Report No. 62587-AM Republic of Armenia Fiscal Consolidation and Recovery (In Two Volumes) Volume I: Synthesis Report November 1, 2011 Poverty Reduction and Economic Management Unit Europe and Central Asia Region Document of the World Bank CURRENCY EQUIVALENTS (Exchange Rate as of May 30, 2011) Currency Unit Armenian Dram US$1.00 375.3 AMD Weights and Measures: Metric System ABBREVIATIONS AND ACRONYMS ADB Asian Development Bank MLSP Ministry of Labor and Social Protection AMD Armenian Dram MTC Ministry of Transport and Communication CIS Commonwealth Independent State MTEF Medium Term Expenditure Framework CIT Corporate Income Tax NaCET National Center for Education Technology DPO Development Policy Operation NILSR National Institute for Labor and Social DSA Debt Sustainability Analysis Research DTI Direct Trade Input NRUF Natural Resources User Fee ECA Europe and Central Asia NSS National Statistical Service ECF Extended Credit Facility OECD Organization for Economic Development EFF Extended Fund Facility OOP Out of pocket EU European Union PCF Per Capita Financing FAD Fiscal Affairs Department PER Public Expenditure Review FBP Family Benefit Program PIT Personal Income Tax FRF Fiscal Rule Framework PPP Public Private Partnership GDP Gross Domestic Product PRSP Poverty Reduction Strategy Paper GFS Government Finance Statistics RAMS Road Asset Management System GIS Geographical Information System SHA State Health Agency HDM-4 Highway Development and Management SME Small and Medium Entreprise Model SRC State Revenue Committee ILCS Integrated Living Conditions Survey STI Sexually Transmitted Infection ILO International Labor Organization TA Technical Assistance IMF International Monetary Fund TIMSS Trends in International Mathematics and LIBOR London Interbank Offered Rate Science Study LRDP Lifeline Road Development Program USD US dollar LTU Large Taxpayer Unit VAT Value Added Tax MoF Ministry of Finance WTO World Trade Organization MoH Ministry of Health WHO World Health Organization Vice President: Philippe Le Houérou Country Director: Asad Alam Sector Director: Yvonne Tsikata Sector Manager: Kazi Mahbub-al Matin Task Team Leader: Souleymane Coulibaly Contents Acknowledgement ....................................................................................................................................... vi Executive Summary .................................................................................................................................... vii Chapter 1. Macroeconomic Context ....................................................................................................... 1 Chapter 2. Debt and Fiscal Sustainability............................................................................................... 7 A. Is Armenia Debt Dynamics Sustainable? ..................................................................... 8 B. Is Armenia Fiscal Stance Sustainable? ....................................................................... 10 Chapter 3. Revenue Potential ............................................................................................................... 13 A. Revenue Potential from Tax Administration .............................................................. 14 B. Revenue Potential from Tax Policy............................................................................ 23 Chapter 4. Expenditure Rationalization and Expenditure Efficiency Gains......................................... 31 A. Expenditure Trends, 2006-2009 ................................................................................. 34 B. Public Sector Wage Bill ............................................................................................. 38 C. Health Expenditures ................................................................................................... 47 D. Education Expenditures.............................................................................................. 56 E. Road Transport Expenditures ..................................................................................... 65 References ................................................................................................................................................... 75 iii List of Tables Table 1.1: Expenditure Ratios in Armenia and Selected Economies ......................................................... 3 Table 1.2: Armenia Fiscal Stance after the 2009 Crisis ............................................................................. 5 Table 2.1: Fiscal Frameworks Assumed under DSA and MTEF............................................................. 10 Table 2.2: Primary Balances to Keep Debt at 42 percent ........................................................................ 11 Table 2.3: Primary Balances to Reduce Debt to 30% in 10 Years........................................................... 11 Table 3.1: Tax rates and tax productivity, 2009 ...................................................................................... 13 Table 3.2: Estimated Gains from Improving Tax Policy and Administration (% of GDP) ..................... 14 Table 3.3: Estimated Additional Tax Potential from Large Taxpayers ................................................... 19 Table 3.4: Estimated Additional Tax Potential from the Non-Agricultural Informal Sector................... 22 Table 3.5: Excise Revenues as a Percentage of GDP, 2009 .................................................................... 25 Table 4.1: Armenia Consolidated Expenditures by Functional Classification ........................................ 32 Table 4.2: Expenditure on General Public Services................................................................................. 34 Table 4.3: Expenditure on Public Order, Security and Judicial Work ..................................................... 35 Table 4.4: Healthcare Expenditure over 2006-09 .................................................................................... 35 Table 4.5: Education Expenditure over 2006-09 ..................................................................................... 36 Table 4.6: Armenia: Number of Recipients of Key Social Protection Programs 2005-2007 .................. 36 Table 4.7: Armenia consolidated expenditure in economic classification ............................................... 37 Table 4.8: Capital Expenditure in Armenia, 2006-2009 .......................................................................... 38 Table 4.9: Overview of Civilian Government Employment in Armenia (2006) ..................................... 40 Table 4.10: Overview of Remuneration Systems for Armenia‘s Public Services ..................................... 42 Table 4.11: Total Health Spending in Armenia, 2010 Estimates ............................................................... 49 Table 4.12: Coverage of Key Health Interventions in Armenia is Low .................................................... 50 Table 4.13: Key Statistics for State General Secondary Schools, 2009/10................................................ 58 Table 4.14: Summary of the Determinants of Academic Performance ..................................................... 62 Table 4.15: Road Network Densities in ECA ............................................................................................ 65 Table 4.16: Main Roads Current Condition and Asset Value .................................................................... 67 Table 4.17: Investment Program on the Road Network from 2005-2011 (US$ million) .......................... 68 Table 4.18: Economic Comparison the Three Scenarios ........................................................................... 69 List of Figures Figure 1.1: Real GDP Growth, Per Capita GDP, and Poverty ................................................................... 1 Figure 1.2: Armenia Public Debt, 2005-2009 ............................................................................................ 1 Figure 1.3: Public Spending and Per Capita Income in Selected Countries, 2009 .................................... 2 Figure 1.4: Tax Revenue and Per Capita Income in Selected Countries, 2008 ......................................... 3 Figure 1.5: Fiscal Stance of the Government ............................................................................................. 4 Figure 1.6: Estimated Potential GDP ......................................................................................................... 4 Figure 2.1: Public Debt Burden and Liquidity Indicators under Baseline and Alternative Scenarios ....... 9 Figure 3.1: Audit Outcome, 2005-2009 ................................................................................................... 17 Figure 3.2: Share of Armenia Informal Sector to Total Gross Value Added, by Sectors ........................ 22 Figure 3.3: Economic Contribution of Mining and Metallurgy ............................................................... 24 Figure 4.1: Allocation of increases in total expenditure to functional divisions ...................................... 32 Figure 4.2: Armenia‘s Average Differences with Benchmark Countries, share of GDP ........................ 33 iv Figure 4.3: Compensation of Employees as a Percent of GDP, 2008...................................................... 39 Figure 4.4: The Size of Public Employment in Armenia and other Transitional Countries .................... 40 Figure 4.5: Monthly Earnings by Position in a Sample Ministry, 2010 .................................................. 43 Figure 4.6: Public and Private Health Spending as a Share of GDP, Europe and Armenia .................... 48 Figure 4.7: Health Spending in Armenia: Actual and Projected .............................................................. 49 Figure 4.8: Probability of Dying Between 15 and 60, Armenia and Comparator Regions,..................... 50 Figure 4.9: SHA budget allocation by inpatient state order program, 2009 ............................................ 51 Figure 4.10: The Impact of Household Spending on Health on Poverty ................................................... 52 Figure 4.11: Eliminating Informal OOP: Impact of the Maternity Voucher Program ............................... 53 Figure 4.12: Composition of Armenia Public Spending on Education, 2009 ............................................ 57 Figure 4.13: Distribution of Educational Inputs by Community Type and by School Size, 2009/10 ....... 59 Figure 4.14: TIMSS Performance in Selected ECA Countries .................................................................. 61 Figure 4.15: Road Network Utilization in Armenia................................................................................... 66 List of Boxes Box2.1: Debt vs. Fiscal Sustainability ......................................................................................................... 7 Box 2.2: Armenia -- Key Macroeconomic Assumptions for Baseline Scenario (2010–30) ........................ 8 Box 3.1: Armenia – Tax Administration Diagnostics and ......................................................................... 16 Box 3.2: Regression Analysis to Estimate Tax Leakages .......................................................................... 18 Box 3.3: Armenia – Transactions and Operations Exempt from VAT ...................................................... 28 List of Annexes Annex 1: Selected Review of previous Studies on Tax Administration and Tax Policy in Armenia ........ 72 Annex 2: Regression Results ..................................................................................................................... 74 v ACKNOWLEDGEMENT This report is the result of joint work between the Armenian Ministry of Finance and the World Bank. On the MoF side, Deputy Minister Vardan Aramyan provided overall guidance on debt and fiscal issues; Deputy Ministers Suren Karayan and Vardan Aramyan as well as Deputy Chairman of SRC Artashes Beybutyan on revenue potential assessment; and First Deputy Minister Pavel Safaryan on expenditure rationalization issues. Five staff from the macroeconomic department (Ruzana Gabrielyan and Vahagn Lalayan) and the debt management department (Artak Marutyan, Artur Hambardzumyan and Marina Melikyan) were involved in data collection, attendance of the training on the World Bank – IMF DSA tool and write-up of key sections of the debt and fiscal sustainability chapter, under the guidance of the World Bank team led by Souleymane Coulibaly (Senior Economist), and comprising Gohar Gyulumyan (Economist) and Tigran Kostanyan (Economist). The Debt Sustainability Analysis included in this report was conducted jointly with the IMF SPR unit. The Bank‘s team benefited from the guidance of Asad Alam (Country Director, ECCU3), Yvonne Tsikata (PREM Sector Director), Kazi Matin (PREM Sector Manager), Pedro Luis Rodriguez (Lead economist and Country Sector Coordinator, ECSP1) , Aristomene Varoudakis and Jean-Michel Happi (Country Managers, ECCU3). Owen Smith (ECSHD) drafted the health expenditure analysis, Sachiko Kataoka and Igor Keyfeits (ECSHD) drafted the education expenditure analysis and Rodrigo Archondo-Callao (ECSSD) drafted the road transport expenditure analysis. Jaime Vazquez-Caro (HQ consultant) contributed to the assessment of the tax administration revenue potential, Jana Orac (HQ consultant) provided the public sector wage bill analysis, Mohamed Diaby (HQ consultant) contributed to the quantitative analysis conducted in the wage bill and education sector analyses, and Artsvi Kachatryan prepared multi-country graphs. Leonardo Fernando Hernandez and Tihomir Stucka (PRMED) conducted the training on the DSA tool with overall guidance on the training structure of Sudharshan Gooptu (PRMED Sector Manager and peer reviewer). Munawer Sultan Khwaja (ECSP4) and Sebastian James (CICRS) provided detailed comments on the revenue chapter. Zakia Nekaien-Nowrouz and Nelli Kachatryan provided an excellent support to the team. The team is grateful for the comments and supports received from the peer reviewers (Sudharshan Gooptu (PRMED), Auguste Tano Kouame (LCSPR) and Marinus Verhoven (PRMPS)) at different stages of this report. vi EXECUTIVE SUMMARY 1. Armenia’s structural reforms since 1999 have led to a strong economic record, including low fiscal deficits and declining public debt over the pre-crisis decade. Between 2001 and 2008 Gross Domestic Product grew at an average annual rate of 12 percent and poverty fell from over 50 percent to about 28 percent of the population. Over this period of rapid growth, prudent fiscal management contained fiscal deficits between 0 and 2.5 percent of GDP and helped to reduce public debt from 49 percent to 16 percent of GDP. 2. This fiscal headroom allowed the Government to respond to the crisis with an appropriately large fiscal stimulus. When GDP contracted by more than 14 percent in 2009 and total revenues fell sharply, nominal public spending was increased by 13 percent to shore up the domestic economy and protect the poor and vulnerable. Despite the severity of the crisis, the Government maintained a sound macroeconomic framework while continuing to undertake social protection expenditures to mitigate the impact of the crisis on the most vulnerable people. This was done by securing sizable external financing. The fiscal deficit rose to 7.8 percent of GDP in 2009 and the public debt to GDP ratio rose from 16 percent in 2008 to 40.2 percent. Efforts at fiscal consolidation reduced the fiscal deficit to 5.6 percent of GDP in 2010, but public debt is projected to reach 42 percent of GDP in 2011. 3. Looking forward, further fiscal consolidation will need to be designed carefully to avoid jeopardizing growth. The relatively high, and rising, debt to GDP ratio calls for a cautious fiscal stance to ensure a sustainable level of debt. Given lower growth prospects in Armenia‘s key economic partners (EU and Russia) in the medium term, it appears wise not only to prepare for slower growth in the future but also to create some fiscal room that would permit active policy in the event of economic shocks. Relatively low tax revenues (excluding social contributions) of 16.2 percent of GDP in 2010 and the pressure to control the fiscal deficit reinforce the need for fiscal retrenchment. The 2012-14 Medium Term Expenditure Framework suggests that the Government is envisaging to reduce expenditure while increasing tax revenue, to bring the fiscal deficit below 3 percent of GDP by 2013 (Table I). Although this projected fiscal deficit reduction is to be lauded, putting the brunt of the burden on expenditures may jeopardize economic recovery. Table 1: Armenia Fiscal Framework, 2006-2014 Actual 2012-2014 MTEF Difference in share of GDP, % 2006 2007 2008 2009 2010 2011 2012 2013 2014 2014-2010 Total revenue, o/w 19.4 20.1 20.5 21.1 22.5 22.3 21.2 21.4 21.7 -0.8 Tax revenue 14.5 16.0 16.8 16.1 16.2 17.0 17.0 17.4 17.7 1.5 Social contributions 2.8 2.7 2.9 3.3 3.0 3.2 3.4 3.4 3.5 0.5 Official grants 0.6 0.8 0.4 0.7 0.9 1.5 0.3 0.2 0.1 -0.8 Total expenditure, o/w 20.9 22.4 22.2 28.9 28.2 26.2 24.3 23.8 24.1 -4.1 Current expenditure 16.2 15.8 18.3 22.7 22.0 21.9 20.1 19.4 20.2 -1.8 Capital expenditure 4.7 6.6 3.9 6.2 6.1 4.2 4.3 4.4 3.9 -2.2 Overall deficit -1.5 -2.3 -1.8 -7.8 -5.6 -3.9 -3.1 -2.4 -2.4 3.2 Public debt 19.2 16.4 16.1 40.2 39.2 42.1 43.2 41.9 40.5 1.3 Real growth, in % 13.2 13.7 6.9 -14.1 2.1 4.6 4.2 4.5 4.8 2.7 Source: Armenian Ministry of Finance. vii 4. The immediate challenge, then, is to develop strategies for fiscal consolidation without undercutting economic growth necessary for both social welfare and long run fiscal sustainability. Spending cuts are constrained by the need to maintain or scale up some growth-sustaining spending while at the same time protecting the poor and vulnerable from the aftermath of the 2009 crisis. Effective education and health services provided equitably to the population builds up the country‘s human capital. An adequately maintained road network preserves the country‘s physical infrastructure and efficiently connects production to market. The short term imperative of reducing the fiscal deficit should not jeopardize the country‘s growth prospects by underfunding critical growth-sustaining expenditures. 5. The Government should increase efficiency of all aspects of public finances – tax policy, tax administration, and public expenditures. Armenia has a relatively small government as illustrated by its low expenditure to GDP ratio compared to income-level peers (Figure I). This reflects on key sectors such as education, health and road transport that have been consistently underfunded. Unless revenues are significantly increased, the Government will have to revisit its expenditure priorities to create the spending headroom needed to at least maintain the spending level in these sectors. Tax revenue increases, not expenditures, should bear the brunt of the required fiscal consolidation, to secure growth and maintain debt at sustainable levels. Indeed, it is estimated that another major GDP contraction in the range of 7 percent in the near future would make Armenia‘s debt level unsustainable. Such a shock is not unrealistic, given current low growth and external volatility, and hence the need to protect growth- sustaining public expenditure. Figure 1: Relationship of Public Spending and Per Capita Income in Selected Countries, 2009 55 Ukraine BiH Hungary Total expenditure (as share of GDP) 50 Slovenia Montenegro Czech Moldova Belarus Poland 45 Estonia Serbia Lithuania Croatia Georgia Bulgaria Turkey Latvia Slovakia 40 Kyrgyz Russia Romania 35 Macedonia Azerbaijan Uzbekistan Albania Turkmenistan Uruguay 30 Kosovo Malaysia Tajikistan Armenia 2010 25 Chile Armenia 2008 Kazakhstan 20 15 10 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 Ln of GNI per capita Atlas method Source: World Bank WDI database, Staff calculation. 6. On the revenue side, reforms can improve outcomes in both the short and the medium run. Armenia can increase collection in the short term by reforming excise and mining-specific taxation, as recommended by previous studies by the International Monetary Fund and the World Bank. If supported by the Government, these tax policy measures would have a positive short term impact on revenues, estimated between 0.9 and 3.2 percent of GDP (Table II). Identifying and closing the tax loopholes would require more time given the need to assess thoroughly the revenue losses created by the various tax viii expenditures scattered in many tax laws and decrees, and to balance them against any developmental impact they may have. Although this study identifies the key tax loopholes, quantifying their revenue impact is deferred to another ongoing tax study. In the medium term, improvements in the tax administration could generate an additional 1.4 to 2.6 percent of GDP in revenue. 7. The expenditure reform should proceed with caution, and avoid spending cuts that will compromise productive capacity and exacerbate poverty. Evidence suggests that health care, education, and physical infrastructure (roads in particular) may require more, not less, spending in order to support the future growth required for sustainable fiscal policy (Table II). Poorly targeted subsidies should be cut, and administration should be made more efficient. However, cuts to education, health, and social security can be justified only if they reduce wasteful programs. If expenditures are cut in ways that jeopardize Armenia‘s growth prospects, compromising the quality of the future workforce and the transportation and communication infrastructure, these cuts may worsen the fiscal situation in the medium run, even if they provide short run fiscal relief. Policy Options to Increase Revenues 8. The Armenian tax system comprises two parallel taxation structures: a regular and a presumptive regime. The regular taxation regime reflects contemporary taxes (corporate income tax, individual income tax, VAT and some standard type of excises), while the presumptive (in lieu of VAT and corporate income tax) covers the bulk of informal merchants. The first structure is friendly to foreign investors: it provides favorable treatment to non-resident taxpayers, combined with a relatively low profit tax rate compared to other countries. The revenue engine seems to be the VAT. The second solves the problem that high compliance standards pose to informal taxpayers and provides a small contribution to revenues. Yet it creates a universe of taxpayers that are neither income nor VAT payers, operating in the informal economy. 9. Tax collection remains relatively low in Armenia compared to other countries. The tax revenue to GDP ratio in Armenia (excluding social contributions) increased by nearly 2 percentage points between 2006 and 2010, to reach 16.2 percent. In comparison, between 2004 and 2008, Georgia‘s tax ratio increased from 18 to 25 percent. The potential rise in tax revenues in Armenia is estimated to be between 2.3 and 5.8 percent points of GDP (Table II). To improve the rate of tax collection, tax administration as well as tax law and tax rates should be reformed. 10. Reform of the tax administration should address two key issues: the informal economy and corruption. In Armenia, tax administration is undermined by the difficulty of enforcing tax laws in the informal economy and by corruption within the enforcement process. To fully tap Armenia‘s revenue potential under the existing tax laws or a new tax code, a strategic rethinking of the tax administration is needed. So far, SRC has focused on filers to review returns and has ignored the existence of the informal economy which is particularly important for its links to the formal economy that can undermine the tax base of the formal sector. To address corruption an ad-hoc program needs to be developed and enforced. The Government Decree N 1972 of October 8, 2009 provides the basis to develop such program, but more need to be done of the enforcement of the Action Plans developed since then. To address the informal economy, constant efforts should be devoted to identifying activities now covered by the presumptive taxation regime that can be brought under the regular taxation regime. This should be complemented by reducing tax compliance costs and closing tax privileges that entice formal businesses to rely on the informal economy, which is estimated currently to produce about 8 percent of GDP (excluding agriculture). 11. Revamping the audit function of the State Revenue Committee will significantly improve revenue collection. Tax administration enforcement efforts need to be better targeted based on the type ix of tax and the size of firms. For instance, very large firms appear to receive privileged treatment in VAT collection. There is therefore the need to thoroughly review the use of tax privileges to assess their developmental goal relative to the revenue foregone, since in some cases it is not clear that there is a developmental purpose. Enforcement also could be enhanced through the promotion of a risk-based audit function of the Large Taxpayers Unit to better target companies evading tax payments. Data suggest that firms with a turnover between 1 and 10 billion AMD (between US$2.6 and 26 million) underpay Corporate Income Tax and Personal Income Tax, while firms with turnover higher than 10 billion AMD underpay VAT. The LTU should focus its audits accordingly to improve efficiency and augment revenues. Tax administration could also be ameliorated by lengthening the statute of limitations (currently only three years) on tax fraud. Armenia requires far more rapid enforcement action than most countries, reducing the likelihood of a successful audit of inaccurate tax returns. 12. Improved tax administration requires improved storage and tracking of information as well as advanced quantitative and qualitative analytical tools. Tax administration must be thought of as a digital institution. It is essentially a complex information flow out of which a number of legally defined products emerge. Today, information management that relies mostly on paper is far less efficient, hindering productive use of the information. As a leading institution in the current digital environment, SRC needs to generate such capacity. The ongoing tax project and the Development Policy Operation are providing support to expand the IT capacity of SRC, and this effort should be continued. 13. On tax policy measures, the excise tax and mining tax system should be reformed promptly and harmonized with the VAT. Excise tax reform should bring tobacco back to the regular excise regime, and raise specific rates to reflect inflation as well as the negative externalities caused by excisable goods. Revamping the excise regime is expected to increase revenue between 0.5 and 2.6 percent of GDP, and the proposed new tax regime for mining is expected to increase revenue between 0.4 and 0.6 percent of GDP. The latter will combine the current natural resource user fee for metallic minerals and royalty into a single royalty with sales as the tax base, and a higher rate on a sliding scale to capture windfall profits. In Armenia, the effective rate of taxation of mining (calculated over the life of a mine) is low by international standards, and the annual effective rate has been falling systematically over the past several years even as volume and prices of mineral exports have increased. The main specific tax for the mining sector, the Natural Resource User Fee, is levied on an outdated base (output extracted) valued at estimated international prices. Combining the current NRUF on metallic minerals with royalty into a single royalty to be managed by the SRC will simplify administration. Also with sales as base, the actual unit-value or export price received by mining companies in Armenia will be part of the base and will reflect both increases in volume and in prices of exports in the tax base, avoiding the systematic under- estimation that had occurred in the past. The higher rate placed on a sliding scale will help to capture windfall profits when there is a boom in international demand and prices. 14. During the review of the tax laws, recent changes must be revisited and reversed, along with other loopholes. In spring of 2009 the Government increased the threshold for VAT exemption for the Small and Medium Enterprise sector from 3 million AMD (US$7,900) annual turnovers to 58 million AMD (US$153,000) to create economic incentives and maintain jobs.1 This change in the tax code undermines data verification (since purchases and sales between large firms and smaller firms, now excluded, will no longer be reported by both parties) and it encourages transfer pricing to shift value added from large to smaller tax units. IMF and World Bank studies estimate this new threshold to be too high. In fact, this study revisits all the changes made in the various tax laws since 1997 and points to wide range of provisions that have almost certainly reduced revenue collection without any assurance that the developmental goals targeted by these provisions, such as job creation, were met. The audit function of 1 Various changes have been made to the threshold for VAT exemption since 1997. The latest change was estimated by the Ministry of Finance to have induced about 24 billion AMD of forgone revenues. x SRC does not have a mandate to control these tax privileges. The very first tax policy measure to increase revenues would therefore be to revisit these provisions and take the opportunity of the drafting of the tax code to close tax loopholes. 15. Successful tax reform will require close collaboration between the SRC in charge of tax administration and the MoF, in charge of developing tax policy. A strengthened Tax Policy Unit, sitting in the MoF but closely collaborating with SRC for data exchange and testing of the tax policy envisaged, will be a major institutional improvement. The unit should be expected to monitor revenue collection, evaluate the economic, structural, and revenue aspects of tax policy, provide regular tax expenditure analysis, consider the revenue impact of non-tax economic policies, and forecast future tax revenues. To perform these tasks, it is necessary to establish an effective information system within the Government, linking SRC and MoF databases. All forecasts – revenue forecasts or tax expenditure analysis, for example – must use a common methodology and a common set of assumptions. These assumptions include economic variables tracked by the MoF, such as growth in the national income, the rate of inflation, interest rates, and the international environment, as well as tax-specific data tracked by the SRC. Policy Options to Rationalize Expenditure 16. Unlike the tax system, the expenditure side of the budget offers fewer clear opportunities for saving. This is in part because areas like defense and pensions were not analyzed. This report examined the wage bill, health, education and road transport expenditure to assess the scope to improve efficiency or to eliminate specific types of spending. While there is clear evidence that efficiency could be increased, some measures supportive of growth may increase expenditure rather than reduce it (Table II). Overall, caution is warranted in cutting spending that supports investment in human and physical capital and that maintains the safety net against impoverishment. Even desirable spending cuts must be phased in gradually so as not to undermine recovery from the economic crisis. 17. The public sector wage structure should be reviewed and reformed. There is a need to harmonize the different public sector remuneration systems and strengthen the analytical foundations for a coherent, affordable public sector remuneration policy. It will be important to design a unified grading structure and job classification system. New performance appraisal and bonus practices should be introduced to end the current discrepancies in wages. These changes will require better data collection and analysis. Analytical work will permit forecasts of the fiscal impacts of any human resource policy changes. Data must be collected and analyzed concerning performance appraisal, recruitment, and staff flows. Wages can be modified where evidence accumulates about difficulties in finding applicants for recruitment and retention of staff in specific positions. Data such as resignation rates by specific job title/grade or profession, and numbers of qualified applicants for advertised vacancies by specific job title/grade or profession will need to be tracked. 18. Health outcomes in Armenia, as well as the relatively high level of out-of-pocket expenditure for health care, suggest that the sector may well be a candidate for greater public spending rather than less. On health spending, a key priority in the short term would be to expand the health poverty program by equalizing eligibility at a score of 30 (instead of the current 36) in line with the Family Benefit Program. There is also a need to reform the disability program for categories II and III disability groups. The first of these reforms is likely to raise expenditure on health care (estimated at 0.1 percent of GDP), the second, to reduce them. One reform that is likely to improve Armenia‘s high adult mortality rate would be to make a core set of essential, generic pharmaceuticals accessible to everyone at low cost by expanding the basic benefit package. In the medium term, the Government needs to decide whether to continue expanding coverage and reducing out-of-pocket health spending, which require xi additional funding to the sector; or to scale down on promises and target healthcare services more to the poor. 19. Education spending in Armenia is inefficient with rural schools producing inferior results at higher cost, but education system reforms may still not result in significant savings. The large network of rural schools serving 40 percent of Armenia‘s student population is not producing high-quality educational outcomes. The system perpetuates a large number of small schools through an overly generous ―fixed‖ per-school component of the funding formula and reallocation of funds among schools. 