Policy Research Working Paper 10570 Fiscal Policy Effects on Poverty and Inequality in Cambodia Wendy Karamba Michal Myck Kajetan Trzcinski Kimsun Tong Poverty and Equity Global Practice September 2023 Policy Research Working Paper 10570 Abstract This study assesses the short-term impact of fiscal policy, inequality, the degree of inequality reduction is small in and its individual elements, on poverty and inequality in international comparison; and (iii) low-income households Cambodia as of 2019. It applies the Commitment to Equity pay more in indirect taxes than they receive in cash ben- methodology to data from the Cambodia Socio-economic efits in the short term to offset the burden. As a result, Survey of 2019/20 and fiscal administrative data from var- the number of poor and vulnerable individuals who, in ious government ministries, departments, and agencies for the short term, experience net cash subtractions from their the assessment. The study presents among the first empir- incomes is greater than the number of poor and vulnerable ical evidence on the impact of taxes and social spending individuals who experience net additions. Fiscal policy can on households in Cambodia. The study finds that: (i) deliver more net benefits to poor and vulnerable households Cambodia’s 2019 fiscal system reduces inequality by 0.95 through expanding social assistance spending. Cambodia Gini index points, with the largest reduction in inequal- has embarked on this expansion during the coronavirus ity created by in-kind transfers from spending on primary pandemic, bringing it closer in line with comparators. education; (ii) while Cambodia’s fiscal system reduces This paper is a product of the Poverty and Equity Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at wkaramba@worldbank.org; mmyck@cenea.org.pl; ktrzcinski@worldbank.org; ktong@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Fiscal Policy Effects on Poverty and Inequality in Cambodia Wendy Karamba 1 Michal Myck 2 Kajetan Trzcinski 3 Kimsun Tong4 Keywords: Fiscal Policy, Social spending, Taxation, Fiscal Incidence, Poverty, Inequality, Cambodia, Commitment to Equity. JEL Codes: D31, H2, H5, I38, O23 1 Economist, Poverty and Equity Global Practice, World Bank. Email: wkaramba@worldbank.org. 2 Director, Centre for Economic Analysis (CenEA). Email: mmyck@cenea.org.pl. 3 Consultant, Poverty and Equity Global Practice, World Bank. Email: ktrzcinski@worldbank.org. 4 Economist, Poverty and Equity Global Practice, World Bank. Email: ktong@worldbank.org. The authors are thankful to Jon Jellema and Mariano Ernesto Sosa for peer-reviewing the paper. We are thankful to Matthew Wai-Poi for useful inputs, comments, and suggestions on the paper and to Rinku Murgai for overall guidance. The authors would like to thank the Ministry of Economy and Finance of Cambodia (MEF) and the Ministry of Planning (MoP), National Institute of Statistics (NIS) for making available data. The authors are also grateful for the valuable comments on the results received from the Ministry of Economy and Finance during consultations held in Phnom Penh in 2022 and on the draft document. Funding for the research was received from the Global Tax Program Multi-Donor Trust Fund generously supported by the governments of Australia, Denmark, France, Japan, Luxembourg, Netherlands, Norway, Switzerland, the United Kingdom as well as Bloomberg Philanthropies. The findings, interpretations, and conclusions in this paper are entirely those of the authors. The findings do not necessarily represent the view of the World Bank Group, its Executive Directors, or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this work. 1. Introduction Sound fiscal policy is central for macroeconomic stability and growth, as well as for poverty and inequality reduction. This can happen both through its influence on macroeconomic conditions (economic growth, employment, and inflation) and through the distributional implications of taxes and public spending. For instance, public spending that is well-targeted to poor and vulnerable households can effectively alleviate the incidence and depth of poverty, support broader distribution of the benefits of economic growth (de la Fuente et. al 2017), while at the same time leaving scope and resources for public interventions in other areas. Fiscal policy thus plays a direct role in poverty and inequality reduction through redistributive effects and how resources are collected and spent influences whether the fiscal toolkit achieves poverty- reducing and equalizing goals. In the case of Cambodia, so far a comprehensive analysis of the distributional implications of fiscal policy has been missing and this study aims to fill this gap. In the decade to 2019, Cambodia sustained macroeconomic stability, including prudent fiscal management, and robust growth. Strong and inclusive growth contributed to substantial poverty reduction and shared prosperity. According to official statistics, poverty nearly halved nationally from 33.4 percent in 2009 to 17.8 percent in 2019/20. Inequality remained moderate in international comparison with a Gini coefficient estimated at 0.32 in 2019/20, although this level of inequality represents a slight increase since 2014. 5 This paper assesses the redistributive effects of Cambodia’s 2019 fiscal policy and its individual elements. We apply an internationally recognized methodology developed by the Commitment to Equity (CEQ) Institute to assess how taxes and social spending affect poverty and inequality. 6 The present analysis relies on the 2019/20 Cambodia Socio-Economic Survey (CSES) and 2019 administrative data on taxes, transfers, and social spending. The CSES is collected by the National Institute of Statistics (NIS) and the fiscal administrative data is compiled by the Ministry of Economy and Finance (MEF). The paper seeks to answer the following questions: How much are social spending, subsidy, and tax policies contributing to redistribution and poverty reduction goals? How are specific taxes and government spending contributing to equalization and poverty reduction? How would fiscal policy reforms that change the size and/or progressivity of a particular tax or benefit affect inequality and poverty? Within the limits of fiscal prudence, what could be done to make taxes and transfers more “pro- poor”? Such evidence can inform policy makers and other stakeholders in assessing existing fiscal instruments and in designing reforms. The 2019 fiscal system in Cambodia, and many of its elements, reduced inequality but also left poor households out-of-pocket in the short term. 7 According to the Fiscal Incidence Analysis using the CEQ methodology, the modeled fiscal system in Cambodia decreases the Gini index by 0.95 percentage points. Spending on primary education shows the largest benefit on inequality, complemented by direct taxes, 5 World Bank (2022a). 6 See Lusting and Higgins (2018) for a description of the CEQ methodology. 7 Note, some of the findings of this paper are featured in the 2022 Cambodia Poverty Assessment. 2 which the CEQ approach estimates to be largely progressive. Despite the decrease in inequality, the degree of redistribution is small by international comparison. At the same time, the analysis estimates that poorer households pay more in indirect taxes than they receive in cash benefits in the short-term. The fiscal system increases the short-term measure of the poverty headcount by 2.1 percentage points. Although direct transfers to poor households have a poverty-reducing effect, they are not large enough to offset the poverty-increasing effect of indirect taxes—notably of value-added tax that is levied on consumption. While poorer households are worse off on average in cash terms in the short-term, it is important to note the analysis does not account for the longer-term economic benefits of health and education services to households which will help reduce poverty. Subject to available funding, cash transfers at a scale large enough to compensate for the burden of taxes has the potential to deliver net benefits to poor and vulnerable households, while improved targeting would help to reduce the overall costs of such expanded program. Cambodia is not alone among developing countries in seeing short-term poverty increase due to the modeled elements of fiscal policy. Because the CEQ methodology has been consistently applied in over 60 countries, the exercise allows Cambodia’s performance to be compared with peer lower-middle- income countries. Several lower-middle-income countries experience net poverty increases due to their fiscal policies which is not surprising given the frequent heavy reliance on indirect taxation. Nonetheless, some low and lower-middle income countries have more poverty-reducing effects of fiscal policy. It is important to note that the presented Cambodia fiscal assessment predates the COVID-19 pandemic. Hence, the results may be different at present due to the scale up of cash transfers to poor and vulnerable households in response to the pandemic. As of January 2023, the COVID-19 cash transfer program has disbursed US$ 932 million since the launch in June 2020. The program has been the largest component of the government’s support package. Spending on cash transfers rose from less than 0.1 percent of GDP in 2019 to 0.7 in 2020 and 1.4 in 2021. The program has reached about 690,000 households and 2.7 million individuals, or about 17 percent of the population, up from 2 percent pre-pandemic. 8 Box A.1 in the Appendix summarizes the government’s fiscal response to the crisis. The next sections are organized as follows: Section 2 provides an overview of the main taxes and transfers in Cambodia, including the size and composition of these fiscal tools. Section 3 describes the methodology, empirical approach, and data sources. Section 4 presents the main findings for Cambodia and in comparison with other countries. Section 5 describes the results from the simulations that examine the poverty and inequality effect of expanding cash transfers to poor households. Section 6 concludes and highlights the implications of the results for Cambodia. 2. Taxes and Social Spending in Cambodia Evolution of government revenues and spending between 2009 and 2019 In the decade leading up to 2019, Cambodia experienced strong economic growth that raised the incomes of the poor and reduced poverty. This took place under a period of macroeconomic stability and prudent 8 See World Bank (2022a) for summary of the government’s fiscal response to the crisis. 3 fiscal measures. In 2009, the fiscal deficit was large, accounting for over 8 percent of the gross domestic product (GDP). But Cambodia progressively reduced its fiscal deficit through increased revenue collection and prudent spending. Since 2015, Cambodia was able to substantially reduce the fiscal deficit and maintain it at less than 3 percent of the GDP (Figure 1). The deficit is often financed by foreign aid grants or development partner financing of investment projects. In 2019, government savings rose to 20 percent of GDP (or KHR 22.2 trillion) after several years of accumulation. The overall fiscal deficit is estimated to have been limited at -0.7 percent of GDP in 2019. Figure 1: Domestic revenue and expenditure in Cambodia (% of GDP), 2009-2019 30 20 Percent 10 0 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 Direct taxes Indirect taxes (incl. trade taxes) Other revenue Domestic revenue Expenditure Source: MEF. Note: Other revenue includes non-tax revenue, capital revenue and provincial revenue. Revenue collection increased because of improvements in tax administration and a broadening tax base. Between 2009 and 2019, tax revenue collection of the central government increased from 9.7 percent of GDP to 20 percent. Various tax administration reforms introduced by the General Department of Taxation (GDT) and the General Department of Customs and Excises (GDCE) supported this increase. Since 2008, the GDCE rolled out the Automated System for Custom Data (ASYCUDA) and enhanced anti-smuggling measures. In 2014, the GDT introduced a shift toward a real regime system of taxation for business profits (mainly of large, incorporated and/or commercial enterprises), improved taxpayer registration and services, and improved tax auditing capacity leading to better compliance (World Bank, 2019). These reforms were implemented in the context of the Revenue Mobilization Strategy for 2014-18. These efforts, coupled with fast economic growth, enabled gains to be realized in most sources of revenues, particularly from value-added tax, excise duties, and corporate tax. In 2019, tax revenues for Cambodia were higher than most ASEAN countries (Figure 2). 4 Figure 2: Tax revenue (% of GDP) of Cambodia and ASEAN countries, 2019 Cambodia Thailand Philippines Singapore Malaysia Indonesia Myanmar 0 5 10 15 20 Source: WDI database. Note: Data for Vietnam, Lao PDR and Brunei Darussalam were not available. Cambodia’s improved fiscal space allowed the government to boost spending. Government expenditure increased from 20.2 percent of GDP in 2011 to 24 percent in 2019, largely driven by spending on social services and on defense and security (World Bank, 2023). However, Cambodia’s expenditure boost partly reflected salary increases for civil servants. Between 2011 and 2019, the public sector wage bill as a share of GDP almost doubled from 4.4 percent of GDP in 2011 to 7.1 percent of GDP in 2019 (World Bank, 2023) and public sector minimum wage rose from around US$ 50 a month in 2012 to US$ 250 in 2018 (World Bank, 2019). Social sector spending increased from 4.6 percent in 2011 to 6.8 percent in 2019 largely due to increased education spending, which almost doubled 1.5 percent of GDP in 2011 to 2.9 percent of GDP in 2019 (World Bank, 2023 and WDI). However, health spending only grew marginally in the years before the pandemic. Electricity subsidies to eligible households and businesses amounted to US$ 33 million (0.12 percent of GDP) in 2019. Tax system in Cambodia Revenue collection in Cambodia is highly centralized with a small number of taxes collected at the municipal or provincial level. In total, central government revenue amounted to 23 percent of GDP in 2019, while provincial government revenue amounted to 2 percent (Table 1). Indirect taxes are the main source of government revenues (13 percent of GDP), followed by direct taxes (5 percent of GDP). Table 1 outlines Cambodia’s government revenue structure. 