Long‑Term Spending Projections in Israel

2. Function-specific spending projections

Abstract

This chapter presents the detailed methodology and projections for key spending categories – education, health, social protection, and interest payments – which together account for about two-thirds of general government expenditure. Exogenous expenditure paths are applied to defence and transport, while all other spending items are projected under a constant-service-level assumption, linked to GDP growth and employment dynamics relative to population trends. In addition to the baseline scenario, the chapter examines how alternative demographic, and labour market integration assumptions affect each of the key spending categories. It also analyses the fiscal implications of selected reform scenarios.

2.1. Education spending

Overall government spending on education at pre-primary to tertiary levels amounts to 6 ½ percent of GDP after a slight decline in the last years (Figure 2.1). This share is higher than in most OECD countries, mainly due to a much younger population. Considering the population's age structure, the Bank of Israel found that spending on education in Israel is about 1% of GDP lower than the OECD average (Bank of Israel, 2019[1]). Additionally, the government spends about 0.25% of GDP on Early Education Development and 0.1% on subsidising studies in Yeshivas and Kollels, which are reported in COFOG under Social Protection and General Services and Religious Services.

Figure 2.1. Spending on public education as a share of GDP has declined recently

Education spending, % of GDP

Source: Israel Central Bureau of Statistics.

Government spending on education reflects demographic developments and other drivers. Demographic change determines the size of future cohorts of students. As mentioned in Chapter 1, the young population (0-17) is projected to increase by about 0.7% a year on average up to 2065, significantly slower than the prime working-age population (25-64, 1.4%). Other key drivers include the involvement of the government in the education system, the duration of mandatory education, the enrolment rates in upper secondary and tertiary education, relative wages in the education sector, the average class size and the scope of the special education system. In recent years, one of the most notable contributors to rising government expenditure has been the limited oversight of transitions from general education to special education frameworks (Blass, 2025[2]). Changes in the relative shares of different population groups would also affect spending, as the government's involvement in the education system varies between groups, and so does the willingness to pursue academic studies.

Consistent with the other expenditure items projected in this report, the education projections are conducted under a ‘no-policy-change’ assumption. Thus, the baseline scenario will primarily assess the impact of demographic change on public spending. Then, a set of sensitivity scenarios is conducted to quantify the responsiveness of the projection results to changes in key underlying assumptions and main policy measures.

2.1.1. Education spending projection methodology

The methodology used to project education expenditures is largely based on the EU Commission Ageing Report method (European Commission, 2023[3]), with several adjustments to Israel. It contains three main steps: (1) projecting the number of students in the system based on the demographic trends and assumptions regarding future enrolment rates; (2) projecting the number of budgeted students based on the current share of students in private and dependent institutions and the levels of government support; and (3) projecting per budgeted student expenditures growth, relying on assumed students to teacher ratio and macroeconomic developments. Figure 2.2 provides a visual description of the method. The projections were conducted separately for the main expenditure and educational categories: early childhood education, pre-primary and primary education, secondary education, post-secondary non-tertiary education, and tertiary education.

Figure 2.2. Schematic overview of the education model

Note: The rubrics are highlighted in colours according to the source used to project the figure: the latest available data (green), the macroeconomic projections based on the OECD Long-Term Model (blue), CBS’s demographic projections (pink) and assumptions regarding the main policy measures (orange).

Source: OECD elaborations.

2.1.2. Approach to projecting education spending

Projecting the number of future students

To project the number of future students, the model relies on the CBS’s mixed demographic projection described in Chapter 1 and estimated enrolment rates for each education level per single age. To simplify, it is assumed that enrolment rates in compulsory levels (pre-primary, primary and secondary education) and tertiary level, except bachelor’s, will remain constant at the level observed in 2022 (last available data points). Enrolment rates in the base year, by age and education level, are sourced from the UNESCO/OECD/EUROSTAT (UOE) database on education.

Future enrolment rates in early childhood education depend on projected female employment rates (aged 30-34). The assumed effect of a change in mothers’ employment is based on the difference between the employment rates of Arab and Jewish women aged 30-34 and the enrolment rates of children in each population group (to impute so-called ‘elasticity’). It implies that if Arab women's employment is equal to that of Jewish women, so will enrolment rates in early childhood education. The effect of this assumption on the enrolment of young Jewish kids is negligible, as both female employment and kids’ enrolment are already high. Expected enrolment rates at the bachelor’s level are estimated according to projected changes in the share of individuals with an academic degree (reported in the Labour Force Survey) by gender, age and population group. These projections were conducted using a simple cohort-based model to capture how an age group maintains or changes its education levels as time passes. So, if, for example, Arab females who were born in 1996 had a higher likelihood of holding an academic degree when they were 27 (this means in 2023), the model assumes that they will also have a higher chance of having an academic degree at older ages, compared to previous cohorts. Relying on current trends, the model projects a continuous increase in participation in higher education among Arab women. In contrast, for the other population groups, it projects that the young cohorts will participate as the cohorts born in the early 1990s. This means that participation among the Haredim (men and women) and Arab men will remain significantly lower than the rest of the population. These results are illustrated in Annex 2.A to this chapter.

Table 2.1 summarises the projection results for the number of students in the baseline scenario. Overall, the number of students is expected to increase by 42% up to 2065, with a slower increase in early childhood and primary education due to lower assumed fertility rates, despite an assumed rise in early childhood education enrolment in the Arab population.

Table 2.1. Projected number of students in the education system

Projected number of students by level (thousand)

	2022	2040	2065	Projected change from 2022 to 2065
Early childhood educational development	307	392	403	+31%
Pre-primary education	585	688	696	+19%
Primary education	1,003	1,241	1,348	+34%
Lower secondary education	457	597	680	+49%
Upper secondary education	426	589	680	+59%
Post-secondary non-tertiary education	15	21	26	+73%
Tertiary education, total	408	530	626	+54%
Bachelor’s or equivalent level	266	337	374	+40%
In total, all levels	3,467	4,395	4,833	+39%

Source: Actual 2022 numbers and OECD elaborations.

Projecting the number of budgeted students

The number of budgeted students is projected based on the demographic projections, the shares of students in private, public and dependent institutions, and the level of government support for each type of institution. The shares of students by institution type are reported in the UNESCO/OECD/EUROSTAT database (Table 2.2). In principle, there is no government support for students in independent private institutions and partial support, depending on the education level and the scope of public regulation, for students in government-dependent private institutions.

The shares of students in private, public and dependent institutions are assumed to remain constant at the early childhood, pre-primary, and tertiary levels throughout the entire projection period. In early childhood education, this means that in the baseline, about two-thirds of students are assumed to be getting zero support throughout the projection period. This assumption has no significant effect at the pre-primary level, as students in public and government-dependent institutions receive similar government support. At the primary and secondary levels, these shares are deemed to develop according to changes in the minors’ population composition. The mix of students by institution type differs between population groups, with a lower share of students in public schools among the Haredim (Table 2.3). At the same time, the mix of institutions within each sub-group is assumed to remain constant. Furthermore, the calculation considers differences in support levels provided to government-dependent institutions according to the extent of government regulation and core curriculum studying (Table 2.4). The budget for students in government-dependent private institutions is about 88% for primary education and about 91% for upper secondary education, considering the demographic composition and the different support levels.

Table 2.2. At some education levels, many study in independent private institutions

Type of educational institution, 2022

	Public institutions	Independent private institutions	Government-dependent private institutions
Early childhood educational development	0%	66%	34%
Pre-primary education	65%	5%	30%
Primary education	76%	0%	24%
Lower secondary education*	82%	0%	18%
Upper secondary education	93%	0%	7%
Post-secondary non-tertiary education	10%	0%	90%
Short-cycle tertiary education	47%	0%	53%
Bachelor’s or equivalent level	11%	14%	75%
Master’s or equivalent level	14%	16%	70%
Doctoral or equivalent level	0%	0%	100%

Source: UNESCO/OECD/EUROSTAT database.

Table 2.3. Share of students by type of educational institution and education stream

Share within each education stream, excluding independent private institutions

	Non-Haredi Jews and others, State + State Religious		Haredim		Arabs
	Public institutions	Government-dependent private institutions	Public institutions	Government-dependent private institutions	Public institutions	Government-dependent private institutions
Primary	98%	2%	4%	96%	82%	18%
Lower-secondary*	100%	0%	100%	0%	100%	0%
Upper-secondary	68%	32%	0%	100%	48%	52%

Note: (*) There is an inconsistency between the share of lower-secondary students studying in government-dependent private institutions, as shown in the UNESCO/OECD/EUROSTAT database (Table 1.2), and the figures provided by the Israeli Ministry of Education. Currently, the model relies on data from the Ministry of Education.

Source: Ministry of Education.

Table 2.4. Schools’ budgeting depends on the extent of supervision

% of government support provided to government-dependent private institutions, by stream

	Non-Haredi Jews and others, State + State Religious	Haredim	Arabs
Primary	70%	87%	70%
Lower-secondary	70%	100%	70%
Upper-secondary	100%	82%	100%

Source: Ministry of Education.

Figure 2.3 illustrates the results of the aforementioned calculation for the upper secondary education level. It indicates that although there is a gap between the total number of students in the system at the projection’s starting point, this gap is anticipated to widen significantly in the future, primarily due to the increase in the proportion of the Haredi population among the total population in the relevant age groups. By 2065, the number of budgeted students is expected to reach 91% of the total student population at this education level.

Figure 2.3. In some education levels, the projected increase in budgeted students is significantly lower than the total number of students

Upper secondary education

Source: OECD elaborations.

Projecting per budgeted student expenditure growth

Projecting per budgeted student expenditures growth, relying on the assumed students-to-teacher ratio and considering macroeconomic developments. Up to 2025, total public expenditure on education reported in COFOG is uprated according to the latest budget proposal (see Section 1.2.2 in Chapter 1). Then, expenditures in the base year are broken down into two main components: expenditure on staff compensation (i.e. gross wages and salaries of teaching and non-teaching staff) and other expenditures, including current and capital expenditures as well as transfers (like student loans). Dividing these figures by the projected number of budgeted students allows to present expenditures per budgeted student in each educational level. Alternatively, expenditures on staff compensation per teacher could be calculated by relying on the current staff-to-student ratios based on the UNESCO/OECD/EUROSTAT database.

Per-budgeted student expenditures on staff compensation grow in line with labour productivity in the total economy, assuming the education staff-to-student ratio (average class size) remains constant over the projection period. This implies that the number of teachers and other staff adjust instantaneously and fully to demographic and macroeconomic changes. Likewise, the average compensation per staff member as a share of GDP per worker is assumed to remain constant. International comparisons of teacher salaries are inconclusive, but overall, they seem to be on par with the OECD average when comparing teachers’ salaries per teaching hour relative to average national wages per work hour (Shoresh, 2020[4]). Non-remuneration expenditures per budgeted student are assumed to grow in line with GDP per capita, implying the government will seek to provide a constant level of public spending per capita in real terms. Finally, total outlays are calculated by multiplying per budgeted student expenditures in period t by the total projected number of budgeted students in the same period.

