Abstract

The key policy challenges facing countries in Central, Eastern, and Southeastern Europe remain broadly unchanged, among them supporting domestic demand, addressing financial crisis legacies, rebuilding buffers against external shocks, and improving the business environment. Country-specific priorities depend on how far along they are in the postcrisis adjustment and their exposure to external risks.

Annex I. CESEE: Growth of Real GDP, Domestic Demand, Exports, and Private Consumption, 2013–16

(Percent)

article image
Source: IMF, World Economic Outlook database.

Weighted averages using 2014 GDP valued at purchasing power parity.

Includes Albania, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Kosovo, Latvia, Lithuania, Macedonia FYR, Moldova, Montenegro, Poland, Romania, Russia, Serbia, Slovak Republic, Slovenia, Turkey, and Ukraine.

CESEE excluding Czech Republic, Estonia, Latvia, Lithuania, Slovak Republic, and Slovenia.

Includes Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic, and Slovenia.

Annex II. CESEE: Consumer Price Index Inflation, Current Account Balance, and External Debt, 2013–16

(Percent)

article image
Source: IMF, World Economic Outlook database.

Weighted averages using 2014 GDP valued at purchasing power parity.

Includes Albania, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Kosovo, Latvia, Lithuania, Macedonia FYR, Moldova, Montenegro, Poland, Romania, Russia, Serbia, Slovak Republic, Slovenia, Turkey, and Ukraine.

CESEE excluding Czech Republic, Estonia, Latvia, Lithuania, Slovak Republic, and Slovenia.

Includes Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic, and Slovenia.

Annex III. CESEE: Evolution of Public Debt and General Government Balance, 2013–161

(Percent of GDP)

article image
Source: IMF, World Economic Outlook database.

As in the WEO, general government balances reflect IMF staff’s projections of a plausible baseline, and as such contain a mixture of unchanged policies and efforts under programs, convergence plans, and medium-term budget frameworks. General government overall balance where available; general government net lending/borrowing elsewhere. Public debt is general government gross debt.

Weighted averages using 2014 GDP valued at purchasing power parity.

Reported on a cash basis.

Regarding the overall balance, this includes fiscal room for donor-financed capital projects (for 2016-2018 period), which might not be fully utilized by year-end. Public debt includes former Yougoslav debt, not yet recognized by Kosovo.

General government balance: the measure reflects augmented balance, which adds to the balance of general government outlays for banks recapitalizations and is related to called guarantees of publicly-guaranteed debt.

Includes Albania, Belarus, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Kosovo, Latvia, Lithuania, Macedonia FYR, Moldova, Montenegro, Poland, Romania, Russia, Serbia, Slovak Republic, Slovenia, Turkey, and Ukraine.

CESEE excluding Czech Republic, Estonia, Latvia, Lithuania, Slovak Republic, and Slovenia.

Includes Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovak Republic, and Slovenia.

Annex IV. Methodology to Obtain Cyclically-Adjusted Revenues/Expenditures17

A cyclical adjustment was applied to revenues/expenditures following the disaggregated approach proposed in Bornhorst and others (2011). The purpose of cyclical adjustment is to decompose the overall balance into cyclical and cyclically adjusted components:

OB=CB+CAB(1)

where OB is the overall balance, CB is the cyclical balance (the part of the fiscal overall balance that automatically reacts to the business cycle), and CAB is the cyclically adjusted balance (the part of the overall balance that is left after cyclical movements are taken out), expressed in nominal terms. The disaggregated approach computes the cyclically adjusted balance as a function of individual cyclically adjusted revenue and expenditure categories:

CAB=(i=1NRiCA)GCA+RNCAGNCA(2)

where RiCA represents the cyclically adjusted component of the i-th revenue category, GCA represents cyclically adjusted primary expenditures, while RNCA and GNCA contain all revenue/ expenditure categories that do not require cyclical adjustment (Girouard and Andre, 2005).

