The Effects of Taxes and Transfers on Inequality and Poverty in Pakistan Beenish Amjad Haydeeliz Carassco Moritz Meyer The Effects of Taxes and Transfers on Inequality and Poverty in Pakistan1 Beenish Amjad Haydeeliz Carrasco Moritz Meyer The World Bank Group The World Bank Group The World Bank Group This version: January 19, 2025 Abstract. This study assesses the impact of fiscal policy in Pakistan - taxes, social expenditures, and subsidy expenditures - on poverty and inequality using the Commitment to Equity (CEQ) Methodology. Results show that fiscal policy increased the national poverty headcount and left inequality largely unchanged during fiscal year 2019 (spanning July 1, 2018, to June 30, 2019). The net effect of the fiscal system is to increase the poverty headcount ratio at the national poverty line by approximately two percentage points. Overall, the combination of direct and indirect taxes paid, cash or near-cash transfers received, and subsidy benefits captured leaves most households as net payers into the fiscal system; only the poorest decile are net recipients in cash terms from the fiscal system. Taxes, transfers, and subsidies have a muted impact on inequality: the Gini coefficient index falls from 29.0 to 28.6 from pre-fiscal to post-fiscal income. The most effective expenditure-side instrument for either poverty or inequality reduction is the BISP social protection program; the revenue-side instruments that shield poor and vulnerable households most effectively are the personal income tax and the urban property tax. The revenue-side instrument most effective at reducing inequality is the personal income tax. Indirect taxes and subsidy expenditures – which are far larger in fiscal magnitude than direct taxes or social transfers (respectively) in Pakistan in fiscal year 2019 – are not particularly effective at reducing poverty or inequality. In-kind transfers of public health and education services also have an insignificant impact on inequality. Keywords: fiscal policy, social expenditure, taxes, fiscal incidence, inequality, poverty, Pakistan JEL classification: H22, I38, D31 1 Beenish Amjad and Haydeeliz Carrasco are Economics Consultants in the Poverty and Equity Global Practice at the World Bank; and Moritz Meyer (corresponding author: mmeyer3@worldbank.org) is Senior Economist in the Poverty and Equity Global Practice at the World Bank. We declare that we have no relevant or material financial interests that relate to the research described in this paper. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of the World Bank Group or any affiliated organizations, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. Acknowledgements Juan Pablo Baquero and Aroub Farooq (Economist, ESAC1) provided research assistance. Jon Jellema (Senior Economist, ESAPV) and Gabriela Inchauste (Lead Economist, EWAPV) provided analytical direction, and Oscar Barriga-Cabanillas (Economist, ESAPV) and Erwin Knippenberg (Economist, EMNPV) shared feedback. The team is grateful for comments received from peer reviewers Arden Finn (Senior Economist, EMNPV), Sailesh Tiwari (Lead Economist, EEAPV), Derek Hung Chiat Chen (Senior Economist, ESAC1); Florian Michael Blum (Senior Economist, EECM2 ); Phillippe George Pereira Guimaraes Leite (Senior Social Protection Economist, HSASP); Saadia Qayyum (Energy Specialist, ISAE1); Nasir Iqbal (Researcher, Pakistan Institute of Development Economics); and Rida Najam (Economist, Planning Commission). The authors would like to thank the Pakistan Government for granting access to the data necessary to compile this report and World Bank colleagues (World Bank Poverty and Equity Global Practice, World Bank Equity Policy Lab, World Bank Development Economics Group, World Bank Pakistan Country Team) and participants from the Pakistan Planning Commission in the Ministry of Planning, Development, and Special Initiatives as well as the Pakistan Institute of Development Economics (PIDE) for critical and insightful feedback provided at all stages in the preparation of this study. Contents Acknowledgements ................................................................................................................................................. 3 Abbreviations .......................................................................................................................................................... 6 Executive Summary ................................................................................................................................................. 1 1. Introduction .................................................................................................................................................. 6 2. Literature Review .......................................................................................................................................... 9 3. Tax Revenues and Public Expenditure ........................................................................................................ 11 3.1 Tax Revenues ............................................................................................................................................. 11 3.2 Social Expenditure ...................................................................................................................................... 15 3.3 Non-Social Expenditure .............................................................................................................................. 20 4. Methodology and Data ..................................................................................................................................... 24 4.1 Methodology: Commitment to Equity ....................................................................................................... 24 4.2 Data Sources for Pakistan .......................................................................................................................... 26 5. Application of the Fiscal Incidence Analysis ...................................................................................................... 28 5.1 Coverage of the Fiscal System.................................................................................................................... 28 5.2 Assumptions behind the Analysis .............................................................................................................. 29 5.3 Comparison between Survey and Administrative Data ............................................................................. 37 6. Main Findings on Fiscal Equity .................................................................................................................... 42 6.1 Relative Incidence ...................................................................................................................................... 43 6.2 Absolute Incidence ..................................................................................................................................... 45 6.3 Progressivity of Fiscal Interventions .......................................................................................................... 46 6.4 Marginal Contributions .............................................................................................................................. 49 6.5 Net Payers and Net Receivers .................................................................................................................... 53 6.6 Overall Impacts on Inequality and Poverty ................................................................................................ 54 6.7 Impact Effectiveness and Spending Effectiveness ..................................................................................... 58 7. International Comparison ........................................................................................................................... 61 7.1 Pakistan and Peer Countries ...................................................................................................................... 61 7.1 Note on International Comparison ............................................................................................................ 62 8. Methodological Innovations ....................................................................................................................... 64 8.1 Incidence of Indirect Taxes in the Context of Consumption Informality ................................................... 64 9. Conclusions ................................................................................................................................................. 67 References............................................................................................................................................................. 69 Annex I. Fiscal Parameters of the Model .............................................................................................................. 73 Annex II. Methodology for the Estimation of Public Sector Employees and Labor Informality in the HIES ......... 84 Annex III. Methodology for Estimating Indirect Effects Based on the IO Matrix .................................................. 88 Annex IV. Relative and Absolute Incidence, Detailed ........................................................................................... 90 Annex V. Fiscal Interventions Excluded from the Model ...................................................................................... 96 Annex VI. Data Limitations and Recommendations .............................................................................................. 97 Abbreviations AIT Agriculture Income Tax ARV Annual Rental Value BISP Benazir Income Support Program CCT Conditional Cash Transfer CEQ Commitment to Equity CIT Corporate Income Tax CNG Compressed Natural Gas COVID-19 Coronavirus Disease 2019 DAP Diammonium Phosphate EU European Union EUROMOD Tax-benefit Microsimulation Model for the European Union FBR Federal Board of Revenue FIA Fiscal Incidence Analysis GDP Gross Domestic Product GIDC Gas Input Development Cess GST General Sales Tax; also known as and referred to as “Goods and Services Tax” HIES Household Integrated Economic Survey IBT Increasing Block Tariff ICT Islamabad Capital Territory IFA Individual Financial Assistance IFPRI International Food Policy Research Institute IMF International Monetary Fund IO Input-Output Matrix KI Kakwani Index KP Khyber Pakhtunkhwa KPP Khushal Pakistan Program LFS Labour Force Survey MC Marginal Contribution MMBTU Million British Thermal Units MoF Ministry of Finance MoFEPT Ministry of Federal Education and Professional Training MoNFSR Ministry of National Food Security and Research MoNHSRC Ministry of National Health Services Regulations and Coordination MoWP Ministry of Water and Power NA National Accounts NEPRA National Electric Power Regulatory Authority NFC National Finance Commission NFDC National Fertilizer Development Centre NSER National Socio-Economic Registry OGRA Oil and Gas Regulatory Agency PBM Pakistan Bait-ul-Mal PBS Pakistan Bureau of Statistics PDC Price Differential Claim PEPCO Pakistan Electric Power Company (Private) Limited PIDE Pakistan Institute of Development Economics PIT Personal Income Tax PITC Power Information Technology Company PKR Pakistani Rupees PMT Proxy Means Testing PPAF Pakistan Poverty Alleviation Fund PSDP Public Sector Development Programme PSLM Pakistan Social and Living Standards Measurement PSO Pakistan State Oil RSPN Rural Support Programmes Network SBP State Bank of Pakistan SDG Sustainable Development Goal SNGPL Sui Northern Gas Pipeline Limited SSGPL Sui Southern Gas Pipeline Limited TCSP Targeted Commodity Subsidy Program TDS Tariff Differential Subsidy UCT Unconditional Cash Transfer WAPDA Water and Power Development Authority WB World Bank WEO World Economic Outlook Executive Summary In 2013, the World Bank established the twin goals of ending extreme poverty by 2030 and boosting shared prosperity and adopted a unified strategy to focus policies and operations on achieving this core mission while supporting the Sustainable Development Goals (SDGs). Central to delivering on this agenda is helping policymakers maximize the development impact of public policy when fiscal space is limited. This became even more important during the COVID-19 pandemic, given that the poor and vulnerable were most affected by the crisis. Fiscal incidence analysis (FIA) measures the poverty, inequality, and social welfare impact of the taxes and transfers that make up fiscal policy. FIA findings can be used to inform pro-poor policy design and to make progress on fiscal equity. Specifically, FIA can identify pro-poor fiscal interventions; monitor progress toward SDG 10.4.2 which measures the redistributive impact of fiscal systems on inequality reduction; or assess the distributional impacts of existing fiscal policies as well as potential policy reforms. Over the past two decades, Pakistan has struggled with low and increasingly volatile economic growth. Although Pakistan lifted about 46 million people out of poverty between 2001 and 2018 and halved the headcount poverty ratio in that time span, rural poverty remains twice as high as poverty in urban areas. Nationwide, the top 10 percent of the population consume on average three times more than the bottom 10 percent, and their incomes are five times larger, with no substantial change over time. These inequalities in income and consumption translate into unequal access to human capital development services and unequal access to economic opportunities, which further limits intergenerational mobility. As Pakistan has recently faced substantial fiscal challenges, natural disasters, and negative economic shocks (such as the COVID-19 pandemic), disparities in income, consumption, and access gaps to services are likely to have widened further for the poor and vulnerable. The FIA for Pakistan for fiscal year 2019 summarized in this document is the first study that models the joint distributional impacts of the country’s main taxes, social expenditures, and subsidy expenditures on households’ welfare. It is based on the Commitment to Equity (CEQ) Methodology and uses the Pakistan Household Integrated Economic Survey (HIES) 2018–19, along with fiscal, budgetary, and administrative data from both federal and provincial governments from the same period. The study presents answers to the following questions: (i) What is the impact of taxes and social expenditure on poverty and inequality? (ii) Which taxes and social expenditure items are progressive or regressive? (iii) Who bears the burden of taxes and receives the benefits of social expenditures? and (iv) Which households are net payers or net receivers of the fiscal system? The study covers 61 percent of federal tax revenues and 45 percent of provincial tax revenues in the fiscal year 2018–19. The lower coverage of provincial tax revenue is due to data limitations in modeling some provincial taxes, such as the agriculture income tax (AIT) and the motor vehicle tax. Taxes included in the model are the general sales tax (GST), customs duties, the federal excise duty, the withholding tax from salaries, the withholding tax on telecommunications, and the property tax. This study covers 8 percent of federal expenditures2 and 23 percent of provincial expenditures in the same period. Government social expenditures covered in the analysis are the main social protection programs (conditional and unconditional cash transfers under the Benazir Income Support Program, or BISP), the indirect subsidies (agriculture tubewells, urea fertilizer, 2 Public pensions expenditures are not counted as a direct transfer in this analysis. There are no specific public pension contributions collected, withheld, or remitted. We treat the public pension system in Pakistan under the CEQ Assessment “Pensions as deferred Income” (PDI) scenario; see section 5.2 for more detail. domestic electricity, and natural gas), health in-kind benefits (inpatient and outpatient services), and education in-kind benefits (pre-primary and primary, secondary, and tertiary). This analysis includes several innovations, such as the inclusion of the zakat paid and received (religious taxes and transfer, respectively) through the federal government’s system and the modeling of informality in household consumption (based on Bachas et al. 2020). The final fiscal incidence model allocates to individuals and households in the HIES more direct and in-kind transfer expenditure relative to direct and indirect taxes than appear in the budget and administrative data (for the same fiscal policy instruments). For example, the ratio between total taxes and total public transfers3 allocated in the HIES is 86 percent, which is only half as large as the analogous ratio using administrative data (163 percent). The allocation of direct and indirect taxes is tied to the taxable incomes and taxable purchases (respectively) that appear in the HIES; we are unable to allocate tax burdens to missing taxable incomes or taxable consumption expenditures.4 Meanwhile, the estimation of social protection benefits and in-kind transfers received depends instead on socioeconomic and demographic characteristics – including eligibility based on poverty status of certain direct transfers – which are more frequently and widely-represented in the HIES. Key findings from the Pakistan fiscal incidence analysis are summarized here: First, the combination of taxes, social expenditures, and subsidies modeled increases the national poverty headcount while leaving inequality largely unchanged in fiscal year 2019 (which spans July 1, 2018, to June 30, 2019). The Gini coefficient index of inequality is reduced slightly from 29.0 at prefiscal income to 28.6 at final income while the national poverty headcount increases from 23.3 percent at prefiscal income to 25.5 percent at consumable income.5 Second, most poor and vulnerable households are net payers into the fiscal system meaning that benefits received are smaller in magnitude than taxes paid. Only individuals from the poorest ten percent of the population can expect to be net recipients – or can expect to receive more in benefits than they will pay in taxes – with a net cash gain estimated 1.2 percent of prefiscal income. All other individuals can expect to be net payers in cash terms, with cash losses ranging from -1.8 percent in decile 2 to -5.5 percent in decile 10 (the richest decile). That deciles 2 and 3 are net payers of the fiscal system is consistent with the increase in the poverty headcount ratio due to fiscal policy while the fact that the cash losses are greater in the richer deciles is consistent with the slight reduction in inequality due to fiscal policy (Figure ES1). 3 Total taxes include direct and indirect taxes. Total transfers refer to direct transfers, indirect subsidies and in-kind benefits from health and education. 4 Note that this applies to subsidy benefits captured also: when the consumption expenditure that potentially carries a subsidy benefit is missing, that subsidy benefit will go unallocated. HIES individuals and households account for approximately 45 percent of total national consumption expenditure according to Pakistani national accounts. This underestimate could be due to misreporting of survey respondents; unit non-response in top-income households; the HIES sampling frame itself; or other reasons. 5 CEQ Consumable Income includes the main taxes, direct transfers, and indirect subsidies; CEQ Final Income is defined as CEQ Consumable Income plus in-kind benefits from health and education. Figure ES1. All deciles are net payers, except for the poorest decile (net receiver) 20.0% 15.0% Tax or transfer as a % of market income plus 10.0% 5.0% 0.0% pensions Poorest 2 3 4 5 6 7 8 9 Richest -5.0% -10.0% Deciles by market income plus pensions, real, per adult equivalent Total in-kind education benefits Total in-kind health benefits Total indirect subsidies Total indirect taxes Total direct transfers Total direct taxes NET CASH POSITION (market income-taxes+ cashable transfers) TOTAL POSITION (Net cash position+ in-kind benefits) Source: World Bank calculations based on HIES 2018–19 and fiscal administrative data. Third, richer households capture a larger share of the subsidy and in-kind benefits available and pay more of total revenues collected from direct and indirect taxes. Estimating each decile’s share of the total subsidy benefits available or the total revenue from indirect taxes provides a summary of how concentrated benefits or taxes are in one group. In Pakistan in fiscal year 2019, the two richest deciles capture 34 percent of total subsidy expenditure, 29 percent of in-kind education benefits, and 27 percent of in-kind health benefits, and pay 40 percent of total revenues from indirect taxes. The subsidy and in-kind benefit concentration shares for the two richest deciles are driven by the electricity subsidy, tertiary education, and inpatient health spending.6 The two richest deciles also pay just less than 90 percent of total direct taxes modeled while the two poorest deciles receive approximately 53 percent of the total direct transfer spending.7 Fourth, the General Sales Tax (GST) has the largest negative impact on the poverty headcount; the Benazir Income Support Program (BISP) has the largest positive impact on inequality reduction. Estimations of the marginal contributions of individual fiscal instruments – or the additional impact that individual fiscal instruments have on poverty or inequality when all other fiscal instruments are included – demonstrate that GST has the largest marginal contribution to the national poverty increase. GST is allocated roughly in proportion to consumption expenditure and therefore is neither strongly regressive or strongly progressive; however, GST payments account for over 7 percent of households’ pretax expenditure which leads to further impoverishment among poor and vulnerable households. The BISP cash transfer meanwhile demonstrates the largest marginal contribution to inequality reduction. The second-largest positive impact on inequality comes from pre-primary and primary education expenditures. 6 97 percent of the two richest deciles have access to electricity compared to 70 percent among the two poorest deciles; 32 percent of total household electricity consumption is accounted for by the two richest deciles compared to 10 percent among the two poorest deciles. While 7 Direct transfers include the BISP unconditional cash transfer (UCT) and conditional cash transfer (CCT) programs. The BISP transfers are targeted via a Proxy Means Testing [PMT] method. Fifth, BISP transfers and direct taxes are the most cost-effective fiscal policy tools for reducing inequality while protecting poor and vulnerable households; in contrast, indirect subsidies are relatively ineffective. Effectiveness indicators demonstrate that BISP transfers are the most cost-effective expenditure policy for reducing both inequality and poverty depth. Effectiveness indicators also show that direct taxes are the most cost-effective revenue policies for inequality reduction, while protecting poor and vulnerable households from further impoverishment. Meanwhile, indirect subsidies (in natural gas, electricity, and agricultural inputs) are relatively ineffective instruments for reducing poverty. The overwhelming majority of the subsidy benefits available are captured by non-poor, non-vulnerable households, so the majority of subsidy expenditures do not significantly impact poor and vulnerable households. Sixth, among comparator middle-income countries, Pakistan has the highest poverty increase and the lowest inequality reduction from the sample (Figures ES2 and ES3). This result can be explained by the fiscal facts (circa fiscal year 2019) in Pakistan, namely: (1) tax collections overall were relatively low at 13.4 percent of GDP and (2) were generated in large part from impoverishing indirect taxes while (3) the vast majority of domestic revenues and other sources of financing were dedicated to rigid expenditures like debt service, the government’s wage bill (including pension payments), and (4) highly inefficient and regressive general subsidies on energy and agricultural inputs which (5) left no fiscal space for either targeted social protection transfers or progressive spending on essential education and health services which more frequently benefit the larger families in the bottom half of the population. In other words, fiscal policy in Pakistan emphasized revenue collection from more frequently-impoverishing indirect taxes as well as regressive and inefficient subsidy expenditures to provide benefits while de-prioritizing progressive direct taxation (which can shield poor and vulnerable households from further impoverishment), direct, targeted transfers to poor and vulnerable households, and the investments in social infrastructure which help poor and middle-income households participate productively in social, civic, and economic life. Moving forward, Pakistan should improve its domestic revenue mobilization and public expenditure efficiency to generate greater fiscal space. The additional fiscal space generated should be prioritized to expand social expenditure and targeted transfers and to improve fiscal equity. Fiscal sustainability could go together with fiscal equity if additional revenues collected from GST harmonization, for example, is used to compensate poor and vulnerable households through well-targeted cash transfers. Similarly, public expenditure efficiency reforms (for instance, moving from generalized ineffective subsidies in the energy sector to targeted social transfers) could improve fiscal sustainability along with fiscal equity. Lastly, expenditure reforms to improve the accessibility and quality of public health and education services in Pakistan could have long-term impacts in terms of poverty and inequality reduction. These fiscal reform areas align with the policy recommendations from the World Bank’s Poverty and Shared Prosperity Report (2022).8 8 World Bank (2022) highlights three sets of fiscal policy priorities to foster a robust and inclusive recovery post-COVID-19: (i) reorienting public expenditure away from subsidies to provide income support to households via cash transfers; (ii) prioritizing public expenditure with long-term impacts; and (iii) when needed, raising tax revenue without making the poor worse off (ages 18–22). Figure ES2. Pakistan has the lowest inequality Figure ES3. At the same time, Pakistan has the reduction (based on the Gini coefficient) in the highest poverty increase (based on the US$3.20 sample of comparator countries international poverty line PPP 2011) in the sample of countries 0.60 45.0 ($3.2 intl poverty line) poverty headcount 0.55 40.0 35.0 Gini coefficient (0-100) 0.50 30.0 0.45 25.0 0.40 20.0 0.35 15.0 0.30 10.0 0.25 5.0 0.0 Market Disposable Consumable Final income income plus income income Market Disposable Consumable pensions income plus income income pensions Brazil (2009) Egypt (2015) Mexico (2014) Sri Lanka (2009) Indonesia (2012) Jordan (2010) Turkey (2014) Pakistan (2018) Mexico (2014) Sri Lanka (2009) Brazil (2009) Turkey (2014) Pakistan (2018) Source: World Bank’s elaboration, based on staff calculations for Pakistan 2018 and secondary studies for Brazil (Higgins and Pereira 2013); Egypt (Lara Ibarra et al. 2019), Indonesia (Jellema et al. 2017), Jordan (Alam et al. 2017); Mexico (Scott et al. 2020); Sri Lanka (Arunatilake et al. 2019); and Türkiye (Cuevas, P. Facundo et al. 2020). The Pakistan CEQ Assessment can serve as a baseline for evaluation of potential policy reforms for fiscal equity and sustainability. The FIA model summarized in this report was expanded to include ex-ante distributional analysis of innovative fiscal policy applications and potential fiscal policy reforms under discussion in Pakistan. Examples of the former include the COVID-19 emergency transfers implemented between 2020 and 2021 and the BISP Targeted Commodity Subsidy Program, which was launched in 2022 to compensate households for high food inflation. Examples of the latter include ongoing tariff and price-setting reforms in the electricity and natural gas utilities. 1. Introduction Over the past two decades, Pakistan has experienced high poverty reduction despite slow and volatile economic growth. On average, Pakistan’s real per capita GDP growth rate between 2000 and 2018 was 2.1 percent, which was below the South Asia regional average. Over the same period, poverty measured using the national poverty line fell from 64.3 to 21.9 percent. This decline in poverty was largely driven by an increase in labor incomes outside the agriculture sector and a rise in remittances. Pakistan also managed to make progress toward eradicating extreme poverty, measured at US$1.90 2011 PPP line (World Bank 2020a), and closed the gap with regional best performers over the first two decades of the 20th century, Bhutan, and Sri Lanka. However, there remain challenges in terms of poverty and inequality reduction. Over time, poverty has increasingly become concentrated among rural households relying on agriculture and among individuals with levels of low human capital endowments. The economic opportunities created by economic growth in the past two decades have not been equally distributed due to low socio-economic inclusion. Consumption inequality as measured by the Gini coefficient increased slightly from 27.5 in 2001 to 28.4 in 2018. The human development landscape of Pakistan’s four provinces is glaringly unequal. Across districts, unequal access to education and health services reinforces the transmission of monetary deprivations, particularly in rural areas. These socio- economic deficits in the population translate into low levels of productivity and limited resilience, adding to the economic and social exclusions (World Bank 2020a). Pakistan also faces substantial fiscal challenges including a high public deficit and increasing levels of public debt. Moreover, low tax revenues are mostly absorbed by rigid expenditures, including salaries, pensions, interest payments, defense, and subsidies (World Bank 2020a), which limit fiscal space to finance social investments. The macroeconomic crisis and the COVID-19 pandemic have put more pressure on public finances, as tax revenue is more limited while social expenditure is increasing. At the same time, the COVID-19 pandemic and the associated economic slowdown have brought negative welfare shocks to the population, slowing the pace of poverty and inequality reduction. The Government of Pakistan is relying on structural economic reforms to enhance fiscal space along with social protection interventions to tackle poverty and inequality. In 2018, the Government’s First 100 Days agenda focused on six themes, including: (i) transforming governance; (ii) strengthening the Federation; (iii) revitalizing economic growth; (iv) uplifting agriculture and conserving water; (v) revolutionizing social services; and (vi) ensuring Pakistan’s national security. Under Theme III, the Government has tried to improve fiscal space by introducing tax reforms for enhanced revenue mobilization and reducing public expenditures.9 Moreover, under Theme V, the focus has been on expanding social safety nets. In this regard, the Government has strengthened the country’s largest social protection program, known as Benazir Income Support Program (BISP), which was originally launched in 2008. The BISP has multiple programs and initiatives aiming at eradicating poverty, promoting financial inclusion, supporting the economic empowerment of women, focusing on human capital formation, and overcoming financial barriers to accessing health and post-secondary education. In addition, the 9 Public finance reforms identified between the Pakistan’s Government and the IMF were: (i) increasing revenues by hike in petroleum levy; reinforcing tax administration efforts; tax reforms particularly in corporate income tax (CIT) approved in March 2021 which simplifies CIT by streamlining numerous exemptions; (ii) improving expenditure efficiency by streamlining subsidies; recalibration of development expenditure; reforms in power sector; (iii) reducing the pension and wage bills; and (iv) improving large public sector enterprises transparency and monitoring to reduce public expenditure. On the other hand, some reforms in tax were introduced to increase the tax net and revenue including GST reforms and Personal Income Tax reforms. Government has launched two major new programs since 2020: the COVID-19 emergency program to provide income support to poor and vulnerable households affected by the pandemic; and the BISP Targeted Commodity Subsidy Program (TCSP), which provides income support for the purchase of basic food items to poor and vulnerable households in the context of food inflation. As the Government tries to balance improving fiscal space along with poverty and inequality reduction, the fiscal equity consideration matters. In effect, fiscal policies have multiple objectives. First, the Government needs to collect enough taxes to ensure the funding of public expenditure and maintain debt sustainability. Second, the Government needs to collect taxes and provide transfers in an efficient manner that minimizes distortions in the economy. Third, the Government could design taxes and transfers with the aim of tackling positive or negative externalities. Lastly, the Government needs to ensure that fiscal policies are consistent with poverty and inequality reduction. This study is interested in this latter aspect—the equity aspect of fiscal policies—while it is acknowledged that there are other objectives of the fiscal system. Moreover, the equity aspect of fiscal policies is so important that it has been included as the Sustainable Development Goal 10.4.2, which measures the redistributive impact of the fiscal system on inequality reduction. Fiscal equity matters because economic growth alone will not be enough to eliminate poverty and inequality in Pakistan. The fiscal system can support this process, through the right mix of investments and redistribution via taxation and social expenditures. Furthermore, pro-poor fiscal policies could lay the foundation for future economic growth and human development. As the country explores the mix of taxation and public expenditure policies to ensure fiscal sustainability, it is important that the design and implementation of reforms maintain a focus on equity. This study provides the first comprehensive distributional analysis of the fiscal system in Pakistan. This fiscal incidence analysis describes the distributional impact of the main tax interventions and public social expenditure items on poverty and inequality for the fiscal year 2018–19. While there are previous fiscal incidence studies of individual fiscal interventions in the country, to the best of our knowledge this is the first fiscal incidence analysis that aims at covering the joint impact of the main interventions of the fiscal system. The analysis is based on the Commitment to Equity (CEQ) Methodology and uses data from the Household Integrated Economic Survey (HIES) 2018–19, which is a nationally representative household survey. Under this approach, taxes and transfers are either directly observed or simulated from the survey data. The analysis assesses the incidence of the main fiscal interventions in Pakistan, including direct and indirect taxes, direct transfers, indirect subsidies, and in- kind benefits from public health and education services. The study aims to respond to the following questions: (i) What is the impact of Pakistan’s fiscal system on inequality and poverty? (ii) Which taxes and social expenditure items are progressive or regressive? (iii) Who bears the burden of taxes and receives the benefits of social expenditures? and (iv) Which households are net-payers or net-receivers of the fiscal system? Moreover, the fiscal incidence model establishes a platform for undertaking ex-ante distributional analysis of potential policy reforms. The main findings from the fiscal incidence analysis show that the combination of taxes and social expenditure items modeled in Pakistan in fiscal year 2018–19 increases the national poverty headcount by 2.3 percentage points, while leaving inequality largely unchanged (only a reduction of 0.4 Gini points). Most of the poverty increase is explained by the General Sales Tax (GST) and most of the inequality reduction comes from the Benazir Income Support Program (BISP) Unconditional Cash Transfer (UCT). An international comparison suggests that Pakistan has the lowest inequality reduction, and the highest poverty increase from the list of selected middle- income countries; these distributional results are explained by a combination of low levels of taxation and social expenditure along with limited progressivity of the fiscal system. This paper is structured in nine sections. Section 2 presents the literature review. Section 3 describes the structure of tax revenues and public expenditure in Pakistan. Section 4 describes the data and methodology, and Section 5 discusses the assumptions and allocation rules used for the modeling of the fiscal incidence analysis in Pakistan. Section 6 presents the main results and Section 7 presents an international comparison. Section 8 describes some methodological innovations of the model, and Section 9 presents the main conclusions. Annexes present more details regarding Pakistan’s fiscal system, the specific methodology applied in the Pakistan country case, as well as data recommendations for future research. 