Abstract

This report analyses the main economic developments and achievements in the Western Balkan countries, and lays out the key macroeconomic policy challenges for the future.

Annex 1.1. Growth Accounting

A standard growth accounting analysis can shed light on key drivers of growth in the Western Balkans relative to those in the New Member States. Growth accounting, based on a standard Cobb-Douglass production function, allows for decomposing real GDP growth into contributions of human capital, physical capital and total factor productivity:1

Yt=AtKtαHt*(1α)(1)and  Ht*=htLt*,(2)

where Yt represents domestic output in period t, Kt the physical capital stock, Lt the employed labor force, ht the index of human capital per worker, and At total factor productivity. The analysis uses Penn Tables 8.0 data on capital stocks; country-specific measures of the labor share in GDP; and an index of human capital per person, based on years of schooling (Barro and Lee 2010) and returns to education (Psacharopoulos 1994; and Inklaar and Timmer 2013).2 Using (1), real GDP growth is decomposed into the contribution of physical capital accumulation, employment growth, and accumulation of human capital per worker. Total factor productivity (TFP) growth is the residual, that is, output growth not explained by either growth in capital, growth in adjusted labor, or:

Y^t=αK^t+(1α)L^t+(1α)h^t+A^t.(3)

Gains in TFP and capital accumulation have been major drivers of growth in the Western Balkans over the past decade. Growth in TFP, reflecting more efficient use of inputs, has long been recognized as an important source of improvements in income and welfare. Cross-country differences in income levels and growth rates are mostly due to differences in productivity (Klenow and Rodriguez-Clare 1997); and Easterly and Levine 2001). Results indicate that gains in TFP in the region—the residual in the growth accounting analysis—explain about half of annual average growth, comparable to New Member States (Annex Figure 1.1.1), in line with other studies on transition economies (Campos and Coricelli 2002; IMF 2009; and EBRD 2013). This TFP improvement has likely reflected the effects of transition to a market economy, including enterprise restructuring and privatization, and increased technology transfer from the European Union.

Annex Figure 1.1.1.
Annex Figure 1.1.1.

Western Balkans: GDP Growth and Contributions

(Percent)

Sources: Inklaar and Timmer (2013); and University of Groningen Growth and Development Centre.Note: * For Bosnia and Hertzegovina, FYI Macedonia and Montenegro employment growth, human capital per worker is estimated based on UNDP dataset on educational attainment and on the methodology from Barro and Lee (2012).

Physical investment has also expanded rapidly and contributed to solid growth performance over the past decade. For example, the massive buildup of productive capacity in Bosnia and Herzegovina—financed by foreign direct investment and an externally financed credit boom—on average accounted for about 60 percent of the observed output growth over 2001–08. Similarly, capital accumulation accounted on average for about 40 percent of the observed output growth in Montenegro and FYR Macedonia, and about 30 percent in Croatia and Albania. The contribution of capital accumulation to Western Balkan state growth is comparable, on average, with that of New Member States.

The low contribution of human capital accumulation—employment growth, adjusted for schooling—constitutes a key difference between the Western Balkans and the New Member States. The contribution of human capital accumulation to growth was more significant only in FYR Macedonia and Montenegro and negative for Albania and Serbia. In Bosnia and Herzegovina and Croatia, human capital accumulation was positive in the boom years, but since 2009 this increase was more than offset by employment losses associated with the global crisis.

Annex 1.2. Convergence Analysis

Convergence is most commonly understood as the process of decreasing differences in income per capita across economies over time. Convergence happens because, in theory, poor countries should grow faster than rich countries due to decreasing returns on capital and increasing costs of productivity advancement for countries already on the production frontier. In reality, however, various structural factors may hinder convergence.

There are two commonly-used ways to test the existence of convergence among a group of countries: (1) calculating whether the dispersion of income per capita across countries is decreasing over time (Annex Figure 1.2.1); and (2) regressing GDP growth on initial income levels to see whether poorer countries grow faster than richer countries (Annex Figure 1.2.2). The literature has emphasized that the two approaches measure different phenomena (Quah 1996). We apply both methods to examine the convergence performance of Western Balkan States (WBS) and compare it with that of the New Member States (NMS).

