ent al s u nd a m PFR F AN CY B UOY TA X PFR Fundamentals TAX BUOYANCY MAY 2025 World Bank Economic Policy Global Department Fiscal Policy © 2025 The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved This work is a product of The World Bank. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. The boundaries, colors, denominations, links/footnotes and other information shown in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. The citation of works authored by others does not mean the World Bank endorses the views expressed by those authors or the content of their works. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Attribution—Please cite the work as follows: “World Bank. 2025. PFR Fundamentals: Tax Buoyancy. © World Bank.” Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e mail: pubrights@worldbank.org. Contents About the Notes and Acknowledgments iv Section 1 Introduction 2 Section 2 Theoretical Framework and Empirical Approaches 5 2.1. Theoretical Approach 5 2.2. Empirical Approaches 7 2.3. Literature Review 8 Section 3 Empirical Approach for Estimating Tax Buoyancy 10 3.1. Data 10 3.2. Estimation Approach 11 Section 4 Results 12 4.1. Country-Income Group and Regional Tax Buoyancy Estimates 12 4.2. Robustness Check of Panel Estimates with Controlling for Inflation 17 4.3. Controlling for Tax Rates 20 4.4. Business Cycle 22 4.5. Country-Level Estimates of Tax Buoyancy 22 4.6. Conclusions 30 References 31 iii About the Notes and Acknowledgments PFR Fundamentals is a series of analytical and how-to notes prepared by the Fiscal Policy Unit to assist task teams in preparing and implementing Public Finance Reviews (PFRs). This note was prepared by Violeta Vulovic. Overall guidance was provided by Emilia Skrok and Tuan Minh Le. Franz Ulrich Ruch and John Nana Darko Francois provided helpful comments and inputs during the preparation process. The peer reviewers were Cristina Savescu, Barbara Cunha, Chadi Bou Habib, Rafael Chelles Barroso, and De- sislava Enikova Nikolova. iv Section 1  Introduction M edium-term fiscal sustainability relies on accurate projections of rev- enue mobilization. Assessing whether a government aligns tax mobili- zation with economic activity is important for prudent spending choices. Revenue projections depend on an assumption on how the revenues will respond to changes in economic growth in the future, and the two concepts of tax elasticity and tax buoyancy are most used to assess this. While closely related, these two concepts are different and produce different outcomes. The objective of this note is to recommend a methodological approach for country economists to estimate buoyancy when analyzing a country’s tax systems, for example when writing a Public Finance Review (PFR). The note presents theoretical under- pinnings and describes alternative econometric approaches and data for estimating short-run and long-run tax buoyancy. Results and practical implementation of the proposed methodology are displayed. Tax elasticity measures an isolated response of tax revenues to changes in economic growth (i.e., automatic changes), assuming everything else, including tax policy, remains constant. Tax elasticity is therefore a better measure of automatic stabilization of taxes than tax buoyancy and is more appropriate for forecasting purposes. However, it is difficult to estimate tax elasticity properly given the need for extensive, detailed data on features and implementation of tax policy and tax administration. In most cases, 2 Section 1 Introduction including dummy variables to account for policy interventions is a good proxy, but it is difficult to observe the size of the discretionary changes. Tax buoyancy measures the total response of tax revenues both to automatic changes to economic growth and to discretionary changes in tax policy. A tax system is buoyant when tax revenue increases more than one-for-one with an increase in GDP. Theoret- ically, the long-run tax buoyancy should equal one, as tax cannot grow faster or slower than GDP indefinitely. The lack of long-run equilibrium increases the risk of an increase in public debt. In the short-run, tax buoyancies can be different from one due to dif- ferent features of the tax systems. For example, if income tax brackets and deductions are not adjusted for inflation, personal income taxes (PIT) may increase faster than income. Similarly, during the rebound period after a recession, corporate income tax (CIT) revenues may increase slower than value-added tax (VAT) revenues due to the loss-carry forward provision (Dudine and Jalles 2018). Tax buoyancy is sometimes a more appropriate measure of responsiveness of taxes to economic growth. As automatic and discretionary changes can be complimentary in the long run, thereby making it difficult to isolate their separate impacts, it is useful in this case to use tax buoyancy instead of tax elasticity as a comprehensive measure of the sustainability of the tax system. The same holds true after a package of reforms has been passed since it would be difficult to adequately measure the size of each policy measure in the package. Furthermore, as certain tax instruments, such as customs and excises, may have lower tax elasticity, an increase in tax revenues is contributed to discretionary rather than automatic changes. Similarly, as tax evasion and/or avoidance reduce the automatic responsiveness of tax revenues to GDP, additional tax revenues are raised through discretionary measures to improve tax compliance. In either of these cases, growth in tax revenues is reflected through tax buoyancy and not tax elasticity. Short-run and long-run buoyancy can vary across countries and tax instruments. Theoretically, long-run tax buoyancy is expected to be greater than one for progressive taxes, such as PIT, and less than one for taxes that are mostly regressive like VAT or sales tax. However, both PIT and consumption taxes may show the short-run buoyancy lower than one because of a lack of indexation, wage rigidity (Stockhammer 2013), and relative persistence in consumption spending. Depending on the VAT rate and con- sumption structure, even VAT can have long-run tax buoyancy greater than one if the standard VAT rate applies mostly to luxury goods,[1] and necessary goods are subject to a reduced VAT rate. Long-run buoyancy of excises depends on whether the excise tax rate annual adjustment is greater than or less than the increase in GDP. Property 1. There is no clear-cut distinction in definition of luxury and necessary goods. In this case, the distinction is in terms of income elasticity (Deaton and Muellbauer 1980; Lancaster 1971), where luxury goods are those whose income elasticity is greater than one, while necessary goods have income elasticity between zero and one. 3 PFR Fundamentals ――― Tax Buoyancy taxes commonly have a lower short-run buoyancy due to counter-cyclical property tax rate adjustments (Dillinger 1991). However, as one of the major sources of municipal own tax revenues, they tend to have a stabilizing role during periods of economic contraction and fiscal distress. Similarly, excises are found to be more buoyant during recessions than during economic growth. Only CIT revenues are buoyant in the short and long run as profits are usually more responsive to fluctuations in business activity (Belinga et al. 2014). An examination of tax buoyancy is crucial for tax policy design for a few reasons. Firstly, it illustrates the role of tax policy in stabilizing the economy over the business cycle in the short run, and in ensuring fiscal sustainability in the long run. Analyzing country-specific individual tax buoyancies also allows a country to determine whether its tax revenue mobilization is in line with economic activity as well as identify the strengths and weaknesses of its tax system (Dudine and Jalles 2018). An understanding of the institutional and structural characteristics that affect tax buoyancy can help adjust expectations about tax buoyancy as these characteristics change. 