2.1.3. Expenditures on Yeshiva and Kollel students

Except for spending on common educational levels, in Israel, the government also pays allowances for Haredi men who study in religious seminaries. In 2023, the number of Israeli Yeshiva (unmarried men) and Kollel (married men) students reached a total of 158 thousand, a 73% increase since 2013 (Cahaner and Malach, 2024[5]). With the increase in the Haredi population, these expenditures are expected to rise significantly. From the age of 20 to 34, the majority of Haredi men studying in these seminars (Figure 2.4). In 2019, about 60% of Haredi men aged 25-34 who studied in Kollels weren’t active in the labour market. Approximately 13% were neither engaged in education nor the labour market. In 2025, the direct government allowances for non-married Yeshiva and Kollel students are about 500 and 1 100 shekels per month, respectively. The model relies on the projected employment rates and participation in higher education described above, and the intensity of full-time studies in Kollels to project the number of students in these seminars in the future. The allowances are expected to rise along with labour productivity growth.

Figure 2.4. The majority of young Haredi men study in religious seminars

Israeli Yeshivas and Kollels students as a share of the respected Haredi male population, by age

Source: OECD calibrations based on the Israel Democracy Institute’s data (2024[6]).

2.1.4. Public spending on education throughout 2065

Under the baseline scenario, education spending as a percentage of potential GDP is projected to decrease from approximately 6.1% of potential GDP in 2025 to 5.5% by 2045 and 4.6% by 2065 (Table 2.5), along with the projected decline in the share of children in the total population. The expected decline is more significant in the pre-primary and primary levels, as the projected increase in the number of kids in the relevant age groups is more moderate. In contrast, expenditures on Yeshiva allowances, which in COFOG are part of recreation, culture and religion expenditures, are expected to triple, along with the increase in the share of the Haredim among the working-age population.

Table 2.5. Pre-primary and primary education will account for most of the spending decline

Spending on education as a % of potential GDP

Education level	2025	2045	2065	Change up to 2065
Pre-primary and primary education	2.8	2.4	1.9	-0.9
Secondary education	1.4	1.3	1.1	-0.3
Post-secondary non-tertiary education	0.1	0.1	0.1	0.0
Tertiary education	1.1	1.1	1.0	-0.2
Other education expenditures	0.7	0.7	0.5	-0.2
All (COFOG 9. Education)	6.1	5.5	4.6	-1.5
Early childhood educational development (part of 10. Social Protection)	0.2	0.2	0.1	-0.1
Yeshiva allowances (part of 9. Recreation and Culture)	0.1	0.15	0.2	0.1
All	6.4	5.8	4.9	-1.5

Note: The unclassified layer includes all expenditures not classified to a specific education level.

Source: OECD elaborations.

2.1.5. Scenario analysis

To test the implications of policy shifts and the sensitivity of the model to changes in assumptions, the baseline results – which assume no policy changes – are compared with several alternative scenarios (Table 2.6). First, education expenditures are projected for alternative scenarios in which population growth is higher or lower, as well as for alternative GDP growth trajectories, consistent with what is done with other expenditure items. The findings underscore the critical role of demographic trends in shaping the trajectory of education spending.

In particular, under a scenario of higher population growth, driven by increased fertility rates, as reflected in the median variant of the most recent CBS population projections (2017), education spending is projected to rise significantly. This contrasts with the baseline scenario, which anticipates a notable decline. By 2065, the difference in spending between the two scenarios could amount to as much as 1.3% of potential GDP. The primary driver of this divergence is a higher proportion of minors in the population, requiring allocating more resources to education, especially at the early childhood, pre-primary, and primary education levels.

The strong correlation between the projected education spending as a share of potential GDP and the proportion of minors in the population is illustrated in Figure 2.5. These results highlight the strategic importance of closely monitoring fertility trends, which are likely to exert a significant influence on the long-term planning of the education system.

Figure 2.5. High correlation between projected spending and the share of children in the total population

Source: OECD elaborations based on customised Israel Central Bureau of Statistics 2017 population projections.

Policy shifts, whether driven by political considerations or other pressures, could introduce significant additional fiscal burdens. In the current modelling framework, only the direct expenditure implications of such policy changes are considered. The potential impact on human capital formation, productivity growth, labour market participation and, consequently, GDP growth is not incorporated.

For instance, expanding fully subsidised early childhood education to 100% from 2027 would increase spending by 0.4% of potential GDP by 2045 and by 0.3% by 2065, without considering the potentially higher uptake. This estimate also does not account for the potential long-term benefits of such a reform, including improved skill acquisition among future cohorts or increased labour market participation among mothers. Nor does it consider the possible drag on GDP growth arising from higher taxation required to finance the expansion.

Similarly, aligning Israel’s average class size, from pre-primary through upper secondary levels, with the OECD current average by 2065 would necessitate a higher teacher-to-student ratio, estimated to raise education spending by 0.8% of potential GDP by 2065. For simplicity, this scenario assumes a 24% increase in the ratio until 2065, reflecting the current gap at the primary level as identified in Education at a Glance 2024 (OECD, 2024[7]), and extrapolates it across all levels. This pace is broadly consistent with the observed decline in average class size between 2013 and 2023 (Bank of Israel, 2024[8]). The scenario can therefore be interpreted as one in which reforms affecting class size, including those related to special education and its impact on teacher demand, continue at the same pace as over the past decade.

Nonetheless, more efficient deployment of the existing teaching workforce could help reduce class sizes with less pronounced fiscal pressure (Tamir and Avraham, 2021[9]). On average, across the OECD, in 2023, there were 14 students per teacher in primary education, 13 students in lower secondary education and 13 students in upper secondary education. In Israel, the corresponding numbers are 15 in primary education, 13 in lower secondary education and 11 in upper secondary education (OECD, 2024[7]). These small gaps in the teachers-to-students ratios compared to the considerable gaps in class sizes suggest that better allocation is indeed likely.

Increased government support for semi-private (dependent) schools could also have notable fiscal implications. If such support is provided without a corresponding requirement to expand the study of core subjects, the additional spending is unlikely to yield meaningful gains in educational outcomes. In this scenario, modelled by assuming that all students from pre-primary to upper secondary are funded at the same level as those in public schools, effective from 2026, spending is projected to rise by 0.14% of potential GDP by 2065 compared to the baseline.

As noted above, one of the key drivers of rising education expenditure in recent years has been the rapid growth in spending on special education. From 2020 to 2025, special education spending doubled. In the baseline scenario, these expenditures are incorporated within the broader education budget and not modelled separately. However, an alternative scenario assumes that special education spending will continue to grow at a faster rate than other education categories through 2035. This reflects mounting parental demand for support services and improved diagnosis of learning disabilities1. In the 2025 budget, allocations for special education reached NIS 17 billion, and these are classified under COFOG function 9.1: Pre-primary and primary education. Assuming that these expenditures will increase at a rate 30% higher than that of the corresponding education expenditures until 2035, total spending on education could be 0.1% higher in 2035 and 2065, respectively.

Finally, the model includes a scenario in which spending on Arab schools rises more rapidly, to equalise the budget per student by 2030, following previous OECD recommendations (OECD, 2023[10]) (OECD, 2025[11]). At present, the budget allocated per Arab student attending a school in the lowest quartile of the nurture index is lower than that in the Hebrew system, particularly at the upper secondary level (Figure 2.6). These findings are relevant to approximately 70% of schools’ resources, which are allocated directly by the Ministry of Education. Part of these funding gaps can be attributed to school characteristics, such as school size, staff tenure, the share of migrant students, the share of students with special needs, as well as additional funding for religious activities in some schools. Nonetheless, even after controlling for relevant characteristics, a funding gap of about 10% remains between the Arab and Hebrew sectors at the primary level (Blass and Bleikh, 2025[12]) and of about 15% at the upper secondary level (Blass and Bleikh, 2024[13]). Assuming these gaps will be closed in the public system by lifting spending on Arab schools without affecting spending on the Hebrew system, total education spending will be 0.1 percentage points of GDP higher than in the baseline by 2045.

Figure 2.6. Resources in the education system are not directed to where they are most needed

Average budget per student by nurture index, 2020, thousand NIS

Note: The nurture index of the Ministry of Education measures the socio-economic background of a school’s student population. Approximately 80% of Arab students are in the first two quartiles of the nurture index. The charts illustrate the first four quartiles, as no Arab schools are present in the upper quartile of the index. In secondary education, Arab schools are found only in the three lower quartiles. The Hebrew education system encompasses all sub-sectors: state, state-religious, and Haredi.

Source: The State Comptroller's office.

Table 2.6. Overview of education spending scenarios

Variable	Baseline model	High population growth	Low population growth	High GDP growth	Low GDP growth	Universal early childhood education*	Lowering class sizes	More generous government support to semi-private institutions	Faster spending growth on special education	Additional funding for Arab schools
Population projection	Mixed variant	Adjusted median variant	Adjusted low variant	Mixed variant	Mixed variant	Mixed variant	Mixed variant	Mixed variant	Mixed variant	Mixed variant
GDP scenario	Baseline	Baseline	Baseline	Melting pot	Frozen rates	Baseline	Baseline	Baseline	Baseline	Baseline
Share of early childhood education students in government-supported institutions	34%	34%	34%	34%	34%	Gradual increase to 100% by 2045	34%	34%	34%	34%
Teachers-to-students ratio in pre-primary to upper secondary education	Constant	Constant	Constant	Constant	Constant	Constant	Rise by 24% by 2065	Constant	Constant	Constant
Government support for semi-private institutions in pre-primary to upper secondary levels	As in 2023	As in 2023	As in 2023	As in 2023	As in 2023	As in 2023	As in 2023	A subsidy of 100% to a semi-private institution	As in 2023	As in 2023
Special Education	Not considered separately	Not considered separately	Not considered separately	Not considered separately	Not considered separately	Not considered separately	Not considered separately	Not considered separately	30% higher annual increase up to 2035	Not considered separately
Funding gaps between sectors	As in 2023	As in 2023	As in 2023	As in 2023	As in 2023	As in 2023	As in 2023	As in 2023	As in 2023	Rise by 10% in primary and 15% in upper secondary
Public education expenditure, % potential GDP 2065 (2045)	4.6 (5.5)	5.8 (5.9)	4.6 (5.3)	4.4 (5.4)	4.7 (5.6)	+0.3* (+0.4*)	5.4 (6.0)	4.7 (5.6)	4.7 (5.7)	4.6 (5.6)

Note: *Early childhood education is part of Social Protection spending according to COFOG, so it is not shown as part of Education spending in this table.

Source: OECD calculations.

2.2. Health spending

Total and public spending on healthcare is comparatively low in Israel. In 2023, total spending accounted for 7.6% of GDP, while public spending accounted for 5.2% of GDP, representing 13% of general government expenditures. While both shares have risen over the past decade, the increase was smaller than in the OECD median, widening the spending gap (Figure 2.7). Part of the explanation lies in Israel’s relatively young population, with only around 13% of the population aged 65 years and above compared to around 18% in the OECD. But adjusting for Israel’s younger population, recent OECD research suggests that per capita health spending is still about 15-20% below the OECD average (Morgan and Mueller, 2023[14]). The efficient delivery of primary care, including through larger health clinics and the organisation of doctors working in the community into teams, which allows them to deliver follow-up support, preventive activities and regular monitoring, are other explanations for the relatively low spending on healthcare (OECD, 2023[10]). In addition, the widespread use of voluntary health insurance (VHI), with over 80% of the population covered, may alleviate some of the burden on the public system. Differences in how long-term care spending is allocated between health and social protection may also contribute to the health spending gap.

As in most OECD countries, health spending is expected to outpace economic growth in the future as the population is ageing. As mentioned in Chapter 1, the share of the population aged 65 or above is projected to increase from about 13% to around 18% by 2065, which will likely cause an increase in public (and private) spending on healthcare. Cost pressures from new technologies, labour shortages, and wage growth keeping track with the overall economy despite lower productivity improvements (known as the Baumol effect) could further push spending upwards. The following section provides a macro perspective on the main determinants of government health expenditure identified in the literature, evaluates how well these determinants have historically predicted Israeli health spending, and projects future health spending in Israel up to 2065.