Annex Figure 1.2.1.
Annex Figure 1.2.1.

Dispersion of the Logarithm of GDP per Capita

(Cross-country standard deviation of log GDP per capita)

Sources: Penn World Table; WEO; and IMF staff calculations.
Annex Figure 1.2.2.
Annex Figure 1.2.2.

Catching up with Advanced Europe

(Average country GDP per capita as percent of average EU17 GDP per capita)

Sources: Penn World Table; and IMF staff calculations.

The first method shows that the dispersion of GDP per capita across the Western Balkans and advanced EU countries (EU15) has steadily declined since 1993, especially during the second half of the 1990s and the boom years in the 2000s. Since the global crisis, however, the decline has mostly stopped. In contrast, the New Member States have continued to close the economic gap with advanced EU economies since the crisis, albeit at a slower pace than during the boom years. As a result, the gap between the Western Balkans and the New Member States has increased.

For the second approach, we estimate the following equation with NMS, WBS, and EU15 in the sample:

growthi,t=β0+β1disti,t1+β2disti,t1×WB+β3disti,t1×NMS+β4WB+β5NMS+β6GEO+ui,t,

where growthi,t is the GDP per capita growth rate for Country i; disti, t-1 is the GDP per capita gap between the average level of EU15 and country i from the previous period; WB and NMS are dummy variables for the Western Balkan and NMS regions; and GEO is the physical distance of Country i to advanced EU economies (proxied by the distance of the capital of country i to Berlin). A larger coefficient for disti,t-1 and its interactions means poorer countries have grown faster than richer countries, that is, it provides evidence of convergence.

The regression results indicated strong convergence for the WBS during 1990–2000 due to the bounce-back and reconstruction after the regional conflict subsided, while there was little convergence during that period in the NMS (Annex Table 1.2.1). For 2000-2007, however, the convergence coefficient was small and insignificant for the WBS. This was due to the fact that the poorer countries in the region, such as Albania and Bosnia and Herzegovina, actually grew more slowly than the richer ones, such as Croatia, during this period. In contrast, the convergence coefficient was positive and highly significant for the NMS during the period. For the period since the onset of the global crisis, the convergence coefficient is positive and significant for the WBS, though it is smaller than that for the NMS.

Annex Table 1.2.1.

Speed of Convergence: Western Balkan States versus New Member States

article image
Note: Standard errors in parentheses. *: p < 0.1, **: p < 0.05, ***: p < 0.01.

Next, we examine if structural factors can affect the speed of convergence in WBS and NMS (Annex Table 1.2.2). For this purpose, we adapt the convergence regression to the following:

growthi,t=α0+α1disti,t1+α2disti,t1×factori,t1+α3factori,t1+α4controlsi,t+vi,t

where factori is the structural factor under examination. If a factor facilitates convergence, we shall expect the associated estimate for α2 to be positive, and vice versa. The regression is estimated using data of nonoverlapping five-year intervals after 2000. The following are the structural factors we examined and their data sources:1

  • Quality of governance (World Bank Governance Indicators)

  • Development of market-oriented institutions (EBRD Transition Indicator)

  • Level of human capital (Human capital indicator from the Penn World Table)

  • Government’s share in GDP (Penn World Table)

  • Unemployment rate (IMF, World Economic Outlook, cyclically adjusted)

  • Financial development level (credit to GDP ratio, IFS, cyclically adjusted)

Annex Table 1.2.2.

The Impact of Structural Factors on Convergence

(Dependent variable: relative growth to EU 15)

article image

The results show that high-quality governance, market-oriented institutions, a strong human capital base, and a more developed financial system facilitate catching-up by poorer countries. In contrast, dominance of the public sector in the economy and high unemployment hinder the catching-up process.