4 Section 2  Theoretical Framework and Empirical Approaches T ax buoyancy is traditionally estimated by means of a regression of a natural logarithm of tax revenue on the natural logarithm of GDP. The difference between regressions for tax buoyancy and tax elasticity is that the latter includes other factors that measure changes in tax policy parameters. In other words, while tax buoyancy estimation could be presented by the following simple regression, log T = γ 0 + γ 1 log GDP + ξ , tax elasticity is estimated as log T = β 0 + β 1 log GDP + ∑ k l =1 θl X l + ε. The difference between these two equa- k tions is factor ∑ l =1 θl X l , which represents a matrix of discretionary policy factors. 2.1. Theoretical Approach A commonly applied theoretical framework for estimating tax buoyancies starts from an autoregressive distributed lag (ARDL) model, p and q, which allows for a dynamic relationship between tax revenue and GDP: p q Ti, t = ∑ αi, j Ti, t ¡ j + ∑ β i, j GDP i, t ¡ j + µ i + ε i, t , ( 1) j =1 j =0 5 PFR Fundamentals ――― Tax Buoyancy where Ti, t and GDPi, t represent the natural logarithms of tax revenue and GDP, re- spectively, for country i in year t, μi is a country-specific fixed effect and εi, t is the error term. Based on equation (1), changes in tax revenues can be explained by its own distributed lag of order p, and a distributed lag of order q of GDP.[2] Number of lags (p, q) is determined based on the Bayesian information criterion, based on which the existing literature on tax buoyancy assumes the number of lags as (1, 1) (Belinga et al. 2014; Cornevin, Corrales, and Angel 2023; Deli et al. 2018; Dudine and Jalles 2018; Gupta, Jalles, and Liu 2022; Lagravinese, Liberati, and Sacchi 2020). This is also confirmed in the analysis for this note. Following that assumption, equation (1) becomes Ti, t = αi, 1 Ti, t ¡1 + β i, 0 GDP i, t + β i, 1 GDP i, t ¡1 + µ i + ε i, t . ( 2) This means that a change in tax revenues can be explained by its value in the preceding year and by GDP in the current and previous year. If changes in tax revenues and changes in GDP are cointegrated, which is confirmed with a cointegration test (Kao 1999; Pedroni 1999; Westerlund 2005), short- and long-run buoyancy can be estimated at the same time. Subtracting the lag dependent variable from both sides of equation (2) transforms it into a single equation error cor- rection model, as ∆ Ti, t = λ i ( Ti, t ¡1 ¡ γ i GDP i, t ¡1 ) + β i, 1 ∆ GDP i, t + µ i + ε i, t , ( 3) where βi, 0 measures the short-run buoyancy (i.e., instantaneous response), λi =¡(1¡αi, 1) measures the speed of adjustment between the short-run and the long-run buoyancy (i.e., speed at which buoyancy converges to its equilibrium), while λ i = ( β i, 0 + β i, 1 ) / (1 ¡ αi, 1 ) measures the long-run tax buoyancy. 2. Number of lags, p and q, should be carefully chosen to address potential issues of serial correlation and multicol- linearity. Including too many lags would reduce the degrees of freedom and increase the probability of multicollinearity between the lagged variables. On the other hand, if the error terms are serially correlated, not including sufficient lags would not resolve the problem of serial correlation, impacting the efficiency of the estimators. 6 Section 2 Theoretical Framework and Empirical Approaches 2.2. Empirical Approaches The ARDL model in equation (3) can be estimated country-by-country or for a group of countries. A country-level analysis, including the benchmarking to its peer countries, should be based on the country-level (time-series) analysis. On the other hand, a coun- ty-group analysis should employ the information provided in the panel data. As dis- cussed above, the country-level estimation provides useful information about the strengths and weaknesses of its tax system, but it can only be applied if there are suf- ficiently long consecutive series of data for the variables of interest. The panel data approach reduces the problem of degrees-of-freedom caused by the short time series for some countries. In addition, in a large panel data setting, discretionary policy changes are captured by time effects or fixed effects. For that reason, in the absence of infor- mation of discretionary tax measures, estimates of tax buoyancy from panel data are considered a good approximate of tax elasticities (Cornevin, Corrales, and Angel 2023). Several econometric techniques have been applied in the literature for estimating the above ARDL model with panel data. A fixed-effects (FE) approach assumes only the intercepts vary across countries, while the slope coefficients are identical for all, which effectively means that the estimator assumes identical short-run and long-run relationships for all countries in the sample. However, this may produce inconsistent results if the actual slopes are not identical. Conversely, the mean group (MG) estimator (Pesaran and Smith 1995) assumes different intercepts and slopes (i.e., short-run and long-run relationships) for each country and calculates a simple average of coefficients for the full sample. A combination of these two extreme approaches is the pooled mean group (PMG) estimator (Pesaran, Shin, and Smith 1999) which allows the intercept and short-run coefficients to differ across countries (as in the MG estimator) but as- sumes identical long-run coefficients (as in the FE estimator). It is reasonable to assume that certain countries may have a similar long-run relationship between GDP and tax revenues, where countries may have similar budget constraints and fiscal rules, or where they belong to the same currency union. However, this assumption should be tested using the Hausmann test (Hausman 1978). As panel data may exhibit extensive cross-sectional dependence in errors due to unobserved common factors, the ARDL model with panel data should be estimated using an augmented estimator. Examples of common factors are changes in commodity prices, technological change, common currency, common fiscal rules, etc. They may affect all units in the sample, but not necessarily in the same way. If these common factors are correlated with the regressors, the FE, MG and PMG estimators are incon- sistent. To correct for that, the common correlated effects (CCE) estimator augments the MG and PMG models by adding cross-sectional averages of the dependent variable 7 PFR Fundamentals ――― Tax Buoyancy and the regressors as control variables to act as proxies for the unobserved common factors (Pesaran 2006). The dynamic common correlated effects (DCCE) estimator goes even further in controlling cross-sectional dependence by adding also lagged cross-sectional averages as control variables (Chudik and Pesaran 2015; Ditzen 2018). The addition of cross-sectional averages and lags transforms equation (3) into ∆ Ti, t = λ i ( Ti, t ¡1 ¡ γ i GDP i, t ¡1 ) pt + β i, 1 ∆ GDP i, t + ∑ ζ t ¡l z ¯t ¡l + µ i + i, t , ( 4) l =0 where z ¯i, t , ln GDP i, t is a vector of cross-sectional averages of dependent ¯t ¡l = ln T and independent variables. In the case of the CCE estimator, l = 0, while for DCCE, l > 0. The ARDL model can also be estimated country-by-country using time-series anal- ysis, in which case cross-sectional dependence is not a problem, as there is only one unit and therefore no others to be correlated with. There may be other shocks to economic growth that should be observed in the analysis, such as Covid, changes in international commodity prices, or shocks in other countries that that may spill over (e.g., conflicts, natural disasters, etc.). For example, if country A is a major importer of goods from country B, sudden change in demand in country A caused by a natural disaster may significantly impact GDP in country B. If such factors are correlated with GDP and not observed, the estimated tax buoyancy may be biased. 