Figure 2.7. Health expenditures are relatively low

Panel A: Total expenditures as a share of GDP, 2006-2023

 

Panel B: Public expenditures as a share of GDP, 2006-2023, COFOG data

Sources: OECD, Health expenditure and financing, COFOG.

2.2.1. Approach to projecting health spending

The methodology used to project health expenditures is based on previous OECD work (Lorenzoni et al., 2019[15]), (Guillemette, 2019[16]) as well as the EU Ageing Report methodology (European Commission, 2023[3]). A few adjustments are made for application to Israel (Box 2.1). The starting point of total health spending, based on COFOG and the 2025 budget proposal, is disaggregated by age groups, relying on the Israeli capitation formula. The formula, last updated in 2010, determines how much government budget is allocated to the four health funds responsible for providing universal health insurance based on age, gender, and place of residency (periphery vs. centre) (Box 2.2 below). However, due to challenges in projecting future residences, the model excludes the periphery/centre distinction.

Box 2.1. Overview of recent long-term models for health expenditure

The chosen approach for projecting health expenditure in Israel draws on methodologies used in studies by the OECD and other institutions (Figure 2.8). Guillemette (2019[16]) projects government health expenditure in all OECD countries and non-OECD G20 countries using a logarithmic projection based on changes in the proportion of the population over the age of 65 and GDP per capita. They also incorporate cost pressures from technological progress and productivity constraints by assuming an excess of healthcare inflation over general inflation of one-half the rate of aggregate labour productivity growth.

Lorenzoni (2019[15]) and the European Commission (2023[3]) use a detailed projection methodology, incorporating population estimates by age and gender, along with country-specific age-spending profiles. This allows them to provide a nuanced breakdown of health expenditures across EU and OECD countries. They also consider the extent to which people are likely to age more healthily in the future and the implications for health expenditure (see Box 2.3 below). Geva (2013[17]) employs a similar, though less detailed, approach and integrates the Israeli capitation formula, which is used to allocate resources to the health funds as per relative client needs.

Figure 2.8. Projection approaches of different long-term health expenditure models

Note: “Partly” refers to situations where the indicators are considered indirectly by proxies.

Sources: European Commission (2023[3]), Guillemette (2019[16]), Lorenzoni et al. (2019[15]), Geva (2013[17]).

The main assumption is that the capitation formula applies to all public health expenditures, even though it is specifically used to compensate the four health funds for goods and services in the universal health care basket, covering about 77% of public health expenditures in Israel. The remaining 23%, which includes, among others, government compensation for severe diseases (like HIV and kidney failure), government-owned hospitals, certain public health activities in mental health, well-baby clinics and admin work, is assumed to be distributed in the same way as the capitation formula. This assumption is used in the absence of a more detailed breakdown of government spending by age-group and can be adjusted as more accurate data become available.

Box 2.2. Overview of the Israeli healthcare sector and the capitation formula

The Ministry of Health is responsible for governing the healthcare system, overseeing the performance of hospitals, health insurance funds, and healthcare professionals. Healthcare is regulated by the National Health Insurance (NHI) law, which ensures universal access to health services for all residents. Each resident can choose from among four competing non-profit health insurance funds. The health funds are financed by the government within the framework of the NHI and are based on prospective payments according to the capitation formula, which was last updated in 2010.

The formula considers age, gender, residence (periphery vs. centre) and a few chronic conditions of the client base. Compared with other OECD countries, per-capita healthcare spending across age cohorts in Israel follows a more pronounced U-shape, with a relatively larger share allocated to children under 14. However, for the oldest citizens aged 85 and above, per-capita spending in Israel increases only moderately, whereas the rise is much steeper on average across the OECD. (Figure 2.9).

Figure 2.9. The health spending profile by age and gender differs from the average OECD country

Estimated per-capita government health expenditure in 2023, NIS

Note: The following age spending weights (2017 estimate) are applied to construct the Israeli per-capita cost curve (women in brackets): 0: 1.9 (1.4); 1-5: 1.0 (0.8); 5-15: 0.5 (0.4); 15-25: 0.4 (0.5); 25-35: 0.5 (0.8); 35-45: 0.6 (0.8); 45-55: 1.00 (1.2); 55-65: 1.8 (1.7); 65-75: 3.1 (2.6); 75-85: 4.1 (3.4); 85+: 4.2 (3.5). The series OECD average illustrates what spending in Israel would have looked like if its age-spending profile was in line with the OECD average.

Source: Israel Ministry of Health. The OECD average curve developed by Morgan and Mueller (2023[14]).

The health funds are by law required to accept every client and obligated to provide their members with a broad government-determined benefits package, which includes hospital care, community-based care and various preventive services. They provide some of these services directly and purchase others from different providers (Rosen et al., 2021[18]).

The projection approach for health expenditures is illustrated in Figure 2.10. As a first step, the weights from the capitation formula are used to allocate the latest available data on total health spending to per capita spending by age group and gender. Secondly, Israel's age-dependent per capita spending profile is adjusted for each projection year by applying four distinct effects:

• Income effect;

• Healthy ageing effect;

• Productivity constraints (“Baumol”) effect;

• Technological progress effect.

These factors collectively influence the per-capita cost of the different age groups. The application and impact of these factors on the model are explained below. Lastly, for each year, the projected number of people in each age group, according to the baseline demographic scenario constructed based on the CBS population projections (described in detail in Chapter 1), is multiplied by the corresponding age cost profile to determine total public health expenditures. This approach has the advantage of capturing the effect of Israel’s specific demographic structure and is adaptable to future changes and refinements to the Israeli capitation formula. A targeted review of the health-spending projection literature by Morgan and Mueller (2023[14]) concluded that using country-specific age-spending profiles was the most reliable way to estimate the impact of demographic changes on health spending, capturing important variation unique to each country.

Figure 2.10. Schematic overview of the health expenditures model

Source: OECD elaborations.

2.2.2. Main drivers affecting health expenditures per capita

The income effect is measured by the income elasticity of health spending, which captures the percentage change in health expenditure in response to a given percentage change in GDP per capita. While earlier studies found income elasticity to be greater than one, indicating that health expenditures grow faster than income, current evidence indicates an income elasticity of around 0.7-0.8 for OECD countries. A recent econometric analysis of high-income countries found an income effect of 0.73, controlling for demographic changes, technical innovation, country, and time fixed effects (Lorenzoni et al., 2019[15]). As part of an econometric analysis conducted for this report, the income elasticity was estimated using a longer period and additional variables perceived to be relevant to Israel, such as the number of doctors related to the size of the old-age population (a proxy for doctor shortages), an interaction variable of Israel with GDP growth and a healthcare system classification (Böhm et al., 2013[19]). However, these potential factors were found to be insignificant. Therefore, the baseline scenario of the current model adopts an income effect of 0.73 (as in previous OECD reports), which means that for a one percent increase in Israel’s projected GDP per-capita, the per-capita health spend for each age group increases by 0.73 percent.

The healthy ageing effect accounts for reductions in age-spending profiles as life expectancy increases. It assumes that as people live longer, the years of morbidity decline, thereby lowering the per-capita costs of each age group. This is modelled by shifting the age-spending profiles downward in proportion to the projected gains in life expectancy2. The magnitude of this shift depends on the assumed ratio of healthy years gained to years gained in life expectancy (herein the ‘healthy ageing ratio’; see Box 2.3 below for details). For example, suppose the life expectancy of a 50-year-old man is projected to increase by 4 years from 2021 to 2065, and the ‘healthy ageing ratio’ is 0.5. In that case, a 50-year-old man in 2065 will have the age-spending profile of a 48-year-old man in 2021 (or 50-0.5*4), after adjusting for the impact of all other drivers (European Commission, 2023[3]). In the baseline scenario, the healthy ageing ratio is assumed to be 0.5, which aligns with recent literature on the topic.

The assumption that only half of future longevity gains will be healthy might be considered conservative. Recent data suggest that a large share of Israel’s gains in life expectancy over the past two decades has been in good health. According to WHO data, between 2000 and 2019, approximately 76% of the increase in life expectancy in Israel was spent in good health, compared with 71% on average across OECD countries (André, Gal and Schief, 2024[20]). However, this estimation did not consider the financial aspect of ageing directly.

Box 2.3. Healthy Ageing

The extent to which people are likely to age more healthily in the future has been widely discussed in the literature. Three general hypotheses have been proposed. The most optimistic is the ‘compression of morbidity’ hypothesis (Fries, 2002[21]). It posits that as life expectancy increases, healthier lifestyles result in a decrease in the share of years lived in poor health or with a disability. In contrast, the ‘expansion of morbidity’ hypothesis (Gruenberg, 2005[22]) supposes that life expectancy gains are added near the end of a person’s life when they are assumed to be in ill health. The ‘dynamic equilibrium’ hypothesis (Manton, 1982[23]) presents an intermediate view. It posits that although the number of years lived with health problems may increase, their severity could diminish due to improved medical treatments.

A recent review on healthy ageing hypotheses (Lindgren, 2016[24]) found that trends in high-income countries have most often been in accordance with the dynamic equilibrium hypothesis, with some evidence of a compression of morbidity and limited to no evidence of an expansion of morbidity. The current model is consistent with the dynamic equilibrium hypothesis. It models healthy ageing by applying health costs of younger age groups to older individuals when life expectancy increases. The outcome is a net decrease in health expenditure (ceteris paribus), which can be understood as older age groups becoming healthier in the future (or that the effects of being ill are less severe). In the baseline scenario, a partial dynamic equilibrium is used, whereby 50% of the gains in life expectancy result in reductions in health costs across most age groups (24 and older), in line with (Lorenzoni et al., 2019[15]).

The productivity constraints effect, also referred to as the “Baumol effect”, captures the impact of lower productivity growth in the health sector relative to other sectors of the economy. This concept, developed by economist William Baumol (1992[25]), highlights how certain sectors, such as health and education, experience slower productivity growth because they cannot substitute labour with technology as effectively as manufacturing or service industries with higher absorptive capacity for intangible capital. In more productive sectors, technological advancements often lead to higher productivity and lower labour costs per unit of output. However, in less productive sectors like healthcare, technological advancements do not significantly reduce the need for human labour.

To compete for labour with more productive sectors, wages in these non-progressive sectors must keep pace with overall wage growth, which is tied to broader productivity gains in the economy3. As a result, the unit costs in these sectors rise over time, driving up the relative prices. Given the inelastic demand for health services, higher costs are not offset by reductions in quantity, leading to overall increases in total expenditures4.

The Baumol effect is modelled by linking per-capita health expenditures to projected increases in overall economic productivity. Since the Baumol effect arises from the gap between productivity growth in the total economy and that in the health sector, it is logical to tie the magnitude of this effect to the projected rate of economy-wide productivity growth. Empirical studies have found evidence of a partial Baumol effect in the health sector, where a one percent increase in economy-wide productivity leads to about a 0.2 to 0.5 percent rise in health spending (Lorenzoni et al., 2019[15]; Colombier, 2017[26]; Hartwig, 2011[27]). Based on the literature and calibration to historical data, the baseline scenario assumes a Baumol effect of 0.3. This means that a one percent increase in aggregate labour productivity results in a 0.3 percent increase in per-capita healthcare spending.