On the revenue side, the elasticity of each revenue category can be decomposed into two factors. The output elasticity of tax revenue (εRi,Y) is the product of the elasticity of tax revenues (Ri), with respect to the relevant tax base (Bi),εRiBi, and the elasticity of the tax base relative to the output gap, εBi,Y:

ɛRi,Y=ɛRiBiɛBi,Y(3)

Applying this decomposition to the computation of cyclically adjusted revenue yields:

RiCA=Ri((Y*Y)ɛB,Y)ɛR,B(4)

Assuming, or deriving, the value of the tax elasticity with respect to its base is the first step. In addition to statutory tax rates, derivation also requires knowledge of the income distribution. The second step is an econometric estimation of the sensitivity of the relevant tax bases with respect to the output gap. This requires specifying macroeconomic proxies for the tax bases. For income taxes and social security contributions a common proxy is the wage bill, for corporate income taxes, the tax base is a measure of corporate profits, whereas private consumption serves as a base for indirect taxes. With these two elasticities at hand, the elasticities of tax revenue with respect to the output gap can be computed. The resulting elasticities of revenue categories with respect to the output gap are usually larger than one for income taxes (reflecting progressivity), around one for indirect taxes (reflecting generally flat indirect (VAT) tax rates), and somewhat smaller than one for social security contributions. For practical reasons, the elasticities of the several revenue components with respect to the output gap were drawn from Girouard and Andre (2005), taking the mid-point of their estimated elasticities. As such, we assume an elasticity of 1.3 for personal income taxes, of 1.5 for taxes on corporate profits, of 0.7 for social security contributions, and of 1 for indirect taxes, including taxes on goods and services and property tax.

Similarly on the expenditure side, the elasticities can also be decomposed into two factors. Current transfers—in particular unemployment benefits—are most likely to display a cyclical behavior owing to the benefit system. In contrast, nominal spending on other items such as wages and goods and services or capital spending is likely to be largely independent of the business cycle, not requiring any adjustment. As with revenues, the elasticities of expenditure with respect to the base can be assumed or derived. For simplicity, we only adjust the expenditure on social benefits, assuming an elasticity of -0.5. For all other expenditure categories, the elasticity relative to the output gap is assumed to be zero. All results in the chapter are robust to changes in elasticities.

Annex V. Country Coverage18

The empirical analysis in this chapter covers 76 countries during 1990-2014 including CESEE, as well as other advanced and emerging economies whose 2014 GDP (constant 2005 USD, PPP) is greater than 3.7 billion USD (20th percentile of all advanced and emerging economies), and for which cyclically-adjusted fiscal variables can be computed.

The sample includes the following countries:

  • CESEE: Bulgaria, Bosnia and Herzegovina, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Moldova, Poland, Romania, Russia, Serbia, Slovak Republic, Slovenia, Turkey, Ukraine.

  • Advanced Europe: Austria, Belgium, Switzerland, Cyprus, Germany, Denmark, Spain, Finland, France, Greece, Ireland, Italy, Luxemburg, Netherlands, Norway, Portugal, Sweden, and United Kingdom.

  • Other advanced: Australia, Canada, Israel, Japan, New Zealand, and United States of America.

  • Asia: China, Indonesia, India, Malaysia, Philippines, Singapore, and Thailand.

  • Latin America and the Caribbean: Argentina, Bahamas, The, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Peru, Uruguay, and Venezuela.

  • Africa: Kenya, Mauritius, and South Africa.

  • Middle East and Central Asia: Armenia, Algeria, Egypt, Georgia, Kazakhstan, Korea, Lebanon, Pakistan, and Tunisia.

Annex VI. Budget Structures and Country Characteristics19

To compare the budget structure of CESEE countries with that of comparator countries, expenditure and revenue categories are regressed on structural country characteristics. Using the sample of countries defined in Annex V, we estimate equations of the form:

Fiscalit=αi+β3Xit+ɛit(1)

where the index i=1,…,N and t=2008, 2014 are respectively country- and time-indicators. In Eq. (1), Fiscal is a set of cyclically-adjusted expenditure and revenue variables, expressed relative to potential GDP, and X is a set economic characteristics. In order to assess changes to the budget structure since the global financial crisis, Eq. (1) is estimated for two separate years: 2008 and 2014.