5 Table 1: Cambodia Domestic Government Revenues, 2019 Amount Percent of Included Percent Source (billion Government in CEQ of GDP KHR) Revenue analysis Domestic Revenues (A+B) 27,730 25 100 A. Central Government Revenue (I+II+III) 25,576 23 92 I. Tax Revenue 22,053 20 80 Direct taxes of which 5,045 5 18 Profit tax 4,197 4 15 n.a. Payroll tax 848 1 3 Yes Tax on interest, royalties, and dividends -- -- -- No Land and property -- -- -- Yes Indirect taxes of which 14,095 13 51 VAT (domestic and on imports) 7,410 7 27 Yes Excise duties (domestic and on imports) 6,478 6 23 Yes Others 207 0 1 Yes Taxes on international trade 2,913 3 11 n.a. II. Non-tax revenue 3,342 3 12 n.a. III. Capital revenue 181 0 1 n.a. B. Provincial revenue 2,153 2 8 n.a. Source: MEF. Note: Social Security Contributions revenue (96.4 billion KHR as of 2018) are recorded under the revenue of the National Social Security Fund. Hence there is no record under Budgetary Central Government Level. Indirect taxes Value added tax (VAT) is the largest component of indirect taxes. It amounts to 27 percent of domestic government revenues. VAT is applied to taxable goods and services at a 10% flat rate. Exempted from VAT are public postal services, hospital, clinic, medical and dental services, the sale of medical and dental goods, the services of transporting passengers using the state-owned public transportation system, educational services, water and electricity, raw agricultural products and liquid and solid waste disposal services. Excise tax is the second largest component of indirect taxes amounting to 17 percent of domestic government revenues. Excises are imposed on the importation or domestic production and supply of certain goods and services with various tax rates including cigarettes, cigars, beer, petroleum products, cars, motorcycle, telephone services, transportation services, hotels, and entertainment services. Direct taxes As is the case in most countries, the fiscal system in Cambodia has various types of direct taxes levied on income or wealth. The system of direct taxes and contributions covers taxes on income, profits, capital gains and property (further details are provided in the Technical Appendix). Direct taxes in Cambodia represent 18 percent of domestic government revenues in 2019, of which corporate income tax and salary tax account for the largest components. 6 Salary tax is the primary source of direct taxes included in the analysis. 9 All individuals, who are either residents or non-residents of Cambodia and currently employed, are liable for salary tax. The salary of resident employees is taxed at progressive rates of 0 to 20 percent, while that of non-resident employees is subject to a 20 percent flat rate. In 2019, the tax system levied salary tax at 5%, 10%, 15% and 20% conditional on the level of earnings. Taxes are levied in the same way on each individual salary, calculated and withheld by employers. They are thus not integrated into a systematic form of a Personal Income Tax. In addition, residents receive a monthly tax deduction of KHR 150,000 for each child (less than 14 years old or up to 25 if he/she is in school full time at a recognized educational institution), and KHR 150,000 for a dependent spouse who is not working. Direct taxes also encompass corporate income tax, which are the largest component of direct taxes, but are not included in the present analysis. Corporate income tax is excluded because the allocation of the tax burden to individual households is generally not possible and is further complicated by the fact that a substantial proportion of businesses in Cambodia operate outside the formal sector. Direct taxes also include: (i) property tax of 0.10 percent levied on the market value of immovable property (land, houses, buildings, and infrastructure) valued at more than KHR 100 million; (ii) rental property income tax of 10 percent on gross rental income; (iii) withholding tax of 15 percent imposed on income from interest, dividends, and royalties; (iv) registration tax of 4 percent imposed on the buyer for transfer of ownership of property or transportation vehicles; (v) tax on means of transportation; (vi) other direct taxes which cannot be modelled in the analysis due to data constraints. Social security contributions Private sector employers in Cambodia with 8 or more workers are required to contribute to the National Social Security Fund (NSSF). The NSSF, established in 2008, was designed to provide both public and private sector employees with: (i) pensions; (ii) injury insurance, and (iii) health insurance. Each month, an employer with 8 or more staff pays 0.8 percent of the average monthly wage of workers towards employment injury insurance, capped at KHR 1,200,000. The employer also contributes 2.6 percent of the monthly wage for health insurance (again capped at KHR 1,200,000). The employee receives full medical assistance and 70 percent of the salary if hospitalized for more than 8 days. Non-tax revenues In addition to tax revenues, the government collects non-tax revenues amounting to 12 percent of total revenues. Non-tax revenues include income derived from state properties; state owned enterprises; autonomous state entities; revenue obtained from the provision of public services; fines and penalties levied against citizens. Government spending While education spending increased over the last few years, social sector spending in Cambodia remains low by international comparison. Cambodia is spending less on education and health as a share of GDP 9 Tax on salary was introduced in 1995. There is no personal income tax, per se, in Cambodia. 7 than other ASEAN countries. Education spending accounted for nearly 3 percent of GDP in 2019, of which primary education absorbed more than half of overall spending. Cambodia’s constitution mandates public schools provide 9 years of free basic education. Health spending was only 1.7 percent of GDP in 2019, only modestly increasing since 2009 from 1.2 percent. Best performers in ASEAN spend nearly twice the amount as a share of GDP. For example, between 2009 and 2019 Malaysia and Vietnam spent, on average, 5 percent of GDP on education. Thailand and Vietnam spent, on average, more than 2 percent of GDP on health. Cambodia, together with Myanmar and Lao PDR, lag among ASEAN countries respectively. Cambodia lags on some human capital indicators due to weaker education attainment, skill utilization, and child nutrition outcomes. 10 Cambodia needs to catchup and align its education and health outcomes with best performers within ASEAN and the world. As the CEQ analysis demonstrates, increased education and health spending to match top performing ASEAN countries could help Cambodia redistribute wealth and support low-income households. Table 2: Central Government Spending, 2019 Type Amount Percent Included in (billion KHR) of GDP CEQ Analysis? Central Government Expenditure 27,317 24.8 Social Benefits 2,310 2.1 Social security 898 0.8 Partial Social assistance to citizens 1,219 1.1 Partial Social assistance to social and cultural entities 138 0.1 No Other social benefits 56 0.1 No Health (MoH spending) 1,924 1.7 Yes Education (MoEYS spending) 3,679 3.3 Yes Other Spending 19,404 17.6 No Source: MEF, MoH, MoEYS. Social protection spending consisting of social assistance and social security amounted to about KHR 2,255 billion, or about 2 percent of GDP, in 2019. About half of Cambodia’s social protection spending was devoted to social security, primarily for former civil servants as pensions. The other half, almost 1 percent, was devoted to social assistance (Figure 3). Social assistance spending in 2019, in particular on cash transfers, was very low in Cambodia albeit better than in some developing countries in East Asia and the Pacific.11 Cash transfer spending was minimal in 2019 despite the full national roll out of the Conditional Cash Transfer for pregnant women and children below 2 years old. However, during the COVID-19 pandemic Cambodia’s social assistance spending increased significantly. 10 World Bank (2020). 11 The National Social Protection Policy Framework (2016–2025) has two main pillars: social assistance and social security. The social assistance pillar includes emergency response, human capital development, vocational training, welfare for vulnerable people. The social security pillar includes pension, health insurance, employment injury insurance, unemployment insurance, disability insurance. 8 Figure 3: Social assistance spending Figure 4: Cash transfer spending (EAP non-HIC) (EAP non-HIC) 6 2 5 Percent of GDP Percent of GDP 4 3 1 2 1 0 0 Papua New… Papua New… Vietnam Philippines Indonesia Philippines Vietnam Timor-Leste Thailand Malaysia Fiji Malaysia Timor-Leste Indonesia Fiji Thailand Samoa Lao PDR Samoa Lao PDR Kiribati China China Kiribati Mongolia Myanmar Myanmar Mongolia Cambodia Cambodia Source: CEQ and World Bank databases and World Bank calculations. Sosa and Wai-Poi (forthcoming). Note: Total SA spending excluding health fee waivers. Cash transfers include CCT and UCT. HIC = High-income country, CCT = conditional cash transfers, UCT = Universal cash transfers. 3. Methodology, Data, and Assumptions CEQ framework This paper applies an internationally recognized methodology developed by the Commitment to Equity (CEQ) Institute to assess how taxes and government spending affect poverty and inequality in Cambodia. The CEQ framework has in recent years emerged as the gold standard to answer questions concerning the implications of fiscal interventions across the income distribution.12 The approach has been applied in over 60 countries, allowing comparisons of Cambodia’s performance with peer lower middle-income countries. The CEQ approach combines detailed household survey data with tax and benefit regulations to estimate what each household paid in taxes and received in benefits. The CEQ approach then describes the distributional effect of the fiscal system sequentially through four income measures. The baseline concept is market income, which is the income that an individual (or a household) earns before any taxes are subtracted and any transfers are included. Three “post-fiscal” income measures—disposable income, consumable income, and final income—are sequentially built by subtracting and adding different forms of fiscal interventions to the pre-fiscal income. Comparisons of the four income aggregates provide information on how fiscal interventions affect the income distribution sequentially from the pre-fiscal income to the final income. Poverty and inequality measures derived from the pre-fiscal and post-fiscal income measures are then compared. A comprehensive incidence analysis is therefore built by 12 See http://commitmentoequity.org/. 9 sequentially quantifying the poverty and inequality impact of direct taxes and transfers, indirect taxes and subsidies, and social spending on health and education. Figure 5: CEQ basic income concepts Source: Excerpted from Lustig (2018). Note: The income concepts are under the scenario of Pensions as Deferred Income. Total consumption obtained from the household survey is treated as a proxy for disposable income and serves as the starting point for deriving the other income measures. 13 This approach is applied in developing countries given the difficulty of directly obtaining a precise measure of disposable income from 13 For poverty measurement purposes, a consumption aggregate was constructed by the National Working Group on Poverty Measurement (NWGPM) under the leadership of the Ministry of Planning with technical assistance from the World Bank. The aggregate includes food, nonfood, imputed housing rents, and the use-value of durable goods. The sum of these items is normalized by household size and by a spatial price index computed for each region (i.e., Phnom Penh, other urban and other rural). 10 the survey data. Second, market income, is computed as disposable income plus direct taxes and contributions minus direct transfers. It includes direct personal income taxes, social security contributions (health and injury), and direct- and near-cash transfers. 14 Market income, which is the primary income concept for the CEQ, is used to rank households into income deciles or quintiles as it represents income before state actions through taxes and transfers. Third, consumable income is computed as disposable income plus indirect transfers/subsidies minus indirect taxes. Finally, final income is consumable income plus imputed values of education and health care services. The monetarized value of education and health care services is conditional on the composition of households, declared use of public services and calculated values of these benefits in specific circumstances. The methodology is based on an accounting approach. Taxes and transfers are added to and subtracted from household per capita income to measure income before and after each fiscal intervention. The per capita household income after taxes and transfers is given by: ℎ = (ℎ − ∑ ℎ + ∑ ℎ )/H (1) where ℎ is income before taxes and transfers, are the taxes paid by the households for range of taxes analyzed, are the transfers received by households for range of transfers analyzed. ℎ and ℎ are the amounts of tax and transfer paid and received by households, respectively, and H is the number of individuals in the household. Data sources The primary data providing individual and household-level information necessary to allocate the fiscal policy elements is the 2019/20 Cambodia Socio-Economic Survey (CSES). 15 This survey is collected by the National Institute of Statistics every two years. CSES 2019/20 includes modules covering education, health, labor market activity, income sources, and household consumption expenditure. CSES also provides a roster of individuals in each household that provides individual and demographic characteristics. The 2019/20 CSES uses the 2019 General Population Census of Cambodia as the sampling frame and is representative at the national level; by Phnom Penh, other urban, and rural areas; and by four geographic zones. The survey was administered to 10,075 households (44,549 individuals). Administrative data for the survey year provides relevant information on the fiscal system. This paper focuses solely on the 2019 fiscal year because it overlaps with the 2019/20 CSES and it is the last year before the fiscal data is affected by the coronavirus disease 2019 pandemic. The Ministry of Economy and Finance is the source for total revenues collected by the government from households (from different taxes) and spending (on subsidies and social programs). Details of Government revenues and spending are also needed for validating the precision of imputations of the fiscal elements to the household sector. 14 Cash and near-cash transfers are transfer income from the government to individuals and families, and include Social Security, unemployment compensation, workers compensation, all means-tested cash transfers, and food and housing benefits. 15 See National Institute of Statistics (2020). 