Annex 1.3. Trade Linkages

The strength of regional integration can be inferred from the augmentation of a standard gravity model that relates cross-border trade flows to population, economic size, and geographical distance. In particular, we follow (Paas and Tafenau 2005) and augment a standard gravity model with an indicator variable that takes the value 1 when the country pair belongs to the Western Balkans (WBS) group of countries. We also test the strength of linkages between Western Balkan and New Member States, as well as the Western Balkan States and main euro zone partners. We expect the coefficient on the dummy to be positive, indicating that belonging to the WBS group increases the size of cross-border trade.1

The following model is estimated:

Tijt=α+β1GDPpcit+β2GDPpcjt+β3POPit+β4POPjt+β5distij+β6I(WB)ij+εijt,(1)

where T denotes bilateral trade (exports and imports, respectively) imports in U.S. dollars, GDPpc nominal GDP per capita in U.S. dollars, POP populations, dist distance between capitals and I(WB) the Western Balkans dummy. All variables are in log form. Because the distance between countries does not vary within the panel unit, we use fixed-effects between estimators for our regressions.

Econometric estimates do not reject the hypothesis that the particular strength of linkages between WB economies is an additional explanation for the size of their cross-border trade. While similar results hold for the links between WBS and NMS, or WBS and eurozone countries, those ties appears weaker. This seems plausible given the improvement of intra-regional relations since 2000 that has led to an increase in intraregional trade on the back of historically similar institutional frameworks and languages, in addition to the growing integration into euro area supply chains.

Annex Table 1.3.1.

Dependant Variable: Nominal Bilateral Trade

(Exports)

article image

WBS: excluding Kosovo

Eurozone core: Germany, Italy

Annex Table 1.3.2.

Dependant Variable: Nominal Bilateral Trade

(Imports)

article image

WBS: excluding Kosovo

Eurozone core: Germany, Italy

Annex 1.4. Structural Reform Gaps and Economic Growth

A. Estimating Reform Gaps

For a number of indicators we estimate country-specific reform gaps by comparing Western Balkan countries with the New Member States and an average EU country, taking into account some other country-specific characteristics. For each indicator, the reform gap—the distance of the indicator’s value for country X from the NMS or the European Union average—is derived first by estimating the following regression:

Iik=α+βXi+i,(A1)

where Iik is indicator k in country i; Xt is the set of controls—GDP per capita, geographical location (dummy for the sub-Saharan region), common history (Emerging Europe dummy), and a dummy for resource-richness.

The structural reform indicator gap k in country i is then simply defined as the difference between the actual indicator in country i and its predicted value for the NMS or EU average from the estimated regression (A1):

gapik=Iikα^β^Xi.(A2)

To make comparisons of indicators between countries possible, each gap is weighted by the inverse of its standard deviation. Therefore, if a k-gap in country i is Z, the indicator k in country i is Z standard deviations from the average of the comparator.

B. Ranking Structural Reforms by Their Importance for Economic Growth

To rank structural reforms in terms of their importance for economic growth, we use the results of growth regressions. The methodology does not identify the causal effect of reforms on growth and does not allow for quantifying the magnitude of the difference between reforms. However, it provides an indicative guide for government reform priorities. Ranking the reforms helps governments align the biggest policy gaps with the most important policies.

For simplicity, suppose the true data generating process for the economic growth is:

gi=α+β0Xi+β1ΔIi1++βKΔIiK+εi,(A3)

where gi is GDP per capita growth in country i over a certain period and Xt is a set of structural macroeconomic controls at the beginning of the period: GDP per capita, stance of structural reform (Global Competitiveness Index score), geographical location (e.g., a dummy for the sub-Saharan region), common historical past (e.g., Emerging Europe dummy), resource-richness. Progress (Δ) in all structural indicators Ii1,,IiK is assumed to affect growth positively (or at least non-negatively).

Ideally, we would like to run regression (A3) and estimate β1,…,βK, which would give guidance analyzing the reform cost-benefit ratio. However, estimating (A3) is not possible, in particular because of the very large number of indicators, and potentially omitted variables. Instead, a separate growth regression for each indicator is estimated:

g˜i=βkΔI˜ik+εi,(A4)

where a tilde over g and I means that the variables were adjusted for constant and Xi.1 Each I˜k is also adjusted for the inverse of its standard deviation. Hence, the interpretation of each βk is the effect on growth in GDP per capita of a reduction of one standard deviation in the structural indicator gap. Another feature of adjusted I˜ks, which we will use later, is that they all have the same variance. Data availability limits the estimation period to 2006–13. Regression (A4) is estimated as a cross-section for 131 countries. It is estimated for all countries and subgroups, that is, countries with similar income—plus and minus one income category according to the World Bank classification. For example, for an upper-middle-income economy, we take all high-income, upper-middle-income and lower-middle-income countries.