27 percent of Armenia‘s schools serve fewer than 100 students each, at a cost per student 270 percent of the national average and 360 percent of the per-pupil cost in the largest schools (with over 700 pupils each). Yet these ―mini-schools‖ produce the worst educational outcomes. To address these issues, there is a need to: (i) conduct a thorough assessment of the quality of education and educational opportunities currently provided by rural schools (initial analyses show that educational opportunities in the smallest schools lag far behind those available to graduates of large urban institutions); (ii) review and reform the current per capita financing formula (there appears to be a need to reduce the fixed component of the PCF in favor of more nuanced adjustment factors); (iii) consider alternative ways of providing education in rural areas (consolidation of small rural schools into hub schools at least at the higher school level may be warranted, though transportation needs must be reviewed concurrently); and (iv) address the issues of low quality of teaching in rural areas (for example introducing a rotation of younger teachers, and appropriate incentives). 20. Currently, too little is spent to maintain and rehabilitate roads; the current situation is not sustainable and comes at a cost. In 2009, the Government spent only US$110 million on rehabilitation, periodic and routine maintenance, and only US$79 million were initially planned in the 2010 budget. The current standards used are too high for the actual traffic and, combined with the lack of maintenance, contribute to high costs, paid by drivers and vehicle owners as well as through the budget. Reforms should include modification of road standards to reflect traffic, use of efficient contracts (such as performance based contracts that include rehabilitation and maintenance), and allocation of sufficient funds to maintain the most used part of the network (Table II). Structure of the Report 21. The report is in two volumes: a synthesis volume and a background volume. The synthesis volume summarizes the macroeconomic context of Armenia (chapter 1), analyzes the recent debt dynamics and its implication for fiscal consolidation (chapter 2), then assess the revenue potential the Government can tap (chapter 3) while ensuring key growth-sustaining and poverty-reducing expenditures are maintained (chapter 4). The background volume provides more details on the assessment of the tax potential of the mining sector (chapter 1), and thoroughly analyzes the efficiency of spending on the public sector wage bill (chapter 2), health (chapter 3) and education (chapter 4). xii Table 2: Policy Recommendations for Fiscal Consolidation and their Fiscal Impacts Fiscal Impact Policy Measures Policy Area (as a share of 2010 GDP) Revenue Short Term Revamp the excise regime (bring tobacco to the excise regime, adjust up rates to reflect inflation, introduce VAT payment on excisable Tax Policy 0.5-2.6 goods) Revamp the mining taxation regime (combine natural resource user fee on metals and royalty into a single royalty with sales as the tax base Tax Policy 0.4-0.6 and a higher rate on a sliding scale to capture windfall profits) Reinforce risk-based audits using regression analysis (firms with turnover higher than 10 billion AMD tend to underpay VAT, while Tax 0.4-1* firms with turnover between 1 and 10 billion AMD tend to underpay Administration CIT and PIT) Medium Term Identify tax expenditures, quantify revenue losses and phase out if Tax Policy not quantified necessary Expand base to informal economy by bringing back the presumptive Tax 1-1.3 regime to regular taxation Administration Tax not quantified Reinforce ICT infrastructure and build analytical capacity Administration Establish a strong Tax Policy Analysis Unit in the MoF, closely Tax Policy not quantified working with SRC Introduce a fully functioning Green Channel by default operation of Tax customs clearance, (backed by robust post release verifications, 0-0.3 Administration controls, audits, and investigations) and strengthen customs preventive and enforcement capacity Expenditure Short Term Prioritize expenditures to reflect Government stated development MTEF Not quantified objectives Expand the health poverty package by aligning the score required to that of the Family Benefit Program (from 36 to 30) Healthcare -0.1 Protect the road maintenance and rehabilitation budget at least at Road -0.3 US$110 million Transport Medium Term Reorganize the delivery of educational services in rural areas Education not quantified Harmonize the various public service pay systems Wage Bill not quantified Introduce co-payment for category II and III disability program Healthcare not quantified Introduce performance-based contracting for road maintenance and Road not quantified rehabilitation Transport Source: World Bank Staff estimates. * This estimate would be higher if report had information to assess the extent of under-declaration of taxable incomes by tax payers. xiii CHAPTER 1. MACROECONOMIC CONTEXT Since 1991, the Armenian economy has faced all the stages of the economic cycle from stagflation to overheating. Exposure to external shocks - both positive and negative - largely explains the volatility of real growth, with average real growth of 2 percent over 1991-2009 contrasting with the double digit real growth over 2001-2007. The long-term real growth is assumed at the level of 4 percent to be somewhat higher than 20-years average (including deep recession of 1991-1993 and 2009), but much lower than the average growth over 1994-2008 (period of uninterrupted growth); inflation is taken at 4±1½ percent, reflecting the inflation targeting framework of the Central Bank; fiscal variables are in line with assumed fiscal consolidation; and external sector variables – in line with global recovery. 1.1. Armenia had a strong record of sustained economic growth before the global economic and financial crisis. From 1994 Armenia registered moderate economic growth, which accelerated to a two- digit level after 2000, supported by a positive external environment and reforms to the business climate. Between 2001 and mid-2008 the economy grew at an average of 12 percent, combined with progressive social safety net programs, led to a fall in poverty from over half of the population in 1999 to about 28 percent (Figure 1.1). During this period, the share of investments in GDP nearly doubled to reach 44 percent in 2008, reflecting private investment fueled by remittances mainly from Russia and channeled to the construction sector. Figure 1.1: Real GDP Growth, Per Capita GDP, Figure 1.2: Armenia Public Debt, 2005-2009 and Poverty 4 50% 4150 55 3.5 45% 3650 40% 3150 35 3 35% 2650 2.5 30% in bln USD 15 2150 2 25% 1650 -5 20% 1.5 1150 15% -25 1 650 10% 0.5 5% 150 -45 0 0% 1990 1992 1994 1995 1997 1999 2001 2002 2003 2004 2006 2008 2010 1991 1993 1996 1998 2000 2005 2007 2009 2005 2006 2007 2008 2009 2010 Per capita GDP, USD Real GDP growth, % (RHS) Poverty rate (% of total population) External debt Domestic debt Debt-to-GDP (RHS) Sources: Ministry of Finance and World Bank Staff. Sources: Ministry of Finance and World Bank Staff. 1 1.2. This record growth led to low fiscal and external deficits as well as public debt. Between 2001 and 2008, fiscal deficit fluctuated between 0 and 2.5 percent. Total public debt-to-GDP ratio constantly decreased from 46 percent in 2001 to 16 percent in 2008, while total public debt nearly doubled to reach US$1.9 billion (Figure 1.2). Net FDI inflows also increased as a share of GDP, from 3.3 percent in 2001 to 8.1 percent in 2008. The current account balance consistently decreased from 8 percent in 2001 to 1.3 percent in 2006, but then quickly increased to reach 11.5 percent in 2008, led by a widening trade deficit. Mirroring this deterioration of trade balance, gross remittances inflows almost doubled between 2006 and 2008 to reach US$1.1 billion. 1.3. The fiscal headroom provided by the growth dividend allowed the Government to expand public expenditure. By doubling total expenditure to nearly AMD 930 billion (US$2.5 billion) between 2006 and 2009, the Government created headroom for a steady increase in social security spending from 4.8 to 7.9 percent of GDP, in defense from 3.0 to 4.2 percent, in general public services from 2.4 to 3.6 percent, in education from 2.5 to 3.5 percent, and in public order and law enforcement from 1.6 to 2.2 percent. On the other hand, infrastructure expenditure has been fluctuating between 1.8 and 4.3 percent of GDP, and health expenditure between 1.5 and 1.8 percent. Although we cannot assess the overall effectiveness of this expenditure expansion, significant poverty reduction due to social spending, a peaceful Armenia as well as overall education achievements indicate that at least these core Government mandates have been relatively effective. Figure 1.3: Public Spending and Per Capita Income in Selected Countries, 2009 55 Ukraine BiH Hungary Total expenditure (as share of GDP) 50 Slovenia Montenegro Czech Moldova Belarus Poland 45 Estonia Serbia Lithuania Croatia Georgia Bulgaria Turkey Latvia Slovakia 40 Kyrgyz Russia Romania 35 Macedonia Azerbaijan Uzbekistan Albania Turkmenistan Uruguay 30 Kosovo Malaysia Tajikistan Armenia 2010 25 Chile Armenia 2008 Kazakhstan 20 15 10 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 Ln of GNI per capita Atlas method Source: World Bank WDI database, Staff calculation. 1.4. Yet, Armenia remains a low-spending country compared to income-level peers. From 2001 to 2008, Armenia‘s total expenditure to GDP fluctuated between 19 and 23 percent, compared to a Europe and Central Asia average higher than 30 percent of GDP (Figure 1.3).Table 1.1 confirms that over 2006- 09, Armenia health/education expenditures ratio was half that of the average ECA upper middle income country. Although spending on the real sector improved from 2001-05 to 2006-09, it remained significantly below that of the average ECA low middle income countries. In contrast, the country appears to have been over-spending on defense compared to the two benchmark regions, with 2 defense/education expenditures ratio at 1.02 over 2006-09 compared 0.63 for ECA upper middle income countries. However, note the defense/education expenditure ratio in Azerbaijan and Turkey estimated at 1.2 and 0.45 over 2006-09. Table 1.1: Expenditure Ratios in Armenia and Selected Economies Armenia ECA low MIC ECA upper MIC 2001-05 2006-09 2001-05 2006-09 2001-05 2006-09 Health/Education 0.46 0.47 0.61 0.63 1.01 0.94 Social/Education 1.75 1.71 2.21 1.95 3.37 2.52 Defense/Education 1.11 1.02 0.28 0.63 0.66 0.40 Economic relations/Education 0.54 0.74 0.69 1.32 1.46 0.99 Source: ECA expenditure database ECA low MIC includes Albania, Georgia, Kosovo, Moldova, and Ukraine. ECA upper MIC includes Belarus, Bulgaria, Kazakhstan, Latvia, Lithuania, Poland, Romania, Russia, and Serbia. 1.5. On the other hand, tax collection performances have fallen short of expectations. Although the Central Government tax revenues (excluding social contributions) to GDP consistently increased from 13 percent in 2001 to 16.8 percent in 2008, they are still low compared to Armenia income-level peers. During 2004-2008, Georgia‘s tax ratio increased from 18 to 25 percent. In OECD countries, the average tax-to-GDP ratio over the same period was higher than 30 percent and it neared 40 percent in EU-15 countries. Figure 1.4 compares more systematically Armenia to some benchmark countries, and confirms that it is at the lower end of tax collection performance. Figure 1.4: Tax Revenue and Per Capita Income in Selected Countries, 2008 50 Taxes and social contributions ( % of GDP) Belarus 45 Hungary 40 Slovenia BiH Serbia Czech 35 Poland Moldova Bulgaria Estonia Ukraine Russia Croatia 30 Macedonia Lithuania Romania Slovakia Georgia Latvia 25 Tunisia Turkey Uruguay Armenia Chile 20 Tajikistan Kazakhstan Kyrgyz Thailand 15 El Salvador Azerbaijan 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 Ln of GNI per capita Atlas method Source: World Bank WDI database, OECD, IFS, and staff calculation. 3 1.6. The recent global crisis put a brake on Armenia’s decade-long economic expansion, exacerbating fiscal and debt pressures. The August 2008 conflict between Russia and Georgia was the first event to affect the Armenian economy, given its dependence on transit through Georgia. A few months later, the effects of the global economic crisis reached the country by means of reduced exports (- 32.8 percent by the end of 2009), imports (-25 percent) and remittances (-35 percent). This had a direct impact on the domestic economy, particularly on the construction sector (-54 percent by the end of 2009). The economy contracted by 14.1 percent, translated into a 13 percent decline in non-grant revenues (tax revenues declined by 14 percent) while public expenditures increased by more than 14 percent to shore up the domestic economy and protect the poor and vulnerable. As a result, the debt-to-GDP ratio increased to 40 percent by the end of 2009 and is projected to reach 42 percent by the end of 2011. 1.7. Although such levels of public debt ratios are not unprecedented for Armenia, the current debt accumulation situation is very different from earlier debt accumulation. Shortly after independence, Armenian debt was mainly driven by the conversion of contingent liabilities of the energy bill arrears of enterprises into sovereign debt. Then, despite a stabilization program with the World Bank and the IMF, a significant share of budget deficits continued to be financed by external credit, resulting in the rapid accumulation of public external debt, especially after 1996. After peaking at nearly 50 percent in 1999, Armenia‘s total public debt to GDP continuously decreased to reach 16 percent in 2007, reflecting the growth dividend over this period. With the wake of the global crisis, the anticipated revenue shortfall and the need to protect the poor and vulnerable as well as maintain economic activities at a reasonable level led the government to secure external deficit financing from both multilateral and bilateral creditors. 1.8. Fiscal consolidation should be central to Armenia’s recovery from the global crisis. So far, the fiscal stance has been pro-cyclical, with an expansionary fiscal stance while the country was operating above its potential between 2005 and 2008 (Figures 1.5 and 1.6). The fiscal stance dramatically changed in 2009. The large GDP contraction forced the Government to adopt counter-cyclical measures with automatic stabilizers and discretionary spending driving the change in overall fiscal balance. The Government needs to reverse its debt accumulation path, and reduce significantly its fiscal deficit from 7.8 percent in 2009 to a more sustainable level below 3 percent over the medium term. Figure 1.5: Fiscal Stance of the Government Figure 1.6: Estimated Potential GDP Sources: Ministry of Finance and World Bank Staff. Sources: Ministry of Finance and World Bank Staff. 1.9. This new situation is exacerbated by a low growth context. In the first quarter of 2010, the real sector sent encouraging signs of recovery, but by the end of the year real growth was a meager 2.1 percent, in line with the 20-year average. Given that the global economy, particularly the EU and Russia which are Armenia‘s main economic partners, is geared to a lower growth path, the country will most likely operate in a similar context over the medium term. Although the projected growth in the 2012-14 MTEF (between 4 and 5 percent) is higher than the previous forecast (between 2 and 4 percent under the 2011-13 MTEF), these growth rates are significantly lower than pre-crisis rates (Table 1.2 and Figure 1.1). 4 Table 1.2: Armenia Fiscal Stance after the 2009 Crisis Actuals 2011-2013 MTEF Difference 2012-2014 MTEF Difference in share of GDP, % 2006 2007 2008 2009 2010 2010 2011 2012 2013 2013-2010 2011 2012 2013 2014 2014-2010 Total revenue, o/w 19.4 20.1 20.5 21.1 22.5 23.1 23.4 23.0 23.3 0.2 22.3 21.2 21.4 21.7 -0.8 Tax revenue 14.5 16.0 16.8 16.1 16.2 17.7 18.2 18.7 19.1 1.4 17.0 17.0 17.4 17.7 1.5 Social contributions 2.8 2.7 2.9 3.3 3.0 3.3 3.4 3.5 3.6 0.3 3.2 3.4 3.4 3.7 0.7 Official grants 0.6 0.8 0.4 0.7 0.9 1.3 1.2 0.3 0.1 -1.2 1.5 0.3 0.2 0.1 -0.8 Total expenditure, o/w 20.9 22.4 22.2 28.9 28.2 29.1 28.3 27.3 27.3 -1.8 26.2 24.3 23.8 24.1 -4.1 Current expenditure, o/w 16.2 15.8 18.3 22.7 22.0 24.0 23.5 22.1 22.0 -1.9 21.9 20.1 19.4 20.2 -1.8 Interest payment 0.3 0.3 0.3 0.5 0.9 1.1 1.4 1.4 1.4 0.3 1.1 1.1 1.1 1.0 0.2 Capital expenditure, o/w 4.7 6.6 3.9 6.2 6.1 5.1 4.7 5.2 5.3 0.2 4.2 4.3 4.4 3.9 -2.2 Overall deficit -1.5 -2.3 -1.8 -7.8 -5.6 -6.0 -4.9 -4.3 -4.0 2.0 -3.9 -3.1 -2.4 -2.4 3.2 Primary deficit -1.2 -2.0 -1.4 -7.3 -4.8 -4.9 -3.5 -2.9 -2.6 2.3 -2.8 -2.1 -1.3 -1.4 3.4 Public debt 19.2 16.4 16.1 40.2 39.2 47.2 48.5 48.1 46.2 -1.0 42.1 43.2 41.9 40.5 1.3 Real growth, in % 13.2 13.7 6.9 -14.1 2.1 1.2 2.5 3.0 3.5 2.3 4.6 4.2 4.5 4.8 2.7 Source: Ministry of Finance. 5 1.10. The Government started the fiscal consolidation process, although heavy reliance on expenditure rationalization may be at odds with the need for stronger growth. The 2011-13 MTEF forecasted a 2 percent deficit reduction. The Government tightened its fiscal stance under the 2012-14 MTEF, projecting a 3 percent deficit reduction to bring the fiscal deficit below 3 percent by 2013. This correction is expected to come mostly from expenditure cuts by nearly 4 percent points of GDP, while tax revenue would contribute less than 2 percent points of GDP. A closer look at the 2012-14 MTEF indicate that the cuts mainly target capital expenditure (-2.2 percent of GDP), construction works in disaster areas (-0.7 percent of GDP), education expenditure (-0.4 percent), health care (-0.3 percent), defense spending (-0.2 percent) and subsidies to cover the loss of enterprises (-0.1 percent). Social spending is projected to increase by 0.5 percent point of GDP and tax revenues by 0.9 percent. Cuts in education, healthcare and capital expenditure may be problematic if they are not targeted to wasteful expenditure. Indeed, as is demonstrated in this report, Armenia education and health outcomes need to be significantly improved, and some infrastructure expenditure (road transport is thoroughly studied) scaled up to sustain long term growth. 1.11. The brunt of the fiscal consolidation should be borne by tax revenues increase. Focusing on a narrow range of sources of revenues namely large taxpayers (CIT, PIT and VAT payments), customs duties, presumptive taxation, excise and mining taxation, this study estimates the potential increase in tax revenues in the range of 2.3 and 5.8 percent of GDP. Of this total, between 1.4 and 2.6 percent derive from tax administration measures (improved audits of CIT, PIT and VAT paid by large taxpayers, improved audit of customs duties and improving the coverage of presumptive taxation), and between 0.9 and 3.2 from tax policy measures (excise and mining taxation). The revenue potential from tax policy is a low hanging fruit expected by bringing tobacco back to the regular excise regime, and raising specific rates to reflect inflation as well as the negative externalities caused by excisable goods. The other measure is to revamp the mining taxation by combining the natural resource user fee on metallic minerals and royalty into a single royalty regime to be managed by the SRC, changing the tax base from extracted outputs to sales which is easier to audit, and setting a higher royalty rate with a sliding scale to capture windfall profits. Revenue potential from tax administration will need a longer term horizon to materialize, and is expected to come from bringing some activities from the presumptive to the regular taxation regime, improving audit and modernizing the IT system to increase the efficiency of the revenue administration. 1.12. Going forward, some debt reduction is desirable to be able to implement a counter-cyclical fiscal policy, if the country were to face another major shock. How quickly the fiscal consolidation is done will affect the country‘s debt trajectory. To address these challenges, the government can consider either refinancing its debt, adjusting faster or growing faster. To grow sustainably, Armenia‘s growth strategy will need to rely less on the domestic market as in the recent past, channeling its resources to diversify its production and export. In the context of this study, the sustainability of macroeconomic framework gets dual importance. First of all, sound macroeconomic situation implies greater repayment capacities and less need for further borrowing. For example, favorable macroeconomic situation allowed reducing public debt / GDP ratio from 49 to 16 percent over 1999-2008. And just oppositely, deterioration of macroeconomic performance resulted in immediate jump of the debt / GDP ratio to 40 percent in 2009. Broad-based economic growth is also necessary for adequate public revenue performance. 6 CHAPTER 2. DEBT AND FISCAL SUSTAINABILITY Although Armenia’s debt management practices have been always considered satisfactory, sharp increase of public debt/GDP ratio back to the level of 2003, as well as uncertainties about the pace of macroeconomic recovery (including global prospects), called for prior analysis to assess the country’s debt management potential. With long-term real growth at 4 percent, inflation stable at 4 percent, fiscal parameters in line with the fiscal consolidation assumed by the MTEF and external variables in line with global recovery, Armenia’s public debt and fiscal sustainability analyses show no major risks. The stress tests, however, show that a depreciation shock of 30 percent or another economic contraction of7 percent can push debt dynamics on to an unsustainable path. To curb down its debt-to- GDP ratio to 30 percent in 10 years time, the government will need to run at most a primary deficit of about 2 percent. 2.1. Armenia’s debt and fiscal situation needs to be considered from a sustainability and solvency point of view. Armenia is solvent if the net present value of the income stream (excluding interest payments) is at least as large as the net present value of expenditure plus any initial debt. This means that the country is meeting its inter-temporal budget constraint. It implies for the government that the net present value of future primary balances must be equal to or greater than the net present value of the public debt stock. Armenia‘s fiscal policy stance is sustainable if public expenditures are in line with available financing sources without needing either a restructuring of debt obligations or a large correction in the balance of income and expenditure. A country does not need to reduce the balance of its debt to remain solvent (Box 2.1). Box2.1: Debt vs. Fiscal Sustainability Strictly speaking, solvency requires that debt grows at a rate less than the rate of interest. If this condition is satisfied, all debt, external or internal, can be serviced. A wedge can be driven between the two questions (fiscal sustainability and external creditworthiness, even in the absence of private foreign debt) when there is insolvency but external debt is considered more important than internal debt. In that case even a situation of fiscal unsustainability could be compatible with external creditworthiness. The possibility of default on internal debt would explain the difference. However, since the country‘s internal debt is relatively low (less than 6 percent of GDP in 2010), this particular configuration is of no practical interest in Armenia. A bigger wedge between the answers to the two questions appears when willingness to pay is introduced. A government may be able to raise sufficient resources to finance its debt service, but may simply not be willing to do so. This is clearly a bigger issue in the case of external debt than in the case of internal debt. A related issue has to do with foreign exchange availability. The Government may be able to produce sufficiently high surpluses but might not be able to transform the internal surplus in externally usable assets; foreigners will insist on payments in dollars, yen or other international currencies, but not in Drams. However this is an exchange rate issue; the problem will only arise if there is a fixed or at least not market determined exchange rate without sufficient reserve availability to meet all funding requirements in foreign currency (an inconvertible exchange rate). This will not be an issue in Armenia if the Dram is allowed to float without too much intervention from the Central Bank, since imbalances in demand and supply would translate into exchange rate movements rather than foreign exchange shortages. Whenever the temptation to intervene dominates, and especially if convertibility is suspended, the foreign exchange shortage problem will emerge as a source of external debt servicing trouble. 7 2.2. Between 2005 and 2010, Armenia’s debt-to-GDP nearly doubled, raising concerns about the debt burden and repayment capacity. As shown in the macroeconomic context chapter (Figure 1.2), public debt was less than 25 percent of GDP in 2005, with more than 95 percent of the stock being external debt. This level decreased to reach 16 percent in 2007, with the share of domestic debt slightly increasing to 12 percent of the debt stock. The 2009 global economic and financial crisis dramatically reversed this trend, putting the debt-to-GDP ratio to 40 percent in 2009. The debt spike is projected to reach 42 percent in 2012, with the share of external debt maintained at nearly 90 percent. Going forward, it is therefore important to properly assess the short and long term impacts of the debt level on the country‘s fiscal sustainability. This chapter addresses the two following questions: (a) Are the debt dynamics implied by the current macroeconomic context sustainable? (b) Given those debt dynamics, what would be the sustainable fiscal stance for Armenia? A. IS ARMENIA DEBT DYNAMICS SUSTAINABLE? 2.3. Projections of the debt-to-GDP ratio based on the most likely macroeconomic framework are essential to assess the sustainability of Armenia’s debt level. The relationship between the solvency condition and the debt-to-GDP ratio stems from the fact that if the debt-to-GDP is either stable or declining, the solvency condition is automatically met.2 This provides a strong rationale for evaluating solvency by looking at the projected behavior of debt ratio. Box 2.2 summarizes the key macroeconomic assumptions used for the baseline scenario. Box 2.2: Armenia -- Key Macroeconomic Assumptions for Baseline Scenario (2010–30) • Real GDP growth projected at around 4 percent over 2010-2030, below the 10-year historical average (9 percent) but above the 20-year historical average (2 percent) • Inflation projected at around 4 percent over 2011-2030 in line with the Central Bank of Armenia target of 4±1½ percent • The overall fiscal deficit projected to remain between 2-3 percent of GDP starting 2013, in line with the ongoing fiscal consolidation • Net domestic financing (NDF) expected to remain around 2 ½ percent of GDP by 2013, rising to about 3 percent of GDP in 2015, declining thereafter to an average of about 2 percent of GDP in 2016-2030, in line with assumed fiscal consolidation • Current account deficit projected to narrow from 16 percent of GDP in 2009 to 12 percent of GDP in 2011, and gradually narrowing to about 8 percent in 2015 as exports and remittances pick up in line with the global recovery • Exports are projected to grow at 11% over 2010-2015 (compared to average 15% over 2000-2009) as new investments become operational, and then stabilize at 8% over 2016-2030 2 See Blanchard, O. and S. Fisher, 1989, Lectures on Macroeconomics, Chapter 2, MIT Press, Cambridge MA. Also see the discussion in Barro, R. and X. Sala-i-Martin, 1995, Economic Growth, McGraw-Hill, New York, NY. 8 2.4. The public debt sustainability analysis focuses on debt incurred domestically and externally by the public sector, with all variables expressed in domestic currency. The analysis identifies different channels that contribute to the evolution of the debt-to-GDP ratio: the primary deficit and the endogenous/automatic factors, which include the real interest rate, real GDP growth, and exchange rate movements. Other debt-creating or debt-reducing flows, e.g., from recognition of contingent liabilities or privatization receipts, are also included. Changes in gross debt arising from other below-the-line operations such as repayment of debt financed by a reduction in financial asset, and cross-currency movement are included in a residual. Baseline Scenario 2.5. The baseline scenario shows the public debt-to-GDP ratio increasing to 43 percent in 2012, then decreasing to reach 26 percent by 2030. The net present value of public sector debt initially would increase gradually from about 33 percent of GDP in 2009 to about 36 percent of GDP in 2012, reflecting the increase in external borrowing. The net present value of the stock of debt is then projected to gradually drop reaching 22 percent of GDP by 2030. The net present value of the debt to revenue ratio would increase from 13 percent in 2008 to a staggering 170 percent in 2011, before dropping again to about 90 percent by 2030, indicating a potential liquidity problem. 2.6. Under the baseline scenario, liquidity indicators point to some medium term pressures with public debt service to revenue increasing from 11 percent in 2011 to 27 percent in 2013 before decreasing to 14 percent by 2030. Note for reference that social safety nets expenditure, which constitutes the largest expenditure component, accounted for 34 percent of revenues in 2009. Gross financing needs (defined here as the primary deficit plus debt service plus the stock of short-term debt at the end of the previous period) increased from 4 percent of GDP in 2008 to 11 percent in 2009 and was estimated at 6 percent in 2011. It is projected to climb to 8 percent in 2013 when major repayments are due, before stabilizing at below 7 percent thereafter. These medium term pressures on the budget clearly indicate some risks as the country is gearing for a fiscal consolidation coming out the crisis. Figure 2.1: Public Debt Burden and Liquidity Indicators under Baseline and Alternative Scenarios 35 80 Public Sector Debt Service to Revenue, % 30 Public Sector Debt to GDP, % 70 25 60 20 50 15 40 10 30 20 5 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 0 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 Baseline Growth sock Depreciation shock Baseline Growth sock Depreciation shock Source: IMF and World Bank staffs calculations. 9 Sensitivity of the Baseline Scenario to Shocks 2.7. Armenia’s public debt outlook would be most adversely affected by another major growth shock or another major depreciation shock. Another major contraction of the economy (-7 percent) in the near future would push Armenia‘s debt to GDP ratio to an unsustainable path around 70 percent and induce an increasing erosion of the fiscal space over time (Figure 2.1). In net present value terms, the public debt to GDP ratio will continuously increase from 46 percent in 2011 to reach 52 percent by 2030. Another major depreciation of the Dram (30 percent) in the near future would dramatically increase the Dram-denominated value of the external debt, and if not accompanied by a strong export supply response, would push Armenia‘s debt to GDP ratio close to 70 percent before it stabilizes around 32 percent by 2030. Such shock would push the net present value of debt to 54 percent in 2011 before slowly bringing it back to 25 percent by 2030. 2.8. These adverse scenarios also worsen the liquidity situation of the country over the medium term. Under the growth shock, the debt service to revenue ratio is likely to increase from 12 percent in 2011 to 27 percent in 2013 before stabilizing between 14 and 15 percent over the medium to long term (Figure 2.2). Under the depreciation shock, the debt service to revenue ratio is projected to increase from 12 percent in 2011 to 33 percent in 2013 and stabilizes below 15 percent by 2022 (Figure 2.2). 2.9. The recent run up in Armenia’s public debt calls for initiating fiscal consolidation over the medium term. Revenues fell and public expenditures rose in the economic crisis for very good reasons. Yet the rise in the debt stock during the crisis requires compensatory policy afterwards. These developments are driven by the public external debt stock, which now accounts for 95 percent of total public debt. Public external debt was less than 14 percent of GDP at end-2008, but reached 36 percent at end-2009, and is expected to peak at 38 percent in 2012, before gradually declining over the medium term. Under the baseline scenario, the projected debt to GDP ratio does not appear excessive, but this stable path is very sensitive to external (depreciation shock) and domestic (another major contraction) shocks. Furthermore, rollovers requirements are high, mainly due to the short maturity of some bilateral and multilateral loans with repurchase obligations during 2012–14. The Government should therefore aim to reduce its debt level promptly, to restore fiscal headroom, particularly given the low growth environment and persistent external volatility the country is heading to in the medium term. B. IS ARMENIA FISCAL STANCE SUSTAINABLE? 