Blanchard, Olivier, and Francesco Giavazzi, 2002, “Current Account Deficits in the Euro Area: The End of the Feldstein-Horioka Puzzle?” Brooking Papers on Economic Activity: 2, pp. 147– 86.
Chamon, Marcos, Paolo Manasse, and Alessandro Prati, 2007, “Can We Predict the Next Capital Account Crisis?” IMF Staff Papers, Vol. 54, No.2, pp. 270– 304.
Egert, Balazs, 2005, “Balassa-Samuelson Meets South Eastern Europe, the CIS and Turkey: A Close Encounter of the Third Kind?” The European Journal of Comparative Economics, Vol. 2, no. 2., pp. 221– 243.
Funda, Josip, Gorana Lukinic, and Igor Ljubaj, 2007, “Assessment of the Balassa-Samuelson Effect in Croatia,” Financial Theory and Practice, Vol. 31, no. 4, pp. 321– 351.
Hilaire, Alvin and Anna Ilyina, 2007, “External Debt and Balance-Sheet Vulnerabilities in Croatia,” IMF Country Report No. 07/82 (Washington: International Monetary Fund).
International Monetary Fund, 2008, “Republic of Croatia: Financial System Stability Assessment Update,” IMF-World Bank, April 2008 (Washington).
International Monetary Fund, 2008, “Regional Economic Outlook, Europe: Reassessing Risks,” World Economic and Financial Surveys, April 2008 (Washington).
Isard, Peter, 2007, “Equilibrium Exchange Rates: Assessment Methodologies,” IMF Working Paper No. 07/296 (Washington: International Monetary Fund).
Lane, Philip R. and Gian Maria Milesi-Ferretti, 2006, “Capital Flows to Central and Eastern Europe,” IMF Working Paper No. 06/188 (Washington: International Monetary Fund).
Mihaljek, Dubravko and Marc Klau, 2003, “The Balassa-Samuelson Effect in Central Europe: A Disaggregated Analysis,” BIS Working Paper No.143.
Moore, David, and Athanasios Vamvakidis, 2008, “Economic Growth in Croatia: Potential and Constraints,” Financial Theory and Practice, forthcoming.
Appendix I.I: Macroeconomic Balance Approach
Appendix I.II: Stabilizing Net IIP/GDP and External Debt/GDP
Using a simple accounting framework, the net IIP of a country can be decomposed as follows:
where bt is the net IIP of a country (expressed in percent of GDP); at and lt denote gross asset and liability positions, respectively (expressed in percent of GDP); nicat is the non-income CA balance (in percent of GDP); nt is the growth rate of nominal GDP;
Focusing on the stock of external debt, (1) can alternatively be rewritten as follows:
Appendix I.III: Application of the Binary Classification Tree for Predicting the Capital Flow Reversals
Prepared by Anna Ilyina.
See, e.g., “Emerging Europe’s Current Account Deficits: Mind the Gap!” by FitchRatings, January 31, 2008.
In particular, the 2007 Decision on Bilateral Surveillance clarifies that the objective of the IMF’s surveillance is to foster stability of the international monetary system by encouraging national policies that do not disrupt or compromise the members’ own “external stability.”
Decision on Bilateral Surveillance, http://www.imf.org/external/np/pp/2007/eng/062107.htm
ULCM stands for “unit labor cost in manufacturing.”
The results seem to be sensitive to the choice of the time period, data frequency and the definitions of the tradable and non-tradable sectors. See Mihaljek and Klau (2003), Egert (2005), and Funda et al. (2007) for more details. For example, Funda et. Al. (2007) found no statistically significant Balassa-Samuelson effect, but using a simple accounting framework, they assess the contribution of the Balassa-Samuelson effect to annual inflation over a period of 1999–2006 to be a maximum of 0.64 percentage points.
For example, Croatia’s export share in world imports of chemicals remained stable since 2001, but the composition has been changing, with the share of “chemical elements and compounds” and “plastic materials” declining and the share of “pharmaceutical products” and “perfume materials, etc.” increasing (based on COMTRADE data), indicating possible shifts towards higher value-added goods.