2.3. Literature Review Various studies have attempted to estimate tax buoyancy and tax elasticity for indi- vidual countries, using either ordinary least squares (e.g., India [Upender 2008], Kenya [Mawia and Nzomoi 2013], Netherlands [Wolswijk 2007], Paraguay [Mansfield 1972]) or cointegration techniques (such as Barbados [Ochieng and Mamingi 2022; Scott-Joseph et al. 2016], Bulgaria [Tanchev and Todorov 2019], Ethiopia [Bayu 2015], Nigeria [Musa et al. 2016], South Africa [Naape and Mahonye 2020]). Several studies have also tried to estimate tax buoyancy using cointegration tech- niques with panel data. For example, Belinga et al. (2014) find that in about half of OECD countries, long-run buoyancy is greater than one between 1965 and 2012, while short-run buoyancy does not significantly differ from one in most countries in the sample. A study by Lagravinese, Liberati, and Sacchi (2020) find short- and long-run 8 Section 2 Theoretical Framework and Empirical Approaches buoyancy lower than one after correcting for cross-sectional dependence, for the same group of countries but with a shorter, more recent time period (1995–2016). Gupta, Jalles, and Liu (2022) find that in most of 44 Sub-Saharan African countries over the period 1980–2017, short- and long-run buoyancy were around one for total taxes. Dudine and Jalles (2018) and more recently Cornevin, Corrales, and Angel (2023) estimate tax buoyancy for advanced, emerging, and low-income economies. Dudine and Jalles (2018) find that, between 1980 and 2014, short- and long-run buoyancy were not different than one in advanced economies, while in emerging and low-income economies, short-run buoyancy exceeds one. They find CIT to be buoyant in the short and long run in all country groups, and PIT is buoyant in emerging economies and taxes on goods and services (TGS) in low-income countries in the long run. While correcting for cross-sectional dependence in panel data and with a larger number of countries, Cornevin, Corrales, and Angel (2023) find that tax revenues are buoyant in the long run across all country-income groups. Short-run buoyancy is less than one for total tax revenues in all country groups, but higher in advanced economies than in emerging and low-income economies. 9 Section 3  Empirical Approach for Estimating Tax Buoyancy THIS SECTION of the note briefly describes the data and the recommended approach for estimating tax buoyancy for country groups and country-by-country. 3.1. Data This note uses annual panel data on total tax revenues, CIT, PIT, VAT, and GDP for 171 countries between 1980 and 2021. Tax revenue data are obtained from IMF’s Government Finance Statistics and GDP data is from the World Development Indica- tors. As with most parametric models, the rule of thumb is to use a large number of observations. Following the existing similar studies, only countries with at least 15 years of observations are included in the analysis (Cornevin, Corrales, and Angel 2023). The panel is unbalanced with an average of 33 observations per country. All data is in current local currency units. Social security contributions (SSC) are not included in total taxes as they have much lower automatic stabilization power and may distort estimates (Cornevin, Corrales, and Angel 2023). There is also a risk of introducing inaccurate year-to-year changes in tax revenues driven by many gaps in SSC data. 10 Section 3 Empirical Approach for Estimating Tax Buoyancy 3.2. Estimation Approach Country-by country estimates are obtained with MG/PMG estimator using time-se- ries data, while for panel data estimation, this note recommends using the CCE estimator.[3] Theoretically, the preferred approach for estimating tax buoyancy with panel data is DCCE, since it accounts for unobserved heterogeneity and cross-sectional dependence (Cornevin, Corrales, and Angel 2023; Lagravinese, Liberati, and Sacchi 2020). However, the inclusion of lagged cross-sectional averages may significantly reduce degrees of freedom and cause the estimates to be biased and unreliable. While the CCE estimator may not fully address the problem of cross-sectional dependence, it does significantly reduce it while not sacrificing large degrees of freedom (Anderson and Raissi 2022). Robustness checks are conducted by adding inflation (measured by CPI) and tax rates as additional control variables. While inclusion of CPI tests whether tax buoy- ancy is independent from price changes, controlling for a change in tax rates accounts for an impact of discretionary changes on tax revenues. As prices increase, a non-in- dexed tax base (e.g., if income tax brackets are not adjusted) causes tax revenues to increase not due to high buoyancy but due to higher prices. If tax buoyancy is not neutral to price changes, long-run and short-run coefficients should be lower when CPI is included. To account for the impact of discretionary changes on tax revenues, standard VAT, top marginal PIT, and statutory CIT rates are added as explanatory variables. As the inclusion of tax rates reduces the sample size because of limited tax rate data, equation (3) is estimated using the MG estimator to limit the loss of the degrees of freedom while preserving the assumed heterogeneity in short-run and long-run relationships by country. In addition, to compare the results with and without tax rates, a balanced panel is employed. As some tax revenue sources may perform stabilization and sustainability roles differently during various phases of the business cycle, tax buoyancy estimates during expansion and recession are compared by including a dummy variable that takes value one in years with negative GDP growth (Gupta, Jalles, and Liu 2022) and zero otherwise. 3. PFR visualization tool provides country-by-country estimates produced by the global FPSG team. The country-group estimates are not provided as each country has its own group of peers. This note provides guidance on estimating country-group estimates for country economists interested in producing those. The FPSG team is available to provide support as needed. 11 Section 4  Results THIS SECTION presents the country-income group and regional tax buoyancy estimates obtained by estimating equation (4) using the CCE approach and country-level estimates obtained by estimating equation (3) with the MG estimator. 