1 Introduction

Taxes, government spending, and public transfers play a crucial role in advancing child rights and welfare and in reducing poverty and inequality. While there is increasing empirical evidence of the distributional effects of public finance in low- and middle-income countries, data and insights on the impacts on children are very limited.

However, it is essential to understand the specific impacts of public finance decisions on children because children have different demands and consumption patterns compared to adults, hence fiscal interventions may impact them differently. For example, children could be disproportionately affected if taxes add to the cost of goods and services particularly relevant for children. This is besides the indirect effects of consumption taxes, such as value-added taxes (VAT), and excises through their parents. Direct income taxes paid by adults could also have indirect effects on children’s welfare. Direct public transfers could have both direct and indirect effects on children’s wellbeing. Similarly, in-kind public transfers (spending on health and education) affect children’s school enrollment and access to basic health services. Moreover, household-level analyses often do not provide a full picture of the distributional effect of fiscal policy, and children may fare poorly in intrahousehold allocation (Dunbar et al., 2013). Recent evidence also suggests that many poor individuals do not necessarily live in poor households (Brown et al., 2017; Belete, 2021).

Children experience poverty differently from adults, as their experience of poverty is determined by material deprivations in the realization of child rights (e.g., health, education) rather than financial means (Alkire & Santos, 2013; Gordon et al., 2003). Therefore, a multidimensional assessment of poverty and wellbeing, in addition to monetary poverty, would be required. This also serves to highlight the essential role public spending on health and education could have on children’s schooling and health access. Finally, a comprehensive analysis of the distributional effects of current fiscal policies on children is essential to estimate the potential impact of policy changes. Simulations of new policy approaches or changes to existing tax or spending regimes provide a clearer understanding of the policy’s impact on poverty and inequality and provide policy makers with essential information to guide evidence-based decision-making on public finance.

The Ethiopia case study offers an opportunity to examine the fiscal space in an environment with high child poverty and high child undernutrition. In 2016, approximately 88% or 36.2 million children in Ethiopia were multidimensionally poor, meaning they were deprived of the fulfillment of multiple rights or needs for basic food or services (CSA and UNICEF Ethiopia 2018). And despite progress over the previous decades, the most recent Demographic and Health Survey shows that childhood stunting in Ethiopia is still as high as 37%, which makes it among the highest in the world (EPHI and ICF 2021).

This study investigates the effects of public transfers, services, and taxes on children’s wellbeing in the context of a sub-Saharan African country, and simulates the impact of potential policy choices aimed at alleviating child poverty. Specifically, the study answers the following questions:

  1. 1.

    How do the burdens of taxation and the benefits from government transfers and spending differ between children living in rural and urban settings, boys and girls, as well as between poorer and richer children?

  2. 2.

    What do government transfers, spending, and taxes contribute to the reduction of child monetary and multidimensional poverty and inequality?

  3. 3.

    What effects do potential changes to the social transfer system have on poverty and inequality among children?

The study applies the Commitment to Equity for Children (CEQ4C) methodology, which is an extension of the Commitment to Equity approach (Cuesta et al., 2021). The methodology compares welfare indicators before (pre-fiscal) and after (post-fiscal) taxes and/or transfers, and ultimately evaluates the distributional effects of fiscal policy (Inchauste et al., 2017; Lustig, 2018). This study specifically examines how children in Ethiopia are affected by fiscal actions following a recent cohort of studies that extended the CEQ4C method to children (Cuesta et al., 2021; Save the Children, 2021, 2022; Bornukova et al., 2020). As a result, in this study, individual children, rather than households (as is often the case in fiscal incidence analyses), are the unit of analysis. The study primarily employs data from Ethiopia Socioeconomic Panel Survey (ESPS) 2018/19, an LSMS-ISA survey which also collected data on certain taxes including business taxes, land use fees, and agricultural income taxes. We also integrate these survey data with administrative data obtained from various ministries and their subsidiary agencies.

The study finds that the fiscal system is progressive, poverty-reducing and equalizing. However, those results vary when the available information is further disaggregated. An analysis by tax type shows that direct taxes are progressive while indirect taxes are regressive. Moreover, indirect taxes account for more than two-thirds of taxes relevant to children. On the transfer side, direct and indirect in-kind transfers are progressive. Transfers are predominantly indirect in-kind transfers, with education spending being by far the largest in-kind transfer. Primary education spending is progressive, while secondary education spending is regressive across levels of child deprivation. Public spending on health is progressive as well. The study does not find significant differences in incidence by gender, however in rural areas, primary education and health spending are neutral, and do not show the progressivity seen in urban areas or overall.