2.10. The fiscal consolidation assumed by the Government 2012-14 MTEF is in line with the one assumed under the DSA, although the government relies more on expenditure to reduce deficit while the DSA relies more on revenues. As can be seen in Table 2.1, the DSA assumes a 3.7 percentage points of GDP cuts of the primary deficit over 2011-13, compared with 3.5 percentage points cut assumed by the 2012-2014 MTEF. The difference is coming from less optimistic revenue collection and more expenditure cuts assumed by the MTEF than by the DSA. Table 2.1: Fiscal Frameworks Assumed under DSA and MTEF Actual DSA MTEF in share of GDP 2010 2011 2012 2013 2011 2012 2013 Total revenue 22.5 22.7 21.8 22.2 22.3 21.2 21.4 Total expenditure 28.2 26.6 24.7 24.6 26.2 24.3 23.8 Interest payment 0.9 1.3 1.3 1.3 1.1 1.1 1.1 Overall balance -5.6 -3.9 -2.9 -2.4 -3.9 -3.1 -2.4 Primary balance -4.8 -2.6 -1.6 -1.1 -2.8 -2.0 -1.3 Sources: Ministry of Finance; IMF and World Bank Staff. 10 2.11. The Government’s lifetime budget constraint can be used to derive a rigorous fiscal sustainability assessment. The Government finances its initial debt by raising seigniorage revenue and running primary surpluses in the future, whose present value is equal to its initial debt obligations. Starting from the government lifetime budget constraint, Burnside (2005) derives a long-term fiscal sustainability indicator that determines the size of the primary balance the government would need to run in order to keep its debt stock constant as a fraction of GDP.3 Using the projected 2010 debt ratio (42 percent) as initial debt ratio and setting seigniorage at its average level over 2000-2010 (3 percent), Table 2.2 simulates long-term fiscal sustainability indicators using various growth and interest assumptions. Table 2.2: Primary Balances to Keep Debt at 42 percent Real interest rate, % Real growth, % 0.00 1.00 2.00 3.00 4.00 5.00 2.00 -3.86 -3.43 -3.00 -2.57 -2.14 -1.71 3.00 -4.28 -3.85 -3.43 -3.00 -2.57 -2.15 4.00 -4.69 -4.27 -3.85 -3.42 -3.00 -2.58 5.00 -5.10 -4.68 -4.26 -3.84 -3.42 -3.00 Table 2.3: Primary Balances to Reduce Debt to 30% in 10 Years Real interest rate, % Real growth, % 0.00 1.00 2.00 3.00 4.00 5.00 2.00 -2.51 -2.16 -1.80 -1.44 -1.08 -0.72 3.00 -2.86 -2.51 -2.15 -1.80 -1.44 -1.09 4.00 -3.19 -2.85 -2.50 -2.15 -1.80 -1.45 5.00 -3.52 -3.18 -2.84 -2.49 -2.15 -1.80 Source: World Bank Staff calculations. 2.12. With a real interest at 3 percent and debt level stable at 42 percent, Armenia’s fiscal stance would be sustainable only if long-term growth is at least 4 percent and the primary deficit is contained below 3.4 percent. Given the currently high concessional level of Armenia debt, the realized real interest over the last 10 years is estimated at 0.1 percent with a 1.3 percent standard deviation. However, the country is scheduled to graduate from IDA during the next IDA round and may have to increasingly rely on commercial markets for its financing needs. If a 3 percent real interest is assumed as the long term financing cost Armenia will face, the baseline macroeconomic framework suggest that Armenia needs to run at most a primary deficit of 3.4 percent to keep its debt-to-GDP ratio stable at 42 percent as in 2010. However, the macroeconomic framework also assumes a path for the country indebtedness: by the 2020, the debt-to-GDP should reach 30 percent. The fiscal sustainability indicator can be revised to accommodate that debt target.4 2.13. If the Government aims to decrease the debt-to-GDP ratio from 42 to 30 percent in 10 years time, it will need to run at most a primary deficit of about 2 percent. If the Government targets a 30 percent debt to GDP ratio in 10 years (as assumed in the baseline scenario of the DSA), it will need to run at most a 2.15 percent primary deficit, hence the need for a tougher fiscal consolidation than what was planned under the 2011-2013 MTEF. The 2012-2014 MTEF is in line with the necessary overall deficit 3 Sustainable primary balance=(real interest-real growth)/(1+real growth)*debt ratio –seignior age. See theoretical background in the Annexes for an elaboration on a more advanced tool, the Fiscal Sustainability analysis tool, that builds on this framework. This model could be developed at the request of the Government. 4 Sustainable primary balance=(real interest-real growth)/(1+real growth)*[f(n)*debt ratio –g(n)*targeted debt ratio]-seignior age, where n is the number of years to reach the target. 11 reduction, although reliance on expenditure rationalization rather than revenue increase may jeopardize the fragile economic recovery. This study proposes a fiscal consolidation relying more on improving tax collection performance. 2.14. The debt and fiscal sustainability analyses show that Armenia’s debt dynamics are very sensitive to economic performance. Various alternative scenarios and stress tests undertaken have shown that variables like real GDP growth, exchange rate movements and primary balance can have a significant impact on the country‘s level of indebtedness over the medium to long term. The magnitude at which the low level of economic growth and exchange rate depreciation might affect the debt profile emphasizes the importance of developing a new model of economic growth for Armenia. Changing the model of economic growth to export-driven might help overcome the above-mentioned constraints. But achieving such policy changes will require a strong commitment of the Government to address entry barriers concerns as well as the perceived negative role of oligarchs on the business environment. 12 CHAPTER 3. REVENUE POTENTIAL Tax productivity in Armenia is below countries with similar tax rates and income levels. This Chapter contributes to the existing literature and policy agenda by: (1) estimating CIT, PIT and VAT compliance difference among the largest 1000 tax payers; (2) assessing actual and statutory compliance for custom duties; (3) quantifying the tax gap due to informality; and (4) quantifying gains from modernizing the mining-specific taxation and excise regimes. It estimates that the revenue could increase by between 2.3 and 5.8 percent of GDP, if a concerted effort to modernize both taxation policies and administration is made. But this estimate could be higher if revenue losses from tax loopholes and under-declaration of taxable income could be estimated. The approach is innovative as it uses publicly available information about the largest 1000 taxpayers. It also provides a tool that the SRC can use to guide audits of large taxpayers. While the Chapter does not attempt to provide a thorough diagnostic of the tax policy or the tax administration systems in Armenia, it does present a summary of revenue potential in key tax loopholes. 3.1. The tax rates applied in Armenia are not unusually lower than regional and income-level peers, but its performance on tax collection is consistently lower. In 2009, the rates for VAT and PIT in Armenia were in line with the ECA average, while the CIT rate was 4 percent higher than the ECA average. Compared with low-middle income countries, Armenia‘s VAT rate is higher by 5 percent, but the CIT and upper PIT rates are about 5 percent lower (Table 3.1). Perhaps more importantly, Armenia‘s VAT tax productivity5 is slightly lower than the average for ECA and low-middle income countries, and the productivity of the CIT and PIT showed even larger gaps relative to peers. This has been a persistent trend in Armenia during the last decade, even when the country was growing at double digit rates (Annex 1). Table 3.1: Tax rates and tax productivity, 2009 VAT CIT PIT rate productivity rate productivity min rate max rate productivity Armenia 20 0.42 20 0.1 10 20 0.08 ECA average 19 0.5 16 0.2 13 19 0.2 Low-middle income 15 0.5 25 0.2 11 27 0.1 countries‘ average Source: Ministry of Finance and USAID (2009). 3.2. To better address this underperformance on revenue collection, it is important to identify the sources of tax revenue leakages. How much can be achieved within the current tax policy framework (e.g., via better administration), and how much by further modernizing tax policy? A number of tax studies for Armenia (Annex 1) have pointed out at tax evasion as the chief cause of low collections in the country: by large taxpayers, by importers, and by firms not reporting or under-reporting their tax liabilities, e.g., in the informal economy. If that is the case, tax administration efforts alone (or with minor amendments to close loopholes) should yield large payoffs in collections. However, many policy issues need to be addressed almost in every tax, so in the end a concerted effort is needed by Armenian policy- 5 Tax productivity is defined as tax collection as a share of GDP divided by the statutory tax rate. The PIT, the weighted average rate is used. 13 makers and the society at large to improve both tax policy and tax administration. The rest of the Chapter focuses on two aspects:  Section A provides estimates of the revenue collection losses due to uneven compliance in CIT, PIT and VAT payments among the largest 1000 tax payers, as well as the yield of bringing the average collected customs duties to the level of the statutory duties. It also estimates the revenue potential from broadening tax bases to ensure proper coverage of the informal sector covered under the presumptive regime.  Section B provides estimates of revenue collection gains from improving the taxation policy and administration regimes for the mining sector and for excisable products. The section also summarizes key loopholes in the tax policy system, but does not provide estimate of revenue potential by closing these loopholes. 3.3. Our estimates show the Armenian authorities could increase the tax collection. Our assessment (Table 3.2) is consistent with those of other authors: Armenian authorities could increase collection by a lower-bound interval of between 2.3 and 5.8 percent of GDP. Indeed, Davoodi et. al. (2007) estimated, and with different methodology, that the tax potential was about 6.5 percent of GDP. Estimates by these and other authors are summarized in Annex 1. Table 3.2: Estimated Gains from Improving Tax Policy and Administration (% of GDP) Total revenue collection gains from all areas, of which 2.3 to 5.8 1. Subtotal losses from tax administration weaknesses, of which 1.4 to 2.6 - Losses from uneven compliance by the 1000s largest tax payers* 0.4 to 1.0 - Losses from the effective customs duties being lower than the actual 0.0 to 0.3 - Losses from informality 1.0 to 1.3 2. Subtotal gains from selected improvements in tax policy, of which 0.9 to 3.2 - Gains from improving policy and administration for the mining sector 0.4 to 0.6 - Gains from improving policy and administration for excisable goods 0.5 to 2.6 Source: World Bank staff estimates *If we were to assume that these enterprises under-report tax bases by 20 percent, then the additional revenue losses would be between 0.9 and 1.6 percent of GDP (see paragraph 40). This estimate is not included in this table as it might overlap with those on ―losses from informality‖. A. REVENUE POTENTIAL FROM TAX ADMINISTRATION 3.4. A number of studies have pointed out the weaknesses of the Tax Administration in Armenia. Diagnostic covered issues of organization and staffing, integrity, large tax payers unit, tax payer services and education, and processes such as audits, collections, appeals—in addition to IT systems. Box 3.1 summarizes the findings of one such study and the efforts made since then to address weaknesses. 3.5. Efforts to address SRC’s weaknesses have been piecemeal and to a large extent donor- driven, although by now a large-tax payer unit is well established, together with other elements of a reform package (e.g., tax payer service centers). Under the Doing Business reform agenda, a number of 14 reporting actions previously expected from firms have been eliminated, and electronic filing for companies has increased significantly since its inception in August 2010 (about 20 percent of the tax payers file electronically). SRC (with USAID support) is in the process of introducing risk-based systems for audit selection. However, the perception of uneven collection efforts is still there among tax payers and the population at large, suggesting a significant pending agenda. The failure of recent efforts to improve perceptions can be attributed to the fact that these have been piecemeal and often driven by donor requests rather than domestic conviction (beyond technicians in the government who understand well the importance and magnitude of the agenda). 3.6. Country ownership over the agenda of improving tax policy and administration has nonetheless increased significantly over the past months. Last year the office of the Prime Minister paid for a thorough assessment of 15 business processes in the tax administration.6 Moreover, the President openly highlighted during a meeting at the Chamber of Commerce and Industry that ―unpaid taxes are money stolen from health and education‖. It is significant to note that revenue mobilization is a Presidential priority. Presidential awareness and leadership during the hyperinflation crisis in Latin America in the late 80s and the 90s was not present when the crisis exploded. It took precious time to make Presidents aware of the need for their involvement as fiscal matters were considered the responsibility of the Ministries of Economy and Finance and/or the Tax Authorities. 3.7. The agenda of modernizing skills, processes, and systems for tax administration demand a sustained commitment and dedicated effort from the authorities. Tax compliance costs, a key instrument to assess efficiency in a country‘s tax collection, are still high in Armenia and confirm the importance of the agenda of modernizing tax administration. According to the latest Cost Compliance Survey conducted by the International Finance Corporation, tax compliance cost by an average firm in Armenia was estimated in 2009 at 400 person-hours (50 business days) in terms of total time spent on tax compliance. Using the average wage level, this corresponded to US$1,363. These results have triggered a series of reforms that are in force since January 2011 (the introduction of system for issuing electronic tax invoices, reducing the frequency of payment of advance profit tax and minimum tax amounts, reduced frequency of filing returns and payments on property tax, Amendments to the Law on Organizing and Conducting Inspections in the Republic of Armenia) that are certainly reducing compliance costs.7 3.8. There is no methodology to assess the opportunity costs to society of an inefficient tax collection system. While the Bank is in the process of developing an integrated assessment model for tax administration (IAMTAX), there is still a need for providing diagnostic tools for key processes. For this reason this chapter builds on the literature of revenue potential8 applied already to the case of Armenia by many others. Our key innovation is the application of regression analysis to the information available on the 1000 largest tax payers in order to estimate unevenness in their compliance for VAT and CIT, as well as in the compliance of their workers for PIT. To complement this analysis, we also assess revenue shortfalls due to lower than statutory effective rates for import duties, and from informality. However, we do not estimate the revenue potential from tax loopholes and under-declaration of taxable income. 6 The business process reengineering was conducted with support of KPMG. The SRC has translated the findings into an Action Plan that is to be supported by USAID, the World Bank, and other donors. 7 Note however that the modernization of tax administration may not be enough since tax compliance costs may also be impacted by factors beyond SRC mandate such as the clarity/quality of tax legislation, tax literacy and awareness of taxpayers, or simply tax culture. 8 See for instance Le, Moreno-Dotson, and Rojchaichaninthorn (2008). 15 Box 3.1: Armenia – Tax Administration Diagnostics and Recent Reforms Implemented by the State Revenue Committee Diagnostics The World Bank Choices in Development Policy report identified the promotion of voluntary compliance as the main strategy to improve the tax-to-GDP ratio in Armenia. In addition to building trust and partnership with taxpayers, the State Revenue Committee can use its processes and technological tools to detect fraud and punish fraudulent behavior. Armenia has anchored its reform efforts on such principles, and has focused efforts on VAT administration, implementing self-assessment systems and strengthening the Large Taxpayer Unit (LTU). The report suggested focusing on the following reforms: (i) Organization and Management, (ii) Operational Efficiency, (iii) Taxpayer Services, and (iv) Technology Upgrading.  Organization and Management should focus on organizational development, training and anti-corruption systems, including the design of surveys to measure the extent of corruption and identify areas with the gravest possibilities of abuse.  Operational Efficiency should focus on efficiency and effectiveness in collection, audit, centralized return processing, adopting new streamlined business processes, and strengthening large taxpayer auditing capacity.  Taxpayer Services would include: (i) establishing a public forum to ensure taxpayer feedback; (ii) developing a central automated call center to provide information and services to taxpayers; (iii) designing and conducting periodic taxpayer satisfaction surveys to assess progress; and (iv) conducting periodic analysis for assessing taxpayer compliance costs.  Technology Upgrading would focus on upgrading existing facilities and information and communication systems, as well as the establishment of specialized tax training centers. Such upgrading would include the hardware and software for a centralized processing unit and a document management system. Recently introduced measures by the SRC include:  Introduction of the ―one-stop shop‖ principle for business registration and registration, in place since April 2011.  Reduction in number and optimization of tax returns – Businesses no longer submit financial returns, quarterly returns on mandatory fees for public services regulation, annual returns, and information on base data for road fee to the SRC. The SRC set a unified deadline for submitting all calculations and returns. Returns on social contribution and calculations for excise tax are to be submitted quarterly.  Reduction of periodicity of tax payments – Advances on profit tax and payments of minimal profit tax will be done on quarterly basis (previously, 12 times per year).  Introduction of a new system of patent payments – According to the new law On Patent Fees, small businesses (providing services to population specified by the Law) with annual turnover not exceeding 6 mln AMD are allowed for choosing patent payment as a tax regime and not submit any return to the SRC.  Introduction of electronic system for VAT invoices – The businesses implementing this system are not required to maintain books of shipment.  E-filing – Currently 27 out of 40 tax returns can be submitted electronically.  Introduction of a risk-based audit system – The SRC revised qualification criteria large taxpayers and expanded the list of such companies (440 for 2011). Not risky taxpayers are eligible for VAT/excise tax refund without audit.  Changing public attitude towards the SRC – To change the image of the SRC from a tax collector to a service provider the SRC established three Tax Payers‘ Service Centers in Yerevan and Vanadzor. Three more centers and the call center to become operational in 2011.  Simplifying Customs administration – The major changes in this area relate to further upgrade of the TWM system, clearer rules for using a contract price to define the customs value, etc. Sources: World Bank, 2008. 16 Quantifying Tax Leakages from Large Taxpayers 3.9. Tax payments reported on Figure 3.1: Audit Outcome, 2005-2009 Armenia’s one thousand largest taxpayers indicate that the first 500 contribute ten times more than the Audits with positive outcome second. This suggests that focusing on 45 40 the largest 500 taxpayers is enough to share of amount found (%) 40 35 assess revenue leakages from large 35 30 taxpayers. This assessment can be done 30 25 bln. AMD by comparing the implicit tax rates of 25 20 20 different subgroups among this first 15 15 half of the largest tax payers. This 10 10 would help identifying the type of 5 5 taxes and or tax-payers that need a 0 0 2005 2006 2007 2008 2009 closer monitoring by the Large Taxpayers Unit to ensure higher Amount found (RHS) of which PIT revenue collections. of which CIT of which VAT 3.10. The SRC is currently developing computer-supported Source: SRC and World Bank Staff calculations. comparisons and crosschecking as the basis for allocating its audit resources. The initial comparisons are internal consistency checks of tax returns regarding their declared turnover, imports and other data. Tax authorities are building taxpayers‘ profiles looking for evasion risk. The initiative is succeeding since it has led to an increasing number of successful audits, collecting an additional 35 billion AMD (US$92 million) in 2008 (Figure 3.1). The largest share of these revenues came respectively from the VAT, CIT, and PIT. 3.11. Our analysis provides further statistical evidence that complements the risk-based algorithms being developed by the SRC. The underlying idea of the analysis is to search for possible structural breaks that may exist above or below a certain level of turnover, and allow us to separate the sample of firms analyzed between taxpayers for whom the correlation between the tax base and the tax payment is higher than for other types of taxpayers (Box 3.2). For easy reference, we call ―optimal taxpayers‖ the group for which the correlation between the tax base and the tax payment is higher, and the other group is referred to as ―sub-optimal taxpayers‖. The estimated coefficients on the ―optimal taxpayers‖ group can then be used to simulate the potential level of tax payment by ―sub-optimal taxpayers‖ group. Finally, these firm-level simulations are aggregated to get an estimation of the tax revenue potential. 3.12. Differentiating between “optimal� and “suboptimal� taxpayers helps to estimate additional revenue from large taxpayers between 0.43 to 1.02 percent of GDP. The estimated coefficients (see Annex 2) indicate that taxpayers with turnover between 1 and 10 billion AMD (between US$2.6-26 million) have an estimated effective tax rate for CIT and PIT lower than that of firms with turnover higher than 10 billion AMD. Conversely, taxpayers with turnover higher than 10 billion AMD (US$26 million) appear to have an estimated effective tax rate for VAT lower than that of firms with turnover between 1 and 10 billion AMD (between US$2.6-26 million). Replacing the estimated coefficients on firms depicting a lower effective tax rates for CIT, PIT and VAT by the confidence interval of the coefficients estimated on the group of ―optimal‖ taxpayers shows on average potential additional tax collection between 0.43 and 1.02 percent of GDP (Table 3.3). 17 Box 3.2: Regression Analysis to Estimate Tax Leakages From the database of 1,000 large taxpayers handled by the Large Taxpayer Unit, we have been able to extract a panel of 457 firms present in the database from 2006 to 2009. More than two third of these firms are ranked among the 500 largest taxpayers. The database contains information on VAT, CIT and PIT payments as well as some tax base information such as profits before tax, social contributions and domestic sales. Although the PIT is, strictly speaking, a liability of individuals, Armenia has not introduce the practice of filing by individuals and thus the liability on labor income is withheld by firms and paid to the SCR on the behalf of their employees. In many ways, the PIT in Armenia is similar to the social security contribution, at least in the manner it is assessed and collected for labor income in the formal economy. We therefore included the PIT because firm‘s might underreport the wage bill to avoid social security contributions and those can be capture in the PIT data. Also because this is a way to measure the compliance of a firm‘s worker with PIT, which is well captured by firm data given that withholding rather than self assessment is used in Armenia (a practice that resembles that in various European countries that differentiate taxation from labor and capital incomes). This balanced panel is used to run regressions of firms‘ tax payments (CIT, PIT and VAT) on proxy tax bases (―profit before tax‖ for CIT, ―social contributions‖ for PIT and ―domestic sales‖ for VAT) and a set of control variables (total assets, number of employees, location, sector of activity). To ensure the robustness of the estimations, we control for multicolinearity by calculating the Variance Inflation Factors (VIF). Since the ―social contributions‖ variable appears to be strongly correlated with the variable ―number of employees‖, this later variable is excluded from the estimation of the PIT equations. To check for consistency of the estimated coefficients, two specifications are used: a panel regression with random effects and a truncated panel regression to control for any selection bias of firms paying zero CIT, PIT or VAT taxes. Finally, taxpayers‘ turnover is used to form three well-identified groups of firms: turnover below 1 billion AMD, turnover between 1 and 10 billion AMD (between US$2.6-26 million) and turnover above 10 billion AMD (Table below). Table: Distribution of Taxpayers by Turnover, 2006-2009 Number of firms Below 1 bln Between 1 and Above 1 bln Year AMD 10 bln AMD AMD Total 2006 262 170 25 457 2007 213 219 25 457 2008 201 222 34 457 2009 193 231 33 457 Source: SRC and World Bank Staff. Our methodology is subject to various caveats. Firstly, the tax base proxy used for the PIT and VAT are imperfect as the file amounts are not publicly available. Indeed, using social contributions as the proxy base for income tax, and domestic sales as the proxy base for VAT can be criticized on many grounds. However, the actual tax bases are not available and the proxy bases are a transformation of these (if not necessarily linear) and, thus, can be interpreted in a straightforward manner with respect to the results. Secondly, the cutoff turnover points are somewhat arbitrary although various ranges were tried and we conducted robustness checks to ensure that the estimated coefficients were stable over different cutoff points. Thirdly, the PIT in Armenia has two bands (10 percent for incomes below 200$ per month and 20 percent for incomes above) which may explain the lower coefficient estimated on firms with a turnover between 1 and 10 billion AMD. Given that the database does not allow sorting out incomes collected from each of these two bands, the result on the PIT revenue leakage should therefore be considered with caution (though the over estimation is likely to be small since the 10 percent rate applies to very low income levels). Finally, there could be firm- or industry-specific conditions that make some companies structurally pay more taxes than others (e.g., the mining sector might have special depreciation allowances compared to other sectors), and thus, the assumption of equal rates of compliance for every firm over the four years examined might be strong. 18 3.13. The biggest payoff appears to come from the improvement in VAT collections, followed by CIT and then PIT collections. On average, as indicated in Table 3.3, additional VAT collections of up to 0.65 percent of GDP can be expected, compared to additional CIT collections of up to 0.23 percent of GDP and additional PIT collections of up to 0.13 percent of GDP. The simulations indicate that taxpayers with turnover higher than 10 billion AMD are paying between 4 and 15 percent less of VAT than firms with turnover between 1 and 10 billion AMD, while these latter taxpayers are paying between 10 and 14 percent less of CIT than firms with turnover higher than 10 billion AMD. Table 3.3: Estimated Additional Tax Potential from Large Taxpayers CIT PIT VAT Sum CIT PIT VAT Sum in billion AMD as a share of GDP (%) Tax collected 46.8 14.7 113.8 175.3 1.76 0.55 4.28 6.60 2006 Estimated additional tax, lower bound 2.2 1.3 5.2 8.7 0.08 0.05 0.20 0.33 Estimated additional tax, upper bound 3.2 3.3 17.2 23.7 0.12 0.13 0.65 0.89 Tax collected 50.6 19.2 150.8 220.5 1.61 0.61 4.79 7.00 2007 Estimated additional tax, lower bound 6.4 2.8 6.2 15.5 0.20 0.09 0.20 0.49 Estimated additional tax, upper bound 9.4 3.8 22.2 35.5 0.30 0.12 0.70 1.13 Tax collected 57.5 22.5 161.6 241.5 1.58 0.62 4.43 6.62 2008 Estimated additional tax, lower bound 6.5 2.5 7.4 16.5 0.18 0.07 0.20 0.45 Estimated additional tax, upper bound 8.5 4.5 24.4 37.5 0.23 0.12 0.67 1.03 Tax collected 49.4 22.1 131.5 202.9 1.59 0.71 4.24 6.54 2009 Estimated additional tax, lower bound 5.6 2.9 5.5 14.1 0.18 0.09 0.18 0.45 Estimated additional tax, upper bound 8.6 4.9 18.5 32.1 0.28 0.16 0.60 1.03 Tax collected 51.1 19.6 139.4 210.1 1.63 0.62 4.44 6.69 Average Estimated additional tax, lower bound 4.9 2.4 5.6 12.9 0.16 0.08 0.19 0.43 Estimated additional tax, upper bound 6.9 4.4 20.6 31.9 0.23 0.13 0.65 1.02 Source: World Bank Staff calculations. 3.14. Our estimates on the revenue leakage may not necessarily be the result of tax evasion. In fact, given the numerous loopholes that exist in almost every tax regime in Armenia, we might be capturing the impact of tax expenditures or concessions that are perfectly legal, or the exploitation of tax loopholes (that could be considered tax evasion, and might or not be illegal). Whether this is the case or not something to be answered through selective audits. More importantly, a rigorous inclusion of tax evasion in our analysis would require estimates on whether all firms in our sample tend to underestimate profits, sales, and wages. Given that the tax evasion has been documented by many studies, it is most likely that the estimated leakages are mainly coming from tax evasion, be it legal through underreporting of tax base or simply underpayments. If we assume that the tax bases used for the estimation were 20 percent higher than reported and that all firms comply at the rate of ―optimal firms‖, the additional revenue to be expected is between 0.9 and 1.6 percent of GDP. 3.15. The simulations suggest that SRC should focus its VAT audits on firms with turnover higher than 10 billion AMD, and focus its CIT and PIT audits on firms with turnover between 1 and 10 billion AMD. This recommendation is in line with the risk-based audit system recently introduced by the SRC. The regression analysis needs to be periodically revised to assess any differential tax payment between various groups of taxpayers, and adjust the audit objectives accordingly. The regression analysis would also be improved by using actual tax bases for VAT and PIT instead of the proxy tax bases used in our analysis. 19 Quantifying Lower than Expected Compliance with Customs Duties 3.16. Two fundamental principles guided reform in customs over the recent past: (i) reduction of interaction between importers and customs officials; and (ii) shift to risk-management. Efforts focused on the development of direct trader input systems that required importers or their agents to directly enter data into the customs system and obtain a validation without the assistance of customs officials and the institution of risk management throughout customs practices. Moreover, the grossly excessive use of reference prices as opposed to invoice prices for valuing imports, contrary to the rules of WTO, created opportunities for discretion by officials, and led to a vicious circle of under-reporting of prices (knowing that they would be adjusted upwards in most cases) and forging of invoices. 3.17. These reforms are complementing Armenia’s relatively liberal trade regime since the early 2000s. Armenia joined the WTO in 2003 and adopted a two-band tariff regime (0 and 10). The import- weighted average applied tariff (including preferences) was 2.3 percent during 2005-2009. With such low applied tariff level, not much revenue collection is to be expected from Custom duties, although we can still assess the revenue potential from Custom duties by estimating the difference with the effective taxation rate of imports. 3.18. Armenia’s customs duties to GDP were on average 0.8 percent over 2005-2009. This amount was collected on imports fluctuating between 39 and 43 percent of GDP over the same period. This implies an effective import tax rate between 1.9 and 2.2 percent, and represents between 0.1 and 0.4 percent point less than the import-weighted applied tariffs including preferences. 