11 These administrative data on taxes and transfers are complemented with information of the Cambodian tax code and laws to understand the design of each fiscal intervention. Additional information on public education and health care services was collected to allow for a more precise assignment of in-kind education and heath transfers to households by the type of service received. The Ministry of Education, Youth and Sport provides enrollment and spending for each level of public education, allowing for an estimation of the spending per student by level of education. For health care, the costing of various health care interventions from Flessa et al. (2018) and Jacobs et al. (2019) is used to assign the monetarized value of health care per person by intervention. We apply this valuation, adjusted by inflation, to health care interventions reported in the 2019/20 CSES data (see Technical Appendix for more details). Empirical Approach and Allocation Overview There are broadly five steps involved in the empirical approach of a CEQ. Step 1: Identify the fiscal interventions that can plausibly be studied using the household survey. Step 2: Allocate the gains and losses from fiscal interventions to individuals and households. In the case of taxes on income, allocations are made to individuals and then aggregated at the household level. In the case of taxes or subsidies on consumption, allocations are made to the household. This process involves (a) identifying who pays taxes and social security contributions, based on the source of income and imputed formality of employment; (b) estimating how much each household paid in taxes and received in transfers using laws, codes, program rules, plausible assumptions, and household survey responses. Step 2 is the most complex and labor- intensive component of the exercise. Step 3: Test the quality of the model by validating the simulated fiscal policy elements against administrative statistics. Step 4: Build the income aggregates according to CEQ concepts. Step 5: Construct poverty and inequality measures for each income concept. In the analysis fiscal policy elements are allocated to individuals or households based on simulation and imputation. Household data usually do not focus specifically on the value of individual taxes paid by households and often collect aggregate information on direct transfers. In the CSES 2019/20 case, the only exception is property tax which is recorded in the data. This means that most of the modeled fiscal interventions need to be simulated using other data provided in the survey. The most important simulation assumptions which have been applied are the following (the Technical Appendix provides a detailed description of the assumptions applied in the analysis): A) Income taxes, health, and social security insurance values have been simulated using the detailed regulations based on the information on employment earnings and self-employment income; this information is combined with an imputation of employment status to ensure that taxes are only levied on formal employment. The tax on property rental is simulated based on the regulations 12 and reported income in the CSES data. The tax on immovable property and the vehicle tax are modeled based on the estimated value of the property which is reported in the data, and the type(s) of car(s) which are reported by the household. There is not enough information in the data to assign the exact value of the vehicle tax, so the tax is imputed based on the characteristics that are available in the CSES. B) Property taxes in Cambodia are levied on means of transportation and on fixed assets. These are assigned to households based on a joint declared sum of property taxes reported in the CSES data, which is disaggregated into means of transportation tax and property tax using information on vehicle ownership and vehicle characteristics. The residual amount is assigned as property tax. C) The CSES data does not provide much detail concerning welfare transfers and itemizes only receipts of scholarships. Therefore, direct transfers are only divided into scholarships and (other) welfare transfers. We adjust transfer receipt values among households eligible to the Conditional Cash Transfer (CCT) to reflect the applicable values of the CCT. Pensions constitute part of overall welfare payments in the CSES 2019/20 data and cannot be separately identified. In Cambodia pensions however are a very small proportion of household incomes – in 2017 for example (when they could be separately identified) they amounted to 3% of total household incomes. D) Indirect taxes and subsidies are calculated based on the general rules applied to consumption taxation and the information provided in the CSES on household expenditures; this applies both to the VAT and excise duties, as well as to the imputation of electricity subsidies calculated based on location and reported cost and usage of electricity. The purchase of food products in rural areas is treated as informal and thus not subject to VAT in the model. E) In-kind transfers are allocated to households based on either age eligibility (education) or declared usage of these services (health). In the case of education values are imputed using information on public school attendance children’s age eligibility. In the case of health care, we use the information on declared use of health care services in the data and impute their value using estimates of intervention costs from Flessa et al. (2018) and Jacobs et al. (2019). Macro-validation of the simulated fiscal policy elements The data used for this study allow us to analyze only a proportion of fiscal interventions since much of government spending and revenues cannot be directly assigned to specific households in the data. The CEQ methodology in Cambodia covers the dimensions of fiscal policy which account for about 19.3 percent of total government revenues and 23.2 percent of total government spending. To test the quality of the survey data and modelled fiscal elements, total imputed/simulated taxes and benefits are compared against administrative statistics. The exercise aims to see how precisely the total taxes (transfers) collected from (paid to) households in the survey matches the total tax collections (spending) reflected in budget documents. Table 3 shows the results of the validation exercise. 13 Consistent with household survey data in many developing countries, our calculations confirm the lack of representativeness of households in the upper tail of the income distribution. This is reflected in the degree to which we can capture the tax base and thus evaluate the allocation of the tax burden across the income distribution and the true implications of the tax and benefit system for redistribution. For example, there is no one in the data with earnings high enough to be paying salary tax in the highest tax bracket. Similarly, underrepresentation of the highest income households suggests some caution is needed in the interpretation of inequality measures under different income definitions and changes in inequality level over time. Several elements of the tax system are simulated relatively well when compared to administrative statistics. For example, the exercise captures about 96 percent of employment injury insurance and 99.2 percent of property tax. Our method of allocating education benefits allows us to assign 99.2 percent of education expenditure and 93.1 percent of health care costs. In the case of education, this is because transfers per student are given their value based on the official statistics. For health transfers, this validates the adopted approach to imputation of the value of health care benefits and indicates that the source of health costs used for the valuation of various procedures is very accurate. It is also worth noting that in this case underrepresentation of top income households may be less damaging for the precision of our simulations since many of the highest income households are likely to opt for the use of private health care. Table 3: Simulation results and validation Amount simulated, Macro data, Proportion of (billion KHR) (billion KHR) simulated to macro (%) Direct transfers 111.6 101.7 109.8 Salary tax 196.1 847.8 23.1 SSC: Employment Injury Insurance 93.4 96.4* 96.9 Health Insurance 303.6 154.2* 196.9 Taxes on interest, royalties, and dividends n.a.** n.a. --- Property tax 148.3 149.5 99.2 Means of transportation tax 91.7 259.6 35.3 Tax on income from rental of movable and 77.0 268.6 28.7 immovable property Registration tax n.a.*** n.a. --- VAT tax 3,707.7 6,259.9 59.2 Specific tax, excise 644.5 1,130.5 57.0 Public lighting tax 43.8 182.5 24.0 Accommodation tax 4.4 20.2 21.8 Indirect subsidies (electricity subsidy) 90.7 134.9 67.2 Education transfers 2,950.9 2,974.8 99.2 Health transfers 1,400.5 1,498.8 93.4 Disposable Income 107,335.6 81,589.7 131.6 Source: Amounts of taxes paid are taken from 2019 Table of Government Financial Operations (TOFE); for SSC values come from NSSF, Report on Ten-Year Achievements 2008-2017 and Action Plans 2018; Education transfers are taken as MoEYS budget, health transfers are taken as MoH expenditures less donor funding. Notes: Values in yearly amounts. *External statistics of SSC: Employment Injury Insurance and Health insurance not available for the most recent years, value provided for 2017 and uprated to account for inflation to 2019. **There are no observations in the CSES data of income from interest, royalties, or dividends. ***Only one case of property sale in the CSES data. 14 The simulations of other fiscal elements are less precise. This applies to such instruments as the salary tax (23.1 percent of the administrative aggregate) or means of transportation tax (35.3 percent). These ratios reflect lack of representativeness of the data at the higher end of income distribution and lack of detailed information in the data with the parameters of the fiscal system. The underrepresentation of richest households has consequences both on the direct side of taxes, given the high tax base and the progressivity of both the salary tax and means of transportation taxes, and the indirect side since expenditure of the top income households might be conducted more specifically in the formal sector and correspond in value with their high incomes. The limited precision in modelling SSC/HI and salary taxes stems from the high degree of progressivity of tax contributions and lack of highest earners in the data. For example, there are only 13 people in the data who would be modeled to pay income tax at the highest rate of tax. While detailed information on the number earners who fall in the top income tax bracket is unavailable for detailed sensitivity analysis, it is clear that information on top incomes is not as precise in household data and the corresponding simulation of taxes simply cannot match the information in administrative statistics. The observed difference between survey estimations and macro data is not unusual for taxes on several of the modeled consumption items.16 The simulated values of indirect taxes less closely match administrative records. Estimations from the survey capture 59.2 percent of the total government revenue from VAT. 17 The undersimulation of some indirect taxes like the public lighting tax or alcohol tax in all sections of the income distribution might reflect underreporting of specific categories of goods such as tobacco and alcohol. Further, the undersimulation for expenditures also reflects the fact that not all final consumption is conducted by private households. VAT is also paid by public institutions and by the corporate sector for which the final sales are exempt from VAT. Additionally, some taxable incomes and a high proportion of some consumption items, like accommodation, are received and paid by visiting foreigners. Indirect subsidies (in this case the electricity subsidy) are simulated with an accuracy of 67.2 percent when compared to official statistics. However, the scale of the electricity subsidy is very small when compared to indirect taxes; according to government statistics, 134.9 billion KHR in 2019 was spent on the electricity subsidy in 2019, as opposed to 6,259.9 billion KHR in revenues from VAT alone. 16 See the appendixes for more details. 17 The total value of household consumption in the CSES 2019/20 is KHR 107,335.6 billion. According to the World Bank, GDP in 2019 was US$ 27.09 billion at official exchange rate and household consumption was 74% of GDP. This means: 27.09*4070*0.74 = KHR 81,590 billion; so less than observed in the data. So, with more consumption we catch less VAT. Usually this can be explained by government, NGO spending, intermediate VAT in zero-rated final goods. However, all of those are low in Cambodia. The only reasonable explanation seems to be that we miss consumption of very rich households which is likely to be subject to VAT. This would suggest though that household consumption in the national statistics would also be too low. Note that we assume that VAT is not paid on food in rural areas – this seems reasonable, but perhaps it is too restrictive. 15 4. The Distributional Impact of Cambodia’s Fiscal System Size effect of fiscal interventions Comparisons of average income per capita at each income concept reveals the size effect of fiscal interventions. The difference between market income and disposable income reflects the net effect of direct taxes and direct transfers. The difference between disposable income and consumable income reflects the net effect of indirect taxes and subsidies. The difference between consumable income and final income reflects the net effect of in-kind transfers. The difference between market income and final income reveals the overall change in average per capita income from the operation of the full fiscal system. Among three main categories of fiscal interventions, indirect and in-kind fiscal interventions have the largest size effect on incomes in Cambodia. There is a sizable decline in average per capita income due to indirect fiscal interventions as revealed in the transition from disposable income to consumable income. The reduction in income indicates that on average indirect taxes paid exceed subsidies received. There is a sizable increase in average per capita income due to the receipt of public education and health services as revealed in the transition from consumable income to final income. Direct taxes and transfers have a small size effect on incomes, reflecting the small size of the formal sector in Cambodia and limited social protection spending as is shown later. We ought to bear in mind though that a high proportion of salary taxes, and other direct taxes, on the highest earners cannot be simulated on the CSES data due to underrepresentation of high-income households, as is common in household survey data. This implies that both the total and the average values of incomes, and the differences between them would have been different if we could account for the top income households correctly. Similarly, average implications of different forms of fiscal interventions would look differently if richest households were fully represented in the data. Figure 6: Average monthly per capita income Figure 7: Net effect of the fiscal system and its (KHR) by income concept elements on average monthly per capita income (%) Market income Full system -0.6 Direct Disposable income -0.6 interventions Consumable income Indirect interventions -3.5 Final income In-kind 3.7 interventions 400,000 500,000 600,000 700,000 -4.0 -2.0 0.0 2.0 4.0 6.0 Source: Authors’ calculations based on CSES Source: Authors’ calculations based on CSES 2019/20 and fiscal data. 2019/20 and fiscal data. 16 Overall Impact of Taxes and Spending on Poverty and Inequality In Cambodia, the overall fiscal system reduces inequality. Bearing in mind the caveat that top income households are often underrepresented in household surveys—a common feature of most surveys, especially in developing countries—the fiscal system in Cambodia reduces inequality by 1 percentage point. Inequality, as measured by the Gini coefficient, falls between market income and final income (Figure 8). Before any fiscal interventions, the market income Gini index is 32.4 percent. Once direct taxes and direct transfers are considered, the Gini index reduces slightly to 32.2 percent at disposable income. Indirect taxes and subsidies have limited effect, and their consideration leaves the Gini index of consumable income remain at 32.2 percent. “In-kind” transfers from health and education, on the other hand, have the largest effect on inequality with final income Gini at 31.4 percent. This reflects the fact that in-kind transfers represent a significant share of pre-fiscal income and proportionally benefit lower- income households more relative to those from the upper end of the income distribution. Figure 8: Gini index before and after fiscal interventions in Cambodia 34 33 32.4 Gini index (%) 32.2 32.2 32 31.4 31 30 Market income Disposable Consumable Final income income income Source: Authors’ calculations based on CSES 2019/20 and fiscal data. While Cambodia does reduce inequality through taxes and transfers, the degree of redistribution is small in international comparison. Figure 9 demonstrates that the redistributive effect of fiscal policy (including in-kind transfers) is low in Cambodia. Some lower-middle-income countries achieve inequality reduction of up to 9 percentage points from the pre-fiscal level. When in-kind transfers are excluded, the redistributive effect in Cambodia is even lower. This is because of the very low impact of direct transfers on inequality. 