Combining both taxes and transfers, the study finds a 21% decrease in the poverty headcount, i.e., from 33% at market income to 26% at final income, a 33% decrease in the poverty gap, and a 17% decrease in monetary inequality. The poverty effect is stronger for girls than boys. Similarly, poverty rates decline at a relatively higher rate for children in rural areas than in urban areas. Those findings show that the overall fiscal system (including in-kind benefits) reduced inequalities in poverty rates between boys and girls, as well as between rural and urban children. However, this overall decrease is driven by in-kind fiscal transfers – mainly government spending on education and health. Excluding these in-kind transfers shows that the fiscal system is not well calibrated to reduce poverty, since poverty rates increase for all groups between market income and consumable income. Only the significant in-kind transfers for education and health result in a decrease in the poverty headcount at final income. This highlights not only the essential role of those public services to deliver on fundamental child rights, but also the importance of investments in education and health to reduce poverty.

We also estimate the potential welfare impacts of four fiscal policy simulations that are relevant to children in Ethiopia, including providing universal public education and changes to benefit levels or distribution of the productive safety net program (PSNP). Across the various specifications, we find significant decreases in poverty headcount and inequality among children. More importantly, we also illustrate the various distributional effects of such changes by gender and location.

This study contributes to the existing literature in multiple ways. First, for the first time, this study estimates fiscal incidence for children in Ethiopia. Recent studies on fiscal policy and wellbeing in Ethiopia looked at the distributional effects of taxes and transfers at the household level (Hill et al., 2017; Mogues, 2013; Tesfaye & Gao, 2020). Most recently, Ambel et al. (2022) used individual-level data to investigate differences in the welfare impact of taxes and government spending on men and women in Ethiopia. However, this study is the first of its kind to analyze fiscal incidence specifically for children in Ethiopia, thereby contributing essential insights into a country with a high prevalence of child poverty. By identifying and assigning public transfers and spending associated specifically with children (such as education, vaccinations, and cash transfers), a child-specific CEQ assessment also gives precise impact estimates.

Second, it adds important empirical evidence to the limited research on fiscal incidence for children in low- and middle-income countries. To the knowledge of the authors, there are currently only four published CEQ4C assessments, covering Uganda, Kenya, Belarus, and Indonesia (Cuesta et al., 2021; Save the Children, 2021, 2022; Bornukova et al., 2020).

Third, this study applies an intersectional approach when analyzing the effects on children, as it systematically highlights differences between boys and girls and children in rural and urban areas, as well as the intersection between both. By doing so, it aims at contributing to the literature on intrahousehold allocation between children.

Fourth, insights from this study are directly relevant for policymakers, development practitioners, civil society organizations in Ethiopia, and beyond. The study examines the current impact of public finance on children, and highlights areas where government spending has the largest impact on reductions in child poverty and inequality. Furthermore, the findings build the basis to analyze the distributional effects of future fiscal policies on children, as illustrated by selected policy simulations.

The rest of the paper is organized as follows. Section 2 presents the methodology. Section 3 describes the data. Section 4 presents and discusses the results. Section 5 illustrates the kind of insights that can be generated via policy simulations on the basis of fiscal incidence analyses. Section 6 summarizes key insights and policy remarks.

2 Methodology

2.1 Measuring Fiscal Incidence

The analytical framework uses the CEQ methodology (Lustig, 2018) as well as its child-specific version CEQ4C (Cuesta et al., 2021) to estimate the distributional impact of fiscal policy on children’s wellbeing. The approach begins with calculating pre-fiscal and post-fiscal income concepts by assigning public transfers, spending, and taxes. Four income concepts are considered: market income, disposable income, consumable income, and final income (Fig. 1). The analysis then estimates monetary and multidimensional child poverty and inequality at different income concepts.

Fig. 1
figure 1

Income concepts of the CEQ framework. Source: Adapted for country context from Lustig (2018)

In this study, the individual child is the unit of analysis. The construction of most other variables and income concepts closely follows those described in Ambel et al. (2022). Disposable income is proxied by consumption expenditure in the underlying household survey data. Other income concepts are therefore computed through backward and forward calculations. Individual level expenditure is estimated based on intra-household resource allocation (Belete et al., 2019; Calvi et al., 2023) and equivalence scales (Browning et al., 2013). The allocation approach of expenditures to household members is based on consumption patterns and the availability of individual-specific information in the data. For example, the 2018/19 Ethiopia Socioeconomic Panel Survey (ESPS) collects clothing expenditures for boys and girls, as well as individual expenditures on education and health. Some expenditures, such as alcoholic drinks and cigarettes, are assignable only to adults. Non-assignable expenditures are allocated to each child based on equivalence scales.