4.1. Country-Income Group and Regional Tax Buoyancy Estimates Estimated long-run buoyancy is around one for high-, upper-middle-, and low- er-middle-income countries (table 1). At significantly less than one, long-run tax buoyancy in low-income countries suggests a high risk of increased indebtedness in the future as their tax revenues are growing slower than their income. At the same time, short-run tax buoyancy estimates suggest that the tax systems in all country-income groups are not well equipped to stabilize the economy in the short-run; this is not surprising due to a trend in reduced progressivity in high-income countries and reliance on VAT in the low- and middle-income countries for revenue generating power. On the other hand, there is greater variation in both short-run and long-run buoyancy across World Bank regions. 12 Section 4 Results TABLE 1 Total tax buoyancy By country-income groups High Upper-middle Lower-middle Low Short-run buoyancy 0.878*** 0.857*** 0.770*** 0.857*** Long-run buoyancy 1.011*** 1.003*** 0.913*** 0.701** Speed of adjustment −0.437*** −0.560*** −0.449*** −0.550*** Observations 1,903 1,456 1,451 553 No. of countries 56 47 48 20 CD 30.70 12.93 10.94 −0.27 CD p value 0.00 0.00 0.00 0.78 By region L. Amer. E. Eur. & Mid. East & E. Asia & & the Cent. Asia N. Africa Africa Pacific S. Asia Caribbean Short-run buoyancy 0.797*** 0.701*** 0.765*** 1.003*** 0.507*** 0.979*** Long-run buoyancy 0.889*** 0.840*** 0.834*** −0.617 0.106 3.048 Speed of adjustment −0.465*** −0.544*** −0.533*** −0.384*** −0.282*** −0.429*** Observations 1,455 519 1,237 778 280 1,016 No. of countries 48 16 41 27 8 29 CD 29.25 −2.16 0.251 9.05 −1.54 6.10 CD p value 0.00 0.03 0.802 0.00 0.12 0.00 Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. The estimates in table 1 are in line with previously published estimates. There are at least two reasons why these results explain the discrepancies with the previous results. Firstly, Dudine and Jalles (2018) do not correct for cross-sectional dependence, while results in table 1 do. Secondly, country-income classification is slightly different, Dudine and Jalles (2018) follow the IMF classification on advanced, emerging, and low-income economies, but this note uses the World Bank classification on high-income, upper-mid- dle-income, lower-middle-income, and low-income economies. Thirdly, difference in tax revenue data used in this note and the previous study may be driving the difference. A more recent IMF study has compared the estimated tax buoyancy by country-income group (Cornevin, Corrales, and Angel 2023) while applying a different estimator. Rep- 13 PFR Fundamentals ――― Tax Buoyancy licating their approach, which consists of applying the CCE estimator on data from 1990–2020 and using their country-income classification, produces comparable results.[4] PIT ensures long-run fiscal sustainability only in high-income countries and ECA region, but it has not been, on average, a good output stabilizer (table 2). Such low short-run tax buoyancy estimates may be a result of a decline in the share of labor income, but also wage rigidity, especially in public sector. Estimates for low-income countries should be interpreted with caution because of the relatively small sample size of only 17 countries. TABLE 2 Tax buoyancy , PIT By country-income groups High Upper-middle Lower-middle Low Short-run buoyancy 0.753*** 0.354 0.545*** 0.603** Long-run buoyancy 1.758* −0.023 0.299 0.670 Speed of adjustment −0.402*** −0.545*** −0.547*** −0.623*** Observations 1,603 1,100 1,048 413 No. of countries 49 40 38 17 CD 19.93 -2.70 -1.20 3.96 CD p value 0.00 0.01 0.23 0.00 By region L. Amer. E. Eur. & Mid. East & E. Asia & & the Cent. Asia N. Africa Africa Pacific S. Asia Caribbean Short-run buoyancy 0.406 0.587** 0.598*** 0.770*** 0.789* 0.923* Long-run buoyancy 0.740 1.313** 12.447 0.705*** 0.664 0.834 Speed of adjustment −0.447*** −0.414*** −0.599*** −0.514*** −0.563*** −0.458*** Observations 1,323 288 935 537 196 807 No. of countries 45 9 34 20 7 27 CD 13.87 0.18 -2.57 -0.32 -1.82 -1.00 CD p value 0.00 0.86 0.01 0.75 0.07 0.32 Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. 4. Results available from the author upon request. 14 Section 4 Results In high- and upper-middle-income countries, CIT is very responsive to GDP, in the short and long run, in line with expectations. High CIT buoyancy (table 3) may be consistent with low PIT buoyancy in high- and upper-middle-income countries and may be due to a decline in share of labor income over the last couple of decades. Similar to PIT results, buoyancy estimates for the low-income countries should be carefully interpreted because of the small sample size. Unsurprisingly, VAT is not a good output stabilizer in most countries. Due to its regressive structure and persistence in consumption, tax buoyancy of VAT is expected TABLE 3 Tax buoyancy, CIT By country-income groups High Upper-middle Lower-middle Low Short-run buoyancy 1.527*** 1.398*** 0.624*** 0.864*** Long-run buoyancy 1.636*** 1.453*** 0.800*** 0.096 Speed of adjustment −0.517*** −0.561*** −0.570*** −0.694*** Observations 1,740 1,183 1,037 405 No. of countries 53 42 38 17 CD 7.50 3.57 1.92 −3.65 CD p value 0.00 0.00 0.06 0.00 By region L. Amer. E. Eur. & Mid. East & E. Asia & & the Cent. Asia N. Africa Africa Pacific S. Asia Caribbean Short-run buoyancy 1.773*** 0.460 0.973*** 0.731** 0.377 1.307*** Long-run buoyancy 0.644 0.631 0.546 0.407 −0.641 1.431*** Speed of adjustment −0.504*** −0.596*** −0.679*** −0.617*** −0.461*** −0.520*** Observations 1,316 435 959 525 215 837 No. of countries 44 15 36 19 7 27 CD 7.46 0.58 −3.93 5.98 −2.32 2.09 CD p value 0.00 0.56 0.00 0.00 0.02 0.04 Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. 15 PFR Fundamentals ――― Tax Buoyancy to be lower than that of other taxes. However, it is not impossible for VAT tax buoyancy to be higher, as in case of low-income countries and SSA region (table 4), where it may be due to the differential tax rate structure and basic goods being subject to a reduced rate of VAT. These estimates may also be driven by the small sample size for low-income countries and should be carefully analyzed and linked with the tax designs of these countries.[5] TABLE 4 Tax buoyancy, VAT By country-income groups High Upper-middle Lower-middle Low Short-run buoyancy 0.772*** 0.991*** 0.790*** 0.936*** Long-run buoyancy 0.838*** 0.962*** 0.617** 1.971*** Speed of adjustment −0.511*** −0.659*** −0.615*** −0.840*** Observations 1,280 690 471 185 No. of countries 42 31 24 9 CD 12.04 7.15 0.42 −1.16 CD p value 0.00 0.00 0.68 0.25 By region L. Amer. E. Eur. & Mid. East & E. Asia & & the Cent. Asia N. Africa Africa Pacific S. Asia Caribbean Short-run buoyancy 0.819*** 1.015*** 0.645*** 0.800*** −0.063 0.950*** Long-run buoyancy 0.764*** 0.279 1.160*** 0.489* −0.662 0.674** Speed of adjustment −0.527*** −0.462*** −0.868*** −0.646*** −0.792*** −0.507*** Observations 1,134 101 471 301 34 557 No. of countries 40 5 24 14 2 20 CD 10.29 −1.30 −0.90 1.51 −1.76 13.29 CD p value 0.00 0.19 0.37 0.13 0.08 0.00 Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. 5. CCE may not be an appropriate estimator for the low-income countries sample due to the small sample size. The MG estimator may be more appropriate in this case, even though it does not correct for cross-sectional dependence. MG estimates for low-income countries are slightly smaller—0.766 for short-run buoyancy, and 1.295 for long-run buoyancy. Based on 20 low-income countries from 1990 to 2020 (435 observations), Cornevin, Corrales, and Angel (2023) estimate short-run buoyancy of 0.208 and long-run buoyancy of 1.117 (higher than for advanced and emerging economies). 