17 Figure 9: Change in Gini after fiscal policy (including in-kind transfers) 0 -5 LIC Point Change -10 -15 HIC UMIC LMIC -20 -25 Iran Ecuador Colombia Colombia Kenya Chile Namibia Ethiopia Uruguay Brazil Panama Romania Russia El Salvador Belarus Armenia Peru Peru Indonesia Comoros Tanzania Nicaragua Egypt Ghana South Africa Ukraine Tunisia Bolivia Cambodia Uganda Poland Croatia Mexico Mexico Mexico Dominican Republic Guatemala Honduras Argentina Costa Rica Venezuela Albania Jordan Sri Lanka Burkina Faso Paraguay High income Upper middle income Lower middle income Low income Source: CEQ and World Bank databases and World Bank calculations (see World Bank, 2022b). Note: HIC = high-income country, UMIC = upper middle-income country, LMIC = lower middle-income country, LIC = low-income country. At the same time, poorer households pay more in tax than they receive in cash transfers; short-term poverty is slightly higher due to indirect taxation. Overall, the modelled fiscal system increases short-term poverty by 2.1 percentage points, as shown by the transition from market income to consumable income (Figure 10). The net transfer achieved through direct transfers and direct taxes on income are too small to change average household incomes, and thus poverty. However, the net transfer resulting after subsidies and indirect taxes on consumption is negative and large, thereby increasing poverty measured today. In other words, all households face higher prices on goods from indirect taxes. For poor and near- poor households, the higher price reduces purchasing power, potentially reducing net expenditures below the poverty line. It is important to note that poorer households are only out-of-pocket in the short-term; this analysis does not include the economic benefits of health and education services which households will enjoy in the longer-term and which will help reduce poverty. Figure 10: Poverty headcount before and after fiscal interventions 25 Poverty headcount (%) 19.8 20 17.7 17.8 15 10 Market income Disposable income Consumable income Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Note: Per CEQ technical recommendations, in-kind health and education transfers are not included when estimating the impact on poverty. 18 Cambodia is not alone in having fiscal policies that leave the poor out-of-pocket in the short-term. Fiscal policy (excluding in-kind transfers) increases the short-term poverty headcount ratio in most developing countries. Most fiscal systems in these countries rely heavily on indirect taxes. For instance, Tanzania generates 60 percent of its revenues from indirect taxes (Younger et al. 2016). Cambodia’s fiscal system is similarly reliant on indirect taxes, with revenues collected through consumption taxes (value-added tax and excise duties) amounting to twice the revenues collected through income taxes. Although excise duties cover socially costly goods such as alcohol and tobacco, these taxes also cover necessary goods, such as bottled water and communication services (telephone and internet). Most necessities such as meat, seafood, grains, and fruit are exempt from VAT. However, nearly every person is affected by the Cambodian indirect tax system because at least one of the items she consumes regularly has an implicit or explicit indirect tax. Net purchasing power for households along the income distribution therefore decreases after receiving transfers (direct and subsidies) and paying taxes (direct and indirect). Although many lower middle-income countries have fiscal systems that increase poverty, others perform better (Figure 11). Figure 11: Fiscal policy’s impact on poverty headcount ratio across countries 10 Market to Disposable Market to Consumable Percentage point change 5 0 -5 -10 -15 Panama Sri Lanka Mexico Armenia Mexico Mexico Jordan Indonesia Russian Federation Russian Federation Colombia Costa Rica Venezuela, RB Ethiopia Poland Paraguay Belarus Peru Peru Colombia Burkina Faso Ecuador South Africa Guatemala Argentina El Salvador Nicaragua Ghana Chile Honduras Tanzania Brazil Iran, Islamic Rep. Albania Uganda Tunisia Cambodia Uruguay Bolivia Dominican Republic High income Upper middle income Lower middle income Low income Source: CEQ and World Bank databases and World Bank calculations (see World Bank, 2022b). Note: Pensions as deferred income treatment. Relevant international poverty line used for income classes: US$1.90 for LIC, US$3.20 for LMIC, US$5.50 for UMIC and HIC. Fiscal policy has larger repercussions on rural than urban poverty in Cambodia. Areas with higher levels of pre-fiscal poverty are disproportionately impacted. In other words, the net benefits of the fiscal system are less concentrated in rural and other urban areas where more individuals are impoverished. This is largely due to the impact of indirect taxes. The modelling indicates that the current (pre-COVID) fiscal system increases short-term poverty in rural areas by 2.5 points (Figure 12), higher than the 1.8 points in non-Phnom Penh urban areas, primarily driven by indirect taxation. The larger tax result reflects the greater rate of near-poor in rural areas before fiscal policy, meaning they are more likely to fall under the poverty line when indirect taxes are paid. Notably, three features of rural life are accounted for the results: (i) home production of food and other staples is not subject to indirect taxation; (ii) current preferential indirect tax rates and exemptions are modelled; and (iii) indirect taxes on food expenditures are modeled 19 in urban areas but not rural ones, reflecting the likely informal nature of food purchases in rural areas. For these three reasons, the greater rural poverty increase is driven by the underlying welfare distribution and not modeling characteristics. Figure 12: Fiscal policy’s impact on poverty headcount ratio by area of residence 3.0 Percentage point change 2.5 2.0 1.5 1.0 0.5 0.0 Full system Direct Indirect interventions interventions Cambodia Phnom Penh Other Urban Rural Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Note: Pre-fiscal poverty headcount; Cambodia=17.7%, Phnom Penh=4.1%, Other urban = 12.4%, Rural = 22.6%. Table 4: Fiscal policy’s impact on inequality and poverty by area of residence Market income Disposable Consumable Final Income income income Gini coefficient (%) Cambodia 32.4 32.2 32.2 31.4 Phnom Penh 34.8 34.5 34.6 34.0 Other Urban 30.9 30.8 30.9 30.2 Rural 28.0 28.0 28.0 27.5 Poverty headcount ratio (%) Cambodia 17.7 17.8 19.8 n.a. Phnom Penh 4.1 4.2 5.0 n.a. Other Urban 12.4 12.6 14.3 n.a. Rural 22.6 22.8 25.1 n.a. Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Fiscal Impoverishment and Fiscal Gains to the Poor Comparisons of poverty before and after taxes and transfers are applied only broadly indicates the gains or losses the fiscal system creates. But this can fail to capture an important aspect: that some poor are made poorer (or non-poor are made poor) by the tax and transfer system. Fiscal Impoverishment (FI) traces the number of people who are impoverished following the execution of the fiscal policy. 20 Almost 20 percent of Cambodians experience fiscal impoverishment. This means that: 1) some individuals whose pre-fiscal incomes were below the poverty line were net contributors to the fiscal system such that this reduced their net cash position, and 2) some individuals whose pre-fiscal incomes were above the poverty line were net contributors to the fiscal system such that their post-fiscal incomes (consumable income) was below the poverty line. The average reduction in income as a proportion of market income among Cambodians fiscally impoverished was 3.7 percent. Conversely, some of the poor gained through the fiscal system. Fiscal gains to the poor (FGP) traces pre- fiscal poor households to determine how many experienced an income increase following fiscal policy execution. Fiscal policy made only about 2 percent of pre-fiscal poor better off; that is, net beneficiaries of the Cambodian fiscal system. Among pre-fiscal poor beneficiaries, the average increase in market income was 10.6%. Table 5: Fiscal impoverishment and fiscal gains to the poor Market income Change in poverty Fiscally Fiscal Fiscally better- poverty headcount impoverished (% impoverished (% off (% of market headcount (percentage points) of population) of consumable income poor) (%) income poor) 17.7 2.1 19.5 98.4 1.9 Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Contribution of Individual Interventions to Poverty and Inequality Reduction In addition to overall effects of fiscal policy, policy making must assess the contribution of each fiscal intervention to poverty and inequality. We can analyze an intervention in terms of its progressivity and marginal contribution to poverty and inequality. To do this, we use the Kakwani index (Kakwani 1977) to analyze whether a tax or transfer is progressive. 18 A Kakwani index for taxes will be positive (negative) if a tax is globally progressive (regressive), and a Kakwani index is positive if a transfer is progressive in relative terms. A tax is considered progressive if the burden proportionally increases with income. For each tax, the Kakwani index is defined as the difference between the concentration coefficient of the tax and the Gini coefficient of market income. Similarly, a transfer is considered progressive if the entitlement proportionally decreases with income. In the case of transfers, the Kakwani index is the difference between the Gini coefficient of market income and the concentration coefficient of the transfer. To analyze whether a tax or transfer is equalizing, we use the marginal contribution to income inequality measured by the Gini coefficient. The marginal contribution measures the additional change in inequality due to a tax or transfer. The marginal contribution is the difference between the Gini without the fiscal intervention and the Gini coefficient of all income components combined. An intervention is equalizing when the marginal contribution is positive. When there is only a single intervention in the fiscal system, the Kakwani index would have been sufficient to determine whether the intervention is unambiguously equalizing. When there are multiple fiscal interventions, the relationship between inequality outcome and 18 Progressivity can also be assessed using Lorenz curves and concentration curves. 21 progressivity is complex. 19 Comparing the marginal contribution and the Kakwani will reveal whether a tax or transfer is equalizing (unequalizing) despite being regressive (progressive). Box 1: Kakwani Index of Progressivity Two related indexes can be used to assess the progressivity of an intervention (tax or benefit) to compare the effectiveness of different interventions and to measure their redistributive effects: the Kakwani index and the Reynolds-Smolensky (RS) index. Here we define the Kakwani index. Let’s define the following terms: xi is the income of household i. T(xi) is the tax liability of household i. X = Σxi is the total pre-tax income. T = ΣT(xi) is the total tax collected. = is the total tax ratio The Kakwani Progressivity Index for a tax T is KT = CT – GX, where CT and GX are, respectively, the con- centration coefficient of the tax and the Gini coefficient of pre-tax income. If KT > 0, then the distribution of the tax burden is more unequal than that of pre-tax income, meaning that low-income households’ share of the tax burden is lower than their share in original income, so the tax is progressive. If KT = 0, the tax is neutral and if KT < 0, the tax is regressive. Taxes and Transfers and their Effect on Inequality Direct taxes are the most progressive and most equalizing type of tax in Cambodia. Taxes on income and property are both progressive and equalizing; the burden of taxation increases with income and contributes to Gini coefficient reduction. Social security on the other hand and health insurance contributions are regressive and unequalizing. Lower-income households face a proportionately higher burden of social security due to the low maximum threshold (KHR 1,200,000). Indirect taxes are neutral and less equalizing. 19 See Enami, Lustig, and Aranda (2017) and Lambert (2001). 