Results suggest that reforms in all areas have a positive impact on growth. Reforms in institutions, financial markets, and infrastructure tend to have a somewhat higher growth impact than reforms in other areas. In addition, there is some variability in the estimated impact on growth among different income groups of countries (Annex Figure 1.4.1). While the estimates of βk may be biased because of omitted variables, the figures show under which assumption it is possible to use the estimates of βk to rank reforms according to their growth impact.

Annex Figure 1.4.1.
Annex Figure 1.4.1.

Results of Growth Regressions: Estimates of Growth Coefficients

Proposition 1. Let β^1 and β^m be the ordinary least square (OLS) estimates of (A4) for indicators l and m correspondingly, where βl and βm are the true parameter values from (A3). Then:

>βmplimnβ^1>plimnβ^m,(A5)

using the following assumption:

Assumption 1: Structural reform indicators that are more growth-enhancing (higher beta) are more correlated with other growth-enhancing indicators. Assumption 1 will be formulated in mathematical terms further in the proof of the proposition.

Proof of the Proposition 1

To simplify notation, and without loss of generality, let us take l=1, and m=2. From (A3) it follows that both β^1 and β^2 are biased and inconsistent, because regressions of type (A4) do not include other indicators, which are almost certainly correlated with both GDP growth and I˜1 or I˜2.

From the standard OLS algebra it follows that:

plimnβ^1=β1+β2γ12+β2γ13++βKγ1K,andplimnβ^2=β2+β1γ21+β3γ23++βKγ2K,(A6)

where γij is the probability limit of OLS coefficient in a regression of I˜j on I˜i. All of the terms in identities (A6) except for the first ones constitute the bias.

Now, since all indicators are adjusted and have the same variance:

γij=Cov(I˜i,I˜j)Var(I˜i)=Cov(I˜i,I˜j)Var(I˜j)=γij=corr(I˜i,I˜j)(A7)

Using (A7) we get:

plimnβ^1plimnβ^2=(β1β2)(1corr(I˜1,I˜2))+Σi=3K(corr(I˜1,I˜2)corr(I˜1,I˜2))(A8)

Now, in mathematical terms Assumption 1 states that:

β1>β2Σi=3Kβi(corr(I˜1,I˜2)corr(I˜1,I˜2))>0(A9)

that is indicator 1 is more correlated with the indicators, which correspond to higher betas, that is, those that are more important for growth.

Since correlation between any two random variables is bounded between -1 and 1, the two terms of the sum on the right hand side of (A8) are either both negative or both positive. Hence, we get (A5).

Under assumption 1, which seems to hold for most indicators in our dataset, it is possible to rank structural reforms by running simple regressions like (A4) and then comparing the OLS estimates. However, it is not possible to identify the true betas or the true difference between betas, because the last term in (A8) is not identified.

Classification of Reforms into Priority Classes

We classify reform areas into three classes of priorities: high, medium and low. The classification rule is based on two criteria—how large the reform gap is and how important the reform is for growth. Annex Table 1.4.1 guides our selection according to the criteria outlined below.

Annex Table 1.4.1.

Classication of Reform Priorities

(Exports)

article image
Source: IMF staff calculations.

On the vertical axis of the table are the reform gaps (measured in standard deviations from the mean of the peer group), sorted from large negative gaps (bad) to large positive gaps (good). On the horizontal axis the reforms are sorted by their importance for growth, that is, by the point estimate βk of corresponding growth regression.

Reforms that we classify as high priority range from reform areas with large negative gaps but minor/medium importance for growth to reform areas where the gap is small but improving the reform area matters a lot for growth. In the lower left corner are reform areas where the country performs better than its peers (a positive gap) and reforms in this area have a low importance for growth.

Reported scores correspond to reform priorities. We have assigned scores to the gaps (from 1 (large positive gap) to 5 (large negative gap) and to the importance of reforms for growth (from 1 (minor) to 4 (major)). The maximum would thus be 9 (a large negative gap (5) plus major importance for growth (4)). Annex Tables 1.4.2a and b report the sum of the two scores. Scores from 7-9 are classified as high priority, from 5-6 as medium priority, and 1-4 as low priority.