Both the macroeconomic balance and external sustainability approaches are used by the IMF’s internal Consultative Group on Exchange Rate Issues (CGER), which provides multilaterally consistent exchange rate assessments for a number of advanced and emerging market countries. While Croatia is included in the sample used in the panel regression estimation of the current account norms, it is not on the list of countries for which the CGER group provides regular assessments of real exchange rate misalignment. The results reported in this section, however, are largely based on the CGER methodology (Methodology for CGER Exchange Rate Assessments (2006)).
See Isard (2007) for an overview and discussion of the equilibrium exchange rate assessment methodologies.
NFA is the difference between a country’s total foreign assets and total foreign liabilities.
See Regional Economic Outlook (2008), Box 9, which presents the CEE CA norms estimated using a variant of the macroeconomic balance approach.
The ERER estimation was performed using two sets of productivity data from Funda et al. (2007), one with “hotels and restaurants” included in the tradable sector (in view of the importance of tourism revenues for Croatia’s current account position) and the other with “hotels and restaurants” included in the non-tradable sector. The results are not significantly different. For more details on the ERER approach, see Methodology for CGER Exchange Rate Assessments (2006).
See, Regional Economic Outlook (2008), Box 9, Chapter 3. This approach involves computing the difference between the actual current account balance and the predicted current account balance based on the estimated regression of the current account deficit on the level of income per capita relative to the peer group average, and a number of other control variables (see Blanchard and Giavazzi (2002) for details).
All numbers are calculated from underlying data in euro terms.
Note that the concept of “net debt” used in this Chapter is different from that used in the staff report. All ratios are calculated from underlying data in euro terms.
After having increased by 60 percent annually in 2006-07, stock prices fell by 30 percent during the first three months of 2008 (see Financial System Stability Assessment Update (2008) for more details on the stock market developments).
This concept is also referred to as “trade balance inclusive of services and transfers” (IMF (2006), page 19).
Notwithstanding the improvements in methodology, possible measurement errors in both the numerator and the denominator of the external debt-to-GDP ratio are yet another reason why focusing on a particular “threshold” level may not be very practical.
See “Republic of Croatia: Financial System Stability Assessment Update,” 2008.
This is because of a high level of financial euroization and significant balance-sheet exposures of the non-financial sector to exchange rate risk. See Hilaire and Ilyina (2006) for detailed discussion of Croatia’s sectoral balance-sheet vulnerabilities.
The average interest rate paid by Croatian banks on their external liabilities may, to some extent, reflect the parent-subsidiary relationship between Croatian banks and their foreign owners (e.g., the “quasi equity” nature of certain liabilities to parent banks).
The external liquidity ratio is defined as liquid external assets (net official reserves plus banks’ gross external assets) divided by liquid external liabilities (short-term external debt on the remaining maturity basis). The reserve cover is the ratio of official reserve to the sum of current account deficit and short-term external debt by remaining maturity. The external debt service ratio is the ratio of debt service to current external receipts.
For example, an adjusted external liquidity ratio (which adds foreign currency deposits in domestic banks to external liabilities) is sometimes used to gauge the adequate level of external liquidity in the context of high financial dollarization/euroization (see, e.g., FitchRatings (March 2007)). This is because the adjusted external liquidity indicator also takes into account the amount of foreign exchange that banks would need to raise in an extreme event of withdrawal of all foreign currency deposits from the banking system. In the case of Croatia, this indicator stood at around 58 percent at end-2007, reflecting the historically high level of financial euroization as well as the size of Croatia’s banking system. However, this indicator should be interpreted with caution: in particular, if one were to draw policy implications taking into account the level of euroization, it would be essential to make realistic assumption with regard to the share of foreign currency deposits that might be withdrawn in an extreme scenario, taking into account historical experience.
Afonso, A., L. Schuknecht, and V. Tanzi, 2006, “Public Sector Efficiency: Evidence for New EU Member States and Emerging Markets,” European Central Bank Working Paper Series no. 581, (Frankfurt: European Central Bank).
Afonso, A. and M. St. Aubyn, 2004, “Non-Parametric Approaches to Education and Health: Expenditure Efficiency in OECD Countries,” mimeo, (Lisbon: Technical University of Lisbon).
Charnes, A., W. Cooper, and E. Rhodes, 1978, “Measuring Efficiency of Decision-Making Units,” European Journal of Operational Research, Vol. 3, pp. 429–44.