Various assumptions, following those made in Ambel et al. (2022), are needed when assessing the fiscal incidence for children. The study assumes that each student enrolled in a public school in each region receives the education benefit per pupil. Education costs for each region are calculated by dividing the total spending by the number of primary and secondary students enrolled. We exclude spending on tertiary education, as they generally serve the non-child population. For health spending, the per-beneficiary benefit is obtained by dividing total health spending by the number of public health service users. We use household survey data to estimate the population of public health service beneficiaries by region and national level. Total government spending on education and health is used to monetize in-kind transfers, and copayments (fees or contributions) are deducted where applicable. The 2016/17 regional and federal spending data are used to estimate the cost of providing health services and primary and secondary education. We derive missing data for 2018/19 by deflating the 2016/17 data using the average annual growth rate of spending for each region.Footnote 1

Tax burdens borne by parents or the household are passed on to children. Indirect taxes on purchased consumption items identified in the household survey are simulated using the 2015/16 social accounting matrix (SAM) framework (Mengistu et al., 2019). For such a purpose, this paper uses a similar list of consumption items annexed in Ambel et al. (2022). Once the price burden of all goods and services is calculated using their effective tax rate, the price burden on consumers resulting from indirect taxes paid for inputs of production is computed to estimate how taxes on petroleum and coal affect the prices of final goods and services. Second-round tax effects are estimated for items exempt from VAT. With regards to indirect subsidies, those on wheat in urban areas and kerosene nationallyFootnote 2 are estimated based on the household’s expenditures on these items.Footnote 3

The study has the following limitations that are relevant to fiscal incidence analysis. First, not all fiscal instruments are included in this study due to either lack of data or difficulties in assigning them to individuals. Corporate taxes and government spending on infrastructure are not included. Second, the analysis does not consider differences in service quality. However, the quality of schools, clinics, hospitals as well as their staff varies in rural and urban areas and in small and big towns.

2.2 Measuring Monetary and Multidimensional Poverty Impacts on Children

The impact of the fiscal policy instruments on poverty is assessed by analyzing the changes in child monetary and multidimensional poverty indices at the different income concepts. Monetary poverty is measured using the FGT indices (Foster et al., 1984),

$${P}_{\propto }= \frac{1}{N}\sum_{n=1}^{M}{\left(\frac{z-Y_i}{z}\right)}^{\propto }$$

where \(\propto\) measures poverty aversion so that \(P_0\),\(P_1\) and \(P_2\) provide poverty headcount, gap, and severity respectively; \(N\) is the total number of children; \(M\) is the number of poor children; \(Y_i\) represents any of the six income concepts; and \(z\) is the poverty line.

However, as discussed above, measuring child wellbeing using only monetary indicators is inadequate. The multidimensionality of wellbeing is crucial for the measurement of non-monetary indicators of child wellbeing, both in the short and long-run. Multidimensional poverty can be measured in different ways, each involving challenges on which dimensions to include, weights, aggregation of dimensions, and cut-offs. In fact, previous CEQ4C assessments have used different multidimensional poverty measurements, depending on individual preferences, available data in the country, or existing definitions already used by governments (Cuesta et al., 2021; Save the Children, 2021; Bornukova et al., 2020).

This study adapts the AF methods (Alkire & Foster, 2011; Alkire & Santos, 2014) to measure multidimensional child poverty. Based on the literature and data availability, four dimensions (child education, child health, water and sanitation, and housing and assets) and ten indicators are used to construct the multidimensional child poverty index (Table 1). Indicators of child education and child health dimensions are specific to each child. Those in the water and sanitation and housing and assets dimensions are common to household members but have implications for children. Weights in multidimensional poverty measurement are debatable (Ravallion, 2011; Thorbecke, 2011). For the sake of producing comparable estimates of child poverty and arguing that child rights are equally important, UNICEF (2021) advocates for equally-weighted indicators. The paper thus applies equal weights as in Cuesta et al. (2021).

Table 1 Dimensions, Indicators, and Deprivation Thresholds of Multidimensional Child Poverty

For the identification of the multidimensionally-poor children, the AF dual cut-off approach is employed. The first cut-off, also called deprivation cut-off for each indicator, is based on national and international standards. The second cut-off, also called multidimensional cut-off, is being deprived in at least 33% of the weighted deprivations (Alkire & Santos, 2014; Belete, 2021; Bruck & Kebede, 2013).