2. Literature Review In Pakistan, there have been previous fiscal incidence studies that analyze the distributional impacts of specific taxes or social expenditure items. However, none of these studies provided a unified assessment of the fiscal system. On the taxation side, some of the previous relevant studies include: • Refaqat (2003) analyzed the incidence of GST using the Pakistan Integrated Household Survey (PIHS) 2001–02. The author found that GST was slightly progressive when ranking households based on consumption by decile (proxy of lifetime income). Under this definition, the author estimated that incidence ranged between 3.4 and 4.2 percent for the poorest and richest deciles of the welfare distribution, respectively. The main reason for this result is that most items consumed by the poor were exempt from GST in Pakistan. • Wahir and Wallace (2008) analyzed the incidence of the main indirect taxes using the HIES 2004–05. The authors found that GST and customs duties had an incidence that was more or less proportional, while excise duties were regressive. • Ahmed and O’Donohogue (2009) found that the Income Tax Ordinance in Pakistan improved the distributive effects of personal income tax (PIT) when comparing the two years 2002 and 2005. Deductions for salaried taxpayers contributed the most toward progressivity. • Jamal and Javed (2013) analyzed the distributional impact of GST based on the HIES 2010–11. The results suggested no strict regressivity or progressivity in GST incidence. When assessing the effective GST rate by decile of per capita expenditure, results indicate proportionality with some progressivity at the upper part of the welfare distribution (e.g. the GST incidence ranged from 4.4 percent in decile 1 to 5.5 percent in decile 10). When looking at the Kakwani Index (KI), findings suggested regressivity in GST on food and progressivity on durable expenditure items. • Ara and Asad Khan (2022) analyzed the direct and indirect effects of the main indirect taxes in Pakistan (GST, customs duties, and federal excise duties) based on the HIES 2018–19 and the Input-Output (IO) Matrix 2010–11. The authors found that the overall incidence of all combined indirect taxes is, on average, 20.7 percent in Pakistan, ranging from 22 percent in the poorest decile to 19 percent in the richest decile. The authors show that all components of indirect taxes portray a regressive pattern except the federal excise duties. In terms of the type of commodity, the results suggest that basic food items, and personal and household items were regressive, whereas progressive commodity groups included non-basic food items, utilities, and transport fuel. On the social expenditure side, there are fewer studies, but some of the relevant literature includes the following: • Trimble, Yoshida and Saqib (2011) analyzed households’ electricity tariffs and subsidies in Pakistan based on the Pakistan Social and Living Standards Measurement (PSLM) 2007–08. They found that, although Pakistan’s tariff structure provided a low price to small users, the largest beneficiaries from the subsidy were the richest quintile of the population. The poorest quintile received only 10 percent of the electricity subsidies paid by the Government. The analyses highlighted that the Incremental Block Tariff (IBT) structure was a relatively inefficient method to protect households in the poorest deciles from increasing electricity costs. It also did not help in improving the redistribution of the subsidy, as a major share of the subsidies was benefiting richer households. Therefore, they concluded that the electricity subsidy structure was highly regressive. • Asghar and Zahra (2012) performed a benefit incidence analysis of the public expenditure on education in Pakistan based on the PSLM 2007–08. Their methodology was based on the imputation of unit cost subsidies based on information of usage of educational services available in the survey. The study found that public expenditure on primary education for Pakistan was pro-poor overall. However, for the secondary and tertiary levels of education it was regressive, both at federal and provincial levels. • Saleem (2019) evaluated the political capture and targeting performance of the BISP in Pakistan, based on the PSLM 2009–10. The author found that most BISP beneficiary households belonged to the poorest three quintiles. The poorest quintile in rural areas received a significantly higher share of benefits compared with the same quintile in urban areas. Targeting was better in Punjab and Sindh and significant leakages were found to richest 4th and 5th quintiles in Khyber Pakhtunkhwa (KP) and Balochistan. The main contribution of the present study is that it documents the first fiscal incidence analysis in Pakistan that analyzes the combined distributional impact of the main taxes and social expenditure items in the country. As such, it provides the first unified assessment of the fiscal system. In addition, this study contributes through several methodological innovations to the literature: • Including the zakat paid/received through government channels as a form of religious tax/transfer (see Sections 5 and 6). • Accounting for the differential patterns of consumption informality across deciles for the modeling of indirect taxes, following Bachas et al. (2020) (see Section 8). 3. Tax Revenues and Public Expenditure Public finance in Pakistan involves different government agencies at both the federal and provincial levels. The Ministry of Finance (MoF) is mainly responsible for fiscal policy formulation and management, including expenditure, utilization of foreign exchange, debt sustainability and macroeconomic stability in coordination with other financial institutions. On the taxation side, the Federal Board of Revenue (FBR), a department within the MoF, is responsible for the formulation and execution of tax policy, including the levying and collection of taxes. However, for some taxes (GST on services, the property tax, the agricultural income tax, the motor vehicle tax, and land tax), the tax policy is designed and implemented at the provincial level. For social protection, the Poverty Alleviation and Social Safety Division is responsible for managing Pakistan’s flagship program, the BISP, which is designed and implemented at the federal level.10 Similarly, there are different indirect subsidies (energy, fertilizers, wheat) that are implemented by federal agencies (the Ministry of Energy Power Division and the Ministry of National Food Security & Research, respectively). Regarding public pensions (mostly non-contributory), these are managed by the federal and provincial governments (Accountant General of Pakistan, provincial Accountant General Departments). Finally, the federal government has a role in the collection and distribution of the zakat religious tax/transfer, which takes place through bank channels. Zakat is distributed through the respective provincial zakat and usher departments. Although the mandate to drive fiscal consolidation rests with the federal government, two constitutional developments—the 18th Amendment and 7th National Finance Commission (NFC) awards—now require the federal government to coordinate closely with provincial administrations. As part of the 18th Amendment to the Constitution in 2010, the responsibility for certain federal functions, including education, health, the environment, agriculture, and social protection, was transferred to the provinces.11 The 18th Amendment enhanced the fiscal autonomy of the provinces without establishing any mechanism to implement consolidated fiscal targets for the general government (federal and provincial governments). Under the Constitution, the NFC was to decide on the nature of fiscal relationship between the federal and provincial governments and to address regional disparities. In this regard, the 7th NFC award, implemented in 2010, reduced the federal government’s share in the divisible pool of revenue without any corresponding reduction in its expenditure mandates, resulting in a structural deficit at the federal level. This mismatch between expenditure responsibilities and revenue assignments contributes to persistently high general government fiscal deficits, insufficient capital expenditure, and adverse incentives for fiscal consolidation at the provincial level. 3.1 Tax Revenues In 2018, total revenue from the general government (federal plus provincial) in Pakistan reached 13.4 percent of GDP.12 Out of this, 11.6 percent of GDP came from total federal revenue and grants,13 with the remainder 10 For the latest program under BISP (BISP Targeted Commodity Subsidy Program [TCSP]), the federal government is proposing a cost-sharing agreement with the provincial governments. 11 However, after the 18th Constitutional Amendment, the federal government is still responsible of planning the sectoral policies of devolved subjects (the Ministry of Federal Education and Professional Training and the Ministry of National Health Services Regulations and Coordination). 12 Source: General government revenue based on IMF/WEO April 2022. 13 Source: Federal government revenue based on official administrative data from Pakistan. coming from the provinces. The provincial-levied taxes include GST on services, property tax, agriculture income tax, stamp duties, vehicle tax and other taxes. Table 3.1. Pakistan’s federal government revenue, 2018–19 Government Revenues Value in admin. Total (% Included in Accounts of GDP) Analysis (PKR billion) (Yes/No) Total Revenue & Grants 4,421.03 11.64 No 1 Tax Revenue 4,057.13 10.68 1.1 Direct taxes of which; 1,431.15 3.77 1.1.1 Personal, Corporate and Association of 18.67 0.05 No Personal Income Tax 1.1.2 Withholding Tax, of which: 960.24 2.53 1.1.2.1 WHT on salaries 76.44 0.20 Yes 1.1.2.2 WHT on telecommunications 17.19 0.05 Yes 1.1.3 Advance Tax 344.33 0.91 No 1.1.4 Collection on demand 102.65 0.27 No 1.1.5 Capital Value Tax 5.26 0.01 No 1.1.6 Taxes on Property in provincial data provincial Yes 1.1.7 Agricultural Income Tax in provincial data provincial No 1.1.8 Other Direct Taxes - 1.2 Indirect Taxes, of which: 2,383.88 6.28 1.2.1 General Sales Tax 1,464.89 3.86 Yes 1.2.2 Customs Duties 685.40 1.80 Yes 1.2.3 Federal Excise Duty 233.59 0.62 Yes 1.3 Other Taxes 242.10 0.65 No Petroleum levy 206.30 0.54 No Natural Gas Development Cess 26.80 0.02 No Airport Tax 0.08 No Others 9.00 0.01 No 2 Non-tax Revenue 363.90 0.95 No 3 Grants 33.02 0.10 No Source: World Bank elaboration based on official administrative data. Note: The table excludes provincial-tax revenue structure. Each province has its own tax revenue report and most of the provincial revenue comes from federal transfers. The revenue structure from the federal government in Pakistan is illustrated in the Table 3.1. Federal tax revenues in Pakistan reached 10.7 percent of GDP in fiscal year 2018–19 (92 percent of total federal government revenue). Almost 60 percent of the tax revenue came from indirect taxes, particularly arising from GST, followed by customs duties and the federal excise duties. Direct tax revenues constituted about 35 percent of total taxes, with 67.1 percent stemming from the withholding tax. Pakistan federal tax revenue is lower than peer low- income countries in South Asia (10.7 vs. 12.1 percent of GDP in 2018).14 A tax gap analysis indicates that Pakistan’s tax revenue could reach 26 percent of GDP if tax compliance were raised to 75 percent.15 14 Calculations for South Asia based on World Bank-World Development Indicators data. 15 World Bank (2019). Pakistan Raises Revenue Project. Report No: PAD3237. Direct Taxes Withholding tax on salaries In fiscal year 2018–19, total withholding tax collection (under all categories) was 2.5 percent of GDP, of which withholding tax on salaries was 0.20 percent of GDP (contributing 5 percent of the total direct taxes collection). Withholding tax on salaries in Pakistan is deducted/withheld at source by the employer, based on the employee’s annual gross salary. Salary16 is the amount received by an employee from employment, whether of a revenue or capital nature. In fiscal year 2018–19, withholding tax on salaries was calculated based on six progressive brackets depending on annual taxable income, starting from a first bracket of 0 percent for annual taxable income below PKR 400,000 up to a sixth bracket of PKR 300,000 plus 15 percent of the amount exceeding PKR 4,800,000 for the excess of income above PKR 4,800,000.17 Property tax Property tax is a provincial subject in Pakistan. Each province has its own laws and rules under which properties are valued, and Annual Rental Value (ARV) is estimated. The tax rates and exemptions applicable vary by province as well. In fiscal year 2018-19, Punjab’s collection on property (PKR 11,217.8 million or PKR 100.1 per capita) remained the highest compared with other provinces.18 In contrast, the lowest property tax collection was from Balochistan (PKR 213.3 million or PKR 16.97 per capita).19 Zakat contribution to the Government Zakat is a form of Islamic tax, collected by both the Government and by private charitable organizations. The publicly administered zakat system operates at the Federal level and collects the Zakat Fund. The money collected in the Zakat Fund is distributed according to the Islamic criteria to poor and destitute through local and provincial committees. The rate of zakat varies with the type of asset and the income source. The common minimum amount for those who qualify is 2.5 percent, or 1/40 of a Muslim’s t otal savings and wealth. The compulsory deduction of zakat on bank accounts is implemented at source on the first day of Ramdan-al- Mubarak (for each zakat year based on the Hijri Calendar). Through government channels, the total collection of zakat during fiscal year 2018–19 was PKR 7,377.7 million. Indirect Taxes Customs duties In fiscal year 2018–19, total customs duty collections reached 1.8 percent of GDP. This constituted around 28.7 percent of indirect taxes and 16.9 percent of federal taxes, respectively. In Pakistan, customs duties are levied 16 It includes leave pay, payment in lieu of leave, overtime, bonuses, commissions, fees, gratuities, work condition supplements, monetary and non-monetary perquisites, any allowance including those paid on a fixed basis or which is not exclusively spent on behalf of the employer, profits in lieu of or in addition to salary, pensions, annuities, and tax reimbursement. 17 Source: Source: Federal Board of Revenue of Pakistan (2021). Withholding Tax Regime (Income Tax), Under the Income Tax Ordinance, 2001. Withholding Income Tax Rates Card (updated up to 30-06-2021). 18 Source: Government of the Punjab (2020). Annual Budget Statement for 2019-2020. Property tax per capita based on own calculations, using property tax collection from Government of Punjab and population estimates from PBS Census 2017, assuming 1.9 percent population growth rate. 19 Source: Government of the Balochistan (2020). Annual Budget Statement for 2019-2020. Property tax per capita based on own calculations, using property tax collection from Government of Balochistan and population estimates from PBS Census 2017, assuming 1.9 percent population growth rate. on an ad-valorem basis: (i) the customs duty ranged from 0.75 to 50 percent; (ii) the regulatory duty ranged from 3.3 to 60 percent; and (iii) the additional duty ranged from 2 to 7 percent.20 The FBR has administrative control of the collection of the custom duties in Pakistan. Around 56 percent of customs duty collection is contributed by 10 major commodities, including vehicles, petroleum oil lubricant products, iron and steel, machinery and mechanical appliances, electrical machinery, edible oil, plastic resins, paper and paperboards, articles of iron and steel, and coffee tea and spices.21 Federal excise duties In fiscal year 2018–19, total federal excise duty collections reached 0.6 percent of GDP.22 This contributed around 6.2 percent of federal tax revenue and 10.0 percent of total indirect taxes. Federal excise duty is payable on: (i) goods produced or manufactured in Pakistan; (ii) goods imported into Pakistan; (iii) such goods as the federal government may, by notification in the official Gazette, specify as being produced or manufactured in non-tariff areas and are brought to tariff areas for sale or consumption; and (iv) services, provided or rendered in Pakistan. Major excise duty collection items include artificial non-alcoholic beverages, cigarettes, cement, and petroleum products.23 The purpose of imposing federal excise duties is not only to collect tax revenues but also to discourage the use of some unhealthy products, such as cigarettes.24 General sales tax In fiscal year 2018–19, total effective (net) GST collections reached 3.8 percent of GDP (domestic and imports), and 38 percent of total net federal revenue collection. About 55.5 percent of net GST collections stemmed from imports and 44.5 percent from domestic sales.25 GST in Pakistan is the final sales tax after the application of customs duties and excise duties. In the case of oil derivatives, an additional tax levied before GST is the petroleum levy. GST is charged incrementally at different stages of value addition and input tax credits are applicable.26 The standard GST rate in Pakistan is set at 17 percent, but there are multiple reduced rates and exempted items. The main items under the reduced GST rate include food (e.g., rice and rice flour, fresh milk, several fruits, cooking oil, tea, others), medicine, jewelry, and textbooks, while exempted items also include other food products (e.g., wheat, cereals, chicken meat, beef, fish, eggs) and household sewing machines. Zero- rated items include garments, textile products, clothing accessories, and carpets/curtains. Regarding GST implementation, GST for goods is collected by the federal government; for each good, one rate is defined at the national level (FBR). GST for services is collected by the provinces and the GST rates vary by each province. The administration of GST in Pakistan varies by administrative level. Besides the GST rates, other aspects such as GST exemptions, eligibility for GST input tax credit and GST registration thresholds also vary at the federal and province levels. Currently, businesses that transact on both goods and services need to send 20 Custom duty rates by household product were received from the World Bank’s Macro Trade and Investment team. 21 Source: FBR Yearbook 2018–19. 22 Historically, the federal excise duty had been an important source of FBR revenues. It has declined from 43.1 percent of FBR revenue in 1967–68 to 6.2 percent in 2018–19. 23 For instance, in 2018–19, examples of excise duty rates include: PKR 1,670 per 1,000 cigarettes; 11.5 percent of retail price for carbonated drinks; PKR 85/million tons of LPG. 24 FBR Yearbook 2018–19, p. 20. 25 Source: Own calculations based on GST collections data from FBR and GDP data from the PBS. Year 2018 –19. 26 For this reason, the GST in Pakistan is treated as a traditional Value Added Tax System, with the only exceptions that there are some services at the provincial level for which the GST credit is not applicable. different tax returns at the federal and provincial levels. To help simplify this procedure, FBR and the provinces are developing a single portal to centralize the receipt of all tax returns.27 According to World Bank estimates, the GST tax gap is estimated at 87 percent. The forgone GST revenue or efficiency gap is due to a combination of high GST exemptions (policy gap) and informality (implementation gap). In 2018–19, GST accounted for 1.5 percent of GDP or 61.5 percent of total tax exemptions in the country.28 In terms of the scope of GST informality, in cities such as Lahore the estimation of the informal economy is between 30 and 35 percent.29 Withholding tax on telecommunications In fiscal year 2018–19, the withholding tax on telecommunications brought in revenues equivalent to 0.05 percent of GDP. The withholding tax on telecommunications has three categories: (i) telephones (landline); (ii) mobiles; and (iii) internet (including prepaid cards). In Pakistan, this tax is deducted/withheld at source by the telecommunications service provider. For telephone and internet subscribers, the withholding tax is 10 percent if the monthly bill exceeds PKR 1,000. Below this amount, no tax is charged. For subscribers of mobile telephones, prepaid internet or telephone cards, the withholding tax rate is 12.5 percent of the amount of the bill or sales price of cards. The withholding tax collected under this category is adjustable against the total payable tax liability30. While the withholding tax on telecommunications is listed as a direct tax along with other withholding taxes in government accounts (Table 3.1), the current fiscal incidence analysis treats it as an indirect tax as it applies to sales transactions and not directly on income. 3.2 Social Expenditure In fiscal year 2018–19, total general government expenditures in Pakistan reached 19.1 percent of GDP.31 Out of this sum, 14.7 percent of GDP corresponded to federal government expenditures32 and the remainder corresponded to provincial expenditures. The expenditure structure from the federal government in Pakistan is illustrated in Table 3.2. In the federal government expenditures, non-social expenditure represented the largest outlay (almost 10 percent of GDP), half of which corresponds to debt interest payments (estimated at 4.8 percent of GDP33). The second-largest outlay was defense expenditure (3.0 percent of GDP). In contrast, federal government social expenditure was as low as 1.8 percent of GDP, explained mostly by social protection, which is largely executed at the federal level (1.3 percent of GDP). In the case of health and education, federal government expenditures are much lower (0.07 and 0.33 percent of GDP, respectively), which is consistent with the 18th Constitution Amendment of 2010 that delegates the execution of these functions to provincial 27 A person transacting in both goods and services must file a return to FBR (for sales tax on goods) and one or more to the provinces (for sales tax on services). However, FBR and the provinces are currently working on the development of a common portal/single return for sales tax for goods and services. The return is expected to be launched by the end of this financial year. 28 Source: Own calculations based on GST data from FBR and GDP data from PBS. 29 Williams et al. (2016). 30 This tax is adjustable against total tax liability of individual. Meaning thereby it is form of advanced tax. Source: Amendment through Finance Act. 2018 31 Source: General government expenditures based on IMF/WEO April 2023 Database. 32 Source: Federal government expenditures based on official admin data from Pakistan. 33 Federal expenditure on debt-interest payments for 2018/19 based on the IMF Pakistan Country Report No. 22/288 governments, while the federal government only focuses on policy planning and coordination. When considering the general government expenditures on health and education (which consolidates federal and provincial expenditure), these figures reached 1.1 and 2.5 percent of GDP, respectively.34 Table 3.2. Pakistan’s federal government expenditure, 2018–19 Federal Government Expenditure Value in Total Included Including admin. (% of in provincial Accounts GDP) Analysis expenditur (PKR billion) (Yes/No) e Total Expenditure 5,599.17 14.74 1.1 Defense Expenditure 1,146.79 3.02 No 1.2 Social Expenditure 666.90 1.76 1.2.1 Social Protection 508.52 1.34 1.2.1.1 Social Assistance, of which: 115.65 0.30 1.2.1.1.1 Conditional Cash Transfers 4.01 0.00 Yes 1.2.1.1.2 Unconditional Cash 104.64 0.30 Yes Transfers 1.2.1.2 Social Insurance, of which: 392.87 1.03 Old-Age Pensions 392.87 1.03 Yes 1.2.1.3 Education of which; 125.20 0.33 2.5 percent of total GDP 1.2.1.3.1 Primary 11.07 0.03 Yes 1.2.1.3.2 Secondary 15.38 0.04 Yes 1.2.1.3.3 Tertiary 94.64 0.25 Yes 1.2.1.3.4 Other expenditure 4.10 0.01 No 1.2.1.4 Health 27.13 0.07 Yes 1.1 percent of total GDP 1.2.1.5 Housing and Urban Development 2.60 0.01 No 1.2.1.6 Other social expenditure 3.45 0.01 No 1.2.2 Non-Social Expenditure 3,785.47 9.97 1.2.2.1 Subsidies, of which: 255.97 0.67 1.2.2.1.1 Energy, of which: 171.20 0.45 Yes Electricity 160.50 0.42 Yes Fuel 10.70 0.03 Yes 1.2.2.1.2 Food 30.99 0.08 No 1.2.2.1.3 On Inputs for Agriculture 5.00 0.01 Yes 1.2.2.1.4 Other subsidies 48.77 0.13 No 1.2.2.2 Infrastructure, of which: 57.00 0.15 Road Transport (Highways, 57.00 0.15 No roads, and bridges) 1.2.2.3 Other non-social expenditure 3,002.80 7.91 No Source: World Bank elaboration based on official administrative data and World Bank-WDI for general government expenditures on health and education. Note: Federal government expenditures on indirect subsidies are the official published numbers and they exclude circular debt. 34 Source: World Bank-WDI Databank for general government expenditures on health and education in Pakistan 2018 –19. Direct Transfers BISP Unconditional and Conditional Cash Transfers The Benazir Income Support Program (BISP) is Pakistan’s largest social protection program. It has two types of cash transfer schemes: (i) an unconditional cash transfer (UCT); and (ii) a conditional cash transfer (CCT). BISP beneficiaries are identified via a Proxy Means Test (PMT), known as the BISP poverty score card, which approximates a household’s level of welfare and poverty status using a set of indicators empirically correlated with monetary welfare. Based on the National Socioeconomic Registry (NSER), the authorities calculate a PMT score, which then determines the eligibility of families for specific social programs.35 The UCT program provides quarterly cash payments in the amount of PKR 3,000 per quarter36 directly to female beneficiaries within households (above 18 years of age). As of fiscal year 2018–19, an amount of PKR 104,641.63 million was disbursed to about 5.5 million families under the BISP UCT program.37 In addition, the BISP CCT or Waseela e Tahleem (WeT) program provided an education stipend of PKR 750 per quarter per eligible enrolled child (between 5 and 12 years old). As of fiscal year 2018–19, an amount of PKR 4,012 million was disbursed to 3.244 million eligible children in 50 districts under the BISP CCT program.38 Since its introduction in 1980, the publicly administered zakat cash transfer system has been the principal component in Pakistan’s formal social protection system. The objective of zakat was to assist the needy, indigent, and the poor (termed as mustahiqeen) by providing them with financial assistance from taxes levied on those who possess wealth (sahib-e-nisab). Targeted beneficiaries were widows, orphans, and handicapped people fulfilling certain religious conditions. Due to the lack of a transparent mechanism for the identification of beneficiaries and politically induced allocation of funds across provinces, the impact of zakat transfers on poverty and inequality remained uncertain. Pakistan Bait-ul-Mal (PBM) was established in 1992 to help those needy and poor groups in Pakistan who were either excluded or were ineligible to receive zakat. PBM was funded mainly by the federal government but also received small grants from provincial and local governments and funding from charities and other non- governmental sources. PBM-administered benefits included the Individual Financial Assistance (IFA) scheme, which provided assistance for education and health. 35 In fiscal year 2018–19, BISP-eligible families had a PMT score (derived from NSER records) equal to or below 24.7, and with at least one married female above 18 years old. In addition, there were other exclusion criteria implemented by the BISP based on other data sources. Eligible households cannot own a car nor have a telephone expenditure bill equal or above PKR 1,000 (at 2018 prices). Also, BISP has filters to exclude households above certain income threshold, with international travel or with application for passport using express option. 36 UCT transfers per family were PKR 3,000 per quarter until June 2018 and PKR 5,000 per quarter since July 2018World Bank. 37 Benazir Income Support Programme official data. 38 Benazir Income Support Programme official webpage. As of 2020–21, total expenditure under IFA was PKR 2.8 billion which reached 28.887 beneficiaries.1 According to Pakistan Baitul Maal, other programs include orphans’ assistance (providing boarding and education), women’s empowerment center (WEC), Pakistan Thalassemia center, shelter homes, sweet home, old home, and aiding persons with disabilities. As regards to PMB’s WEC programs: “these centers are providing free vocational training to widows, orphan & poor girls in modern professional skills like, dress designing, embroidery, Basic & Advance Computer Courses, beautician course, Tie & Dye and fabric painting”. However, the impact of PBM is limited due to the financial constraints of the institution, limited coverage of population, and bureaucratic procedures in applying and receiving assistance. [See: IMF 2017 and Laila. U. 2021] The social protection system has included public works programs: the Rural Works Programme (1962–72), the People’s Works Programmes (1972–83) and the former Khushal Pakistan Programme (KPP) for physical infrastructure (2003–07). Typical implementation starts at the provincial level where funds are allocated to districts. The selection of projects and their completion involves local communities. Several factors have influenced the impact of these programs on poverty and inequality, such as a lack of information on the duration of employment, a lack of clear targeting instruction on the selection of districts (either based on infrastructure needs or poverty) and limited fiscal space of the federal government under the Public Sector Development Programme (PSDP) to fund infrastructure projects. [See: Sheikh 2015; Pakistan Center for Philanthropy 2016; Nazir 1996] The Government has also used microfinance programs to alleviate poverty through three different institutions, namely, the national and provincial Rural Support Programmes Network (RSPN), the Pakistan Poverty Alleviation Fund (PPAF), and the Microcredit Bank. The ultimate benefit of microfinance appears to be limited due to the large degree of subsidization and utilization of micro-credit for consumption-smoothing instead of asset creation. The social protection system has included public works programs: the Rural Works Programme (1962–72), the People’s Works Programmes (1972–83) and the former Khushal Pakistan Programme (KPP) for physical infrastructure (2003–07). Typical implementation starts at the provincial level where funds are allocated to districts. The selection of projects and their completion involves local communities. Several factors have influenced the impact of these programs on poverty and inequality, such as a lack of information on the duration of employment, a lack of clear targeting instruction on the selection of districts (either based on infrastructure needs or poverty) and limited fiscal space of the federal government under the Public Sector Development Programme (PSDP) to fund infrastructure projects. [See: Sheikh 2015; Pakistan Center for Philanthropy 2016; Nazir 1996] The Government has also used microfinance programs to alleviate poverty through three different institutions, namely, the national and provincial Rural Support Programmes Network (RSPN), the Pakistan Poverty Alleviation Fund (PPAF), and the Microcredit Bank. The ultimate benefit of microfinance appears to be limited due to the large degree of subsidization and utilization of micro-credit for consumption-smoothing instead of asset creation. Public pensions First introduced as a pensions-cum-gratuity scheme in 1954, the current pension system in Pakistan covers most government jobs through non-contributory pensions. Pensions constitute a monthly payment made by the Government in consideration of past services rendered by a government servant. In 2019, federal government expenditure on public pensions was PKR 392.9 billion, with the total number of pensioners (previously civil servants hired by the federal government) at 1.27 million.39 The sum of provincial pension expenditure across all four provinces was PKR 438.2 billion, whereas the number of pensioners in all provinces remains fewer than 1 million;40 provincial pensions represent an important share of provincial government’s revenues.41 Zakat transfer received through the Government The collections from the religious zakat tax are redistributed to those in need. Those eligible to receive zakat include the poor, especially widows and orphans, and the handicapped. Like the zakat collections, the redistribution of zakat transfers can take place through private charitable organizations or through government channels. The existing zakat government system is composed of one Central Zakat Administration at the federal level, one Provincial Council in each province, a District Zakat Committee in each district, a Tehsil Zakat Committee in each tehsil (sub-division) and a Local Zakat Committee. Out of total funds releasable in one year, 60 percent are allocated for direct payment to individuals for subsistence and rehabilitation through the Local Zakat Committees, while the remaining 40 percent are utilized for stipends for students enrolled in religious educational institutions (8 percent), educational stipends (18 percent), vocational, social welfare stipends (4 percent), for the treatment of mustahiq patients through health institutions (6 percent), and for defraying the expenses of marrying poor girls (4 percent).42 Public Education expenditure Public expenditures on education executed by federal and provincial governments in fiscal year 2018–19 was estimated at 2.4 percent of GDP.43 In Pakistan, education is free and compulsory for all children from ages 5 to 16 years. After the 18th Amendment, education is primarily the responsibility of the provincial governments, except in federally administrated areas, but the federal government remains responsible for planning and formulating national education policies. In 2017–18, the public sector education accounted for 28.5 million (56 percent) of all students enrolled and 22.7 million (44 percent) were enrolled in private sector education.44 Public Health expenditure Public expenditures on health executed by federal and provincial governments in fiscal year 2018–19 were 1.1 percent of GDP.45 According to National Health Accounts 2017–18,46 out of total public sector health expenditures, federal government is funding 16.7 percent, provincial government is funding 65.2 percent and district government/ local bodies are funding 15.1 percent. After the 18th Amendment, health is primarily the responsibility of the provincial governments, except in federally administrated areas, with the federal government being responsible for planning and formulating national health policies. Increasing population demands, insufficient public health facilities, and poor service delivery have led to the increased presence of private health-care providers in the health sector. The public 39 PIDE Knowledge Brief: The Pension Bomb and Possible Solutions, November 2020. 