16 Section 4 Results 4.2. Robustness Check of Panel Estimates with Controlling for Inflation The long-run tax buoyancy is, on average, mostly neutral to price changes, while the short-run buoyancy coefficients become smaller.[6] Comparing estimated total tax buoyancy coefficients while controlling for inflation (table 5) allows to determine whether tax buoyancy estimates are neutral to price increases or not. This can indicate TABLE 5 Total tax buoyancy, controlling for inflation By country-income groups  High Upper-middle Lower-middle Low Short-run buoyancy 0.867*** 0.728*** 0.638*** 0.866*** Long-run buoyancy 0.995*** 1.011*** 1.060*** 0.313 Speed of adjustment −0.592*** −0.634*** −0.598*** −0.635*** Observations 1,855 1,356 1,307 504 No. of countries 56 45 45 20 CD 25.270 9.266 6.695 −1.771 CD p value 0.000 0.000 0.000 0.077 By region L. Amer. E. Eur. & Mid. East & E. Asia & & the  Cent. Asia N. Africa Africa Pacific S. Asia Caribbean Short-run buoyancy 0.776*** 0.669*** 0.728*** 0.975*** 0.406** 0.892*** Long-run buoyancy 0.842*** 0.771*** 0.924*** 1.144*** 1.252* 1.673*** Speed of adjustment −0.573*** −0.827*** −0.646*** −0.618*** −0.337*** −0.603*** Observations 1,398 456 1,148 720 274 948 No. of countries 47 15 41 25 8 28 CD 33.750 2.203 2.072 3.125 −1.292 5.356 CD p value 0.000 0.028 0.038 0.002 0.196 0.000 Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. 6. Also, inflation coefficients are not statistically significant. 17 PFR Fundamentals ――― Tax Buoyancy whether an increase in tax revenues is a result of the price increases due to certain features of the tax system. Such features include: lack of or irregular indexation of income tax brackets and VAT threshold, taxation of nominal rather than real household income and profits, and lag between tax collection and refund payments. For PIT, short-run and long-run buoyancy remain mostly unchanged, except for high-income countries where short-run buoyancy becomes smaller when CPI is included, suggesting a potential lack of income tax bracket indexation (table 6). Including inflation does not reduce CIT buoyancy in most country groups (table 7), except for short-run buoyancy TABLE 6 PIT buoyancy, controlling for inflation By country-income groups High Upper-middle Lower-middle Low Short-run buoyancy 1.111** 0.375 0.624*** 0.678* Long-run buoyancy 1.002*** 0.597 0.585** 1.477*** Speed of adjustment −0.526*** −0.632*** −0.703*** −0.786*** Observations 1,587 1,045 1,029 372 No. of countries 49 39 38 16 CD 20.18 −1.36 −1.27 7.72 CD p value 0.00 0.17 0.20 0.00 By region L. Amer. E. Eur. & Mid. East & E. Asia & & the Cent. Asia N. Africa Africa Pacific S. Asia Caribbean Short-run buoyancy 0.305 0.381 0.160 1.171*** 0.715* −0.190 Long-run buoyancy −1.750 1.427*** 1.675 1.051** 1.256 −9.992 Speed of adjustment −0.561*** −0.405*** −0.701*** −0.616*** −0.754*** −0.544*** Observations 1,306 281 874 536 195 763 No. of countries 45 9 33 20 7 26 CD 12.21 2.62 −0.62 0.01 8.69 −2.00 CD p value 0.00 0.01 0.54 1.00 0.00 0.05 Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. 18 Section 4 Results TABLE 7 CIT buoyancy, controlling for inflation By country-income groups High Upper-middle Lower-middle Low Short-run buoyancy 1.567*** 1.496** 0.735*** 0.441 Long-run buoyancy 2.006*** 1.706*** 1.561*** −0.812 Speed of adjustment −0.593*** −0.730*** −0.668*** −0.882*** Observations 1,713 1,135 994 364 No. of countries 53 41 36 16 CD 4.91 10.11 3.46 −4.43 CD p value 0.00 0.00 0.00 0.00 By region L. Amer. E. Eur. & Mid. East & E. Asia & & the Cent. Asia N. Africa Africa Pacific S. Asia Caribbean Short-run buoyancy 1.875*** −0.501 0.849*** 1.064*** 0.576 1.066*** Long-run buoyancy 1.484** 0.553 0.446 1.419** 3.499 1.684*** Speed of adjustment −0.580*** −0.790*** −0.825*** −0.688*** −0.568*** −0.629*** Observations 1,309 394 891 524 214 796 No. of countries 44 14 34 19 7 26 CD 8.68 0.64 −3.95 1.63 −1.59 0.25 CD p value 0.00 0.52 0.00 0.10 0.11 0.80 Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. 19 PFR Fundamentals ――― Tax Buoyancy TABLE 8 VAT buoyancy, controlling for inflation By country-income groups High Upper-middle Lower-middle Low Short-run buoyancy 0.585*** 0.662*** 0.574*** 1.647*** Long-run buoyancy 0.748* 0.468* 0.777*** 1.937** Speed of adjustment −0.639*** −0.777*** −0.824*** −1.027*** Observations 1,280 688 463 184 No. of countries 42 31 24 9 CD 8.28 4.29 −0.64 −1.18 CD p value 0.00 0.00 0.52 0.24 By region L. Amer. E. Eur. & Mid. East & E. Asia & & the Cent. Asia N. Africa Africa Pacific S. Asia Caribbean Short-run buoyancy 0.701*** 0.400 −0.080 0.832** −0.299*** 1.078*** Long-run buoyancy 1.981* −0.992 −2.174 0.596 −1.108 0.905** Speed of adjustment −0.632*** −0.550*** −0.997*** −0.739*** −0.897*** −0.700*** Observations 1,134 101 468 301 34 549 No. of countries 40 5 24 14 2 20 CD 5.15 −1.88 0.08 −0.72 −1.45 4.53 CD p value 0.00 0.06 0.93 0.47 0.15 0.00 Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. in EAP and LAC regions. On the other hand, VAT buoyancy is not neutral to price changes in all but low-income countries (table 8). 4.3. Controlling for Tax Rates Controlling for changes in tax rates slightly reduces long-run tax buoyancy in gen- eral, while short-run buoyancy remains mostly unchanged, but the effect varies by country-group. Controlling for top marginal PIT rates only has a small impact on 20 TABLE 9 Tax buoyancy, with and without controlling for tax rates By country-income groups By region L. Amer. Upper- Lower- E. Eur. & Mid. East E. Asia & & the Buoyancy Full sample High middle middle Low Cent. Asia & N. Africa Africa Pacific S. Asia Caribbean PIT Not controlling for PIT rates Short-run 0.861*** 0.712*** 1.030*** 0.956** 0.870*** 0.659*** 0.149 0.944*** 0.829*** 0.367 1.756*** Long-run 0.986*** 0.727 1.390*** 0.985*** 1.015*** 1.229*** 1.164*** 0.903*** 1.280*** 1.230*** 0.012 Controlling for PIT rates Short-run 0.872*** 0.718*** 0.993*** 1.004*** 0.989*** 0.651*** 0.189 1.060*** 0.801*** 0.329 1.763*** Long-run 0.884** 0.512 1.391*** 0.935*** 1.065*** 1.288*** 1.162*** 0.848*** 1.216*** 1.246*** −0.659 CIT Not controlling for CIT rates Short-run 1.379*** 1.557*** 1.537*** 0.890*** 1.224*** 1.901*** 0.709*** 1.216*** 1.055*** 0.662 1.369*** Long-run 1.305*** 1.297*** 1.254*** 1.213*** 1.869** 1.331*** 1.089*** 1.423*** 1.200*** 1.766*** 1.254*** Controlling for CIT rates Short-run 1.368*** 1.553*** 1.511*** 0.874*** 1.258*** 1.921*** 0.639*** 1.273*** 1.074*** 0.504 1.256*** Long-run 1.256*** 1.286*** 1.256*** 1.190*** 1.293*** 1.302*** 1.055*** 1.253*** 1.214*** 1.579*** 1.263*** VAT Not controlling for VAT rates Short-run 0.888*** 0.843*** 0.994*** 0.836*** 0.865** 0.897*** 1.155*** 0.739*** 0.833*** 0.377*** 1.075*** Long-run 1.127*** 1.227*** 0.972*** 1.072*** 1.345*** 1.027*** 0.994*** 1.200*** 1.620*** 0.838** 1.002*** Controlling for VAT rates Short-run 0.886*** 0.839*** 0.989*** 0.829*** 0.894*** 0.878*** 1.106*** 0.756*** 0.823*** 0.378*** 1.100*** Long-run 1.090*** 1.114*** 0.995*** 1.076*** 1.348*** 1.004*** 0.980*** 1.202*** 1.318*** 0.838** 1.039*** Section 4 Results Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. 21 PFR Fundamentals ――― Tax Buoyancy short-run tax buoyancy in low-income countries (table 9), while inclusion of standard CIT rates reduces long-run buoyancy in the same group. Standard VAT rates have the highest impact on long-run buoyancy in high-income countries. These results suggest that there is a very small correlation between tax rates in GDP in certain groups of countries. 