Finally, technological advances – including new medicine, equipment, processes, and knowledge – generally lead to increased overall health spending. While some technological innovations make the treatment of existing medical conditions more cost-efficient, others are expansionary, extending healthcare services to previously unmet needs. Expansionary advancement in medical technologies, such as MRI scanning, prosthetic limbs, and new medications, enhances the capabilities of the health sector, but also increases total spending. Empirical evidence suggests that medical technologies do increase health spending, but accurately measuring the magnitude of this effect is complex due to the interaction between technological innovation and other variables such as GDP and productivity, and the lack of reliable proxies (Lorenzoni et al., 2019[15]).

One approach to estimate the net impact of technological progress on a country's health spending is to measure the average annual increase in health expenditures across high-income countries, controlling for other relevant determinants like GDP, demographic changes and country-specific factors. Since medical technologies often diffuse internationally (e.g., Israel benefits from medical advancements developed in the U.S.), this common effect, unexplained by other determinants, can be partly attributed to technological progress shared among countries. As it might be pessimistic to assume that technology is forever going to push costs up, in the current model, the positive impact of technological advances is assumed to be rather small at 0.0005, implying an annual increase of 0.05% in national per-capita health spending, holding all else constant.

Considering the combined effect of the various health cost drivers, real per capita spending is projected to more than double at both ends of the age curve. For the adult population below age 85, healthy ageing is expected to slightly moderate cost increases, keeping them between 70 and 100 percent. Overall, the income effect is projected to be the most significant driver of increases in per capita health spending (Figure 2.11).

Figure 2.11. The income effect will be the main driver of per-capita cost increases

Female per-capita cost curve (in constant 2023 NIS)

Note: Healthy ageing cost curve (2065) reflects the expected per-capita costs in the Baseline case. The results are equivalent for the male per-capita cost curve, with a slightly different healthy ageing profile. The compounding residual captures the portion of projected health spending growth that is not explained by the individual modelled components but reflects the interaction between these components.

Source: OECD calculations.

Finally, by multiplying the per-capita cost in each projection year by the age-specific population according to the baseline demographic projections and summing across all age groups, the total health expenditure for each projection year is derived.

2.2.3. Assessing the model using a back-testing exercise

The model's performance is evaluated by comparing its projections of total public health expenditure with historical COFOG data for the period 1995 to 2021. This constitutes an out-of-sample evaluation, as the coefficients of the health expenditure drivers were not estimated using historical data from Israel. The projections are constructed in a one-period-ahead manner, where each year’s prediction is based on information available up to that point in time, including GDP and productivity data of the respective year.

Prior to the impact of the COVID-19 pandemic in 2020, the model closely tracked actual spending trends throughout most of the evaluation period (Figure 2.12). In 2019, health expenditures predicted by the model were only around 0.7% higher than the actual figure. Most deviations in the annual expenditure growth rate are minor and oscillate around zero, with a few years, such as 2015, exhibiting larger errors. The mean error of the growth rate deviations between 1995 and 2019 is close to zero (-0.03 percentage points). A simple t-test confirms that this deviation from zero is not statistically significant (p = 0.96), indicating that the model does not systematically over- or underestimate expenditure growth over the sample period. The 95% confidence interval for the average growth rate deviation ranges almost symmetrically around zero spanning from – 1.08 to 1.02 percentage points.

Figure 2.12. Actual health expenditures vs. predicted expenditures by the model, 1995-2021

Note: Actual data are based on total government consumption (COFOG). They are compared with results based on a model that assumes income elasticity of 0.73, healthy ageing ratio of 0.5, a Baumol effect of 0.3 (which means that expenditures will rise by 0.3% for each 1% increase in labour productivity) and technology advances effect of 0.0005. The annual healthy ageing effect for the historical series is estimated based on UN life-expectancy data, which deviates from the CBS data.

Source: OECD calculations.

2.2.4. Projecting public health expenditure throughout 2065

Public spending on healthcare is projected to increase from 4.9% of potential GDP in 2025 to 5.2% by 2045 and 5.4% by 2065. While higher than historical trends, the increase in spending as a share of potential GDP is still lower than the OECD median, which has an anticipated annual increase of 0.05 percentage points of GDP by 2060 (Guillemette, 2019[16]).

The primary drivers of the projected rise in health spending are demographic changes and the income effect, linked to increasing GDP per capita. The spending pressure resulting from demographic change can be attributed to the growing elderly population, which more than offsets savings from a declining number of children. Together, these two effects account for over 80% of the overall projected expenditure growth by 2025 and 2065. The Baumol effect and advancements in medical technology also contribute to the long-term rise in health expenditure relative to potential GDP, though their influence is relatively small compared to the demographic and income effects. Spending increases are slightly moderated by a degree of compression of morbidity.

2.2.5. Scenario analysis

To test the model’s sensitivity to changes in the assumptions, the baseline results are compared to high and low scenarios for population growth and GDP variation due to labour supply convergence of different population groups. Three other scenarios assume different magnitudes of the income effect, meaning different elasticities of health spending to GDP-per-capita growth, cost pressures, meaning labour productivity gains in the economy translating into higher or lower cost pressures in the healthcare sector, and healthy ageing, meaning higher or lower compression of morbidity for the projected gains in life expectancy. Lastly, the effect of a counterfactual age cost curve according to the average OECD per-capita health spending instead of the Israeli capitation formula is tested.

Faster population growth reduces health spending by 0.1% of potential GDP by 2065 relative to the baseline. The modest decrease is the result of different forces at work. If the population grows more rapidly, spending initially increases slightly due to a higher share of minors in the population, who incur greater per capita costs in the early years of life (see Figure 2.11 above). However, higher population growth, primarily driven by groups with low labour market attachment, also leads to lower GDP per capita and a moderation in productivity growth. This reduces the income effect and the Baumol effect in the health sector.

Slower population growth, by contrast, increases health spending by 0.1% of potential GDP by 2065 relative to the baseline. This is because the share of the working-age population and population groups with strong labour market participation rises in the coming decades, so that stronger income and Baumol effects drive up health expenditure. As the demographic dividend fades in the second half of the century and the old-age dependency ratio continues to rise after 2065, health spending could deviate further from the baseline owing to higher per-capita costs for the growing number of elderly.

Better labour market integration of population groups with weak participation could exert mild downward pressure on health expenditure, amounting to 0.1% of potential GDP by 2065. The reason is that higher labour force participation among Haredim and Arabs would boost GDP through increased labour supply, thereby increasing health expenditures, but this would be overshadowed by the moderation of aggregate productivity growth, reducing cost pressures in the health sector. By contrast, if employment rates among all population groups remained at current levels, the changing population composition would lead to lower labour supply and reduced GDP, while the smaller workforce would be more productive, intensifying the cost disease. The impact could result in health spending being 0.1% of potential GDP higher by 2065.

Compared to the relatively small impact of the population and macroeconomic scenarios on health spending, alternative assumptions regarding the elasticity of health expenditure to aggregate income levels could have a substantial long-term impact. Assuming an income elasticity of 1, rather than 0.73 as in the baseline, meaning a 1 percent increase in health spending for every 1 percent increase in GDP per capita (as supported by some literature), could raise health spending to 5.7% of potential GDP by 2045 and 6.5% by 2065. In contrast, a lower elasticity of 0.5 would imply that spending remains close to 4.8% until 2045 and falls slightly to 4.7% by 2065. Monitoring this elasticity in the future will therefore be crucial to identify emerging spending pressures in the decades ahead.

Increasing (or decreasing) the Baumol effect relative to the baseline simulates health wages rising faster (or slower) than productivity growth in the sector, implying greater (or lesser) responsiveness to overall productivity growth in the economy. An alternative interpretation is a decline (or increase) in the supply of medical professionals in Israel. A reduction of the Baumol coefficient from 0.3 in the baseline to 0.1 simulates a scenario in which almost all wage increases in the economy are offset by productivity gains in the health sector. AI technologies, for example, have the potential to revolutionise healthcare by improving diagnostics and enabling providers to serve more patients, thereby raising productivity. In this scenario, health spending is projected to remain broadly stable throughout the projection period. Should the Baumol effect turn out worse than assumed in the baseline (simulated with a coefficient of 0.5), spending pressures could rise to 5.5% by 2045 and 6.1% by 2065. Strengthening the supply of healthcare professionals (while mitigating risks related to the supply-induced demand phenomenon) and investing in productivity gains in healthcare provision could therefore prove cost-effective.

Increasing the healthy ageing variable represents a more optimistic view of the relationship between life expectancy and morbidity, where each additional year of life expectancy is lived in good health (a healthy ageing ratio of 1:1). This puts downward pressure on average health spending, as the population requires fewer medical services, resulting in spending of 5.1 percent and 5.3 percent of potential GDP by 2045 and 2065 respectively. Should healthy ageing not materialise (assuming a coefficient of 0), spending pressures could increase to 5.3% and 5.7% of GDP in the reference years. Strengthening preventive care programmes and policies that promote a healthy lifestyle (like Pigouvian taxes) could help realise the gains from healthy ageing and the implied fiscal savings.

Assumptions on age-spending profiles in the baseline model are based on the Israeli capitation formula. However, the latest available data on this formula dates back to 2010, leaving some scope for changes in the cost profile of individuals in the meantime. A sensitivity analysis is therefore conducted, assuming that spending by age group has since aligned with the average OECD cost curve. The overall effect of substituting the cost curve is small, increasing spending by 0.1 percentage points by 2065. Therefore, uncertainty surrounding per capita cost does not appear to be a major concern for the health spending projections.

Table 2.7. Overview of health spending scenarios

Variable	Baseline model	Faster population growth	Slower population growth	Frozen rates	Melting pot	Baumol	Income elasticity	Compression of morbidity	OECD average age-spending profile
Population projection	Mixed scenario	Adjusted median variant	Adjusted low variant	Mixed scenario	Mixed scenario	Mixed scenario	Mixed scenario	Mixed scenario	Mixed scenario
GDP scenario	Baseline	High population growth	Low population growth	Frozen rates	Melting pot	Baseline	Baseline	Baseline	Baseline
Baumol effect	0.3	0.3	0.3	0.3	0.3	0.5 (0.1)	0.3	0.3	0.3
Income effect	0.73	0.73	0.73	0.73	0.73	0.73	1.0 (0.5)	0.73	0.73
Healthy ageing ratio	0.5	0.5	0.5	0.5	0.5	0.5	0.5	1.0 (0.0)	0.6
Age-spending profiles	Israel capitation formula	Israel capitation formula	Israel capitation formula	Israel capitation formula	Israel capitation formula	Israel capitation formula	Israel capitation formula	Israel capitation formula	OECD average
Technological advances	0.0005	0.0005	0.0005	0.0005	0.0005	0.0005	0.0005	0.0005	0.0005
Public health expenditures (% potential GDP), 2045	5.2	5.2	5.2	5.2	5.1	5.5 (4.9)	5.7 (4.8)	5.1 (5.3)	5.2
Public health expenditures (% potential GDP), 2065	5.4	5.3	5.5	5.5	5.4	6.1 (4.9)	6.5 (4.7)	5.3 (5.7)	5.5

Source: OECD calculations.