40 Pakistan Pay and Pension Commission reports. 41 In 2019, civil service pensions absorbed 12.3 percent of provincial general revenues in Punjab and 11.8 percent in Sindh, while civil service employees only represented 2.0 to 3.5 percent of the labour force (World Bank 2020a). 42 Khalid Nazir (1996). 43 Source: Own calculations based on federal and provincial expenditure data. 44 Source: Pakistan Education Statistics 2017–18. 45 Source: Pakistan Economic Survey, Health and Nutrition Chapter, 2018–2019. 46 Pakistan National Health Accounts, 2017–18. sector only represents 30 percent of the health facilities whereas the private sector owns nearly 70 percent.4748 There is no national health insurance, and only a small portion of the population had access to health insurance through formal employment. Moreover, Pakistan’s households’ out of pocket expenditure on healthcare (53 percent of total health expenditure)49 is above the South-East Asia Region average (40 percent); this is far greater than the recommended benchmark (15-20 percent)50. Recently, the federal government has launched “Sehat Sahulat Program” in collaboration with provincial governments (Punjab, Islamabad Capital Territory [ICT], and Khyber Pakhtunkhwa [KP]) under the BISP to improve the access of the poor population to good quality medical services, through a micro health insurance scheme. There are user fees paid by patients, but these are very minimal.51 3.3 Non-Social Expenditure Indirect Subsidies Agriculture tubewell subsidies In fiscal year 2018–19, the agriculture sector accounted for 5.8 percent of total electricity consumption52; electricity is critical for operating tubewells for groundwater extraction used for agricultural purposes. Currently, 1.2 million private tubewells are operating in Pakistan, out of which 85 percent (or 1.02 million) are in Punjab, 6.4 percent (76,800) in Sindh, 3.8 percent (45,600) in KP, and 4.8 percent (57,600) in Balochistan.53 The price of electricity in Pakistan is the result of tariff and non-tariff components. The electricity tariff structure for agricultural use is determined by the National Electric Power Regulatory Authority (NEPRA). In Pakistan the tubewells’ owners pay a highly subsidized price for the electricity that they use for agricultural production.54 In the current document the indirect electricity subsidy to agriculture tubewells is defined as the difference between the average cost of supply of electricity and the average monthly tariff paid by the electric tubewells users. In addition, in Balochistan there is a direct subsidy for the installation of agriculture tubewells (PKR 4.8 billion in 2018–19). 47 UK Home Office (2020). Country Policy and Information: Pakistan Medical and Healthcare Provision. 48 For this study, the team did not have complete admin information on total health attendances in the public vs private sector. 49 Source: Pakistan NHA 2019-20 50 South East-Asia region average and World Health Organization international benchmark retrieved from Tandon, A. et al. (2021). 51 According to the PBS National Health Accounts 2019-20 (P. 53): “The OOP health expenditure for access to government hospitals (16.55 percent) is lower than those for access to private hospitals (23.50 percent) because government hospitals provide services at lower rates”. 52 NEPRA State of Industry Report 2018–19”. (P. 53) 53 Qureshi, A.S. (2020). 54 This is a subsidy only for tubewell owners who are operating their tubewells using electricity. Fertilizer subsidies The Government spends about PKR 200 billion annually on subsidies to essential agricultural products. These include a fertilizer subsidy of PKR 150 billion,5556 tax relief and provincial subsidies and incentives57 to the fertilizer industry to provide low-cost inputs for agricultural production.58 The fertilizer subsidy is mainly provided for urea and diammonium phosphate (DAP). The subsidy on urea is a production subsidy provided in the form of cheap gas to the local fertilizer industry producing urea.59 In fiscal year 2018–19, the gas subsidy was PKR 865 per urea bag of 50 kg60. The gas subsidy received by urea producers is equivalent to 50 percent of the urea retail price per bag of 50 kg in the domestic market. In contrast, the final subsidy to urea buyers (agricultural producers) is estimated at 25 percent of the retail price when using international urea prices as a benchmark.61 62 In addition, the Government subsidizes the importation and distribution of fertilizers to keep domestic prices at a reasonable level. This subsidy amounted to PKR 2.7 billion in fiscal year 2018–1963 that goes to mainly subsidize the retail prices of DAP, with the subsidy amounts equivalent to PKR 405 per 50 kg bag of local DAP. Lastly, the Government also provides tax relief in the form of GST rate reductions on urea and the Gas Input Development Cess (GIDC). According to the National Fertilizer Development Centre (NFDC), this tax expenditure amounted to PKR 59.4 billion in fiscal year 2018–19. Domestic electricity subsidies In Pakistan, electricity provided to domestic consumers is subsidized. Single-phase domestic consumers account for 85 percent of all consumers, 47 percent of total consumption, but only 29 percent of revenues billed. 64 The subsidy to domestic consumers is mainly provided in the form of a Tariff Differential Subsidy (TDS) and a cross- subsidy65. In fiscal year 2018–19, a major share of total electricity subsidies was provided to WAPDA/PEPCO (PKR 189 billion out of total outlay of PKR 254 billion). Most of the resources were spent on the TDS (PKR 130 billion).66 The electricity tariff for single-phase domestic consumers is articulated into six consumption slabs: (i) slab 1 for consumers from 0 to 50 kWh/month; (ii) slab 2 for consumers from 51 to 100 kWh/month; (iii) slab 3 for consumers from 101 to 200 kWh/month; (iv) slab 4 for consumers from 201 to 300 kWh/month; (v) slab 5 for consumers from 301 to 700 kWh/month; and (vi) slab 6 for consumers exceeding the 700 kWh/month. All units consumed up to 300 kWh per month receive a TDS, corresponding to the difference between Determined Tariff (DT), and the prevailing tariff. In addition to the TDS, electricity customers above 100 kWh per month also benefit 55 To ascertain the quantum, total PSDP expenditure allocation for 2020 –21 was PKR 650 billion and PKR 150 billion (fertilizer subsidy) amounts to 23 percent of total PSDP allocation. 56 Source: Pakistan Institute of Development (2021). Fertilizer Subsidy an Ineffective Policy Tool to Offer Low Prices of Basic Food Commodities. 57 The cost component of "other incentives in the form of distribution, utilization of local fertilizer” were not published in government accounts. But these types of incentives could include lower transportation cost or packaging bags. 58 PIDE (2021). 59 The fertilizer industry (largest consumer of natural gas in Pakistan) uses gas as both feed stock and fuel (for electricity generation, steam). 60 Source: Administrative data. 61 Own calculations based on 2018–19 local urea price (PKR 1,745 per bag) and international urea price (PKR 2,193 per bag). 62 In other words, due to inefficiencies in the allocation of the subsidy in the production and distribution chain, price difference of final fertilizers bought by agricultural producers is smaller than the fiscal cost of the subsidy. 63 Source: National Fertilizers Development Board. 64 NEPRA State of Industry report 2019. 65 TDS is to cover for prices below cost recovery, and cross subsidy is across different consumer groups. 66 Pakistan Economic Survey 2018–19. from an Incremental Block Tariff (IBT) structure, whereby the tariff of the previous slab is granted to all the units consumed up to the lower bound of the slab in which their consumption falls. As of July 31, 2018, the total circular debt for the domestic electricity subsidy amounted to PKR 421.8 billion.6768 More details on the domestic electricity tariff structure are provided in Annex I. Fuel subsidies Fuel subsidies are provided by the Government to the consumption of petrol, gas, and other fuel products. According to official data published in the government fiscal operations, total fuel subsidies granted by the federal government reached PKR 10.7 billion during the fiscal year 2018–19 (excluding circular debt in the respective sector). Natural gas is a major contributor in Pakistan’s energy mix and 22 percent of it is consumed by the domestic sector. The two major transmission and distribution companies of natural gas in Pakistan are Sui Northern Gas Pipeline Limited (SNGPL) and Sui Southern Gas Pipeline Limited (SSGPL)69. Based on the revenue requirements of these companies, the Oil and Gas Regulatory Agency (OGRA) determines the prescribed price. Both SSGPL and SNGPL have quoted different subsidy amounts (difference in prescribed price and Oil and Gas Regulatory Authority [OGRA] approved price) in their annual reports.70 In the year 2000, the federal government initiated a pricing reform as part of which the guaranteed return formula of the refineries was changed to an Import Parity Price (IPP) formula established by OGRA. OGRA was authorized to review, fix and announce the prices of petroleum products every two weeks in accordance with the approved pricing formula. To protect domestic consumers from international oil price fluctuations, the Government established petroleum subsidies either as a price reduction (Price Differential Claim [PDC]) or in the form of forgone revenue (reductions in the petroleum levy and GST rates charged to petroleum products). In fiscal year 2018–19, there was no PDC provided to consumers, mainly due to lower and relatively less volatile international prices. However, minor adjustments were made using GST and the petroleum levy to provide relief to consumers. 67 Source: The World Bank’s Energy team, based on official data. 68 The TDS captures the flow of government’s subsidies to the electricity sector whereas the circular debt captures the cumulative stock. Circular debt significantly adds to the fiscal burden of the Government and is always included in the IMF conditionalities be it in the electricity sector or gas. 69 Sui Southern Gas (SSGPL) and Sui Northern Gas (SNGPL). 70 SSGCL reported that the subsidy unit they provided was PKR 109.03/MMBTU and the SNGPL reported that the subsidy unit they provided was PKR 350/MMBTU. The design and structure of the wheat subsidy in Pakistan was aimed at supporting farmers and consumers alike. The Government of Pakistan started wheat procurement for the first time in 1968, mainly in response to the large harvests that occurred in the two consecutive preceding years and that put downward pressure on local wheat prices. Gradually, the Government became the main procuring agency of wheat.1 Every year, the federal government announces a “wheat support price” at which the wheat is procured from the market. It is estimated that out of total wheat production of 24.3 million tons during 2018 –19 (PBS), around 60 percent was retained on farms for village, household food consumption and seed, while the Government procured about 20–25 percent. Key players in this subsidy program are farmers, middleman (beopari), government, flour mills, banks, and consumers. For farmers, selling wheat to the Government involves a four-step process; (i) being assigned to a wheat Procurement Centre; (ii) procure gunny/jute bags from the designated Procurement Centre; (iii) delivering the wheat filled bags to the Procurement Centre; and (iv) receiving payment from the bank. For the Government, the procurement is a five-step process involving: (i) setting procurement targets; (ii) establishing centers; (iii) providing gunny bags; (iv) storage; and (v) transportation. Procurement targets are set based on historical data. There are losses at various centers due to storage inefficiencies and pilferage. Small farmers with low marketable quantities face large administrative costs which absorb most of the benefit that the program seeks to provide. Only those farmers who own land are eligible to receive a brown bag (also called jute bags or gunny bag or bardana) through which the Government buys wheat. Most of the small farmers are not landowners and therefore cannot benefit from this subsidy. Storage and procurement weaknesses lead to crop losses. This has led to adulteration of wheat and, consequently, a reduced quality of the final product, which is not beneficial to end consumers. Most of this subsidy is spent on bank mark-ups due to the accumulation of debt. This institutional problem in wheat procurement increases transaction costs and uncertainty, discourages marketing investment and participation and, ultimately, leads to a negative fiscal impact for the Government (Ahmad et al. 2005). It is not possible to simulate the wheat subsidy in the current fiscal incidence analysis due to the lack of data on the amount of subsidy retained by each actor of the supply chain, complexity of the distribution system, and the non- availability of the administrative information. However, actual subsidies benefit reaching to genuine farmers and end consumers is only a small share of the total amount, since subsidies are shared by all actors involved in the supply chain process, not just end consumers. 4. Methodology and Data 4.1 Methodology: Commitment to Equity The design and implementation of fiscal policies should consider different dimensions: (i) fiscal sustainability; (ii) economic efficiency; (iii) externalities; and (iv) equity (Figure 4.1). The objective of this fiscal incidence analysis is to analyze the impact of taxes and social expenditure on poverty and inequality in Pakistan. The motivation for performing such an analysis is threefold: (i) there has never before been a comprehensive distributional assessment of the country’s fiscal system; (ii) the current Government has prioritized enhancing tax revenue mobilization and the rationalization of subsidies to create fiscal space and to provide additional targeted social transfers to the poor; and (iii) given the above, findings from the fiscal incidence analysis create an evidence base on how to promote tax and expenditure reforms that are consistent with both fiscal sustainability and fiscal equity. The fiscal incidence analysis measures how the Government’s taxes and social expenditure affect households’ welfare. The goal of such an analysis is to answer questions such as: (i) What is the impact of taxes and social expenditure on poverty and inequality? (ii) Which taxes and social expenditure items are progressive or regressive? (iii) Who bears the burden of taxes and receives the benefits of social expenditures? and (iv) Which households are net payers or net receivers of the fiscal system? Moreover, by developing detailed and parametrized microsimulation models for fiscal incidence analysis, it creates a platform that can be used not only to assess the distributional impacts of existing fiscal policies but also to simulate the ex-ante distributional impact of potential policy reforms. These models can therefore be particularly useful to inform evidence-based fiscal policy design, considering equity considerations both before and after implementation. Figure 4.1. Key considerations of fiscal policy Source: Authors’ elaboration. The fiscal incidence analysis for Pakistan (2018–19) is developed using the Commitment to Equity (CEQ) Methodology,71 which provides a systematized framework to determine the impacts of the fiscal system on poverty and inequality. The standard CEQ model covers several fiscal interventions that affect households’ welfare (direct taxes, indirect taxes, direct transfers, indirect subsidies, and in-kind benefits from health and education) and simulates how fiscal systems work in practice. Other fiscal incidence methodologies, such as the Tax-benefit Microsimulation Model for the European Union (EUROMOD), cover fewer fiscal interventions (e.g., for instance, in-kind benefits are typically excluded). EUROMOD also simulates the different fiscal interventions de jure, or based on the design rules, which could be a reasonable approach in developed countries. However, in developing countries, facing tax informality and social program implementation challenges, it makes more sense to aim for de facto modeling, as embraced by the CEQ Methodology. The CEQ Methodology has been implemented in over 70 countries, which facilitates the production of results that are internationally comparable. As part of the dissemination efforts of these fiscal incidence initiatives, the equity aspect of fiscal policies has become more policy-relevant and it has been included as the Sustainable Development Goal 10.4.2, which measures the redistributive impact of the fiscal system on inequality reduction. Figure 4.2. Definitions of income underpinning the CEQ fiscal incidence analysis - Direct transfers + Indirect subsidies + In-kind transfers BISP UCT, BISP Electricity subsidy, Education & Health CCT, others natural gas subsidy Disposable Pre-fiscal income income= = Market income Net market Consumable Final income Official + pensions income income consumption from HIES - Indirect taxes + Direct taxes GST, custom duties, Withholding tax withholding tax on on salaries, telecommunications property tax Source: Adapted from Lustig (2018). Commitment to Equity Handbook. Building a fiscal incidence model based on the CEQ Methodology requires understanding how a country’s fiscal system works (based on legislation and administrative data), and then modeling how taxes and social expenditure are allocated across households and individuals, using micro data from a representative socio- economic household survey. Once all taxes and transfers are modeled, the CEQ Methodology calculates different income concepts for each household to assess how fiscal policy affects households’ income at various stages of redistribution (Figure 4.2). For each household, the analysis starts with “market income” or “market income plus pensions” as the pre-fiscal income (before taxes and transfers), and then subtracts taxes and adds transfers to obtain households’ final income or post-fiscal income (after government’s taxes and transfers). 71 The Commitment to Equity project (CEQ) is led by Nora Lustig at Tulane University. The latest CEQ Handbook was published in 2018 and is available online. Limitations. Fiscal incidence analysis is “both methodology and art” (Lustig 2018). Every country poses its own challenges given the complexity of fiscal systems in general and country-specific data availability (or non- availability): (i) some indicators such as inequality could be misestimated, given that household surveys typically fail to capture top-income households well72 ; (ii) there are certain fiscal interventions (for instance Corporate Income Tax73, public expenditure on infrastructure74, defense or debt interest payments) which are not included in the fiscal incidence analysis; (iii) it does not consider the quality or long-term impacts of the public health and education services. The CEQ Assessment framework for fiscal incidence analysis is a static and retrospective accounting exercise without behavioral, lifecycle, or general equilibrium effects. This means the incidence results represent an “overnight impact” counterfactual75 ; and the direction of bias of these “overnight impact” estimates is unknown. In practice, there are no standard errors calculated which would allow a statistical assessment of the allocations made to individuals and households.76 Nevertheless, the CEQ Assessment framework for fiscal incidence analysis provides a standard methodology (which also enables international comparisons) for estimates of the impact of fiscal policy on poverty, inequality, and social welfare more generally. 4.2 Data Sources for Pakistan The Fiscal Incidence Analysis in Pakistan uses the following data sources: • The Household Integrated Economic Survey (HIES) 2018–19 (Pakistan Bureau of Statistics [PBS]), which surveyed 24,809 households and is representative at the federal and province levels. This survey has household and individual-level information on income, employment, consumption, health, and education. • Administrative data on direct and indirect taxes from the Federal Board of Revenue (FBR) (for federal taxes) and from provincial revenue authorities (for provincial taxes) complemented with a review of relevant tax legislation. • Government expenditure (executed) at the federal level (MoF) and the provincial level (provincial governments). • Social protection expenditures (executed), number of beneficiaries, and allocation rules for the main social programs (the BISP and the World Bank’s Social Protection team). • Indirect subsidies for agriculture tubewells (Ministry of Water and Power [MoWP] and Power Information Technology Company [PITC] and Pakistan Bureau of Statistics [PBS]), fertilizers (Ministry of National Food Security and Research [MoNFSR] and National Fertilizer Development Centre [NFDC]) and domestic electricity distribution (Ministry of Water and Power [MoWP], Pakistan Information Technology Company [PITC] and the World Bank’s Energy team). Indirect subsidies for gas and petrol 72 Household surveys also do not capture some low-income households very well either: institutional populations (prisons, old-age care facilities, youth care facilities) and households without a domicile address (informal housing. 73 On one hand, CIT is better modelled with tax administrative data. Furthermore, data on CIT cannot help us allocate CIT burdens to households in the standard CEQ FIA framework. 74 Examples of the relevant infrastructure investments that we are missing and that could affect households differently in the income distribution could be connectivity infrastructure, water infrastructure (publicly provided water), local roads. The methodology for including infrastructure expenditure in fiscal incidence models is currently under development by the CEQ Institute. 75 Meaning distributional (static) impacts, not casual impacts. 76 There are, however, CEQ Assessment procedures for assessing the statistical significance of the estimated impact of a fiscal policy (or set of fiscal policies) on poverty, inequality, and other indicators. were collected from Oil and Gas Regulatory Authority (OGRA), the National Electric Power Regulatory Authority (NEPRA), Sui Southern Gas Company Limited (SSGCL), Sui Northern Gas Pipeline Limited (SNGPL), Pakistan State Oil (PSO), the Federal Board of Revenue (FBR), the State Bank of Pakistan (SBP), and the MoF. • Government education expenditure (executed) and students’ enrolment (Ministry of Federal Education and Professional Training [MoFEPT], provincial education departments). • Government health expenditure (executed) and health cases (Ministry of National Health Services Regulations and Coordination [MoNHSRC], National Health Accounts, provincial health departments). • Input-Output Matrix for fiscal year 2013–14 (developed by International Food Policy Research Institute [IFPRI]77).78 • Labour Force Survey (LFS) 2017–2018 to simulate the labor informality and public sector employees in the HIES. 77 See: Debowicz, D. et al. (2012). 78 While the PBS continues to use an IO Matrix which was constructed in the late 1990s, this study has used the IO Matrix developed and published by IFPRI in 2013 to model indirect effects. By using the IFPRI IO Matrix from 2013 –14, the underlying assumption is that the production technologies in each sector did not change significantly in 2018 –19 relative to 2013–14. The team is aware of the ongoing work program in the ADB to prepare a multi-regional IO Matrix. However, the ADB MRIO only covers 35 sectors (in contrast to IFPRI, which has 60 sectors), hence the IFPRI IO Matrix was preferred because it allows for a better estimation of the indirect effects of taxes disaggregated by sectors. 5. Application of the Fiscal Incidence Analysis 5.1 Coverage of the Fiscal System The fiscal incidence analysis in Pakistan (2018–19) modeled the core fiscal interventions from the country’s tax system and public social expenditure. On the tax side, the study modeled GST (federal and provincial), customs duties, federal excise duties, withholding tax from salaries (first occupation), withholding tax on telecommunications, and the property tax. Together, the total direct and indirect taxes included for potential allocation to individuals and households in the microdata represent about PKR 2,443.2 billion, equivalent to 61 percent of federal tax revenues and 45 percent of provincial tax revenues in fiscal year 2018 –19. The lower coverage of provincial tax revenues is due to data limitations to model some provincial taxes, such as the agriculture income tax (AIT) and the motor vehicle tax.79 On the social expenditure side, the analysis covers the main social protection programs (BISP UCT and BISP CCT), indirect subsidies (agriculture tubewells, urea fertilizer, domestic electricity, natural gas, and petrol), health expenditure (inpatient and outpatient services), and education expenditure (pre-primary and primary, secondary, and tertiary). Altogether, the total social expenditure included for potential allocation to individuals and households in the microdata represent PKR 1,500 billion, which represents about 8 percent of federal government expenditures and 23 percent of provincial government expenditures in fiscal year 2018–19. In addition, the study also includes the zakat collected and received through federal government channels.80 The total expenditures and taxes actually allocated in the microdata using the current incidence model produce an tax-to-expenditure ratio that is higher than the ratio observed in the administrative data for the same expenditure and taxes (see Section 5.3): the ratio of total taxes to total public transfers81 was 86 percent or only half of the administrative ratio of 163 percent for the same collection of taxes and expenditures. The reason behind this discrepancy is that the allocation of taxes in the survey depends on correct information about income and consumption patterns. However, the HIES only covers 45 percent of the total consumption in national accounts, which could be due to missing top-income households in the survey. In contrast, the estimation of public transfers performs better in the survey model given that the allocation is based on socio-economic characteristics and many transfers aim to cover the poorer segments of the population (which is better represented in the HIES). On the other hand, due to limitations of the HIES data and fiscal administrative data, the fiscal incidence model focuses on the four provinces plus the capital (Balochistan, KP, Punjab, Sindh, and ICT) only.82 Also, due to further 79 For example, for Punjab motor vehicle tax we need to know the engine capacity of the vehicle and this information is not available in the HIES. 80 The zakat (collected and transferred) by the government is considered outside of the official social protection system in this fiscal incidence analysis. The government of Pakistan only acts as an intermediary in collecting zakat, but it redistributes it to zakat and usher departments. 81 Total taxes include direct and indirect taxes. Total transfers refer to direct transfers, indirect subsidies and in-kind benefits from health and education. 82 The fiscal and administrative data available for other regions and administrative units is not detailed enough for the inclusion in the CEQ Assessment. For social protection, the benefit incidence analysis of this study uses total executed data limitations, the analysis excluded taxes such as CIT and AIT, 83 and withholding tax from other sources84 (see Annex V). The analysis also excludes infrastructure expenditure, defense, debt interest payments, small social programs (without clear allocation rules), and the wheat subsidy program (due to its complexity and lack of information on the subsidy distribution between the multiple actors involved). 5.2 Assumptions behind the Analysis The analysis follows the CEQ Framework (Figure 4.2) for the definition of income concepts. Given that the official welfare aggregate in Pakistan is based on consumption expenditure, the calculation of income concepts starts with disposable income which for this report is defined as the official consumption aggregate available in the HIES 2018–19.85 This means that the income concepts are all derived from our primary observed variable, consumption expenditure (Figure 4.2). The CEQ income concepts above disposable income are calculated backwards: net market income equals disposable income minus direct transfers, while market income plus pensions equals net market income plus direct taxes. The CEQ income concepts below disposable income are calculated in the standard way: consumable income equals disposable income minus indirect taxes plus indirect subsidies, while final income equals consumable income plus in-kind benefits from health and education. This study treated public expenditures on the civil servant pension system in Pakistan according to the “Pensions as Deferred Income” (PDI) scenario. The PDI scenario implies: (i) the contributions to private social security schemes are considered mandatory savings (of income) rather than a tax on income; (ii) pension income for pension recipients is treated as current “market income plus pensions”. As public civil servant pensions in Pakistan are funded entirely from current expenditures – there is no system of explicit or implicit civil servant pension contributions, nor any pension assets held in a fund or an administrator from which pension payments can be drawn - they are considered in essence part of the total wage bill for the active and retired civil service sector86 and therefore outside of the set of social expenditures and subsidy expenditures that we wish to analyze in this fiscal incidence analysis. Hence, public pensions in this analysis are treated as deferred income (embedded in the prefiscal income concept “market income plus pensions”). The main assumptions used to model each fiscal intervention are described below and further details on the fiscal parameters used are presented in Annex I. expenditure and total beneficiaries at the national level. For health and education expenditure, it was possible to exclude the portions of the federal PDSP expenditure allocated to FATA territories and other regions, but the PDSP is only a small fraction of the federal expenditure. For instance, for the case of education, in 2018 –19 the PDSP expenditure was about 5 percent of the general government’s expenditure on education (and only 5 percent for this PDSP expenditure was allocated the FATA territories and other regions). 83 This study tried to model AIT but the tax collections from the model were overestimating the official administrative tax collections; this is explained due to high informality in this tax which could not be accounted in the model. For this reason, the AIT was excluded from the model. 84 This study tried to model the withholding tax from electricity but all households in the HIES reported consumption levels that were below the eligible taxable threshold. Also, for the withholding tax on education, it was not possible to disaggregate the eligible expenditures in the HIES, so the tax collections in the survey were overestimated. For these reasons, these two taxes were excluded from the model. 85 Using consumption as the welfare measure (instead of income) is typical in low-income countries because the former is easier to report or less subject to fluctuations. (Burtles, G., cited in Lustig, 2018). 