Aggregating into multidimensional poverty indices then follows. The deprivation count or sum of weighted deprivations \(C\) for each child \(i\) is

$$C=\sum_{j=1}^{D}{w}_{j}{I}_{\left(\mathrm{0,1}\right)}\left({y}_{ji}\le {z}_{j}\right)$$

where \({w}_{j}\) is the weight of indicator \(j\), and \(D\) is the total number of indicators. A child is identified as multidimensionally-poor if she is deprived in at least 33% of the weighted deprivations, i.e., \({C}_{i}\ge 0.33\). Using this cut-off \(k\), multidimensional poverty headcount ratio (\(H\)) is

$$H=\frac{1}{N}\sum_{i=1}^{N}{I}_{\left(\mathrm{0,1}\right)}\left({C}_{i}\ge k\right)$$

The weighted deprivations as a proportion of the maximum of the weighted deprivations suffered by the multidimensionally-poor children give the average intensity of deprivations as

$$A=\frac{1}{N*D*{h}_{j}}\sum_{i=1}^{N}{I}_{\left(\mathrm{0,1}\right)}\left({C}_{i}\ge k\right)*{C}_{i}$$

Finally, the adjusted multidimensional poverty index is given as \(M=H*A\).

The effect of the fiscal system for children is ultimately assessed by analyzing the incidence of the various fiscal interventions over the space of multidimensional deprivations and expenditure quintiles. The changes in poverty headcount, gap and severity at the different income concepts are also analyzed. Concentration coefficients and Kakwani indices for progressivity and pro-poorness of taxes and transfers are alternatively employed.

2.3 Measuring Inequality Impacts

For gauging inequality, the study uses the Theil index which is a family of the generalized entropy inequality measures. The Theil index is given by

$$I=\frac{1}{N}{\sum }_{i=1}^{N}\frac{{Y}_{i}}{\overline{Y} }ln\left(\frac{{Y}_{i}}{\overline{Y} }\right)$$

where \({Y}_{i}\) is the income of child \(i\); \(\overline{Y }\) is the average income; and \(N\) is the number of children. \(I\) varies from 0 (perfect equality) to \(ln\left(N\right)\) (maximum inequality). One advantage of the Theil index is that it has the property of additive decomposability into inequality within and between subgroups. For gender, the total inequality is the sum of within-child-gender inequality and between-child-gender inequality. The within-child-gender inequality is \({I}_{w}={\sum }_{g=1}^{h}{S}_{g}{I}_{g}\), and the between-child-gender inequality is \({I}_{b}={\sum }_{g=1}^{2}{S}_{g}\left(ln\left(\frac{{S}_{g}}{{P}_{g}}\right)\right)\), where \({S}_{g}=\frac{{\sum }_{j=1}^{{N}_{g}}{Y}_{j}}{{\sum }_{i=1}^{N}{Y}_{i}}\) is gender g’s income share of total income, \({P}_{g=}\frac{{N}_{g}}{N}\) is the share of the child gender g’s population from the total child population. The same decomposition formula is applied to children’s residence (rural/urban). To evaluate how the fiscal system affects inequality among children, we calculate the inequality indices and compare for each income concept.

3 Data

Fiscal incidence analyses such as those presented in this study rely on integrating two sources of data. First, administrative data mainly provide key insights into public revenues and expenditures, but can also provide information on subsidy schemes, transfer systems and users of public services. Second, household survey data is crucial in identifying individuals, both as taxpayers for different kinds of taxes and as users of publicly funded services. The next subsections present details on these datasets used in this study.

3.1 Administrative Data: Taxes and the Child-Relevant Budget in Ethiopia

Administrative data used in this study include the following: (i) public revenue and expenditure data for the 2018/19 fiscal year and regional education and health spending from the Ministry of Finance, (ii) school enrollment information from the Ministry of Education, (iii) kerosene subsidy from the Ethiopian Petroleum Supply, and (iv) wheat subsidy from the Ethiopian Trading Businesses Corporation.

Table 2 shows Ethiopia’s tax revenues in 2018/19. Revenue collection was equivalent to 13.5% of GDP; 43% of those through direct taxes (mainly business profit tax followed by personal income tax) and 57% through indirect taxation. The last two columns show the tax burden per child. Domestic indirect taxes are the most important followed by personal income taxes and business profit tax.

Table 2 Annual Tax Revenues, Share of GDP and Per Child Burden, 2018/19

Ethiopia’s 2018/19 public spending with child-relevant components is shown in Table 3. Thirty-nine percent of government spending goes towards social development, followed by economic development (33%) and general services.

Table 3 Annual Government Spending, Share of GDP and Child Benefits, 2018/19

3.2 Survey Data: Consumption, Utilization of Services and Child Poverty

The survey data are from the 2018/19 Ethiopia Socioeconomic Panel Survey (ESPS). ESPS is a nationally representative survey implemented by the Ethiopia Statistical Service in collaboration with the World Bank under the LSMS-ISA project. The survey interviewed 6,700 households out of which 4,992 households had at least one household member between 0-17 years old at the time of the interview. A total of 13,820 members in this age group are included in the analyses.

