumption TaxesIncome TaxesPayroll TaxesOther Taxes
Advanced economies
Japan17.89.03.25.8−0.1
Switzerland9.52.63.14.0−0.2
Korea7.43.92.71.1−0.3
United States6.13.71.21.3−0.1
Singapore5.44.1−0.32.9−1.3
Greece4.52.02.81.0−1.3
New Zealand4.2−1.0−4.68.11.7
Canada3.32.9−1.63.6−1.6
Germany3.12.50.9−1.41.0
Spain2.74.40.0−1.5−0.2
Portugal2.1−0.6−0.20.91.9
Estonia1.70.41.1−0.30.4
Ireland1.50.1−0.10.11.5
United Kingdom0.70.7−2.14.7−2.5
Italy0.74.9−4.72.0−1.5
Emerging market economies
Latvia10.13.81.24.60.5
Bulgaria8.9−0.13.06.1−0.2
Kazakhstan5.94.31.10.6−0.1
Mexico5.93.12.6−1.01.2
Lithuania5.12.12.9−1.11.2
Indonesia5.03.00.41.60.1
Saudi Arabia4.51.32.30.30.6
Thailand3.91.2−0.33.00.0
Jordan1.9−1.92.80.90.2
Egypt1.01.7−0.5−1.00.9
Low-income countries
Sudan8.52.64.20.71.1
Madagascar8.53.73.70.70.4
Haiti5.23.61.61.0−0.9
Yemen4.61.62.30.40.3
Nepal4.31.32.40.8−0.3
Armenia4.22.8−0.42.4−0.6
Cambodia4.10.92.00.60.6
Georgia3.6−1.3−3.98.40.4
Côte d’Ivoire3.53.92.2−1.0−1.6
Chad3.31.91.40.4−0.4
Uganda3.2−0.42.30.50.8
Ghana1.01.5−1.70.70.6
Congo, Rep. of1.0−0.71.10.50.0
Source: IMF staff estimates.

Stochastic frontier analysis

Stochastic frontier analysis69 instead models revenue potential explicitly, taking revenue to be a function

where M denotes maximum revenue, dependent on observables exogenous to policy, and U denotes “effort,” lying between 0 and 1 and depending on variables that are, to at least some degree, choice variables, as well as on wider social preferences. Put most simply, peer analysis finds the best fit to the observations, whereas stochastic frontier analysis aims to put a frontier around them (Figure A.2.1).70 The stochastic frontier analysis approach has the considerable advantage of not inherently implying that some countries are raising more than their “potential” and fits neatly into the conceptual framework for gap assessment in “Finding, and Minding, the Gap” in Section 2 (with effort reflecting rate choices, policy gaps, and compliance gaps). A weakness in applications so far is that relatively little attention has been paid to the determinants of effort.

Figure A.2.1.Peer and Stochastic Frontier Analysis Estimation of Tax Potential

Source: IMF staff estimates.

Results using the same data set and controls as Torres (2013) and—in the absence of good measures of, for instance, the breadth of tax bases—treating z i as unobserved71 are presented in Table A.2.2. With a few notable exceptions (such as Greece), results are in line with priors and previous estimates (IMF, 2011).72 They are highly positively correlated to the peer analysis gap estimates presented previously (as in Cyan, Martinez-Vasquez, and Vulovic, 2013). These results show that

  • Countries with similar revenue levels can have very different levels of effort. This is the case for Ireland and Switzerland, for example, and for Armenia, Nicaragua, and Mozambique.

  • There are wide variations across countries, but average effort is fairly similar across advanced and emerging market economies and low-income countries.

  • Estimated tax efforts are consistent with priors on social preferences: Denmark and Norway, for instance, figure among those with the highest effort.

Table A.2.2.Estimated Tax Effort, 2012
Tax Revenue1Tax Effort2Tax Revenue1Tax Effort2Tax Revenue1Tax Effort2
Advanced economiesEmerging market economiesLow-income countries
Switzerland28.50.52Saudi Arabia1.10.05Madagascar10.90.33
Korea19.30.48Kazakhstan12.40.39Sudan6.10.34
Estonia32.80.55Latvia25.50.43Cambodia11.00.39
Singapore13.90.55Bulgaria26.80.47Chad5.50.40
Germany40.00.57Lithuania27.90.51Haiti12.70.40
Sweden44.20.62Mexico15.70.50Ghana17.10.46
Ireland27.80.74Peru18.00.63Nepal13.10.49
Japan30.00.43Jordan15.00.64Moldova31.90.66
Israel34.00.75Philippines15.30.69Uganda12.20.57
Slovak Republic29.00.78Thailand17.90.63Armenia20.50.53
Netherlands39.20.75Malaysia16.10.72Tanzania16.10.64
United States25.10.61Romania28.30.72Georgia25.20.53
Austria44.10.73Poland33.20.77Cameroon13.80.71
Iceland36.30.80Turkey26.70.90Nicaragua21.40.72
Spain33.10.71Ukraine40.00.76Congo, Rep. of8.70.70
Finland43.80.75Chile21.60.69Bolivia20.60.71
New Zealand29.50.62Egypt15.80.72Zambia17.80.74
Slovenia36.60.75Russia35.00.85Lao P.D.R.16.20.78
United Kingdom35.50.75Hungary38.40.79Yemen6.80.73
Czech Republic35.00.79South Africa24.20.89Congo, Dem. Rep. of the16.70.77
Italy44.20.68Colombia22.20.91Honduras17.60.76
Canada30.20.67Argentina36.20.87Côte d’Ivoire17.60.75
Portugal34.90.74Morocco24.10.93Mozambique21.00.78
Norway43.20.91Nigeria16.40.94Burkina Faso14.90.81
Denmark49.70.86Brazil29.60.96Mali17.30.88
France44.70.85Senegal19.70.88
Belgium46.20.85
Greece35.50.80
Average35.20.7023.30.6915.90.63
Source: IMF staff estimates.

What these results do not shed light on, however, is precisely how effort can be increased. The results in Torres (2013) are somewhat more informative on this point, but would require considering country specifics of both design and implementation.

    Other Resources Citing This Publication