Coelli, Tim, Mathieu Lefebvre, and Pierre Pestieau (2007), “Measurement of Social Protection Performance in the European Union,” mimeo.
Cucilić, Judita, Michael Faulend, and Vedran Šošic (2004), “Fiscal Aspects of Accession: Can We Enter the European Union With a Budgetary Deficit?” in Croatian Accession to the European Union ed. by Katarina Ott, Chapter 3 in Vol. 2, pp. 49–77.
Davies, M., M. Verhoeven and V. Gunnarsson, 2006, “Wage Bill Inflexibility and Performance Budgeting in Low-Income Countries,” (unpublished; Washington: International Monetary Fund).
Ederer, Peer, Philipp Schuler, and Stephan Willms, 2007, “The European Human Capital Index: The Challenge of Central and Eastern Europe.” Lisbon Council Policy Brief. Brussels: The Lisbon Council for Economic Competitiveness and Social Renewal.
Farrell, M., 1957, “The Measurement of Productive Efficiency,” Journal of the Royal Statistical Society, Series A, Vol. 120, No.3. pp. 253–90.
Gupta, S., and M. Verhoeven, 2001, “The Efficiency of Government Expenditure: Experiences from Africa” Journal of Policy Modeling, no. 23, pp. 433–67.
Herrera, S., and G. Pang, 2005, “Efficiency of Public Spending in Developing Countries: an Efficiency Frontier Approach,” World Bank Policy Research Working Paper 3645, (Washington: World Bank).
Mattina, Todd, and Victoria Gunnarsson, 2007, “Budget Rigidity and Expenditure Efficiency in Slovenia,” IMF Working Paper 07/131, (Washington: International Monetary Fund).
Mihaljek, Dubravko (2007), “Health Care Policy and Reform in Croatia: How To See the Forest for the Trees,” in Croatian Accession to the European Union ed. by Katarina Ott, Chapter 11 in Vol. 4, pp. 277–320.
Mossialos, Elias, Anna Dixon, Josep Figueras, and Joe Kutzin (2002), Funding Health Care: Options for Europe, European Observatory on Health care Systems Series, Open University Press.
Simar, L., and P. Wilson, 2007, “Estimation and Inference in Two-stage, Semi-parametric Models of Production Processes,” Journal of Econometrics, 136, pp. 31–64.
Verhoeven, Marijn, Victoria Gunnarsson, and Sergio Lugaresi, 2007 “The Health Sector in the Slovak Republic: Efficiency and Reform,” IMF Working Paper 07/226, (Washington: International Monetary Fund).
World Bank, 2007, “Croatia: Restructuring Public Finance to Sustain Growth and Improve Public Services – A Public Finance Review,” World Bank Report No. 37321-HR.
Appendix. Data Envelopment Analysis (DEA)23
The DEA technique is a non-parametric method of estimating production possibility sets, which can be used to evaluate the efficiency in the use of inputs in producing outcomes for a sample of production units.24 It is mostly used for estimating relative efficiency in business applications, but it has recently also been used to assess the relative efficiency of public expenditure. In the context of government expenditure efficiency, indicators of public production are typically used to measure outcomes, for example, life expectancy and infant mortality rates (in health care), youth literacy rates and test scores (in education), and the number of roads and telephone lines (in infrastructure). Inputs used to produce these outcomes are public and private expenditure on health, education, and infrastructure, as well as intermediate outputs and resources such as the number of doctors and hospital beds (in health care) and enrollment rates and student-teacher ratio (in education). The production units in this case are often countries, but could also be sub-national regions.25
Figure II.A1 illustrates a stylized example of DEA based on a single input and outcome indicator across countries. The efficient frontier connects countries A to D as these units dominate countries E and G in the interior. The convexity assumption allows an inefficient country (point E) to be assessed relative to a hypothetical position on the frontier (point Z) by taking a linear combination of efficient unit pairs (points A and B). In this manner, an input-based technical efficiency score that is bounded between zero and one can be calculated as the ratio of YZ to YE. The score corresponds to the proportional reduction in inputs that is consistent with relatively efficient production of a given output, and can be interpreted as an indicator of the cost savings that could be achieved from efficiency enhancement. Similarly, an output-based technical efficiency score can be calculated as the ratio of FX to EX, which reflects the improvement in outputs for given inputs that could be achieved from efficiency enhancement. This paper focuses on output-based efficiency scores, since Croatia will need to improve outcomes without increasing expenditures.26 27
DEA is a powerful tool to assess the relative efficiency of spending, but also has important caveats. For example, it does not require an assumption about unknown functional forms for the efficiency frontier or complex distributional properties for econometric analysis. However, it is also subject to the following caveats:
Results are highly sensitive to sample selection and measurement error. As a result, outliers exert large effects on the efficiency scores and the shape of the frontier. For this reason, proper sample selection is the key to ensuring that cross-country input-output combinations are comparable.