86 According to World Bank (2020b): “Pakistan’s civil servants’ pension scheme is a non -contributory defined-benefit scheme providing relatively generous benefits for career civil servants. A worker is entitled toa n annuitized benefit at age 60 or upon completion of 25 years of service”. (P. 1) Direct Taxes Withholding tax on salaries Withholding tax on salaries was modeled based on the HIES data for the individuals that reported being paid employees at their main occupation.87 From the HIES, annual income from first occupation was grossed up based on PIT and social security contributions. However, the HIES did not have any variable to identify public sector employees or any proxy of labor informality (e.g., contract type or contribution to taxes). Hence, the study leveraged on the LFS 2017–18, to build regression models that allow to predict employees’ formality (probability of having a contract) and public sector employees, based on employment characteristics that were common in both the LFS and the HIES (see Annex II).88 Employees in the main occupation identified in the HIES were classified as formal if they were public sector employees or if their probability of having a contract was above 80 percent (both predicted variables in the HIES). For those formal individuals, the withholding tax on salaries was calculated based on the six-bracket schedule available in 2018–19. Two limitations of this model are the exclusion of second-job employees in the model (which should self-report taxes voluntarily) and the missing information on top-income households in the HIES data. Property tax The property tax was simulated in the HIES based on the following variables: (i) property ownership and property type (residential and commercial); (ii) self-reported annual rental value (ARV) received from the property (cash/in kind) if rented out; and (iii) province (since tax rules vary by province). Based on these variables and the provincial tax rules, the property tax was calculated for each province. The analysis was restricted to urban areas, given the application rules of this tax. One limitation for modeling this tax is that it was not possible to proxy evasion levels informality. Zakat payment to the Government The zakat contribution paid through government channels was included in the model and considered as a direct tax. The analysis was based on direct identification in the HIES. In the HIES there is one variable of “transfers paid out last year” by type of transfer and value. In that variable it was possible to identify the amount paid in “zakat/usher” through government channels. One limitation is that, since “zakat/usher” are aggregated under the same category, there could be slight overestimation of zakat payments. Indirect Taxes 87 This study tried to model the withholding tax on salaries for individuals that have a second occupation. However, there was a complexity for predicting tax compliance since individuals with multiple income sources should be self-reporting and filing their income taxes to the tax authority. It was not possible to predict a formality proxy for the second occupation; labor informality regression coefficients from LFS 2017–18 were based on first occupation. 88 The variables of “public sector employee” and the “probability of having a contract” were predicted in the HIES for the main occupation of individuals, based on employment characteristics and regression coefficients from the LFS 2017 –18. The model focuses on individuals that reported being “paid employees” for the main occupation (restricted sample). The covariates used in both models were: urban dummy, male dummy, education level, industry classification, and occupation type. Employees were classified as being from the “public sector” if their predicted probability was equal or above 35 percent; this allowed to replicate the same share of public sector employees available in the LFS (17 percent). Employees were classified as “formal” if their predicted probability was equal or higher than 80 percent or if they were public employees; this allowed to have a total number of potential PIT salaried taxpayers equal to 906,281.27, closer to the administrative data (1.13 million). Customs duties The main customs duties (customs duty, regulatory duty, and additional duty) were simulated based on the expenditures by goods reported by households in the HIES and effective customs duty rates available at the product level.89 Since the HIES does not allow to identify imported items (as in most household surveys), the simulation of custom duties is based on the Law of One Price (LOOP)90. The analysis calculates the direct effects of custom duty effective rates for all taxable products (assuming the LOOP)91. Goods with controlled prices were excluded from customs duty simulation, i.e., petroleum products, wheat, and sugar. The final simulation applies downscaling to the survey-based custom duty estimations, to match the macro-effective rate (total customs duty collections relative households’ consumption). The study acknowledges that the LOOP assumption comes with several caveats92. On another hand, other limitations are that the current custom duties’ simulation does not account for informality in households’ purchases nor differences in consumption patterns of imported goods between poor and rich households. General sales tax GST was simulated in the HIES based on the expenditures by goods and services reported by households.93 The analysis uses federal GST rates for goods, province-differentiated GST rates for services, and rules on GST input tax credit eligibility differentiated at the federal and province-level. For the tradable goods that have different GST rates (domestic vs imported), the study assumes GST for imported products (under the LOOP). Total households’ purchases were divided between formal share and informal share, where the informality purchase ratio was estimated based on an out-of-sample prediction from Bachas et al. (2020), given that the HIES does not have information on households’ place of purchase. A sensitivity analysis of the GST estimations (with and without informality) is presented in Section 8.1. Two types of effects were calculated in the GST model: (i) direct effects (statutory rates) for the formal purchase share of taxable goods and services; and (ii) indirect effects from taxed inputs were applied for the informal share of taxable goods and services, for exempt goods/services and for goods/services not eligible for the GST-input tax credit.94 Indirect effects were calculated based on the 89 The analysis focuses on purchased expenditure (excluding home production, gifts and salaries in-kind consumed). The netting down of purchased expenditure was calculated as = expenditure/[(1+gst rate)*(1+excise rate)*(1+all duty rates)]. 90 Following the Law of One Price (LOOP), the custom duties are applied to all taxable products, even if domestic and imported goods cannot be differentiated. Under the LOOP and a competitive economy, the prices of domestic goods would converge to the prices of equivalent imported products. 91 The analysis excludes the indirect effects of custom duties from the analysis since the methodology is currently under development. 92 This study acknowledges that there are strong assumptions/limitations that come from using the LOOP in the current model: (i) it assumes perfect market competition (domestic and imported goods prices equate); (ii) it assumes that domestic and imported products are perfect substitutes for households. (iii) It does not account for general equilibrium effects since this is a static model. Nevertheless, the LOOP (and scaling-down to match the macro effective rate) has been used to analyze the distributional incidence of custom duties in other CEQ’s, such as Armenia (Younger et al.). 93 The analysis focuses on purchased expenditure (excluding home production, gifts and salaries in-kind consumed). The netting-down of purchased expenditures was done as = expend/(1+gst rate). 94 For exempt, informal GST taxable purchase or GST taxable purchases that are not eligible for GST input tax credit, the indirect effects are applicable since final sellers cannot recover the GST they paid on their inputs. The indirect effect of GST is also known as the cascading effect (i.e., the tax-on-tax effect). When producers and distributors cannot claim GST credits on inputs, there is a cascading effect as they will try to recoup the hidden VAT (part of their costs) into higher final prices to be paid by the consumer. Augmented IO Matrix95 and the Cost-Push Model,96 following Jellema and Inchauste (2018) and Inchauste et al. (forthcoming) (see Annex III). Meanwhile, the GST model incorporates the Government’s differential tax policy with respect to petroleum products. As previously mentioned, to protect domestic consumers from international oil price fluctuations, the Government provides petroleum subsidies either as a price reduction or indirect subsidy (PDC) or in the form of tax expenditure (reductions in the petroleum levy and GST rates). In fiscal year 2018 –19, the Government applied the latter approach (reductions in the GST rate, below the 17 percent standard rate). Hence, the GST model incorporates the reduced GST rates (differentiated by month) to petrol. Also, the GST model assumes 100 percent formality for these petroleum products, given that households cannot purchase these products in informal markets. Federal excise duties The federal excise duties were simulated in the HIES based on the expenditures by goods reported by households.97 The taxable item list is shorter than other taxes and includes carbonated drinks/juices, cigarettes, cooking oil (vegetable, mustard), tobacco and cigarettes, compressed natural gas (CNG) expenses and gas charges. Two types of effects were modeled: (i) direct effects (statutory rates) were applied for the share of formal purchases of excisable items; and (ii) indirect effects98 from excisable products that are used as production inputs (petroleum products and cement) were calculated and applied to all households’ purchases. The indirect effects were calculated based on the IO Matrix and Cost-Push Model following Jellema and Inchauste (2018) and Inchauste et al. (forthcoming) (see Annex III). The share of households’ formal purchases was calculated based on an out-of-sample prediction using cross-country regression coefficients from Bachas et al. (2020) (see Section 8.1). Withholding tax on telecommunications Withholding tax on telecommunications is treated as an indirect tax in the model because it is based on households’ consumption.99 This tax was simulated in the HIES based on the following variables: (i) annual households’ expenditure reported on telephone and internet; (ii) access to internet in the household; and (iii) type of telephone connection (mobile owner vs landline owner). Based on these variables and the applicable tax rates for internet and telephone, withholding tax on telecommunications was calculated. 95 For the GST, the IO Matrix is augmented for the IO sectors that have a mix of exempt and non-exempt items. Given the lack of data on production shares, households’ consumption shares calculated in the HIES are used to weight what portion of the technical coefficient is exempt vs non-exempt in the augmented IO sector. See Annex III. 96 Similarly, to the direct effects, for the IO sectors that had a different domestic and imported GST rate, the latter was applied under the Law of One Price. 97 The analysis focuses on purchased expenditure (excluding self-production, gifts and salaries in-kind consumed). The netting-down purchased expenditure was calculated as= expenditure/[(1+excise rate)*(1+gst rate)]. 98 The indirect effect from inputs that pay excises affects all final goods that use these inputs since sellers cannot claim a credit for the excises they paid on their inputs. Hence, the indirect effect happens because the sellers pass-through a hidden excise (in the form of higher final prices) to consumers to recover part of their costs. 99 In contrast, in the Pakistan’s governments’ accounts, the withholding tax on telecommunications is reported as a direct tax (see Table ES1). Direct Transfers BISP Unconditional Cash Transfer program The BISP UCT program was modeled based on a combination of direct identification and simulation in the HIES. Although the BISP is implemented at the family level100, the analysis is performed at the household level given the socio-demographic structure of the HIES. Based on administrative data, the equivalence of 1.00 beneficiary household being equal to 1.27 families is used in the model. About half (48 percent)101 of the total beneficiaries in fiscal year 2018–19 (target 3,888,011 households)102 were self-reported in the HIES and the rest were completed based on a random selection from the pool of eligible households. For the random selection of additional beneficiaries, the eligible pool of households was defined as those that combined the following characteristics: (i) having a simulated PMT score below or equal to 24.7; (ii) having at least one female married above 18 years old; and (iii) not self-reported as beneficiaries in the HIES. The replication of the PMT algorithm and households’ scores based on the HIES were received from the World Bank’s Social Protection team . The PMT is based on the 2013 adjusted formula and has been replicated in the HIES 2018 –19. Based on additional eligibility criteria from the BISP, the analysis excludes households that own personal car or that have a telephone bill above PKR 1,000 monthly (deflated to 2018–19 values).103 Lastly, for all beneficiary households (self-reported and simulated), the annual statutory transfer of the BISP UCT was imputed (equivalent to PKR 18,000104 for fiscal year 2018–19). Total BISP UCT beneficiary households identified in the HIES (self-reported and simulated) were equal to 3,888,011, equivalent to the same number targeted in the administrative data. BISP Conditional Cash Transfer program The modeling of the BISP CCT program was based on simulation (random selection of eligible households in the HIES). Again, while the BISP CCT is allocated at the family level, the analysis in the HIES is allocated at the household level using the equivalence provided in the administrative data (1.00 beneficiary household = 1.27 families). The eligible population of the CCT was defined as households that complied with the following criteria: (i) being a beneficiary of the UCT transfer; (ii) having at least one child eligible for the transfer (between 5 and 12 years old, currently attending school); and (iii) living in one of the 50 districts where the CCT was being implemented in 2018. From this eligible sample of households, the model made a random selection, which resulted in 825,545 CCT beneficiary households (compared with 1,087,000 CCT beneficiary households in the administrative data). The reason why the algorithm in the HIES reached fewer beneficiary households than the administrative target is because of the restriction requiring households to belong to one of the 50 beneficiary districts. For all simulated BISP CCT beneficiary households, the annual statutory transfer for fiscal year 2018– 19 was imputed (PKR 750 per quarter per child), which means that the transfer was multiplied by the number of eligible children per beneficiary household. 100 BISP defines a family as a group composed by: (i) husband, wife and unmarried children; (ii) husband and wife without any children; (iii) divorced/separated woman with or without her unmarried children, living alone or with her parents/relatives; or (iv) a widow with or without her unmarried children, living alone or with her parents/ relatives. 101 In the HIES, 2,572,492 BISP beneficiary households were identified based on the variable “annual income received from the BISP program”. 102 In the official administrative data total BISP UCT beneficiaries was 5,054,414 families. 103 Given data limitations from the HIES, other exclusion criteria applied by the BISP could not be implemented, such as: households’ income (since this variable is noisy in the HIES), international travel or application for passport using express option. 104 Beneficiary families receive quarterly transfers. The BISP UCT transfer was PKR 3,000 per quarter per family until June 2018 and then PKR 5,000 per family per quarter since July 2018. Hence, the assumption applied was that households received PKR 3,000 for the first quarter of fiscal year 2018–19 and PKR 5,000 for the following three quarters. Zakat transfer received through the Government The modeling of the zakat transfer was based on direct identification. In the HIES, there is one variable of “transfers received last year” by type of transfer and value. With this information, it was possible to identify the amount received in “zakat/usher” through government channels. One limitation is that this HIES category includes usher aggregated along with zakat, so the latter could be slightly overestimated. Indirect Subsidies Fertilizer subsidy The analysis focused on the gas subsidy that domestic producers of urea fertilizer receive. The price differentials (lower domestic price relative to the international price) were used as evidence of the urea subsidy to the fertilizer users (small farmers) and to calculate the indirect effects to end consumers. The modeling of the subsidy was based on simulation using variables from the HIES agriculture module. For calculating the size of the subsidy amount, the metric was the difference between the domestic retail price and the international price of the urea fertilizer.105 Based on the price differential, the implicit subsidy was equivalent to 25 percent of the domestic retail price.106 It is worth noting that, when comparing the urea fertilizer subsidy to consumers (25 percent of the domestic retail price) with the urea fertilizer subsidy to producers (50 percent of the domestic retail price),107 the comparison suggests that half of the total subsidy is kept by urea producers. Then, the urea quantity was calculated by dividing households’ total expenditure on local urea (estimated)108 by the average domestic retail price of urea. For allocating the benefits of the urea fertilizer subsidy across households, one needs to consider two things. First, the subsidy is provided as an agricultural input subsidy (the gas used by urea producers). Second, some households in the survey are both “farmers” and “final consumers” (subsistence farmers), so they could benefit from the urea fertilizer subsidy directly and indirectly. Following Lustig (2018), in the case of indirect subsidies to agricultural inputs, both direct effects and indirect effects need to be calculated in the fiscal incidence analysis. Direct effects of the agricultural input subsidy would be treated as direct transfers (not as a subsidy) for small- scale farmers in the household survey.109 At the same time, indirect effects are calculated for all consumers since the price subsidy of urea fertilizer as a production input will be passed to the final goods consumed by households. This also builds on the assumption that there is a direct pass-through to consumers, which depends on production elasticities and market power. 105 The ideal counterfactual for the subsidy calculation would have been the market price of urea in the domestic market in the absence of the subsidy. Given that these data were not available, the international retail price of urea was used as the counterfactual. 106 The domestic price for urea was PKR 1,745 per bag vs. PKR 2,193 per bag internationally, hence the implicit subsidy was 25 percent of the domestic retail price. 107 The government provides a direct gas subsidy to the domestic urea fertilizer producers, this subsidy was equivalent to PKR 865 per bag in 2018–19 and when divided by the domestic price of urea (PKR 1,745 per bag), the implicit subsidy is 50 percent. 108 The households’ total expenditure on fertilizers reported in the HIES was multiplied by the share of urea in total fertilizer s (74 percent) and the share of urea that is local (95 percent). These shares were calculated based on macro data of fertilizer s’ usage. 109 Lustig (2018): “When production and consumption decisions are intertwined, as happens with small subsistence farmers in developing countries, subsidies to inputs should be treated as direct transfers rather than a subsidy”. In summary, for the urea fertilizer subsidy, two types of effects were modeled in the HIES: (i) direct effects (direct transfers to farmers that own-consume their crops) were calculated based on the estimation of households’ consumption of local urea. The subsidy amount (direct transfer) was calculated as the subsidy unit times the urea quantity purchased by the household adjusted by the share of own-consumed crops estimated;110 and (ii) indirect effects (subsidized urea fertilizer inputs that lower prices of final goods) were calculated and applied to all households’ purchases. The indirect effects were calculated based on the IO Matrix (where the fertilizer sector receives the price-subsidy) and the Cost-Push Model.111 Agriculture tubewell subsidies The analysis focused on the electricity subsidy that is provided to the agriculture tubewells in Pakistan. The study explored the subsidy by taking the difference between cost of electricity per unit supplied to the agriculture sector and the average supply electricity cost. For the agriculture tubewell subsidies, two types of effects were modeled: (i) for direct effects (direct transfers to farmers that used the electricity for agriculture tubewells), the subsidy was calculated as the subsidy amount times the electricity quantity (estimated) used by farmers times the share of own-consumed crops (estimated); and (ii) indirect effects (subsidized electricity in the agriculture sector that lowers prices of final goods) were calculated and applied to all households’ purchases. The indirect effects were calculated based on the IO Matrix (subsidy applied to the electricity distribution sector) and the Cost-Push Model.112 Electricity subsidy to domestic consumers The analysis focused on the electricity subsidy that is provided to domestic consumers in Pakistan. The analysis explored the subsidy by taking the difference between the cost of electricity per unit supplied and the actual electricity price paid by domestic consumers.113 For calculating the direct effects of the domestic electricity subsidy,114 the following steps were taken: (i) the amount of households’ expenditure on electricity based on the “electricity charges” reported in the HIES household expenditure module was identified; (ii) the quantity of electricity consumed was estimated by dividing total expenditure on electricity by the monthly price of electricity in 2018; and (iii) the amount of the direct effect of the subsidy was calculated as the subsidy amount (difference between the cost of electricity per unit supplied and the actual electricity price115 paid by the domestic 110 For the numerator of own-consumed crops share a selection of raw crops or mildly processed crops where selected, based on households reported self-consumption. For the denominator, the total value of crops produced (available in the agriculture module of the HIES) was used. 111 In the IO Matrix, the input sector that receives the price shock is the “fertilizer” sector. It was assumed that 75 percent of the fertilizer sector would be urea (subject to the subsidy shock) based on the national figures of fertilizers’ usage. T he size of the urea subsidy used in the Cost-Push Model was the same used for direct effects (25 percent of the domestic retail price). 112 It was assumed that 5.8 percent of the “Electricity distribution to the agriculture sector” in the IO Matrix would be subject to the subsidy shock based on the national figures of energy usage by the agricultural sector. The size of the subsidy used in the Cost-Push Model was the same used for direct effects (79.4 percent). 113 The average electricity cost for domestic consumers was estimated at PKR 15.12/KW and the actual electricity tariff ranged from PKR 2.00/KW for the first consumption slab (0–50 KW consumed monthly) until PKR 16.00 for the fifth consumption slab (301–700 KW consumed monthly). Source: The World Bank’s Energy team. 114 According to Lustig (2018): Calculating direct effects is sufficient if household energy subsidies are provided for domestic consumption only. 115 The tariff was inclusive of Base Tariff (2018 –19) + FC surcharge + NJ surcharge + Tariff rationalization surcharge + Fuel Price Adjustment (FPA) + GST. The information on the energy price-buildup and tariff structure was received by the World Bank’s Energy team. consumers)116 times the electricity quantity used by the domestic consumers. The analysis of electricity subsidies focuses on the households that pay a tariff below the electricity cost of supply. The price build-up structure and consumption slabs are presented in Annex I. Natural gas subsidy The analysis focused on the government subsidy provided to natural gas consumed by households. For calculating the direct effects of the natural gas subsidy on households’ welfare, the following steps were taken. First, the average subsidy unit (difference between final consumer price and prescribed price) for natural gas was calculated for fiscal year 2018–19, based on OGRA data and differentiating between SSGCL (which provides gas to Sindh and Balochistan) and SNGPL (which provides gas to KP and Punjab). According to OGRA, in 2018–19 the average subsidy unit for natural gas was PKR 47.98 /MMBTU for the SSGCL and PKR 95.62/MMBTU.117 Second, the quantity of natural gas consumed by each household was estimated in the HIES by dividing total households’ expenditure on gas by final consumer price differentiated according to interview month and province (to bifurcate the SSGCL and SNGPL consumers). Third, the direct effect of the gas subsidy was calculated as the subsidy unit multiplied by the quantity of gas consumed by each household, the subsidy amounts were differentiated by province according to the gas provider. It is important to note that natural gas purchased by industries (production input) is not subsidized by the Government. On the contrary, companies pay a higher price relative to the market price to finance the cross-subsidy to households. This means that the higher price of natural gas used as a production input by companies could have a negative indirect effect on households’ welfare. These indirect effects could not be calculated given that this requires a disaggregation by product of the “Petroleum sector” in the IO Matrix. In-kind Health Benefits The in-kind benefits from access to public healthcare were modeled based on the direct identification of beneficiaries in the HIES and imputation of average benefits (measured at the average government cost of providing the health services). The analysis covered inpatient and outpatient public health services. First, the beneficiaries (individuals that accessed inpatient and outpatient services in public health facilities) were identified in the HIES for each province and healthcare service level.118 The model includes “predicted public employees” that accessed private health-care facilities as beneficiaries, since they can benefit from government reimbursements. Second, the average in-kind benefits on public health were calculated for each province and type of health service (inpatient and outpatient). For the intersection of each province and type of health service, the calculation was performed by dividing executed public expenditure in the province/health service (based on administrative data) by the total number of health beneficiaries in the province/health service (observed in the HIES).119 Third, the average benefit by province and health-care service level was imputed to the target public health beneficiaries identified above. 116 In the model, the households that are paying above the electricity cost are not considered as subsidy-recipients. 117 However, gas distribution companies published different subsidy units in their annual reports for fiscal year 2018 –19. SSGCL reported that the subsidy unit they provided was PKR 109.03/MMBTU and SNGPL reported that the subsidy unit they provided was PKR 350/MMBTU. 118 The HIES asked about access to health services in the past three months. 119 Ideally, the number of beneficiaries by province and type of health service (inpatient/outpatient) to calculate the “health transfer per service” would come from the administrative data but this information was not available. In-kind Education Benefits The in-kind benefits from access to public education were modeled based on direct identification of beneficiaries in the HIES and imputation of average benefits (measured at the average government cost of providing these services). The analysis covers pre-primary and primary, secondary, and tertiary education. First, the average in- kind benefits on public education were calculated for each province and education level (pre-primary and primary, secondary, and tertiary) based on official administrative data. For the intersection of each province and education level, the calculation was performed by dividing executed public expenditure by the total number of students enrolled in each province and education level according to official administrative data. Second, the beneficiaries (students currently enrolled in public education) were identified in the HIES for each province and education level. Third, the average benefit by province and education level was imputed to the target public education students identified above. Fourth, for all public students, the out-of-pocket expenditures on registration fees (self-reported in the HIES) were deducted from the value of the public in-kind education benefit.120 The estimation of in-kind education benefits per education level and the comparison of public- students identified in the HIES 2-18-19 with administrative data are presented in Annex I. In the case of Pakistan, both the urea fertilizer and the electricity tubewell subsidy constitute subsidies to agricultural inputs. There are challenges and limitations of modeling agricultural input subsidies based on the HIES 2018–19: • The HIES is a survey representative at the household level. While there are some households that are also farmers (many of them small-scale own producers), the survey is not representative of all farms in Pakistan. Hence, it is likely that the subsidy allocated in the survey is underestimated with respect to the official administrative amount. However, it was not possible to compare survey-estimates of the fiscal incidence model with official numbers due to the lack of administrative data on total government expenditure on these subsidies. • Another challenge is that for the urea fertilizer, since the subsidy is given to the gas used by urea producers, there is part of the subsidy that does not go to neither farmers nor consumers but kept by urea producer firms themselves (the latter not observed in the HIES). • The HIES 2018–19 has data limitations for the modeling of agricultural indirect subsidies, so the current study had to be performed based on several assumptions. For instance, the agricultural module of the HIES had the variable of “total expenditure on fertilizers” so the share allocated to the purchase of local urea had to be estimated based on macro data of fertilizers usage by origin and type. Moreover, for the case of the electricity subsidy given to agriculture tubewells, there was no variable in the HIES to identify such expenditure. The electricity quantity used in agriculture was calculated by dividing households’ expenditure on “Water, electricity and all other fuel charges” by the average monthly price of agriculture electricity in 2018. Hence, since this expenditure item includes other categories beyond agricultural electricity, it could be overestimating the subsidy among thebetween 5.3 Comparison small-scaleSurvey in the survey. Data and Administrative farmers observed After finalizing the fiscal incidence analysis, one step prior to analyzing the results is to compare survey-estimates from the model with actual values from official administrative data on taxes and social expenditure in the year of analysis. Table 5.1. presents this comparison: the first two columns represent the estimates in PKR million 120 For those observations where the net education benefit is negative (because households pay more out of pocket than what they receive from the government), this study has replaced the net education benefit to zero. (from the survey and the administrative accounts, respectively), the third column presents the ratio between the absolute survey estimate (column 1), and the absolute administrative data (column 2). Columns 4 and 5 present the results of columns 1 (survey data) and 2 (administrative data) as a share of their respective total households’ consumption. The main takeaways from the comparison of the survey modelled estimates and fiscal administrative data can be summarized as follows: - Total household consumption: The total household consumption provided in the HIES 2018–19 covers 45.2 percent of the aggregate households’ consumption in the national accounts (NA).121 The CEQ Methodology does not allow for reweighting of household survey consumption to match the NA.122 In contrast, the CEQ Methodology recommends that the comparison of fiscal estimates should be done in relative terms, i.e., as a share of consumption (as in columns 4 and 5). Potential reasons to explain the large underestimation of households’ consumption in the HIES with respect to the NA include: (i) different methodological approaches to measure households’ consumption in a household survey with respect to NA;123 and (ii) household surveys typically have under-coverage of top-income households (which consume more in absolute terms). - Tax coverage: When comparing the ratios of tax estimates to consumption (columns 4 and 5), the effective withholding tax on telecommunications in the current model is 0.1 percent (similar to the admin ratio). At the same time, the effective rates from customs duties and GST124 estimated in the model (2.2 and 4.3 percent, respectively) are close to their respective administrative ratios (2.2 and 4.6 percent, respectively). However, for withholding tax on salaries, the model presents a significant underestimation since the effective rate in the survey was 0.0 percent compared with 0.1 percent in the administrative ratio.125 Two potential reasons behind the underestimation of withholding tax on salaries are that the model focuses on salaries from main occupation and that the HIES does not include income information from top-income households. Lastly, for excise duties and the property tax, the model had a slight overestimation. For excise duties, the effective rate is 1.0 percent, slightly above the administrative ratio (0.7 percent), while for the property tax the effective rate in the survey is 0.1 percent compared with 0.0 percent in the administrative accounts. Property tax collection is slightly overestimated in the model because it is challenging to account for the informality of this tax in the household survey. - Social expenditure coverage: When looking at the third column, which looks at the absolute ratios between survey estimates and administrative values in the official data (in PKR), many of the social expenditure items 121 The aggregate households’ consumption used as the administrative benchmark is defined as “Household and non -profit institutions serving households, final consumption expenditure (current local currency units)”, from World Bank World Development Indicators. 