3.19. Between 0 and 0.3 percent of GDP additional revenue can still be expected from custom duties. Indeed, if the effective import tax rate is 2.3 percent with an import to GDP ratio between 36 and 43 percent, customs duties to GDP should be between 0.8 and 1.1 percent. Therefore, deducting the current 0.8 percent of GDP collected on customs duties gives a rough estimation of revenue leakage.9 3.20. This relatively low expected additional revenue from custom duties reemphasizes the fact that trade and transport facilitation has to be the main focus of the Customs Administration. The Customs Administration is currently implementing a fully functioning ―green channel by default‖ release as one of the key DPO-supported policy reform. While the Bank has recommended in recent years a bottom-up approach to customs reforms, which at the time was the only way to improve customs operations (introduction of direct trade input to minimize interaction between customs‘ officials and importers, gradual reduction in the number of repetitive and non-productive physical examinations, development of a post-release control capacity), a few new instruments have been introduced which now allow some step forward (such as a green channel release by default). Recent developments indicate that there are now opportunities for a holistic approach to customs reform. The steps would be to (i) develop a real strategic approach to a government-supported integrated border management and control policy (extending beyond the purely customs domain and working with respective structures of a neighboring country), (ii) consolidate recent progress under an objective-driven action plan, and (iii) moving towards international best practice standards. Quantifying Revenue Leakage from Informality 3.21. A major value in modern tax administration is to protect the tax base. Is Armenia really protecting its tax base? The answer emerging from this study is that the tax policy and administration authorities could protect the base much better by expanding their control and redefining various self- imposed limits. Firms currently can evade taxation by under-reporting or by shifting activities to the 9 If firms avoiding customs duties also skip the VAT on imports, the revenue impact of our estimates becomes significant. 20 shadow economy to be beyond the control of the tax administration authorities. For instance, integrated agro-industries could shift their profits to their agricultural business and avoid profit tax that way. Firms in excisable products could also break their sale points in small units that fall below the VAT thresholds and thus avoid VAT payments. 3.22. Ensuring that tax policy and administration does not incentivize moving to the informal economy would not only be revenue efficient, but would also deliver more equitable and comprehensive tax administration. While tax policy must be simpler and cover all sectors, the tax administration efforts should not only focus on compliers. Firms in Armenia might have already learned to play with the narrow policy and administrative focus, thus enlarging the informal economy and shrinking the ―captive‖ base. 3.23. Research on the informal economy in Armenia suggests that it is significant. The formal National Accounts measured that 17 percent of the GDP in Armenia comes from agriculture, while 52 and 31 percent comes from industry and services, respectively. The informal sector‘s average contribution to these 3 broad sectors has been estimated at 22 percent for agriculture, 10 percent for industry and 9 percent for services. (These are the results of aggregating the estimations presented in Figure 3.2). This suggest that the GDP in Armenia is about 11 percent higher than it is currently reported in the statistics— clearly a conservative estimates of under-reporting sales, profits, income and wages. These estimates are conservative. For instance, Schneider (2002) estimates that Armenia‘s informal economy is about 48 percent of the reported GDP in 2004-05, which is significantly higher than that Albania (34 percent) and Bosnia and Herzegovina (35 percent) but lower than that of Ukraine (55 percent), Georgia (66 percent), Azerbaijan (59 percent) and Georgia (66 percent). 3.24. We can use the estimates of the informal economy to derive further revenue losses. Given that the agriculture sector is by and large tax-exempted, this implies that the contribution of the informal sector to Armenia taxable GDP is about 8 percent. A back of envelope calculation using the 20 percent average statutory tax rate of the regular tax regimes (VAT, CIT and PIT), generates an estimate for foregone tax revenue of 1.6 percent of GDP (Table 3.4). 3.25. Between 1 and 1.3 percent point of GDP could be expected by adding the non-agricultural firms of the informal economy to the formal tax base. If we assume that the presumptive and simplified tax payments are entirely coming from the informal sector, its average tax burden over 2005- 2009 is estimated between 0.3 and 0.6 percent of GDP. Subtracting this from the expected tax collection of 1.6 percent of GDP suggests that additional revenues equal to between 1 and 1.3 percent point of GDP could be expected if tax administration were to extend throughout what now is the non-agricultural shadow economy. This expansion of the tax base might come at an additional administrative cost, but the gains from preventing tax evasion through the informal economy would much higher. 21 Figure 3.2: Share of Armenia Informal Sector to Total Gross Value Added, by Sectors Agriculture, hunting, forestry 22.35 and fishing Other service activities 16.62 Construction 15.43 Wholesale and retail trade; 14.82 repair Real estate, renting 8.74 Education 6.95 Health and social work 4.94 Manufacturing 4.27 Transport and 3.70 communication 0 5 10 15 20 25 Source: ADB TA ―Measuring the Contribution of Informal Sector to Total Economy in Armenia‖, 2010. Table 3.4: Estimated Additional Tax Potential from the Non-Agricultural Informal Sector Contribution to Estimated Taxable Current tax Tax burden Revenue GDP Informality Production burden on on informal loss informal sector if 20% sector statutory rate used As share of GDP Agriculture 17 22 0 0 Industry 52 10 5.2 1.04 Services 31 9 2.8 0.56 Total 100 11 8 0.3 – 0.6 1.6 1.0 to 1.3 Source: WB staff estimates based on ADB-supported study. 3.26. Recapitulating, we estimated that better policy and administration over the basis could yield between 1.4 and 2.6 percent of GDP in new revenues. Harmonizing compliance rates across large taxpayers for VAT, PIT, and CIT could provide between 0.4 and 1 percent point of GDP in revenues. Bringing customs implicit duties to statutory levels is estimated to bring between 0 and 0.3 percent of GDP in new revenues and ensuring broad focus by the tax administration to cover the informal sector would provide between 1 and 1.3 percent point of additional revenues. 3.27. Addressing the formalization of firms operating in the informal sector appears to be the first tax policy and administration measure to pursue. Note that the unreported revenue or tax base could come from companies: 22  Entirely operating outside the formal sector (e.g., fully informal, even illegal)  Partly operating in the formal sector and partly underreporting  Fully operating in the formal sector, but exploiting major gaps in tax policy legislation (e.g., lack of transfer pricing provisions for agriculture/food industries or mining industries) 3.28. It should be stressed that our estimates do not double count evasion. Our estimates of unequal compliance rates among the 1000 largest tax payment is based strictly on what they report. If some of these firms underreport their imports, sales, profits or their wage-bill, we should have capture part of that in our estimates for tariffs duties and out estimates on informality. In all cases, we do not capture the universe of lack of compliance and, from that perspective, our interval estimates should be taken as a lower bound of the potential losses for the country and for society. B. REVENUE POTENTIAL FROM TAX POLICY 3.29. Various reports have conducted a tax by tax analysis and identified tax policy gaps in Armenia. The Bank‘s Choices in Development report (2008) suggested focusing primarily on improving the VAT registration system so that all taxable businesses are registered and books well kept. The reliance on presumptive and simplified tax regimes (e.g., for excisable products) was found to be fostering informality and plain evasion. More recently, IMF (2011) pointed out a number of specific tax issues. For the VAT, it found the design sound, but its revenue potential is low because of the shrinkage of the tax base by excluding a large number of potential taxpayers such as those below the VAT threshold or covered by presumptive regimes. Taxpayers who stopped filing upon introduction of the new threshold used to contribute a large amount of revenue. Furthermore, they estimated that excise revenues and productivity were low in comparison to the yield nine years ago. In this section, we complement these two studies (and the mining taxation regime that the authorities are carrying out under the auspices of the Bank‘s Development Policy Operation) by focusing on the assessment of the revenue potential from improving the tax policy for the mining sector, excisable goods. The section concludes by enumerating a number of tax loopholes that might well be exploited extensively by small and large firms alike. Quantifying the Revenue Potential from the Recently Proposed Mining Taxation 3.30. The mining GDP and exports reached almost US$500 million in 2007. Armenia mining GDP has been consistently growing from 2003 to 2010, except during the crisis in 2008 and 2009 (Figure 3.3). In 2010, mining GDP grew by 16 percent, while mining exports reached 60 percent of the total exports. Furthermore, comparing unit prices to international prices, Armenia appears to be getting higher prices, compared to international standards, in copper and zinc (due to the fact that concentrates contain poly- metals). Furthermore, discussions with businessmen in the mining sector revealed that important mines are still to come on stream. Unfortunately, what does not seem to be keeping up the pace is the taxes paid by the metallic mineral industry. This is so, even after discounting the effect of the devolution of non- refunded VAT payments—a backlog fully addressed during 2010. 23 Figure 3.3: Economic Contribution of Mining and Metallurgy 900 800 700 600 500 400 300 200 100 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 Value added of Mining (mln. USD) Output of mining and metallurgy (mln. USD) Exports (mln. USD) Taxes paid (mln. USD) Source: Armenian Authorities and World Bank Staff estimates. Note: Data on tax revenues for the mining sector is not available for 2002 through 2004. 3.31. Armenia has a two-tier royalty regime. In addition the profit and other taxes noted above, the mining sector in Armenia is currently taxed by two additional mining-specific tax provisions: a natural resource use fee (NRUF) under the December 1998 Law on Nature Protection and Nature Utilization Payments, and a mining royalty and mining concession fee under the November 2002 Concession Law, amended in 2006. The NRUF is administered by the SRC as a fixed-rate fee levied on the estimated value of the in situ geological estimate of the ore that has been extracted and evaluated at a reference price. The rate is specific to each mineral (e.g. 1.3 percent for copper and molybdenum and 1.5 percent for gold and silver) as defined in a government decree. The royalty regime under the Concession Law is administered by the Concession Agency and calculated in the amount of 0.1 percent of the aggregate netback value of mineral sales per each 0.8 percent exceeding the profitability index. Since the base allows some deductions from netback sales,10 and starts at a 25 percent profitability rate, only relatively large metallic mines usually pay this additional royalty. The effective royalty (including both the NRUF and the profit- based royalty) rate has therefore remained at 1 percent for most mining companies, unless metal or mineral prices become quite high. 3.32. The authorities are in the process of modernizing their Mining Code, including the mining specific taxation regime. This work has been supported by the Bank under a Development Policy Operation, and most recently by the IMF under an FAD technical assistance. For taxation, the proposal is to unify the NRUF and the profit-based royalty into a single royalty regime to capture both the welfare losses to the state for the non-renewable resources that are being depleted as well as windfall profit. The administration of these two taxes, currently split between the Concession Agency and the SRC, will be fully transferred to the latter one. A number of loopholes in the profit taxation regime are also to be addressed, including the introduction of transfer pricing rules, effective thin-capitalization rules, rules for transfer of an interest in a mining license, ring-fencing of projects (if surcharge profit tax system were to be introduced)b and tax treatment of hedging /forward transactions. 10 Essentially recurrent cost as financial and capital expenses cannot be deducted. 24 3.33. Between 0.4 and 0.6 percent of GDP additional tax can be expected from the mining sector. During 2006-2008 when the international copper price was at a plateau of $7,000 per ton, Armenia‘s implicit tax rate without VAT was between 11 and 15 percent. In 2010, international copper price was $7,500, for an estimated implicit tax rate of only 6 percent hence a tax rate gap has arisen between 5 and 9 percent. If 2010 mining export is used as the proxy tax base, this represents between 0.4 and 0.6 percent of GDP additional tax to be expected from the mining sector. A more rigorous assessment of the revenue potential from the proposed new mining taxation regime is underway. 3.34. For both the policy and administration of mining specific taxes, the key advice is to develop a well thought-through plan to strengthen capacity for policy formulation (in MoF) and implementation (in SRC). Synergies between these two bodies must also be enhanced—if necessarily via the signing of a formal Memorandum of Understanding which would transcend the synergies, or lack of, of current management structures. It is also important that both the SRC and the MoF develop the analytical capacity to create, and maintain, financial models for each of the existing and potential future mining projects so as to be able to assess the tax liabilities on the bases of information received, published, or estimated, even before companies file their quarterly report. Quantifying the Revenue Potential from Excise Tax 3.35. The first Armenian excise tax system rightly focused on only a few key commodities. However, subsequent revisions of the excise law eroded the tax base removing two important categories - fuel and tobacco – to a presumptive regime. Recently, fuel was brought back to the normal excise regime, but tobacco is still under the presumptive regime. Furthermore, while it is common practice that goods subject to excise taxation are also subject to VAT at the normal rate, this is not the case in Armenia. Most excise taxes in Armenia are imposed in the form of specific rates, which is simple to administer. However, real revenues may fall in the face of price increases, which have been the case in Armenia, leading to a continuing erosion of the excise tax base. 3.36. Armenia’s tax yield on excisable goods is very low compared to Central and Eastern European countries. Excise taxes for the comparator countries typically raise one to two points more in revenue as a percent of GDP than in Armenia in 2009, the bulk of this difference stemming from the large collections on diesel (see Table 3.5). Grote et al. (2011) have estimated that Armenia‘s excise revenue to GDP decreased by 2.1 percent over the period 2002-2010. Table 3.5: Excise Revenues as a Percentage of GDP, 2009 Petroleum fuels Tobacco Alcohol Total Petrol Diesel LPG Armenia 0.81 0.34 0.61 0.12 1.88 Average Central and Eastern Europe 1.19 0.52 0.83 1.23 0.07 3.83 Average EU 0.83 0.35 0.71 0.92 0.03 2.83 Source: Ministry of Finance and IMF Staff (2011). 3.37. Improvements in the excise regime could yield additional revenues in the range between 0.5 and 2.6 percent of GDP. There is a need to introduce price indexation to reflect inflation on the specific rates used. This will induce the increase in the specific rates to properly compensate for the negative externalities imposed on healthcare costs by excisable goods. Since the IMF has estimated that Armenia excise revenue to GDP decreased by 2.1 percent of GDP over the period 2002-2010, we can expect 2.1 percent of GDP in additional revenue from this adjustment of tax rates. In addition, excisable goods 25 should all be subject to regular VAT taxation, which is estimated by the IMF to yield an additional 0.5 percent of GDP. The revenue potential of excise taxation is therefore estimated between 0.5 and 2.6 percent of GDP. 3.38. The above calculations suggest that these two improvements in tax policy, while technically simple but presumably difficult to carryout politically due to strong vested interest, could yield between 0.9 to 3.2 percent of GDP and additional revenues. These, again, are conservative estimates for these two specific improvements. In particular mining taxation should end up yielding more revenues during the times in which commodity prices boom. However, since the details for the new royalty regime have not been finalized, our estimates have the strong assumption that the implicit rate of taxation would at least be that observed in previous years—when in fact it could be higher. Identifying Tax Loopholes in Armenia Tax Laws 3.39. We reviewed the taxation legal system by reading and classifying the existing versions of the laws provided by the web page of the Armenian Parliament, attempting at identifying loopholes but not at quantifying the losses they might generate.11 Each individual tax law was converted into a matrix which summarized its contents by concept (taxpayer, threshold, base, exemptions and exclusions—some of them called ―tax privileges‖ in Armenia) and excluded the articles that were considered of procedural nature.12 These articles were assembled in a procedures matrix based on the law on taxes. The following nine laws were reviewed under the approach described:  Law on Taxes (which is the framework of the system)  Law on Profits Tax  Law on Individual Income Tax  Law on VAT  Law on Excise Taxes  Law on Land Tax  Law on Property Tax  Law on Presumptive Payments  Law on Organizing and Conducting Audits 3.40. Unfortunately, the tax system in Armenia comprises two parallel taxation structures: a regular and a presumptive regime. The regular taxation regime reflects contemporary Western taxes (CIT, PIT, VAT and some standard type of excises), while the presumptive (in lieu of VAT and CIT) covers the bulk of informal merchants. The regular regime appears to be a friendly system to foreign operators: some favorable treatments given to non-resident taxpayers, combined with a relatively low profit tax rate compared to other countries. The revenue engine seems to be the VAT. The second structure is a standardized way to estimate economic activity for some trades.13 It is based on parameters that have no visible connection to modern tax law of Western-like systems of taxation. It simplifies the problem of compliance for informal taxpayers, but it yields only a small contribution to revenues. The presumptive tax structure creates a universe of taxpayers, operating in the informal economy, that pay neither income nor VAT taxes. The revenue cost of these de facto internal tax havens, which reduce the tax rate of any formal-sector firm that manages to shift activity into them, should be measured. 11 http://www.parliament.am/legislation.php?sel=alpha<ype=3&lang=eng. 12 Consultant study produced by Jaime Vazquez-Caro, 2010, ―Public Expenditure Review: Revenue Component‖. 13 Initially, about 20 trades were covered. The National Assembly of the Republic of Armenia passed the RA Law on Amendments and Supplements to the RA Presumptive Fees Law on May 11, 2011 specifying 5 types of activities to be moved to the general taxation regime effective January 1, 2012. 26 3.41. Equity considerations do not seem to be a concern in the existing laws. The existence of a single rate VAT as a dominant revenue item reinforces this idea. The combination of a single rate VAT with a relatively low PIT suggests that the prevalence of the VAT has led to lower personal income tax rates. On the other hand, an aspect that is immediately clear is the equity issues implicit in the differential rate treatment of residents and non residents who face differential privileges. In the long run, the equity issue will affect the perception of fairness of the tax system, becoming a strong compliance consideration. 3.42. Tax base erosion through tax expenditure does not seem to be a concern of the Government. Tax law in Armenia creates many explicit and costly privileges and, perhaps, hidden ones. For instance, some special legal provisions of the profit tax law are introduced for the Deposit Guarantee Foundation, the Pan-Armenian Bank, and even the Gyumri province. As it has been well highlighted in the fiscal literature, not only are tax expenditures a direct fiscal cost, they are also an invitation to expand their impact by hard-to-detect combinations and expanded interpretations well beyond the original intent of exemptions or tax credits. These legal loopholes enable tax avoidance and evasion without ensuring that the expected developmental outcomes, which ostensibly motivated the tax break, will be fulfilled. When this happens, by inducing evasion, tax expenditure create a revenue loss larger than the fiscal cost of a direct subsidy, without the assurance of a positive result.. 3.43. In Armenia, no effort is made to measure either the costs in foregone revenue or the benefits in developmental outcomes, and to evaluate the net results of tax exemptions and privileges. The VAT law shows quite a long list of exemptions that prima facie are hard to monitor or control (Box 3.3). Any exemption should be an immediate candidate for quantitative analysis and monitoring as well as for a special audit program. With respect to the CIT , revenue consequences of exemptions to the agricultural sector and those on Free Trade Zones have not been measured and their use is not questioned for audit purposes, despite the major risk that these exemptions represent in terms of the under pricing of transactions. For instance, an agro-industry company may use internal transfer prices to buy from a farm it owns at very high price, generating a loss at the processing stage, while transferring a large profit to the farm which is tax-exempted. Rigorous quantitative assessment of such activities should be part of a permanent risk analysis. 3.44. The VAT legal structure has some flaws and the threshold is too high. VAT refunding is a structural characteristic of most VAT systems, introduced in the 1970s in many countries as a substitute to the VAT exemption approach, which was found to be more prone to fraud and corruption. Over time the IMF has accumulated evidence of ill-performance of the VAT refund system in Armenia. Another legal issue is that the VAT threshold has been deemed too high, which creates discontinuities in the VAT chain of information. This problem was exacerbated in 2009 when Article 3 of the VAT Law was revised to increase the VAT turnover threshold from 30 to 58 million AMD (US$79,000-153,000) for VAT payment exemption, combined with a complex set of rules for which the threshold is overruled. It is not advisable to keep the threshold too high for two reasons: (i) it maintains a large segment of producers without reporting obligations and this weakens the VAT information flow, which depends on the built-in opposition of interest between buyers and sellers and, (ii) tax neutrality breaks down when small producers are taxed under one set of rules and large producers under another (the VAT). For these reasons, this analysis concurs with IMF recommendations to lower the threshold to about half of the existing one. 3.45. The short statute of limitations for tax audits is a flaw in the process designs. At first sight, our review of the processes defined by law did not reveal specific problems derived from the text of the laws. In an indirect way, however, the tax procedures have established a major vulnerability of the tax system: the three year statute of limitations to perform investigations on filed returns. General practices around the world place the statute of limitations at 5 years and some others leave it open in fraud cases. The brief three year period for audits reduces the expected cost of tax fraud in Armenia by lowering the probability of being caught. 27 Box 3.3: Armenia – Transactions and Operations Exempt from VAT Sales of: 1. copy-books, music books, drawing albums, children‘s and school literature, school educational publications; scientific/educational editions published by higher educational institutions, specialized scientific organizations; 2. scientific and research works; 3. poisonous chemicals/fertilizers for agriculture, agricultural plants, perennial plantations seeds and planting material to the producers of agricultural production; 4. goods produced and agricultural products, produced in Armenia; 5. newspapers and magazines; 6. lottery tickets at the face value; 7. donor blood, breast milk, prosthetic and orthopedic items, medical assistance services (prophylactic diagnostic measures), goods related to treatment prepared by the patients within medical assistance frames in prophylactic enterprises/organizations, and services rendered by them; 8. precious and semiprecious stones indicated in the list specified by the Government; Other exempt transactions and operations: 9. Secondary/vocational schools for qualification and re-qualification, specialized -secondary and higher educational institutions - in the part of the fee for education 10. Care of children in preschool institutions, of persons in boarding- for disabled children and invalids services rendered by the persons living at the expense of these institutions; 11. Provision of services by undertaker‘s offices, cemeteries, as well as other services of ritual character and sales of respective accessories related to the death and burial; 12. Religious ceremonies, sales of religious items to religious organizations, sales of these items by the religious organizations; 13. Insurance and reinsurance operations, including services related to them, which are rendered by insurance mediators and agents; 14. Pension insurance related operations, including related services, rendered by mediators and agents; 15. Delivery of goods and provision of services, according to the Government procedure established for preparation facilities of credit/ grant programs of international financial organizations; 16. Free consumption provided by public (including benevolent) and religious organizations. 3.46. The 2008-2009 attempts to produce a Tax Code never materialized although a full text was drafted. The production of a unified procedural tax code as a self-standing clear document would be of great value. This task would need to define uniform procedure rules for all taxes and recognize tax specificities that require unique procedures in a separate section of the code. 3.47. The findings point out at the important agenda of improving tax policy and administration in Armenia. Addressing this agenda would require:  Strong political commitment as, clearly, there are many economic agents that profit significantly from the existing situation. Forging a Social Compact, by which an agenda to improve revenue collection is accompanied by an agenda to improve the effectiveness of public spending (particular on social protection and infrastructure) and a plan to simplify the business environment could provide political headroom—if credible. Although not desirable from the efficiency of spending policies point of view, credibility could be enhanced via the introduction of golden rules or earmarking.  Steady investments in the modernization of methods, skills and systems for both tax policy and administration. o Armenia‘s capacity to understand and address tax policy problems needs significant investments. Efforts should go into hiring more and greatly qualified staff. Training in tax policy analysis for the current staff is also essential. Even with more and better trained staff, the unit will need to rely on a strong cadre of foreign consultants if it decides to 28 undertake the development of a full fledge modern Tax Code, which, in our view, is the first way to address the many inconsistencies created in several pieces of regulation o SRC‘s weakness has been better studied in various reports and assessments and, thus, there is no need to repeat them. Ongoing support by various donors (including USAID and the Bank) should continue.  Stronger and more formal collaboration needs to be established between the MoM and the SRC. This includes granting access to the MOF staff to all SRC databases as up-to-date data is essential for conducting rigorous analysis. This also includes the creation of working groups in areas that might be of interest (such as excise policy and administration, etc.). Finally, a more rigorous differentiation of roles and responsibilities needs to be established, for instance in terms of preparation and updating of manuals or administrative regulations. It is suggested that a new MoU between the SRC and the MoF be signed, modeled after best international practice. 3.48. Armenia’s revenue potential is estimated between 2.3 and 5.8 percent of GDP, but it could be more given that we only focused on a limited number of sources of tax revenue leakages. We did not estimate the effect of removing tax privileges and loopholes, nor did we estimate the impact of under- declaration of taxable income. The revenue potential from tax administration is estimated between 1.4 and 2.6 percent of GDP, covering leakages from large taxpayers VAT, CIT and PIT payments (between 0.4 and 1 percent of GDP), leakages from customs duties (between 0 and 0.3 percent of GDP) and leakages from firms operating under the presumptive regime (between 1 and 1.3 percent of GDP). We were not able to quantify the revenue potential from improving the legal system and closing the identified loopholes. We also estimated between 0.4 and 0.6 percent of GDP additional revenue from adopting and enforcing the new mining taxation, and between 0.5 and 2.6 percent of GDP from excise taxation. This makes for at least 0.9 to 3.2 percent of GDP additional revenues to be expected from an improved tax policy. Revamping the excise taxation regime by bringing tobacco back to the regular excise regime, adjusting up the specific rates to reflect the negative externalities caused by excisable goods and submitting them to the regular VAT regime appears to be the tax policy reforms with the highest revenue impact. The second in line is the new mining taxation proposing to combine the natural resource user fee and royalty into a single royalty regime with a higher rate and a sliding svcale to capture windfall profits. 29 CHAPTER 4. EXPENDITURE RATIONALIZATION AND EXPENDITURE EFFICIENCY GAINS Improving the efficiency of expenditure of the state budget is another option to accommodate debt sustainability pressures. However, despite the fiscal space created during the years of double-digit growth, the Armenian Government’s expenditure to GDP remained relatively small, and with total expenditures to GDP contained below 30 percent. During the crisis, the Government protected its priorities: social security, defense, general public services and education. Expenditure rationalization cannot bear the whole brunt of the fiscal consolidation, although efficiency can be improved by harmonizing the public service pay regimes, streamlining the co-payment policy in healthcare, reorganizing the provision of educational services in rural areas and prioritizing the maintenance of the most used segments of the road network through performance-based contracting. Increase the spending allocation on education, health and roads may even be needed to sustain the country’s fragile growth. 4.1. Social security, defense, general public services and education have been the main priorities of the Armenian Government over 2006-09. From 2006 to 2009, public expenditure increased by 81 percent, creating headroom for a steady increase in social protection, defense, education and general public services. Defense and general public services have attracted a consistently increasing share of GDP over this period, with their combined share reaching 7.8 percent in 2009 (Table 4.1). However, the largest share of public expenditure has been allocated to social security whose share of GDP increased from 4.8 percent in 2006 to 7.9 percent (higher than defense and general public services combined). The education sector has fared relatively well with an average annual increase of 11 percent up to 3.5 percent in 2009. The 2010 and 2011 budgets illustrate the adjustments the Government has been making since the crisis. Although expenditure continued to increase in nominal term, the Government shifted expenses from defense, public order, real sector (economic relations), education and social security to general public services (driven by interest payments), environment protection and house building. 