Spending attributes that are difficult to quantify are not easily incorporated in the analysis, such as the quality of spending.
The outcome indicators against which inputs are evaluated may not actually be targeted by policy makers.
Large differences across countries in private health care or education spending could bias the efficiency scores of public spending, as the outcomes targeted by policy makers are also impacted by private spending.
Factors beyond the direct control of policy makers can also affect relative efficiency scores. For instance, a high incidence of AIDS would reduce the measured efficiency of health spending in African compared to other countries.
Moreover, simple DEA estimation produces biased estimates of the efficiency scores that need to be corrected. In particular, the best-practice frontier can move outward, if efficient pairs/countries are added in the sample, but cannot move inward. This one-sided error means that estimating the best-practice frontier with a finite sample is subject to bias. Since output–oriented efficiency scores are measured in relation to the frontier, the estimated scores are subject to the same finite sample downward bias (i.e., the level of efficiency is overestimated unless a correction is made for the bias). This bias stems from the fact that since we only observe a sub-sample of the possible outcomes representing all feasible combinations of spending and outcomes, we do not know the exact position of the best-practice frontier. Where appropriate, corrections are made for the estimation bias in the best-practice frontier and efficiency scores through bootstrapping, as suggested by Simar and Wilson (2000).28
DEA results can be disaggregated to assess at what stage of the spending process inefficiencies arise. This is done as by comparing spending efficiency (the overall measure of efficiency from spending to outcomes as discussed above) and system efficiency (the measure of efficiency from intermediate outputs to outcomes; Tables II.5 and II.9). Figure II.A2 illustrates how it is done in the analysis of efficiency of health care spending. First, cost efficiency is assessed using health care spending and intermediate output indicators such as hospital beds, immunizations, physicians, health care workers and pharmacists per capita. Second, efficiency scores are calculated, using the intermediate output index as an input and associated outcomes (infant, child, and maternal mortality rates, as well as HALE, standardized death rates and the incidence of tuberculosis). Third, the resulting system efficiency rankings are averaged, and expressed as a ratio of the average OECD ranking, and compared with similar ratios for spending efficiency.
Prepared by Etibar Jafarov (EUR) and Victoria Gunnarsson (FAD).
This pressure is related to the use of EU structural funds, contributions to the EU budget, and an upgrading of environmental standards. Funck (2003) suggested that implementing National Programs for the Adoption of the Acquis of the new member states was going to have entailed additional annual spending of (on average) about 3½ percent of GDP for these countries. Cucilić, Faulend, and Šošic (2004) estimated net fiscal costs (netting out transfers from the EU) of Croatia’s EU accession for 2007, the year the authors had expected accession to take place at the time of writing, at 1.1 percent of GDP.
The projection does not include spending related to the use of EU structural funds.
EU-10 countries are new EU members and comprise the Czech Republic, Estonia, Latvia, Hungary, Lithuania, Poland, Slovakia, Slovenia, Bulgaria, and Romania. EU-15 countries comprise Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden, and the United Kingdom.
Old-age pensions will not be a subject of this study, since this component of social spending does not lend itself to analysis of efficiency in the same way as the other components that are analyzed.
Results for the EU-10 are heavily influenced by the results for Bulgaria and Romania, which have significantly worse results than the other new EU members. But Croatia’s performance is still slightly better than the averages for the other EU-10 countries.
Twenty groups of people, including pensioners, unemployed, and students, are exempt from paying contributions. Only around 35 percent of the population pays contribution.
See Funding Health Care by Mossialos et al. (2002) for a description of cost sharing in Europe. Several countries, including Australia, Canada, and Switzerland, do not allow supplementary insurance to cover co-payments associated with services paid for by the health insurance fund.