122 Also, this reweighting could affect the official poverty estimates. 123 On the one hand, for NA, private consumption is usually a residual, hence not well estimated. On the other hand, there could be a downward bias stemming from the design of the household survey, since the latter typically covers some (not all) consumption items. 124 Some reasons that could explain why the GST effective rate (GST/consumption) is close in the survey relative to the admin figure (4.3 vs 4.6 percent) despite the fact that this study accounts for households’ consumption informality: (i) The model does not match absolute GST collections, it only gets close to the GST effective rate for the portion of households’ consumption (45 percent) that is observed in the HIES. (ii) The model includes indirect effects (GST paid throughout the production chain) which could capture part of the GST paid by firms. (iii) For the case of products that had a domestic and an imported GST rate, the latter (which is higher) was used assuming the Law of One Price. 125 Also, when comparing the absolute values, the withholding tax on salaries in the model only covered 15.5 percent of the admin value, being this one of the fiscal interventions with the largest estimation gaps. in the model have good coverage of the official data. In effect the major social protection transfers (BISP UCT126 and BISP CCT127), the domestic electricity subsidy, the gas subsidy and the in-kind health and education benefits simulated in the survey all covered above 60 percent of the absolute value of executed expenditure in the administrative accounts. However, the ratios of these social expenditure items relative to consumption was overestimated in the survey with respect to the administrative ratios (given that consumption is underestimated in the HIES). Lastly, it was not possible to macro-validate the urea fertilizer or the agriculture tubewell subsidy due to lack of bifurcated expenditure data on these categories. Taken altogether, while the Pakistan fiscal incidence analysis was able to estimate a total of PKR 1,100 billion in taxes (45 percent of the administrative target), in the case of social expenditures, the survey model was able to allocate PKR 1,277 billion (85 percent of the administrative target). In effect, the ratio of taxes/benefits simulated in the survey is 86 percent (vs. 163 percent in the administrative accounts). The survey model performs better in the allocation of benefits relative to the allocation of taxes. The underestimation of taxes is consistent with the fact that the HIES has under-coverage of total income and total consumption, possibly due to a combination of underreporting by survey respondents and the missing top-income households. In contrast, the benefits are better allocated in the survey model since they are based on the socio-economic characteristics of households and the HIES has better coverage of households at the bottom of the distribution. 126 The absolute expenditure of the BISP UCT estimated in the survey covered 80.5 percent of the absolute admin expenditure. While the survey model matches the number of beneficiary households in the admin accounts, the reason of the gap is that the survey model uses the statutory BISP UCT transfers per beneficiary (not the effective transfers). The statutory transfer is the transfer defined by the operational rules of the program whereas the effective transfer can be calculated by dividing total program’s executed expenditure by actual beneficiaries. 127 The absolute expenditure of the BISP CCT estimated in the survey covered 135.6 percent of the absolute admin expenditure. One reason behind this is that, following operational rules, the survey model does not impose a limit on the number of children that receive the transfer per household. Table 5.1. Pakistan’s fiscal incidence analysis: Survey-estimates vs. fiscal administrative data PKR billion Survey: Admin: Ratio Ratio (A) Value Ratio Federal/ (B) Value over over in Survey (A/B) Provincial in Admin. consum consum (estim.) (%) ption ption (%) (%) Total consumption 14,222.0 31,461.0 45.2 Total direct taxes 18.3 56.3 32.4 0.1 0.2 Withholding tax on salaries (1st occupation) Federal 5.2 33.5 15.5 0.0 0.1 Total property tax (commercial and residential) Provincial 10.0 15.4 64.5 0.1 0.0 Total zakat payment to the Government Federal 3.1 7.4 42.0 0.0 0.0 Federal/ Total indirect taxes provincial 1,081.8 4,489.2 24.1 7.6 14.3 Federal/ Total GST provincial 616.7 1,459.2 42.3 4.3 4.6 Total custom duties Federal 306.0 676.9 45.2 2.2 2.2 Total excises Federal 148.3 2,335.9 6.3 1.0 7.4 Withholding tax on telecommunications 10.9 17.2 63.2 0.1 0.1 Total direct transfers Federal 320.2 1,168.8 27.4 2.3 3.7 BISP UCT transfer Federal 84.2 104.6 80.5 0.6 0.3 BISP CCT transfer Federal 5.4 4.0 135.6 0.0 0.0 Zakat transfer from the Government Federal 6.3 n.a . 0.0 . Urea fertilizer subsidy, direct transfers to farmers (own consumers) Federal 2.4 n.a. . 0.0 . Agriculture tubewell subsidy, direct transfers to farmers (own consumers) Federal 0.2 n.a. . 0.0 . Indirect subsidies 440.6 Urea fertilizer subsidy, indirect effects to consumers Federal 3.4 n.a . 0.0 . Agriculture tubewell subsidy, indirect effect to consumers Federal 4.8 n.a . 0.0 . Total electricity subsidy to final consumers Federal 390.9 421.8 92.7 2.7 1.3 Total gas subsidy Federal 41.4 62.0 66.8 0.3 0.2 Total health transfers, inpatient and Federal/ outpatient (*current) provincial 288.5 289.1 99.8 2.0 0.9 Federal/ Inpatient health transfers provincial 240.6 241.0 99.8 1.7 0.8 Federal/ Outpatient health transfers provincial 47.9 48.0 99.8 0.3 0.2 Federal/ Total education transfers provincial 437.6 468.0 93.5 3.1 1.5 Education transfers, primary Federal/ education provincial 85.0 109.0 78.0 0.6 0.3 Education transfers, secondary Federal/ education provincial 183.1 152.9 119.7 1.3 0.5 Education transfers, tertiary Federal/ education provincial 169.6 206.1 82.3 1.2 0.7 Source: Authors’ elaboration based on HIES estimations and official admin data. Notes: 1/For the indirect subsidies (electricity and natural gas), the admin figures for the macro-validation include the stock of the circular debt and the data source was the World Bank’s Energy team and the Pakistan Economic Survey. 2/ For gas subsidy the admin value of total expenditure has been scaled down by 22 percent (the share that domestic consumers represent in total gas consumption). 6. Main Findings on Fiscal Equity This section presents the main findings from the fiscal incidence analysis in Pakistan for fiscal year 2018 –19. It aims to answer four questions: (i) What is the impact of taxes and social expenditure on poverty and inequality? (ii) Which taxes and social expenditure items are progressive or regressive? (iii) Who bears the burden of taxes and receives the benefits of social expenditures? and (iv) Which households are net payers or net receivers of the fiscal system? The main indicators used in this section are described in Box 6.1. In this section, the main indicators used for the fiscal incidence analysis are the following: • Relative incidence is defined as the total taxes (transfers) that each decile pays (receives) as a share of total income in that decile. It should not be interpreted as the average share of taxes (transfers) relative to households’ income in each decile, since it is based on the total taxes (transfers) paid in each decile. In the case of social protection direct transfers, since relative incidence is based on the aggregate transfers of each decile, it does not allow to distinguish if coverage or generosity is driving the results. • Absolute incidence is calculated as the total taxes (transfers) that each decile pays (receives) as a share of total taxes (transfers) modeled. In the case of social protection direct transfers, since absolute incidence is based on the aggregate transfers of each decile, it does not allow us to distinguish if coverage or generosity is driving the results. • Relative Progressivity is measured by the Kakwani Index (KI), which compares the concentration coefficient of a tax/transfer (with respect to a reference income) to a Gini coefficient measured over the same reference income. A positive KI means that the fiscal intervention is more equally distributed than is the reference income, and a negative KI means that the fiscal intervention is less equally distributed than is the reference income and a KI close to zero indicates that the distribution of the fiscal intervention is approximately as equal as the distribution of the reference income. Results are typically considered significant when the KI is higher than 0.10 in absolute value. The KI measures progressivity in relative terms by comparing concentration coefficients of taxes/transfers relative with the Gini of a reference income. • Marginal contributions. Marginal contributions (MC) summarize the individual impact of a tax or transfer in poverty and inequality reduction when the impact of all other fiscal interventions in the model are held constant. For example, the marginal contribution to inequality reduction is calculated as the difference between the Gini coefficient of the reference income (market income plus pensions in our case) relative to the Gini coefficient of the reference income minus (plus) the specific tax (transfer). The procedure is similar for calculating the marginal contribution to poverty reduction. A positive MC means that the fiscal intervention contributes to poverty or inequality reduction; a negative MC means that the fiscal intervention contributes to an increase in the poverty or inequality indicator. Marginal contributions depend on the size, progressivity, and coverage of fiscal interventions. • Net cash position by decile is calculated by aggregating the relative incidence of taxes and cashable transfers (excluding in-kind benefits) with respect to pre-fiscal income (market income plus pensions). When net cash position is positive (negative) it means that the decile is a net receiver (payer) of the fiscal system • Inequality is measured by the Gini coefficient, which ranges in the scale from 0 to 100: zero means perfect equality and 100 means perfect inequality. For the fiscal incidence analysis, the impacts of taxes and transfers on inequality is measured as the difference in the Gini coefficient between market income plus pensions (pre-fiscal income) and the Gini at final income (post-fiscal income after including all taxes, transfers and in-kind benefits from health and education). • Poverty headcount is measured by the share of population below the national poverty line in Pakistan. The poverty line is different for rural and urban areas. For the fiscal incidence analysis, the impacts of taxes and transfers on poverty is measured as the difference between the poverty headcount at market income plus pensions (pre-fiscal income) and the poverty headcount at consumable income (post-fiscal income excluding in-kind benefits). Poverty headcount is not measured at final income since in-kind health and education benefits are not part of monetary poverty measurement. • Impact effectiveness. The impact effectiveness indicator is a summary of how well a fiscal policy instrument reduces inequality (or the poverty gap) relative to the maximum possible reduction achievable with the same expenditure. Higher numbers indicate greater effectiveness, and one (1) is a "perfect" score that would indicate that the fiscal policy instrument in question is perfectly distributed to reduce pre-fiscal inequality by the maximum possible amount. Negative numbers indicate that the fiscal policy instrument increases the indicator in question. • Spending effectiveness. This indicator presents a summary of how much money a fiscal policy instrument spends relative to how little it could spend to achieve the same impact (on inequality or the poverty gap). Higher numbers are better and one (1) is a “perfect” score that indicates that the fiscal policy instrument in question spends not a single Pakistani rupee more than is necessary to achieve the inequality/poverty gap reduction achieved by that instrument. Indicators less than one (1) show that the fiscal policy instrument is spending more than needed to achieve the same impact (on poverty gap or inequality reduction). Source: Authors’ elaboration. 6.1 Relative Incidence Relative incidence summarizes total taxes and transfers allocated to each decile as a share of total income in each decile. Annex IV presents further details of the different fiscal intervention. • For direct taxes and direct transfers, we measure relative incidence as a share of market income plus pensions. The relative incidence of total direct taxes increases by decile, ranging from 0.0 percent in decile 1 up to 0.3 percent in decile 10. In the case of total direct transfers, their relative incidence ranges from 6.1 percent in decile 1 up to 0.0 percent in decile 10. • For indirect taxes, indirect subsidies, in-kind health, and education benefits, we measure relative incidence as a share of disposable income. Total indirect taxes represent 7.5 percent of households’ disposable income in decile 1, which is slightly higher than the 7.4 percent in decile 10. Total indirect subsidies represent 3.0 percent of households disposable income in decile 1, slightly higher than the 2.2 percent received by decile 10. Total in-kind public education benefits represent 4.7 percent of households’ disposable income in decile 1, higher than the 1.9 percent in decile 10. Lastly, total in-kind public health benefits represent 4.3 percent of households’ disposable income in decile 1, higher than the 0.8 percent received by decile 10. It is important to complement the relative incidence measures with measures of absolute incidence and progressivity. While most of the social expenditure items have a larger relative incidence among the poor, this result is driven by the fact that poor households have lower incomes. Absolute incidence shows that, for some social expenditure items, the share from the total program’s expenditure is larger among the rich, which shows expenditure allocation inefficiencies. To understand which indirect taxes, create the most significant economic burdens on households, the study disaggregated the relative incidence by indirect tax instrument (see graphic below). For excise duties, the results are disaggregated between direct and indirect effects. For GST, the results are disaggregated between GST on goods (direct vs. indirect effects) and GST on services (direct vs. indirect effects). For customs duties, the results represent the total effect (only direct effects). The differentiation between direct and indirect effects is important. On one hand, the direct effects are the (statutory) price increases on taxed final goods. On another hand, the indirect effects capture how sellers increase prices of (untaxed) final goods due to the tax that they paid on inputs; the rationale is that the indirect price increase from input to output happens so that sellers that cannot claim their input tax credit can recover their costs (Inchauste and Jellema 2018). In addition, for GST, the differentiation between GST on goods vs. services is important, since the tax rates are defined at the federal and provincial levels, respectively. The results show that the indirect tax components with the largest relative incidence are the effective custom duties (direct effects) and GST on goods (direct effects and indirect effects). For customs duties, the relative incidence as a share of households’ disposable income declines from 2.6 percent in decile 1 to 1.6 percent in decile 10. For customs duties, the results suggest that the tax burden is regressive in relative terms, since it represents a larger share of households’ disposable income among the poor. For GST on goods (direct effects), the relative incidence goes from 1.2 percent in decile 1 up to 2.0 percent in decile 10. For the indirect effects from GST on goods, the relative incidence as a share of households’ disposable income ranges from 1.9 percent in decile 1 to 1.2 percent in decile 10, which shows that this effect is regressive in relative terms. This finding is very important, since GST indirect effects in the current model have been applied to exempt goods, GST taxable goods without GST input tax credit and informal taxable goods. The first two are determined by the federal GST policy and the third is determined by households’ informal consumption patterns. Hence, the results show that although the federal GST policy could have a rationale for exempting goods and services from the GST (e.g., protecting the poor) or for excluding taxed goods and services from the GST input tax credit (e.g., for tax administration purposes), these policies could still have significant indirect effects (GST cascading) that hurt the poor when they purchase final goods and services that have higher prices because sellers could not claim their input tax credit. At the same time, even though the poor could evade the direct effects from GST of goods and services by purchasing in informal markets, they could still bear the burden from the hidden indirect effects from GST. Relative incidence of indirect taxes (% of disposable income) 10.0% 8.0% 6.0% taxes as a % of disposable income 4.0% 2.0% 0.0% 1 2 3 4 5 6 7 8 9 10 Deciles by disposable income, real, per adult equivalent GST on services (direct effects) GST on services (indirect effects) GST on goods (direct effects) GST on goods (indirect effects) Excises (direct effects) Excises (indirect effects) Total custom duties Withholding tax on telecommunications 6.2 Absolute Incidence The absolute incidence shows the share of total taxes (transfers) that each decile pays (receives) (Figure 6.1). Further details by fiscal intervention are presented in Annex IV. • For total direct taxes, most of the tax collection is paid by the richer deciles. In effect, deciles 9–10 (the richest) pay 87.9 percent of total direct taxes modeled. The share paid by deciles 9–10 is higher for the withholding tax on salaries (88.0 percent of total withholding tax on salaries) and the zakat payment to the Government (96.1 percent of total zakat payments). These taxes represent about 45 percent of total direct taxes simulated in the model. For taxes such as withholding tax on salaries, the fact that richer households pay the largest share of direct tax collection is consistent with the fact that these households are more likely to be formal employees and have higher taxable income. • Similarly, for total indirect taxes, most of the tax collection is paid by the rich since deciles 9 –10 (the richest) pay 40.4 percent, whereas deciles 1–2 (the poorest) pay 9.2 percent of total indirect taxes. Indirect taxes that are paid in higher proportion by deciles 9–10 include the total excises (42.9 percent of total excise duties) and the total GST (43.1 percent of total GST), representing about 71 percent of total indirect taxes modeled. The fact that richer households pay the largest share of indirect taxes is consistent with the fact that these households consume more in absolute terms. • For total direct transfers, most of the distribution of benefits is received by poorer deciles. In effect, deciles 1–2 (the poorest) receive 52.9 percent of total direct transfers. For some interventions such as the BISP UCT and the BISP CCT (which jointly represent 91 percent of total direct transfers128), deciles 1– 2 receive 53.4 and 64.8 percent of each transfer’s expenditure, respectively. Both the BISP UCT and the BISP CCT are targeted via a PMT score. It is important to note that the definition of absolute incidence in the present study is based on total transfers received by each decile and does not allow to distinguish if coverage or generosity is driving the social protection results.129 128 The BISP UCT alone represents 85 percent of total direct transfers modelled. 129 For a review of the performance indicators in the social protection literature, see: Grosh, M. et al. (2022) • For total indirect subsidies, there is evidence of benefits going in a relatively higher proportion to the rich. In effect, deciles 9–10 (the richest) receive about 34.0 percent of total indirect subsidies modeled, about three times larger than the share of benefits received by deciles 1–2 (the poorest). This result is explained by the domestic electricity subsidy (88.8 percent of total indirect subsidies), where deciles 9– 10 receive about 32.8 percent of total benefits, whereas deciles 1–2 receive 9.5 percent. Moreover, for the gas subsidy, deciles 9–10 receive 47.3 percent of total benefits compared with only 5.4 percent received by deciles 1–2. • For total in-kind public education benefits, the richest deciles 9–10 receive a larger share of benefits compared with the poorest 1–2 deciles (29.3 vs. 13.1 percent). This is explained by the fact that tertiary education benefits (the largest one) provide a larger share to the deciles 9–10 (53.0 percent), whereas in contrast pre-primary and primary education benefits provide a larger share to the deciles 1–2 (32.3 percent). Lastly, for total in-kind health benefits, deciles 9–10 receive 27.0 percent of total benefits and deciles 1–2 receive about 17.0 percent, and most of this result is driven by the inpatient health benefits. Figure 6.1. Absolute incidence: share of total taxes/benefits paid/received by each decile 100% 80% 60% 40% 20% 0% Total direct taxes Total direct Total indirect Total indirect Total education Total health in- transfers taxes subsidies in kind benefits kind benefits Deciles 1-2 (poorest) Deciles 3-8 Deciles 9-10 (richest) Equal share to bottom 20% Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: Deciles ranked by market income plus pensions. The gray dashed line is set at 20 percent (how much would be received by the two poorest deciles if the allocation was based on proportional population shares). 6.3 Progressivity of Fiscal Interventions This section assesses the progressivity of taxes and transfers of Pakistan’s fiscal system as of 2018–19. To assess progressivity, the Kakwani Index (KI) is calculated for each fiscal intervention. This is an aggregate indicator of relative progressivity, and it is calculated by comparing the concentration coefficient of a tax/transfer with the Gini of a reference income.130 A positive KI means that a tax/transfer is progressive relative to the concentration 130 The KI for taxes is calculated as the difference between the concentration coefficient of the tax and the Gini of the income concept. The KI for a transfer is calculated as the difference between the Gini of the income concept and the concentration coefficient of the transfer. This allows that in both cases, a positive KI means that a tax or transfer is progressive; a negative KI means that it is regressive, and a zero KI means neutral (Lustig 2018). of a reference income, while a negative KI means that a tax/transfer is regressive and a KI close to zero means it is neutral. The results are typically considered significant when they are above 0.10 in absolute value. The results for the KI are presented in Figures 6.2 to 6.4. The KI of direct taxes and direct transfers is calculated with respect to market income plus pensions. The KI of indirect taxes and indirect subsidies is calculated with respect to disposable income, and the KI for in-kind health and education benefits is calculated with respect to consumable income. The results on progressivity are as follows: • All direct taxes and direct transfers are progressive relative to market income plus pensions. In terms of the size of the KI, the BISP CCT and the BISP UCT transfers (targeted via the PMT) appear to be the most progressive, followed by the zakat distributed by the Government131. In terms of direct taxes, the zakat payment collected by the Government appears to be the most progressive, followed by the withholding tax on salaries and the property tax. • Most indirect taxes (GST and excise duties) are neutral relative to disposable income, while withholding tax on telecommunications and customs duties are regressive. The neutrality of GST in Pakistan is also documented in previous incidence studies from the country, such as Wahid and Wallace (2008) and Jamal and Javed (2013), although these studies did not account for the differential impacts of consumption informality nor include indirect effects. A contribution of the current study is that it accounts for the differential patterns of consumption informality across deciles for both GST and excise duties, following informality estimates from Bachas et al. (2020). In addition, the current study accounts for the indirect effects from taxes on production inputs to the final prices of goods and services consumed by households. The current study finds that, after accounting for informality, both the relative and absolute incidence of GST and excise duties are higher among richer deciles, consistent with the literature on informality. However, GST and excise duties in Pakistan are still neutral, since their concentration coefficients are not large enough to offset the inequality in disposable income. The results with this study contrast with Ara and Asad Khan (2022), given that the latter study analyzed the direct and indirect effects of the main indirect taxes in Pakistan and concluded that the indirect taxes in Pakistan had a regressive pattern (except for the federal excise duty). However, it is likely that the latter study overestimated the indirect taxes’ effective rates in their model, which could affect the regressivity.132 • Indirect subsidies on domestic electricity, agricultural tubewells and urea fertilizer are all progressive relative to disposable income, except for the natural gas subsidy (which is regressive). However, having a progressive subsidy based on the KI does not mean that the policy is ideal, since there are other factors to consider, such as expenditure inefficiencies due to lack of targeting and administrative costs. As noted, based on absolute incidence, 32.8 percent of the domestic electricity subsidy is received by the two richest deciles (three times higher than the share received by the two poorest deciles). Hence, this could be considered as expenditure inefficiency due to lack of targeting.133 Similarly, in the case of the urea fertilizer subsidy, although the subsidy is progressive according to the KI, there are inefficiencies in expenditure allocation, since this study estimates that a significant share of the urea subsidy, being 131 However, the zakat transferred through Government channels could not be macro-validated due to lack of administrative data. 132 Ara and Asad Khan (2022) found that the overall incidence of indirect taxes (custom duties, excises, and GST) in Pakistan for fiscal year 2018–19 was around 20.7 percent of households’ consumption. In contrast, the current study finds that the same indicator was around 7.6 percent, which matches the value of the administrative data (see Table 5.1) 133 This is in line with previous studies in Pakistan that document that most of the domestic electricity subsidy is concentrated among the rich and hence it is an inefficient means of redistribution (Trimble, Yoshida and Saqib 2011). provided through gas as a production input, is kept by the fertilizer producers.134 Lastly, this study finds that the indirect subsidies to domestic consumers of natural gas are regressive relative to disposable income. This is a pattern also found in other countries, since richer households are more likely to benefit from direct fuel consumption (e.g. owning personal transport or using cooking stoves/heater based on gas).135 • In-kind health and education benefits are all progressive relative to consumable income, except for tertiary education benefits, which are regressive. Like Asghar and Zahra (2012), this study also finds that primary education benefits are the most progressive and pro-poor. This is consistent with poor households having more children, so they are more likely to benefit from access to public primary education. Meanwhile, the regressivity of tertiary education is consistent with the literature and the country context, since poorer children have less access to higher levels of education, as they have a higher opportunity cost that pushes them out of school at older ages. Lastly, the fact that outpatient benefits are more progressive than inpatient benefits is consistent with the fact that the former benefits are more concentrated among the poor relative to the latter. Although there is widespread opting out to private services in Pakistan, poor households still have a higher probability of accessing public health- care facilities relative to richer households, and the latter have a higher probability of opting out to private health-care facilities (World Bank 2020a).136 134 Authors’ calculations. See Section 5.2. 135 For gas subsidy, it is regressive because nationally only 40.0 percent of the households have access to gas. There are inequalities in access to gas based on socio-economic characteristics: 17.2 percent of the poorest quintile has access to gas, compared to 63.8 percent for the richest quintile. Source: Own calculations based on the HIES 2018 –19. 136 Based on own calculations from HIES 2018–19: The probability of accessing a private doctor clinic is relatively higher for the richest decile relative to the poorest decile (66.5 vs 54.9 percent). Similarly, the probability of accessing a government hospital is relatively higher in the poorest decile relative to the richest (21.8 vs. 11.4 percent). Progressivity of selected fiscal policy instruments, Pakistan 2018-19 Figure 6.2. Direct taxes and direct transfers (with respect Figure 6.3. Indirect subsidies and indirect to market income plus pensions) taxes (with respect to disposable income) -0.12 Natural Gas subsidy 0.12 Tubewell subsidy (direct… 0.07 Electricity subsidy… 0.20 Urea subsidy (direct… 0.21 Tubewell subsidy… 0.80 Zakat Govt. transfer 0.23 Urea subsidy… 0.91 BISP CCT transfer 0.06 Total indirect subsidies 0.81 BISP UCT transfer -0.25 Withholding tax on… 0.80 Total direct transfers 0.02 Excises 0.62 Zakat payment to Govt 0.50 Property tax 0.03 GST 0.55 Withholding tax on… -0.07 Custom duties 0.53 Total direct taxes -0.01 Total indirect taxes 0.00 0.50 1.00 -0.50 0.00 0.50 1.00 1.50 Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Figure 6.4. Progressivity of in-kind health and education benefits, based on the KI (with respect to consumable income) -0.19 Tertiary edu. 0.24 Secondary edu. 0.57 Pre-primary and primary edu. 0.15 Total education benefits 0.19 Inpatient health 0.27 Outpatient health 0.20 Total health benefits -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. 6.4 Marginal Contributions In a fiscal system with multiple taxes and transfers, assessing progressivity alone is not sufficient to determine if a fiscal intervention has a positive redistributive effect. To assess which fiscal interventions have a greater individual impact on poverty and inequality reduction, the study estimates their marginal contributions. The marginal contribution (MC) is defined as the difference in the poverty or inequality indicator with and without the fiscal intervention, measured for a specific income concept.137 A positive MC means that the specific fiscal intervention reduces poverty or inequality, and a negative MC means that it increases poverty or inequality. The MC is a function of three characteristics: (i) how progressive a tax/transfer is; (ii) how large a tax/transfer is relative to households’ income; and (ii) how extensive is the coverage of the tax/transfer in the population. The interaction of these three characteristics determines the final impact on poverty and inequality. By calculating the MC, one can respond to a key policy question: what would the inequality (poverty) level be if the fiscal system did not have a particular tax (or transfer)? (Inchauste and Lustig 2017; Lustig 2018). Table 6.1 presents the marginal contributions of each fiscal intervention to inequality reduction (column 4), results are presented with respect to consumable income and with respect to final income. Most fiscal interventions had small marginal contributions meaning that they left inequality largely unchanged. When looking at consumable income (market income plus pensions minus all taxes plus direct transfers plus indirect subsidies), the estimations show that the BISP UCT transfer is the one that has the highest contribution to inequality reduction (MC = 0.5 of a Gini point). In addition, when looking at final income (consumable income plus in-kind benefits from health and education), the intervention with the second-largest contribution to inequality reduction is pre-primary and primary education benefits (MC = 0.3 of a Gini point). Both the BISP UCT and the pre-primary and primary education benefits are progressive when looking at their Kakwani Indexes; however, their marginal contribution to inequality reduction is limited by their small size (0.7 and 0.6 percent of households’ market income plus pensions). On another hand, the two interventions with the largest contributions to inequality increase are the inpatient health benefits (MC = -0.5 of a Gini point) and tertiary education benefits (MC = -0.4 of a Gini point). Hence, these interventions offset part of the inequality reduction stemming from the BISP UCT transfer and the pre-primary and primary education benefits. It is noteworthy is that in the case of Pakistan most of the inequality reduction comes from the BISP UCT transfer, whereas in other countries with fiscal incidence studies under the CEQ Methodology it has been found that in-kind health and education benefits are the major driver of inequality reduction.138 Two fiscal facts explain this pattern in Pakistan: (i) public expenditure levels on health and education are low compared with other countries; (ii) most of the public expenditure on health and education goes to inpatient services and tertiary education, respectively , both benefiting mostly richer households according to absolute incidence measures.139 The Zakat transfer managed by government is progressive relative to the prefiscal income distribution (Figure 6.2) while its marginal contribution to inequality reduction overall (at postfiscal income) is relatively small (Table 6.1). The marginal contribution of a fiscal instrument depends on both the progressivity of the instrument in the presence of all other fiscal instruments (i.e., at postfiscal income) as well as its relative magnitude, and the Zakat 137 For the calculation of the marginal contribution to inequality reduction, the following formula is used: Gini of market income plus pensions- Gini of market income plus pensions including the specific tax or transfer (minus for tax and plus for transfer). A similar formula is used for the calculation of marginal contributions to poverty reduction, based on the poverty difference without and with the tax or transfer. 