4.2. Although total public expenditure consistently increased over 2006-09 at an average annual growth rate of 22 percent, it was unevenly distributed to sectors. An intuitive way to examine this is, to take, for each functional division, the ratio between the difference in amount devoted to that division and the change in total expenditure from one year to another.14 Figure 4.1 plotting this indicator shows that:  the 31 percent growth of public expenditure in 2007 was mainly targeted at to economic relations, followed by social security, education and defense;  the 20 percent expansion in 2008 was mainly devoted to social security, followed by general public services and defense;  the 15 percent increase in public expenditure in 2009 was mainly dedicated to the real sector (reflecting the fiscal stimulus) and social security. 14 Precisely writing the budget as . The budget difference is then . For each division, represents the share of the budget extension allocated to division k. The sum of those shares is equal to 1. 31 The allocation of fiscal increments shows that the Government has always regarded social security and social protection as a key priority –i.e., before and during the crisis. The calculations do however suggest that there was a shift towards economic relations and away from education and defense before and after the crisis. Table 4.1: Armenia Consolidated Expenditures by Functional Classification 2006 2007 2008 2009 2010* 2011** Total (in billion AMD) 514.6 674.1 807.5 929.1 959.2 998.6 in percent of GDP General public services 2.4 2.5 2.9 3.6 4.3 4.1 Defense 3.0 3.1 3.3 4.2 3.9 3.7 Public order, security and judicial work 1.6 1.7 1.7 2.2 2.0 1.6 Economic relations 1.8 3.3 2.4 4.3 3.3 2.6 Environmental protection 0.1 0.1 0.1 0.1 0.2 0.3 House-building and public utilities 0.5 0.6 0.4 0.7 1.3 1.2 Healthcare 1.5 1.5 1.4 1.8 1.7 1.7 Recreation, culture and religion 0.5 0.5 0.4 0.5 0.5 0.5 Education 2.5 2.8 2.8 3.5 2.9 2.9 Social security 4.8 4.7 5.8 7.9 7.0 7.2 Other 0.7 0.7 0.9 1.1 0.6 0.7 Source: Ministry of Finance. Note: *Revised budget; **Approved budget. Figure 4.1: Allocation of increases in total expenditure to functional divisions Reserve funds not covered by main divisions Social security 2007/2006 Education Recreation, culture and religion Healthcare 2008/2007 House-building and public utilities Environmental protection Economic relations 2009/2008 Public order, security and judicial work Defense General public services -0.05 0.00 0.05 0.10 0.15 Sources: Ministry of Finance and World Bank Staff. 32 4.3. Compared to some regional benchmarks, Armenia appeared to under spend on general public service, real sector, social security and health, while over-spending on defense, education and housing.15 Figure 4.2 shows that Armenia spends on average more on: defense (1.5 percent of GDP more), education (0.3 percent more) and house building and public utilities (0.2 percent more). At the same time, the country spends on average less on social security (4.1 percent of GDP less), general public service (3.4 percent less), economic relations (1.5 percent less) and healthcare (1.2 percent less). These results have caveats. Firstly, the ECA countries spend significantly more on social security inherited from the Soviet era and have been struggling to contain the increase of their social security expenditures. Hence Armenia relative under-spending on social security should rather be seen as a major achievement of the Government. Secondly, while education expenditures in Armenia are only financed from the central Government, many ECA countries have a broader source of education financing at the community level which may explain why Armenia appears to over-spend in the sector. The analysis confirms that more, not less, may be needed to improve Armenia rural education outcomes. Figure 4.2: Armenia’s Average Differences with Benchmark Countries, share of GDP 4.00 Social security 2.00 Education 0.00 2006 2007 2008 2009 Average Recreation, culture and religion -2.00 Healthcare -4.00 House-building and public utilities -6.00 Environmental protection Economic Relations -8.00 Public order, security and judicial work -10.00 Defense -12.00 General public services -14.00 Sources: Ministry of Finance, IMF and World Bank Staff calculations. 4.4. Expenditure rationalization has been integrated in the 2012-2014 MTEF. Although some revenue potential in the range of 2.1 to 5.8 percent points of GDP has been estimated, measures to tap this potential may take some time to materialize in revenue increase. Expenditure should bear its share of the needed fiscal consolidation. We need to take a closer look at the long-term expenditure to identify low hanging fruits the Government could rely on to speed up its fiscal consolidation. These include the wage bill, a scaled up effort to improve the efficiency of social spending , and a new look at capital expenditure 15 The report uses the IMF GFS data to benchmark Armenia against Belarus, Bulgaria, Czech Republic, Georgia, Kazakhstan, Latvia, Moldova, Russia, Slovenia, Ukraine, China and Tunisia. For each of the benchmark countries, the GDP shares of all functional classification divisions are subtracted from equivalent shares for Armenia, and then these differences are averaged to obtain Armenia‘s deviation from the set of benchmark countries. 33 as a driver of future growth while keeping a cautious stance vis-à-vis large infrastructure projects that may considerably increase the Government‘s implicit contingent liabilities. 4.5. These tradeoffs are carefully analyzed using detailed budget and expenditure data covering 2006-09. Working in close collaboration with the MoF to harmonize the 2006-07 budget and expenditure based on GFS1986 with the 2008-09 data based on GFS2001, a multidimensional disaggregated budget and expenditure database was built. The database allows looking at public expenditures combining functional, economic and administrative classifications at the most disaggregated level available in Armenia. Such user-friendly access allows better assessing allocation efficiency of public expenditure, and easily identifying priority programs that need to be ring-fenced to effectively protect the poor and vulnerable if the effects of the crisis are protracted. The rest of the chapter reviews first expenditure trends over 2006-2009 and then delve into public sector wage bill, health, education and road transport expenditures to better inform the fiscal consolidation process. A. EXPENDITURE TRENDS, 2006-2009 4.6. Spending in general public services was driven by subsidies and grants, personnel expenses and interest payments during the pre-crisis and crisis period (2006-2009). The authorities have been consistently supporting regional equalization. The share of subsidies and current grants has been increasing reflecting increases in equalization grants and other current grants to sub-national entities (Table 4.2). The headroom for this regional development policy was created by reducing the share of personnel expenses at least during the 2006-08. The allocation for salaries and other payments was slightly adjusted upwards in 2009. This study analyzes in greater detail the public administration wage bill, including civil servants in both central and sub-national entities. During the pre-crisis years, increase in interest payments was driven by greater issuance of domestic securities. Table 4.2: Expenditure on General Public Services In Share of Total, % 2006 2007 2008 2009 Subsidies and Current Grants 31.7 32.0 49.5 46.2 Personnel Expenses 26.8 25.1 20.7 21.3 Interest Payments 13.9 12.8 10.0 14.4 Other Recurrent Costs 9.8 9.8 7.9 7.1 Acquisition of Goods and Services 8.6 7.3 6.6 6.7 Capital Expenditures 9.3 13.0 5.4 4.4 Sources: Ministry of Finance. 4.7. Spending in public order was mainly driven by personnel expenses during the pre-crisis and crisis period. The large increase from 1.7 percent of GDP to 2.2 percent between 2008 and 2009 was due to an equivalent increase in personnel expenses (mainly salaries and other remunerations). In fact, the share of personnel expenses in total public order expenditures increased by 15 percent over 2007-2009 (Table 4.3). At the same time, acquisition of goods and services consistently decreased from 31 percent of total public order budget in 2006 to 18 percent in 2009, due to a major decline in expenses in personal and catering materials. Capital expenditures have been fluctuating between 15 and 21 percent of total public order expenditure over 2006-2009. Whether or not the spending in public order has improved security for businesses and citizens in general is outside the scope of this PER. 34 Table 4.3: Expenditure on Public Order, Security and Judicial Work In Share of Total, % 2006 2007 2008 2009 Personnel Expenses 50.3 48.6 56.0 63.5 Acquisition of Goods and Services 31.0 26.7 22.2 18.1 Capital Expenditures 14.9 20.9 19.1 14.8 Other Recurrent Costs 3.4 3.6 2.6 2.7 Subsidies and Current Grants 0.3 0.1 0.1 0.9 Sources: Ministry of Finance. 4.8. Real sector expenditure has experienced the largest variations from year to year. This sector has been used as the adjustment variable of public expenditure, at least judging from the recent trends the share of these spending to GDP increased by nearly 80 percent between 2006 and 2007, but then decreased abruptly by about percent 30 percent in 2008 before expanding again by 80 percent again in 2009. The 2010 and 2011 budgets indicate a declining allocation to the real sector, which reflects the phasing out of the fiscal stimulus injected in infrastructure construction. To support the export-led growth strategy the Government is now aiming at, some growth-sustaining expenditure on transport, agriculture and energy may need to be scaled up. 4.9. Healthcare expenditure equally funded outpatient and hospital services as top priorities An average of nearly 37 percent of the healthcare expenditure has been allocated to outpatient and hospital services over 2006-09 (Table 4.4). Medical goods, devices and equipment received an average 9 percent of the total while an average 5 percent was dedicated to public healthcare services. Very little has been allocated to research and development while other healthcare areas received 12 percent, although half of it was allocated to capital expenditure (Table 4.4). Table 4.4: Healthcare Expenditure over 2006-09 2006 2007 2008 2009 average Total health expenditure (in billion AMD) 39.4 46.8 50.0 56.2 48.1 in percent of total health expenditure Medical goods, devices and equipment 6.9 12.9 9.5 6.3 8.9 Outpatient services 35.9 37.1 37.3 36.4 36.7 Hospital services 36.5 32.7 38.7 38.5 36.6 Public healthcare services 5.6 5.3 5.6 5.1 5.4 Research and development 0.2 0.1 0.0 0.1 Healthcare (not belonging to other classes) 15.0 11.8 8.9 13.8 12.4 Source: Ministry of Finance. 35 Table 4.5: Education Expenditure over 2006-09 2006 2007 2008 2009 Average* Total education expenditure (in billion 67.4 89.8 103.5 107.5 92.0 AMD) in percent of total education expenditure Pre-school and general elementary 15.3 18.1 16.7 education secondary education General 50.0 41.3 42.4 48.5 45.5 Primary and secondary vocational 4.0 3.8 3.7 4.4 4.0 Higher education 7.5 6.1 6.5 7.0 6.8 Education not classified by levels 2.0 1.9 2.4 2.9 2.3 Auxiliary services provided to education 19.3 29.5 29.3 18.7 24.2 Education (not belonging to other 0.6 0.5 0.4 0.5 0.5 classes) Ministry of Finance. Sources: Note: The average of preschool and general primary education from general secondary education in 2006 and 2007 is subtracted for consistency. Table 4.6: Armenia: Number of Recipients of Key Social Protection Programs 2005-2007 2005 2006 2007 Pensions 540,298 533,288 527,496 Social insurance pensions 486,015 479,791 473,612 Old-age 358,886 355,003 342,669 Disability 75,024 77,757 88,117 Survivors‘ 14,755 14,731 14,600 Social pensions 44,657 44,858 46,233 Military 17,374 15,575 13,889 Family benefit program Regular assistance Number of families 127,167 130,406 124,689 Number of people 467,534 484,551 461,359 One-time benefit Number of families 8,342 9,264 n/a Number of people 20,560 19,865 n/a Birth grant 25,237 29,257 38,917 Child care allowance for employed mothers 6,320 5,145 Unemployment benefit 8,645 12,500 n/a Cash for work Source: Administrative statistics as reported by the Nationals Statistical Service of Armenia and the Ministry of Labor and Social Issues. Note: The number of beneficiaries of paid maternity and temporary sick leave is not available, because the monitoring system tracks days of leave, not the number of persons using it. 36 4.10. Education expenditure has mainly focused on general primary and secondary general education as well as capital expenditure. As presented in Table 4.5, the education expenditure is allocated to eight groups: preschool and general elementary education16 (average of 17 percent of the total), general secondary education (46 percent), primary and secondary vocational education (4 percent), higher education (7 percent), education not classified by levels (2 percent), auxiliary services provided to education (24 percent) and other expenditures (less than 1 percent). A closer look at the post ―auxiliary services provided to education‖ shows that 80 percent on average has been allocated to capital expenditures. This sector is analyzed in more detail in this study. 4.11. Social security expenditure targeted old age pension and the family benefit program. During 2006-2009, more than two third of social security expenditures were allocated to old age pension. The Family Benefit program received on average 19 percent of the social security envelope. The social protection programs cover common types of social risks and largely correspond to the Armenia‘s current poverty and vulnerability profile (Table 4.6). They also reflect Armenia‘s social contract, particularly pertaining to the provision of social insurance benefits, as well as the national priorities such as the increasing birth rate. For instance, different types of pensions and allowances are aimed at providing income support in situations when it declines because of old age, disability, sickness or loss of a bread winner. Maternity leave, birth grants, and child care allowance are meant to provide incentives for families to have more children, but also to give mothers an opportunity to stay in the labor market. 4.12. Current expenditure has Table 4.7: Armenia consolidated expenditure in economic been fluctuating between 75 and classification 90 of total expenditure before 2006 2007 2008 2009 and during the crisis. This trend Total (in percent of GDP) 19.4 21.4 22.1 29.9 reflects more the large variations registered by expenditures on fixed in percent of total expenditure assets: from 18 percent of total Current expenditure, o/w 84.4 76.1 88.5 77.3 expenditures in 2006, their share Wages and salaries 8.9 7.6 7.9 8.1 rose by 8 percent in 2007, and then Acquisition of goods and services 20.9 18.1 16.8 16.4 declined by 5 percent in 2008 Interest payment 1.8 1.5 1.4 1.8 before increasing by 4 percent in 2009 (Table 4.7). This echoes the Subsidies 3.6 2.9 10.9 2.0 large variations observed on the Grants 8.5 8.1 10.2 8.6 real sector‘s expenditure. Wages Social allowances and pensions 23.2 20.3 25.7 25.9 and salaries have been contained Other costs 17.6 17.7 15.6 14.5 between and 8 and 9 percent of Capital expenditure, o/w 15.6 23.9 11.5 22.7 total expenditure with a slight Fixed assets 18.2 26.4 20.9 23.8 decreasing trend. Interest payment has been contained below 2 Strategic reserves 0.4 0.3 1.6 0.0 percent although the service of the Acquisition of land 0.0 0.0 0.0 0.0 debt contracted during the crisis Receipt from alienation of non- will impose an upward trend over financial assets -3.1 -2.8 -11.0 -1.1 the medium term. Grants to sub- national entities have attracted on Source: Ministry of Finance. average 8.5 percent of total expenditure, with a 10 percent share registered in 2008. 4.13. An average of 5 billion AMD (US$13 million) have been allocated every year to water user associations as irrigation subsidies. In 2009, this represented 38 percent of total subsidies provided to 16 Note that this group was lumped to general secondary education before 2008. 37 the real sector and 69 percent of the subsidies to the agriculture sector. Except in 2008 when massive subsidies were provided to mineral oils and natural gas, irrigation has received the highest share of subsidies allocated to the real sector, at more than 30 percent. In the context of the needed fiscal consolidation, a closer look at the efficiency of this expenditure is necessary. 4.14. Capital expenditures have been mainly on construction and reconstruction of buildings and edifices. Of the 161 billion AMD (US$424 million) allocated on average to capital expenditures during 2006-2009, 72 billion AMD (US$ 180 million) were allocated to new construction of building and edifice while 64 billion AMD (US$ 168 million) went into capital reconstruction (Table 4.8). Acquisition of buildings and edifices has been very volatile, ranging between a low 99 million AMD (US$261 million) in 2008 and a high 25 billion AMD (US$66 million) in 2009 (debited from the Government‘s reserve funds). On average, 17 billion AMD (US$45 million) were allocated every year for the purchase of equipment. Likewise, every year (between 2 and 3 billion AMD (US$5-8 million) have been allocated to purchase equipment distributed as follows: 15 percent for transport equipment, 32 percent for administrative equipment and 53 percent for other machinery. Receipts from assets reported were received mainly from the alienation of land which varied from 14 to 19 billion AMD (US$37-50 million) during 2006-2009. Table 4.8: Capital Expenditure in Armenia, 2006-2009 in million AMD 2006 2007 2008 2009 Average Construction of buildings and edifices 32,348.8 74,355.9 81,411.5 100,205.3 72,080.4 Capital reconstruction of buildings and edifices 41,137.2 73,066.1 64,596.8 78,812.6 64,403.2 Acquisition of buildings and edifices 991.4 1,624.3 99.2 25,418.1 7,033.3 Equipment 15,292.6 23,608.5 18,026.7 12,728.7 17,414.1 Other, incl. receipts from assets -9,685.8 -11,675.4 -71,169.8 -6,406.0 -24,734.3 Total capital expenditure 89,769.9 172,654.9 164,134.2 217,164.7 160,930.9 Source: Ministry of Finance. 4.15. In summary, from 2006 to 2009 public expenditure increased by 81 percent, creating headroom for a steady increase of expenditures on social protection, defense, education and general public services. This trend was not altered by the 2009 crisis when the Government opted to stimulate the economy by raising spending on social protection, general public services (driven by interest payment which increased by 57 percent y-o-y), economic expenditure (driven by construction of transport and irrigation infrastructures) whose share in total expenditure increased from 11 to 14 percent, and housing expenditure whose share jumped from less than 2 percent in 2008 to 2.3 percent by the end of 2009. Four sectors are analyzed in further detail to assess the efficiency gains that could be expected. B. PUBLIC SECTOR WAGE BILL 4.16. Armenia does not appear to be an outlier in terms of compensation of employees. Compared to other low- and middle-income countries in the ECA region, consolidated general government wage bill costs – as a percent of GDP -- fall at the lower end of the spectrum (Figure 4.3). In 2008, Armenia spent 4.9 percent of GDP on compensation of employees, compared to 5.3 percent in Georgia and 9.2 percent in Moldova. However wages in the education and health sectors are not included since they are classified under the provision of goods and services. Note, however that comparator countries might also be excluding some personnel from ―compensation of employees‖ line item. Figure 4.4 also show estimates 38 for compensation of employees that include Armenia‘s education sector wage bill.17 Even with this adjustment, Armenia remains at the lower end, with an estimated 6.6 percent point of GDP spent to pay employees from the public sector. 4.17. Furthermore, the country’s public employment is comparable to other ECA countries. 76 per 1000 population in Armenia are public employees (Figure 4.5), in line with countries like Croatia and Slovakia.18 Armenia appears however to be an outlier when considering public employment as a share of total employment. The second panel of Figure 14 shows Armenia at the far right with 34 percent of total employment in the public sector. This may be due to the fact that official total employment numbers do not include the large Armenian informal sector, estimated to contribute 52 percent of employment (Asian Development Bank, 2010). Figure 4.3: Compensation of Employees as a Percent of GDP, 2008 14% 12% 10% Percent GDP 8% 6% 4% 2% 0% Source: IMF GFS and IFS, and World Bank calculations. 17 The estimates are the sum of 2008 compensation of employees as reported in GFS plus the 2009 approved budget for elementary, basic and higher general education wages and salaries and actual social security contributions. Data for 2008 were not available for health sector. 18 Public Employment includes the following standard categories used in the International Labor Organization‘s Laborsta database: Public administration and Defense, compulsory Social Security; Education; Health and Social Work. All data are for 2008 except Bulgaria, which refers to 2007. 39 Figure 4.4: The Size of Public Employment in Armenia and other Transitional Countries Public Employment per 1000 Population, 2008 Public Employment as Percent Total Employment, 2008 120 40% 35% 100 30% 80 Employees 25% 60 20% 15% 40 10% 20 5% 0 0% Source: ILO Laborsta, IMF IFS for population data, and World Bank Staff calculations. 4.18. Education and health account for the bulk of public servants. In 2006, Armenia‘s civil servants totaled roughly 185,000 individuals with an estimated 52 percent working in the education sector and 18 percent working in the health sector (Table 4.9). The bulk (110,624 employees) was working in sub-national entities, with 69 percent of them employed in education and 16 percent working in healthcare. Education expenditure is allocated through the per-capita financing mechanism, while health expenditure is partly allocated through state orders to the State Health Agency (SHA). The analysis of expenditures on health and education in the following sections provide a more detailed assessment of these two major sources of public employment. Table 4.9: Overview of Civilian Government Employment in Armenia (2006)19 Central Sub-national Total Staff Percent Staff Percent Staff Percent General Government, o/w 74,586 110,624 185,210 100% Education 20,733 28% 76,149 69% 96,882 52% Health 15,636 21% 18,175 16% 33,811 18% Other 38,217 51% 16,300 15% 54,517 29% Civil Servants (2010) .. .. 7,960 .. Source: World Bank and Ministry of Labor. Civil service data are from Civil Service Council. 4.19. The civil service – which comprises core white-collar administrative staff - is a small component of the public sector employment: approximately 8,000 individuals. The 2001 Law on Civil Service specifies forty-odd government bodies in which civil service positions may be located. 20 The Law requires a university education for all job groups except the lowest, and defines in some detail the required years of experience for each job group. The Civil Service Council approves the general 19 Since the official government data on employment do not follow the format commonly used for international analyses and comparisons, obtaining figures in this format requires a time-consuming process of sorting and classifying individual records. This was last done in 2007, with 2006 the most recent available year. 20 Article 4 of the Law, specifies that civil service positions are in the President‘s Office, Prime Minister‘s Office, executive bodies, state bodies in administrative operations of ministries, regional offices (Marzpeterans) and Yerevan municipality, permanent bodies created by law (except for the Central Bank). 40 description of the four groups of civil service jobs (Junior, Lead, Chief and Senior) as well as the overall roster of civil service positions and the group and grade to which each position is assigned.21 Each of the four groups comprises two or three grades. Data from the Civil Service Council show that the Chief group is the largest (43% of total) followed by Lead (34%). Given the low level of base salaries (from an entry monthly base salary of 104$ to the highest base salary of 610$), this apparent inflation in grades may suggest that staffs are mechanically promoted to higher grades over time to adjust up their remuneration levels. This hypothesis is confirmed by anecdotal evidences within the civil service, and the MoF is conducting a grade inflation study to quantify the scope of the problem. 4.20. The bulk of civil servants are 40-65 years of age (60%), with 26 percent aged 30-40 and 14 percent under 30. This age structure suggests a challenge as well as an opportunity for the Government: to attract and retain a new, younger cohort of civil servants. The current complement retires in the coming years, as well as to reduce employment, where that is deemed desirable, through attrition rather than separations. Replacement of retiring employees will require a thorough assessment of the public sector labor compensation policy. We examine the various pay regimes in force, then assess the adequacy of the current pay levels of civil servants, and finally explore some pay and employment policy simulations to evaluate the fiscal cost of some policies the Government is currently considering. Pay Regimes in Armenia’s Public Services 4.21. Appropriate compensation, especially for civil servants, is a prominent concern in Armenia. The issues go beyond the mere adequacy of pay levels and the extent to which they help attract, retain and motivate skilled staff to the civil service. Policy makers note the absence of a comprehensive public sector pay policy, and the perceived inconsistencies in salaries offered by the different pay regimes that operate within the public sector. 4.22. In addition to the civil service, Armenia has a number of “parallel� public services regulated under separate remuneration systems. Pay policy experts and policy-makers within the Government are concerned by the large number of pay systems, the perceived lack of coherence and consistency among earnings for similar positions, and the resulting potential inequities. There is also a perception that salary differentials drain applicants away from lower paid services to higher paid ones, though other factors may also be at play. The National Institute of Labor and Social Research, and the Ministry of Labor‘s Wage Policy Unit carry out analysis of these systems, whose broad outlines are summarized in Table 4.10. This information suggests the scope for variation - with twelve pay regimes covered by distinct legislations, base salary scales and additions. A more in-depth review that examines earnings for equivalent positions, and factors in the impact of allowances (funded from budgetary and extra-budgetary sources) would be necessary to draw robust conclusions as to the degree of any inequities in the system. 21 By law, the Council also approves the number of positions in each body although in practice, decision-making authority on this point rests with the MoF. 41 Table 4.10: Overview of Remuneration Systems for Armenia’s Public Services Service Staff Legal Act Base Salary, Maximum Base Salary For Regular Additions To Base Salary AMD/month Position (excludes bonuses) Management positions n/a Law on Base Salaries for 120,000 400,000; – Legislative and Positions of Managers in Executive branch Legislative, Executive and 680,000 (Chair of Judicial Branches Constitutional Court) Judicial system 220 Judicial code 440,000 660,000 For court chairpersons; Allowance for (Cassation Court, years of experience Appeals Court, Court (Judges of of First Instance; general Criminal, Civil & jurisdiction Administrative Court court) judges) Prosecutorial system 337 Law on Procuracy 392,000 Allowance for rank, defined as a percentage of chief prosecutor‘s base salary Civil Employees n/a Based on annual budget 720,000 approved by National Assembly Civil servants (except 7,603 Law on Civil Service 40,000 235,000 (for the highest For working in mountainous areas, Public Services Remuneration coefficient specified by law) hazardous work; especially hazardous Regulatory work and demanding work; for higher Commission and personal rank than required for the Chamber of Control) position Public Services 78 Law on Public Service 100,000 495,000 Regulatory Regulatory Commission Commission Chamber of Control 126 Law on the Chamber of 111,000 653,790 Control State Revenue Service 2,921 Law on Taxation, Law on 55,500 394,000 For special working conditions, Customs geographic environment; holiday weekend and night shift work; for special titles; for scientific degree Diplomatic service 178 Law on Diplomatic Service, 72,600 277,200 For diplomatic rank (for service and government decrees abroad); for third foreign language; for rank Rescue Service 973 Law on the Armenian 30,000 85,600 For working in mountainous regions‘ Rescue Service ranks; scientific degrees; for temporarily acting for other staff; years of service; overtime; weekend and holiday work ―Power‖ ministries n/a Government decree no. 32,348 116,451 Can receive pension while continuing (National Security, 1554 of 13 December 2007 to serve (as decided by government) Police, Ministry of Defense, Officers and Junior Commanders) Municipal Service 16,180 Law on Municipal Service 30,000 270,000 Source: Adapted from Civil Service Pay Strategy, the MoL and NILSR. Employment data from State Service for Social Insurance and World Bank calculations. 42 4.23. Bonuses appear to significantly augment earnings in the civil service. An analysis of earnings for civil service jobs in a sample ministry, one of its agencies and regional facilities shows that the compression ratio can be significantly altered by bonus payments. In its 2010 earnings survey, the NILSR used data to calculate average earnings for jobs with at least three incumbent staff. Three subsets of earnings were used for the analysis: base pay only; base pay plus additions (i.e. all regular monthly supplements and allowances, excluding bonuses); and finally, base pay and additions plus one-twelfth of the total annual value of bonuses. Vacation ―bonus‖, which equals roughly one month‘s base salary, is not captured in the data. As per the official pay scale, average base pay increases gradually with grade, while regular additions to base pay appear to be limited in value and incidence, contributing little to total earnings. In contrast, the value of bonuses – shown prorated over 12 months – considerably augments earnings in the ministry‘s headquarters as well as in the agency and other affiliated body (Figure 4.5). In the sample ministry, average bonuses for the selected positions range from the equivalent of 3.1 months of base salary per year to a high of 4.6. The highest bonus to a jobholder was 5.5 times monthly base salary, whereas some jobholders (including recent hires) received no bonus. Although it was not possible to analyze earnings by job title for multiple ministries and their affiliated bodies, preliminary analysis of the maximum reported value of total bonus payments to an individual jobholder in seven ministries found that these varied from a low of 1.9 in one to a high of 6.4 in another. Figure 4.5: Monthly Earnings by Position in a Sample Ministry, 2010 Average, Maximum & Minimum Monthly Earnings by Position in a Sample Ministry (2010) 300,000 250,000 Drams per month 200,000 150,000 100,000 50,000 0 4.2 1st class 4.1 1st class 3.3 Key 3.2 Key 3.1 Sr. spec. 2.3 Sr. spec. 2.2 Chief of 2.1 Chief of spec. spec. spec. spec. unit dept. Grade & Job Title Base Base + additions Base, additions + 1/12 bonus Source: Ministry of Labor and Social Issues and World Bank Staff calculations. Note: Vertical lines indicate maximum & minimum total earnings. Are Public Servants Underpaid? 4.24. Several analytical approaches can be used to assess the adequacy of the pay level of public servants. The most basic is to analyze pay compression ratios. Compression ratios are often used in civil service analyses as a quick indicator of the formal financial incentives for individuals to accept positions of greater responsibility. Ratios of earnings to consumption go a step further by relating pay to a measure of living expenses. Finally, two others approaches – jobs and worker analyses – provide more detailed, and sometimes divergent, insights that compare earnings for similar jobs or workers across sectors. 