Annex Table 1.4.2a.

Reform Priorities in Western Balkan Countries: Compared to NMS Average, 2013

article image
Sources: World Economic Forum; and IMF staff calculations.Note: Analysis for Kosovo not included as the relevant data are not available.
Annex Table 1.4.2b.

Reform Priorities in Western Balkan Countries: Compared to EU Average, 2013

article image
Sources: World Economic Forum; and IMF staff calculations.Note: Analysis for Kosovo not included as the relevant data are not available.

To classify reforms by the importance for growth we use estimates from the income-specific growth regressions (see above for details). We also did the exercise with the estimates from the uniform regression. The difference to our main results is minor.

Annex Figure 1.4.2.
Annex Figure 1.4.2.

Reform Gaps: Institutions

(Standard deviations)

Sources: World Economic Forum; and IMF staff calculations.Note: Global Competitiveness Report, sub-indicators of pillar “Institutions”: 1 - Property rights, 2 -Intellectual property protection, 3 - Diversion of public funds, 4 - Public trust in politicians, 5 -Irregular payments and bribes, 6 - Judicial independence, 7 - Favoritism in gov’t decisions, 8 -Wastefullness of gov’t spending, 9 - Burden of gov’t regulation, 10 - Efficiency in setlling disputes, 11 - Efficiency in challenging regs, 12 - Transparency of policymaking, 13 - Business cost of terrorism, 14 - Business cost of crime, 15 - Organized crime, 16 - Reliability of police, 17 - Ethical behavior of firms, 18 - Strength of reporting standards, 19 - Efficacy of corp. boards, 20 - Protection of minority shareholders, 21 - Strength of investor protection.
Annex Figure 1.4.3.
Annex Figure 1.4.3.

Reform Gaps: Infrastructure

(Standard deviations)

Sources: World Economic Forum; and IMF staff calculations.Note: Global Competitiveness Report, sub-indicators of pillar “Infrastructure”: 1 - Quality of overall infrastructure, 2 - Quality of roads, 3 - Quality of railroad, 4 - Quality of ports, 5 -Quality of air transport infrastructure, 7 - Quality of electricity supply, 8 - Mobile telephone subscriptions, 9 - Fixed telephone lines. Excluded is sub-indicator 6 - “Availability of airline seats”, as it is related to the size of a country.
Annex Figure 1.4.4.
Annex Figure 1.4.4.

Reform Gaps: Goods Market Efficiency

(Standard deviations)

Sources: World Economic Forum; and IMF staff calculations.Note: Global Competitiveness Report, sub-indicators in pillar “Goods Markets Efficiency”: 1 - Intensity of local competition, 2 - Extent of market dominance, 3 - Effect. of anti-monopoly pol., 4 - Effect of taxation on incentives to invest, 5 - Total tax rate, 6 - # of proc. to start business, 7 - # of days to start business, 8 - Agricultural policy costs, 9 - Prevalence of trade barriers, 10 - Trade tarrifs, 11 - Prevalence of foreign ownership, 12 - Business impact of rules on FDI, 13 - Burden of customs procedures, 14 - Imports, %GDP, 15 - Degree of customer orientation, 16 - Buyer sophistication.
Annex Figure 1.4.5.
Annex Figure 1.4.5.

Reform Gaps: Labor Market Efficiency

(Standard deviations)

Sources: World Economic Forum; and IMF staff calculations.Note: Global Competitiveness Report, sub-indicators in pillar “Labor Market Efficiency”: 1 - Cooperation in labor employer relations, 2 - Flexibility of wage determination, 3 - Hiring and firing practices, 4 - Redundancy costs, 5 - Effect of taxation on incentives to work, 6 - Pay and productivity, 7 - Reliance on professional management, 8 - Country capacity to retain talent, 9 - Country capacity to attract talent, 10 - Women in labor force.