Table 4 presents the descriptive statistics of the children included in this study. The profile shows that both boys and girls have similar demographic and socioeconomic characteristics. The average age is about 8.5 years. The household size is over the national average because this sub-sample of households includes only those with children. About one in five children live in urban areas, and the share is slightly higher for girls. The profile, however, differs by place of residence. For example, children in rural areas are way more deprived than those in urban areas. This difference is strongly associated with child deprivations in housing conditions including water and sanitation facilities, access to electricity, number of rooms per household member, and access to information.

Table 4 Descriptive Sstatistics of the Sample

In this study, survey data are not only used to estimate the incidence of spending and revenue raising activities but are also the basis on which poverty and inequality amongst children are estimated. Table 5 shows the results of multidimensional child poverty, monetary poverty and inequality indices. On average, children are deprived in about 3.5 out of 10 measures of multidimensional deprivations included in this study. The indicator is similar for both boys and girls. On average, urban children are deprived in 1.5 measures, compared to about 4 for rural children. Over half of children are multidimensionally poor with no boy/girl differences. This incidence reaches as high as 66% and 10%, respectively, for rural and urban children. Over a third of children are monetarily poor with girls slightly poorer than boys. Though inequality is generally low, within-group inequalities outweigh between-group inequalities. Monetary child poverty and inequality profiles show substantial rural-urban differences.

Table 5 Overall Estimates of Child Poverty and Inequality, by Gender and Location

Analyzing multidimensional and monetary child poverty separately hides the significant differences between these measures, while the intersection could provide refined information. Across the whole sample, 24% of children are poor, both multidimensionally and monetarily. A larger share of children (29%) is multidimensionally poor despite not living below the monetary poverty line. In fact, many of those children live in the richest quintiles (44% of children in the fourth and 31% of children in the fifth quintile are found to be multidimensionally poor). Likewise, 9% of children are found to be monetarily poor, but not considered to be multidimensionally deprived.

4 Results and Discussion

4.1 Fiscal Incidence

4.1.1 Incidence Across Child Multidimensional Poverty

Taxes or public spending do not affect all children equally. Analyzing the fiscal incidence will allow us to differentiate the effects by multidimensional poverty, providing important insights into the potential role individual fiscal policies can play in reducing poverty and inequality for children.

Figure 2 compares the distribution of taxes and transfers as a percent of market income by child multidimensional deprivation counts. In all scenarios (all children, children by gender or location), direct taxes are progressive, i.e., their value relative to market income decreases with the average number of child deprivations. For example, direct taxes constitute 6% of market income among households of non-deprived children, while this is only 1% when a child experiences five or more deprivations. Indirect taxes, comprising of VAT and excise taxes, are regressive. There are also differences by gender. The share of both direct and indirect taxes relative to market income is slightly higher for girls (with one or two deprivations) than for boys in the same situation. However, the reverse is the case for children with four or five deprivations. Rural-urban differences also exist. In rural areas, we find that both direct and indirect taxes are low across multiple child deprivations, i.e., they are neither progressive nor regressive. For urban children, excise taxes are regressive.

Fig. 2
figure 2

Taxes and transfers as percent of market income by child deprivation status. Source: Authors’ calculations; Data from ESPS 2018/19. Note: Shaded areas represent 95% confidence intervals

Transfers are almost completely in the form of indirect in-kind transfers, with direct transfers accounting for only 0.3% (in the case of no deprivations) to 1.9% (in the case of five or more deprivations) of market income. Direct transfers are relatively equal across the various disaggregation groups, although they are slightly higher for urban children with four or more deprivations than similarly deprived children in rural areas. Primary education is the largest in-kind transfer and is progressive overall, constituting about 3.9% of market income of non-deprived children and rising to 13.2% for those with four deprivations. Though no differences exist between boys and girls, public spending on primary education is progressive in urban areas while neutral in rural areas. On the other hand, secondary education is regressive overall and in all disaggregation groups.

These findings are in line with recent studies in Kenya (Save the Children, 2021) and Uganda (Cuesta et al., 2021). We also find that healthcare is generally progressive among urban children and neutral among their rural counterparts.

In all cases, the pre-fiscal income (market income) is higher than the disposable and consumable income (see Tables 9 and 10 in Annex A). However, this changes when indirect in-kind transfers (government spending on education and health care services) are considered. Comparing the pre-fiscal income (market income) and final income, the incidence is negative only for children without any deprivations, i.e., post-fiscal income is less than pre-fiscal income. In all other scenarios, where there is at least one deprivation, a child received more transfers and subsidies than paid in taxes and co-payments. This holds for both boys and girls and for children in rural and urban areas. An exception is in rural areas where final income is always greater than market income including the scenario of no deprivation (C=0) (Table 10).