2.3. Interest payments

Government spending on servicing public debt includes interest payments and expenses for underwriting and floating government loans. In Israel, this function also includes subsidies to pension and insurance institutions to reduce their exposure to capital market fluctuations. According to COFOG, in 2024, the general government spent about 3.4% of GDP (68.5 billion Shekels) on public debt transactions – a low amount from a historical perspective (see Figure 2.13) but still above-average debt servicing costs across OECD countries (OECD, 2025[28]). The historic decline in interest payments reflects both the global reduction in sovereign borrowing costs over recent decades and Israel’s prudent fiscal policy, which has contributed to lowering the public debt burden5. However, looking ahead, the Bank of Israel’s nominal interest rate is expected to remain at 4.5 percent through 2026 as the economy recovers from years of a large output gap (OECD, 2025[29])6. Rising net borrowing requirements, combined with an elevated country risk premium on long-term debt due to the war, are accelerating the refinancing and issuance of new debt at marginal borrowing costs significantly above Israel’s current implicit interest rate (OECD, 2025[28]).

Figure 2.13. Debt servicing costs have declined

Government spending on servicing public debt, share of GDP

Note: Interest payments are recorded in COFOG on an accrual basis. When bonds are issued at a premium or discount, the difference between their issue price and their face or redemption value when they mature measures the interest the issuer is obliged to pay over the bond's life.

Source: Israel Central Bureau of Statistics.

The long-term evolution of debt servicing needs depends on the size and structure of the current debt stock, primary spending (all non-interest spending), economic growth and a forecast of gross interest expenditures. This section focuses on the latter. Interest expenditures on marketable debt are projected to evolve based on the average borrowing cost of outstanding maturing debt and an endogenous projection of marginal borrowing costs at different maturities. To this end, a yield curve is constructed based on forecasts of short- and long-term interest rates. The short-term interest rate is primarily influenced by the business cycle, potential GDP growth, and a measure of inflation convergence towards a target (explained below). Long-term interest rates, in turn, reflect expected future short-term interest rates as well as a term premium and a risk premium. In addition to marketable debt, the servicing needs of non-marketable debt issued to support pensions and life insurance are projected based on an exogenous trajectory, reflecting the phase-out of pension fund subsidies through earmarked bonds and the yield guarantee mechanism replacing them.

The rest of this section is structured as follows: first, it presents Israel’s debt structure. Then, it presents expected spending on servicing unmarketable debt and subsidising pensions, reported in COFOG under function 1.7 (Public debt transactions). Lastly, it discusses the methodology used to project the marginal interest rates and the implicit average effective nominal interest rate on the government marketable debt.

2.3.1. Israel’s government debt structure

In 2024, about 59% of Israel’s government debt was held domestically, denominated in local currency and tradable. Another 15% was denominated in foreign currency, and 26% was non-tradable domestic debt (Figure 2.4). The latter was mainly issued as earmarked bonds with a fixed real interest rate of 4.86% and a maturity of 15 years for pension funds to limit the exposure of expected pension benefits to capital market fluctuations. However, in recent years, the offered yield was much higher than that of corresponding bonds issued in the market and constituted a generous subsidy for private pensions. In addition, the issuance volumes increased rapidly, generating challenges to debt management. Therefore, since October 2022, the issuance of these bonds has ceased, and a new yield-guaranteed mechanism was introduced (see below). The debt stock will gradually be affected by this reform. Each time a designated bond is redeemed, the mechanism is replaced. Thus, full implementation could be expected in 2037/8, 15 years after the new mechanism was first applied. At the same time, the share of marketable debt will rise towards 100%. The following section describes the method for estimating interest payments on un-marketable public debt and the costs associated with the new yield-guaranteed mechanism.

Figure 2.14. Detailed composition of Israel’s government debt, 2024

Source: OECD calibrations to the Israeli Ministry of Finance data.

2.3.2. Interest payments on un-marketable public debt

Interest payments on unmarketable debt and pension-related subsidies are calculated by estimating the future unmarketable debt stock, multiplying this stock by the average interest rate paid to bond holders, evaluating the pension funds’ assets covered by the guaranteed-yield mechanism, and projecting spending on this guarantee, relying on assumptions regarding future returns in capital markets.

The starting point is the level of untradable debt. It was 338 billion shekels in 2023, about 2/3 in bonds issued to DC pension funds (“Arad bonds”7). As mentioned above, as of October 2022, the yield guarantee mechanism replaced the issuance of Arad bonds. Thus, the stock of the earmarked bonds is assumed to diminish gradually. Because more significant amounts of these bonds have been issued in recent years, the decline is supposed to accelerate over the years. For the sake of simplicity, the remaining 1/3 – primarily bonds issued in the past to the old DB funds and insurance companies8 – is assumed to diminish at the same speed. Overall, the un-marketable stock is expected to decline from 18% of potential GDP in 2023 to 7.8% in 2030 and 0% in 2038 onwards. In 2023, the average real interest rate paid on unmarketable bonds was 4.8%. This rate is multiplied by the projected stock of unmarketable bonds (adding inflation compensation) in each period to reach the approximation of the annual interest payments related to the earmarked and other untradable bonds (Figure 2.15). Interest payments as a share of potential GDP are declining in line with the untradable debt stock and the return of inflation to the central bank target.

Figure 2.15. The stock of the earmarked bonds is expected to diminish gradually

Untradable debt stock (primary axis) and interest payments on untradable debt (secondary axis), % of potential GDP

Source: OECD calculations.

Under the reformed yield guarantee mechanism, the government pledges a 5.15 percent real annual return on the portion of annual pension fund investments previously allocated to Arad bonds. A reconciliation between the government and the pension funds occurs every five years to determine whether the actual investment returns on this portion fall short of or exceed the guaranteed rate. If returns fall short, the government covers the difference; if returns exceed 5.15 percent, the surplus is deposited into a dedicated account at the Bank of Israel, which serves as a reserve to cover possible future losses. To build a cushion for potential expenditures arising from reconciliation, the government contributed an amount equal to 3.15 percent of covered assets into this reserve in 2023. This annual contribution rate will gradually decrease to 1.95 percent by 2029, at which level it is expected to stabilise (Ministry of Finance, 2024[30]).

Government expenditures on the yield guarantee mechanism will depend on the level of assets managed by pension funds and future returns in capital markets. Pension fund assets are expected to increase rapidly over the next two decades due to the mandatory pension law implemented in 2008 (applying to all individuals younger than 55 at that time). The relatively recent introduction of mandatory pensions implies a rapidly growing number of savers, while withdrawals from these funds remain limited. A Ministry of Finance report evaluated that, assuming a real yield of 5.5 percent annually, covered assets would grow by approximately 9 percent per year (in constant prices) until 2030, before gradually slowing to around 5.5 percent annual growth in the long term (Ministry of Finance, 2016[31]).

The baseline scenario assumes a more conservative yield of 4.0 percent annually, which is closer to the long-term GDP growth projection. The lower yield implies slower asset growth and a larger expected government subsidy, calculated as the difference between the guaranteed 5.15 percent yield and the assumed 4.0 percent market yield. Although the reconciliation with pension funds occurs every five years, expenditures are accounted for annually, consistent with the accrual approach of projected COFOG expenditures.

Moreover, the mechanism implies a risk to the government budget due to the volatility of market returns. In substantial downturns in the capital market, the government may have to compensate institutional investors with significant amounts within a short time. At the same time, the likelihood of more significant primary deficits in such a situation increases. The required government savings to manage the contingent liability are modelled as an extra-budgetary fund, assuming an additional amount (buffer) of 1% of all covered assets to be saved (spent).

Figure 2.16 shows projected spending for different assumptions regarding the long-run average yield in the financial market. It reveals an interesting pattern: while spending is expected to be higher in case of lower returns in the next two decades due to the gap between the target yield (5.15%) and actual yields, in the very long run, expenditures on this mechanism will be higher in case of high returns. The reason is that the high returns increase the level of covered assets, determining the subsidy coverage. Overall, the highest spending is expected for returns of 3 to 4%.

Figure 2.16. Higher returns are associated with higher government spending needs in the long run

Government expenditures on the yield-guaranteed mechanism, % GDP

Note: Assumptions on savings are aligned with the Ministry of Finance report (Ministry of Finance, 2016[31]).

Source: OECD calculations.

2.3.3. Interest payments on marketable public debt

The tradable debt amounted to 827 billion shekels in 2023. The average maturity was 8.2 years, with 9% of all tradable debt expected to mature in less than a year. This tradable debt level is assumed to evolve in line with the budget balance and the increase in the share of tradable debt in the total debt, leaving aside the potential impact of stock-flow adjustments and exchange rate effects on external debt for simplicity (some financial transactions, such as land privatisation receipts, are rather considered in the calculation of general government spending). The cost of servicing this debt is calculated in two steps: first, interest paid on the old (initial) debt is considered by multiplying the unmatured initial debt in each period by the average implicit interest rate in the base year (2023), after deducting payments on untradable debt9. In 2023, interest payments on the tradable debt amounted to 40.5 billion Shekels, implying a nominal interest rate of 4.9% (40.5 divided by 827, which is the total stock of tradable debt).

${IP}_{initial\mathrm{ }debt,t}={\stackrel{-}{Implicit\mathrm{ }IR}}_{base}\mathrm{*}{Unmatured\mathrm{ }initial\mathrm{ }debt}_{t-1}$

Then, gross financial needs or newly issued debt (GFN) in each period are projected based on the overall deficit (primary deficit plus interest expenses), debt amortisations – which depend on the maturity structure of the initial tradable debt and the debt issued after the base year – and other debt-creating flows assumed to equal zero in the baseline. The maturity structure of the initial debt is based on data from the Accountant General. In the baseline scenario, the maturity structure of newly issued debt is assumed to mirror that of the initial debt (but with a different interest rate).

${GFN}_t={Deficit}_t+\mathrm{ }{Amorotisations}_t+\mathrm{ }{Other\mathrm{ }debt\mathrm{ }creating\mathrm{ }flows}_t$

Debt servicing costs on newly issued debt are calculated based on the yield curve at the time of its issuance. The yield curve is approximated each year through quadratic interpolation based on a projected short-term interest rate (IRS) and a 10-year interest rate (IRL). In the baseline scenario, the yields at all maturities represent a linear combination of the IRS and the IRL. The method used to project the short and long-run interest rates is based on the interest rate function of the OECD long-term model, described in detail in (Guillemette, 2019[16]; Guillemette and Turner, 2017[32]; Johansson et al., 2013[33]):

1. The short-term interest rate (${IRS}_t$) includes a neutral nominal interest rate component that is equal to the potential growth rate plus an inflation anchor. This inflation anchor is equal to the inflation target of the United States (2%) plus an adjustment based on the trend labour efficiency growth ($EFFLABS$) differential between Israel and the United States. This is to account for the Harrod-Balassa-Samuelson effect that makes inflation higher in fast-growing countries and also tends to make their neutral interest rate higher:

${IRS}_t^{neutral}=g_t\left(GDPVTR\right)+0.02+0.3\left[{dlog}_t\left({EFFLABS}^{Israel}\right)-{dlog}_t\left({EFFLABS}^{USA}\right)\right]$

Second, the actual short-term interest rate converges to the neutral one over time and as the output gap closes. This is a standard error correction equation with the speed of closure of 0.28, the two other parameters (the output gap and its first difference) being zero in the long run:

$∆{IRS}_t=0.13\mathrm{*}∆{GAP}_t-0.28({IRS}_{t-1}-{IRS}_{t-1}^{neutral}-0.55\mathrm{*}{GAP}_{t-1})$

2. The long-term interest rate in year t (${IRL}_t)$ is based on the ten-year forward-convolution of short-term interest rates ($IRS$) plus a term premium (${TERMPREM}_t$) and a fiscal risk premium (${FRP}_t$). The term premium is assumed to transition smoothly from current market conditions (currently negative for Israel) to a value of 70 basis points, in line with the equilibrium assumption used in the OECD long-term model for all countries. When net lending is negative, each percentage point of GDP in deficit is assumed to be associated with an additional fiscal risk premium ($FRP$) of 9.9 basis points10.