Annex 1.5. Labor Market Outcomes—Regression Analysis

Our empirical analysis links labor market outcomes at the individual level with a number of key macroeconomic and country-level structural and institutional indicators. Specifically, transitions between employment, unemployment, and non-participation in the labor force are linked by means of a micro-econometric multinomial logit model to various demographic characteristics of the labor force (age, disability, education, and marital status, as well as employment status from a year ago), macroeconomic factors (overall economic growth rate, investment level, credit growth, as well as indicators of fiscal stance, public expenditures, and remittances inflows), and structural factors (indicators of institutional rigidities in the labor market and those reflecting the country’s stage of transition to market economy). The micro-level data are derived from labor force surveys of four Western Balkan countries (Bosnia and Herzegovina, Kosovo, FYR Macedonia, and Serbia) as well as Bulgaria, Poland, and Romania for 2006–13, thus covering periods of the precrisis boom, the crisis bust, and the postcrisis recovery for a diverse group of countries in the region.

Annex Table 1.5.1.

Determinants of Labor Market Outcomes: Multinomial Logistic Regression Estimates

article image
Sources: National Labor Force Surveys; and IMF staff estimates.

This variable includes graduate education in cross-country regressions.

Excludes Bulgaria.

Annex 2.1. Estimating the Determinants of Expenditure Policy

This annex examines how responsive expenditure policy is to the business cycle, how much of spending is determined by inertia (for instance, on account of mandatory spending), and how much remains unexplained after accounting for these factors. This residual—an unexplained component—could for instance reflect the political business cycle rather than economic considerations. We focus here on the expenditure side because, with the exception of discretionary tax rate changes and lump-sum receipts, revenues generally reflect their cyclical tax bases (Coricelli and Fiorito 2013).

We extract this unexplained component of fiscal policy by estimating a fiscal policy rule, quantifying the unexpected variation in fiscal policy. In line with the work of Fatás and Mihov (2003, 2006), Afonso, Agnello and Furceri (2010), and Agnello, Furceri, and Sousa (2013), we estimate the following rule for each country i (i = 1,… N):

git=θi+λigit1+βiΔyit+γiΔyit1+ΓiXit+εit,(1)

where g is the logarithm of real government spending, y is the logarithm of real GDP and X is a set of controls including inflation and the logarithm of real public debt. We then examine how much of government spending is explained by persistence (captured using lagged spending), how responsive it is to the business cycle (captured using current and lagged GDP growth), inflation and the level of debt, and how much of it remains unexplained. Thus λi is a measure of persistence; βi and Υi gauge the responsiveness of fiscal policy to the business cycle; and εit is the unexpected variation in fiscal policy that could capture the impact of, for example, elections. We include country fixed effects to account for the impact of country-specific factors. We use a panel dataset including 34 countries (Western Balkans, EU).

Across all countries, government spending exhibits a high degree of persistence, with lagged expenditure explaining most of the variation in current expenditure. Current GDP growth has a negative impact on spending, possibly capturing effects through lower revenues requiring corresponding spending cuts. Debt has a significant constraining impact on spending only in the Advanced EU economies and the Baltics, but not in the Western Balkans and Central Europe. The Western Balkans appear to be less responsive to the business cycle than the New Member States or Advanced EU economies. In fact, the unexpected variation appears to be somewhat larger for the Western Balkans, as less of the variation in spending over time for a given country is explained by cyclical factors and inertia.

Annex Table 2.1.1.

Fiscal Policy Responsiveness, Persistence and Discretion

article image
Note: The dependent variable is log real expenditure. Standard errors in parentheses, * denotes significant at the 5 percent level, ** at the 1 percent level, *** at the 0.1 percent level.

Annex 2.2. Estimating the Determinants of the Recession and the Recovery

We estimate a cross-sectional regression of the size of the peak-to-trough decline in real GDP on explanatory factors designed to measure the degree of overheating and imbalances in the boom period (Annex Table 2.2.1).1 We find that having a fixed exchange rate (captured by using a dummy variable) is associated with a larger decline in growth. Wider current account deficits in the boom years are also associated with bigger recessions. Higher precrisis capital inflows, however, appear to moderate the drop, possibly reflecting the beneficial effects of predominately foreign direct investment inflows during the early boom years. The regression does not yield significant coefficients on the variables for fiscal policies, wage growth, and credit.2 Western Balkan economies experienced a smaller fall than New Member States (as captured using a dummy variable), even with other conditioning factors.