Revenues and expenditure, taken from the World Economic Outlook (WEO) Database, refer to consolidated general government, and are cyclically adjusted in percent of potential GDP. In addition, the IMF’s Fiscal Affairs Department Revenue database and Article IV consultation reports were used to fill gaps in the data series. For spending categories, four categories are considered in addition to total spending: public consumption of goods and services, compensation of employees, transfers, and capital spending. For revenues, four categories are considered on top of total revenues: taxes on corporate profits (CIT), taxes on goods and services, social security contributions, and the personal income tax (PIT). Structural country characteristics are from the World Development Indicator Database: GDP per capita (in 2005 USD, PPP), log of GDP, dependency ratio (the ratio of dependents—people younger than 15 or older than 64—to the working-age population—those ages 15-64), trade openness (sum of exports and imports of goods and services to GDP), population density (number of people per sq. km. of land area), and natural resource rents (in percent of GDP).

Annex Table VI.1 below presents the results for spending and revenue categories. Per capita GDP is generally found positively and statistically significantly correlated with expenditure and revenue components, which is supportive of the Wagner’s law. A higher GDP per capita potentially increases the demand for public services in reflection of a higher degree of economic and institutional sophistication, also requiring higher government revenues. The negative coefficient on public investment could, in turn, reflect structural economic transformation (e.g. Turrini (2004)). Economy size (log GDP) appears to be negatively correlated with public investment, but has strong positive correlation with revenues from social contributions and expenses related to transfers. In addition a high dependency ratio tends to be associated with lower expenditure and lower revenues from consumption taxes. Coefficient estimates on trade openness are mostly statistically insignificant, except for revenues from social contributions, which may reflect higher demand for social insurance against external risks in more open economies (e.g. Rodrik (1998)). Population density matters according to the empirical results: higher population density could imply higher urbanization of the economy which helps improve efficiency of public sector operation. Resource rich economies have different budget structures that are broadly reflected in the estimated coefficients: on the revenue side, resource-rich economies tend to collect less from non-resource tax revenues (Crivelli and Gupta, 2014), except for the CIT tax on companies operating in natural resource sectors. On the expenditure side, there is usually high infrastructure investment needs from natural resource sectors.

Annex Table VI.1:

Government Budget Structure and Structural Characteristics

article image
Robust p-values in parentheses*** p<0.01, ** p<0.05, * p<0.1

The residuals from the estimated equations on revenue and expenditure components are used to assess budget structures in CESEE relative to their structural characteristics. On the expenditure side, current spending (wages and consumption of goods and services) is relatively high in CESEE as compared to the level consistent with the country’s structural characteristics, except for Turkey. When assessing the changes over time, however, both categories of spending—in particular the public sector wage bill—shows a significant drop in SEE, CIS, and Baltic countries from 2008 to 2014. As for public investment, its level in 2008 appears particularly high in SEE and Turkey, with SEE, CIS, and Baltic countries experiencing a significant drop over time. Transfers, however, have remained high among CEE, SEE, and CIS countries. On the revenue side, many CESEE countries except for CIS countries managed to increase their reliance on taxes on goods and services since 2008, although some countries, such as CEE and SEE, already had relatively high levels. Additionally, CESEE economies have on average brought down CIT revenues, expect for Turkey. Nevertheless, changes in labor income taxes—including social contributions—are moderate, and levels remain high compare to country’s structural characteristics.

Annex Figure 1.
Annex Figure 1.

Residuals from Expenditure Benchmarking Regressions

(Percent of potential GDP)

Sources: IMF World Economic Outlook database; and IMF staff calculations.
Annex Figure 2.
Annex Figure 2.

Residuals from Revenue Benchmarking Regressions

(Percent of potential GDP)

Sources: IMF World Economic Outlook database; and IMF staff calculations.