4.1.2 Incidence Across Child Monetary Poverty

A similar picture emerges when analyzing fiscal incidence results across relative monetary poverty measures – consumption expenditure quintiles (see Table 11 and Table 12 in Annex A). In all scenarios, disposable and consumable income are lower than that of the pre-fiscal or market income, i.e., direct transfers and consumption subsidies did not fully compensate for the effect of direct and indirect taxes. However, when in-kind transfers (government spending on education and health) are added, income increases across the board for all but children in the richest quintile.

Similar to the fiscal incidence analyzed above for multidimensional poverty, indirect taxes (VAT and excises) are mostly constant as a share of market income between the poorest and richest quintiles. In contrast to levels of deprivations though, direct taxes are not progressive when income groups are measured. Instead, they are highest for the poorest quintiles at 3.3% of market income, 1.4% for the middle quintile, and 2.6 % for the richest quintile. This u-shape is also slightly more pronounced for boys than for girls. Furthermore, in rural areas, direct taxes as a share of market income are low, and continue to decrease as households get richer, therefore making them consistently regressive. In urban areas, the poorest households face the relative highest burden of direct taxes by far, with 16.3% of market income, compared to 3.5% for the middle quintile and 4.1% for households in the richest quintile.

When it comes to transfers, the analysis for monetary poverty mostly mirrors that of multidimensional poverty; all transfers, with the exception of secondary education, are uniformly progressive and the transfers as a share of market income decrease as households get richer. The analysis reveals some gender differences similar to the analysis for multidimensional poverty: spending on primary education as a share of market income represents a larger share for poor girls compared to poor boys. Also, secondary education spending represents a slightly larger share for girls in the richer quintiles (compared to boys in higher quintiles). Primary education spending is progressive for rural children, representing 20.1% of market income for children in the poorest quintile and 6.7% for children in the richest quintiles. While the overall trend is similar for urban children, primary education transfers represent a smaller share of market income for the poorest children in urban areas compared to the second quintile. Secondary education spending does not benefit the poorest children in rural areas (representing only 0.1% of market income) and is relatively constant across the other quintiles at around 2% of market income. In contrast, in urban areas, transfers on secondary education represent 5.2% of market income for the poorest quintile and 4.3% in the richest quintile, but this is most pronounced for children in the middle quintile (7.8% of market income). These progressivity and pro-poorness results are also confirmed by Kakwani indices and concentration coefficients (see Tables 13 and 14 in Annex B).

4.2 The Effect of the Fiscal System on Poverty and Inequality

A thorough understanding of how fiscal policies affect poverty and inequality requires an analysis of the distributional effects of the full fiscal system, i.e., combining revenue-raising activities with public spending and transfers. In line with the CEQ methodology, this can be achieved by determining poverty and inequality at different income categories (Lustig, 2018) and following individuals through the various steps of the fiscal system (see also Figure 1). While this provides crucial insights into the distinct role of different fiscal policies to reduce monetary poverty, this approach unfortunately does not allow for a similar analysis of multidimensional poverty. Tables 6 and 7 show key statistics on monetary poverty and inequality for all four income concepts.

Table 6 Monetary Child Poverty Across Income Concepts, by Gender and Location
Table 7 Monetary Child Inequality Across Income Concepts, by Gender and Location

First, market income (pre-fiscal income) includes private market or non-market earnings, e.g., what families earn through employment (before tax), any pensions, or other income they may receive (remittances, interests on savings etc.). At this stage, we find 33% of children live in monetary poverty, with poverty rates a little higher for girls than boys (36% versus 31%). Children in rural areas are significantly more likely to be poor than those in urban areas, with poverty headcounts of 39% and 14%, respectively. Monetary inequality (as measured by the Theil index) is 0.32 for all children. Inequality is higher for girls and children in urban locations.

Second, disposable income is derived by adding direct transfers (PSNP and non-PSNP) and subtracting direct taxes (e.g., income tax, agriculture income tax and land use fee, property tax) from market income. Poverty headcounts remain broadly constant to those at market income, increasing by one percentage point for all children and those living in urban areas. This is partly due to the lack of progressivity in direct transfers: while direct taxes are progressive overall (i.e., the tax burden increases as households are getting richer), direct transfers are lowest for the poorest 20% of children and highest in the middle quintile. Monetary inequality decreases slightly between market and disposable income.