${IRL}_t=\mathrm{ }{TERMPREM}_t+\sum _{n=0}^9\frac{{IRS}_{t+n}}{10}+{FRP}_t$

Figure 2.17 shows the calculated yield curves for 2025 (the base year), 2045, and 2065 as projected in the baseline scenario. It presents yields for maturities from one to ten years and the average maturity of bonds issued for more than ten years, which in Israel is approximately 18 years. According to these projections, short-term yields are expected to increase until mid-century due to several factors. First, Israel’s sizeable output gap, at -4% of potential GDP in 2025, is projected to fully close only until 2040, boosting economic activity and increasing money demand. Second, inflation is expected to return to the 2% target only gradually. Third, real potential growth is the highest in the first two decades of the projection period. The short-term nominal interest rate is projected to peak at 5.6% in 2030, before moderating and gradually aligning with nominal potential output growth once the output gap has closed.

The long-term interest rate is expected to rise more sharply in the coming decade. This is mainly due to the forward-looking incorporation of short-term interest rates over a ten-year horizon, as well as the normalisation of the term premium over short-term rates, which is currently negative. The slope increase is partially mitigated by the moderation of the fiscal risk premium, which is projected to decrease by roughly 13 basis points between 2026 and 2065, reflecting the government's improving net lending position in the context of a stable debt-to-GDP ratio (implying improvements in the primary balance) and reduced interest expenditures.

Figure 2.17. The yield curve is projected to shift upward over the coming decade, but moderate thereafter

Government bond yields as a function of maturity, percent

Source: OECD calculations.

2.3.4. Government spending on servicing public debt

Under the baseline scenario, interest payments as a percentage of potential GDP are projected to increase sharply to 3.8% by 2030 and 4.8% by 2045. Given the assumption of a constant debt-to-GDP ratio and an unchanged maturity structure, the increase in interest payments primarily reflects the refinancing of outstanding debt (including non-tradable debt) and new borrowing at higher interest rates. Marginal borrowing costs for the average maturity, which is assumed to remain constant at 8.2 years, are projected to rise by more than one percentage point by 2038, when they are projected to peak. As in the model, nominal interest rates are tied to nominal GDP growth, this expected rise in interest rates can be explained by the expected economic recovery, relatively high potential GDP growth over the next two decades and only gradual convergence of inflation towards the target. Interest payments are expected to return to 4.5% of potential GDP by 2065, reflecting a moderation in potential GDP growth and, therefore, interest rates.

2.3.5. Alternative specifications of the fiscal risk premium

There is considerable uncertainty surrounding how markets will assess Israel’s fiscal risk in the future. In the baseline scenario, the fiscal risk premium ($FRP$) is modelled solely as a function of the government’s net lending position, which the model projects in detail. However, other factors – such as the level of public debt, geopolitical risks and conflict, and investors’ comparative assessment of sovereign risk – could also influence $FRP$ over time. To reflect this uncertainty, several alternative specifications are explored (Table 2.8).

One specification links the $FRP$ to the debt level rather than the net lending position. In this case, the current premium increases or decreases by two basis points for each percentage point that gross government debt increases or decreases from its current levels11. It is relevant only for the representation of the fiscal aggregates in which debt evolves endogenously – i.e. where government receipts remain constant and the debt-to-GDP ratio is not stabilised (see Section 1.4.4 on fiscal aggregates in Chapter 1). The increase of debt close to 80% of GDP in the absence of fiscal adjustments would lead to higher marginal borrowing costs of 32 basis points and higher government interest expenditures of 0.4% of potential GDP by 2065.

A second specification reflects the potential for persistently elevated geopolitical uncertainty, such as the risk of conflict or prolonged regional tensions. In this case, the $FRP$ is assumed to remain constant at its 2024 level – estimated as a 45-basis-point spread between the 10-year bond yield and the short-term interest rate – throughout the projection period to 2065. Borrowing costs would be between 10 and 16 basis points above the deficit-based Baseline case and increase annual interest expenditures by between 0.04 and 0.08% of potential GDP between 2035 and 2065.

Conversely, Israel’s long-term marketable debt risk premium could decline over time, converging towards that of low-risk sovereign assets. This specification assumes a gradual convergence from Israel’s baseline $FRP$, modelled as a function of the net lending position, to Germany’s risk premium, as projected in the OECD long-term model. The weight is assumed to shift linearly from 100% Israeli premium in 2025 to 100% German premium in 2065. This would have a modest moderating effect of up to 15 basis points on the 10-year marginal borrowing costs by 2065, easing the general government interest bill by 0.07% of potential GDP.

Table 2.8. Effect of alternative fiscal risk premia on government financing costs

Difference from the baseline scenario

Variable	2035	2045	2055	2065
10-year borrowing cost (in basis points)
Based on debt level*	+7	+23	+35	+34
2024 level persists	+10	+13	+15	+16
Safe asset convergence	-6	-8	-11	-15
General government interest expenditures (% of potential GDP)
Based on debt level*	+0.02	+0.12	+0.30	+0.43
2024 level persists	+0.04	+0.06	+0.08	+0.08
Safe asset convergence	-0.01	-0.04	-0.05	-0.07

Note: * To estimate the impact of the debt-level-based risk premium on expenditures relative to the deficit-based risk premium applied in the Baseline, the general government receipts as a percentage of potential GDP are held constant, and the debt is calculated endogenously. It follows that debt-level based fiscal risk premium is not directly comparable to the other two specifications in the table which have been estimated assuming a constant debt-to-GDP ratio.

Source: OECD calculations.

2.4. Social protection spending

Social protection is the largest function of public expenditures. In 2024, it accounted for 26.5% of total general government expenditures or 11.6% of potential GDP, after increasing by 1.2 percentage points over the last decade (Figure 2.18). Nonetheless, this spending is relatively low compared to other OECD countries, which spent an average of 17% of GDP on this function. A key factor behind this difference is Israel’s favourable demographic structure, with a relatively small elderly population. In addition, the state’s involvement in pension financing has been reduced with the development of a mandatory private defined-contribution insurance system. Public expenditure on pensions is thus relatively low, while tax exemptions to so-called incentivise savings have been increased.

In 2020, the COVID‑19 pandemic led to a temporary surge in social protection expenditures relative to economic activity. More recently, the war has caused a sharp rise in housing expenditures, reflecting the costs of accommodation and compensation for evacuees. Over the past decade, there have also been notable shifts in the composition of spending. In particular, sickness and disability expenditures have increased by more than 30%, accounting for over two‑thirds of the total rise in spending.

Two-thirds of social protection expenditures are not part of the central government budget but are instead allocated through Israel’s National Insurance Institute (NII), Israel’s public social security institution. The remaining one-third of social protection expenditures not covered by the NII is largely attributed to central and local government budgetary pensions and services for people with disabilities, alongside various smaller expenditures.

Figure 2.18. Social protection spending has increased

Government expenditures on social protection, COFOG, % of GDP

Source: Israel Central Bureau of Statistics (2025).

2.4.1. Approach to projecting social protection spending

The projections for social protection spending follow a threefold strategy (Figure 2.19): Major expenditure items, mostly the NII’s allocation, accounting for about 60% of spending, are projected using seven sub-models described below in further detail. The models rely on the demographic and macroeconomic projections but are further calibrated to integrate specific policy features and trends in eligibility and the generosity of allowance levels. The methodology is similar to that used by the Bank of Israel to forecast NII’s expenditures and deficits (Finkelstein, 2019[34]) (see Box 2.4).

Central and local government spending on budgetary pensions for former civil servants is projected based on the forecast of the General Accountant. These expenditures accounted for around 16% of spending in 2024 but are projected to decline towards zero in the coming decades. Various smaller spending items across different sub-vectors are indexed according to a constant level of service quality (see Section 2.5.3 and (Guillemette and Turner, 2017[32])), or to other modelled spending components if these provide a more plausible fit.

Figure 2.19. Detailed modelled projections for around 60% of expenditures

Social projection spending 2024, sorted by modelling approach (% of potential GDP)

Source: OECD elaborations based on Israel Central Bureau of Statistics data (2025).

Box 2.4. Bank of Israel’s forecast of the National Insurance system

Bank of Israel conducted a long-term forecast for Israel’s National Insurance Institute (NII) benefit system in 2019, covering all major social insurance expenditures, including old-age and survivors’ pensions, disability and long-term care benefits, child allowances, unemployment and work-injury insurance, income support, and other NII allowances. The study also examined the revenue side by incorporating the NII’s funding sources.

The Bank of Israel’s long-term macroeconomic model integrates CBS demographic projections with economic growth assumptions to forecast GDP and average wage growth until 2065 (the final year of the population projections). For each major NII benefit category, the projection uses specific socio-economic drivers and policy parameters to estimate future spending.

In the baseline scenario, total NII benefit expenditures were projected to rise slightly faster than GDP until 2065. NII spending was forecast to increase from 6.8% of GDP in 2018 to 7.6% by 2065. This gradual rise is largely due to population ageing (leading to more retirees and elderly care recipients) and a continued increase in disability claimants, partially offset by the erosion of CPI-indexed benefits such as old-age pensions. The study concludes that Israel’s ageing-related fiscal burden will increase only slowly over the century while remaining relatively low in international comparison.

Source: Finkelstein (2019[34]).

2.4.2. Demography and allowance growth are the main drivers of social protection spending

Demographic change has heterogeneous effects on social protection spending. Due to declining fertility rates, the share of minors is projected to decrease by eight percentage points by 2065, lifting pressures from family and children expenditures. The working-age population is expected to increase slightly by three percentage points, absorbing much of today’s large minor population. The impact of demographic change on sickness and disability expenditures will largely depend on the age-cost curve and recipiency rates within the working-age population. At the same time, the share of the population eligible for unemployment benefits and basic income support will grow. Lastly, the elderly population is projected to increase by five percentage points, exerting demographic pressures on old-age pensions, long-term care and old-age survivors' pensions.

The indexation of allowances is the second main driver of social protection expenditures. Over the past decades, allowances for different social support instruments have grown at different rates (Figure 2.20). Legislated indexation does not fully explain the development of all allowances (Table 2.9). Discretionary policy decisions on adequacy have influenced allowance levels for certain social support instruments. For instance, despite being indexed to CPI growth, the basic amount of the senior citizen allowance has effectively grown along with the average wage. Similarly, the disability benefit for a recipient with a partner and a child has grown faster than the average wage, even though it is legally indexed to it. Means-tested income supplements for old age and survivors’ pensioners have also grown more rapidly than the average wage. Therefore, assumptions on allowance indexation are made for relevant spending items in each sub-function, considering both the legislative basis and historical trends.

Figure 2.20. Allowance levels have not always grown at the legislated rates

Allowance levels growth, 2007-2023

Note: Due to a reform in the allowance mechanism, there is a break in the series around 2018. Nonetheless, the figure compares the average allowance before and after the break.

Source: OECD calculations based on National Insurance Institute data (2025).