4.4. Business Cycle For most country groups, short-run and long-run tax buoyancies are slightly smaller during the contraction than the expansion period. Lower short-run buoyancy during the contraction period means that the tax system is not performing well as an automatic stabilizer. However, for most groups of countries, the change in short-run tax buoyancy during different phases of the business cycle phases is not significantly high. In case of individual tax categories, PIT has higher tax buoyancy during a recession in upper-mid- dle-income countries and LAC region, while CIT has been a good automatic stabilizer in the MENA region. While previous studies showed mixed results on the effect of the business cycle on tax buoyancy, results in this note are mostly in line with the most recent studies. Belinga et al. (2014) finds that total tax buoyancy was larger during contraction in OECD countries between 1965 and 2012, but only SSC, excises and property taxes were good economic stabilizers. Dudine and Jalles (2018) found that only CIT was a good automatic stabilizer in advanced and emerging economies from 1980 to 2014, conversely. Contrary to Dudine and Jalles (2018), Gupta, Jalles, and Liu (2022) found that short-run and long-run buoyancies are smaller during economic slowdown for Sub-Saharan Af- rican countries. Short-run buoyancy is especially lower for TGS and trade taxes, while long-run buoyancy is lower for CIT. 4.5. Country-Level Estimates of Tax Buoyancy This note uses Serbia as an example to illustrate the application of country-level tax buoyancy results for benchmarking tax revenue mobilization to its peers, Serbia is one of the revamped PFR pilot countries. 22 TABLE 10 Tax buoyancy over the business cycle By country-income groups By region L. Amer. Upper- Lower- E. Eur. & Mid. East E. Asia & & the High middle middle Low Cent. Asia & N. Africa Africa Pacific S. Asia Caribbean Total taxes Short-run Expansion 0.749*** 0.752*** 0.699*** 0.954*** 0.753*** 0.675*** 0.765*** 0.909*** 0.442*** 0.783*** Contraction 0.743*** 0.741*** 0.698*** 0.947*** 0.748*** 0.691*** 0.758*** 0.908*** 0.422*** 0.764*** Long-run Expansion 1.015*** 0.994*** 1.000*** 1.211*** 0.976*** 1.095*** 1.054*** 1.067*** 1.037*** 1.009*** Contraction 1.010*** 0.917*** 0.900*** 1.330*** 0.930*** 1.387*** 1.061*** 1.020*** 0.909*** 0.778*** PIT Short-run Expansion 0.739*** 0.993** 0.669*** 0.892*** 0.461** 0.408** 0.608*** 0.896*** 1.159*** 1.894*** Contraction 0.731*** 1.032** 0.649*** 0.884*** 0.455** 0.399** 0.586*** 0.891*** 1.169*** 1.964*** Long-run Expansion 1.039*** 1.186*** 1.177*** 1.344*** 1.042*** 0.928*** 1.706*** 1.069*** 1.374*** 1.454*** Contraction 0.906*** 1.183*** 1.136*** 1.288*** 0.965*** 0.688*** 1.589*** 0.994*** 1.284*** 1.796*** CIT Short-run Expansion 0.718*** 1.139*** 0.945*** 0.765*** 1.394*** 0.760*** 0.807*** 0.878*** 0.620 1.050*** Contraction 0.707*** 1.122*** 0.936*** 0.727*** 1.374*** 0.788*** 0.778*** 0.864*** 0.576 1.017*** Long-run Expansion 1.052*** 1.153*** 1.337*** 1.989*** 1.195*** 1.195*** 1.540*** 1.122*** 1.584*** 1.336*** Contraction 0.900*** 1.057*** 1.335*** 2.338*** 1.074*** 1.253*** 1.679*** 1.072*** 1.310*** 1.145*** VAT Short-run Expansion 0.698*** 0.964*** 0.880*** 0.724** 0.830*** 1.238*** 0.623*** 0.748*** 0.466*** 1.019*** Contraction 0.690*** 0.956*** 0.880*** 0.715** 0.825*** 1.241*** 0.615*** 0.744*** 0.460*** 1.007*** Long-run Expansion 0.935*** 0.910*** 1.041*** 1.274*** 1.025*** 0.993*** 1.164*** 0.659** 0.881*** 0.887*** Contraction 0.908*** 0.954*** 1.038*** 1.252*** 1.008*** 0.982*** 1.150*** 0.641** 0.875*** 0.952*** Source: World Bank staff calculations. *** p < 0.01, ** p < 0.05, * p < 0.1. Section 4 Results 23 24 PFR Fundamentals ――― Tax Buoyancy TABLE 11 Total tax buoyancy in southeast European countries 1980–2021 2000–21 Without CPI With CPI Without CPI With CPI Short-run Long-run Short-run Long-run Short-run Long-run Short-run Long-run Country buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy Albania −0.036 0.979*** 0.439 −0.647 0.574* 1.001*** 0.733** 1.530*** Bosnia and Herzegovina 0.719*** 0.939*** 1.186*** 0.942*** 1.267*** 1.045*** 1.186*** 0.942*** Bulgaria 0.905*** 0.966*** 0.503*** 1.137*** 1.602*** 1.060*** 1.554*** 1.313*** Croatia 0.980*** 0.957*** 0.975*** 0.976*** 1.365*** 1.075*** 1.412*** 1.037*** Czechia 1.355*** 1.055*** 1.420*** 1.168*** 1.123*** 0.991*** 1.121*** 0.563 Hungary 0.679*** 0.952*** 0.614*** 0.855*** 0.650*** 0.954*** 0.695*** 0.877*** Montenegro 1.325*** 1.088*** 1.334*** 1.428*** 1.325*** 1.088*** 1.334*** 1.428*** North Macedonia 1.943*** 1.056*** 2.208*** 1.443*** 1.943*** 1.056*** 2.208*** 1.443*** Romania 0.613** 1.164*** 0.631*** 0.707*** 0.526*** 0.641* 0.687*** 0.118 Serbia 1.334*** 1.184*** 1.777*** 1.297*** 1.334*** 1.184*** 1.777*** 1.297*** Slovak Republic 0.983*** 1.271*** 0.970*** 0.759 1.025*** 1.241*** 0.883*** −4.102 Slovenia 0.849*** 0.889*** 0.893*** 0.865*** 0.923*** 0.796*** 1.028*** 0.857** Peer average 0.971*** 1.042*** 1.079*** 0.911*** 1.138*** 1.011*** 1.218*** 0.609 EU-27 0.757*** 1.010*** 0.714*** 0.863*** 0.857*** 1.037*** 0.840*** −1.148 EU-11 0.967*** 1.026*** 0.918*** 0.992*** 1.042*** 1.011*** 1.090*** 0.599 Source: World Bank staff calculations. Note: For Montenegro, North Macedonia, and Serbia, data available from 2005. *** p < 0.01, ** p < 0.05, * p < 0.1. Section 4 Results Serbia’s total tax revenue mobilization is more responsive to changes in GDP than its regional and aspirational peers. Based on population size, structure of the economy, and intergovernmental relations, Serbia’s selected aspirational peers are Bulgaria, Cze- chia, Hungary, and Slovakia. Table 11 shows that estimated short-run buoyancy for Serbia, Montenegro, and North Macedonia are, contrary to expectations, based on the structure of their tax revenues, well above one and above the estimates for their neigh- bors, whose tax systems are similar in terms of heavy reliance on consumption taxes. Table 11 also shows that in most countries, excluding Bulgaria, Hungary, Romania, and Slovakia, tax buoyancy estimates are neutral to price changes. Tax buoyancy estimates of Serbia’s peers changed after the 1990s. To account for the potential impact of inflation in the 1990s, alternative results for the period 2000–2021 are presented in the table. This allows a more appropriate comparison of Serbia’s tax buoyancy estimates, as data for Serbia is only available from 2005. In five out of eight countries for which data is available prior to 2005, tax buoyancy is higher in more recent periods. When focusing on the period after 2000, Serbia’s total tax buoyancy estimates are more in line with those of its peers. Further investigation should be done to understand the drivers of such high tax buoyancy. In addition to understanding the structure and responsiveness of individual tax categories, there may have been external shocks during this period that could have impacted GDP growth. In addition to the global financial crisis 2007–08, this region suffered from one of the worst droughts ever in 2012 and floods in 2014. Short-run PIT buoyancy is slightly higher than peer average, but in the long run, PIT revenues do not ensure fiscal sustainability. Bearing in mind that Serbia’s PIT structure is flat, high short-run PIT buoyancy is not expected. However, this fast in- crease in PIT collection over the observed period might be contributed to an increase in formal employment. On the other hand, long-run PIT buoyancy in Serbia (table 12) is lower than of its peers, which may be due to wage rigidity, especially in the public sector. In line with expectations, CIT in Serbia is very buoyant in short and long runs. High CIT buoyancy in Serbia (table 13) may be contributed to increased attractiveness of Serbia for investments and increased profitability over much of the past decade. While more buoyant in the short-run, Serbia’s VAT long-run buoyancy is similar to that of its peers. VAT buoyancy is expected to be lower due to assumed regressivity of VAT, but high buoyancy is not unusual in countries with differentiated VAT rate struc- tures. In line with the EU Directive, Serbia and its peers levy a reduced VAT rate on selected goods, including basic goods, while goods that are subject to standard VAT rate are usually those which are very income elastic. 