Third, consumable income adds indirect subsidies (kerosene and wheat subsidies) to disposable income and subtracts indirect taxes (VAT and excise). With indirect taxes significantly higher than direct taxes (although broadly progressive), and indirect subsidies being both small as well as benefiting mostly richer households, the fiscal system up to this point leads to an increase in poverty headcounts across all groups included in this analysis. This results in a 9% increase in poverty headcounts (and 8% increase in the poverty gap) between market and consumable income. This increase is slightly more pronounced for boys than for girls (10% vs. 8%). The highest increase in relative terms can be found in urban areas, where the poverty headcount increases by 21% and the poverty gap by 25% between market and consumable income. Also noteworthy is a significant increase of the poverty gap for girls (17% increase between market and consumable income). This contrasts with small decreases in monetary inequality for almost all groups, with an average decrease of 2.5% in the Theil index.

In other words, the combined effect of fiscal policies in Ethiopia (taxes and direct transfers) increases poverty among children when comparing market and consumable income. While this finding is similar to the observation made for the Kenya (Save the Children, 2021), it is relatively uncommon when compared to most other countries in which similar studies have been carried out.Footnote 4 Finally, these combined effects do not incorporate benefits from education or healthcare, as those cannot be directly used to reduce monetary poverty. However, if we monetize the value of in-kind services in education and health, and subtract co-payments and user fees (as done when computing the final income), those amount to the largest contributions to monetary child poverty, reducing the poverty headcount to 26% for all children. This represents a 21% decrease in the poverty headcount from market income to final income, and a 33% decrease of the poverty gap. The effect is even stronger for girls than for boys (a decrease of 25% between market income and final income for girls, compared to a 23% decrease for boys). Similarly, poverty rates decline relatively more significantly for children in rural areas (23%) than those in urban areas (14%). Those findings suggest that the overall fiscal system (including in-kind benefits) leads to convergence, i.e., reducing inequalities in poverty rates between boys and girls as well as rural and urban children. This is somewhat mirrored in the fiscal system's impact on monetary inequality; while monetary inequality for all children decreases by 17% between market income and final income, those decreases are more pronounced for girls over boys and for rural children over their urban peers. As inequality was more pronounced between urban children, this slightly increases the relative gap between inequality in rural and urban areas.

In summary, this analysis suggests that the overall fiscal system is not well calibrated to reduce monetary poverty, with poverty rates increasing for all groups between market income and consumable income. Only the significant in-kind transfers for education and health result in a decrease in the poverty headcount at final income. This highlights not only the essential role of those public services to deliver on fundamental child rights, but also the importance of investments in education and health in reducing poverty.

5 Policy Simulations

Fiscal incidence analyses – such as this particular study for Ethiopia – do not only provide crucial insights into the current impact of the fiscal system on poverty and inequality, they are also an integral in simulating the potential effects of new fiscal policies or changes to the existing tax and spending regimes. They are therefore useful for providing policy makers with important information on fairness and effectiveness of public finance proposals. This study conducts four fiscal policy simulations that are relevant to children in Ethiopia, focusing on universal education and the flagship productive safety net program (PSNP) (Table 8). The simulations are chosen based on their policy and political relevance (e.g., achieving universal education), periodical discussions such as on improved targeting (e.g., PSNP retargeting), and analytical relevance through multiple policy simulations (e.g., PSNP coverage and retargeting). These criteria are similar to those used by Cuesta et al. (2021).

Table 8 Welfare Effects (in Percentage Change) of Child-Relevant Policy Simulations in Ethiopia

We estimate the fiscal impact of universally enrolling out-of-school children in Ethiopia’s public educational institutions in Simulation 1. Such a policy change will have an impact, both on monetary child poverty and inequality, as well as on our multidimensional poverty index, since child enrolment is included as an indicator in that measure. If all school-age children who are currently deprived of education are enrolled, the multidimensional child poverty headcount ratio would decrease by 2.2%, with a larger effect of 3.2% on monetary child poverty headcount ratio.

These represent moderate decreases in the multidimensional poverty headcount rate from 53.7% to 52.5%, and monetary headcount rate from 25.8% to 25%. The multidimensional poverty reduction is slightly higher among boys (2.6%) than girls (2.1%). This is reversed for monetary poverty, with larger effects seen among girls. For both monetary and multidimensional child poverty, the effects of closing the educational gap are higher for children in urban areas than in rural areas. Monetary child inequality measured by the Theil index also falls by 0.9% from its pre-simulation value of 0.267. The inequality reduction also shows that universal education is more equalizing for boys than girls. Excluding its additional administrative infrastructural costs, this policy change of enrolling currently unenrolled students would cost the government ETB 4.2 billion. For Uganda, Cuesta et al. (2021) simulate that universal education would reduce multidimensional and monetary child poverty headcount ratios by 2.5 and 1.3 percentage points, respectively.