43 4.25. Armenia compression ratio is estimated to be between 3 and 5.9. Although compression ratios are easy to construct, they are very sensitive to the way the ratios are calculated. For instance, do we include the base salary only or also allowances and bonuses? How are the top and bottom positions defined? Using different cutoff points, Armenia compression ratio is estimated to be between 3 and 5.9. In addition to within-country variation due to calculations methods, ratios of highest to lowest salaries can vary substantially even among advanced economies. A glance at selected OECD civil services shows the following variations (World Bank, 2000):22  Belgium: 5.5 (base salaries) in the administrative stream  Canada: Estimated at 5.2  France: Estimated at 4.6 (base salaries) and 5.7 (with allowances)  Netherlands: 8.2  United Kingdom: estimated at 9.8  United States: 7.3 general white-collar civil servants, 8.9 including Senior Executive Service 4.26. Pay levels in Armenia appear to be sufficient for a reasonable standard of living only for the top two civil service groups (senior and chief). Using the above earnings data from a sample ministry and data from the National Statistical Service on average per capita consumption, it is possible to estimate how much of the average consumption expenditure for a four-member household (the average household size in 2008, according to NSS) could be financed from the net earnings of a civil servant. This is, of course, a simplification notably because, while the base salary and regular additions are likely to be comparable across ministries, as noted earlier the size of bonuses varies across ministries. Furthermore, in this example, total earnings (including bonuses) are underestimated, as they exclude vacation pay. Nevertheless, this perspective suggests that, while take-home pay for the top two civil service groups (senior and chief) is roughly sufficient to fund consumption for a family of four, in the third group (lead) it covers something around two-thirds of the expenditure and in the junior group less than half. 4.27. These types of comparisons, however, still leave open key questions. Are earnings appropriate to the requirements of the job? Are they sufficient to attract and retain the required skills, and reasonably competitive with alternatives (in conjunction with other factors such as benefits, working conditions, job security, etc.)? More nuanced analyses comparing earnings for similar jobs and for similar workers in the public and private sectors can shed some light on these questions. Both types of analyses are available for Armenia, and as in other countries, the two approaches yield differing but complementary perspectives on the competitiveness of public sector earnings vis-à-vis private and other sectors. 4.28. Job-based surveys show that non-public employees earn more than public employees, with the exception of those with an educational background in health. The NILSR conducts an annual earnings survey covering both the public and private sectors, although the sample is biased toward senior staff (managers) and white-collar workers. The analysis compares broader skill groups, using approximate positioning in the organizational hierarchy as well as the jobholder‘s educational qualifications (which may differ from the actual field of activity). The survey also does not address the types of benefits provided to workers (which may be higher in the public sector). With these caveats in mind plus the usual subjective value of any survey, the analysis shows that respondents in the non-public sphere consistently earn more than civil service counterparts, with the exception of those with an educational background in health. The gap is greatest for those with social sciences degrees (almost three times greater), while individuals with health credentials earned 26 percent more in the public sector. Overall, civil service salaries tend to be about 60 percent lower than reported earnings by non-public entities, which is a significant difference although not as dramatic as sometimes cited in anecdotal accounts. 22 While the figures are somewhat dated, the underlying point remains valid. . 44 4.29. On the other hand, comparisons using household data show that the Government is not necessarily a low payer relative to other employers. The 2008 Integrated Living Conditions Survey (ILCS) highlights the gap between private and public earnings comparing similar workers (in terms of education, gender, age, location). The analysis estimates the private-public pay gap in earnings in the ranges of 18 to 24 percent for all full-and part-time workers including earnings from primary and secondary employment, and 23 to 28 percent for full-time workers‘ earnings from the primary job only. For university-educated workers – the category most relevant for the civil service – the gap is 18 percent for all workers (all income sources) and 23 percent for full-time workers (primary job only). In rural areas, the private sector premium is greatest (47 percent) for secondary school and college-educated workers, and modest for those with a university education. In urban areas, the private sector premium is greatest for university-educated workers, with a 25 percent pay gap overall. These gaps are markedly smaller than the private sector premiums found in the career-based survey, which range from 45 to 287 percent by field of education (except for the health sector) and 11 to 99 percent by job level. Furthermore, while virtually all government jobs are permanent (99 percent overall), only 70 percent of private sector workers reports having a permanent job. Government workers have a shorter workweek overall (41.4 hours versus 47.3 for private sector) and even more so for university-educated workers (39.7 versus 48.7 hours). The value of benefits such as sick leave and paid annual leave should not be neglected, and appears to be greater in the Government than in a private sector. Some Policy Issues to Consider 4.30. The Government has been championing some initiatives related to the pay policy. As of mid- 2010, the Government was planning to link the size of bonuses paid to civil servants with scores from a new performance appraisal system, with the aims of enhancing the performance orientation of staff and units, as well as increasing the earnings of individuals deemed to be high performers while at the same time continuing to restrain the overall wage bill. A parallel initiative, approved on an annual basis for 2009 and 2010, offered organizations an incentive to cut positions by allowing them to retain the resulting ―savings‖ for distribution to remaining staff in the form of bonuses. However, this idea seems to have little credibility with staff and managers. The main concern appears to be that, without a permanent pay policy, any apparent savings will eventually be clawed back to the budget (since the wage bill allocation is determined based on the number of staff and their grade), leaving staff with a higher workload but not additional pay. In addition, in one ministry that acted on the incentive, staff were motivated to accept the trade-off by hopes for large bonuses; the sums ultimately paid out disappointed these expectations and damaged the morale. 4.31. We developed two quantitative models to assess the fiscal impact of pay initiatives being considered by the Government. The first model simulates the decompression of the base pay for the top four grades (both ―senior‖ grades and two of three ―chief‖ grades) which contain approximately 25 percent of civil servants in pilot bodies and their affiliated units. It also combines an increase to the cornerstone. The second model combines the reductions of employment in selected grades and in the subsequent year, a modest increase to the cornerstone to raise base salary for all staff while keeping the cost of wages and additions near the pre-staff cut level. The models focus on seven ministries and their affiliated agencies and other dedicated units using data from the NILSR‘s salary survey. 4.32. The first model shows that decompressing the top-4 pay categories under some conservative assumptions will cover only 5 percent of the employees of the 7 pilot ministries while inducing a 20 percent cost increase compared to the baseline. Coefficients for targeted grades increase by 7.5 percent in year 2, and a more modest 3 percent in year 3, when the cornerstone is raised from 40,000 to 45,000 AMD (US$105-118). In year 4, targeted grades receive an additional 5 percent increase to their coefficients. With these conservative assumptions, coefficient values and base pay in the targeted grades increase as described, with an especially large rise in base pay in year 3 due to the increase in the cornerstone. Because the number of staff in the targeted grades is relatively limited – with progressively 45 fewer staff in each higher grade – the additional costs to base pay and additions are restrained (only 2.4 percent increase in year 2 relative to the status quo); the increase to the cornerstone, which affects all civil servants, has a considerably larger impact on costs, as evident in the 20 percent increase in the cost of base pay and additions in year 3. 4.33. The decompression model confirms that basing pay policy on compression ratios has many drawbacks. The model increases the base pay coefficients (and thus base salaries) for several higher- level grades in order to achieve a target compression ratio. The actual coefficient increase and the choice of grades are more or less arbitrary, guided only by the mandate to decompress the salary scale, and the assumption that recruitment and retention problems are more severe for higher-graded jobs and are driven by pay rather than other considerations such as job satisfaction or professional development. Furthermore, this approach lumps together very different jobs (managers, professionals in different fields) in different types of bodies (ministries, agencies, regional units, Marzes) and does not consider any empirical evidence as to whether these jobs are, in fact, at risk (i.e. high resignation rates, low number of qualified applicants, inability to fill vacancies). Targeting only certain grades creates a gap in the structure of base pay, and also ignores the role of bonuses that – as described earlier – can play a significant role in total earnings. Finally, by looking only at the civil service, this approach ignores the concerns about inconsistent and uncoordinated pay regimes across Armenia‘s civil service and parallel public services. 4.34. The second model shows that reducing employment by 5 percent in the 5 largest pay categories and increasing the base salary by 5$ the following year will reduce employment by 4 percent while increasing cost by 0.5 percent in 5-year time. Simple trade-offs between employment reductions and salary increases are often proposed in countries that are struggling to keep employment levels and wage bills in check while addressing perceptions of inadequate salaries. Thus the second model simulates a reduction in employment in certain grades, followed by an increase to the base pay cornerstone that raises salaries for all remaining staff while keeping wage bill costs (excluding bonuses) near the pre-staff cut level. Specifically, in year 2, employment in grades that contain more than 100 staff – grades 4.1, 3.2, 3.1, 2.3 and 2.2 – is reduced by 5 percent. In year 3, the cornerstone is raised from 40,000 to 42,000 AMD (US$105-111). The 5 percent employment reduction in these grades results in a 4 percent fall in overall employment. The costs of base pay and additions as well as employer contributions fall by 4 percent relative to the status quo for year 2 (which assumes no cuts to employment or change in the cornerstone). However, the cost of any severance payments must be incorporated into any assessment of savings. With the increase to the cornerstone in year 3, cost of base pay and additions rises to levels slightly higher than the year 3 ―status quo‖; employer contributions are slightly lower, reflecting the fact that the staff cuts were in higher-paying grades with correspondingly higher contribution rates. In subsequent years, costs increase, reflecting a slight flux in staff numbers as well as the movement of staff up the pay scale due to the automatic time-based progression through steps. The ―trade-off‖ between lower employment levels and higher base pay adds 2,000 ADM (about $5) to monthly base pay at the bottom of the scale and 11,780 ADM (about $31) to the top grade (which affects only 1 person in the pilots). 4.35. As with the decompression model, the trade-offs model raises a number of issues. First, the apparent ―savings‖ depend on the types of jobs that are eliminated: a given number of reductions in higher grades (with higher base salaries) free up more funds, but also risk eliminating skilled professionals; in contrast, the same number of reductions in lower grades will free up fewer resources. This numbers-driven approach neglects important substantive questions related to higher objectives that, presumably, include improving administrative performance. How will staff cuts affect the fulfillment of organizational objectives? Is there any evidence that these particular posts are unnecessary? In contrast to targeting specific grades or distributing reductions across all bodies, it may be that sustainable change to employment comes, instead, from policies that, as part of achieving substantive objectives, shift staffing needs gradually over time (as in the case of the education sector and teacher employment) or the removal 46 of entire functions from the public sector‘s mandate. At the very least, a careful review of the functions – as is currently being done for seven pilot bodies in Armenia under the Public Sector Modernization project – is a much more solid basis for considering how to adjust staffing than simple reductions targets such as those modeled here. As with the first example, this model too ignores pay relativities with other public services, the role of bonuses, and does not use empirical evidence to inform and refine policy measures. 4.36. While the modeling exercises presented here are, admittedly, simplifications of complex dynamics, they highlight potential pitfalls and information needs that deserve attention. Many of these can be addressed with analytical measures that are well within the Government‘s reach. Briefly, these include:  Using ―actionable indicators‖ to better identify specific positions where attracting and retaining staff is a problem. Example indicators are resignation rates by job title/grade or profession, numbers of applicants for advertised vacancies by job title/grade or profession). Existing data (from ministries or the Civil Service Council) or pilot studies can be used to construct such indicators.  Conducting regular analyses of total compensation that take into account regular monetary compensation, bonuses and benefits as well as non-pecuniary factors (job security, work environment). These would be used to compare government earnings with private sector and other alternative labor markets. Two different types of analyses are recommended – one comparing earnings for similar jobs (including realistic alternatives, not elite organizations) and another comparing similar workers (with comparable levels of education).  Including parallel public services in the analysis, again with a focus on actual total compensation for specific jobs. C. HEALTH EXPENDITURES 4.37. Value for money will be ensured in the health sector if public expenditures help to: (1) to improve population health outcomes and (2) prevent households from falling into poverty because of ill health. The first objective of any health system should be to reduce mortality and morbidity, usually with a particular focus on the health outcomes of the poorest segment of the population. The second objective aims to prevent people from facing medical bills that are either ‗impoverishing‘ (driving them below the poverty line) or ‗catastrophic‘ (exceeding some threshold of household resources) as a result of an episode of ill health. These two overarching health system goals correspond to the objectives highlighted in major international reports.23 4.38. Health expenditure patterns in Armenia should be reviewed through the lens of these two major objectives. We first provide an overview of health expenditure in Armenia, including major sources of funds and the main spending categories. We then focus on the achievements and shortcomings of health expenditure with regard to long-term trends in health outcomes in the country, before examining the interplay between public spending on health and the degree to which households are protected against high out-of-pocket spending. Finally, we discuss potential avenues for achieving better value for money in the health system. 23 For example, the WHO‘s World Health Report (2000) and the World Bank‘s Health, Nutrition, and Population sector strategy (2007). 47 Overview of Health Expenditure in Armenia 4.39. Total health spending in Armenia is within the range commonly seen in lower-middle income countries around the world, but it is an outlier with respect to the balance between public and private sources. Almost all low and middle-income countries spend about 4 to 8 percent of GDP on health in total (public and private). With total health spending of about 4.6 percent of GDP, Armenia is at the lower end of this range. However, Armenia is an outlier with respect to the share of total health financing from private out-of-pocket sources, which at about 66 percent is roughly twice the norm within the CIS region. Figure 4.6 shows some international comparisons. Armenia‘s health spending profile is part of a global pattern whereby the share of health spending that is not covered by the government budget will be mostly out-of-pocket. Figure 4.6: Public and Private Health Spending as a Share of GDP, Europe and Armenia 12% Private Public 10% 2.2% 8% % of GDP 6% 1.7% 2.3% 4% 7.5% 3.0% 4.7% 2% 3.2% 1.6% 0% Source: WHO and WB EU-15 EU-10 CIS Armenia Source: WHO and World Bank. 4.40. The government budget accounts for about one-third of total health spending, and is largely administered by the State Health Agency (SHA) through its contracts with providers. The equal balance between inpatient and outpatient care represents an important achievement in terms of allocative efficiency that has been sustained for several years, and contrasts favorably with some other ECA countries where inpatient spending predominates. Estimates of out-of-pocket spending (Table 4.11) are based on new data and are significantly higher than past estimates. This largely reflects higher drug spending captured in the ILCS. 4.41. Health spending in Armenia has also consistently fallen short of medium-term expenditure projections. Figure 4.7 shows the gradually declining share of the government budget allocated to health, from nearly 8 percent in 2005 to about 6 percent in 2010-11. This compares to a CIS average of over 8 percent; an ECA average of nearly 12 percent; and an EU-15 average of over 15 percent. This is despite a series of MTEFs (and the 2008 Sustainable Development Program) in which the health share was expected to increase. 48 Table 4.11: Total Health Spending in Armenia, 2010 Estimates Category 2010 spending (AMD) % of GDP Government budget 55.2 billion 1.6% In-patient care (SHA) 21.2 billion Out-patient care (SHA) 21.0 billion Other programs (MoH) 13.0 billion Drug spending (OOP) 45.9 billion 1.3% In-patient spending (OOP) 29.1 billion 0.8% Out-patient spending (OOP) 32.3 billion 0.9% TOTAL 162.5 billion 4.6% Source: Ministry of Finance, Ministry of Health and World Bank Staff calculations. Figure 4.7: Health Spending in Armenia: Actual and Projected 14% Health spending as share of budget 12% 2005-07 10% 2006-08 8% 2007-09 6% 2008-10 2011-13 4% SDP 2% ACTUAL 0% 2005 2006 2007 2008 2009 2010 2011 2012 2013 Source: Ministry of Finance and Ministry of Health. Health Spending and Health Outcomes 4.42. Over the last several decades, Armenia’s health outcomes have improved little relative to other regions of the world. Figure 4.8 shows the historical trends of adult mortality (defined here as the probability of death between the ages of 15 and 60) in Armenia and selected comparator regions. In 1970, adult mortality in Armenia was close to the level in Western Europe, and well ahead of East Asia and Latin America. However, since then, outcomes in Armenia have changed very little, while the middle-income regions have caught up and Western Europe has continued to make steady progress. A similar picture emerges for related indicators such as life expectancy. In 1970, Armenia was about two years behind Western Europe and ten years ahead of East Asia; it is now about eight years behind Europe and on par with Asia. Moreover, health outcomes tend to be worse among the poorer segments of the population. 4.43. International experience suggests that medical care becomes an increasingly important contributor to improved health outcomes in middle and higher-income countries. At earlier stages of development, major health gains typically arise from nutrition, public health measures such as clean water, sanitation, and immunization, and safe child delivery. Armenia had made significant achievements through such interventions by the 1960s. On the other hand, most of the improvements in health outcomes during the late 20th century in Western Europe and the US have been due to tobacco cessation and especially to improved medical care in the area of cardiovascular disease, neo-natal health, and certain cancers. 49 Figure 4.8: Probability of Dying Between 15 and 60, Armenia and Comparator Regions, 1970-2008 250 East Asia Latin America Armenia 200 Adult mortality (15-60), per 1000 Western Europe 150 100 50 0 1970 1990 2008 Source: Rajaratnam et al. (2010) Source: Rajarathnam et al. (2010). 4.44. In Armenia, low levels of health care utilization reflect low public spending on health, leading in turn to stagnant health outcomes. Overall utilization rates in Armenia, whether measured as outpatient contacts per person or acute care hospital admissions per 100 people, are about half the level observed in Western and Eastern Europe or the CIS. With regard to specific interventions, Table 4.12 shows the currently low coverage levels in Armenia of the five interventions that have had the largest impact on improved life expectancy in the West since the 1960s. Although not all of these interventions – such as wide coverage of heart bypass operations – may be affordable for Armenia at present, many of the others (e.g., basic pharmaceuticals) could be covered if Armenia devoted a larger share of its budget to the health sector. Meanwhile, raising cigarette taxes closer to the level prevailing in neighboring countries, or enforcing a ban on smoking in public places, are among the most cost-effective interventions available to improve health outcomes. Rapid approval and implementation of anti-tobacco legislation would be an important first step in this regard. Table 4.12: Coverage of Key Health Interventions in Armenia is Low Five Interventions that Have Had Largest Impact Assessment of Implementation In Armenia on Life Expectancy In West Anti-tobacco policies Cigarette tax (as % of price) is much lower in Armenia (cigarette taxes, smoking bans, etc.) than Europe, Turkey, or Georgia; see Figure below Cardiovascular disease drugs About 93% of total drug spending is out-of-pocket; about (ACE-inhibitors, beta-blockers, statins, etc.) 80% of hypertension cases are undiagnosed Cardiac surgeries About 250 people covered by SHA in 2009, or a very (angioplasty, bypass) small fraction of total need Early diagnosis and treatment of breast and colon No coverage of diagnostic tests in basic benefit package; cancer see Figure below Neo-natal intensive care Few hospitals have capacity to provide this; high OOP by family is often required Source: World Bank Staff. 50 4.45. Escaping the long-term trend of stagnant health outcomes in Armenia will require an expanded benefit package for key interventions beyond the currently minimal coverage. Figure 4.9 shows the breakdown of SHA inpatient spending by category as of 2009. Most spending is currently directed to maternal and child health, infectious diseases (including tuberculosis), emergencies (such as trauma, appendicitis, or heart attack cases), and life-saving hemodialysis. These may be considered the ‗minimum‘ that an inpatient healthcare benefit package should cover, and even here there have been significant out-of-pocket payments by the population to complement SHA payments. There are few obvious areas for potential savings in the current benefit package, although a couple of issues – how TB hospitals are paid and the disability programs – are discussed further below. Figure 4.9: SHA budget allocation by inpatient state order program, 2009 35% 28.9% 30% 25% % of SHA hospital spending 20% 14.4% 15% 13.8% 10.2% 10% 7.5% 6.2% 4.8% 5% 3.4% 2.2% 2.1% 1.9% 1.1% 1.0% 0.8% 0.6% 0.6% 0.5% 0% Source: Armenian authorities. Health Spending and Poverty 4.46. Concerns about large household expenditures for health care stem from the fact that such spending is typically different from most other household purchases. At least three reasons stand out. First, the spending is often not ‗voluntary‘ (for example, if arising due to an unwanted health shock), and may have more to do with overcoming illness than adding to a household‘s overall consumption or living standards, as is usually the case with other goods. Second, the uncertainty and potentially high cost associated with health expenditures (we often do not know when we will become sick and how much it will cost if we do) make them amenable to prepayment and risk-pooling arrangements. Third, access to health care is often viewed as a ‗merit good‘ which should not be determined primarily by ability to pay. For these reasons, a more desirable counterfactual to high out-of-pocket spending on health would be some form of pooling mechanism (whether through general taxes or a contributory insurance scheme), as well as cross-subsidization, to provide financial protection against health shocks. This has been achieved in many countries, but not yet in Armenia. 4.47. High out-of-pocket payments for health may cause a household to fall below the poverty line. If a household has total consumption expenditures (pre-OOP) above the national poverty line, but their total non-medical spending (post-OOP) is below the poverty line, they could be considered to have 51 suffered impoverishment due to OOP for health (Figure 4.10). Households are ranked along the horizontal axis by total consumption. The vertical drip lines represent OOP for health, and the poverty threshold is indicated by the horizontal line. Applying this approach to household survey data, it has been estimated that an additional 6 percent of Armenian households were poor as a result of OOP for health.24 Consistent with Armenia‘s high reliance on OOP as shown above, this level of poverty impact is notably higher than most other countries either in the region or around the world (Xu et al, 2003). 4.48. Addressing the problems associated with high out-of-pocket spending and poor outcomes for health in Armenia will require tackling the major causes, including the relatively low proportion of the Government’s budget for health. A detailed study comparing the actual cost of services with SHA prices found that SHA reimbursements of services covered under the basic benefit package are frequently inadequate to achieve cost recovery by those (hospitals/doctors) who provide care, which results in ―informal‖ OOP by patients. However, OOP also occurs because the population demands health goods and services that are not included in the basic benefit package, such as out-patient pharmaceuticals. In both cases, the way to address this source of OOP is through a larger health budget, subject to the availability of additional funds in a difficult fiscal environment. 4.49. An additional reason for out-of-pocket spending in some cases is rent-seeking by providers. In any health system, individual patients are usually in a weak position to negotiate with hospitals or doctors, because (i) they typically do not have the knowledge or expertise to diagnose and treat their illness, nor to judge the quality of care received; and (ii) in a state of illness they are usually not in a position to shop around for lower prices. As a result, providers can take advantage of their superior knowledge and advantageous position to extract higher payments from individual patients. Figure 4.10: The Impact of Household Spending on Health on Poverty 140 Household consumption per adult equivalent (AMD 000s per month) 120 100 80 60 40 Poverty line 20 0 1 251 501 751 1001 1251 1501 1751 2001 2251 2501 2751 3001 3251 3501 3751 Households ranked in ascending order of total consumption Source: ILCS and WB calculations 24 Whether or not this is an accurate way to evaluate the true poverty impact of OOP is a matter for debate, since in reality households are likely to draw on several possible coping mechanisms that would allow for consumption smoothing, such as drawing down savings, borrowing, or selling assets. Thus, a costly illness episode in one period would not necessarily have an immediate and commensurate impact on total consumption in the same period. Against this, families who experience not a catastrophic but a chronic shock to health care costs may over time exhaust all their reserve stocks and not show up among the population we identify here. Nevertheless, the concept can be useful for international comparisons of financial protection. 52 Figure 4.11: Eliminating Informal OOP: Impact of the Maternity Voucher Program 60000 Maternity voucher program introduced 50000 Average payment for delivery (AMD) 40000 30000 20000 10000 0 Source: ILCS 4.50. There are at least three possible approaches to address rent-seeking by providers. The first is through better information on hospital finances, in particular through mandatory financial audits. These have long been discussed in Armenia but have yet to be implemented on a wide scale. A second approach is by providing better information to patients themselves. The official co-payment policy introduced in February 2011 (discussed further below) should help in this regard. The successful maternity voucher program introduced in July 2008, which included the availability of hotline phone numbers and public information campaigns, is also a very good example of how to reduce the opportunities for rent-seeking. Thirdly, rent-seeking by health care providers is addressed de facto in mature health systems through purchasing power: by pooling resources for health care purchasing at SHA, the ability of providers to extract higher prices for their services will be weakened. 4.51. Armenia already has a successful demonstration of how to reduce out-of-pocket payments due to both cost recovery and rent-seeking in the form of its maternity voucher program. This program roughly doubled the SHA reimbursement of hospitals for child deliveries, and stipulated how these additional funds should be allocated within hospitals (to ensure that they were not fully captured by head doctors). It also undertook key ‗demand-side‘ steps to achieve a well-informed population with avenues for recourse (via an active phone hotline service) in the event that doctors did try to extract informal payments. Figure 4.11 shows the impact of this initiative on reported OOP for child deliveries, with a steep decline soon after the July 2008 launch. A similar program for pediatrics was introduced in early 2011. As additional funds become available, this provides a strong model for how to reduce OOP for health going forward. Some Policy Issues to Consider 4.52. In a constrained fiscal environment with limited funds for the health sector, one potential spending priority would be to achieve better targeting of the poor. At present there is a state order program for socially vulnerable groups, including three categories of disabled groups and households that are classified as poor according to a proxy means test (the same one used to determine eligibility for the family poverty benefit program, or FPB). However, as indicated in Figure 19, a relatively small portion of these funds are allocated to the poor. Moreover, the cut-off score to qualify as poor under this health program is 36, which includes about 100,000 individuals, whereas the eligibility threshold for the FPB 53 program is 30, which includes about 425,000 individuals. Thus, qualification under the health program is much more restrictive. 4.53. Various options exist for improving financial access to health care services among the poor. Perhaps the best option would be to unify the eligibility criteria for gaining vulnerability status under health programs with the one used for the family poverty benefit program – i.e., a score of 30 and above instead of 36 and above. This would raise the number of individuals eligible for exemption to about 425,000. Estimates suggest this might cost about AMD 4 billion (US$11 million) per year, or about 5 percent of the annual health budget and within the typical annual increment. Thus, it is affordable if it is made into a key priority. 