Annex VII. Fiscal Policy Instruments and Economic Growth: A Panel Data Analysis20

The relationship between revenue and expenditure variables and economic growth is estimated using unbalanced panel data for 76 developing and advanced countries during 1990-2014. The empirical methodology follows that in Afonso and Alegre (2011), in estimating a dynamic growth equation as follows:

yit=αi+β1yit1+β2Fiscalit+β3Xit+μt+ɛit(1)

where the index i=1,…,N and t=1,…,L are respectively country- and time-indicators (so that αi and μt are country- and time-specific effects). In Eq. (1), y indicates the growth rate of per capita output of country i during year t (yit = ΔlnGDPit), Fiscal is a set of fiscal variables expressed as a percentage of GDP, and X is a set of non-fiscal control variables. Eq. (1) is estimated using ordinary least squares with country and time fixed effects (OLS-FE). The long-term effect of different fiscal variables on growth is captured by θ=β2(1β1). We also investigate whether the effect of revenues and expenditures on growth differs across regions by interacting the variable of interest with regional dummies for CESEE and Advanced Europe.

As described in Kneller et al. (1999), when estimating the impact of each spending category on output growth, an omitted variable represents the underlying assumption about how to finance the additional expenditure. In all cases, the omitted variables are the reminder of the public expenditures. Similarly for revenues, the omitted variables are the remaining public revenues. Data are cyclically adjusted, inter alia to prevent reverse causality from growth to revenue/expenditure categories. Feedback effects cannot be excluded as cyclical adjustment is difficult to get right, especially for corporate taxes. Results could also capture things like governments adjusting budgets in response to high/low growth. This said, results do not change materially when varying the set of elasticities used in cyclical adjustment, and neither when using Blundell and Bond (1998) system generalized method of moments estimator (GMM) instead of OLS-FE.

The fiscal variables, taken from the World Economic Outlook (WEO) database, refer to consolidated general government and are expressed as ratios to GDP. In addition, the IMF’s Fiscal Affairs Department Revenue database and Article IV consultation reports were used to fill gaps in the data series. For spending categories, five categories are considered: public consumption (purchase/use of goods and services), compensation of employees, transfers, capital spending, and other expense. For revenues, seven categories are considered: taxes on corporate profits (CIT), taxes on labor income (PIT), taxes on goods and services, taxes on property, social contributions, other tax revenues, and other (non-tax) revenue.

In addition, we have included six control variables: labor force (as a growth rate), private investment (in percent of GDP), terms of trade (as a growth rate constructed from an index series in which the year 2000 takes the value 100), population (growth rate), oil rents (in percent of GDP), and the overall budget balance (in percent of GDP). The inclusion of the production factors related to capital increase (proxied by private investment) and labor force growth follows from the related literature. Population growth may explain output growth. Several studies have suggested the relevance of terms of trade (Odedokun, 2001; Bose et al., 2003; Gupta et al., 2005). Oil rents are included to capture potential negative influence of natural-resource revenues on domestic revenues and expenditures (Crivelli and Gupta, 2014). The unit specific term in our panel model αi takes into account the effect of time-invariant idiosyncratic characteristics of each country, whose impact has been suggested in previous studies, such as the initial levels of GDP or human capital, etc.

Annex Table VII.1 and VII.2 below present the long-term results for different government spending and revenue categories. The coefficients shown there are the estimated computed as explained above. Annex Table VII.1, Column 1 presents the results for public consumption spending. The computed long-run effects for CESEE as well as for advanced Europe show a negative and significant relationship with economic growth. Also a statistically significant and negative coefficient has been estimated for transfers (column 3), while capital spending appears to have a positive and significant relationship to long-term economic growth (column 4). As such, for instance, an increase in public investment by 1 percentage point of GDP, financed by an equivalent decrease in current public expenditure (omitted variables), would increase long-term real per capita GDP growth by about 0.4 percentage points among CESEE countries. For the revenue categories, the estimated coefficients in Annex Table VII.2 point to a negative and significant impact of taxes on corporate profits (column 2) and social contributions (column 4) on growth, while a much smaller and not statistically significant impact of broad-base consumption taxes (column 1) or property tax (column 3).