Annex Table 2.2.1.

Determinants of the Size of the Recession

article image
Source: World Economic Outlook, national authorities and staff calculations.Note: Sample includes Western Balkans and New Member States. Constant not reported, robust standard errors. *** denotes significant at 0.1% level, ** at 1%, * at 5%. Export boost is defined as the increase in exports as a share of GDP, comparing the 2009-2013 average with the 2006-2008 average (positive if exports as a share of GDP are higher after the crisis). Import compression is defined as the fall in imports as a share of GDP, again comparing the 2009-2013 average with the 2006-2008 average (positive if imports as a share of GDP are lower after the crisis).
Annex Table 2.2.2.

Determinants of the Recovery

article image
Source: World Economic Outlook, national authorities and staff calculations.Note: Sample includes Western Balkans and New Member States. Constant not reported, robust standard errors. *** denotes significant at 0.1% level, ** at 1%, * at 5%. Export boost is defined as the increase in exports as a share of GDP, comparing the 2009-2013 average with the 2006-2008 average (positive if exports as a share of GDP are higher after the crisis). Import compression is defined as the fall in imports as a share of GDP, again comparing the 2009-2013 average with the 2006-2008 average (positive if imports as a share of GDP are lower after the crisis).

We also examine the determinants of postcrisis recovery using a cross-sectional regression of real GDP growth since the trough (Annex Table 2.2.2).3 A fixed exchange rate is found to support the recovery, as does stronger export performance. Surprisingly, stronger deleveraging also appears to be associated with stronger growth. Discretionary fiscal policy (as constructed in Annex 2.1) again does not have a significant effect, and neither do capital inflows.

Annex 2.3. Fiscal Policies in the Western Balkans

Looking forward, the Western Balkan countries face important structural challenges as they strive to adjust to a postboom environment. Annex Table 2.3.1 summarizes some of the challenges and presents policy recommendations. Efforts aimed at containing deficits and debt levels are also needed in light of aging populations, which over time will add to expenditure pressures. And while in some countries substantial adjustment has already taken place, additional consolidation would be needed to achieve further deficit reduction. Crucially, fiscal consolidation should be complemented by compositional changes, reducing in particular the share of current expenditures. Controls in the broader public sector should also be improved as off-budget operations and the legacy of social and subsidy spending continue to complicate budget planning in several Western Balkans countries. In most countries revenue measures should be seen as a complement to adjustment on the expenditure side. Structural fiscal reforms should focus on broadening the tax base and fighting the grey economy.

Annex Table 2.3.1.

Challenges and Policy Recommendations for the Western Balkan Countries

article image

High persistence reflects a common practice of incremental budgeting with a one-year time horizon. Longer budgeting horizons and fiscal rules can help contain spending pressures during good times and a medium-term strategy would also facilitate the planning of large investment projects. Empirical studies suggest that fiscal rules have been generally associated with improved fiscal performance (IMF 2009), though of course they are only successful if there is sufficient political commitment to adhere to them. While some countries in the region have recently adopted fiscal rules (Croatia, Kosovo, Montenegro, and Serbia), enforcement is weak in some of them. It should be acknowledged that running large surpluses during boom times may be politically difficult, in particular in catch-up economies, with large demands for improvements in infrastructure (particularly Albania, Kosovo, FYR Macedonia, and Montenegro).

Fixed exchange rates and a high dependence on external financing make fiscal consolidation even more crucial. Fiscal consolidation and fiscal buffers are particularly important in the context of unilaterally euroized economies (Montenegro, Kosovo), currency boards (Bosnia and Herzegovina) and exchange rate pegs (Croatia, FYR Macedonia). The prospect of tighter global financing conditions ahead could increase vulnerabilities in countries with a high reliance on external financing.