25 26 PFR Fundamentals ――― Tax Buoyancy TABLE 12 PIT buoyancy in southeast European countries 1980–2021 2000–21 Without CPI With CPI Without CPI With CPI Short-run Long-run Short-run Long-run Short-run Long-run Short-run Long-run Country buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy Albania 0.057 1.449** −0.096 0.233 2.080 1.753*** 2.496** 4.785** Bosnia and Herzegovina −4.288** 1.496 −4.130 1.361 −4.288** 1.496 −4.130 1.361 Bulgaria 1.011*** 0.933*** 0.438** 1.296*** 0.793** 1.188*** 0.844* 1.937*** Croatia 0.327 0.694*** 0.294 0.771** 0.767** 0.966*** 0.734* 1.003*** Czechia 1.780*** 0.948*** 1.899*** 0.945 1.354*** 1.198** 1.234*** 15.188 Hungary 0.491 0.733*** 0.445 0.897 0.544 0.555*** 0.189 1.412*** Montenegro North Macedonia 0.646 2.383 0.889 −22.385 0.646 2.383 0.889 −22.385 Romania 0.486** 1.391*** 0.757** 1.050*** 0.354 0.883*** 0.564 0.619 Serbia 2 1.945*** 0.822*** 1.966*** 1.026*** 1.945*** 0.822*** 1.966*** 1.026*** Slovak Republic 0.856** 0.920*** 1.343*** 2.755*** 1.160*** −1.318 1.163*** 16.145 Slovenia 0.794*** 0.912*** 0.707*** 1.474*** 0.875*** 0.894*** 0.621*** 1.498*** Peer average 0.373 1.153*** 0.410 −0.961 0.566 0.984*** 0.597 2.053 EU-27 0.699*** 0.988*** 0.662*** 1.432*** 0.719*** 1.053*** 0.714*** 0.457 EU-11 1.095*** 0.973*** 1.049*** 1.743*** 1.004*** 0.914*** 1.007*** 3.860** Source: World Bank staff calculations. Note: For Montenegro, data on PIT not available. For North Macedonia and Serbia, data available from 2005. *** p < 0.01, ** p < 0.05, * p < 0.1. TABLE 13 CIT buoyancy in southeast European countries 1980–2021 2000–21 Without CPI With CPI Without CPI With CPI Short-run Long-run Short-run Long-run Short-run Long-run Short-run Long-run Country buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy Albania 1.049 1.170*** 1.086 −0.452 −0.061 0.546 −0.322 −1.214 Bosnia and Herzegovina Bulgaria 1.068*** 0.876*** 1.277** 1.629*** 1.961* 0.814*** 0.595 1.866** Croatia 2.308*** 1.429*** 1.689*** 3.125*** 2.840*** 1.482*** 2.373*** 3.061*** Czechia 1.369** 1.089*** 1.920*** 0.866 1.852*** 0.619*** 1.808*** −0.048 Hungary 0.656 0.876*** 0.852 −6.517 −0.150 −0.512 −0.376 −1.444 Montenegro North Macedonia 7.898*** 1.926** 7.504** 4.145** 7.898*** 1.926** 7.504** 4.145** Romania 1.143*** 0.949*** 1.293*** 0.698*** 1.375*** 0.916*** 1.774*** 0.898** Serbia 3.162*** 2.202*** 3.975*** 1.185 3.162*** 2.202*** 3.975*** 1.185 Slovak Republic 1.637*** 1.370*** 1.655*** 0.701 1.723*** 1.369*** 1.432** −4.858 Slovenia 2.523*** 1.371*** 3.155*** 2.961 2.847*** 1.024 3.968*** 3.555* Peer average 2.281*** 1.326*** 2.440*** 0.834 2.345*** 1.039*** 2.273*** 0.714 EU-27 1.699*** 1.280*** 1.785*** 1.149*** 1.942*** 0.981*** 1.724*** 1.212*** EU-11 1.597*** 1.183*** 1.729*** 1.231 1.803*** 0.986*** 1.798*** 1.482* Source: World Bank staff calculations. Note: For Bosnia and Herzegovina and Montenegro, data on CIT not available. For North Macedonia and Serbia, data available from 2005. *** p < 0.01, ** p < 0.05, * p < 0.1. Section 4 Results 27 28 PFR Fundamentals ――― Tax Buoyancy TABLE 14 VAT buoyancy in southeast European countries 1980–2021 2000–21 Without CPI With CPI Without CPI With CPI Short-run Long-run Short-run Long-run Short-run Long-run Short-run Long-run Country buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy buoyancy Albania 0.279 0.719 0.645 1.950*** 0.279 0.719 0.645 1.950*** Bosnia and Herzegovina Bulgaria 0.921*** 1.037*** 1.635*** 1.161*** 1.855*** 1.100*** 1.875*** 0.967* Croatia 1.083*** 1.195*** 1.217*** 0.114 1.098*** 1.250*** 1.374*** 0.347 Czechia 0.992*** 1.241*** 1.003*** 1.169** 0.916*** 1.359*** 1.160*** 0.627** Hungary 0.993** 1.092*** 0.995** 0.942*** 0.982*** 1.260*** 1.200*** 0.822*** Montenegro North Macedonia 1.064* 0.767*** 0.560 0.700*** 1.064* 0.767*** 0.560 0.700*** Romania 0.665*** 0.966*** 0.534* 0.658*** 0.583 0.836*** 0.814*** 0.008 Serbia 1.502*** 1.057*** 1.852*** 1.272*** 1.502*** 1.057*** 1.852*** 1.272*** Slovak Republic 1.164*** 0.986*** 1.003*** 0.669 0.983*** 0.880*** 0.856* 0.892 Slovenia 1.085*** 0.975*** 1.162*** 0.869*** 1.064*** 0.974*** 1.153*** 0.863*** Peer average 0.975*** 1.003*** 1.060*** 0.950*** 1.033*** 1.020*** 1.149*** 0.845*** EU-27 0.820*** 1.085*** 0.871*** 0.860*** 0.871*** 1.063*** 0.929*** 1.446** EU-11 1.031*** 1.068*** 1.099*** 0.882*** 1.101*** 1.094*** 1.223*** 0.919*** Source: World Bank staff calculations. Note: For Bosnia and Herzegovina and Montenegro, data on VAT not available. For North Macedonia and Serbia, data available from 2005. *** p < 0.01, ** p < 0.05, * p < 0.1. Section 4 Results Measures to broaden the tax base and reduce informal sector and cost of compliance should continue to strengthen the tax system. Given the structure of Serbia’s tax system, fast revenue growth over much of the past decade can be attributed more to discretionary than automatic changes. Policy measures that have been implemented and that have contributed to a reduced informal sector, as well as simplified compliance and administration should be applauded, the tax system itself is not necessarily able to stabilize the economy in the short run, or to ensure fiscal sustainability in the long run. While public debt remained fairly stable during the last year, low tax buoyancy puts the country at risk of increased indebtedness in the future. Features of the tax system and the economy should be improved to make it more progressive and equitable. The regressive structure of PIT and wage rigidity, especially in the public sector, limit the ability of PIT to perform its stabilizing role and support sustainability. Additional reductions in the informal sector would make the tax revenues more responsive to changes in economic output, while other tax instruments, such as taxation of capital income, including property and wealth, should be strengthened, to increase progressivity of the tax system. Buoyant CIT revenues may be contributed to the low starting point and to increased productivity. Continued improvement should be made to maintain the attractiveness of Serbia for investment, with a focus on infra- structure and quality of institutions to support continued growth in profitability. Con- sumption taxes are generally not as buoyant as a result of its regressive structure, but buoyancy of certain consumption taxes, like excises, could be easily improved by regular increases in tax rates. 