Results are qualitatively similar for Advanced Europe. Note though that for the “rest of the world”—i.e. countries outside Europe, captured by the base coefficient for each spending/revenue category—some results deviate. Differences to Europe are notable especially on the revenue side, where for the “rest of the world” no category has a significant coefficient. This may reflect heterogeneity—the group contains a diverse set of advanced and emerging economies—or structural characteristics of these revenue systems that are insufficiently captured by the model specification.

In order to assess the overall growth friendliness of CESEE government budget structures, additional assumptions are needed to indentify the marginal impact of revenue and expenditure components on long-term growth. Given the omitted variable assumption used in estimating Eq. (1) above (controlling for total revenue and total expenditure, as well as for the overall budget balance), the parameter estimates of the expenditure and revenue components cannot be directly interpreted as the marginal impact on growth. For example, the impact of a one percentage point increase of a given expenditure component implies in the regressions, a drop of equivalent size in other expenditure components. As a result, the estimated coefficients can only be interpreted as the net impact (net of the impact of reducing other expenditure components) on long-term growth.

We make the following assumptions in order to identify the marginal impact of expenditure and revenue components on long-term growth21:

  • When computing the net growth impact, the compensating change of other expenditure (revenue) components is equally distributed among other expenditure (revenue) components.

  • The sum of the marginal impact of expenditure (revenue) components on long term economic growth is zero.

With these additional assumptions, and the estimated coefficients from the regressions on expenditure and revenue components above, a system of equations can be solved separately for revenue (expenditure) components that allows identification of the estimated marginal effects. This transformation results, as expected, in marginal effects on long-term growth that are consistently lower (in absolute value) than the estimated coefficients above.

Annex Table VII.1.

Government Spending and Economic Growth

article image
Notes: OLS, with country fixed effects. Full set of controls and year dummies in all regressions. Robust standard errors, in parenthesis; ***(**,*) indicate significance at 1(5,10) percent.
Annex Table VII.2.

Government Revenue and Economic Growth

article image
Notes: OLS, with country fixed effects. Full set of controls and year dummies in all regressions. Robust standard errors, in parenthesis; ***(**,*) indicate significance at 1(5,10) percent.

Annex VIII. Large Fiscal Consolidations: Country Experiences22

Several CESEE countries underwent sizeable consolidation in the wake of the global financial crisis. This Annex summarizes the main policies taken in countries with the largest fiscal efforts. Overall, many measures went into a growth-friendly direction, broadly corroborating the findings of the quantitative analysis. That said, they also comprised spending measures that came with hardship for certain segments of the population. Some revenue measures—such as sector-specific taxes—may be turn out to be harmful to growth.

Romania

Fiscal consolidation started in 2009 in the context of an IMF supported program. Adjustment focused on expenditure consolidation, seeking to contain entitlement and wage costs through cuts in wages and social transfers, parametric reforms to the pension system, including the removal of special pension regimes, and the introduction of a unified wage setting system.

Structural fiscal reforms complemented consolidation, including on the pension system, tax administration, and the public wage framework. A Fiscal Responsibility Law was enacted to streamline budgeting. A multi-year public financial management structure was introduced while limiting intra-year budget revisions. Fiscal rules were introduced to regulate spending, public debt, and the primary deficit. A framework for managing guarantees and other contingent liabilities was approved. The local public finance law was amended to bolster fiscal discipline. Reform implementation was uneven though, particularly in the areas of revenue administration and health care reform.

Main measures:

  • Wage bill cut (2 percent of potential GDP, cyclically adjusted): A key objective was to reduce the public sector wage bill back to the 2007 level through a 25 percent cut in public wages. Public employment was reduced by about 16 percent. The resulting public wage bill in 2014 was consistent with the 7 percent of GDP cap set in Romania’s Fiscal Strategy.

  • Capital spending cut (2½ percent of potential GDP, cyclically adjusted): efficiency of Romania’s high capital spending had been compromised by the lack of a robust framework and capacity for developing, prioritizing, and executing public-investment projects. Cuts to capital spending focused on non-priority, inefficient projects. In recent years the focus has shifted to improve absorption of EU funds, by strengthening targeting rules and public procurement.