4.54. Among the major state order programs included in the basic benefit package financed by SHA, the rationale for disabled categories II and III deserves the closest scrutiny. This is a sub- category of the broader program to cover the socially vulnerable, including the poor and category I disabled. Due to problems with the disability certification process, it is widely known (and survey data confirm) that many of those who obtain category II or III status are not necessarily any more deserving of subsidized health care than the average individual or household. Total spending on these groups accounts for about AMD 2 to 3 billion (US$5-8 million) annually, or about 5 percent of the total health budget. Restructuring the program will require the consensus of the Ministry of Labor and Social Protection. Savings could be re-invested in the same program by expanding coverage of the poor. 4.55. In order to address the persistent problem of informal payments for health care, the Government launched a formal co-payment policy nationwide in February 2011. At present, it only applies to two major state order programs, emergency and gynecology. On average, the co-payment level is AMD 64,000 (US$168) or 40 percent of the full price (SHA price plus co-payment) for Marz hospitals: about AMD 120,000 (US$316) or 55 percent of the full price for Yerevan general services, and AMD 153,000 (US$403) or 55 percent of the full cost for Yerevan specialist services. For emergency services, there are in total 16 co-payment levels (in AMD) in the Marzes, 10 for Yerevan general services, and 20 for Yerevan narrow specialist services. In gynecology, there are 3 different co-pay levels within Marz and Yerevan hospitals. The introduction of co-payments was accompanied by new stipulations on how revenues should be distributed to salaries within hospitals, similar to what was done when SHA prices were increased under the maternity voucher program in 2008. For the population, fee schedules have been posted in all hospitals, a public information campaign has been conducted and a phone hotline established. 4.56. Implementation details highlight a number of issues that warrant attention going forward. First, the co-payment levels are quite high, ranging between approximately 40 and 60 percent of total cost. Elsewhere a co-payment would be a much smaller percentage of the total cost. This approach was necessitated by the decision not to significantly increase SHA prices for these programs in 2011, while aiming to eliminate informal payments altogether (i.e., by filling the gap between SHA reimbursement and estimated costs). In essence, the co-payment policy means that informal barriers to access have been replaced by more predictable, transparent, and formal barriers to access. This highlights the medium-term challenge to increase hospital funding as the fiscal environment permits, in order to correspondingly lower co-pay levels. Second, the number of actual co-pay amounts across specific services is quite high, implying a lengthy fee schedule, which may cause confusion. It will be important to ensure proper communication to patients and the population in order to minimize opportunities for manipulation by providers. In the future, a more streamlined fee schedule may be considered. Third, while emergency services that require intensive care are excluded from co-payments (e.g., heart attack, appendectomy), it will be important to monitor implementation. It is crucial that the population is not deterred from seeking emergency care when needed due to affordability issues, as this could jeopardize health outcomes. 54 4.57. Despite the extensive preparations to introduce the co-payments, several uncertainties affecting implementation will merit close monitoring and evaluation. Most significantly, the behavioral response of providers to the new funding environment (i.e., SHA prices and official co- payments, from which previously unpaid taxes will be subtracted) and the price elasticity of patient demand are both uncertain. These will in turn affect: (i) whether providers face higher or lower (net) revenues as a result of the co-payment regime; (ii) whether patients face higher or lower overall OOP; and (iii) whether utilization of services increases or decreases. If providers believe that their overall revenues have declined as a result of the co-payment policy, they may continue to seek OOP from patients, who will then feel that they are being asked to ―pay twice‖. In a worst-case scenario, already low utilization rates could fall. The WHO has developed a simulation tool to model these potential impacts, and a joint WHO-World Bank survey is being undertaken to help gather information on trends. These can be used to inform adjustments to the policy going forward. 4.58. Moving forward, there are a few remaining state order programs that are suitable for co- payment introduction. The obstetrics program is already largely free of informal payments thanks to the July 2008 reform discussed earlier (higher SHA prices and clear information for the population). A similar program was implemented for pediatrics as of January 2011. Programs for tuberculosis, infectious diseases and STIs are not good candidates due to the classic issue of externalities. Hemodialysis is an expensive and life-saving program for which significant co-payments could be prohibitively expensive for most patients. Remaining programs for which co-payments may be appropriate in the short-term are oncology (although improved access to early diagnosis of certain cancers should be a priority from a health outcome standpoint), and the program for category II and III disabled groups. 4.59. As few out-patient drugs are covered by the basic benefit package, a survey of retail pharmacy prices and availability for 43 common drugs was conducted in 2010 to assess the affordability of medicines. Overall, the survey results suggest that access to low-cost drugs is a problem in Armenia for reasons of both price and availability, for which various policy options may be explored. One option would be to regulate drug prices, as is done in all EU countries, where combined wholesale and retail margins are usually set at about 30 to 40 percent. However, this may not be helpful in Armenia if import prices are already relatively high. Ongoing efforts to obtain customs price data may help clarify this issue. This would also help shed light on wholesale/retail mark-ups. In general, price regulation of drugs would also require a robust institutional structure and strong information flows. 4.60. Probably the best policy option for improving access to drugs would be to add a core list of (generic) out-patient drugs to the basic benefit package, and use government purchasing power to help obtain lower prices. As noted earlier, OOP by individual patients makes it difficult to obtain low prices for health care goods and services, and this is equally true for drugs and medical care. Government purchasing does not need to entail direct procurement, but it could also take the form of negotiated prices with pharmacies and reimbursement of patients. A generic drug policy should be pursued in tandem with measures to improve the prescribing behavior of physicians. Better access to drugs by adding them to the basic benefit package would help to lower OOP. It would also help improve outcomes because a certain family of drugs, particularly for cardiovascular disease management, has proven to be among the most cost-effective interventions for improving health. If Armenia chooses to explore this option, the experience of Kyrgyzstan‘s outpatient drug benefit package may be a useful case study for learning lessons. 4.61. A comprehensive review of healthcare spending indicates that this is not an area where reductions in public expenditure should be sought. There certainly is some inefficient public expenditure on healthcare, most notably, perhaps, in the area of disability payments. However, Armenia‘s stagnant and mediocre health outcomes, already high private OOP expenditure, and low service utilization rates all underline the risks inherent in cutting public spending on health. Indeed, in addition to some reallocation of resources within healthcare spending this review has highlighted some notable gaps 55 in public healthcare spending, most clearly with respect to reaching the poor and in ensuring the availability of a core of essential generic pharmaceuticals at reasonable cost. Austerity applied to public healthcare spending would very likely harm the provision of services in an area that is receiving little support, by international standards, and that is vital for social welfare. D. EDUCATION EXPENDITURES 4.62. Armenia’s enrollment rates at the basic level are close to 100 percent with little geographical or socio-economic diversity. In 2009 the net enrollment rates for primary education were 99.0 percent for the poorest decile and 98.6 percent for rural areas; and for lower secondary education, 96.2 percent and 99.4 percent, respectively. The enrollment rates drop, however, after the lower secondary level, particularly among poor children. The net enrollment rates for upper secondary education ranged from 54 percent for the poorest decile to 86 percent for the richest or 53 percent for rural areas while 75 percent for urban areas. The net enrollment rates for tertiary education further widened to 7.4 percent for the poorest, while they are 33.4 percent for the richest. 4.63. Despite the introduction of the per capita financing mechanism in 2003, which resulted the merger of small classes and fewer teachers, Armenia still has a low student-teacher ratio. Armenia, like many other countries in the region, has experienced a dramatic birth rate decline since the early 1990s. The student-teacher ratio and average class size for general secondary education continued to drop until the early 2000s due to these demographic factors. Both the student-teacher ratio and average class size increased slightly between 2003 and 2005 as a result of the introduction of the per capita financing (PCF) mechanism and school rationalization program as schools with small classes were compelled to merge classes and reduce the number of teachers. However, in recent years the problem has increased once again, leaving the system with an average class size (ACS) of 17.7 and a student-teacher ratio (STR) of 9.8 in 2009, down from 21.6 and 11.5, respectively, in 2006. Armenia‘s ACS and STR are considerably lower than OECD averages – 20.8 and 16.4 for primary and 21.6 and 13.7 for secondary education, respectively. A key reason for the extremely low STR is the government‘s explicit policy to maintain small schools in remote areas. 4.64. As the Government was sharply cutting overall education spending in 2009, one notable exception was the substantial rise in per capita financing (PCF) formula allocations to general secondary schools. In 2009, the formula‘s per-school and per-student allocation coefficients were raised substantially – by 35 and 34 percent, respectively–double the previous year‘s increase. This occurred because of the Government‘s inability, or unwillingness, to reduce the expenditure on school personnel costs in response to the crisis. Personnel costs constitute more than 90 percent of the total recurrent budget for schools. As a result, the proportion of the PCF in total school funding increased from 52-53 percent of total sector spending in 2007 and 2008 to 65 percent in 2009 (Figure 4.12). A year later, in 2010, however, the Government took several measures to reduce the schools‘ recurrent budgets. The funding formula was adjusted with a 13 percent decrease in the per-school coefficient and a flat per- student coefficient; the Government also froze the average teacher salary at the 2009 level. 56 Figure 4.12: Composition of Armenia Public Spending on Education, 2009 (a) By Subsector (b) By Economic Category Primary PCF Lower education allocations secondary 18% to schools education 65% 35% Upper secondary education Miscellaneo 14% Capital us education Higher and Vocational expenditures Other activities postgraduat education 15% recurrent 22% e education 4% expenditures 7% 20% (c) By Spending Authority Marz authorities 67% Other Ministry of ministries Ministry of Education and agencies Urban and Science 6% Development 17% 10% Source: World Bank, Armenia BOOST government expenditure database. 4.65. A closer look at inputs and outcome of general secondary education and the relationships between spending, student performance, and various socio-economic factors is needed to assess the efficiency of education spending. Indeed, the 2011-2013 MTEF shows the education budget continuously declining from 3.0 percent of GDP in 2010 to 2.7 percent in 2013 or from 10.3 percent of total government expenditure in 2010 to 9.8 percent in 2013. This is a reversal of the Government's long- term vision for developing the education sector. However, the Government may wish to take the tight budgetary constraints it faces in the coming years as an opportunity to review the composition of education spending and allocate it more efficiently and effectively. For instance, any saving from efficiency gains in general education could be reallocated to the tertiary or pre-school education sectors currently severely underfunded. Efficiency and Equity of Education Spending 4.66. The main distinction in educational efficiency within Armenia can be drawn between small rural schools and large urban ones. As shown in Table 4.13, 63 percent of Armenia‘s schools, 57 enrolling 29 percent of the students, have 300 or fewer students each. These institutions receive 42 percent of all PCF funding and employ 43 percent of staff in general secondary education. Yet the utilization of educational inputs in these schools is strikingly inefficient. For example, among the 369 schools with 100 or fewer students (27 percent of all schools in Armenia), the average class size is 5.6 with 3.7 students per teacher. These schools utilize only 27 percent of their available building capacity, compared to 56 percent nationwide. Table 4.13: Key Statistics for State General Secondary Schools, 2009/10 Summary Statistics Summary Statistics Efficiency Measures PCF Number Number Number Number Number Non- Number Number Number Non- Allocation Average Average Student- % Non- % of of of of of Teaching of of of Teachin per Student School Class Teacher Teachin Capacity Teacher Teacher Schools Students Classes Staff Schools Students Classes g Staff (AMD Size Size Ratio g Staff Utilization s s '000) Republic of Armenia 1,368 381,286 21,546 38,984 16,889 100% 100% 100% 100% 100% 167 279 17.7 9.8 30% 56% By marz Aragatsotn 123 19,321 1,746 2,878 1,205 9% 5% 8% 7% 7% 221 157 11.1 6.7 30% 54% Ararat 112 35,238 2,152 3,263 1,660 8% 9% 10% 8% 10% 163 315 16.4 10.8 34% 55% Armavir 119 36,889 1,787 3,671 1,578 9% 10% 8% 9% 9% 162 310 20.6 10.0 30% 50% Gegharkunik 126 33,781 2,024 3,689 1,815 9% 9% 9% 9% 11% 187 268 16.7 9.2 33% 48% Kotayk 104 34,655 1,622 3,331 1,554 8% 9% 8% 9% 9% 156 333 21.4 10.4 32% 55% Lori 163 33,341 1,962 3,901 1,687 12% 9% 9% 10% 10% 188 205 17.0 8.5 30% 51% Shirak 166 34,313 2,612 3,896 1,658 12% 9% 12% 10% 10% 188 207 13.1 8.8 30% 68% Syunik 121 16,996 1,200 2,342 909 9% 4% 6% 6% 5% 224 140 14.2 7.3 28% 42% Tavush 81 16,909 957 2,071 812 6% 4% 4% 5% 5% 189 209 17.7 8.2 28% 47% Vayots dzor 51 7,550 520 1,044 399 4% 2% 2% 3% 2% 225 148 14.5 7.2 28% 43% Yerevan 202 112,293 4,964 8,898 3,612 15% 29% 23% 23% 21% 131 556 22.6 12.6 29% 69% By community type Rural 849 147,544 10,846 18,500 8,418 62% 39% 50% 47% 50% 213 174 13.6 8.0 31% 49% Urban 519 233,742 10,700 20,484 8,471 38% 61% 50% 53% 50% 139 450 21.8 11.4 29% 62% Highly mountainous 197 40,226 2,914 4,968 2,290 14% 11% 14% 13% 14% 209 204 13.8 8.1 32% 47% Mountainous 189 30,683 2,215 4,005 1,836 14% 8% 10% 10% 11% 216 162 13.9 7.7 31% 47% Non-mountainous 982 310,377 16,417 30,011 12,763 72% 81% 76% 77% 76% 157 316 18.9 10.3 30% 59% Not only school in 635 283,232 13,156 24,947 10,415 46% 74% 61% 64% 62% 140 446 21.5 11.4 29% 62% community Only school in a community 733 98,054 8,390 14,037 6,474 54% 26% 39% 36% 38% 247 134 11.7 7.0 32% 44% with < 400 students By school size 1-100 students 369 18,445 3,318 5,010 2,294 27% 5% 15% 13% 14% 444 50 5.6 3.7 31% 27% 101-300 students 491 93,096 6,524 11,434 5,291 36% 24% 30% 29% 31% 203 190 14.3 8.1 32% 47% 301-500 students 295 115,524 5,471 10,733 4,768 22% 30% 25% 28% 28% 150 392 21.1 10.8 31% 55% 501-700 students 131 76,003 3,616 6,251 2,601 10% 20% 17% 16% 15% 132 580 21.0 12.2 29% 69% 701 or more students 82 78,218 2,617 5,556 1,935 6% 21% 12% 14% 11% 120 954 29.9 14.1 26% 86% Sources: NSS, NaCET, ATC, and MoF. 4.67. In 2009/10, rural schools had an average of 174 students, with an average class size of 13.6 and a student-teacher ratio of 8.0. In contrast, urban schools educated an average of 450 students, grouping them 21.8 per class and 11.4 per teacher. Rural schools have been able to keep smaller classes because they have ample funding (rural schools receive 53 percent more per student than urban ones) and do not run into infrastructure constraints (nationwide, rural schools utilize less than half of their building capacity). Similar trends are evident when comparing schools in mountainous/highly mountainous locations to non-mountainous ones, as well as the only schools in small communities versus all other ones. Figure 4.13 further illustrates that the distribution of educational inputs, as well as the efficiency and equity of education spending, varies considerably across schools located in different types of communities and with different numbers of students. The most striking finding is that schools in the smallest communities receive four times as much funding per student as schools in Yerevan. They have one-twelfth as many students as an average school in Yerevan, four times as many teachers and classes for a given number of students, and utilize less than half as much of their physical capacity. Also, the proportion of teachers with higher education varies considerably between rural and urban schools. 58 Figure 4.13: Distribution of Educational Inputs by Community Type and by School Size, 2009/10 (a) Average PCF Allocation per Student (b) Average School Size (000s AMD) (Number of Students) 300 500 250 400 200 300 150 200 100 50 100 0 0 Republic of Armenia Republic of Armenia (c) Average Class Size (d) Student-Teacher Ratio 25.0 12.0 20.0 10.0 8.0 15.0 6.0 10.0 4.0 5.0 2.0 0.0 0.0 Republic of Armenia Republic of Armenia 59 (f) School Capacity Utilization (e) % of Teachers with Higher (Student Enrollment as % of Total Education Building Capacity) 100.0% 70.0% 60.0% 90.0% 50.0% 40.0% 80.0% 30.0% 70.0% 20.0% 10.0% 60.0% 0.0% Republic of Armenia Republic of Armenia (a) Average PCF Allocation per Student (b) Average School Size (000s AMD) (Number of Students) 500 1200 400 1000 800 300 600 200 400 100 200 0 0 1-100 101-300 301-500 501-700 701 or more 1-100 101-300 301-500 501-700 701 or students students students students students students students students students more students Republic of Armenia Republic of Armenia (c) Average Class Size (d) Student-Teacher Ratio 35.0 16.0 30.0 14.0 12.0 25.0 10.0 20.0 8.0 15.0 6.0 10.0 4.0 5.0 2.0 0.0 0.0 1-100 101-300 301-500 501-700 701 or 1-100 101-300 301-500 501-700 701 or students students students students more students students students students more students students Republic of Armenia Republic of Armenia 60 (f) School Capacity Utilization (e) % of Teachers with Higher (Student Enrollment as % of Total Education Building Capacity) 100.0% 90.0% 80.0% 90.0% 70.0% 60.0% 50.0% 80.0% 40.0% 30.0% 70.0% 20.0% 10.0% 60.0% 0.0% 1-100 101-300 301-500 501-700 701 or 1-100 101-300 301-500 501-700 701 or students students students students more students students students students more students students Republic of Armenia Republic of Armenia Source: World Bank calculations based on data from NSS, NaCET, ATC, and MoF. Education Spending and Student Performance 4.68. Despite low level of public spending on education, Armenian students score well on standardized international assessments. In spite of below-average government spending on education Armenian students outperformed most of their regional peers on the Trends in International Mathematics and Science Study (TIMSS) in 2007, showing vast improvements from the 2003 round of assessments, as shown in Figure 4.14. Given the modest levels of public financing and better-than-average student performance, Armenia‘s education sector as a whole appears to be relatively efficient.25 Figure 4.14: TIMSS Performance in Selected ECA Countries 560 (a) TIMSS Mean Score 560 (b) TIMSS Mean Score 8th grade mathematics 8th grade science 520 2003 520 2003 2007 480 2007 480 440 440 400 Bosnia and… 400 Bosnia and… Ukraine Lithuania Serbia Romania Turkey Bulgaria Latvia Moldova Armenia Russian Federation Georgia Macedonia, FYR Bulgaria Romania Ukraine Turkey Lithuania Serbia Moldova Armenia Latvia Georgia Russian Federation Macedonia, FYR Source: World Bank, EdStats database. 4.69. Unfortunately, the relatively high student performance is not evenly observed throughout the country. Despite the lower average class size and higher per-student spending, students in rural 25 Private investment in education in Armenia increased from 0.9 percent of GDP in 2001 to 1.9 percent in 2005 (World Bank, 2008). The figure is more likely to have increased in recent years as private tutoring is a key to succeed in university entrance examinations. The large private spending on education might be part of the reasons why Armenia performs relatively well despite the low public spending. 61 schools are significantly less likely to take and pass the unified university entrance examinations (which are taken by students in grade 11 who wish to enroll in university) and score lower on these examinations than students in urban schools. The very low participation rates in the entrance examinations also suggests that the anticipated direct and indirect costs of attending higher education may be a discouraging factor for many students in rural areas. The same pattern is true for students in small schools with less than 300 students, single schools in a community, and schools located in mountainous and highly mountainous areas. Particularly striking is that only 9 percent of all students graduating from Armenia‘s smallest schools (fewer than 100 students) took and passed the entrance examination in mathematics in 2010. As a comparison, the proportion for graduates of the largest schools (701 or more students) is 24 percent, and 15 percent for the nation as a whole. 4.70. Further analyses on student performance by various socioeconomic characteristics suggest that students in communities with the highest levels of adult education and wages perform better than children who come from less-educated and lower-paid families. However, only those in the top- quintile schools significantly outperform the national average on university entrance exams; students in the bottom four community wage quintiles and bottom three community education quintiles perform significantly worse than the country as a whole. A slight positive relationship is also observed between the proportion of ethnic and linguistic minorities in the community and student performance, though children in the top and bottom quintiles of minority communities score lower than those in the middle of the distribution. Table 4.14: Summary of the Determinants of Academic Performance Performance Measure1/ Educational Input (a) (b) (c) (d) (e) (f) (a) (b) (c) (d) (e) (f) Larger classes + + 0 0 + 0 + + 0 0 + 0 More educated teachers + + + 0 + + + + + 0 + 0 More part-time teachers 0 0 0 + 0 0 + 0 + + 0 + More experienced teachers 0 0 0 0 0 0 0 0 0 + + + Older teachers – 0 0 0 0 0 – 0 0 0 0 0 More non-teaching staff 0 0 0 0 0 – 0 0 0 0 0 0 More computers 0 0 + 0 0 0 0 + + 0 + 0 Larger PCF allocation + + 0 0 0 – + + 0 0 0 – All Schools Vulnerable Schools2/ Notes: 1/ Key: Performance measures: + Positive effect (a) More students taking the unified entrance exam in Armenian Language 0 No effect (b) More students taking the unified entrance exam in Mathematics – Negative effect (c) Higher pass rate on the unified entrance exam in Armenian Language (d) Higher pass rate on the unified entrance exam in Mathematics (e) Higher average score on the unified entrance exam in Armenian Language (f) Higher average score on the unified entrance exam in Mathematics 2/ Vulnerable schools include rural schools, small schools, and only schools in small communities. Source: World Bank Staff. 4.71. Regression analyses of student performance on various educational inputs lead to several unambiguous conclusions on the relationship between performance and educational inputs. The results of the regression analyses are summarized in Table 4.14. First, highly educated teachers are associated with better student academic performance, on average, over the full sample of schools. Second, small class sizes in remote rural schools are linked to lower academic performance (however, no statistically significant effect of average class size is observed for students in urban areas). Moreover, greater reliance on part-time and more experienced teachers is shown to increase student performance to a statistical significant degree in small, rural, and isolated schools. The same is true for the prevalence of computer technology in the school. On the other hand, a larger share of older teachers (those past 62 retirement age or within two years of retirement) and more non-teaching staff in the schools is seen as having a slight negative effect or no effect. Finally, the per-student amount of the PCF allocation to each school is positively correlated with the proportion of grade 11 students who choose to take the unified entrance exam in the Armenian language and in mathematics, but has no bearing on the pass rate or average exam score. 4.72. The first implication of the results of the regression analyses is that highly educated teachers are needed to improve educational outcomes. As mentioned before, the proportion of teachers with higher education degrees varies considerably between rural and urban schools (78 percent vs. 90 percent, respectively). By increasing to the urban average the share of an average rural school‘s teachers who have completed higher education, pass rates on the Armenian Language exam can be expected to increase by 3 percentage points, holding all else equal. 4.73. The second implication is that classes in many rural schools are too small to provide high- quality education. Although intuition may suggest that small classes contribute to better learning outcomes, it is possible for classes to be too small to effectively provide high-quality education. Herbst and Herczyński (2004) suggest that ―in some circumstances smaller classes, especially in rural areas, may be associated with actually reduced student access to various school resources, including qualified teachers and school equipment.‖ Our research indicates that Armenia‘s rural schools (average class size: 13.6) may deprive their students of access to the aforementioned resources (technology, quality teachers, etc.) and lower the quality of education through peer effects, compared to what on average is achieved in urban schools (average class size: 21.8). Though worries about large classes impairing learning are legitimate, most of Armenia‘s schools (average class size: 17.7 nationwide) are nowhere near the level at which such concerns would be realized. In fact, if Armenia‘s rural schools were to raise their average class sizes to the level of their urban counterparts, an average increase in test scores of 0.23 standard deviations can be expected on the Armenian Language exam, holding all else equal. 4.74. Thirdly, the amount of funding a school receives has no bearing on its students’ test scores. Though a 10 percent rise in an average school‘s PCF allocation is associated with a 1 percentage point increase in the proportion of its students taking the university entrance exam, there is no positive relationship between school funding and exam pass rates. In fact, a 10 percent increase in school funding corresponds to a slight decline of 0.03 standard deviations in the average score on the unified exam in mathematics, holding other factors constant.26 Policy Issues to Consider 4.75. Armenia has shown significant improvement in the overall student performance in recent years through international assessments, but the achievement has not been evenly shared among the country’s students. Those in small rural schools and single schools in small communities are not receiving the same quality of education as their peers in bigger urban schools. In theory, the potential advantage of small classes is that students could receive more attention from teachers; our analysis indicates that small classes in small rural schools do not on average contribute to higher quality education in practice. Of particular concern is the relatively large amount of resources directed at financing small rural schools coupled with the fact that students in these schools fail alarmingly to take and pass university admission tests. The factors for poor performance among students in rural areas are multiple and closely interlinked, and therefore are not easily separable. The potential advantage of small classes dissipates when small rural schools are staffed with teachers who are often less qualified than those in urban areas, or equipped with poor teaching-learning facilities, and when students who come from poor 26 See the education chapter in the background volume for detailed OLS regression results on the determinants of student performance. 63 families do not have access to private tutoring and lack opportunities for peer learning in extremely small classes. 4.76. Specific efforts are needed to improve educational outcomes in rural schools. Our analysis suggests that students in small rural schools probably will be better served, educationally, in larger schools in neighboring towns or villages, where more qualified teachers and better learning facilities (such as computer facilities) are available. The current policy to protect small rural schools from closing by allocating higher per-student funding may actually be depriving rural children of opportunities to receive high quality education. The policy response, therefore, must address the quality, efficiency and equity concerns to provide better educational opportunities for children in rural areas. This may offer an opportunity to improve educational equity and average quality without increasing costs (although transportation costs have not been considered). 4.77. First, there is a need to conduct a thorough assessment of the quality of education and educational opportunities currently provided by rural schools. As shown throughout this section, there are reasons to believe that the large network of rural schools serving 40 percent of Armenia‘s student population is not producing high-quality educational outcomes. In order to ensure the equity of access to quality education, the effectiveness of small rural schools must be thoroughly evaluated. Initial findings show that educational opportunities for students in Armenia‘s smallest schools lag far behind those available to graduates of large urban institutions. 4.78. Second, there is a need to review the current PCF formula and adjust it in line with current financing needs. During the early stages of the PCF rollout, some schools were ―protected‖ from formula financing in an explicit attempt to maintain certain small rural schools. After the abolition of the protected category in 2007, the perpetuation of a large number of small schools has been achieved implicitly through an overly generous ―fixed‖ per-school component of the funding formula and some reallocation between schools at Marz level. A reevaluation of the formula is necessary to determine whether its simplicity still meets the needs of the changing educational landscape. In particular, a reduction of the fixed component in favor of more nuanced adjustment factors such as distance to the closest neighboring school, feasibility and estimated costs of busing students and community socioeconomic characteristics may be warranted. 4.79. Third, there is a need to address the low quality of teaching in rural areas. It must be a top priority to strengthen the quality of teaching in rural areas by developing policies to ensure that more qualified teachers are assigned there. Options may include targeted training of rural teachers, tightening of the hiring, dismissing and retirement regulations, and potentially establishing a rotation system for young teachers to serve in rural areas for a certain period, possibly accompanied with appropriate incentives. 4.80. Fourth, there is a need to reconsider alternative ways of providing education to students in rural areas. If the assessment described above determines that the current village school network is not adequately serving the needs of rural students in providing high quality education, alternatives must be considered for reforming education provision in rural areas. Options may include consolidating the vast network of village schools into a smaller number of rural hub schools—at least at the higher school (upper secondary) level, which has been discussed in the past. As has been demonstrated by other countries in the region that have experienced demographic declines similar to Armenia‘s, the optimization of rural school networks is a necessary step to ensuring future quality and fiscal sustainability in education. 4.81. Fifth, there is a need to analyze the impact of private spending on student performance. Even though private spending, particularly on private tutoring to prepare high school students for the university entrance examinations, is playing an increasingly important role in education, our analysis did not examine private spending due to data unavailability. An analysis of the impact of private tutoring on 64 student performance will help identify the shortcomings of the existing education system and improve it so that students without access to private tutoring could receive better education within the formal system. 