Simulations 2 through 4 focus on PSNP transfers: While in Simulation 2 the amounts of PSNP transfers are doubled for all beneficiaries, they are retargeted to children found to be monetarily-poor at final income in Simulation 3. Simulation 4 combines both simulations simultaneously. The fiscal costs of each simulation are also computed.

Doubling the amount of PSNP transfers to existing beneficiaries (Simulation 2) reduces monetary poverty by 1.9% and inequality by 0.9%, with higher effects for girls than boys. Given that PSNP is currently rural-oriented, the effects of doubling its transfers are also more pronounced among children in rural areas than in urban areas. The overall cost of this fiscal action is estimated to be ETB 3.1 billion. Simulation 3 considers a scenario where PSNP transfers that reach non-poor children (estimated as ETB 2.5 billion) are redirected and equally distributed to monetarily poor children based on their status at the final income. This policy – at zero fiscal cost – amounts to a reduction in monetary poverty by 6.9% and inequality by 2.5%, with almost no difference between boys and girls. The poverty reduction effect of redistribution transfers from non-poor to poor children is larger for urban than for rural children, though more equalizing for rural children. The effects of retargeting PSNP transfers are substantially larger than those that result from doubling them. The joint welfare effects of doubling and retargeting PSNP transfers (Simulation 4) are sizeable, as monetary child poverty would fall by 8.8%. That translates to a slight reduction in the child poverty headcount rate from 25.8% to 23.5%. With this joint policy change, girls and children in urban areas would benefit slightly better than boys and those in rural areas. The policy change is also associated with the largest drops in inequality for all children (3.1 %) and other child groups, compared with other simulations.

Two assumptions are made a priori about the simulations. First, the extra transfers can induce a reduction in the labor supply by beneficiaries. Such changes in behavior are ruled out by our simulations. Second, the simulations do not take into account the additional administrative costs related to enrolling currently unenrolled students and increasing the amount and redistribution of PSNP transfers.

6 Concluding Remarks

The study investigates the fiscal space for children in Ethiopia using the Commitment to Equity for Children (CEQ4C) methodology. The analysis is based on 13,820 children (0-17 years old) from the 2018/19 Ethiopia Socioeconomic Survey. Individual and household level information collected from the survey is combined with budget figures and administrative data on programs and subsidies. The study then examines the burdens of taxation and the benefits from government transfers and spending in rural and urban settings, boys and girls, as well as poorer and richer children. It also analyzes the effect of these taxes and transfers on poverty and inequality.

The incidence analyses show that the fiscal system on average is progressive and mainly driven by direct taxes and indirect in-kind transfers. However, important differences in the distribution of some of the elements of taxes and transfers exist. For example, indirect taxes are regressive while public spending on primary education is by far the largest in-kind transfer and is generally progressive across levels of child deprivation. Secondary education spending is regressive, while public spending on healthcare is progressive across all children. However, in rural areas spending on primary education and health is neutral, in sharp contrast to strong progressivity in urban areas. Regarding impacts on poverty and inequality, the fiscal system reduces poverty by 21% from market income to final income, and the poverty gap by 33%. The effects are stronger for girls and children in rural areas than for boys and those living in urban areas. However, this is only the case once the significant in-kind transfers for education and health are considered. Poverty rates increase between market income and consumable income, which implies that the overall fiscal system up to this point has impoverished both boys and girls. The findings in this study highlight the fact that public services are not only essential in delivering fundamental child rights, but also in reducing poverty amongst children.

Child-focused fiscal incidence analyses provide essential insights into the distribution of taxes, direct transfers and public spending, and allow for a better understanding of the impact of fiscal policies on poverty and inequality amongst children. These insights are relevant for a wide range of decision makers, including policy makers in local and national governments, international financing facilities and other multilateral organization, as well as civil society organizations. Furthermore, indicators on both pro-poor public spending on social services as well as the distributive impacts of fiscal policies are now part of the global indicator framework for the Sustainable Development Goals.

Finally, while this study offers an analysis of fiscal incidence in 2018/19, CEQ4C assessments can be used to simulate the effects of potential policy interventions, offer an important toolkit to assess the effects on poverty and inequality of new policy proposals. Our four fiscal policy simulations that focus on universal education and the PSNP improve child welfare. Closing the education gap in Ethiopia in particular is associated with modest reductions in monetary inequality as well as multidimensional and monetary poverty, with varying gender and location effects. PSNP transfers, if doubled, would have a modest reduction effect on monetary poverty and inequality. PSNP transfers, if redistributed from non-poor to poor children, would have larger poverty and inequality effects. Doubling and redistribution jointly result in the largest welfare improvements for all children groups.