138 See: Botswana (2010); Lesotho (2017); Eswatini (2017); Namibia (2015); and South Africa (2014 –15). Source: CEQ Institute. 139 There is widespread evidence of households (including the poor) opting out to private providers of health and education services in Pakistan (World Bank 2020a). If the opt-out rate is higher among richer households, this would improve the marginal impact (on inequality) of the publicly provided service in question. government transfer is relatively small in magnitude compared to other fiscal items considered here (Table 6.1).140 Table 6.1. also presents the results of marginal contributions to the national poverty headcount reduction by fiscal intervention (column 5). When looking at consumable income, the estimations show that the fiscal interventions with the largest marginal contributions to the national poverty headcount increase are the GST (MC = -3.3 percentage points) and customs duties (MC = -2.1 percentage points). This shows how the marginal contributions depend both on the size, progressivity, and coverage of taxes/transfers. For instance, the poverty increase is mostly driven by GST (neutral intervention) because this tax covers a large share of the population and has a high burden on households’ market income plus pensions (4.4 percent); and the custom duties represent a smaller burden (2.2 percent of market income plus pension), but they are mildly regressive. Meanwhile, the domestic electricity subsidy and the BISP UCT both had positive marginal contributions to poverty reduction (MC 2.9 and 1.2 percentage points, respectively), which is consistent with the fact that all subsidies and transfers increase households’ income. However, the poverty reduction from these interventions was not large enough to offset the poverty increase stemming from indirect taxes (see Section 6.5 and 6.6). The domestic electricity subsidy has a larger MC to poverty reduction than the BISP UCT, because it represents a larger share of households’ market income plus pension (2.8 percent vs. 0.7 percent), partly because the electricity subsidy has a higher coverage of the overall population for being a generalized subsidy141. Nonetheless, the domestic electricity subsidy also has significant leakage of resources to non-poor households for being a generalized subsidy, whereas the BISP UCT has less expenditure inefficiency given that the latter is targeted via the PMT score.142 Lastly, Table 6.1 presents the results of marginal contributions to the national poverty gap reduction by fiscal intervention (column 6), which are qualitatively similar to the marginal contributions to the national poverty headcount. In effect, the fiscal interventions with the largest marginal contributions to the national poverty gap increase are GST (MC = -0.8 of a percentage point) and the customs duties (MC = -0.5 of a percentage point). Similarly, the fiscal interventions that contributes the most to the poverty gap reduction are the domestic electricity subsidy (MC=0.7 of a percentage point) and the BISP UCT (MC = 0.6 of a percentage point), yet these effects are not large enough to offset the increase in the poverty gap caused by indirect taxes (see Section 6.5 and 6.6). 140 Re-ranking of households from prefiscal to postfiscal income also reduces the marginal impact of the fiscal intervention in question (Enami, 2018 and Enami et al. 2021). Reranking happens when two individuals at, for example, the 51 st and 53rd rank of the income distribution swap places after the execution of fiscal policy, so that the individual who was previously in the 51st position moves to the 53rd position while the individual who was previously in the 53rd position moves to the 51st position. Reranking could happen if, for example, only one of two similarly situated families is eligible for a social transfer while both families face the same effective indirect tax rates on consumption expenditure. In this scenario, a redistribution of resources from the 53rd-ranked household (at prefiscal income) to the 51 st-ranked household (at prefiscal income) did not reduce income inequality. 141 As explained in Box 6.1, since the size or relative incidence is calculated as total transfers/total income, results reflect both coverage and generosity (it is not possible to determine which one of both is driving the results). 142 The simulations in this study show that the domestic electricity subsidy has a higher coverage of households in the three poorest deciles relative to the BISP UCT (75 vs 29 percent, on average). However, at the same time since the domestic electricity subsidy is generalized it also covers about 93 percent of households in deciles 4 –10 whereas the BISP UCT (targeted via the PMT) only covers on average 7 percent of households in deciles 4 –10; this shows the importance of targeting for reducing the inefficiency in the allocation of public resources. Source: Own calculations based on the HIES 2018–19. Table 6.1. Marginal contributions to inequality and poverty reduction Size wrt To To national To national Market inequality poverty Concentration Kakwani poverty gap income reduction headcount coefficient Index reduction plus (Gini reduction (p.p.) pensions points) (p.p.) Total from Market Income plus pensions to Consumable Income Direct Taxes -0.1% 0.80 0.52 0.06 0.00 0.00 Withholding tax on salaries 0.0% 0.83 0.55 0.02 0.00 0.00 Property tax -0.1% 0.75 0.47 0.03 0.00 0.00 Zakat payment to Govt 0.0% 0.90 0.62 0.01 0.00 0.00 Direct Transfers 0.8% -0.37 0.66 0.58 1.25 0.72 BISP UCT transfer 0.7% -0.44 0.73 0.53 1.16 0.63 BISP CCT transfer 0.0% -0.53 0.82 0.04 0.10 0.05 Zakat Government transfer 0.0% 0.61 -0.33 0.01 0.02 0.03 Subsidy urea (direct transfer to farmers) 0.0% 0.10 0.18 0.00 0.03 0.00 Subsidy tubewells (direct transfer to farmers) 0.0% 0.18 0.11 0.00 0.00 0.00 Indirect Taxes -7.7% 0.27 -0.01 -0.03 -6.26 -1.43 GST -4.4% 0.31 0.03 0.14 -3.34 -0.79 Custom duties -2.2% 0.21 -0.07 -0.16 -2.12 -0.53 Excises -1.0% 0.27 -0.02 0.00 -0.98 -0.23 Withholding tax on telecommunications -0.1% 0.03 -0.25 -0.02 -0.10 -0.03 Indirect Subsidies 3.2% 0.24 0.04 0.17 3.13 0.74 Subsidy urea (indirect effect to consumers) 0.0% 0.05 0.23 0.01 0.04 0.01 Subsidy tubewells (indirect effect to consumers) 0.0% 0.07 0.21 0.01 0.05 0.01 Total electricity subsidy 2.8% 0.23 0.06 0.19 2.90 0.68 Natural Gas Subsidy 0.3% 0.41 -0.13 -0.04 0.18 0.04 Total from Market Income plus pensions to Final Income 0.00 0.00 0.00 Direct Taxes -0.1% 0.79 0.51 0.06 -0.01 0.00 Direct Transfers 0.8% -0.35 0.63 0.53 1.37 0.63 Indirect Taxes -7.7% 0.27 -0.02 -0.10 -5.51 -1.19 Indirect Subsidies 3.2% 0.23 0.06 0.21 2.83 0.63 In-kind benefits 5.3% 48.22 -0.20 -0.36 n/a n/a Total Health benefits 2.1% 69.22 -0.41 -0.45 n/a n/a Inpatient services 1.7% 81.72 -0.53 -0.53 n/a n/a Outpatient services 0.3% 6.95 0.22 0.08 n/a n/a Total Education benefits 3.2% 34.69 -0.06 0.12 n/a n/a Pre-primary and Primary education 0.6% -7.17 0.36 0.28 n/a n/a Secondary education 1.4% 23.63 0.05 0.17 n/a n/a Tertiary education 1.2% 69.75 -0.41 -0.38 n/a n/a Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: Total education in-kind benefits – including regressive tertiary education benefits - are approximately neutral with respect to final income (KI=-0.06) while exhibiting a marginal contribution to inequality reduction of 0.12 Gini points or approximately zero. 6.5 Net Payers and Net Receivers To examine how the combination of taxes and social expenditure items modeled in Pakistan affects households across the income distribution in fiscal year 2018–19, the study calculates the net cash position across different groups of the welfare distribution (Figure 6.8). The analysis divides the population into 10 deciles ranked by market income plus pensions (pre-fiscal income).143 For each decile, the stacked bars show the incidence of the fiscal intervention with respect to market income plus pensions. All the fiscal interventions that represent an income gain to the household are above the zero axis (direct transfers, indirect subsidies and in-kind education and health benefits), while all the fiscal interventions that represent an income loss for the household are below the zero axis (direct and indirect taxes). The net cash position (red dashed line) shows the aggregate sum of all cash-based interventions (all taxes, direct transfers, and indirect subsidies) for each decile. The total fiscal position (blue dashed line) includes all cash-based interventions plus in-kind benefits (such as education and health benefits) valued at the government’s cost of provision. According to the net cash position indicator (red dashed line), only the poorest decile of the population is expected to be net cash recipients of the fiscal system, with a cash gain equivalent to approximately 1.2 percent of market income plus pensions. In contrast, the rest of the households are net payers of the fiscal system, with the cash loss ranging from -1.8 percent in decile 2 to -5.5 percent in decile 10 (the richest). The fact that deciles 2 and 3 are net payers of the fiscal system is consistent with the national poverty increase (Figure 6.9) while the fact that the net cash loss is relatively higher in richer deciles is consistent with the slight inequality reduction (Figures 6.10). 143 The size of the effects represents the overall picture by decile, but there could be further heterogeneity of the effects within deciles (Lustig 2018, p. 36) When looking at the total fiscal position of households (blue dashed line), which includes the in-kind benefits from health and education, the results show that deciles 1–9 are net receivers, while only decile 10 is a net payer of the fiscal system. The net cash position provides a better measure of households’ purchasing power and therefore may be a better income construct when estimating the impacts of fiscal poverty on monetary poverty. Another way to see this is that the net cash position allows us to tell whether the Government has enabled an individual to be able to purchase private goods and services above his/her original market income (Lustig 2018). Figure 6.8. Net cash position of households after taxes and transfers 25.0% Tax or transfer as a % of market income plus 15.0% 5.0% pensions -5.0% Poorest 2 3 4 5 6 7 8 9 Richest -15.0% Deciles by market income plus pensions, real, per adult equivalent Total in-kind education benefits Total in-kind health benefits Total indirect subsidies Total indirect taxes Total direct transfers Total direct taxes NET CASH POSITION (market income-taxes+ cashable transfers) TOTAL POSITION (Net cash position+ in-kind benefits) Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. 6.6 Overall Impacts on Inequality and Poverty Impacts on inequality This subsection looks at how inequality – as measured by the Gini coefficient index – changes as we account for the impacts of fiscal policies on household incomes. The Gini coefficient index ranges from zero (perfect equality) to 100 (perfect inequality). The Gini coefficient at disposable income (28.4) is similar to the official estimate in Pakistan 2018-19 since this study’s definition of disposable income is equivalent to the official consumption aggregate estimated with the HIES. The rest of the CEQ income concepts are calculated based on disposable income (consumption) minus/plus the relevant taxes/transfers (see Section 5.2). The estimates show that the combination of taxes and social expenditure items modeled in Pakistan leaves inequality unchanged in fiscal year 2018–19 (Figure 6.9): the Gini coefficient falls by about 0.4 points from when going from market income plus pensions (prefiscal income) to final income (postfiscal income). Most of the inequality reduction occurs from net market income to disposable income suggesting that social protection direct transfers are driving the slight inequality reduction (see also Section 6.4). It is important to note that the small change in the Gini coefficient is unlikely to be significant compared with other countries.144 While the CEQ Methodology focuses on point estimates, standard errors from survey-based estimations could be large.145 Another source of bias is that household surveys do not capture well “top incomes” which could bias estimates of the impact of fiscal policy on inequality reduction. Figure 6.9. Inequality impacts of the taxes + social expenditure items modeled (measured by the Gini coefficient) 35.2 29.0 29.0 28.4 28.3 28.6 30.2 Gini coefficient (0-100) 25.2 20.2 15.2 10.2 5.2 0.2 Net market income Market income plus pensions Disposable income Consumable income Final Income Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. (1) Disposable income=Official consumption aggregate. Net market income=Disposable income- Direct transfers. Market income plus pensions (pre-fiscal income) = Net market income plus Direct Taxes. Consumable income= Disposable income – indirect taxes + indirect subsidies. Final income= Consumable income + in kind benefits from public health and public education. Impacts on poverty The study also assesses how the national poverty headcount changes across various stages of the fiscal redistribution. For this, it calculates what percentage of the total population falls below the national poverty line for each income concept of the fiscal incidence analysis. The estimations are based on the 2018 national poverty line (monthly real consumption, per adult equivalent),146 meaning that every individual with a welfare level below this threshold is considered poor. The poverty headcount at disposable income (21.9 percent) is equal to the official poverty headcount in Pakistan since this study’s definition of disposable income is equivalent to consumption. The rest of the CEQ income concepts are calculated based on disposable income (consumption) minus/plus the relevant taxes/transfers (see Section 5.2). First, Figure 6.10 presents the poverty estimations for different income concepts. The results show that the combination of taxes and social expenditure items modeled in Pakistan increases the national poverty headcount in 2018–19 by 2.3 percentage points (from 23.3 percent to 25.5 percent) from market income plus pensions (prefiscal income) to consumable income (postfiscal income). Most of the poverty increase (+3.6 percentage points) takes place between disposable income and consumable income, suggesting that indirect taxes are driving the poverty increase, which is confirmed in the section on marginal contributions (see Section 6.4). In addition, when looking at the fiscal impoverishment indicator, which measures what percentage of individuals become impoverished after paying taxes and receiving direct transfers and indirect subsidies (comparing market income plus pensions vs. consumable income), the estimations show that: (i) 17 percent of 144 For a list of 67 countries where the CEQ methodology has been applied, the average change in the Gini coefficients (from Gini market income plus pensions to final income) is -8.5 Gini points. Source: CEQ Standard Indicators 145 For instance, standard errors could be large given the uncertainty stemming from survey data and estimations. 146 Poverty headcount measured at national poverty line (monthly real consumption per adult equivalent) of PKR 3,768.457 (urban) and PKR 3,741.232 (rural). the total population became poor when going from market income plus pensions to consumable income; and (ii) 68 percent of the consumable income poor a were fiscally impoverished; that is, they paid more taxes into the fiscal system than they received from it in direct transfers and indirect subsidies. Figure 6.10. Poverty headcount impacts of the taxes + social expenditure items modeled, total national (based on national poverty line) 26.0 25.5 25.0 24.0 23.3 23.3 23.0 21.9 22.0 21.0 20.0 Market Income plus pensions Net Market Income Disposable Income Consumable Income Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: ((1) Disposable income=Official consumption aggregate. Net market income=Disposable income- Direct transfers. Market income plus pensions (pre-fiscal income) = Net market income plus Direct Taxes. Consumable income= Disposable income – indirect taxes + indirect subsidies. (2) Poverty headcount measured at national poverty line (monthly consumption per adult equivalent) of PKR 3,768.457 (urban) and PKR 3,741.232 (rural). (3) Poverty headcount is not measured for final income, because final income includes in-kind benefits (health and education), and the analysis only focuses on cashable interventions. Second, the study disaggregated the poverty headcount impacts by region to assess the differential impacts of the combination of taxes and social expenditure items by geography (Figure 6.11). To analyze this, the poverty headcounts are compared between market income plus pensions and consumable income for the urban and rural regions, respectively. The results show that: (i) the poverty headcount at market income plus pensions (pre-fiscal income) is 2.7 times larger in rural areas relative to urban areas; and (ii) the increase in the poverty headcount due to the combination of taxes and social expenditure items modeled is higher in rural areas (+2.7 percentage points) relative to urban areas (+1.4 percentage points). Figure 6.11. Poverty headcount impacts of the taxes + social expenditure items modeled, by region (based on the national poverty line) Market income plus pensions 32.9 Consumable income 35.0 30.1 30.0 25.0 20.0 11.4 12.7 15.0 10.0 5.0 - rural urban Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. (1) Disposable income=Official consumption aggregate. Net market income=Disposable income- Direct transfers. Market income plus pensions (pre-fiscal income) = Net market income plus Direct Taxes. Consumable income= Disposable income – indirect taxes + indirect subsidies. Similarly, the study disaggregated the poverty headcount impacts by province (Figure 6.12) by comparing the poverty headcounts between market income plus pensions and consumable income for each province. The results show that the increase in the poverty headcount due to the combination of taxes and social expenditure items modeled is the highest in Balochistan (+6.7 percentage points) and the lowest in Islamabad Capital Territory (+0.7 percentage points). Figure 6.12. Poverty headcount impacts of the taxes + social expenditure items modeled, by province (based on the national poverty line) 60.0 50.0 40.0 30.0 20.0 10.0 - KP Punjab Sindh Balochistan Islamabad Market income plus pensions Consumable income Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: (1) Disposable income=Official consumption aggregate. Net market income=Disposable income- Direct transfers. Market income plus pensions (pre-fiscal income) = Net market income plus Direct Taxes. Consumable income= Disposable income – indirect taxes + indirect subsidies. Third, the study calculated the impacts on the poverty gap at the national level by income concept (Figure 6.13). The results show that the poverty gap (average distance to the poverty line in percentage points) increases due to the combination of taxes and social expenditure items modeled. In effect, the poverty gap increases by 0.3 of a percentage point, from 4.4 percent at market income plus pensions to 4.6 percent at consumable income. Figure 6.13. Poverty gap impacts of the taxes + social expenditure items modeled, total national (based on the national poverty line) 5.0 4.4 4.4 4.6 the poverty line (% of 3.7 average distance to 4.0 the poverty line) 3.0 2.0 1.0 0.0 Net Market Income Market Income plus pensions Disposable Income Consumable Income Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. (1) Disposable income=Official consumption aggregate. Net market income=Disposable income- Direct transfers. Market income plus pensions (pre-fiscal income) = Net market income plus Direct Taxes. Consumable income= Disposable income – indirect taxes + indirect subsidies. 6.7 Impact Effectiveness and Spending Effectiveness Impact effectiveness on inequality This study calculated the CEQ Institute’s Impact Effectiveness (IE) indicators (Enami 2018). The IE indicator demonstrates how well fiscal policy instruments reduce inequality relative to the maximum possible reduction achievable with the instrument in question (with the expenditure or revenue collection level held constant). A positive (negative) IE score shows that a fiscal instrument reduces (increases) inequality. Higher (lower) IE scores are better and one (1) represents a “perfect score”, which means the instrument could not be distributed any more efficiently to reduce inequality (given the instrument’s fiscal magnitude and given the underlying characteristics of the population in Pakistan). Another way to interpret IE indicators is that they show how well fiscal instruments are targeted relative to a benchmark in which taxes or transfers are perfectly targeted to achieve maximum inequality reduction. IE estimates for Pakistan fiscal policies are presented in Figure 6.14. The results show that the personal income tax (PIT) is the most effective instrument for reducing inequality.147 An IE score of 75 in urban areas (69 in rural areas) indicates that PIT reaches 75 percent (69 percent) of the maximum impact on inequality achievable with this instrument in urban (rural) areas in Pakistan. In contrast, GST and all other indirect taxes reach between 3 and -22 percent in rural areas (between 2 and -48 percent in urban areas), which means that indirect taxes just as often increase inequality as reduce it. The telecommunications tax – which accounts for less than 1 percent of all revenues from taxes in fiscal year 2019 – contributes to increasing inequality in both urban and rural areas. On the expenditure side, social protection transfers (BISP) constitute the most effective instrument to reduce inequality available to the Government, reaching between 52 and 68 percent of a maximum possible inequality reduction achievable given BISP’s fiscal magnitude. In-kind public primary- and secondary-level education benefits (non-tertiary) are also effective in reducing inequality, and more so in urban than in rural areas. The distribution of tertiary-level education and general healthcare service benefits increases inequality at final income, and more so in rural than in urban areas where such services are less frequently available and less frequently accessed. Energy subsidies are not effective in redistributing resources. This is predictable given what we know about the absolute incidence of energy subsidies captured. The spending on agricultural subsidies in urban areas is the most effective subsidy instrument for reducing inequality in Pakistan, while in rural areas the agricultural subsidy expenditure achieves very little inequality reduction. 147 This study shows an IE score for Zakat payments collected and received by the government; however, these payments do not contribute to general government expenditures, so they are not included in the ranking of general-purpose revenue- generation instruments. Figure 6.14. Impact Effectiveness (IE) of taxes and transfers, at final income Total benefits -1% 11% Healthcare -30% -9% Tertiary education -34% -15% Primary, secondary education 14% 32% Agri. input subsidies 5% 16% 5% Urban Natural Gas subsidies -9% Elec subsidies 2% 14% Rural All subsidies 2% 13% Agri. inputs-direct transfer 1% 16% Zakat transfer -12% 29% BISP (UCT+CCT) 52% 68% Total taxes -2% -1% Telecomms -48% -22% Customs -15%-9% Excise -2% 0% GST 2% 3% Property tax 0% 45% Zakat payments 84% 82% PIT 75% 69% -0.60 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 1.00 Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: 1/The impact effectiveness indicator is a summary of how well a fiscal policy instrument reduces inequality (or poverty headcount/poverty gap) relative to the maximum possible reduction achievable with the same expenditure. 2/Impact effectiveness in this figure is measured at consumable income. Poverty gap effectiveness This study generated estimates of the CEQ Institute’s Poverty Gap Effectiveness (PGE) indicator as well (Enami 2018). Similar in concept to the IE indicator, the PGE provides an estimate of the portion of an expenditure policy’s expenditure that went to reducing the poverty gap (also known as the “depth” of poverty). For taxes, the PGE indicator is defined as the share of revenues collected that did not increase the depth of poverty (measured by the poverty gap). For either expenditures or taxes, higher PGE scores indicate a more effective policy (vis-à-vis the poverty gap). A perfect PGE score of one (1) indicates that the policy in questions uses all its expenditure to reduce the poverty depth or the tax in question collects all its revenues without increasing the depth of poverty. PGE estimates shown in Figure 6.15 demonstrate that the BISP is the most effective program at reducing the poverty gap in either urban or rural areas. For reducing the poverty gap, agricultural input, power, and energy subsidies are relatively ineffective. On the tax side, direct taxes are far more effective at raising revenues without causing an increase in the depth of poverty. In other words, direct taxes are far better at protecting poor and vulnerable households’ expenditures. GST and other indirect taxes are not particularly good at protecting rural poor household expenditures. If energy and agricultural subsidies were eliminated without a compensating transfer, revenues collected from indirect tax instruments would create even larger burdens for poor and vulnerable households. The fiscal system is far less effective at providing transfers to poor households in urban areas than in rural areas. At the same time, indirect taxes cause less impoverishment in urban areas. That leaves the fiscal system equally effective in urban and rural areas at shielding poor and vulnerable households from fiscal impoverishment: Figure 6.15. Poverty Gap Effectiveness (PGE) of taxes and transfers, at consumable income Total benefits 11% 32% Agri. input subsidies 10% 21% Natural Gas subsidies 6% 15% Elec subsidies 8% Urban 20% All subsidies 8% Rural 20% Agri. inputs-direct transfer 10% 16% Zakat transfer 10% 42% BISP (UCT+CCT) 51% 62% Total taxes 88% 73% Telecomms 94% 78% Customs 95% 83% Excise 94% 81% GST 95% 83% Property tax 99% 0% Zakat payments 100% 100% PIT 100% 98% 0.00 0.20 0.40 0.60 0.80 1.00 1.20 Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: 1/The PGE effectiveness indicator shows what portion of expenditure contributed to poverty gap/depth reduction; for taxes, PGE shows the share of revenues that did not increase the poverty gap/depth. 2/PGE in this figure is measured at consumable income. 7. International Comparison 7.1 Pakistan and Peer Countries This section contextualizes the findings from the fiscal incidence analysis in Pakistan by comparing them with seven other middle-income countries148 where similar fiscal incidence studies have been applied under the CEQ Methodology: Brazil (Higgins and Pereira, 2013); Egypt (Lara Ibarra et al. 2019); Indonesia (Jellema et al. 2017); Jordan (Alam et al. 2017); Mexico (Scott et al., 2020); Sri Lanka (Arunatilake et al. 2019); and Türkiye (Cuevas, P. Facundo et al. 2020). For international comparability, the inequality is measured using income per capita and the poverty headcount impacts are calculated based on the international poverty line of US$3.20 PPP 2011 per day; the inequality figures are available for all comparator countries whereas the international poverty figures at US$3.20 PPP 2011 are only available for five comparator countries including Pakistan. The international comparison includes regional peers (Sri Lanka), structural and aspirational peers (Türkiye, Brazil, and Mexico). The comparison of findings across countries on the impact of taxes and transfers on poverty and inequality can be summarized as follows: • Impact on inequality: Pakistan is the country in the sample with the lowest inequality reduction. In all comparator countries, the combination of taxes, transfers and in-kind benefits from health and education reduces inequality measured by the Gini coefficient. The largest reduction in inequality is estimated for Brazil (-14.3 Gini points) and the average inequality reduction in the sample is -6.4 Gini points (Figure 7.1). • Impact on poverty: At the same time, Pakistan exhibits the largest poverty increase in the sample. In two out of five countries (Pakistan and Brazil) the fiscal system increases the poverty headcount (measured at the US$3.2 PPP 2011 lower middle-income class line) after accounting for taxes and transfers that affects households’ cash position (from prefiscal income to consumable income). The poverty increase in Pakistan (+2.9 percentage points) is the largest, followed by Brazil (+0.8 of a percentage point). In the rest of the countries with international poverty figures available, the fiscal system reduces the poverty headcount: Mexico (-2.6 percentage points), Türkiye (-0.2 of a percentage point), and Sri Lanka (-0.2 of a percentage point) (Figure 7.2). The comparison of findings across CEQs from different countries comes with several caveats: (i) different countries could have different coverage of taxes and social expenditure in their models; and (ii) although all countries follow the CEQ framework for defining income concepts and calculating indicators, the methodologies used for the simulation of different fiscal interventions can vary. For instance, the fiscal incidence analysis in Pakistan includes the indirect effects of indirect taxes, which can increase the impoverishment effects of those taxes modeled. From the list of comparator countries only Brazil and Indonesia include indirect effects. One coincidence is that Brazil is one of the countries in the sample that also experiences a poverty increase due to the combination of taxes and social expenditure items modeled along with Pakistan. 148 Middle-income countries according to the World Bank income classification. Figure 7.1. Peer countries: Impact of taxes and Figure 7.2. Peer countries: Impact of taxes and transfers on inequality (by Gini coefficient) transfers on poverty (at US$3.20 per day PPP 2011 lower middle-income class line) 45.0 ($3.2 intl poverty line) poverty headcount 40.0 0.55 35.0 Gini coefficient (0-100) 30.0 0.45 25.0 20.0 15.0 0.35 10.0 5.0 0.0 0.25 Market Disposable Consumable Market Disposable Consumable Final income income plus income income income plus income income pensions pensions Brazil (2009) Egypt (2015) Mexico (2014) Sri Lanka (2009) Turkey (2014) Pakistan (2018) Indonesia (2012) Jordan (2010) Brazil (2009) Mexico (2014) Sri Lanka (2009) Turkey (2014) Pakistan (2018) Source: World Bank’s elaboration, based on authors’ calculations for Pakistan 2018 and secondary studies for: Brazil (Higgins and Pereira 2013); Egypt (Lara Ibarra et al. 2019); Indonesia (Jellema et al. 2017); Jordan (Alam et al. 2017); Mexico (Scott et al. 2020); Sri Lanka (Arunatilake et al. 2019); and Türkiye (Cuevas, P. Facundo et al. 2020). Note: For the calculation of international poverty in Pakistan, the income concepts are per capita, and the US$3.20 PPP 2011 poverty line is calculated at PKR 3,850.75 monthly. For the calculation of the international Gini in Pakistan, the income concepts are converted to real per capita. 7.1 Note on International Comparison The redistributive effects of countries’ fiscal systems depend on three characteristics: (i) the overall magnitude of taxes and social expenditures; (ii) the composition of taxes and social expenditures; and (iii) the allocation rules of taxes and social expenditures: • Fiscal systems that are inequality-reducing and poverty-increasing are also found in other developing countries (Beegle and Luc 2019). This study finds that Pakistan has no impact on inequality while it does increase poverty. The fiscal impoverishment and poverty increase is a typical result found in countries that have large indirect taxes (which everyone must pay) combined with limited social protection systems with low coverage of poor households. • Pakistan’s fiscal system has three characteristics that limit its effectiveness in reducing poverty and inequality: (i) low levels of taxation and social expenditure; (ii) large reliance on indirect taxes; and (iii) small size of progressive fiscal interventions. In effect, from the list of eight comparator countries, Pakistan’s general government tax revenue-to-GDP ratio (13.4 percent) is the second lowest and its social expenditure-to-GDP ratio (5.0 percent) is the third lowest (Figures 7.3 and 7.4). The most effective and progressive instruments on the tax and transfer side (the direct taxes and BISP transfers) are relatively small. The BISP UCT transfer (targeted via the PMT) still has still limited coverage among the poor. According to the simulations performed in this study, out of 6 million households monetary poor based on market income, only 3.7 million of them are covered by the BISP UCT transfer. Figure 7.3. Peer countries, ranked by general Figure 7.4. Pakistan and Peer countries, ranked by government tax revenues (% of GDP) general government social expenditures (% of GDP) 40.0 18.0% Tax Revenue as a % of GDP Social expenditure as a % of GDP 35.0 16.0% 14.0% 30.0 12.0% 25.0 10.0% 20.0 8.0% 15.0 6.0% 10.0 4.0% 5.0 2.0% 0.0 0.0% Brazil (2009) Jordan (2010) Mexico (2014) Turkey (2014) Indonesia (2012) Pakistan (2018) Egypt (2015) Sri Lanka (2009) Brazil (2009) Egypt (2015) Mexico (2014) Jordan (2010) Sri Lanka (2009) Turkey (2014) Pakistan (2018) Indonesia (2012) Source: World Bank’s elaboration based on different sources. Data on general government tax revenues from IMF-WEO (April 2022). Data from general government social expenditures based on WB-WDI for Pakistan (validated with official admin data) and CEQ data center for other countries. Note: For Pakistan and most comparator countries (except Egypt and Jordan, where it could not be verified), government social expenditure includes social protection (including old-age pensions), health and education. 8. Methodological Innovations This section presents results on the “Incidence of Indirect Taxes in the Context of Consumption Informality” which have been introduced as a methodological innovation in the fiscal incidence analysis of Pakistan. 