These lists were introduced in 2006. For drugs on the B list, the HZZO pays a reference price for drugs on the A list and consumers pay the difference between the sale and reference prices. As a result of strong bargaining, pharmaceutical spending was reduced by about 2 percent in 2007, despite a 6 percent increase in consumption of drugs.
Over a third of total health care spending in Croatia finances hospital (in-patient) care.
Croatia ranked 26th on the PISA science scale, ahead of some EU countries (e.g., Italy and Spain).
The sequencing of possible reforms and related political economy issues are beyond the scope of this paper.
This analysis does not provide estimates of causality. It is possible that causality goes the other way around or both ways. The small sample size precludes regression analysis in the second-stage.
Given the close relationship of spending and outcomes with income levels, correlations of efficiency scores and associated factors are conditional on GDP. GDP per capita is adversely related to efficiency since many of the factors that are associated with efficiency are also closely related to income level. In order to avoid attribution of factors whose effects on the variation in efficiency cannot be separated from the effect of GDP, only GDP per capita and factors that are correlated with efficiency independently of GDP per capita are considered in the second-stage analysis of this chapter. The association with efficiency of factors that are strongly correlated with GDP is assessed by regressing the efficiency score on both GDP and the associated factor.
The Croatian government adopted the National Health Care Development Strategy 2006–11 to enhance and secure better-quality health care for citizens. The strategy includes both system reforms and financing reforms.
The replacement rate is the ratio of benefits to (previously received) income.
About 6 percent of the labor force was on sick leave in 2005; anecdotal evidence suggests that sick leave is used to deal with excess employment at the business level.
Moreover, restricting the basic benefit package would stimulate private participation in the provision of additional insurance.
The share of obese people in Croatia is almost double the average of the EU-15. Mihaljek (2007) mentions an unhealthy lifestyle (high alcohol and tobacco consumption, and prevalence of physical inactivity) as the likely reason for the difference in mortality rates for non-communicable diseases between Croatia and EU-15 countries.
Efficiency in secondary education is estimated using both a combined set of secondary intermediary outputs and outcomes, and PISA scores only.
System efficiency was estimated only for the secondary education level, where PISA test scores were used as education outcome. The overall public sector efficiency (quartile) rankings in the primary and secondary levels presented in Table II.7 are for the first stage of the production process (spending to intermediary outputs), since no education outcomes such as test scores are available at these levels.
See Annex I for description of caveats of DEA.
There is well-established literature using DEA to assess the relative efficiency of public expenditure. Gupta and Verhoeven (2001) studied the relative efficiency of education spending in a broad sample of African countries during the 1984–95 period. Afonso and St. Aubyn (2004) applied DEA and a related frontier-based approach on health and education spending in a sample of OECD countries. Herrera and Pang (2005) studied the relative efficiency of spending in 140 countries using DEA. Afonso, Schuknecht and Tanzi (2006) applied DEA in a sample of EU and emerging market countries. An important contribution of their work was to apply truncated regression models based on procedures developed by Simar and Wilson (2007) to control for exogenous factors that impact efficiency but that are not directly controlled by policy makers. Coelli, Lefebvre, and Pestieau (2007) applied DEA to study social protection performance in the EU.
An output-based efficiency score of one corresponds to a relatively efficient country operating on the frontier. Scores exceeding one imply that spending could achieve better output performance. This differs from input-based efficiency scores that range between zero and one.
The input- and output-based efficiency scores are equal assuming constant returns to scale. However, the DEA models considered in this chapter permit variable returns to scale.
A key issue is how quickly the estimated efficiency scores converge to their unbiased true values if the sample of observations is expanded. This convergence speed is n2/(p+q+1), where p is the number of inputs and q is the number of production items. In the 1 input / 1 product examples of this Appendix, the convergence speed is n-2/3. This is faster than the convergence speed for a standard parametric regression of n-1/2, suggesting that reasonable estimates of efficiency scores and confidence intervals can be reached with a lower number of observations than would be needed for standard regression analysis. However, the convergence speed declines exponentially as the number of inputs and production items is increased, and already at two inputs and production items, the speed of convergence is markedly slower than for a parametric regression. This implies that an expansion in the numbers of inputs and production items comes at a significant cost in terms of the ability to draw conclusions on efficiency from a limited number of observations.