Table 2.9. Assumptions for allowance indexation

Allowance	Legal basis	Indexation assumption
Work injury basic amount (COFOG 10.1)	CPI	Wage growth
Disability basic amount (COFOG 10.1)	Wage growth	Wage growth
Senior citizen basic amount (COFOG 10.2)	CPI	Wage growth
Survivors basic amount (COFOG 10.3)	CPI	Wage growth
Birth grant (COFOG 10.4)	CPI	50% Wage growth, 50% CPI
Child allowance (COFOG 10.4)	CPI	50% Wage growth, 50% CPI
Parental leave allowance (COFOG 10.4)	Wage growth1	Wage growth
Unemployment insurance (COFOG 10.5)	Wage growth	Wage growth
Basic income support (COFOG 10.7)	CPI	CPI

Note: ¹The maternity allowance paid to qualifying mothers is calculated based on wage-dependent contributions. As a result, its average level tends to rise in line with wage growth, even though it is not formally indexed to wages by law.

Source: National Insurance Institute and OECD elaborations.

The following sections describe the projection approaches for the different COFOG level two items within the social protection vector.

2.4.3. Sickness and disability (COFOG function 10.1)

In 2024, the general government spent about 3.3% of potential GDP (66.8 billion Shekels) on sickness and disability, mainly on in-kind or cash incapacity-related benefits. In international comparison, spending on incapacity is high. Before the pandemic in 2019, Israel’s spending on incapacity-related benefits was 2.8% of its potential GDP, compared with an OECD average of 2.0%. About 50% of expenditures on incapacity were on disability allowances paid by the National Insurance Institute (NII). Other notable expenditures included NII spending on work accidents and miscellaneous incapacity-related spending covered by the central government budget.

Box 2.5. NII disability pensions

Disability pension

To adults who cannot work full-time due to a significant health impairment, the NII pays a monthly disability pension. Eligibility for the general disability benefit requires meeting several conditions, including being an Israeli resident between age 18 and the legal retirement age, a permanent medical impairment of at least 60%, or 40% if there are multiple conditions (with one impairment ≥25%), a work capacity loss reducing the ability to earn a living by at least 50% and an income test, whereby the person’s earnings must remain below 60% of the average wage.

The monthly disability pension amount is a flat sum set by law according to the assessed incapacity, not the person’s prior wage. As of January 2025, a 100% incapacity (full disability) pension pays about NIS 4,556 per month. Those with partial capacity receive a proportionally lower pension. The pension is increased for up to two children and for a low-income spouse. Separately, individuals whose disability causes them to need help with daily activities can receive an attendance allowance (a caregiver supplement) in addition to the pension – depending on the level of dependence, this adds roughly 50% to 188% of the full pension rate for those requiring constant attendance. Following a reform, benefit levels are adjusted annually according to the national average wage since 2022.

Disabled child allowance

The NII provides a monthly allowance to support families raising a child with disabilities or chronic illnesses. The benefit is generally payable for children up to age 18 who are Israeli residents and meet medical criteria defined in regulations on severe conditions or impairments. Benefit levels are tiered by the severity of the child’s condition. There are five rate levels (50%, 100%, 112%, 188%, or 235% of the base allowance) corresponding to the child’s care needs. The base full allowance for the 100% level is set at NIS 3,694 per month as of 2025. Additional supplements are provided in cases of special needs. Like the adult disability pensions, the disabled child benefits are indexed annually to wage growth.

Source: National Insurance Institute, 2025.

Disability-related NII spending

The projection of NII spending on disability allowances is estimated by multiplying the projected number of recipients by assumptions for average allowances. The recipiency rates for the basic disability allowance, the special services supplement and the disabled child allowances at the starting point are estimated separately for men and women (Figure 2.21). These rates are derived by forecasting NII administrative data for total recipients from 2022 to 2024 and dividing them by the size of the respective reference population groups according to the mixed population variant. By relying on actual COFOG spending data up to 2024 for the starting point, the model considers the rapid increase in recipiency rates over the last few years for the child disability allowance (Box 2.6) and the special services supplement.

Box 2.6. Rapid growth in child disability allowances: the role of autism diagnoses

Spending on disability benefits has increased rapidly in recent years. Between 2017 and 2024, spending on general disability allowances rose by 84% (in constant prices), and spending on special services allowances by 73%. The increase reflects a mix of demographic trends (including ageing), higher benefit levels and income thresholds, greater awareness of mental health, fewer housekeepers (who face stricter eligibility), the retirement age increase for women and policy changes that made access to benefits easier (e.g., by reducing the requirement to appear in person before medical committees).

Most notably, the number of recipients of the “Disabled Child Allowance” increased by 146%, reaching 142,855 in 2024, while spending rose by 226%. A key factor behind this trend is the rapid rise in eligibility based on autism spectrum diagnoses. The NII has granted eligibility to more than six times as many autistic children over the past decade, with an average annual growth of about 20%. The Ministry of Finance estimated autism-related government spending in 2024 at NIS 9.15 billion. Several systemic factors contribute to this surge. First, eligibility for benefits, including allowances and other social services (such as educational support and individual treatment by therapists), depends solely on clinical diagnosis, without factoring in support needs and entitlement to enhanced benefits. This aligns with the broader autism spectrum criteria in DSM-5 and NII’s acceptance of co-diagnoses. Growing social awareness and reduced stigma have further encouraged diagnoses. Additionally, a largely unregulated private diagnosis market, supported partly by HMOs, which can entitle individuals to a full public benefits basket, has expanded. The NII has also expanded remote assessments and proactive support for benefit claims. While these developments enhance accessibility and inclusion, they also raise concerns about targeting, equity, and the financial sustainability of disability-related expenditures.

Sources: Ministry of Finance; NII’s annual report (2023[35]).

Figure 2.21. Basic allowance recipiency rates increase exponentially with age

Recipiency rates, 2022 (% of population in the respective age groups) as a function of age

Source: OECD calculations based on National Insurance Institute data (2022).

Future developments of the recipiency rates in each age group are projected based on various factors. For adults aged 20 and above, projections for the basic disability allowance incorporate healthy ageing assumptions aligned with those used in the health expenditure vector. The assumption is that as people live longer, the likelihood that they will become disabled in a given year declines, shifting the recipiency rate curve to the right. For female cohorts aged 63–65, the legislated increase in the retirement age is considered by extending eligibility to this group. For recipients of the child disability allowance, the current upward trend in eligibility is projected to taper off by 2030, reflecting medium-term pressures driven by improved diagnosis and greater awareness of entitlements, which are expected to level off at elevated levels over the longer term12. The increase in recipiency rates for children aged 14–17 will create a lagged effect on the rates for the young adult group (ages 18–24).

Overall, the average population-weighted recipiency rates for the basic allowance are expected to remain largely stable at their 2024 level until 2030, estimated at 7.6% (6.1%) of the male (female) working-age population, before declining to 7.0% (6.0%) by 2065, reflecting anticipated improvements in the health status of the working-age population. Meanwhile, the population-weighted average for the child disability allowance is projected to rise from 4.4% to 4.8% by 2030 and remain at that level until 2065. Expenditures per recipient are projected to grow in line with average wage growth, reflecting the reformed indexation mechanism in place since 2022.

Work injury-related NII spending

Work injury–related spending is projected for both injury and permanent disability benefits, as well as for related expenditures on an aggregate level. Current recipiency rates for men and women are estimated using NII administrative data on total recipients as a share of the respective workforce. In 2024, the recipiency rate for the male workforce was 2.1%, compared to 1.0% for the female workforce. The recipiency rates are indexed according to the population-weighted recipiency rates of the basic disability allowance (described above) to project their future evolution. To estimate total current spending per recipient (including related expenditures and administration), the sum of the work-injury-related expenditures is divided by the number of recipients. Allowance growth is assumed to track wages.

Other sickness and disability-related expenditures

Other spending items classified under sickness and disability are projected either by relying on exogenous expenditure paths or through relevant indexation. This approach applies to disability pensions paid by the central government (treatment for injured soldiers), support for hostile action injuries (paid by the NII but reimbursed by the central government), assistance to Holocaust survivors, residential care and rehabilitation services for the disabled, and other disability-related services. Together, these items account for 1.2% of potential GDP and 38% of all disability expenditures in 2025.

2.4.4. Old age (COFOG function 10.2)

Total government expenditure on old age amounted to 5.1% of potential GDP in 2024. Over 75% of the COFOG old-age expenditures were allocated to old-age pensions. Pension expenditures were almost evenly split between the senior citizen benefit, covered by the NII, and budgetary pensions. However, budgetary pension payments are a legacy of the old pension system for civil servants, which has not accepted new entrants since the early 2000s. As a result, total expenditures within this system, which is part of the government budget, are projected to peak in 2040 and decline thereafter, while spending relative to GDP is already on a downward trajectory. In contrast, a very high share of Israel’s elderly population is eligible for old-age pensions under the National Insurance Law (Box 2.7). Already the largest single spending item in social protection, its relative importance is expected to increase further.

Apart from pensions, 18.5% of old-age spending was allocated to long-term care (LTC) expenditures covered by the NII in 2024. A detailed cohort model is used to project LTC expenditures, which have surged drastically in recent years.

Box 2.7. Public old-age pensions in Israel

Eligibility: The Israeli public old-age pension (Senior Citizen Benefit) is available to Israeli residents who have contributed to National Insurance for a minimum qualifying period and reached the statutory pension age. Generally, potential retirees need at least five years of insurance contributions in the last decade before pension age (or 12 years total) to qualify. Eligibility begins at retirement age – currently age 67 for men, and about 63.3 for women, with a 2021 reform gradually raising women’s retirement age to 65 by 2032. Between the retirement age and the higher age of entitlement (age 70), an income test applies. If a potential pensioner’s earnings exceed certain limits, the pension may be reduced or deferred. Upon reaching the age of entitlement, the pension becomes payable regardless of the individual’s income. Potential pensioners must be Israeli residents as non-residents, and those who immigrated past pension age generally do not qualify for the regular pension. Special provisions exist for those who immigrated at older ages or certain women without insurance contributions, ensuring they can receive support even if they do not meet the standard contribution record.

Allowance levels and increments: The old-age pension is paid as a flat monthly amount, with standard base rates set by law. The basic pension is roughly NIS 1,800 per month for a single pensioner (slightly higher for those aged 80 or above). This base rate is adjusted by family status: a pensioner with a dependent spouse (who has no pension of their own) receives an extra spousal allowance, and there are additional supplements for up to two dependent children. If both spouses independently qualify for a pension allowance, each receives their own individual rate.

Two key increments can further raise the pension amount based on one’s work history:

1. The seniority increment increases the pension by 2% for each full year of coverage up to a maximum 50% boost for long contribution histories. Minimum contribution thresholds that existed in the past have been abolished.

2. The pension deferral increment rewards those who continue working past retirement age and delay taking the pension: for each year the pension’s start is deferred (due to earned income) between the retirement age and 70, the eventual monthly pension is increased by an additional 5% (this is calculated on top of any seniority increment already earned).

Supplements, special cases, and adjustments: Low-income senior citizens may be eligible for an income supplement on top of the basic pension, which is a means-tested benefit aimed at guaranteeing a minimum income. This supplement depends on household status (higher for couples or those with children) and is slightly higher for older pensioners. Individuals who were receiving disability allowances before reaching pension age transition from the disability benefit to the old-age pension upon reaching the retirement age. However, their benefit amount is protected as the NII will pay whichever is higher of the prior disability pension or the new old-age pension they are entitled to.

Indexation of allowances: By law, pension amounts are uprated annually in line with changes in the Consumer Price Index, helping pensions maintain their purchasing power amid inflation. Periodic policy decisions can also affect benefit levels. For example, the recent gradual increase in women’s retirement age, or occasional legislative increases to the pension rates and supplements. As a result, allowance levels largely followed wage growth in the last two decades.

Source: National Insurance Institute, 2025.