22 Figure 13: Progressivity and redistributive effect of taxes Marginal contribution Kakwani coefficient Marginal contribution to equality 0.2 0.8 0.6 (Change in Gini Points) 0.1 0.4 Kakwani 0.2 0.0 0.0 -0.2 -0.1 -0.4 Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Note: SSC=Social Security Contributions, HI=Health Insurance. All social transfers are progressive in Cambodia, except electricity subsidy and tertiary education spending. Direct transfers are the most progressive but contribute little to reducing inequality due to their small amount and poor targeting as discussed below. Direct social transfers in Cambodia constitute a very small proportion of total incomes, even among the poorest households. For example, among households in the first decile of the market income distribution the average value of direct social transfers amounts to 0.94 percent of disposable income. Moreover, the CSES data collects only aggregated information on welfare transfers, thus making it impossible to distinguish between different programs. Education spending reduces inequality the most. Except for tertiary education, all other education spending is both progressive and inequality-reducing. Primary education has the most equalizing effect. Figure 14: Progressivity and redistributive effect of transfers and subsidies Marginal contribution Kakwani coefficient Marginal contribution to equality 1.0 1.0 0.8 0.8 (Change in Gini Points) 0.6 0.6 Kakwani 0.4 0.4 0.2 0.2 0.0 0.0 -0.2 -0.2 Source: Authors’ calculations based on CSES 2019/20 and fiscal data. 23 Taxes and Transfers and their Effect on Poverty The degree of poverty reduction from direct transfers is not sufficient to fully offset the poverty-increasing effect of taxes. Taxes have the strongest impact on poverty, and indirect taxes lead to higher overall increase in poverty than do direct taxes. VAT has the highest poverty-increasing effect, increasing poverty by about 1.76 percentage points. On the direct tax side, social security contributions and health insurance together have the highest poverty-increasing effect, increasing the poverty rate by respectively 0.13 and 0.33 percentage points. Despite direct transfers reducing poverty, the degree of poverty reduction is limited given the modest generosity of support through this channel. Direct transfers (pre-COVID) reduce poverty by only 0.06 percentage points. Figure 15: Marginal contribution of taxes and transfers to poverty reduction Salary tax SSC HI Property tax MoT tax Rental tax Registration tax All contributions All direct taxes All direct taxes and contributions VAT Specific tax, excise Accommodation tax Public lighting tax All indirect taxes Total direct transfers Electricity subsidy -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 Point change Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Heterogeneity of Effects across Deciles The burden of taxes and benefits varies across market income distribution deciles. In absolute terms, the burden of direct taxes, in particular salary tax, concentrates at the upper end of the income distribution (Figure 16A). In relative terms, the direct tax burden as a share of market income does not vary considerably across the distribution except for the poorest and richest deciles (Figure 16B). However, cash transfers are not sufficient to offset the burden of direct taxes across all deciles of the distribution except the poorest. As such, the net cash position for most households is negative before considering indirect taxes and subsidies. After indirect tax and subsidies, the net cash position is negative for each market income decile. Cash transfers do not offset the tax burden. While in absolute terms, households in the upper end of the income 24 distribution bear a greater burden of direct and indirect taxes, relative to market income, the proportional tax burden does not vary considerably across the market income distribution. Strong progressive, in-kind education and health transfers boost the final position of low-income Cambodian households. The bottom 40 percent are net beneficiaries of the fiscal system, while richer households are net contributors. The reduction in the Gini from 32.4 to 31.4 is due to in-kind transfers (Figure 8). Figure 16: Distribution of direct taxes and direct transfers A. Direct tax burden/benefit by decile B. Direct tax burden/benefit by decile (% of (KHR/month) market income) 5000 1.00 0 Proportion of income (%) 0.50 -5000 0.00 KHR / month -10000 -0.50 -15000 -1.00 -20000 -25000 -1.50 -30000 -2.00 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Income decile Income decile SSC: EIC - ER HI: ER Salary tax SSC: EIC - ER HI: ER Salary tax Witholding tax Rental tax Registration tax Witholding tax Rental tax Registration tax MoT tax Property tax Gov. stipends MoT tax Property tax Gov. stipends Welfare (inc. pensions) Total Welfare (inc. pensions) Total Figure 17: Distribution of direct and indirect taxes, direct transfers, subsidies and in-kind transfers A. 0verall tax burden/benefit by decile B. 0verall tax burden/benefit (% of market (KHR/month) income) 40000 15.00 20000 Proportion of income (%) 10.00 0 KHR / month -20000 5.00 -40000 0.00 -60000 -5.00 -80000 -100000 -10.00 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Income decile Income decile Education transfers Health transfers Total transfers - cash Education transfers Health transfers Total transfers - cash Electricity subsidy Total direct taxes Total indirect taxes Electricity subsidy Total direct taxes Total indirect taxes Total Total Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Welfare transfers have limited poverty-reducing and redistributive effects, however, because they provide low coverage, are small, and are poorly targeted. In 2019, about 10 percent of Cambodian households possessed an IDPoor “equity” card and only 2 percent received conditional cash transfers 25 (CCT) available for pregnant women and children under age 2. The welfare transfers are small and could be better targeted, as Figure 18 shows. Non-poor households also possess the IDPoor equity card to access CCT benefits if they have small children, although at a lower incidence than poor households. The CCT for pregnant women and children provided total benefits US$ 190 during pregnancy and a child’s first 2 years, or approximately US$ 63 each year. This translates to just 6 percent of the national poverty line. Figure 18: Percent of households with an Equity Card and CCT recipients by decile 40% 30% 20% 10% 0% 1 2 3 4 5 6 7 8 9 10 Market income decile Equity Cardholders (IDPoor) CCT Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Because in-kind tertiary education benefits increase with income, they are regressive and not equalizing. Enrollment in tertiary public education is not favorable for poorer segments of the population; about 60 percent of enrollment in tertiary education is among the top 20 percent income households in comparison to 4 percent among the bottom 40 percent. Poor children clearly benefit from having basic education and appear to access primary and secondary public education slightly more than non-poor children. However, due to direct costs and related foregone incomes of remaining in school, these households do not benefit much from tertiary education. In-kind health benefits are progressive but not equalizing. While poorer households use public health more than richer households, they capture relatively smaller shares of in-kind public health benefits. 26 Figure 19: Number of users of public education and health B. Percent of households visiting a public health A. Distribution of publicly enrolled students facility by decile 70% 25% 60% 20% 50% 40% 15% 30% 10% 20% 5% 10% 0% 0% 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Market income decile Market income decile Preschool Primary Secondary Tertiary Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Figure 20: Distribution of in-kind transfers by decile 50B 40B Riel (billions) 30B 20B 10B 0B 1 2 3 4 5 6 7 8 9 10 Market income decile Education Health Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Most of the electricity subsidies accrue to non-poor households. Non-poor households are more likely to be connected to the electricity grid and thus consume more electricity. The poor are more likely to live in neighborhoods without access to an electricity network, and even where electricity is present, they often cannot afford connection or monthly fees. Even if they can afford the connection fees, their electricity consumption tends to be low. Overall, the implicit subsidy embedded in the tariff structure is regressive. 27 Figure 21: Percent of households consuming electricity by decile 100% 80% 60% 40% 20% 0% 1 2 3 4 5 6 7 8 9 10 Per capita market income decile 0–100 kWh 101–200 kWh 201+ kWh Not connected to grid/not consuming Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Note: A kilowatt hour (kWh) is a measure of how much energy is used per hour. Limitations and interpretations The CEQ analysis provides useful insights about the effects of Cambodia’s tax and benefits system, but we must acknowledge some limitations of the CEQ methodology and data, and their implications on interpreting the CEQ results. First, the CEQ looks at only part of the fiscal system and does not fully capture some taxes paid by households or benefits received from public goods. For instance, public infrastructure could be progressive through its impacts on employment and economic activity. Second, while the data are nationally representative, they are not designed to fully represent high-income households, a phenomenon standard with household surveys in all countries. While missing top incomes does not affect our ability to measure poverty, it has important consequences for inequality measures. If many richer households are not captured in the survey, “true” inequality will be higher. This might attenuate distributional implications and redistributive effects of various instruments since the higher burden of taxes on non-represented rich households is not included. This is not a flaw in the survey data, which are not designed to accurately measure the very top of the income distribution, but it does have implications for fiscal incidence analysis. Another limitation is the challenge of assigning direct taxes to individuals when the data does not explicitly distinguish between those in formal employment from those in informal employment. As such, we needed to impute formal employment based on proxy characteristics from the data to model the incidence of direct taxes, as these are not paid by individuals working in the informal economy. As suggested by the macro-validation of the CEQ exercise for Cambodia, discrepancies exist between official statistics and modelled aggregate values of several elements of the tax and benefit system. 20 These 20 Several elements of the tax system are simulated relatively well when compared to administrative statistics. For example, the exercise captures about 96 percent of employment injury insurance, 99.2 percent of property tax, 99.2 percent of education expenditure, and 93.1 percent of health care costs. Simulations of other fiscal elements are less precise. This applies to instruments such as the salary tax (23.1 percent of the administrative aggregate) or means of transportation tax (35.3 percent). The limited precision in modeling social security contributions, health 28 discrepancies could be lessened with more detailed micro-level household data and more detailed administrative information on how particular elements of the system operate. While acknowledging the limitations of the CEQ methodology with the existing data for Cambodia, the analysis does provide useful insights into the operation of the tax and benefit system in Cambodia. 5. Potential Reforms and Policy Direction To enhance the poverty-reducing effect of fiscal policy, the following Conditional Cash Transfer (CCT) simulations provide examples of potential policy reforms which could address the challenges of the current system. Each reform scenario has a specific cost and scale of poverty reduction. As presented earlier, the IDPoor equity card presents a useful device for targeting CCT towards households at the lower end of the income distribution. The card can be combined with other features of households, such as their demographic characteristics, location, or access to facilities to further narrow down the groups of recipients. In countries where such a program is not in place, targeting often relies only on the latter information which often makes targeting less precise and means that significant proportions of benefits are often allocated towards richer households. Increasing the amount of targeted cash transfers to households with children under 3 has a sizable poverty-reducing effect. Simulations demonstrate increasing cash transfers targeted to IDPoor households with young children from about US$ 63 per year (US$ 190 over 1,000 days) to US$ 200 per year would reduce the poverty rate by 0.11 percentage points. Similarly, a targeted cash transfer of US$ 400 per year would reduce the poverty rate by 0.23 percentage points. Policies calibrated to the same total costs but implemented in a universal fashion (i.e., without the use of the IDPoor equity card to identify eligibility), thus allocating lower average values to the households which fulfill the demographic requirements, have a much smaller poverty-reducing effect. Table 6: Poverty reduction and cost of four CCT scenarios aimed at IDPoor households CCT scenario Design Poverty Cost (millions reduction (pp) US$/year) CCT Version 1 200 US$ per year/child under the age of 3 in -0.113 16.5 households identified by the IDPoor equity card CCT Version 2 20 US$ per year/child under the age of 3 -0.062 15.7 CCT Version 3 400 US$ per year/child under the age of 3 in -0.230 32.9 households identified by the IDPoor equity card CCT Version 4 40 US$ per year/child under the age of 3 -0.113 31.4 Source: Authors’ calculations based on CSES 2019/20 and fiscal data. insurance, and salary taxes potentially stems from the high degree of progressivity of tax contributions and lack of highest earners in the data. 29 Figure 22: Absolute gains of four CCT scenarios aimed at households with young children A. Absolute gain per HH by market income decile B. Absolute gains per HH by demographic structure 3,000 1,500 2,500 2,000 1,000 KHR/month KHR/month 1,500 500 1,000 500 0 0 No children No children, 1 child 2 children 3+ children 1 2 3 4 5 6 7 8 9 10 or elderly with elderly CCT Version 1 CCT Version 2 CCT Version 1 CCT Version 2 CCT Version 3 CCT Version 4 CCT Version 3 CCT Version 4 Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Increasing the amount of targeted cash transfers to households regardless of whether they have young children has an even greater poverty-reducing effect. Simulations demonstrate that allocating cash transfers targeted to IDPoor households on a per-capita basis regardless of the age structure of the households at the value of US$ 200 per year would reduce the poverty rate by 1.4 percentage points. Table 7: Poverty reduction and cost of two CCT scenarios aimed at IDPoor households CCT scenario Design Poverty reduction (pp) Cost (Millions US$/year) CCT version 5 IDPoor households: 150 -0.996 236.8 US$ per person CCT version 6 IDPoor households: 200 -1.418 315.7 US$ per person Source: Authors’ calculations based on CSES 2019/20 and fiscal data. 30 Figure 23: Cost and poverty reduction of four CCT scenarios aimed at small children A. Absolute gain per HH by market income decile B. Absolute gains per HH by demographic structure 21,000 12,000 18,000 9,000 15,000 KHR/month KHR/month 12,000 6,000 9,000 6,000 3,000 3,000 0 0 No children No children, 1 child 2 children 3+ children 1 2 3 4 5 6 7 8 9 10 or elderly with elderly CCT Version 5 CCT Version 5 CCT Verison 6 CCT Version 6 Source: Authors’ calculations based on CSES 2019/20 and fiscal data. 6. Conclusion This paper performs a fiscal incidence analysis for Cambodia using the CEQ approach. The present analysis relies on the 2019/20 Cambodia Socio-Economic Survey (CSES) and 2019 fiscal data. The CEQ approach is a tool that helps assess the redistributive effect of a fiscal system and its individual interventions, offering invaluable insights for policy. Moreover, it offers a platform to simulate policy scenarios and assess their distributional implications. The 2019 fiscal system in Cambodia, per the CEQ model, reduces inequality yet lowers incomes of poor households in the short-term. Note, these results predate the substantial