4.82. There is also the need to increase non-salary recurrent spending to improve quality. Despite the recent increase, government spending on education is still very low by regional standards. As the economy recovers and the Government regains its fiscal space, the government needs to start increasing education spending as planned before the economic crisis. Given that the level of teacher salaries relative to GDP per capita has reached international standards, and that the proportion of salaries in the total recurrent budget is so high, it is time for the authorities to increase allocations to non-salary recurrent and capital investment to improve the quality of general secondary education, as well as other subsectors. 4.83. Finally, the Government needs to estimate enrollments at different levels of education based on enrollment rate targets and school-age population projections. Armenia has experienced sharp demographic changes in the last two decades, but it has started to see the recovery of its young population. The demographic changes will continue impacting on the composition of enrollments at different levels of education. It is crucial for the Government to project enrollments by education level and shift resources to meet the needs of the respective subsectors flexibly. 4.84. Armenian primary education is doing a good job by international standards, and secondary education outcomes, too, are impressive. Despite the relatively low public spending on education, Armenia has shown significant improvement in the overall student performance in recent years through international assessments. Unfortunately, the achievement has not been evenly shared among the students across the country. While the unit cost of educating students in small rural schools is substantially higher than that for large urban schools, rural students lack access to high quality education. The factors for their poor performance are multiple and closely interlinked. It is very important for the Government to address the low quality of education in rural schools as a priority reform agenda and strengthen the overall education system as the foundation for economic growth E. ROAD TRANSPORT EXPENDITURES Table 4.15: Road Network Densities in ECA 4.85. In 2009 Armenia had a total of 7,704 km of roads classified into Country Km per 1000 sq Km per 1000 people the three categories based on the km road’s functions and administration. Albania 657 3.5 Interstate roads (1,686 km) connect to Armenia 279 2.6 other countries road network providing international access. Republican roads Azerbaijan 223 2.2 connect districts and cultural centers. Estonia 1,320 41.2 Local roads connect villages to the FYR of 342 4.3 republican roads network. Table 4.15 shows that road network density of Macedonia Georgia 318 5.1 Armenia, about 2.6 km per thousand Hungary 1,733 15.7 population and 279 km per thousand square km, is at the low end of the ECA Kazakhstan 40 7.1 countries‘ networks. However, the share Kyrgyzstan 170 6.3 of paved roads in Armenia (90 percent) Slovenia 1,007 10.2 is high compared to other developing countries and in line with European Source: World Bank Staff estimates. countries. Furthermore, although the country lacks a bypass for Yerevan, the present road network and capacity are considered sufficient by road experts. 65 4.86. The republican roads network increased significantly in 2009 with the reclassification of Lifeline Roads. In 2008, the Government of Armenia developed a Lifeline Road Development Program in order to stimulate economic growth and contribute to poverty reduction by improving a selected network of ―lifeline roads‖. The network comprised about 3,000 km, of which 2,300 km were Local roads that were reclassified into Republican roads in 2008 and 700 km were already Republican roads. In this way the Republican roads increased in length from 1,747 km in 2008 to 4,056 km in 2009. The Lifeline roads were put under the direct responsibility of the Ministry of Transport and Communication, although their management in practice still remained with the Marzes. At the same time, where these local lifeline roads pass through communities, these communities are responsible for their maintenance. The complicated way the Lifeline roads are managed and maintained could compromise the financial sustainability of road maintenance of all Republican roads. 4.87. Interstate and Republican Figure 4.15: Road Network Utilization in Armenia Main Roads are the most utilized by road users, while lifeline roads 4000 newly classified as Republican Network Utilization (Mil veh-km) 3500 roads are under-utilized. Figure 4.15 3000 presents a rough estimate of roads network utilization, in million vehicle- 2500 km by network type. By breaking 2000 Republican roads into initial 1500 Republican Main roads and 1000 Republican Lifeline roads that were reclassified in 2008, it appears that the 500 second group of Republican Roads is 0 currently under-utilized. This means Armenia that unless activities likely to use Interstate Roads Republican Main Roads these roads (agriculture mostly) are Republican Lifeline Roads Local Roads promoted at the same time, rehabilitating and maintaining these Source: HDM-4 evaluation. roads may not yield a growth dividend. 4.88. Only 30 percent of Republican Roads are considered to be in a good or fair condition. Although most of the Interstate network in Armenia has been rehabilitated since its construction, some roads are still waiting for rehabilitation works and many roads are in need of periodic maintenance. The majority of Republican and Local roads have deteriorated since independence, with almost no maintenance for many years. There is no defined objective or consideration of long-term maintenance strategy. As a result, only about 39 percent of the total road network is in good or fair condition; 49 percent of Main roads (Interstate and Republican roads) are in good and fair condition. This is well below the developing international benchmark of 70 percent of main roads in good or fair condition 4.89. As a result, the current value of road assets under the Central Government’s oversight is at less than three quarters of its potential. The current asset value27 of Interstate and Republican roads in Armenia is estimated to be US$ 2,433 million, which is 28 percent of GDP and 73 percent of the maximum possible asset value (US$ 3,340 million), assuming that all roads are in good condition (Table 27 Current asset value computed with HDM-4 accounting for the reduction of the network maximum asset value, which is the asset value considering all roads in good condition, function of the replacement value needed to bring fair and poor condition roads to good condition. 66 4.16). Despite representing only 31 percent of the road network length, Interstate roads carry most of the traffic. Their maintenance should be the top priority of road transport policy. Table 4.16: Main Roads Current Condition and Asset Value Length Utilization Good/Fair Max Current Current Average Network (M veh- Asset Asset Value Value Traffic (km) Roads (%) km/year) Value (US$ Mil) per GDP (AADT) Interstate 1,686 2,571 85% Mil) (US$1,846 1,548 (%) 17.8% 3,816 Republican Main 1,697 629 51% 1,066 774 8.9% 899 Republican Lifeline 2,359 572 17% 898 582 6.7% 732 Total 5,742 3,772 49% 3,810 2,903 33.3% 1,751 Source: HDM-4 Evaluation. Trends in Road Expenditure and Sources of Finance 4.90. Expenditure in the transport sector is allocated under three main headings: (a) rehabilitation of roads, (b) rehabilitation of transport objects (tunnels, bridges); and (c) maintenance of roads (winter and routine maintenance). Road maintenance expenditures in Armenia remained flat and very low in the last five years, standing at about 0.2 percent of GDP. No new construction was carried out during that period. High income countries with mature road networks in good condition normally allocate about 1 percent of GDP on maintenance. Information on road expenditures classified by capacity increase improvements, rehabilitation, periodic maintenance and routine maintenance is not available. This makes it difficult to evaluate the efficiency of current allocation of resources. Table 4.17 presents the consolidated expenditures on the road network compared to GDP in Armenia from 2005 to 2009 and the planned expenditures for 2010 and 2011. Between 2005 and 2007, 100 percent of Armenia‘s capital investments in roads were fully financed by the domestic budget or by grants for rehabilitation (mostly the latter). From 2008, IFIs started investing in roads infrastructure in Armenia but their investments are concentrated on road rehabilitation and do not include maintenance. Remedial road maintenance has been financed mainly by grants. About 13 percent of 2009‘s spending was allocated to routine maintenance and 83 percent to the rehabilitation of the road network. 4.91. Almost no periodic maintenance is being carried out in Armenia; estimates show that routine maintenance funding for Local and Republican Lifeline roads will need to be tripled in order to ensure proper maintenance of these roads in the long run. A total of US$15 million was allocated to routine maintenance of Interstate and Republican roads in 2010, forming 15 percent of the total road sector budget and resulting in an average allocation of US$4,380 per km. This allowed all 3,383 km of Interstate and original Republican roads to be put under summer and winter maintenance. Of this amount per km, approximately 25 percent is spent on winter maintenance and 45 percent on patching (the latter can require as much as 70 percent of the surface for roads in poor condition). For the state-owned local road network, however, the available maintenance funding was US$2 million for 3,319 km of local roads (including Republican Lifeline roads), forming only 2 percent of the road sector budget and resulting in an average allocation of US$620 per km. These allocations for local and Republican Lifeline roads are grossly insufficient, resulting in only approximately half the roads receiving winter maintenance and only 5 to 10 percent receiving summer maintenance. Estimations by the HDM-4 model show that routine maintenance funding for Local and Republican Lifeline roads will need to be tripled in order to ensure proper maintenance of these roads in the long run, assuming that road conditions will continue to improve as a result of rehabilitation. 67 Table 4.17: Investment Program on the Road Network from 2005-2011 (US$ million) Actual Planned 2005 2006 2007 2008 2009 2010 2011 GDP (US$ million) 4,900 6,384 9,206 11,917 8,714 8,269 9,673 Improvement Works Government Budget - - - - - IFIs - - - - - 29 Total Improvement Works 29 GDP share % 0% 0% 0% 0% 0% 0% 0.35% Maintenance and Rehabilitation Works Government Budget 27 35 57 41 36 30 11 Rehabilitation of interstate roads 5 16 25 11 16 9 3 Rehabilitation of republican roads 17 16 7 24 14 11 6 Rehabilitation of local roads 4 3 25 7 6 10 2 GDP share % 0.6% 0.5% 0.6% 0.3% 0.4% 0.4% 0.1% Government Budget 10 13 21 22 14 17 18 Maintenance of inters. and rep. roads 9 12 18 19 13 15 18 Maintenance of local roads 0.7 1.1 1.5 1.5 - 2 - Others 0.4 0.5 0.6 0.9 0.6 0 - GDP share % 0.2% 0.2% 0.2% 0.2% 0.2% 0.2% 0.2% IFIs - - - 6 63 58 43 Rehab of Int. & Repub. roads or new - - - 6 63 58 43 con. Maintenance - - - - - - - GDP share % 0.0% 0.0% 0.0% 0.1% 0.7% 0.7% 0.5% TOTAL 37 48 78 69 113 105 72 GDP share % 0.8% 0.8% 0.8% 0.6% 1.3% 1.2% 0.9% Source: Ministry of Transport and Communication, and World Bank. 4.92. A revamping of the fuel excise regime will improve revenues collected from the transport sector. Fuel excise taxes in Armenia are well below European averages. Indeed, the average gasoline excise tax in Europe is 0.63 US$ per liter, while the current excise tax in Armenia is 0.31 US$ per liter. Similarly, the average diesel excise tax in Europe is 0.45 US$ per liter, while the current excise tax in Armenia is 0.09 US$ per liter. Although not be earmarked for road maintenance, this increase in revenue collection would give some fiscal headroom to increase the road maintenance budget. Tradeoff between Current Budget Constraints and Future Transport Costs 4.93. Maintenance and preservation are essential to avoid high rehabilitation and construction costs in the future, but they are expensive for the current budget. Using the HDM-4 model, the performance of Interstate and Republican roads (main roads) under several maintenance and rehabilitation standards was evaluated to roughly determine the optimal requirements for recurrent maintenance, periodic maintenance and rehabilitation in Armenia. Since detailed network road condition and traffic data are not available in Armenia, the HDM-4 evaluation is based on engineers‘ estimates of the current macro characteristic of the network. The HDM-4 evaluation provides indicative figures of the maintenance and rehabilitation needs of the network and an example of the type of evaluation that is required in Armenia. 68 4.94. Three different scenarios to preserve main roads were considered for this tradeoff analysis: (a) minimize total transport costs; (b) maintain expenditure at US$ 110 million per year; and (c) eliminate backlog in 10 years. The HDM-4 assessment was based on a number of assumptions and four different scenarios. Basic assumptions included 5 percent traffic growth rate per year; a 12 percent discount rate; a 20-year evaluation period; unit costs of road works based on current average road work costs in Armenia; and average unit road user costs based on current average vehicle fleet characteristics for Armenia. The ―minimize transport costs‖ scenario achieves a target of attaining 72 percent of main roads are in good or fair condition by 2020. The ―maintain expenditure at US$110 million‖ scenario takes into consideration maintenance and rehabilitation expenditures on the network in 2009, with 67 percent of roads in good or fair condition by 2020. The ―eliminate backlog in 10 years‖ scenario eliminates the rehabilitation backlog in 10 years, with 65 percent of roads in good or fair condition by 2020. Keeping the road condition (last scenario) takes into consideration the current condition of the road network and assumes that this condition will be maintained in the future, with 50 percent of roads in good or fair condition by 2020. Table 4.18 summarized the schedules of public expenditures for the three scenarios. Table 4.18: Economic Comparison the Three Scenarios Present Value of Costs Network Asset Value in 2020 Scenario condition US Road Agency Road Users Total (%) Percentage million Minimize total transport costs 920 14,308 15,229 72 3,343 15% US$110 million per year 799 14,441 15,240 67 3,338 15% Eliminate backlog in 10 years 729 14,524 15,253 65 3,300 14% Keep current condition 658 14,704 15,362 50 3,084 6% Source: HDM-4 model evaluation. 4.95. Maintaining the main road network in its current road condition will minimize road expenditure, but increase road user costs and reduce the road asset value in the future. While the optimal scenario is the minimize transport cost scenario, the budget constrained scenario, with annualized expenditures of US$110 million over the evaluation period, leads to significantly better road conditions than what would be achieved under the keep current condition scenario. Whereas total expenditure on Interstate and Republican roads was projected to reach only US$68 million in 2011, the HDM-4 analysis suggests that US$98 million should be spent annually in the next five years to eliminate the rehabilitation backlog in 10 years. Thus, the funding gap is US$30 million. 4.96. Road transport expenditure does not appear to be a low hanging fruit for fiscal consolidation. The Government of Armenia spent only US$110 million in 2009 on rehabilitation, periodic maintenance and routine maintenance. On the other hand, reflecting the required fiscal consolidation, $79 million instead have been allocated for 2010 and $39 million for 2011. Such a decline in spending on roads would lead to severe degradation of the transportation system, raising transportation costs to all users and undermining the profitability of enterprises that depend on road transport. Therefore, it is necessary to prioritize expenditure towards periodic and recurrent maintenance in order to maintain the road asset to its current value. Then key rehabilitation works needed to improve the road condition of the most used roads (Interstates and Main Republican roads) could be considered. New road construction projects should be postponed, particularly if their future utilization rate is in doubt. 4.97. Furthermore, financing options need to be identified to keep at least the 2009 spending level of US$112 million necessary to maintain the road network in its current condition. The following could be considered: enforcing the new law bringing gasoline and diesel back to the normal excise 69 regime; borrowing to finance the rehabilitation of high returns portions of the road network; and mobilizing private sector resources through PPP for the rehabilitation of high returns portions of the road network. The following institutional reforms may also be needed to improve intra-sectoral expenditure efficiency: clarify the monitoring responsibility between sub-national and central governments, particularly for the Lifeline Roads recently converted into Republican Roads; monitor road conditions, traffic and accidents on all Republican and Interstate roads on an annual basis; expand the use of performance-based contracts to cover routine maintenance, periodic maintenance and rehabilitation. Policy Issues to Consider 4.98. It is important to realize that road transport is a very poor candidate for fiscal consolidation. Indeed, the Government spent only US$110 million in 2009 on rehabilitation, periodic maintenance and routine maintenance, compared to the $140 million needed under the road financing scenario that would minimize transport costs for users, or the $112 million needed under the financing scenario that maintains the road stock quality as it is. The current situation is not sustainable and comes at a cost to the economy. Current traffic standards are too high and, combined with the lack of routine and periodic maintenance contribute to high costs to the budget (not just the users). A viable strategy should be to adapt the standards to traffic, use efficient contracts (such as performance based ones that include rehabilitation and maintenance), and ensure that the budget allocates enough to maintain the network. 4.99. There is limited scope for private sector investment in the road sector but private contractors can improve the efficiency of public spending through use of the "asset management approach." Under this approach, a contractor receives a multi-year, performance-based contract that assigns it the responsibility to maintain at a certain level of service a small network of roads. Such a contract builds in the funding requirements for maintenance during the duration of the contract, avoids costs overruns, and yields overall better quality road works and road users‘ satisfaction. On the current performance-based maintenance contracts, annual agreements should be abolished and procurement carried out in a staggered manner. 4.100. There is a need to better institutionalize a Road Asset Management System. This requires improving data collection and evaluation processes, ensuring that the public sector group in charge of the system has clear obligations, proper resources and sufficient technical capacity. The weakest elements of the current system are the lack of a GIS interface, the limited network data collection for Republican roads, and the limited network traffic counts program. There should be a commitment to produce an annual report with the information required to make informed decisions in order to enable proper planning and programming of road asset management. 4.101. Far from being a candidate for budgetary saving, road transport requires an infusion of additional resources. Armenia‘s road system, especially its local roads, is depreciating away. This raises the cost of production of any activity that depends on transportation. At a minimum, the current road network needs to be preserved in its current state. Even this would cost more than was spent in 2009. To maintain and rehabilitate the road network at a level that minimizes transport costs for users would require maintenance and rehabilitation spending to increase by more than 25 percent. Road transport spending needs to move in this direction. 4.102. The analysis of the Government’s expenditure showed that the government did not have much, if any, room for savings. The sectors studied require additional allocations over the medium term to get in line with the actual needs and ensure delivery of high quality public goods. In the health sector the Government needs to continue expanding coverage and reducing out-of-pocket spending (otherwise, the authorities might need to target poorer groups of population). Public spending in the education sector needs thorough re-assessment to link better financing to actual outputs. For example, the large network of rural schools serving 40 percent of Armenia‘s student population does not seem to be producing high- 70 quality educational outcomes. In the field of road transport spending, prioritizing expenditure towards periodic and recurrent maintenances is needed to maintain the road asset at its current value, while the Government currently under-spends US$2-40 million in those sectors. In the public sector wage bill, it is necessary to harmonize the different public sector remuneration systems and to strengthen the analytical foundations for a coherent, affordable public sector remuneration policy. This will increase the motivation of public employees and result in better performance of the public sector. However, the analysis showed that reforming the remuneration system may result in increased wage bill for public servants. Tapping the revenue potential - estimated to be in the range of 2.3 and 5.8 percent of GDP - could help to scale up expenditures as needed. 71 Annex 1: Selected Review of previous Studies on Tax Administration and Tax Policy in Armenia Authors Methodology Findings Recommendations Hamid R. Davoodi, David A. Cross-country econometric  Armenia‘s tax effort falls short of  Institutional improvements as well as policy measures designed to Grigorian (2007) tax potential model, using its potential by about 6.5 percent of reduce the size of the shadow economy. It puts a lot of emphasis regressors such as GDP on the need to improve the general tax moral and suggests several institutional quality and recommendations to improve the general legislation, including shadow economy provisions on the conflict of interest. characteristics IMF Article IV Consultations, Assessing Armenia‘ tax gap  Income tax potential falls behind On tax policy side: December 7, 2010 Report by comparing the actual tax Armenia‘s comparator/peer  Eliminate tax policy loopholes, such as tax exemptions (the entire collection (as share of GDP) countries by 3 percentage point of agriculture sector) with its peers in the CIS GDP  Moving from application of presumptive taxes on fuel and tobacco region  In addition, taxes on goods fall imports to regular taxation regime short behind by 1 percentage point On tax administration side: of GDP  Improve tax audits and introduce risk-based approach  Curb the growth of tax arrears  Enhance capacities at large taxpayers inspectorate  Address tax credits practice  Improve interpretation of tax legislation Le, Moreno-Dotson, and Employs a cross-country  Armenia is classified in the group  Undertake comprehensive reforms of both tax policy and tax Rojchaichaninthorn (2008) study to estimate tax of developing countries administration capacity from a sample of characterized with low collection  Revenue enhancement in line with efficiency and equity criteria. 104 countries during 1994- and low tax efforts  Broaden the tax base, rationalize the tax rate structure and 2003.  The tax capacity estimated at 21.06 incentive schemes percent of GDP against 16.4 actual  Remove tax-induced economic distortions with focus on Tax effort is defined as an average for 1999-2003 period enhancing revenue productivity of major taxes, particularly the index of the ratio between (estimated gap - about 5.2 percent broad-based consumption VAT the share of the actual tax of GDP). Thus the tax effort of collection in gross domestic 0.78. product and the predicted  The predicted Revenue capacity is taxable capacity higher than the tax capacity and is estimated at 22.36 and the revenue effort 0.73 Miles K. Light (2004) Quantitative analyses of Cost of public funds in Armenia ranges  The existing profits, income and value-added taxes all tend to Armenia‘s tax potential from 1.3 to over 5 depending on tax discriminate against investment in the formal sectors. based on General base and the model horizon. The Equilibrium Model estimate potential in the short-run is  Tax policy should be targeted to reduce tax evasion and informal between 1.3 to 2 percent. activities. Taxes on capital can raise revenue in the short run but these taxes are costly in the long run due to their distortion of 72 incentives for investment. Value-added and income taxes can be costly sources of additional revenue due to their impact on the level of informality. Excise taxes are efficient but offer limited revenue potential. In the long-run, better enforcement of existing tax laws is essential. UN-funded report (1999) Assessing Armenia‘ tax gap In 1997, tax to GDP ratio in Armenia  Make taxes more transparent and stable to facilitate tax by comparing the actual tax was 17 percent, whereas the average for compliance collection (as share of GDP) the Newly Independent States (NIS)  Adopt simple taxes with lower rate, broad base and few with its peers in the CIS group was 25 percent and for the exemptions region Central and Eastern European/NIS group – 32 percent. The percentage of revenues to GDP in Armenia was half of that in countries like Estonia, Latvia or Poland. Assessing incentive effect of  The overall marginal effective tax  Improve investment incentives of taxation in Armenia David Joulfaian, Lilit Melikyan taxes (property and rates are well below the statutory enterprise profit tax) on rates levied by the government. The (2004) investment in Armenia in the rate varies between 0 and 13 long run. percent  A reduction of one percentage point in the discount rate reduces the cost of capital by more than the repeal of the entire profit tax 73 Annex 2: Regression Results Dependant Variable: Profit Tax Payment of firm i in year t Random Effect Panel Regression Tobit Panel Regression Group 1 Group 2 Group 3 Group 1 Group 2 Group 3 Regressors Coeff. Coeff. Coeff. Marg. Eff. Marg. Eff. Marg. Eff. Profit before tax 0.039*** 0.023*** 0.209*** 0.040*** 0.023*** 0.219*** Total assets 0.001** 0.002*** 0.001 0.001** 0.002*** -0.001 Nb. Employees 8.588 33.878** 25.122 7.763 29.203** 57.309 Yerevan 40003* 17683 234211 5293** 18541 258791 Activity -216*** -844*** -727 -234*** -846*** 165 Const. 18072 66805*** 111607 Nb. Panel group 277 275 41 277 275 41 R2 0.2 0.1 0.7 P-value 0.0 0.0 0.0 0.0 0.0 0.0 Dependant Variable: Income Tax Payment of firm i in year t Random Effect Panel Regression Tobit Panel Regression Group 1 Group 2 Group 3 Group 1 Group 2 Group 3 Regressors Coeff. Coeff. Coeff. Marg. Eff. Marg. Eff. Marg. Eff. Social contribution 0.595*** 0.452*** 0.651*** 0.596*** 0.455*** 0.664*** Total assets 0.000 0.000 0.000 0.000 -0.000 0.000 Nb. Employees Yerevan 2685.118* 10961.620** 48079.400 2764.892* 12242.830** 47663.44 Activity -72.219* -622.499*** -5732.368*** -72.502* -624.740*** -5633.541*** Const. 1585 27423*** 241290*** 1475 26741*** 234109*** Nb. Panel group 277 275 41 277 275 41 R2 0.7 0.5 0.8 P-value 0.0 0.0 0.0 0.0 0.0 0.0 Dependant Variable: VAT Payment of firm i in year t Random Effect Panel Regression Tobit Panel Regression Group 1 Group 2 Group 3 Group 1 Group 2 Group 3 Regressors Coeff. Coeff. Coeff. Marg. Eff. Marg. Eff. Marg. Eff. Domestic sales 0.123*** 0.089*** 0.073*** 0.124*** 0.090*** 0.073*** Total assets 0.145*** 0.004*** 0.034*** 0.015*** 0.004** 0.033*** Nb. Employees -64.915** -38.468 -384.845* -68.469** -34.519 -379.868* Yerevan -17864* -100983*** -836236 -102198** -102199** -836789* Activity 885*** 3944*** -12517 864*** 3926*** -12359 Const. -16229 -31843 217120 Nb. Panel group 277 275 41 277 275 41 R2 0.2 0.1 0.7 P-value 0.0 0.0 0.0 0.0 0.0 0.0 74 REFERENCES ADB TA, 2010 Measuring the Contribution of Informal Sector to Total Economy in Armenia Burnside, C. 2005. Fiscal Sustainability in Theory and Practice: A Handbook. The World Bank: Washington, DC. Casanegra de Jantscher, M., 1990, Administering a VAT, in M. Gillis, C.S. Shoup, and G.P. Sicat, eds., Value Added Taxation in Developing Countries (World Bank, 1990). Ms. Casanegra de Janttscher was the head of the Tax Administration Division of the Fiscal Affairs Department of the IMF Corfmat, Francois, Firestone, Allan, and Michael Welling, 2004, Armenia – Key Areas for Further Tax and Customs Reform‖, IMF FAD Report. Davoodi, H.R. and D.A., Grigorian, 2007, Tax Potential vs. Tax Effort: A Cross-Country Analysis of Armenia’s Stubbornly Low Tax Collection‖, IMF Working Paper WP/07/106 Djankov, S., Ganser, T., McLiesh, C., Ramalho, R., Shleifer, A., 2010, The Effect of Corporate Taxes on Investment and Entrepreneurship‖, American Economic Journal: Macroeconomics 2, 31-64 European Union, 2009, Report on the Tax Law (VAT) Priority Area, EU Project on Baseline Measurement and Reduction of Administrative Costs, Final Report Fedelino, Ivanova and Horton. 2009, Computing Cyclically Adjusted Balances and Automatic Stabilizers, IMF: Technical Notes and Manuals, Washington, DC. Fenochietto, R., 2010, VAT and Income Tax GAP, IMF Fiscal Affairs Department Tax Policy Division. Gertler, P. and others. 2000. ―Health.‖ In M. Grosh and P. Glewwe, (eds.), Designing Household Survey Questionnaires for Developing Countries: Lessons from 15 years of the Living Standards Measurement Study. Grote, M., Caner, E. Hutton, 2011, Continuing Tax Policy in Armenia: Planning for a Simpler, Fairer and More Efficient Tax System, IMF Fiscal Affairs Department. Harrison, Graham, Masters, Andrew and Kevin Woodley, 2008, Armenia – Implementation of the Tax Administration Reform Plan, IMF. Herbst, Mikołaj, and Jan Herczyński, 2004, November 2004 Working Paper. Available online at: http://www.esep.pl/is%20large%2011.pdf. IMF, 2008, Debt Sustainability Analysis for Market Access countries, IMF: Staff guidance note. Joulfarian, D. and L. Melikyan, 2004, ―Taxes, Investment Incentives, and Cost of Capital in Armenia‖ Armenian International Policy Research Group, Working Paper No. 04/03. Junquer, R., Fenochietto, R. and D. Alvarez, 2010, The Revenue Administration and Policy Thematic Group (RAPTG) Methodologies to Estimate Tax Compliance Gap, The Revenue Administration and Policy Thematic Group. Kataoka, Sachiko, 2011, "Chapter 2: Per Capita Financing of General Education in Armenia" in Reforming Education Finance in Transition Countries: Six Case Studies in Per Capita Financing System, Juan Diego Alonso and Alonso Sanchez, eds., The World Bank: Washington, DC. Kheyfets, Igor, 2011, Background Paper for the Armenia Public Expenditure Review in Education. 75 KPMG LLP (UK) and KPMG Advisory Ltd (Hungary) and the State Revenue Committee, 2010, Business Process Re-engineering, Private and Confidential Report prepared for the Armenian Authorities. Le, T.M., Moreno-Dodson, B. and J. Rojchaichaninthorn, 2008, Expanding Taxable Capacity and Reaching Revenue Potential: Cross-Country Analysis. World Bank Policy Research Working Paper No 4559. Light, M.K. and T.F. Rutherford, 2004, Taxation and Economic Efficiency in Armenia, mimeo, World Bank. Masters, Andrew, Kevin Woodley, and Blaine McDonald, 2009, Armenia: Advancing Tax Administration Reforms, IMF, FAD. National Statistical Services of the Republic of Armenia, Statistical Yearbook, various years. OECD, 2010, Education at a Glance 2010. Rajaratnam, J.K., et al. (2010) Worldwide mortality in men and women aged 15–59 years from 1970 to 2010: a systematic analysis. Lancet375: 1704–20 Schaffer, M.E. and G. Turley, 2001, Effective versus Statutory Taxation: Measuring Effective Tax Administration in Transition Economies, European Bank for Reconstruction and Development Working Paper No. 62. Schneider, F., 2002, The Size and Development of the Shadow Economy and Shadow Economy Labor Force of 22 Transition and 21 OECD Countries. Invited Papers prepared for the Round Table Conference held in Sofia, Bulgaria, April 18-20, 2002. Schneider, F. and J. Kepler, 2007, Shadow Economies and Corruption All Over the World: New Estimates for 145 Countries. Shukla, G.P. ,2010, Fiscal Regime for Mining Sector in Armenia. Draft Report prepared for the World Bank and the Ministry of Finance of Armenia. Vazquez-Caro, J., 2010, Public Expenditure Review: Revenue Component. Background paper for this study. WHO‘s World Health Report (2000) and the World Bank‘s Health, Nutrition, and Population sector strategy (2007). World Bank (2008). Armenia Programmatic Public Expenditure Review, June 11, 2008. The World Bank: Washington DC. World Bank, Ready for Europe: Public Administration Reform and EU Accession in Central and Eastern Europe. World Bank Technical Paper 466. May 2000. World Bank, 2008, Armenia: Choices in Development Policy 2008-2012. Unpublished document. Xu, K. and others. 2003. ―Household catastrophic health expenditure: a multi-country analysis.‖ Lancet362: 111-117. 76