8.1 Incidence of Indirect Taxes in the Context of Consumption Informality In the year of analysis, GST is the main source of tax revenue collection in Pakistan, representing 38.1 percent of tax revenue collection, or 3.8 percent of GDP. There are no previous estimates of GST informality in Pakistan, but the World Bank estimates that the country’s GST tax gap is at 87 percent due to a combination of high GST exemptions (policy gap) and informality (implementation gap). In a cross-country study of 31 countries where household expenditure surveys contain information on households’ place of purchase, Bachas et al. (2020) find that: (i) the share of informal consumption is higher among poor households relative to the rich, thus making consumption taxes more progressive; and that (ii) higher consumption informality among poor households makes the consumption tax exemptions (granted by policy design) redundant, since informality acts as a de facto exemption. Given the findings from Bachas et al. (2020), accounting for households’ consumption informality in Pakistan is important to assess the distributional impacts of GST on poverty and inequality. The HIES 2018–19 does not have a place of purchase variable, so this study makes an out-of-sample prediction of households’ consumption informality in Pakistan based on the cross-country regressions available in Bachas et al. (2020). Specifically, this study uses the regression specification 5 (R-squared = 0.43) from Bachas et al. (2020),149 where the dependent variable is the informal share of households’ total consumption and the covariates of the prediction model include: the logarithm of households’ expenditure150 (income proxy), household size, household head’s age, gender of household head gender (dummy if male), and a dummy for food items. For Pakistan, two scenarios were tested: (i) under Scenario A the regression coefficients are based on the average from the 31 countries from Bachas et al. (2020); and (ii) as a sensitivity analysis, under Scenario B the regression coefficients are based on the average of a sub-sample of 12 countries from the reference paper (countries with similar GNI per capita relative to Pakistan).151 The average regression coefficients from both scenarios are presented in Table 8.1. Using the coefficients from Bachas et al. (2020), households’ consumption informality in Pakistan was predicted in a three step process : (i) a household-purchase level dataset was prepared based on the HIES 2018–19, including the socio-demographic variables of interest; (i) the informality share of households’ total consumption was estimated for each household in the HIES based on the informality model and regression coefficients from Bachas et al. (2020) (Table 8.1); and (ii) based on own calculations, the informality share in households’ 149 Table 2, Page 47 from Bachas et al. (2020). 150 Expenditure deflated by the PPP conversion factor for private consumption, based on World Bank-WDI data. 151 According to the World Bank Income Classification, the 12 countries from Bachas et al. (2020) with similar GNI per capita to Pakistan include: Benin, Bolivia, Cameroon, Comoros, Congo Republic, Eswatini, Morocco, Papua New Guinea, So Tome, Senegal, Tanzania and Tunisia. purchases was calculated for each household (e.g., excluding self-consumption), since only the consumption component that corresponds to market purchases is the taxable base for consumption taxes such as GST.152 Table 8.1. Average regression coefficients, regression 5, Bachas et al. (2020) Scenario A Scenario B Regression coefficients (31 countries) (12 countries) Logarithm of households’ consumption (income proxy) -6.8860 -6.4908 Food dummy 29.5978 29.9289 Households’ head age 0.0003 0.0102 Households’ head male (dummy) -0.6384 -0.8692 Households’ size -0.7128 -0.5192 Constant 102.3732 105.2744 Source: Authors’ elaboration based on Bachas et al. (2020). First, the out-of-sample prediction for Pakistan based on the HIES 2018–19 dataset and the informality model from Bachas et al. (2020) suggests that the average informality share in households’ purchases is estimated at 67.8 percent in Scenario A and at 76.4 percent in Scenario B. In both scenarios, the informality share in households’ purchases is higher for poor households and decreases with households’ welfare (i.e.., negative slope of the Informality Engel Curve). For instance, in Scenario A, the informality share in household purchases ranges from 77.3 percent in the poorest decile to 57.0 percent in richest decile,153 whereas in Scenario B the informality shares ranges from 85.9 percent in the poorest decile to 65.6 percent in the richest decile. Also, in both scenarios the informality share in households’ purchases is higher in rural areas than in urban areas. 154 These distributional patterns of informality predicted for Pakistan are qualitatively consistent with findings from Bachas et al. (2020). Second, the analysis incorporates the informality estimates for reassessing GST incidence in Pakistan155 (Figure 8.1). The analysis shows that in the Scenario 0 (without informality), GST relative incidence is almost uniform across deciles, being 6.3 percent in the poorest decile and 6.4 percent in the richest decile. In contrast, when accounting for informality, GST incidence is lower for poorer households and increases for richer deciles, which is consistent with poor households having a higher share of informal purchases. For instance, in Scenario A, the relative incidence of GST (with informality, full sample of 31 countries from Bachas et al. 2020) is estimated at 3.8 percent in the poorest decile and increases up to 4.6 percent in the richest decile. Similarly, in Scenario B (with informality, sub-sample of 12 countries from Bachas et al. 2020), the relative incidence of GST ranges from 3.6 percent for the poorest decile to 4.4 percent for the richest decile. Third, the analysis calculates the GST progressivity index suggested in Bachas et al. (2020), which is defined as the ratio between the GST effective rate of the top quintile and the GST effective rate of the bottom quintile (tax being progressive when this ratio is larger than 1). According to this progressivity index, accounting for the distributional patterns of GST 152 The “informality share in households’ purchases” is estimated by subtracting own consumption ratio in the HIES from the “total informality consumption ratio” (predicted). 153 Deciles ranked by real expenditure per adult equivalent. 154 In Scenario A, the informality share in households’ purchases was estimated at 65.9 percent for rural areas and 64.9 percent for urban areas (78.5 vs. 73.2 percent in Scenario B). 155 As noted in Section 5.2., the GST model in the present fiscal incidence analysis assumes that exempt and informal households’ purchases carry GST indirect effects (e.g. cascading effects from production inputs that paid GST) and zero direct effects. In contrast, Bachas et al. (2020) only focuses on direct effects and thus assumes zero consumption taxes for exempt and informal purchases. informality in Pakistan slightly improves GST progressivity. In effect, in Scenario 0, the GST progressivity index is estimated at 1.01, whereas in Scenarios A and B this index increases slightly to 1.2. However, the latter index is still close to 1, which means that GST in Pakistan is close to neutral, even after accounting for informality. Similarly, when measuring GST progressivity by the KI (relative to disposable income) in Scenario A, the KI is close to zero, which validates that GST is neutral in Pakistan for fiscal year 2018–19 (see Section 6.3). Figure 8.1. Pakistan: GST relative incidence by decile, with and without informality 7.0% 6.3% 6.4% 6.3% 6.4% 6.4% 6.5% 6.4% 6.5% 6.5% 6.4% 6.0% 5.0% 4.4% 4.5% 4.6% 4.0% 4.0% 4.1% 4.2% 4.2% 4.3% 3.8% 4.0% 3.0% 2.0% 1.0% 0.0% 1 2 3 4 5 6 7 8 9 10 Scen. 0: Without informality Scen. A: With informality (full sample of 31 countires) Scen. B: With informality (sub-sample of 12 countries) Deciles by real expenditure, per adult equivalent Source: Authors’ calculations based on HIES 2018–19 and fiscal administrative data. Prediction of informality in households’ purchases based on Bachas et al. (2020). The prediction of households’ consumption informality in Pakistan following Bachas et al. (2020) and the HIES 2018–19 dataset suggests that the informality is higher for poorer deciles and decreases with households’ welfare, consistent with findings from the original study. While accounting for informality slightly improved GST progressivity, the results still suggest that GST in Pakistan was close to neutral for fiscal year 2018 –19. This suggests that finding a progressive GST (via informality) requires larger differences in the share of informal purchases between poor and richer households. One possible reason why the differences in the predicted informality in households’ purchases between the poor and the rich are not wide enough to result in a progressive GST could be that the HIES 2018–19 dataset does not allow is to adequately capture the gap between the poor and the rich, due to the missing top-income households. 9. Conclusions The results from the fiscal incidence analysis for Pakistan in fiscal year 2018–19 show that the combination of taxes and social expenditure items modeled left inequality largely unchanged, with a slight reduction of four tenths of a Gini point from 29.0 at market income plus pensions (prefiscal income) to 28.6 at final income (postfiscal income), but also increased the national poverty rate by 2.3 percentage points from market income plus pensions (prefiscal income) to consumable income (postfiscal income). After paying taxes and receiving cashable transfers, only the poorest decile was a net receiver of the fiscal system, while all the other deciles were net payers. Marginal contribution estimates demonstrate that most of the inequality reduction came from the BISP UCT program. In-kind education transfers – in particular pre-primary and primary education benefits - were another source of inequality reduction. The increase in the poverty headcount and the poverty gap were driven by indirect taxes (GST followed by customs duties). GST was neutral and the customs duties mildly regressive, although they still represented a large share of households’ disposable income. Effectiveness indicators demonstrate that BISP transfers are the most cost-effective expenditure policy for reducing both inequality and poverty depth. Effectiveness indicators also show that direct taxes are the most cost-effective revenue policies for inequality, while also protecting poor and vulnerable households from further impoverishment. Meanwhile, subsidies are relatively ineffective instruments for reducing inequality or poverty in the Pakistani context. There are limitations to these findings. On the one hand, the survey model performs better in the estimation of transfers relative to the estimation of taxes. One of the reasons for the underestimation of taxes is that their allocation is based on income and consumption data, and the HIES demonstrates undercoverage of top-income households. In contrast, the transfers are better allocated in the current fiscal incidence model, since they are based on socioeconomic characteristics and the HIES has better coverage of poorer households. On the other hand, the small change in the Gini coefficient in Pakistan is unlikely to be significant (particularly when compared with the international evidence of similar studies). One of the reasons is that there could be large standard errors in survey-based estimations.156 Comparing the results from Pakistan with seven other middle-income countries where similar fiscal incidence studies have been applied, the study finds that the combination of taxes and social expenditure items modeled in Pakistan achieves the lowest inequality reduction and the highest poverty increase across the sample of countries. This shows that Pakistan has room for improving the redistributive effect of its fiscal system. Moreover, estimations show that 68 percent of the poor at consumable income represent individuals that were fiscally impoverished between prefiscal and postfiscal income. To further reduce poverty and inequality, the Government of Pakistan could expand targeted and progressive social protection transfers (such as the BISP UCT) so that direct transfers offset the impoverishing effects of indirect taxes.157 Moving forward, Pakistan’s will need to increase fiscal space either via domestic revenue mobilization or public expenditure efficiency to be able to expand social expenditure and redistribution measures. Domestic revenue 156 Standard errors could be large given the uncertainty arising from survey data and estimations. 157 Since 2020, the Government has implemented new direct cash transfers programs to offset the negative impacts of the COVID-19 economic crisis and food inflation among the poor and vulnerable (the COVID Emergency Cash Transfers and the BISP TCSP, respectively). mobilization reforms - for instance, the GST harmonization reform - could align the objectives of fiscal sustainability along with fiscal equity if additional revenues are used to compensate targeted households with social protection transfers. Meanwhile, public expenditure efficiency reforms could align fiscal sustainability and fiscal equity if the Government moves from general regressive subsidies (e.g., fuel subsidies) toward targeted and progressive spending. Nonetheless, the federal government still has significant expenditure that is concentrated in other non-social expenditure (71.4 percent of total expenditure in fiscal year 2018–19), which limits the redistributive role of the fiscal system. Ahmad, M., Croraton, Caesar, Qayyum, A., Iqbal, M. and Dorosh, P. 2005. “Impact of Domestic Policies towards Agricultural Trade Liberalization and Market Reform on Food Security in Pakistan.” Ahmed, A. 2018. Measuring Income Inequality in Pakistan: Constructing Distributional National Accounts. Alam, Shamma, and Umar Serajuddin. 2017. “The Distributional Impact of Fiscal Policy in Jordan.” CEQ Working Paper 44, Commitment to Equity (CEQ) Institute, Tulane University, and the World Bank, May 2017. Arunatilake, N., Inchauste, G. & Lustig, N. 2017. "The Incidence of Taxes and Spending in Sri Lanka," Commitment to Equity (CEQ) Working Paper Series 63. Tulane University, Department of Economics. Atkinson, A. B., Piketty, T., & Saez, E. 2011. Top incomes in the long run of history. Journal of Economic Literature 49(1), 3-71. Bachas, P., L. Gadenne, and A. Jensen. 2020. “Informality, Consumption Taxes and Redistribution.” World Bank Policy Research Working Paper 9267. Washington, DC: World Bank. Bargain, Olivier. 2022. "Income sources, intra household allocation, and individual poverty." Commitment to Equity (CEQ) Working Paper Series 121. Tulane University, Department of Economics. Bargain, O. and Donni, O. 2012. "The Measurement of Child Costs: A Rothbarth-Type Method Consistent with Scale Economies and Parents Bargaining." European Economic Review 56(4), 792–813. Beegle, K. and C. Luc. 2019. “Accelerating Poverty Reduction in Africa.” Washington, DC: World Bank. Blanchet, T., Flores, I., & Morgan, M. 2022. “The weight of the rich: Improving surveys using tax data.” The Journal of Economic Inequality 1–32. Bourguignon, F. and P.A. Chiappori. 1992. “Collective Models of Household Behavior: An Introduction.” European Economic Review 36 (2–3): 355–64. Cuevas, P. Facundo, Lucchetti, L. and Nebiler, M. 2020. “What are the Poverty and Inequality Impacts of Fiscal Policy in Türkiye?” CEQ Working Paper 100, Commitment to Equity Institute, Tulane University, May. Deaton, A., & Zaidi, S. 2002. Guidelines for constructing consumption aggregates for welfare analysis (Vol. 135). World Bank Publications. Debowicz, D. et al. 2012. A 2007–08 Social Accounting Matrix for Pakistan. IFPRI & USAID. Diaz Bazan, T. V. 2015. “Measuring inequality from top to bottom.” World Bank Policy Research Working Paper, No. 7237. Enami, A., Larroulet, P. and Lustig, N. (2021). “How Accurate is the Kakwani Index in Predicting Whether a Tax or a Transfer is Equalizing? An Empirical Analysis” ECINEQ Working Paper Series 2022 601 Federal Board of Revenue. 2019. “FBR Yearbook 2018–19.” Finance Division, Government of Pakistan. “Pakistan Economic Survey 2018–19.” Grosh, Margaret, Phillippe Leite, Matthew Wai-Poi, and Emil Tesliuc, editors. 2022. “Revisiting Targeting in Social Assistance: A New Look at Old Dilemmas.” Washington, DC: World Bank. doi:10.1596/978-1-1814-1 Higgins, Sean and Claudiney Pereira. 2013. “The Effects of Brazil’s High Taxation and Social Spending on the Distribution of Household Income.” CEQ Working Paper 7. CEQ Institute, Tulane University and Inter -American Dialogue. Hlasny, V., and Verme, P. 2018. Top incomes and the measurement of inequality in Egypt. The World Bank Economic Review 32(2), 428–455. Inchauste, G., K. Goraus, and H. Carrasco. Forthcoming. “The Distributional Incidence of VAT in the Context of Informality.” World Bank (mimeo). International Monetary Fund. 2017. “Pakistan: Selected Issues”. IMF Country Report No. 2017/213. Jellema, J., Wai-Poi, M., and Afkar, R. 2017. “The Distributional Impact of Fiscal Policy in Indonesia.” CEQ Working Paper 40, Commitment to Equity (CEQ) Institute, Tulane University, and the World Bank, May 2017. Korinek, A., Mistiaen, J. A., and Ravallion, M. 2006. “Survey nonresponse and the distribution of income.” The Journal of Economic Inequality 4(1), 33–55. Ume Laila, Najma Sadiq, Tahir Mehmood, & Khansa Almas. 2021. Impact of Pakistan Bait-Ul-Mal on the Socioeconomic Conditions of the Beneficiaries. Journal of Business and Social Review in Emerging Economies, 7(1), 69-80. Lara Ibarra, G., Sinha, N., Nayer, R., Fayez, S. and Jellema, J. 2019. "Impact of Fiscal Policy on Inequality and Poverty in the Arab Republic of Egypt". Policy Research Working Paper No. WPS 8824 (Washington, DC: World Bank). Lustig, N. 2018. “Commitment to Equity Handbook: Estimating the Impact of Fiscal Policy on Inequality and Poverty.” Brookings Institution Press. Meyer, B. D., and Sullivan, J. X. 2003. Measuring the well-being of the poor using income and consumption. Ministry of Planning Development and Special Initiatives & Pakistan Bureau of Statistics. 2021. “Pakistan Social and Living Standards Measurement Survey PSLM/HIES 2018–19.” National Education Management Information System, Academy of Educational Planning & Management and Ministry of Federal Education & Professional Training. 2021. “Pakistan Education Statistics 2017–18.” Nazir, K. 1996. “Notes and Comments: Zakāt and 'Ushr System in Pakistan.” Islamic Studies 35(3) (Autumn 1996), pp. 333–343 (11 pages). National Electric Power Regulatory Authority (NEPRA). 2020. “State of Industry Report 2019.” Nazir, K. 1996. “Zakat and Ushr System in Pakistan”. Ocampo, J. A. & Gomez-Arteaga, N. 2016. “Social protection systems in Latin America: An assessment.” International Labor Organization. ESS Working Paper No. 52. Pakistan Bureau of Statistics (PBS). YYYY. “National Health Accounts 2019-20 Pakistan.” Pakistan Bureau of Statistics (PBS). 2021. “National Health Accounts 2017–18: Pakistan.” Pakistan Bureau of Statistics (PBS). 2019. “Labor Force Survey 2017–18.” Pakistan Institute of Development Economics (PIDE). 2021. “Fertilizer Subsidy an Ineffective Policy Tool to Offer Low Prices of Basic Food Commodities.” Knowledge brief, August 2021. Pakistan Institute of Development Economics (PIDE). 2020. “The Pension Bomb and Possible Solutions.” Knowledge Brief, November 2020. Pakistan Centre for Philantropy (2016). "The State of Individual Philanthropy in Pakistan”. Phillips, D., R. Warwick, M. Goldman, K. Goraus, G. Inchauste, T. Harris, and J. Jellema. 2018. “Redistribution Via VAT and Cash Transfers: An Assessment in Four Low- and Middle-Income Countries.” CEQ Working Paper Series 78. Tulane University. Piketty, T., & Saez, E. 2003. Income inequality in the United States, 1913–1998. The Quarterly Journal of Economics 118(1), 1–41. Piketty, T. 2018. Capital in the twenty-first century. Harvard University Press. Piketty, T., Yang, L., & Zucman, G. 2019. “Capital accumulation, private property, and rising inequality in China, 1978–2015.” American Economic Review 109(7), 2469–96. Qureshi, A.S. 2020. “Groundwater Governance in Pakistan: From Colossal Development to Neglected Management”. Water 2020, 12, 3017. Redaelli, S. 2019. “Pakistan at 100: From Poverty to Equity.” World Bank Policy Note. Scott, John, Sandra Martinez-Aguilar, Enrique de la Rosa, and Rodrigo Aranda. 2020. “CEQ Master Workbook: Mexico (2014).” CEQ Data Center on Fiscal Redistribution (CEQ Institute, Tulane University). July 28, 2020. Sheikh, S.A. 2015. "Welfare Potential of Zakat: An Attempt to Estimate Economy wide Zakat Collection in Pakistan"/ Pakistan Development Review 54(4):1011-1027 Tandon, A. et al. 2021. “Financing health care in the WHO South-East Asia Region: time for a reset”. WHO South- East Asia Journal of Public Health | February 2021 | 10(Suppl. 1) UK Home Office. 2020. “Country Policy and Information: Pakistan Medical and Healthcare Provision.” Williams, C., Shahid, M. and Martinez, A. 2016. “Determinants of the Level of Informality of Informal Micro - Enterprises: Some Evidence from the City of Lahore, Pakistan.” World Development 84(August 2016), pp 312– 325. World Bank. 2022. Poverty and Shared Prosperity 2022: Correcting course. World Bank. 2020a. “Leveling the playing field: Systematic Country Diagnostic.” World Bank. 2020b. “Pakistan: Assessment of Civil Service Pensions.” Report No: AUS0001350. World Bank. 2009. “Pakistan Tax Policy Report Tapping Tax Bases for Development.” World Bank. Report No. 50078-PK. 1.1. Direct Taxes 1.1.1. Withholding tax on salaries Annex Table A1.1. PIT schedule in Pakistan, 2018–19 Lower PIT Minimum PIT Upper PIT threshold PIT rate (% over the excess of threshold amount (PKR, annual) the lower threshold) (PKR, annual) (PKR, annual) Bracket 1 0 - 400,000 - Bracket 2 0 400,000 800,000 1,000 Bracket 3 0 800,000 1,200,000 2,000 Bracket 4 5 1,200,000 2,400,000 Bracket 5 10 2,400,000 4,800,000 60,000 Bracket 6 15 4,800,000 300,000 Source: Authors’ elaboration based on official tax legislation. 1.1.2. Withholding tax on telecommunications • 10 percent for telephone (landline) and internet subscribers exceeding PKR 1,000 monthly. • 12.5 percent for prepaid mobile. 1.1.3. Property tax • Property tax in Pakistan is a provincial tax levied on banded rental value of the property (both commercial and residential), based on Property Tax Laws of respective provinces. Tax rates vary by province. It is either a flat rate or a percentage of the annual rental value (ARV) and varies if the property is rented out or self- occupied (except for Sindh). Each province determines its exemptions to the property tax, some of which include: exemptions to widow/orphans/disabled (several countries); exemptions for retired government servants (Islamabad); exemptions to public parks/public charities (KP); and exemptions to self-occupied property below a certain area or not owning another property (Punjab, Islamabad). • Property tax rates: These are different for each province. From 2002, in Punjab (and later in other provinces) determination of ARV has been made using “valuation tables” prepared by the Excise and Taxation Department based on a house-to-house rental value assessment. For tax-base assessment purposes, four major factors are considered: use, quality, location, and size. The calculation of ARV and exemptions in respective provinces are listed below. - Islamabad: Capital Development Authority collects the property tax within the jurisdiction of Islamabad. A 5 percent rebate is allowed on payment of current tax up to September 30. A non-waivable surcharge of 1.5 percent per year is payable on the outstanding tax after the end of the fiscal year. - Punjab: Under the provisions of the Act, the property tax is levied on the ARV of buildings and land located in the rating area. It is levied at the rate of 5 percent of ARV at which the property may be let out on a year- to-year basis. - Sindh: The property tax is levied and collection under Sindh Urban Immovable Property Tax Act, 1958. A uniform rate of tax is levied on all categories of properties at 25 percent of ARV. To ascertain the ARV of a property unit, the Government of Sindh has rationalized and simplified the system by notifying a valuation table for different localities.158 - Khyber Pakhtunkhwa: Property tax (adopted “THE WEST PAKISTAN URBAN IMMOVABLE PROPERTY TAX ACT, 1958” with modifications and Finance Bill) applied on urban areas and commercial/residential. The annual value of any land or building is ascertained by estimating the gross annual rent. Buildings and land within the limits of urban areas are divided into categories A1, A, B, C and D. Buildings for commercial use falling within A1, A and B category are assessed and taxed at 18 percent of ARV. - Balochistan: Property Tax in Balochistan is levied and collected based on ARV of properties as follows: (i) properties having an ARV below PKR 12,000 are charged at 10 percent; and (ii) properties having an ARV of PKR 12,000 and above are charged at 15 percent. A person having one self-occupied residential unit with a covered area of up to 5,000/sq feet is exempt from property tax. However, application for a tax exemption needs to be submitted by the owner on an annual basis. ARV is determined after deduction of 10 percent from the gross annual rental value (GARV). 1.2. Indirect Taxes 1.2.1. General Sales Tax • GST in Pakistan is applied at the federal level (for goods) and at the provincial level (for services). General application of GST input tax credit: (i) For goods, they are eligible for input tax credit if subject to GST or zero-rated. Exempted items are not eligible for input tax credit. (ii) For services, it depends on the rules for each province. This suggests that at the federal level the GST input tax credit mechanism works in a similar way to VAT.159 • The standard GST rate in Pakistan is set at 17 percent. However, there are multiple reduced rates and exempted items. For goods, some of the items under the reduced GST rate include food (e.g., rice and rice flour, fresh milk, several fruits, cooking oil, tea, others), medicine, jewelry and textbooks; exempted items also include other food products (e.g., wheat, cereals, chicken meat, beef, fish, eggs) and household sewing machines; zero-rated items include garments, textile products, clothing accessories, carpets/curtains (Annex Table A1.2). 158 For this purpose, all cities (notified as rating areas) in the Province of Sindh have been divided into five groups and these groups are further bifurcated into four zones according to their socio-economic condition of the localities. The different rates by category have been specified for each zone of every group, for the size of plot n sq. yards and covered area in sq. feet of the property unit. 159 “Input tax adjustment is the deduction of input tax from output tax to arrive at the net amount of sales tax payable by the taxpayer. Since sales tax is a value added tax, it is to be charged at each incremental stage of value addition, otherwise there may occur double taxation. Input tax is adjusted against output tax to avoid such double taxation and to calculate the correct amount of tax due to the government”. Source: Sales Tax Guide from FBR. • GST rates for services and the applications of input tax credit vary by province. Below is a non-exhaustive list of services that are not eligible for input tax credit for each province (e.g., could have embedded indirect effects)160 (Annex Table A1.3). Annex Table A1.2. GST at the federal level Examples of exempted goods Examples of zero-rated goods - Rice and rice flour - Woven fabrics of carded wool or of carded - Pulses fine animal hair, of silk, or synthetic staple - Meat (bovine, beef, chicken) - fibers fresh or chilled - Garments - Eggs - Clothing accessories n.e.s. parts of garments - Crustaceans (in all forms. Live or or accessories other than those of heading cooked, in shell or not, frozen or - Textile products fresh) - Curtains (including drapes) and interior - Milk, yogurt blinds curtain or bed valances / Bed linen, - Milk and cream concentrated or table linen, toilet linen and kitchen linen; containing added sugar or other Blankets and travelling rugs / Mattress sweetening matter supports articles of bedding (e.g., - Butter and other fats and oil mattresses, quilts, eiderdowns, cushions derived from meat pouffes and pillows), fitted with springs or - Fresh fruit stuffed, whether or not covered - Vegetables, fresh or chilled - Carpets and other textile floor coverings - Onions, shallots, garlic, leeks knotted, whether or not made up and other alliaceous vegetables - Petroleum oils, HSD/LDO/Kerosene/motor fresh or chilled spirit/ others - see relevant sheet - Leguminous vegetables shelled - Registers, account books, diaries and similar or unshelled, fresh or chilled albums for samples or collections, of paper - Vegetable oil (palm oil, or paperboard sunflower oil, sunflower oil) - Pencils (not of heading no. 9608), crayons, - Salt, spices (ginger, saffron, pencil leads, pastels, drawing charcoals, curry) writing or drawing chalks and tailors chalks - Cane/beet sugar - Pens, ball-point, felt-tipped, other porous - Fruit juices tipped pens, fountain pens, stylograph pens - Pharmaceutical goods duplicating stylos, propelling or sliding - Printed books and material pencils parts of the forgoing, excluding - Typewriters and word- those of heading no. 9609 processing machines (laptops, - Trunks, suit, camera, jewelry, cutlery cases computers etc.) travel, tool, similar bags wholly or mainly - Registers, account books, covered by leather, composition leather, diaries and similar albums for plastic sheeting, textile materials, samples or collections, of paper vulcanized fiber, paperboard or paperboard - Tissue, towels, napkin stock or similar for - Bicycles and other cycles household or sanitary uses, cellulose including delivery tricycles, not wadding, webs of cellulose fibers, motorized. - Bicycles and other cycles including delivery - Sewing machines tricycles, not motorized. - Fertilizer 160 Indirect effects = Price increase in final output due to taxed input. It is passed from producers to final consumers when they cannot claim the GST input tax credit (e.g., exempted items, informal items, or items not eligible for GST input tax credit). Milk and cream concentrated or containing - added sugar or other sweetening matter Source: Authors’ elaboration based on official tax legislation. Annex Table A1.3. Services not eligible for GST input tax credit at the provincial level Punjab Sindh - Owner occupied accommodation - Tailoring charges/ Embroidery, alterations etc. (Market value) excluding self-hiring - Shoes Repair/Polish/ shoes shining /Shoes - Doctor Fee (Hakeem/ Brush homeopathic/specialist/general etc.) - Owner occupied accommodation (Market - Dental care, teeth cleaning, value) excluding self-hiring extraction, charges - Bread, Paratha, Nan, Roti, etc. - Other expenditure not mentioned - Hair Cut Charges/ Shaving Charges (for men) elsewhere (ambulance services, - other (Services for maintenance and repair of speech therapist, midwives, household dwelling) Accompanying Person Cost, Tips, - Stitching charges on household clothing Room Charges etc. thread and tailoring expenses. - Subsidized house rent (Hiring/Self - Private vehicle (Bus, Wagon, Taxi etc.) fare hiring) (Market value) (Outside City) - Rent free accommodation (Market - Other Transport Charges (tanga charges, value) camels, donkeys, ferries, bicycles and rent of - By Air, by ship, boat, ferry Travelling Garage, etc.) Charges etc. - Motorcycle/Scooter Service /tuning /overhauling charges/ Tire Puncture etc. charges KP Balochistan - House rent (house rent, owner - Owner-occupied accommodation (Market occupied, subsidized) value) excluding self-hiring - Doctor Fees - Subsidized house rent (Hiring/Self hiring) - Dental care, teeth cleaning, (Market value) extraction, charges - Rent free accommodation (Market value) - Other expenditures not mentioned elsewhere (ambulance services, speech therapist, midwives, Accompanying Person Cost, Tips, Room Charges etc. - Rent of TV/VCR/video cassettes, cd’s and cable etc. - Service and repair charges of household effects, etc. mentioned above - Car Service Charges/ tuning /repair charges/ - Repair charges of personal effects - Beauty Parlor Hair Style Charges - Legal expenses not related to business - Expenditure on photography - Recreational park charges - Marriage Hall Charges - Cinemas/theaters/concerts/music hall/circuses/museums (ticket only) Islamabad - House rent (owner occupied, subsidized rent) - Washing/ Dry Cleaning charges - Service and repair charges of household effects, etc. mentioned above - Expenditure on cleaning of carpets, blankets, quilts, linen etc. - Car Service Charges / tuning /repair - Repair charges of personal effects (watches, clocks, glasses, etc.) - Beauty Parlor Hair Style Charges Source: Authors’ elaboration based on official tax legislation. 1.2.2. Custom duties • Base and rates: Pakistan’s overall average applied tariff in 2018 was 10.09 percent. However, there are different types of customs duties levied on an ad-valorem basis: (i) customs duties ranged from 0 to 50 percent (with the minimum rate being 0.75 percent); (ii) regulatory duty ranged from 0 to 60 percent (with the minimum rate being 3.33 percent); and (iii) the additional duty ranged 0 to 7 percent (with the minimum rate being 2 percent). Annex Table A1.4. Examples of customs duties in Pakistan 2018–19 Examples of rates and goods (non-exhaustive list) Custom duty • 0.75%: Fertilizers • 3.00%: Wheat, wheat flour, pulse, beans, beef, mutton, potato, onion, tomato, garlic, cabbage, cardamom, and other food products. Jewelry, miserableness personal articles, medical apparatus, calculators, personal computers; literature books, others. • 20%: milk, chicken meat, sugar (all kinds), unstitched cloths, pants/shirts (boy/gentlemen); school uniform; shoes/sandal and other footwear; cigarettes; toilet soap and shampoo; cement/blocks/iron bars/marbles/tiles/door/windows; stationary items like pen/pencil/glaze paper/chart paper. • 50%: Motorcycles Regulatory duty • 3.33%: transport and travelling vehicles • 30%: Footwear • 60%: fruit juice (fresh or packed); squash/sharbat. Additional duty • 2%: pulses, cereal grains, onions, tomatoes, cucumber, garlic, black tea, other food products; coffee; tobacco; LPG, electrical energy, kerosene, natural gas, motor spirits; paracetamol; disinfectants; thermometers; electro-mechanical tools, others. • 7%: milk (fresh and tetrapack), chicken meat, cooking oil; cigarettes; unstitched cloths; school uniform; pants/shirts (buys/gentlemen); cigar shoes/sandals/other footwear; toilet soap, shampoo; cement/blocks/iron bars/marbles/tiles, others. Source: Authors’ elaboration based on custom duty data provided by World Bank’s Macroeconomics Trade and Investment team Note: Complete list of items is available in master dataset provided to the authors. 