increase in cash transfers to poor and vulnerable households in response to the COVID-19 pandemic. The informal sector accounts for a large share of the economy in Cambodia, hence taxes are largely collected indirectly. Poorer households pay indirect taxes as part of the purchase price of goods and services, which reduces their purchasing power, even with tax exemptions on necessities. In the absence of offsetting transfers, their net cash position is negative on average. Short-term poverty can increase following execution of fiscal policy because direct transfers are too small to compensate for the impact of indirect taxes on poor and vulnerable households. Note, households also benefit from in-kind services from spending on health, education, and infrastructure, but they do not receive these services in cash and therefore do not increase their incomes today. However, it is possible for fiscal policy to reduce poverty in the long run through investments in human capital and infrastructure. Other components of spending not captured in the CEQ exercise, such as public infrastructure, can make the overall system more progressive through their indirect effects on inclusive economic growth. Cambodia is not alone among developing countries in seeing a short-term poverty increase due to fiscal policy. Several peer lower middle-income countries also experience net poverty increases due to their 31 fiscal policies. This is not surprising due to their heavy reliance on indirect taxation. Nonetheless, some low- and lower-middle-income countries apply fiscal policies more effectively, achieving both inequality and short-term poverty reduction. The findings for Cambodia suggest some fiscal system reforms are needed to improve its distributional effect, drawing on the international lessons for progressive fiscal policy in World Bank (2022b). Without reform, poorer households will continue paying more into the fiscal system than they receive from it in cash terms. Cambodia must look to spend more and spend better to enhance the redistributive effect of the fiscal system. Higher transfers to poorer households and better targeted transfers would help improve the short-term poverty impact and cost-effectiveness of these transfers. Higher transfers will offset losses in purchasing power from indirect taxation, while improved identification of poor and vulnerable households ensures those most in need benefit. Some of these gains could be achieved by reallocating spending away from regressive subsidies such as electricity. For example, electricity subsidies cost 0.1 to 0.15 percent of GDP and do not benefit poorer households very much; if spent instead on targeted cash transfers, this would double the pre-pandemic budget. Further, maintaining some of the increased pandemic spending will bring Cambodia’s social assistance budget closer to regional and international levels while offsetting much of the short-term increase in poverty from fiscal policy. As a rough calculation, the poverty-reducing effects of Cambodia’s COVID-19 response, if sustained after the pandemic, would largely eliminate the short-term impact of fiscal policy on poverty. 21 Even if social assistance spending was not maintained at COVID-19 levels but sustained at half the level, the impact of fiscal policy on short-term poverty would put Cambodia closer to the middle of the upper-middle-income country range (Figure 24). Figure 24: Impact of fiscal policy on short-term poverty in LMICs 6 Pre-COVID impact 5 Percentage point change Half COVID 4 COVID response* response* 3 2 1 0 -1 -2 -3 -4 Cambodia could also spend more on health and education. Countries that see significant reductions in inequality usually do so through health and education spending. Of all the fiscal interventions modeled for Cambodia, spending on primary and secondary education contributed most to inequality reduction. Allocations to primary and secondary education to improve access and quality of services would be pro- 21 As World Bank (2022a) documents, Cambodia increased its social assistance spending in response to the COVID- 19 crisis from 0.1 percent to 0.7 percent of GDP between 2019 and 2020. The report estimates that this mitigated the poverty-increasing impact of the pandemic by 1.9 percentage points, which is applied to the 2019 CEQ results of this paper to roughly update Cambodia’s fiscal policy impact on short-term poverty. This is a rough approximation only. 32 poor. Improving spending, access, and quality of health care would also be pro-poor, helping to reduce out-of-pocket expenditures for poorer households and improve their longer-term prospects for better health and productivity. 33 References Jacobs B, Hui K, Lo V, Thiede M, Appelt B, Flessa S. 2019. Costing for Universal Health Coverage: Insight into Essential Economic Data from Three Provinces in Cambodia. Health Economics Review 9, 29. https://doi.org/10.1186/s13561-019-0246-6. De La Fuente, Alejandro; Rosales, Manuel; Jellema, Jon. 2017. The Impact of Fiscal Policy on Inequality and Poverty in Zambia. Policy Research Working Paper; No. 8246. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/28907. Flessa S, Jacobs B, Hui K, Thiede M, Appelt B. 2018. Costing of health Care Services in Three Provinces of Cambodia. Final Report. Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) GmbH. Sosa, Mariano, and Matthew Wai-Poi. Forthcoming. Fiscal Policy and Equity in Developing Countries: A Survey of International Patterns and Lessons. Background paper for Poverty and Shared Prosperity 2022, World Bank, Washington, DC. World Bank. 2020. The Human Capital Index 2020 Update: Human Capital in the Time of COVID-19. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/34432. World Bank. 2022a. Cambodia Poverty Assessment: Toward a more inclusive and resilient Cambodia. World Bank, Washington, DC. © World Bank. World Bank. 2022b. Changing Course: 2022 Poverty and Shared Prosperity Report. World Bank, Washington, DC. © World Bank. https://openknowledge.worldbank.org/handle/10986/37739. World Bank. 2023. Cambodia Economic Update - Post-COVID-19 Economic Recovery : Special Focus - From Spending More to Spending Better : Toward Improved Human Development Outcomes (English). Washington, D.C. : World Bank Group. http://documents.worldbank.org/curated/en/099051523221517821/P17734003f2bcf02 b0a89500f61b1f3ff7b. 34 Appendix Appendix A CEQ Technical Details Additional Figures and Tables Box A.1: Royal Government of Cambodia Interventions in Response to COVID-19 In response to COVID-19, the RGC introduced emergency interventions to mitigate the adverse COVID-19 impacts on the economy and households and to ensure economic and social stability in 2020. The interventions include social assistance measures, measures to stabilize the livelihoods of businesses and workers, measures to finance businesses, and fiscal-related measures to increase financing and restore and promote growth in the post- pandemic context. In addition, the RGC is implementing a COVID-19 masterplan to address the health response to the crisis that has been directly supported through World Bank financing, including procurement of medical supplies and equipment for treatment of infected cases and testing. Announced measures include: • introduction of COVID-19 cash transfers to poor and vulnerable households; • tax relief for the tourism and garment, footwear and travel (GFT) goods manufacturing sectors; • retraining and upskilling programs for laid-off workers in tourism and GFT goods sectors; • unemployment benefits for suspended workers in GFT goods sector of US$ 70 per month (US$ 40 paid by the government and US$ 30 paid by the factory); • unemployment benefits of 20 percent of minimum wage for suspended workers employed in the tourism sector (paid by the government); • exemption of property registration tax for purchases below US$ 70,000; • additional capital injection for the Rural Development Bank (RDB) to support agro-processing firms; • establishment of a new SME Bank designed to support SMEs through co-financing and risk sharing with commercial banks; • establishment of a new Credit Guarantee Corporation of Cambodia (CGCC) that has launched the first guarantee scheme; • establishment of a new cash for work program targeting small infrastructure improvement projects; • measures to improve the ease of doing business; • actions to improve trade facilitation, including post audit clearance; and • measures to inject liquidity into the financial sector through the temporary lowering of capital and reserve requirements as well as regulatory forbearance. The total fiscal cost of RGC’s COVID-19 response amounted to US$ 823 million in 2020 and US$ 689 million in 2021, out of which US$ 882 million is expected to be spent on social protection programs (US$ 500 million for cash transfers, US$ 260 million for cash for work, US$ 122 for wage subsidies and training). SME support (RDB, SME Bank, and CGCC) amounts to another US$ 600 million over 2020-2021. 35 Figure A.1: Fiscal Policy’s Impact on Poverty Headcount Ratio (bars) by Area of Residence; Pre-fiscal Poverty Headcount Ratio (dots) by Area of Residence 3.0 30 Percentage point change 2.5 25 2.0 20 Percentage 1.5 15 1.0 10 0.5 5 0.0 0 Cambodia Phnom Penh Other urban Rural Poverty headcount Poverty gap Poverty severity Source: Authors’ calculations based on CSES 2019/20 and fiscal data. Figure A.2: Government Expenditure on Education and Health (% of GDP), 2009–2019 Average A. Education B. Health Malaysia Thailand Vietnam Vietnam Thailand Brunei Darussalam Brunei Darussalam Malaysia Indonesia Singapore Philippines Cambodia Singapore Philippines Lao PDR Indonesia Cambodia Lao PDR Myanmar Myanmar 0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Source: World Development Indicators. Figure A.3: Government Expenditure on Education and Health (% of GDP), 2019 A. Education B. Health Malaysia Thailand Vietnam Singapore Philippines Brunei Darussalam Vietnam Thailand Malaysia Cambodia Philippines Indonesia Cambodia Singapore Indonesia Lao PDR Lao PDR Myanmar Myanmar 0.0 1.0 2.0 3.0 4.0 5.0 0 0.5 1 1.5 2 2.5 3 36 Source: World Development Indicators. Technical Details At the national level, there are two major departments that collect domestic and trade taxes (General Department of Tax and the General Department of Customs and Excise respectively), and two non-tax departments (General Department of State Property and Non-Tax Revenue and the General Department of Financial Industry) that collect non-tax revenue. These tax and non-tax administrations are under the Ministry of Economy and Finance. Provincial offices have the discretion to collect stamp tax on motor vehicles, excise on public lighting and wealth transfer tax. Cambodia’s tax system has gone through several reforms over two decades. These include removing many of Cambodia’s tax and custom duties (early 1990), introducing 10% Value-Added Tax (VAT) (1999), and introducing 0.1% property tax (December 2009). Most recently, the exempted salary threshold has also been increased from KHR 500,000 (effective before January 2015) to KHR 1,200,000 (effective from January 2018), while the allowance for children and spouse has also been increased from KHR 75,000 to KHR 150,000 (Law of Financial Management, 2017). More importantly, the simplified and estimated tax regime was abolished in 2016, for which small taxpayers likely paid profit tax at approximately 2%, which was lower than the 20% flat rate paid by large taxpayers under self-declaration or real tax regime. The only remaining tax regime in Cambodia is self-declaration or real regime. Additionally, transfer pricing rules based on the OECD’s arm’s length principal – guideline for multinationals to record and report all transactions between two companies within the same corporate group – were also introduced in October 2017 to prevent transfer pricing abuse or transfer mispricing, which is an approach that corporation use to manipulate market prices and underreport actual profits that in turn have negative effects on tax collection. The Revenue Mobilization Strategy 2014–2018 focused on efficiency more so than introducing new taxes or increase tax rates. Increased revenue collection has been underpinned by government efforts to modernize tax and customs administration, together with strong economic activity. The 2019–2023 revenue mobilization strategy has been adopted, of which the key objective is to further modernize tax and customs administration and policy. The Royal Government of Cambodia adopted the National Social Protection Policy framework (2016– 2025) which has two main pillars: social assistance and social security. The first pillar on social assistance includes emergency response, human capital development, vocational training, welfare for vulnerable people. The second pillar on social security includes pension, health insurance, employment injury insurance, unemployment insurance, disability insurance. The National Social Protection Policy framework (2016–2025) defined a set of public interventions to assist households and individuals to better manage risks, shocks and reduce their vulnerability by providing them with a better access to social service and employment. Social protection is beyond simple assistance as it provides investment in building human capital and bolstering long-term productivity and promote inclusive growth. Social protection interventions include cash benefits, direct in-kind transfer of goods and services, and tax breaks with social 37 purposes. The benefits are primarily targeted to low-income households and vulnerable individuals such as disabled, elderly, unemployed and so forth. Before the National Charter on Health Financing was adopted in 1996, the general population had access to free medical care at public health facilities. The health sector reform in the context of the Charter allowed public health facilities to collect user fees from all patients, except those who are poor and aimed primarily to reduce unofficial charges and household out-of-pocket expenditure, improving quality of health care, and providing incentives to medical staff. Modeling direct taxes and social security contributions on CSES 2019/20 data This section highlights the main characteristics of the tax and social security system in 2019, and the assumptions made in estimating taxes and benefits for each household in the 2019/20 CSES. Earnings reported in the CSES data are net of taxes and social security contributions (SSCs). Consequently, the modeling sequence of tax and SSC calculations needs to be reversed. Normally, social security contributions are subtracted from gross earnings to derive taxable income on which income taxes are levied to derive net income. In the process of imputing of taxes and SSCs, salary taxes are first computed on net incomes and then further SSCs are calculated with respect to taxable income. The key parameters of the direct tax system are presented in Table A.1. Corporate taxes (e.g., profit and turnover tax) are not modelled as part of the CEQ exercise as the allocation of these taxes to individual households is generally extremely difficult and made additionally complicated by the fact that a substantial proportion of small businesses operate outside of the formal sector. 