29 PFR Fundamentals ――― Tax Buoyancy 4.6. Conclusions Although tax buoyancy and tax elasticity are related concepts, they produce different outcomes. Tax elasticity measures the responsiveness of tax revenues to only automatic changes in income while assuming that discretionary changes remain constant, while tax buoyancy measures changes in revenues as a response to both automatic and dis- cretionary changes in income. There are at least three important reasons why measuring tax buoyancy is useful. Firstly, tax buoyancy illustrates how well the tax policy is performing its role in stabi- lizing the economy over the business cycle in the short run, and in ensuring fiscal sustainability in the long run. Secondly, country-specific tax buoyancy estimates of certain tax categories are useful for identifying strengths and weaknesses of a country’s tax system. Thirdly, understanding the institutional and structural determinants of tax buoyancy helps shape the expectations of how tax buoyancy would respond as these characteristics change. This part of the analysis is not included in this note, but is planned for the next phase of this work. Despite certain theoretical expectations about tax buoyancy, short- and long-run buoyancy can vary across countries and tax instruments. It is commonly expected that progressive taxes like PIT or CIT to have higher and regressive taxes like con- sumption taxes have lower tax buoyancy. However, depending on the design and im- plementation of each tax category, as well as the phase of business cycle, actual tax buoyancy may be quite different than theoretically expected. To interpret tax buoyancy estimates, it is therefore important to understand the structure of a country’s tax system to have a good understanding of potential drivers. Controlling for price changes mostly impacts the short-run buoyancy coefficients, while including tax rates reduces the long-run PIT buoyancy. Reduction of short-run tax buoyancy after controlling for price changes suggests that the tax system is not regularly indexed, tax is levied on nominal rather than real household income and on corporate profits and similar features. Sensitivity of long-run PIT buoyancy to inclusion of PIT rates is a signal that changes in PIT policies, measured by changes in top mar- ginal PIT rates, are impacting GDP growth. Short- and long-run buoyancies are just slightly smaller during the contraction period than during the expansion period. 30 References Anderson, Gareth, and Mehdi Raissi. 2022. “Corporate Indebtedness and Low Productivity Growth of Italian Firms.” In Essays in Honor of M. Hashem Pesaran: Panel Modeling, Micro Applications, and Econometric Methodology, Advances in Econometrics, eds. Alexander Chudik, Cheng Hsiao, and Allan Timmermann. Emerald Publishing Limited, 205–28. doi:10.1108/S0731-90532021000043B009. Bayu, Tadele. 2015. “Analysis of Tax Buoyancy and Its Determinants in Ethiopia (Cointegration Approach).” Journal of Economics and Sustainable Development 6 (3). Belinga, Vincent, Dora Benedek, Ruud A De Mooij, and John Norregaard. 2014. “Tax Buoyancy in OECD Countries.” IMF Working Paper WP/14/110. Chudik, Alexander, and Mohammad Pesaran. 2015. “Common Correlated Effects Estimation of Hetero- geneous Dynamic Panel Data Models with Weakly Exogenous Regressors.” Journal of Econometrics 188 (2): 393–420. Cornevin, Antoine, Juan Sebastian Corrales, and Juan Pablo Angel. 2023. “A Deep Dive into Tax Buoyancy: Comparing Estimation Techniques in a Large Heterogeneous Panel.” IMF Working Paper WP/23/71 (2023–2071). Deaton, Angus, and John Muellbauer. 1980. Economics and Consumer Behavior. Cambridge: Cambridge University Press. doi:10.1017/CBO9780511805653. Deli, Yota, Abian Garcia Rodriguez, Ilias Kostarakos, and Petros Varthalitis. 2018. “Dynamic Tax Revenue Buoyancy Estimates for a Panel of OECD Countries.” ESRI Working Paper, No. 592. Dillinger, William. 1991. “Urban Property Tax Reform: Guidelines and Recommendations.” World Bank PRE Working Paper 710. https://EconPapers.repec.org/RePEc:wbk:wbrwps:710. Ditzen, Jan. 2018. “Estimating Dynamic Common-Correlated Effects in Stata.” The Stata Journal 18 (3): 585–617. Dudine, Paolo, and João Tovar Jalles. 2018. “How Buoyant Is the Tax System? New Evidence from a Large Heterogeneous Panel.” Journal of International Development 30 (6): 961–91. Gupta, Sanjeev, João Tovar Jalles, and Jianhong Liu. 2022. “Tax Buoyancy in Sub-Saharan Africa and Its Determinants.” International Tax and Public Finance 29 (4): 890–921 . Hausman, Jerry A. 1978. “Specification Tests in Econometrics.” Econometrica 46 (6): 1251–71. Kao, Chihwa. 1999. “Spurious Regression and Residual-Based Tests for Cointegration in Panel Data.” Journal of Econometrics 90 (1): 1–44. Lagravinese, Raffaele, Paolo Liberati, and Agnese Sacchi. 2020. “Tax Buoyancy in OECD Countries: New Empirical Evidence.” Journal of Macroeconomics 63 (C): 103189. Lancaster, Kelvin. 1971. Consumer Demand : A New Approach. Columbia University Press. https://cir.nii. ac.jp/crid/1130000797872087296. Mansfield, Charles Y. 1972. “Elasticity and Buoyancy of a Tax System: A Method Applied to Paraguay.” IMF Staff Papers 19 (2): 425–46. 31 PFR Fundamentals ――― Tax Buoyancy Mawia, Mary, and Joseph Nzomoi. 2013. “An Empirical Investigation of Tax Buoyancy in Kenya.” African Journal of Business Management 7 (40): 4233–46. Musa, Ojonago Daniel, Andenyangtso Bulus, Christopher Chukwudi Nwokolo, and Denis Nfor Yuni. 2016. “Tax Buoyancy and Elasticity in Nigeria: The Case of Aggregate Tax.” International Journal of Devel- opment and Economic Sustainability 4 (4): 20–31. Naape, Baneng, and Nyasha Mahonye. 2020. How Buoyant Is the South African Tax System: An ARDL Approach. Economic Research Southern Africa Working Paper 835. Ochieng, Brandon, and Nlandu Mamingi. 2022. “Estimating the Sizes of Buoyancy and Elasticity of the Tax System in Barbados Over the Period 1990 to 2019.” Central Bank of Barbados Working Paper 22 (2). Pedroni, Peter. 1999. “Critical Values for Cointegration Tests in Heterogeneous Panels with Multiple Re- gressors.” Oxford Bulletin of Economics and Statistics 61 (S1): 653–70. doi:https://doi.org/10.1111/ 1468-0084.0610s1653. Pesaran, M. Hashem. 2006. “Estimation and Inference in Large Heterogeneous Panels with a Multifactor Error Structure.” Econometrica 74 (4): 967–1012. Pesaran, M. Hashem, Yongcheol Shin, and Ron P. Smith. 1999. “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels.” Journal of the American Statistical Association 94 (446): 621–34. doi:10.2307/2670182. Pesaran, M. Hashem, and Ron Smith. 1995. “Estimating Long-Run Relationships from Dynamic Hetero- geneous Panels.” Journal of Econometrics 6 8(1): 79–113. doi:https://doi.org/10.1016/0304-4076(94) 01644-F. Scott-Joseph, Ankie, Tracy Maynard, Nichelle Yearwood, and Jamila Beckles. 2016. “Analysis of Tax Buoy- ancies and Revenue Performance in Barbados.” Central Bank of Barbados Working Paper 16 (7). Stockhammer, Engelbert. 2013. “Why Have Wage Shares Fallen? An Analysis of the Determinants of Functional Income Distribution.” In Wage-Led Growth: An Equitable Strategy for Economic Recovery, Springer, 40–70. Tanchev, Stoyan, and Ivan Todorov. 2019. “Tax Buoyancy and Economic Growth: Empirical Evidence of Bulgaria.” Journal of Tax Reform 5 (3): 236–48. Upender, Mulakala. 2008. “Degree of Tax Buoyancy in India: An Empirical Study.” International Journal of Applied Econometrics and Quantitative Studies 5 (2): 59–70. Westerlund, Joakim. 2005. “New Simple Tests for Panel Cointegration.” Econometric Reviews 24 (3): 297–316. doi:10.1080/07474930500243019. Wolswijk, Guido. 2007. “Short-and Long-Run Tax Elasticities: The Case of the Netherlands.” European Central Bank Working Paper No. 763. 32