1.2.3. Excise duties Annex Table A1.5. Excise rates by item in the HIES 2018–19 Item name (from HIES 2018–19) Excise rate Vegetable Ghee (Tin/Loose) 16 percent of retail price Cold Drink (Carbonated) 11.5 percent of retail price Cigarettes PKR 1,670 per thousand cigarettes Cooking Oil (Tin/Loose) 16 percent of retail price Cement, Brocks/Blocks/Bajri/Sand/Timber PKR 1.25 per kilogram Log/Iron bars/Marbles/Tiles/Chips /door/windows, etc. Gas Charges PKR 10 per million British Thermal Units Carbonated Drinks, Juices, Lemon Soda, etc. 11.5 percent of retail price Carbonated Drinks, Juices, Lemon Soda, Mineral 11.5 percent of retail price Water, etc. By Air, by ship, boat, ferry Travelling Charges, etc. PKR 5,000 Glucose, Energile, Tang, Lemon Pani, etc. 50 percent ad valorem Sonf Suparee, Chewing tobacco, Gutka, etc. 65 percent of retail price Other (in the category of mineral water, juices and 11.5 percent of retail price soft drinks) Mustard Oil (Cooking purpose) 16 percent of retail price Raw Tobacco PKR 10 per kg CNG Expenses PKR 17.18 per hundred cubic meters. LPG (Liquid Petroleum Gas) PKR 85/million tons Transport and travelling vehicles (Bicycle, 10 percent ad valorem Motorcycle, Scooter, Car, horses, camels, tongas, etc.) Source: Authors’ elaboration based on official tax legislation. 1.3. Direct Transfers 1.3.1. BISP Unconditional Cash Transfer Annex Table A1.6. BISP UCT Eligibility criteria Target population Annual transfer imputed, per household (PKR) - PMT score (below 29) 3,888,010 beneficiary PKR 18,000161 - At least one female married households above 18 years old (equivalent to 5,054,414 -Not self-reported beneficiary beneficiary families) Source: Authors’ elaboration based on official data from BISP program and World Bank’s Social Protection team. 1.3.2. BISP Conditional Cash Transfer (Wet) 161 Beneficiary families receive quarterly transfers. The BISP UCT transfer was PKR 3,000 per quarter per family until June 2018 and then PKR 5,000 per family per quarter since July 2018. Hence, the assumption applied was that households received PKR 3,000 for the first quarter of fiscal year 2018–19 and PKR 5,000 for the following three quarters. Annex Table A1.7. BISP CCT Eligibility criteria Target population Annual transfer imputed, per eligible child (PKR) -Being a beneficiary of the UCT 1,087,000 beneficiary PKR 3,000 transfer households -Having at least one child eligible for the transfer (between 5 and 12 years old, currently attending school -Living in one of the 50 districts where the CCT was being implemented in 2018 (see below) Source: Authors’ elaboration based on official data from BISP program and the World Bank’s Social Protection team. 1.3.3. Memorandum: Public pensions (not part of direct transfers) Annex Table A1.8. Calculation of average public pension, by administrative unit Pensioners Expenditure executed Avg. annual pension Avg. monthly pension (latest data 2019–20 (PKR) (PKR, estim.) (PKR, estim.) available) SINDH 130,665,424,000 211,465 617,905.68 51,492.14 KPK 69,913,020,000 157,628 443,531.73 36,960.98 PUNJAB 207,600,000,000 508,166 408,527.92 34,043.99 BALOCHIS TAN 30,084,955,000 75,900 396,376.22 33,031.35 TOTAL provincial 438,263,399,000 953,159 466,585.39 38,882.12 Federal 2018* 421,870,000,000 1,279,300 329,766.28 27,480.52 Source: Authors’ elaboration based on pensions data from Institutional Reform Cell, Government of Pakistan. 1.4. Indirect subsidies Annex Table A1.9. Indirect subsidies Indirect subsidy Calculation of subsidy per unit 2018–19 Urea fertilizer subsidy International urea price = PKR 2,193 per bag of 50 kg Local urea price = PKR 1,745 per bag of 50 kg Implicit subsidy (price difference) = PKR 448 per bag of 50 kg = 25.6 percent of the local urea price Agriculture tubewell subsidy Average monthly electricity tariff (for the agriculture sector) = PKR 8.425 per kW Average cost of electricity (general sector) = PKR 15.12 per kW Implicit subsidy = PKR 6.695 = 79.5 percent of the monthly tariff paid by the agriculture sector Source: Authors’ elaboration based on official sources. Domestic electricity subsidy Subsidy to electricity sector is defined as the tariff differential subsidy (TDS), which is the difference between the Determined Tariff (DT), which proxies for the cost of supply, and the Notified Tariff (NT),162 which indicates the tariff passed onto the consumers. The determined and notified tariffs include quarterly tariff adjustments levied on consumers above 300 kWh per month. The electricity tariff of single-phase domestic consumers is articulated into six slabs. All units consumed up to 300 kWh per month receive a TDS, corresponding to the difference between the DT and the slab’s tariff, while units above 300 kWh are priced above the DT, hence cross- subsidizing the lower units. The first slab, identifying the consumption of customers paying highly subsidized lifeline tariff, aimed to protect the poorest consumers. In addition to the TDS, electricity customers above 100 kWh per month also benefit from an Increasing Block Tariff (IBT) structure, whereby the tariff of the previous slab is granted to all the units consumed up to a lower bound of the slab in which their consumption falls. The final price paid by consumers is the combination of base tariffs plus surcharges, a price adjustment to reflects changes in fuel prices, fixed costs, as well as 17 percent GST on the cost of electricity, inclusive of all surcharges and fixed costs. Surcharges include the financial cost (FC) surcharge, a tariff rationalization (TR) and the Neelum– Jhelum (NJ) surcharge.163 All surcharges apply to all consumers except those falling in the first slab (lifeline users).164 The TR has been applied until January 1, 2019, as a surcharge or subsidy to the tariff approved by NEPRA for the different distributors, with the aim, among others, of maintaining uniform tariffs across the country for each of the consumer category, and it was on average PKR 1.02/kWh. The NJ surcharge, which was formally eliminated in June 2018, is nevertheless still levied at a rate of PKR 0.1/kWh. All consumers, including 162 The determined and notified tariffs include quarterly tariff adjustments levied on consumers above 300 kWh per month. 163 The surcharge is intended to finance the Neelum Jhelum Hydroelectric Project, completed in April 2018. The initial project proposal cost of PKR 84,502.260 in 2002 has increased fivefold to PKR 506,808.610 in 2018 (WAPDA,). 164 Lifeline consumers are defined as residential consumers having single-phase electric connection with sanctioned load up to 1kW. At any point of time, if the floating average of past six months' consumption exceed 50 units, then the said consumer would not be classified as lifeline for the billing month even if its consumption is less than 50 units (Government of Pakistan, Ministry of Energy, 2019). lifeline users, are also charged two fixed costs: a PKR 23.4 meter-rent-fee, and a PKR 35 license-fee to finance the Pakistan Television Corporation (PTV). The fuel price adjustment is determined on a monthly basis by NEPRA and reflects changes in production costs associated to fuel price fluctuations. Fuel price increases are passed on to all consumers, except lifeline users. Considering the base tariff, price surcharges and GST, the effective monthly cost of electricity for consumers is summarized in Annex Table A10. Except for lifeline users, all other consumers experienced a fluctuation of electricity cost, driven by fuel price adjustments. The direct impact of the electricity subsidy on domestic consumers is estimated as the difference between the average determined tariff and notified tariff multiplied with the total electricity consumption of the households in each month. Annex Table A1.10. Effective consumer electricity cost 2018 2019 Avera Averag Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun ge e (Base Tariff + FC surcharge + TR surcharge + NJ + FPA) * GST, PKR/kWh 0 - 50 2.34 2.34 2.34 2.34 2.34 2.34 2.34 2.34 2.34 2.34 2.34 2.34 2.34 2.34 51 - 9.00 9.95 8.82 9.14 8.59 9.25 9.13 10.6 8.34 7.39 8.04 7.51 7.39 8.21 100 101 - 11.7 12.6 11.5 11.8 11.3 11.9 13.3 11.0 10.1 10.7 10.2 10.1 11.84 10.93 200 1 6 4 6 0 7 1 6 1 5 3 1 201 - 14.1 15.1 13.9 14.3 13.7 14.4 15.7 13.5 12.5 13.2 12.6 12.5 14.29 13.37 300 6 1 8 0 5 1 6 0 5 0 7 5 301 - 20.9 21.8 20.7 21.0 20.1 21.2 24.4 22.1 21.1 21.8 21.3 21.0 21.01 22.00 700 5 9 7 9 4 0 2 6 7 6 3 6 700 23.2 24.2 23.1 23.4 22.4 23.5 28.0 25.7 25.4 24.9 24.6 and 23.35 24.8 25.63 9 3 1 3 8 4 4 9 8 6 9 more Fixed Cost per month * GST, PKR Meter 23.4 23.4 23.4 23.4 23.4 23.4 23.4 23.4 23.4 23.4 23.4 23.4 23.4 23.4 Fee PTV 35 35 35 35 35 35 35 35 35 35 35 35 35 35 Source: NEPRA and WB Energy GP. 1.5. In-kind Education Benefits Annex Table A1.11. In-kind education benefits Admin data 2018–19 Own Estimation Total public Total enrollment in Annual benefit per student, ICT expenditure public schools*, # PKR (executed), PKR of students Pre-Primary and Primary 9,021,348,365 128,160 70,391.32 Secondary 12,306,348,365 132,954 92,560.72 Tertiary 55,797,723,029 821,859 67,892.07 PUNJAB - Pre-Primary and Primary 2,277,334,151 10,891,413 209.09 Secondary 53,147,334,151 5,056,540 10,510.61 Tertiary 51,254,686,012 659,599 77,705.78 SINDH - Pre-Primary and Primary 69,600,873,266 3,822,966 18,205.99 Secondary 57,083,873,266 1,656,230 34,466.14 Tertiary 43,666,682,800 296,029 147,508.01 BALOCHISTAN - Pre-Primary and Primary 22,082,924,570 943,240 23,411.79 Secondary 20,645,924,570 307,853 67,064.23 Tertiary 16,373,020,242 58,573 279,532.55 KP - Pre-Primary and Primary 6,006,186,313 4,335,980 1,385.20 Secondary 9,715,186,313 1,604,968 6,053.20 Tertiary 39,036,764,584 200,407 194,787.83 Source: Authors’ elaboration. Estimations based on official sources. Note: Estimations of enrollment under the cross-classification of province-education level based on total enrollment levels by province in 2018–19 and disaggregated enrollment by province-education level in 2017– 18. Annex Table A1.12. Public education students: Survey vs. administrative estimates Public education students (PPR) HIES/Admin Admin data 2018–19 HIES 2018–19 % Pre-primary 5,912,519 3,504,684 59 Primary 14,209,240 14,824,662 104 Secondary 8,758,545 11,065,398 126 Tertiary (public) 1,972,049 1,743,670 88 Source: Authors’ elaboration. Estimations based on official sources and HIES 2018–19. 1.6. In-kind Health Benefits Annex Table A1.13. In-kind health benefits Public health expenditure Number of patients, Average benefit per patient executed in 2018–19 based on HIES 2018–19 (PKR, annual) (PKR MM) Inpatient Outpatient Inpatient Outpatient Inpatient Outpatient 8,217 20,016 485,064 410,511 3,377 Islamabad 1,638 128,906 1,390,616 13,701,204 92,697 1,875 Punjab 25,695 57,789 419,004 3,834,220 137,920 3,004 Sindh 11,519 32,779 662,888 5,938,872 49,449 1,100 KP 6,534 12,899 20,568 1,091,100 627,145 2,356 Balochistan 2,571 Source: Authors’ elaboration based on official sources. Notes: 1/ The estimation of public health expenditure is based on total expenditure 2018–19 (disaggregated by province) from MoF and shares of expenditure inpatient vs. outpatient based on National Health Accounts 2015– 16. 2/Number of patients based on individuals that self-report accessing public health facilities for inpatient and outpatient services in the HIES 2018–19 (recall period of three months annualized by 4). Number of patients includes public employees (predicted) that accessed private health facilities since they are eligible to receive reimbursement from public health insurance. The HIES 2018–19 did not have information on type of employer (public/private), or on proxies of informality (e.g., workers having a contract or paying social security contributions). For these reasons, the variables of “public sector” (dummy) and “having a job contract” (dummy) where estimated based on OLS regression coefficients estimated at the Labour Force Survey (LFS) 2017–18, which had information on both types of employers and if employees had a contract. The models focus on individuals that report being “paid employees” as their main occupation, since this is the target population for the withholding tax on salaries modeled in the current fiscal incidence analysis. There are some differences in the figures of paid employees-main occupation reported in the LFS 2017–18 (23.5 million or 43 percent of respondents) and the HIES 2018–19 (33.9 million or 56 percent of respondents). So, for the out of sample predictions in the HIES based on LFS, this study targets the relative values of the variables respectively (share of paid employees having a main occupation working in the public sector and share of paid employees having a contract in the main occupation). For predicting “public employee” and “having a contract” in the HIES, the covariates selected for the models were employment and individual characteristics that were available both in LFS (for the OLS regression) and HIES (for the out-of-sample prediction based on LFS OLS coefficients). These covariates were: urban dummy, male dummy, education level, industry classification, and occupation type. Based on these variables, the OLS regression specification was the same for predicting both “public employee” and “having a contract”. However, in each model, the regression coefficients vary (see below). Annex Table A2.1. Methodological approach for estimating public sector employees a. Public sector employee (OLS model in the LFS 2017–18) . *5.2. OLS . reg public_emp_1 i.industry_class i.type_occup i.educ male urban [weight=int(weight)] (analytic weights assumed) (sum of wgt is 5.5477e+07) Source SS df MS Number of obs = 78,402 F(34, 78367) = 4648.43 Model 3524.01312 34 103.647445 Prob > F = 0.0000 Residual 1747.37181 78,367 .022297291 R-squared = 0.6685 Adj R-squared = 0.6684 Total 5271.38492 78,401 .067236195 Root MSE = .14932 public_emp_1 Coef. Std. Err. t P>|t| [95% Conf. Interval] industry_class_1 Mining and quarrying .135557 .0114713 11.82 0.000 .1130733 .1580407 Manufacturing .0259738 .0029549 8.79 0.000 .0201822 .0317654 Electricity, gas and air cond. .803589 .0088105 91.21 0.000 .7863204 .8208576 Water sypply, sewage and waste mgmt .8794453 .0097682 90.03 0.000 .8602997 .8985909 Construction .0148135 .0029404 5.04 0.000 .0090504 .0205766 Wholesale and retail; motor vehicle repair .0049304 .0034174 1.44 0.149 -.0017677 .0116284 Transport and storage .1009388 .0037537 26.89 0.000 .0935817 .108296 Accommodation and food services .0112318 .0050424 2.23 0.026 .0013488 .0211148 Information and commmunic. .0987544 .0080709 12.24 0.000 .0829354 .1145734 Financial and insurance .2657778 .0080577 32.98 0.000 .2499848 .2815708 Real estate -.0149425 .0087079 -1.72 0.086 -.03201 .002125 Professional, sci and technical .280403 .0080449 34.85 0.000 .264635 .296171 Admin and support .0265316 .0078839 3.37 0.001 .0110792 .041984 Public admin and defense .9845801 .0044622 220.65 0.000 .9758343 .993326 Education .6243873 .0045179 138.20 0.000 .6155322 .6332424 Health and social work .4798213 .0053852 89.10 0.000 .4692663 .4903764 Arts and entertainment .1261555 .0130021 9.70 0.000 .1006714 .1516396 Other services .0233197 .0046544 5.01 0.000 .0141971 .0324423 HH's as employers, other undifferentiated goods .0145096 .0051345 2.83 0.005 .004446 .0245732 Activities of extraterritorial orgs .0049859 .0306298 0.16 0.871 -.0550483 .0650202 type_occup_1 Professionals -.0037733 .0048118 -0.78 0.433 -.0132043 .0056578 Technicians and associate professionals .0642139 .0047443 13.53 0.000 .054915 .0735127 Clerks .1057222 .0058278 18.14 0.000 .0942996 .1171447 Service workers and shop and market sales workers .0476944 .0041241 11.56 0.000 .0396112 .0557777 Skilled agricultural and fishery workers .0504072 .0046737 10.79 0.000 .0412467 .0595676 Craft and related trades workers .0382864 .0041909 9.14 0.000 .0300722 .0465005 Plant and machine operators and assemblers -.0065173 .0046851 -1.39 0.164 -.0157 .0026654 Elementary occupations .0540986 .0042366 12.77 0.000 .0457949 .0624023 educ Primary .003472 .001607 2.16 0.031 .0003224 .0066217 Lower secondary .0066236 .0017975 3.68 0.000 .0031004 .0101467 Upper secondary .0170942 .0016204 10.55 0.000 .0139183 .0202701 Post-secondary .0496773 .0027207 18.26 0.000 .0443448 .0550098 male .0113815 .0011712 9.72 0.000 .009086 .013677 urban -.0102479 .0014048 -7.29 0.000 -.0130012 -.0074945 _cons -.062047 .0047227 -13.14 0.000 -.0713033 -.0527906 Source: Own estimations based on LFS 2017–18 data. b. Public sector employee (out of sample prediction in the HIES 2018–19) Results. Based on the coefficients presented in table above from LFS and covariates from HIES, the estimation of “public sector employee” was performed for those paid employees-main occupation in HIES. For defining “public employee” = 1, the threshold was established as those employees with an estimated probability of having a contract equal or higher than 35 percent. This results in a formality rate in HIES 2018–19 of 17.02 percent similar to that of LFS 2017–18 (17.04 percent). Annex Table A2.2. Share of public employees in the HIES 2018–19 (main occupation) Public sector employee Freq. % Cum. Private 28,163,204 82.98 82.98 Public 5,778,039 17.02 100 Total 33,941,243 100 Source: Own estimations in HIES 2018-19 based on regression coefficients from LFS 2017-18 Annex Table A2.3. Methodological approach for estimating “having a job contract” c. Having a job contract (OLS model in the LFS 2017–18) . reg formal i.industry_class i.type_occup i.educ male urban [weight=int(weight)] (analytic weights assumed) (sum of wgt is 2.3527e+07) Source SS df MS Number of obs = 31,798 F(34, 31763) = 902.18 Model 2977.68455 34 87.5789574 Prob > F = 0.0000 Residual 3083.38532 31,763 .097074751 R-squared = 0.4913 Adj R-squared = 0.4907 Total 6061.06987 31,797 .190617664 Root MSE = .31157 formal_1 Coef. Std. Err. t P>|t| [95% Conf. Interval] industry_class_1 Mining and quarrying .0900217 .0256278 3.51 0.000 .0397903 .140253 Manufacturing .1226684 .0075162 16.32 0.000 .1079363 .1374004 Electricity, gas and air cond. .746848 .0193495 38.60 0.000 .7089221 .7847738 Water sypply, sewage and waste mgmt .746774 .0216344 34.52 0.000 .7043697 .7891784 Construction -.0038068 .0072058 -0.53 0.597 -.0179305 .0103169 Wholesale and retail; motor vehicle repair -.0346341 .0096734 -3.58 0.000 -.0535944 -.0156738 Transport and storage .1025615 .01027 9.99 0.000 .0824319 .1226912 Accommodation and food services -.0003889 .0146752 -0.03 0.979 -.0291529 .0283751 Information and commmunic. .2918727 .0183444 15.91 0.000 .2559168 .3278285 Financial and insurance .5140701 .018169 28.29 0.000 .4784581 .5496821 Real estate .159745 .0325134 4.91 0.000 .0960174 .2234726 Professional, sci and technical .3879357 .0227634 17.04 0.000 .3433186 .4325527 Admin and support .1225985 .0215135 5.70 0.000 .0804313 .1647657 Public admin and defense .7710495 .0106121 72.66 0.000 .7502493 .7918496 Education .4685353 .0108969 43.00 0.000 .447177 .4898936 Health and social work .5048839 .0131626 38.36 0.000 .4790847 .5306832 Arts and entertainment .0165595 .03908 0.42 0.672 -.0600388 .0931578 Other services -.0235767 .0142508 -1.65 0.098 -.0515087 .0043554 HH's as employers, other undifferentiated goods -.0153606 .0119506 -1.29 0.199 -.0387844 .0080631 Activities of extraterritorial orgs .5067508 .0654974 7.74 0.000 .3783734 .6351282 type_occup_1 Professionals -.1405277 .0141443 -9.94 0.000 -.1682511 -.1128043 Technicians and associate professionals -.0838369 .0145667 -5.76 0.000 -.1123882 -.0552855 Clerks -.0664963 .0156369 -4.25 0.000 -.0971453 -.0358473 Service workers and shop and market sales workers -.1544525 .0142449 -10.84 0.000 -.182373 -.1265321 Skilled agricultural and fishery workers -.0670041 .0291018 -2.30 0.021 -.1240448 -.0099635 Craft and related trades workers -.2604403 .0142149 -18.32 0.000 -.288302 -.2325786 Plant and machine operators and assemblers -.2012962 .0149945 -13.42 0.000 -.2306859 -.1719065 Elementary occupations -.2054462 .014149 -14.52 0.000 -.2331787 -.1777137 educ Primary .0233479 .0054767 4.26 0.000 .0126133 .0340825 Lower secondary .0485597 .0060537 8.02 0.000 .0366942 .0604252 Upper secondary .0892388 .0054694 16.32 0.000 .0785185 .0999591 Post-secondary .183134 .0082231 22.27 0.000 .1670165 .1992516 male .0071837 .0036439 1.97 0.049 .0000414 .0143259 urban -.0032023 .0041255 -0.78 0.438 -.0112884 .0048837 _cons .2234143 .0153051 14.60 0.000 .1934157 .253413 Source: Own estimations based on LFS 2017–18 data. d. Having a job contract (out of sample prediction in the HIES 2018–19) Results. Based on the coefficients presented in table above from LFS and covariates from HIES, the estimation of “having a contract” was performed for those paid employees-main occupation in HIES. For defining “having a contract” = 1, the threshold was established as those employees with an estimated probability of having a contract equal or higher than 25 percent. This results in a formality rate in HIES 2018–19 of 25.4 percent similar to that of LFS 2017–18 (25.6 percent). Annex Table A2.4. Overall labor formality rate in HIES . ta formal [fw=int(weight)] formal Freq. Percent Cum. 0 25,313,527 74.58 74.58 1 8,627,716 25.42 100.00 Total 33,941,243 100.00 Source: Own estimations in HIES 2018–19 based on regression coefficients from LFS 2017–18. Furthermore, it was necessary to apply additional criteria for withholding tax compliers, since the total number of paid employees with a contract was an estimated at 8.6 million (see above), whereas withholding income taxpayers were 1,132,900 according to administrative data. To restrict further the sample of potential taxpayers among paid employees-main occupation, the final criteria were: (i) all public sector paid employees (predicted); and (ii) all private paid employees with predicted probability of having a contract equal or higher to 80 percent. This allowed a total number of potential PIT salaried taxpayers of 906,281, closer to the administrative data. 3.1. Definition Indirect effects are the (hidden or embedded) increases in prices of final outputs explained by taxed inputs or taxed intermediate goods. If producers pass on some of the input tax to final consumers, households bear a higher tax incidence beyond the direct effects (e.g., observable statutory rates that apply on final taxable items). Indirect effects depend on the input/cost structure of a good/service, and they are calculated via an Augmented Input-Output (IO) Matrix and the Cost-Push Model (see Section 4.2. on Methodology). In this study, it is assumed that indirect effects arise in the following circumstances of each indirect tax (see table below). In all these circumstances, the final seller is unable to claim the tax paid on inputs. Hence, to recover part of the costs, the seller will pass some of this input tax embedded on to final consumers in the form of higher final prices (embedded input tax or indirect effect). Annex Table A3.1. Indirect tax effects Indirect tax Circumstances where indirect effects apply GST - Exempt items - Taxable items not eligible for GST input tax credit - Informal items Excise tax - All items Source: Authors’ elaboration based on Jellema and Inchauste (2018) and Inchauste et al. (forthcoming). 3.2. Methodology The Cost-Push Model is a price-shifting model that allows to quantify the magnitude of sectoral changes in producer and retail prices resulting from any exogenous shock. For instance, using a cost-push model, the change in post-VAT prices ( ) depends on both direct and indirect effects: = + ∙ ∙ ( − ) − where is the vector of statutory VAT rates, is the matrix of technical coefficients in the Input-Output Matrix, and is the VAT exemption matrix. 3.3. Application in STATA • Data preparation: - The sectors in the IO Matrix need to be mapped to the COICOP sectors from the HIES 2018–19. - The symmetrical matrix of technical coefficients from the IO Matrix needs to be prepared. In the case of GST, for those sectors that have a mix of exempt and non-exempt items, an Augmented IO Matrix needs to be prepared. The Augmented IO is prepared by duplicating rows and columns of augmented sectors and the original technical coefficient is divided between the exempt and non-exempt IO sector based on consumption shares calculated with the household survey.165 - The indirect tax rates (GST, excise tax) need to be mapped to each sector in the IO Matrix. For those sectors with multiple rates, a simple average was calculated. • Running the Cost-Push Model in STATA - The goods with controlled prices are specified (e.g., final prices unaffected by indirect effects). - For each IO sector, the size of the price shock (dp) is defined as the indirect tax rate as a percentage of the final price including the tax (t/(1+t)). - The Cost-Push Model is calculated in STATA by running the costpush.ado developed by the World Bank Equity Policy Lab. 165 The households’ consumption shares for the exempt and non -exempt Augmented IO sectors are calculated based on the mapping between the IO sectors and the households’ product codes. This means that if the IO sector of “Fruits” needs to be augmented and if the IO -COICOP mapping suggests that 30 percent of total households’ consumption in “Fruits” is exempt and 70 percent is non-exempt, then these shares are used for augmenting the IO sector of Fruits. The calculation of households’ consumption shares of exempt vs non-exempt needs to be done for each augmented IO sector. Relative Incidence Annex Table A4.1. Relative incidence of direct taxes/transfers, indirect taxes/subsidies and in-kind health/education benefits Relative incidence of direct taxes (% of market income plus pensions) 0.4% 0.4% 0.3% 0.3% taxes as a % of mktypp 0.2% 0.2% 0.1% 0.1% 0.0% 1 2 3 4 5 6 7 8 9 10 Deciles by market income plus pensions (mktypp), real, per adult equivalent Withholding tax on salaries (1st occupation) Property tax Zakat payment to government Relative incidence of direct transfers (% of market income plus pensions) 7.0% 6.0% transfers as a % of mktypp 5.0% 4.0% 3.0% 2.0% 1.0% 0.0% 1 2 3 4 5 6 7 8 9 10 Deciles by market income plus pensions (mktypp), real, per adult equivalent BISP UCT transfer BISP CCT transfer Zakat Govt. transfer Urea fertilizer subsidy (direct effect) Agriculture tubewell subsidy (direct effect) Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Relative incidence of indirect taxes (% of disposable income) 9.0% 8.0% 7.0% 6.0% taxes as a % of disposable income 5.0% 4.0% 3.0% 2.0% 1.0% 0.0% 1 2 3 4 5 6 7 8 9 10 Deciles by disposable income, real, per adult equivalent GST on services (direct effects) GST on services (indirect effects) GST on goods (direct effects) GST on goods (indirect effects) Excises (direct effects) Excises (indirect effects) Total custom duties Withholding tax on telecommunications Relative incidence of Indirect Subsidies (% of disposable income) 4.0% 3.5% transfers as a % of disposable income 3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% 1 2 3 4 5 6 7 8 9 10 Deciles by disposable income, real, per adult equivalent Natural gas subsidy (direct effects) Domestic electricity subsidy (direct effects) Subsidy on agriculture tubewell (indirect effects) Subsidy on urea (indirect effects) Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Relative incidence of In-Kind Education Benefits (% of disposable income) 5.0% 4.5% transfers as a % of disposable income 4.0% 3.5% 3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% 1 2 3 4 5 6 7 8 9 10 Deciles by disposable income, real, per adult equivalent Pre-primary and primary education Secondary education Tertiary education Relative incidence of In-Kind Health Benefits (% of disposable income) 5.0% transfers as a % of disposable income 4.0% 3.0% 2.0% 1.0% 0.0% 1 2 3 4 5 6 7 8 9 10 Deciles by disposable income, real, per adult equivalent Outpatient health services Inpatient health services Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology Absolute Incidence Annex Table A4.2. Absolute incidence of direct taxes/transfers, indirect taxes/subsidies and in-kind education/health benefits Absolute incidence of direct taxes 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Withholding tax on salaries (1st Property tax Zakat payment to Govt. occupation) Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Absolute incidence of direct transfers 100.0% 90.0% 80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0% BISP UCT transfer BISP CCT transfer Zakat Govt. transfer Urea fertilizer subsidy Agriculture tubewell (direct effect) subsidy (direct effect) Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: Deciles ranked by market income plus pensions. Absolute incidence of indirect taxes 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Total Custom duties Total GST Total excises Withholding tax on telecommunications Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Absolute incidence of indirect subsidies 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Subsidy urea (indirect Agriculture tubewell (indirect Domestic electricity Natural gas subsidy effects) effect) Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: Deciles ranked by market income plus pensions. Absolute incidence of in-kind education benefits 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Pre-primary and primary education Secondary education Tertiary education Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Absolute incidence of in-kind health benefits 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Health outpatient services Health inpatient services Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10 Source: Authors’ calculations based on the HIES 2018–19, fiscal administrative data and the CEQ Methodology. Note: Deciles ranked by market income plus pensions. Withholding tax on electricity In 2018–19, total withholding tax collection on electricity consumption was PKR 35 billion, which is about 3.6 percent of all withholding tax collected. Withholding tax on electricity in Pakistan is deduced at the time of payment of other consumption charges on electricity by the consumers. For domestic consumers withholding tax rate is 0 percent if the amount of monthly bill is less than PKR 75,000 and 7.5 percent if the amount of monthly bill exceeds PKR 75,000. The rates vary for industrial and commercial consumers. We modeled it using HIES 2018–19 and none of the respondents had reported paying more than PKR 75,000 under monthly electricity expenditures. Therefore, it is not included in the final analysis. Withholding tax on education In 2018–19, total withholding tax collection on education was PKR 4.27 billion, which is about 0.4 percent of all withholding tax collected. Withholding tax on education in Pakistan is deduced by the educational institutions at the time of payment of fee. The tax rate is 5 percent if the amount of fee (inclusive of tuition fee and all charges received by the educational institution, by whatever name it is called, excluding the amount which is refundable) exceeds PKR 200,000. In the absence of required variables in the HIES, we could not differentiate the fee and other charges. The results were overestimated and therefore is not included in the final analysis. Fiscal intervention Data limitations/Recommendations Withholding tax on salaries The following variables were missing in the HIES for the three types of occupation: employer type (to differentiate federal public/provincial public/private sector); having a contract; and paying social security dummy. Property tax The analysis was based on self-reported Annual Rental Value (ARV). It is recommended that the variables required to calculate the ARV of both residential and commercial properties may be included in the HIES. These variables include locality of the house or commercial property, area, covered area, widow or government servant. Zakat contribution to the It is recommended that the separate variables may be added for zakat and usher. Government Customs duties We have used the Law of One Price, since we cannot identify imported goods in the survey (standard limitation of household surveys). GST/Excise tax Include the variable on “place of purchase” to be able to proxy informality in households’ purchases. BISP transfers (UCT/CCT) (i) The question on the “annual transfer from BISP program” should be more specific, perhaps could be disaggregated to ask the amount households received from UCT vs. the CCT transfer; (ii) there are two variables that need to be added in the HIES to allow for a better replication of the PMT (public sector employment information and district information); and (iii) the HIES could ask about number of families that lives in the household and are BISP beneficiaries, since the BISP definition of family is not aligned with the HIES and there can be more than one beneficiary in an HIES family. Public pensions The variable of “annual income from pensions” in the HIES should be disaggregated to ask differentiated questions for “public pensions” vs. “private pensions”. Transfer received from It is recommended to separate zakat from usher transfers in the HIES. zakat (government) Urea fertilizer subsidy In the HIES agricultural sheet: (i) Include questions on type of fertilizer (and quantity, imported/local); and (ii) we also needed the estimation of the counterfactual market prices of subsidized fertilizers. Agriculture tubewell In the HIES agricultural sheet: (i) disaggregate the questions on “water, electricity subsidy and fuel charges”; and (ii) include a question to ask if the tubewell is public or private. Fuel subsidy (petrol) HIES expenditure module includes the household expenditure on fuel but does not specify which type of fuel was used (High Speed Diesel [HSD], Motor Gasoline [MS], High Octane Blending Component [HOBC], Light Diesel Oil [LDO]), etc. It is assumed that the GST rate should be 17 percent therefore forgone revenue is calculated as the difference between the assumed GST rate and the implemented GST rate (fortnightly or monthly, whichever was applicable). . Education in-kind transfers (i) The federal PDSP expenditure was disaggregated by province but not by education level; it would be ideal to have the cross-classification; and (ii) the enrolment levels by level of education had to be distributed based on 2018–19 levels and 2017–18 provincial-education level shares. Health in-kind transfers (i) The health expenditure by type of health service was only disaggregated/available for current health expenditure (not capital) and for 2017–18 (not 2018–19); and (ii) we did not have admin data on patients by type of health service (inpatient/outpatient), so we had to use self-reported patients from the HIES.