38 Table A.1: Direct tax system parameters in Cambodia, 2019 Group Name Amount/rate SSC parameters: SSC maximal wage cap 1 200 000 KHR SSC minimum wage cap 200 000 KHR SSC rate, private sector employer 0.80% contributions HI – employer contribution 2.6% HI – employee contribution 0% Salary tax parameters: Salary tax rate 1 5% Salary tax threshold 1 1 200 000 KHR Salary tax rate 2 10% Salary tax threshold 2 2 000 000 KHR Salary tax rate 3 15% Salary tax threshold 3 8 500 000 KHR Salary tax rate 4 20% Salary tax threshold 4 12 500 000 KHR Salary tax: child & spouse credit 150 000 KHR Other tax parameters: Capital tax/withholding tax rate 15% Property tax rate 0.1% - property tax threshold 100 000 000 KHR Property rental tax rate 10% Registration tax rate 4% Means of transportation tax: Motorcycle 0 KHR Van – old 250 000 KHR Van – new 141 000 KHR Car – old 141 000 KHR Car – new 100 000 KHR Source: NSSF for SSC parameters (http://www.nssf.gov.kh). Notes: Salary tax and SSC parameters are monthly values; other values are expressed in yearly amounts; Values for means of transportation tax estimated from CSES data. The tax system parameters were essentially the same in 2019 and 2020 to clarify. Salary Tax To impute the salary tax individuals paid on their declared earnings, we first identify which individuals pay taxes by specifying the formality status of their employment. Formal employment is defined using information on occupation (ISCO codes) in the CSES and the degree of informality has been calibrated to match the levels reported in the Cambodia Labor Force Survey (NIS, 2021). People working in agriculture or trade, as well as Cambodians working abroad are assumed not to be formally employed, and consequently not to pay salary tax, health insurance or Social Security Contributions. Incomes from the main and secondary employment are handled separately, as there is no system of aggregating salary tax from different income sources. Because residents are entitled to a child and spouse tax credit, children are then assigned to a parent within the household. For tax purposes, children are identified as those aged less than 14 or up to 25 who are attending school. Partners are defined as married spouses without income from employment. Salary tax is calculated by reversing the tax schedule (Table A.1). This can be 39 understood as finding the tax amount that – after subtracting from taxable income - would result in the reported net wage in the data, given salary tax rates, thresholds, and the family composition. Social Security Contributions Under Social Security Contributions (SSCs), the following three areas of the fiscal policy are examined here: (i) pensions; (ii) injury insurance; and (iii) health insurance. (i) Pensions: Unlike in earlier years of the CSES data, the 2019/20 dataset does not allow us to specifically separate out pension benefits and thus to treat them correctly in accordance with the CEQ terminology as deferred income (and thus constitute an element of Consumable, Disposable, and market income), since there is no system of pension contributions these are not accounted for in the calculations. Pension benefits in the 2019/20 data are declared together with other welfare benefits and are as a result treated as such. However, since pensions are primarily received by former civil servants, and have been very rarely reported in the CSES data in the past, this omission should not constitute a serious source of bias of the overall results. The National Social Security Fund for Civil Servants and the National Fund for Veterans, the two key institutions involved in the provision of pension benefits are financed from the central budget, and current benefits are not related to the history of contributions. Civil servants thus accumulate entitlements to future benefits without explicitly paying contributions towards these benefits, which implies that such pensions constitute a form of deferred earnings. (ii) Injury insurance: Employment Injury Insurance exists for private employees managed by the National Social Security Fund. Employers who employ 8 or more persons are obliged to pay 0.8% of the average monthly salary in the enterprise as individual SSC contribution which covers all employees in the enterprise. The minimum cap for wages is KHR 200,000 and the maximum cap is KHR 1.2 million. In 2017 the system covered 1.183 million employees in 8507 enterprises (NSSF, 2018). Unfortunately, the CSES data does not provide any information on the size of company in which the respondents are employed. Several criteria related to the sector of employment, occupation and earnings have been used in the simulation to narrow down the likelihood of being covered by the system of contributions. The systemic thresholds and rates are applied to the declared earnings of respondents identified as covered by the system after accounting for salary tax payments. In a similar way to taxation, the SSC schedule needs to be reversed to obtain an estimation of the contribution to injury insurance. (iii) Health insurance: In September 2016, the Ministry of Health of Cambodia initiated a Health Insurance scheme for private and public employees to cover their health care expenditure beyond employment injuries covered by injury insurance. The scheme in 2019 set a contributions rate of 2.6% of the employees’ average monthly salary for health insurance. The contributions are paid entirely by the employer. The Ministry of Health reported that 1,156,682 persons were registered under the new health insurance scheme and paid KHR 154.8 billion in contributions in 2017 (See: NSSF report 2018). 40 Health insurance contribution is computed in a similar way as injury insurance. Employer health insurance contributions are calculated separately based on grossed-up wages. Health insurance is calculated for the same minimum and maximum thresholds as SSC injury insurance. Other direct taxes paid by households The following five other direct taxes are also modeled in the exercise: (i) capital gains/withholding tax; (ii) tax on property rental; (iii) registration tax; (iv) tax on means of transportation; (v) property tax. (i) Withholding Tax: Interest and dividends income from the CSES data are used to impute the values of the withholding tax paid by individual household, with a fixed 15% rate based on 12-month recall of household income coming from other sources. Tax withholding applies to income from a selection of services, rent, and savings including: (a) Payments for services e.g., consulting, interest payment and royalties (15%); (b) Interest paid by a domestic bank or savings institution on deposits (varies depending on the recipient). (ii) Tax on Property Rental: A 10% rate is applied to the monthly household rental revenues from the section on building ownership to estimate the amount paid as tax on property rental. (iii) Registration Tax: Information on the value of income from the last 12 months from sale of vehicles, land and property is used to calculate the registration tax paid. The main caveat here is that the tax is assigned to the seller, not the purchaser (who is not identified in the data). Thus implicitly – from the distributional point of view – we assume that the buyer and the seller are placed in a similar section of the income distribution. (iv) Tax on Means of Transportation: This tax is imposed on selected means of transportation and its value depends on the parameters of the vehicle. The CSES data captures the state (new/old) of the durable good when it was purchased. The state of the vehicle that was purchased is used in the simulation. Since the survey did not capture all the necessary parameters of the vehicle which determine the value of taxes (like year of production and engine size), the values of taxes are imputed using information on the total value of property tax paid in the last 12 months (a single total value including both movable and immovable properties). The rates are derived using imputation from a regression of the value of the total tax on the number of new and old cars, and the area of the house subjected to the tax. From 2017 onwards, motorcycles were not taxed under the Means of Transportation Tax, so the tax on means of transportation imputed in the 2019/20 data is only for cars. (v) Property tax: Information on self-assessed property value could be used to calculate the total tax. However, the CSES data also includes a total value of tax on all property (transportation and other). Property tax paid is therefore imputed after subtracting the transportation tax (given the number of vehicles in the household and the procedure described above) from the total value of the tax reported in the data. Direct taxes which cannot be modeled due to data constraints are as follows: 41 (vi) Corporate income taxes (CIT): Previously known as annual tax on profit (TOP), corporate income taxes are the largest component of direct tax but are not modeled for reasons explained above. A 20% flat rate of TOP is applied to medium and large enterprises, while a progressive rates scale ranging from 0 to 20% is applied for small enterprises. 22 Enterprises engaged in the production or exploitation of oil, gas and natural resources are subjected to TOP at the rate of 30%. Insurance companies are taxable at 5% on the gross premium income and 20% on other income derived from non-insurance or reinsurance activities. TOP rates are set at zero percent for enterprises engaged in a Qualified Investment Project (QIP) during the period of tax exemption 23 and 9% for 5 years of transitional period as determined by the Council for the Development of Cambodia (CDC). (vii) Tax on unused land: Tax levied annually at 2% of land value are not modeled as the CSES does not include details on the use of land owned by households. (viii) Stamp tax: A uniform rate is applied for legal documents. Not modelled due to lack of information in the data. Modeling indirect taxes and subsidies on CSES 2019/20 data Three main categories of indirect taxes existed in Cambodia in 2019: (i) value added tax (vat); (ii) specific tax (excise); and (iii) public lighting and accommodation taxes. Below, we briefly describe how they operate and how the values of these taxes are imputed to individual households in the CSES data. The system of indirect taxes in Cambodia (i) Value Added Tax: The Value Added Tax (VAT) is applicable to real-regime enterprises and is charged at 10% on the value of the supply of most goods and services. A 0% rate is used for suppliers of goods and services for export-oriented garment enterprises as well as domestic suppliers of paddy rice (PWC, 2015). There are several VAT-exempted products. The main categories of exempted goods and services are public postal services, medical and dental services, electricity, transportation of passengers by wholly state-owned public transport systems, insurance services, primary financial services, and land. (ii) Specific Taxes: Excise duties are levied on some goods (VBD-LOI, 2019): such as • Air tickets sold in Cambodia (10%) • Entertainment services (10%) • Cigarettes and cigars (20%) • Alcoholic beverages excluding beer (35%) • Beer (30%) • Non-alcoholic beverages (10%) 22 The classification of large, medium and small enterprises is based on their turnover, legal form and other criteria. 23 Tax exemption period for QIP is usually “trigger period” + 3-year period + priority period. Trigger period usually ends after first profit is derived considering from the date of the receipt of the QIP certificate, while priority period varies by type of business or industry that the QIP is involved and size of capital. The maximum total years could reach 9 (see link to CDC’s website http://www.cambodiainvestment.gov.kh/investment-scheme/investment- incentives.html for more detail). 42 • Telecommunications services (3%) • Automobiles and spare parts (15, 25, 45%) – 15% is assumed in the simulations • Petroleum, diesel, and gasoline (4, 10, 25, 33%) – 10% is assumed in the simulations (iii) Public Lighting and Accommodation Tax: Public lighting tax is charged at 3% and is an indirect tax levied on all types of alcohol and tobacco. It is levied on all steps of the supply chain. Accommodation tax is levied on the supply of accommodation services by hotels, hostels, camping grounds, etc. It does not include private rental and is levied at rate of 2%. Modeling of indirect taxes on CSES 2019/20 data The 2019/20 CSES data does not include a detailed expenditures diary and most of expenditure items are reported in the form of recall expenditures. Therefore, recall values are used to model all indirect taxes in this exercise. Food expenditures are recorded as recall expenditures during last 7 days. Based on the frequency at which households purchase various goods and services, expenditures on non-food items are recorded as 1-, 6- and 12-month(s) recall values. For simulation purposes all values are converted to monthly equivalents. Table A.2 provides an overview of the parameters used to compute VAT, Specific Tax (Excise), Public Lighting and Accommodation tax for 115 product categories available in the data. The VAT rate is generally uniform at 10% for most products, except for exempted goods and services. CSES data captures postal services expenditures under a broader “Communication and postal services” category. Full 10% VAT rate was assigned to this category, assuming that postal services comprise only a small portion of the whole group (the category includes phone charges and internet service). For transportation services, it was not possible to distinguish if the service used was provided by a private or public company. As a result, a decision was made to assign 0% VAT assuming that many private service providers constitute part of the informal economy. It was also assumed that unprocessed food, bought in the rural area is beyond the scope of VAT, excluding it entirely from VAT taxation, while still taxing processed food products including tea, spices, beverages, alcohol, and tobacco. The tax base for Public Lighting and Accommodation was identified by reported expenditures on accommodation services, alcohol, and tobacco accordingly. For obvious reasons, Public Lighting tax is computed only for the final value of sold products. For Specific Tax, a uniform rate is assigned for whole groups of products in the data. For example, it is not possible to assign the tax for fuels only, because fuels are part of a category “Operation of transport equipment” which also includes parts, repairs and driving lessons. But we assume that fuel/gasoline composes most of these expenditures and therefore, the tax rate for gasoline is assigned to the entire category. Similarly, because of lack of granularity in the data, a uniform tax rate is assigned to alcohol and tobacco (35% and 20% respectively). Table A.2: VAT and specific tax rates Code Description VAT Specific Public Accommodation rate Tax lighting tax rate tax 1 rice, quality 1 (kg) 0 0 2 rice, quality 2 (kg) 0 0 3 rice noodles/ fried noodle (kg) 0 0 4 Chinese noodle/ Khmer noodles (kg) 0 0 5 other cereals or flour and other bakery products (kg) 0.1 0 43 6 bread (piece) 0.1 0 7 mudfish (kg) 0 0 8 catfish (kg) 0 0 9 other inland fish (kg) 0 0 10 shrimp/lobster (kg) 0 0 11 crabs (kg) 0 0 12 other seafood (kg) 0 0 13 preserved or processed fish/seafood (kg) 0 0 14 pork (kg) 0 0 15 beef (kg) 0 0 16 duck (kg) 0 0 17 chicken (kg) 0 0 18 other meat products (kg) 0 0 19 eggs and egg-based products (piece) 0 0 20 milk or yoghurt(can) 0.1 0 21 oils or fats (kg) 0 0 22 banana (set) 0 0 23 mangoes (kg) 0 0 24 longan (mien) (kg) 0 0 25 papaya (kg) 0 0 26 tamarind (kg) 0 0 27 coconut (piece) 0 0 28 nuts and edible seeds (kg) 0 0 29 maize and corn crop (piece) 0 0 30 other fresh fruits (kg) 0 0 31 dried and preserved fruits (kg) 0 0 32 trakun (watercress marsh cabbage) (kg) 0 0 33 spring onion/ garlic/ leeks leaves (kg) 0 0 34 cabbage/ leaves (kg) 0 0 35 gourd, cucumber, pumpkin, eggplant (kg) 0 0 36 other fresh vegetables (kg) 0 0 37 prepared and preserved vegetables (kg) 0.1 0 38 tubers (potato, sweet potato, carrot, radish) (kg) 0 0 39 mushrooms/ dried mushrooms (kg) 0 0 40 pea, bean/ soybean/ bean sprout (kg) 0 0 41 sugar cane/ palm sugar (kg) 0 0 42 sweets (kg) 0 0 43 salt (kg) 0 0 44 pepper (kg) 0.1 0 45 monosodium glutamate (kg) 0.1 0 46 fish sources/ soy sources/ chilly sources (liter) 0 0 47 other ingredients (kg) 0.1 0 48 nutritive tablets (kg) 0.1 0 49 coffee, tea, and chocolate (kg) 0.1 0 50 bottled/mineral water (liter) 0.1 0.1 51 soft drinks, orange juices, fruit juices (liter) 0.1 0.1 52 ice cream (roll) 0.1 0 53 beer at home (liter) 0.1 0.35 Yes 54 wine at home (liter) 0.1 0.35 Yes 55 other alcohol not in bar or restaurant (liter) 0.1 0.35 Yes 56 cigarettes and other tobacco (roll) 0.1 0.2 Yes 57 food at school (Riels) 0.1 0 58 drinks at school (Riels) 0.1 0 59 food at work (Riels) 0.1 0 60 drinks at work (Riels) 0.1 0 61 food/snacks at restaurant, pub or café (Riels) 0.1 0 62 drinks at restaurant, pub or café (Riels) 0.1 0 44 63 prepared meals bought outside and eaten at home (Riels) 0.1 0 64 other food expenses (Riels) 0 0 101 Clothing 0.1 0 102 shoes, slippers 0.1 0 103 household textiles (cotton thread, cotton scarf, belt) 0.1 0 104 raincoat, umbrella 0.1 0 105 toothpaste, toothbrush, and tooth care 0.1 0 106 hair soap, cloth soap, lotion, powder, perfume 0.1 0 107 jewelry, watch, and clock 0.1 0 108 gasoline, diesel, and lubricant 0.1 0 109 local travel (Last 3 months) 0 0 110 hotel, guesthouse, and other accommodation (Last 3 months) 0.1 0 Yes 111 foreign travel 0.1 0.1 112 postal services/ package 0.1 0.03 113 car and travel insurance 0 0 114 costs for motorbikes (other than gasoline and purchase) 0.1 0.15 115 costs for cars (other than gasoline and car purchase) 0.1 0.15 116 telephone service (exclude telephone accessories) 0.1 0.03 117 internet service) 0.1 0.03 118 games of chance (lottery, football betting) 0.1 0.1 119 other recreation (movie, karaoke) 0.1 0 120 newspapers, magazine 0.1 0 121 books, papers and other stationaries 0.1 0 122 salary/wage for housekeeper 0 0 123 expense for children look after 0.1 0 124 spoon, fork, knife, broom, chopsticks 0.1 0 125 gardens, plants and flowers (not for agriculture) 0.1 0 126 pets and related costs 0.1 0 127 toys, games and hobbies 0.1 0 128 dwelling insurance and maintenance (excl improvements) 0 0 129 drugs bought with prescription or over the counter 0 0 130 medical products and assistive products 0 0 131 medical or dental consultation without overnight stay 0 0 132 medical or dental treatment with overnight stay 0 0 133 traditional medicine 0 0 134 health insurance 0 0 135 taxes on income (tax on salary) 0 0 136 taxes on property (e.g., houses, cars) 0 0 137 bank payback, other financial service or tong tin 0 0 138 wedding gift 0.1 0 139 other gift (funeral, bonkathen, bonpka) and other contribution to other household 0.1 0 140 other expenditure (specified) 0.1 0 201 water charges last month 0 0 202 sewage or waste disposal last month 0 0 203 garbage collection last month 0 0 204 electricity last month 0 0 205 gas - LPG last month 0.1 0 206 kerosene last month 0.1 0.1 207 firewood last month 0 0 208 charcoal last month 0.1 0 209 battery last month 0.1 0 210 other fuel last month 0 0 211 how much paid for rent last month; 0 0 Source: MEF. 45 Electricity subsidies Electricity is not directly subsidized in Cambodia, but there exist special lower tariffs for poor households identified by their low levels of electricity consumption. Due to the licensing system operating in Cambodia and due to electrical network still being developed, there is no uniform tariff for the whole country. Tariffs vary locally depending on the energy source used and are being approved by a central agency, Energy Authority of Cambodia (EAC). For Phnom Penh, Kandal Province, and Provincial Town of Kampong Speu, there is a volume differentiated tariff system, where special reduced tariffs are set for poorest households consuming less energy. There are also special tariffs applicable to households in rural areas. These tariffs are subsidized by the government. Table A.3: Electricity energy prices depending on consumption and geography kWh/month Phnom Penh and Provincial towns Other areas Kandal provinces <= 10 480 380 380 > 10 & <= 15 480 480 480 > 15 & <= 50 480 480 480 > 50 & <= 200 610 720 720 > 200 740 740 740 Source: The Council for the Development of Cambodia (CDC). The 2019/20 CSES captures information on household spending on electricity. Respondents also provide information on whether electricity is provided from the public network. Subsidies are estimated only for those households drawing power from the public grid. For each area we assign a standard, non-subsidized price per kWh of consumed electricity. From expenditures on electricity, we calculate consumption in kWh, depending on geography (special provinces/other urban/other rural) and actual consumption. We calculate the subsidy as the difference in standard price less subsidized price times consumption in kWh. Modeling social transfers In recent years, Cambodia has increased social spending. However, social transfers remain very limited. In 2019, the government implemented a broad program for supporting pregnant women and infants in the form of a Conditional Cash Transfer (MOE and MOH, 2019). This new Conditional Cash Transfer (CCT) program is included in the CEQ analysis based on the eligibility criteria of recipients and total expenditure on the program (which was 11,737.64 million Riels or 2.87 million US$ in 2019).24 In 2019, most assistance focused on specific targeted programs to groups such as AIDS patients, victims of natural disasters and people requiring emergency food assistance. Most support programs consisted of in-kind support and were implemented on small scale (often with the help of development funds). In the 2019/20 CSES data we can only identify one specific type of transfer support: government scholarships. All other benefits have been classified in the data as “general welfare transfers”. 25 On top of 24 Official exchange rate for 2019 was 4092.78 per US$. 25 Other forms of transfers to civil servants operated in Cambodia in 2017 including family allowances. In the latter case they were paid jointly with civil servant salaries and are thus included jointly with their labor income. Civil 46 those, since in 2019 the Conditional Cash Transfer was an element of the Cambodian fiscal system, we model the CCT as part of the baseline system by granting US$ 63 per year for each child under the age of three in families that are holders of the so-called IDPoor equity cards, an important element of the Cambodian social support system, which identifies poor households. This imputation is based on the general rules of the CCT program which grants US$ 190 over the course of three years to pregnant mothers and children under the age of 2. Since we do not have information in the data on pregnancy, the simulation targets the benefits towards households with children aged 0-2. In-kind education benefits Presently, formal education in Cambodia comprises of a 12-year system divided into a 6-3-3 format, which includes six years of primary schooling, followed by three years lower-secondary and three years of upper- secondary schooling. Public education in Cambodia is officially free of additional charges, but it is not uncommon for schools and teachers to request additional fees, often officially to finance additional costs of educational materials or to contribute towards after-school extra classes. Public tertiary education is not free of charge and students are charged fees for their courses (e.g., Songkaeo and Yeong, 2016). In parallel, there are private education institutions providing instruction at all levels of education including universities. The 2019/20 CSES provides information on whether a child is attending a public or a private school. To distribute the benefits of the public education system, we use the age and school attendance of the child available in the CSES data and average schooling costs for each level of education. Average schooling costs are derived from the total government expenditures on education at each level (i.e., pre-primary, primary, secondary, tertiary) divided by the number of students attending this level of public education in 2019 (cost values are based on the data from the Ministry of Education, Youth and Sport: Education strategic plan: 2019-2023, number of students taken from Education Statistics and Indicators 2018-2019). Our calculations cover also pre-primary and tertiary education expenditures (Table A.4). Table A.4: Education expenditures in 2019 Number of students MoEYS exp. Expenditure per (billion KHR/year) student (thousand KHR/month) Pre-primary 217,509 246.8 94.5 Primary 2,040,257 1969.5 80.4 Secondary 931,406 1093.6 97.8 Tertiary 226,731 367.5 135.1 Total 3,415,903 3677.4 89.7 Source: Ministry of Education, Youth and Sport (2014, 2018). servants and their families are also eligible for support through sickness benefits, employment injury benefits, maternity benefits, invalidity benefits, and death benefits. These benefits are financed from the central budget by the National Social Security Fund for Civil Servants (NSSFC) and are not based on previous contributions. War veterans, army and police personnel are entitled to similar set of benefits paid by the National Fund for Veterans (NFV). 47 To simulate in-kind education transfers, the theoretical monthly per capita expenditure for each level of education is assigned to all children in the data attending public education. Out-of-Pocket (OOP) education expenditures are considered. There is a strong rationale to include OOP expenditures because households are contributing to the fiscal system due government under-funding of education costs. Sensitivity analyses suggest that the addition or exclusion of OOP expenditures does not have significant implication on findings. In-kind health benefits The health care system in Cambodia combines both public and private provision. While financing of public health care consists of public funds, patient contributions take place in the form of user fees charged by public hospitals and health centers. Individuals who have been assigned the Health Equity Fund card and those covered by Health Insurance may be exempt from paying the user fees. The Ministry of Health provides a detailed breakdown of its expenses, which allows to compute the average costs of both inpatient and outpatient care. Table A.5: Ministry of Health expenses in 2018 by type Cases Total Cost Cost per case (US$) (US$) I. Total medical services 2,810,697 12,452,344 4.43 Outpatient Services 2,623,842 5,480,861 2.09 General treatment 127,800 3,169,514 24.80 Surgical Treatment 12,700 2,793,963 220.00 Regular births 30,489 620,511 20.35 Rescue vehicles 15,866 387,496 24.42 II. Non-medical services 286,526 1,929,084 6.73 Travel expenses 121,987 836,623 6.86 Food 163,415 1,075,611 6.58 Funeral 1,124 16,850 14.99 Total 6,194,446 28,762,857 4.64 Source: Ministry of Health (2018). The 2019/20 CSES provides detailed information on health care usage during the 30 days prior to the interview. The survey collects information on frequency of health care visits, type of facility visited (public/private), length of hospitalization, and total out-of-pocket expenses incurred by each person in the household. Each recorded visit to a health care center is treated as an outpatient case and hospitalizations are treated as inpatient interventions. When aggregated to population totals, the 2019/20 CSES reports overall 7,485,161 visits and 6,521,216 days spent in public hospitals for 623,854 patients during 2019. 48 Given the level of health utilization details in the CSES data and information on costs of health care interventions in Cambodia provided in Flessa et al. (2018) and Jacobs et al. (2019), it is possible to apply a more precise method of imputation of health care expenditure based on individual use of health care services rather than impute average gains to all households. 26 Given the available data and information, it is not possible to assign a cost to each specific medical procedure/service, but the data and information on health care costs allow to distinguish cost of a medical consultation Out-Patient Day (OPD) and hospitalization In-Patient Day (IPD). These can be calculated separately for Health Centers, and CPA3, CPA2 and CPA1 facilities. Flessa et al. (2018) also provide details of the percentage of publicly subsidized cost by health facility type. Based on information from the CSES data, one can easily build variables for OPD visits during a month in public health care facilities by type of facility, and number of IPD nights in facilities by type of facilities. Since there is no detailed information on all visits regarding whether this was a public or private center, assumptions are made on the place used based on the first and the last visit. When both the first and the last visits took place in a public facility, it is assumed that all health visits reported were in a public facility. When either the first or the last visit was in a private facility, one visit is subtracted, and all remaining visits are treated as public. If both the first and the last visit reported were in a private facility, all the visits are treated as private. In case of IPD, the hospitalized nights are considered to have been in a public facility if any of the OPD visits was in a public facility. Based on these numbers and the detailed costs per day, the total health care transfers per person is computed and divided into a public sector cost and a user cost. In the case of people who report having a Health Equity Card and for those whom we model as contributing to health insurance we set the subsidy to the full cost of the reported treatment. For people who are not covered by health insurance we subtract the non-subsidized part from their benefit. In this way we incorporate the out-of-pocket user fees that need to be paid by patients over and above the level of the public subsidy. 26 In the calculations we were able to use the latest, updated version of cost estimations from Flessa et al. (2018) available directly from the authors. We are very grateful to Flessa for making this data available. 49