  • Transfer cuts (0.1 percent of potential GDP, cyclically adjusted): A parametric reform to the pension system was an essential pillar for long-term fiscal sustainability, including also the removal of special pension regimes. Short-term budgetary impact was only marginal, however, with the focus on preserving the scope and size of the safety net system (social protection).

  • Revenue increases: Only 1/3 of the adjustment was achieved through revenue measures, mostly explained by an increase to the VAT rate from 19 to 24 percent that triggered an increase in yields of about 1 percent of GDP. The modest revenue increase relative to the large VAT rate hike is explained by a weak revenue administration.

Estonia

Reflecting the early onset of the crisis in Estonia, fiscal consolidation started early in 2008. Supplementary budgets in February and June of 2009 contained consolidation measures of 7½ percent. Many measures were on the revenue side, such as VAT increases and higher excise taxes. Other measures included social benefit reductions, cuts in operational spending, as well as land sales and discretionary spending cuts. Overall, the efforts led to a 2009 fiscal deficit of 1.7 percent of GDP (in ESA terms), which helped paving the way for euro adoption in 2011.

Main measures:

  • Wage bill and consumption spending cuts (2 percent of potential GDP, cyclically adjusted): Large increases, particularly in current spending, resulted in an ill-timed loosening of fiscal stance in boom years. Fiscal consolidation was aimed at reversing these earlier spending. Cuts in spending in Estonia were possible due to lower rigidities as compared to other Baltic countries.

  • Transfer cuts (1 percent of potential GDP, cyclically adjusted): Cuts in social transfers reflect lower replacement rates for unemployment benefits.

  • Revenue collection increases (1½ percent of potential GDP, cyclically adjusted): mostly base-broadening measures. Tax collection held up very well, as a result of improvements in tax administration at the onset of the crisis. Estonia has a revenue-productive and cost-effective tax system, characterized by an internationally acclaimed low compliance burden (with some 90 percent of taxpayers filing electronically). As a result, they achieved a sizeable increased in revenue collection despite only marginal increases to VAT rates (2 percent) and excise taxes.

Bosnia and Herzegovina

Financing constraints triggered fiscal consolidation in 2010-2011. A temporary financing rule restrained spending on capital goods to offset overruns in wage and other current spending. Since 2012, adjustment efforts have increasingly focused on rationalizing public expenditures and on improving the composition of spending. There have also been structural fiscal reforms, including strengthening tax administration and tax compliance, reforms of the system of rights-based benefits; a comprehensive overhaul of the health sector and pension systems; strengthening the medium term budget framework; and streamlining public administration.

Main measures:

  • Transfer cuts (2½ percent of potential GDP, cyclically adjusted): The largest contribution came from reduction in war-related benefits. As a result of a new privileged pension law in the Federation, benefits of existing war veterans were reduced substantially.

  • Capital spending cuts (1½ percent of potential GDP, cyclically adjusted): In addition, a temporary financing rule restrained non-priority spending on capital goods, in part explained by delays in official foreign financing.

  • Tax revenue increases (4 percent of potential GDP, cyclically adjusted): Mostly due to better tax administration and base-broadening measures. Large gains in revenue collection have been explained by a significant effort to strengthen tax administration. Tax compliance measures including among others, a broader exchange of information between collection agencies in the Federation, and compulsory registration of farmers to broaden the tax base for social security contributions have resulted in increased collection of all main taxes without major changes to tax rates.

Czech Republic

The Czech Republic reacted to the global financial crisis at first with a fiscal stimulus: in 2007-09, the structural fiscal balance widened by more than 3 percent of GDP, while public debt increased to 38 percent of GDP by 2010. As the crisis intensified, stimulus was withdrawn and followed by large structural consolidation of more than 4 percent of GDP in 2010–13.

Main measures:

  • Revenue: Mostly due to tax rate changes. Fiscal stimulus at the onset through cutting the CIT rate, resulting in 1 percent of GDP lower CIT revenue collection. Afterwards, graudual increase in the VAT standard rate from 19 to 21 percent and increases in reduced VAT rates, resulting in higher revenue collection by 1½ percent of GDP.

  • Capital spending cuts (1½ percent of poten