Annex 3.1. Benchmarking Financial Development and Explaining Gaps

When comparing levels of financial development across countries, it is useful to take into account the level of economic development and structural characteristics, regardless of the policies or institutions of a country (Beck and others, 2008). In particular, structural characteristics such as income per capita, population (or market) size, population density and age profile, and whether the country is a transition economy, fuel exporter or offshore financial center, have generally been found to be associated with indicators of financial development.1

Controlling for these structural, or policy-invariant, variables in a regression provides us with a country’s structural “benchmark,” that is, an expected average (if using least-squares regressions) or median (or other percentile, if using quantile regressions). The World Bank’s Finstat database provides estimates of these benchmarked indicators for 183 countries. The pooled regression assumes a common path of development (as financial systems fulfill similar functions and face similar frictions) and includes time dummies to control for the effect of global conditions on all countries.

The empirical estimates for bank deposit and private credit depth show a positive relationship between income per capita (though it levels off at high income levels) as well as size of market. In contrast, age dependency (i.e., in particular a relatively greater share of young people in population), and being a transition economy tends to be associated with lower depth. Population density is associated with higher deposit depth but lower credit depth.

Annex Table 3.1.1.

Median Regression Results for Bank Depth Indicators

article image
Source: Feyen and Kibuuka, FinStat 2013.

It is important to note that that the structural depth is not the “frontier,”2 as it does not take into account institutional and other long-term policy variables that can affect depth either positively or negatively. Rather, the gap between actual and benchmark levels of financial depth can be compared to institutional and policy factors to see if these explain either an overperformance or underperformance gap.3 Thus, it may be possible to assess why a county has an overperforming gap even though it has a lower absolute level of an indicator than another country—for example compare Country A, at point A, with Country B, at point B.

The position of B (underperformance) may be because of policy (macroeconomic) instability or institutional weaknesses (for example, the result of weak information or protection for creditors). In contrast, the position of A may indicate a relatively stronger macroeconomic or market structure environment. However, if there is a very large gap that cannot be well explained by policies, then this gap could indicate an “excess” or “boom” that may eventually be followed by a bust (for example point C in Annex Figure 3.1.1 above).

Annex Figure 3.1.1.
Annex Figure 3.1.1.

Stylized Financial Possibility Frontier

Source: Barajas and others (2013).

Recent work has found that gaps, or changes in them, can be affected by macro-financial variables (such as inflation rates, remittances, and growth) as well as by the enabling or institutional environment, including market structure (such as foreign bank entry and competition) in addition to regulatory factors (such as strength of banking regulation and supervision) and institutional factors, particularly creditor rights and enforcement costs.

Annex Table 3.1.2.

Impact of Macro and Enabling Environment on Private Credit Depth and Depth Gap

article image
Source: Compiled by staff from papers cited above.Note: *, **, *** shows whether variables were significant at the 10, 5 or 1 percent level.

Estimate of depth.

Looks at factors affecting gap between actual and benchmark (as estimated in Finstat). Thus a positive gap signifies overperforming gap.

Gap, measured as benchmark minus actual; thus a negative gap signifies overperforming. Specification 1 reported unless otherwise noted.

For Barajas and others (2013), the inverse of inflation.

For De la Torre, Feyen, and Ize (2011), average annual growth over the sample period. For Barajas and others (2013), lagged GDP growth over the previous five years.

For De la Torre, Feyen, and Ize (2011), fraction of sample years in which a country experienced an annual decline in domestic private credit to GDP of 20 percent or more; For Barajas and others (2013), banking crisis dummy (1 if had a crisis in the last decade).

Index ranges from 0 (hard peg) to freely floating (8)

Specification 6 reported, which is similar to specification 8. Other specifications exclude it or it does not appear as significant.

Domestic financial system and capital account liberalization.

For Cottarelli and others (2003), index of stringency of specific legal requirements for obtaining a license to operate a bank. For Barajas and others (2013), foreign entry restrictions index.

Other specifications for market structure/competition variables show that foreign entry restrictions decrease the gap, while government ownership increases the gap, and the asset concentration and foreign bank share are not significant.

Other specifications for regulatory variables show privatization and supervision decrease the gap, while other variables (e.g., geographic diversity requirements and credit controls) are not significant.

Other specifications for institutional variables find credit rights decreases the gap, while financial, political and economic risk indexes are not signification.

Index is principal component of Doing Business Indicators on contract enforcement costs, number of days to enforce a contract (in logs), and number of procedures to enforce a contract.

Specification 6.