Senior citizen benefits

Senior citizen benefit spending is projected in three steps. First, the number of eligible individuals is determined. Second, average entitlements are projected based on several key characteristics. Third, current average allowance levels are estimated and projected using indexation assumptions.

The number of senior citizen benefit recipients is projected based on the population projections, an assumption about the share of eligible elderly individuals, and the legislated age threshold provisions. According to the baseline population scenario, the number of people above the current retirement age (67 years for men and 63.3 years for women) is projected to increase from 1.3 million to 3.0 million by 2065, a 129% increase. Most of these individuals are expected to be eligible for an old-age pension. In the past, lower eligibility rates were largely due to old-age migration to Israel, particularly following the dissolution of the Soviet Union, and a higher share of women without National Insurance coverage. But already today, both factors play a minor role. For simplicity, the share of Israel’s elderly population eligible for an old-age pension from the NII is assumed to remain constant at 95% in the coming decades. In the baseline scenario, age threshold provisions, including the gradual increase in the retirement age for women, are projected in line with current legislation. People deferring retirement in a given year are deducted and reallocated to later years based on projections of the average pension deferral period (see below). Based on these parameters, the total number of pension recipients under the National Insurance Law is projected to increase from 1.24 million in 2025 to 2.73 million by 2065, a 120% rise (Table 2.10). The main reason for the difference between old-age population growth and the number of pensioners is the increasing retirement age for women.

Table 2.10. Projected number of pensioners by age group

Age group	Gender	2025	2045	2065	% change between 2025 and 2065
80 and above	Female	170,417	373,304	636,505	273%
	Male	117,664	275,407	494,562	320%
Age of entitlement (70) to 80	Female	328,321	445,127	568,786	73%
	Male	273,001	395,764	529,332	94%
Retirement age to 69, reallocating people who are deferring their pensions	Female	255,670	267,489	318,407	25%
	Male	97,276	148,452	184,387	90%
Total		1,242,349	1,905,544	2,731,978	120%

Source: OECD calculations based on CBS and NII data (2024).

The average entitlement of pensioners in gender-specific age groups is projected based on various characteristics, including the number of dependents, eligibility for increments, and income supplement needs. The average number of dependents per recipient is estimated for the six gender-age groups, using administrative data from the NII. Most pensioners do not have dependents. For those who do, a weighted average number of dependents is estimated. For simplicity, this average number of dependents is assumed to remain constant over time.

The seniority and pension deferral increments are estimated using a two-step approach: first for new pensioners, and then for the total pension stock. The average increment sizes for new retirees are modelled as reaction functions, accounting for employment trends and changes in retirement age (Box 2.8). For new female retirees, the seniority increment is projected to increase from 42.6% in 2024 to 45.1% in 2032 as the retirement age rises. It will increase more gradually afterwards. The pension deferral increment for newly retiring women is projected to decline from 4.8% in 2024 to 4.0% in 2032, due to the increase in the retirement age. Thereafter, driven by rising female old-age employment rates, it is projected to re-enter a slow upward trend, reaching 4.2% by 2065. For new male retirees, the increment is projected to increase gradually due to increasing employment among non-NHJO men, rising from 1.3% in 2024 to 2.3% in 2065. Due to the slow-moving effect on the pension stock, the average increment size is projected to increase from 2.2% in 2024 to 2.3% in 2065.

Box 2.8. Projecting the seniority and pension deferral increments

The average size of the seniority increment is projected separately for men and women. First, the seniority increment for new pensioners in a given year is projected. Second, the seniority increment for the entire pension stock is determined.

Seniority increment

The contribution period $CP$for new retirees (n) each year (t) is determined by an autoregressive element ${CP}_{t-1}^n$, the change in average lifetime employment $∆E_t$ of a retiring cohort and the change in retirement age in a given year ${∆R}_t$. $E_t$ adjusts very gradually so that it does not fully capture the rapid increase in the contribution period of women retiring during the retirement age reform under implementation until 2032. Therefore, $∆R_t$ is added as an explanatory variable to capture the rapidly increasing exit ages.

${CP}_t^n={\beta }_1{CP}_{t-1}^n{+\beta }_2{∆E}_t+{\beta }_3∆R_t$

The projected contribution period of pensioners is multiplied by a 2% annual contribution bonus. In line with current regulations, the maximum total bonus is capped at 50%:

${SI}_t^n={CP}_t^n*{Contribution bonus}_t\le 50\%$

The seniority increment for the total pension stock in year t, denoted as ${SI}_t$ is calculated as the sum of:

1. The seniority increment for new retirees ${SI}_t^n$, weighted by the number of new pensioners $\theta$, and

2. The seniority increment of the pension stock from the previous period ${SI}_{t-1}$ weighted by the number of people in the pension recipient stock in period t, excluding new joiners $1-\theta$ (assuming for simplicity that the seniority increment size within the pension stock is uniform):

${SI}_t=\theta {SI}_t^n+(1-\theta ){SI}_{t-1}\le 50\%$

The approach for projecting the pension deferral increment is equivalent. First, the size of the increment for new retirees is estimated, and then its effect on the pension stock average is assessed.

Pension deferral increment

The average number of years for which people defer their pension (${YPD}_t^n$) is a function of an autoregressive element $Y{PD}_{t-1}^n$ and the change in the old-age employment rate ${∆emp.}_t^{old age}$. To consider the fact that increases in the retirement age ${RE}_t$ shorten the period in which pension deferral is possible (as the allowance becomes universal from the age of entitlement at 70), the reaction function is subject to a multiplier $\frac{70-{RE}_t}{70-{RE}_{t=0}}.$

${YPD}_t^n=\left({\beta }_{1}Y{PD}_{t-1}^n{+\beta }_2{∆emp.}_t^{old age}\right)\frac{70-{RE}_t}{70-{RE}_{t=0}}$

The deferral increment ${DI}_t^n$ is equal to the ${YPD}_t^n$ times the annual deferral bonus of 5%.

${{DI}_t^n=YPD}_t^n*{Deferral bonus}_t$

The average increment size of the pension stock ${DI}_t$ is estimated equivalently to the approach of the seniority increment, with the same simplifying assumption:

${DI}_t=\theta {DI}_t^n+(1-\theta ){DI}_{t-1}$

The income supplement to an old-age pension is projected separately based on age threshold provisions and whether pensioners have dependents. Eligibility for the income supplement is significantly higher among pensioners above 80 and those with dependents. However, NII data reveals that the number of income supplement recipients as a share of all senior citizens has been gradually declining over the last three decades. These historical trends are assumed to continue due to low old-age migration, insurance coverage and increasing contributions periods. Eligibility for pensioners without dependents is projected to decline from 13.9% in 2023 to 9.7% in 2065 and from 33.4% to 29.2% for those with dependents over the same period.

Average allowances for the basic amount are estimated starting from 2024 NII pension rates. Various age- and dependent-specific scenarios are summarised using simplifying family composition assumptions, derived from NII administrative data on the prevalence of different family structures. The results yield four cases for the basic amount, distinguishing pensioners below and above 80 and those with and without dependents. The average seniority increment is added to the basic amount, while the pension deferral increment is applied to the sum of the two. Additionally, average income supplement amounts are incorporated based on actual allowance levels, also adjusted for age and dependent structure.

According to the National Insurance Law, since the early 2000s pension allowances are indexed to price inflation. However, in practice, allowance growth has tracked the average wage growth rate (see Figure 2.20 above). Therefore, under the no-policy-change assumption in the baseline scenario, it is reasonable to assume that allowances will continue to track wage growth. The income supplement, which grew much faster than average wages, is also assumed to follow wages.

Long-term care

Long-term care (LTC) services help people live as independently and safely as possible when they can no longer perform everyday activities on their own. General government spending on long-term care accounted for about 0.85% of GDP in 2023. This constitutes in-kind or cash benefits provided by the National Insurance Institute (NII) to every elderly person whose daily activities are restricted and who passes the means test and the test of dependency on others in daily activities. Spending on LTC, reported under the old-age function in COFOG, has recently increased, reflecting demographic developments and a policy shift in 2018 that significantly eased eligibility for LTC benefits (Box 2.9). The number of elderly individuals who receive LTC benefits from the NII increased by 88% between 2018 and 2023, while at the same time, the number of individuals aged 75+ grew by only 25%. Recent OECD analysis showed that the share of older people who receive any kind of LTC in Israel is the highest among OECD countries (Figure 2.22). Already before 2018, the rate of eligibility for public LTC insurance was high, since the threshold for non-functionality to receive benefits was low (Llena-Nozal, Araki and Killmeier, 2025[36]; Haran Rosen, Ramot-Nyska and Friedmann, 2018[37]). A previous Israeli study projected reaching these high levels of public spending on LTC (0.85%) only in 2045 (Geva, 2013[17]).

Figure 2.22. Coverage of long-term care in Israel is the highest among OECD countries

Share of elderly people with LTC needs who receive any formal care (including public and private care)

Note: The OECD average is unweighted.

Source: Llena-Nozal, Araki and Killmeier (2025[36]).

Future public expenditure on LTC will be determined by various factors that impact the demand for and supply of these services. First, as the population ages, unless there is a corresponding improvement in health and reduced disability, there will be an increase in the number of elderly people in need of LTC. According to the baseline population projections (the adjusted mixed variant), the weight of the 75+ age group will grow from 5.8% in 2023 to 10.3% in 2065. Second, the LTC system heavily depends on informal care provided by family and friends, often women, for the elderly in need. However, the increase in women's participation in the labour market suggests that the availability of informal care may decline over time. Additionally, as countries become wealthier, there is likely to be an increased demand for higher-quality and more accessible long-term care. Institutional arrangements for the provision and financing of LTC, such as the proportion of care financed by national insurance and the setting in which care is provided (home versus institutional care), will also play a role. Lastly, the supply of formal care workers, which highly depends on government policy as many workers in the sector are work migrants, will also impact expenditures on LTC services (European Commission, 2023[3]).

Box 2.9. NII support for long-term care

Provision of long-term care by the NII

NII spending on long-term care is mainly directed towards in-kind services or, in some cases, cash benefits for elderly individuals who remain at home or, under certain conditions, reside in long-term care institutions and nursing homes. The benefit covers personal care assistance at home, adult day-care centre services, and other related expenditures (e.g. incontinence supplies and emergency alarm devices). Eligibility is limited to Israeli residents who have reached retirement age and live at home or in qualifying assisted living settings, and who require substantial assistance with activities of daily living or close supervision for their safety. Private organisations and Non-Profit Institutions (NPIs) provide long-term care services.

Determination of service and benefit levels

Benefits and services are provided in six levels based on the severity of the person’s dependence (according to dependency points based on a functioning capacity examination), with higher levels granting more hours of care per week or a larger cash equivalent. The examination can be conducted in several ways, including remotely. The amounts granted are means-tested against income: seniors with a monthly income below a threshold receive the full benefit, those above a certain limit receive only 50%, and those exceeding an upper income cap are not entitled. The income test also considers whether recipients live alone or as part of a couple.

2018 reform of the long-term care system

The 2018 reform increased the number of levels from three to six to better align the number of care hours with the level of need. It also expanded eligibility to seniors with moderate initial care needs and introduced the option of cash benefits (full cash benefits at the lowest level and up to one-third of the benefit in cash at the higher levels). The eligibility threshold at the lowest level remained unchanged. The reform has led to a significant increase in approval rates and spending.

Source: National Insurance Institute, 2025.

The model used to project LTC spending is a cell-based or macro-simulation model consisting of three main parts (Figure 2.23): (1) projecting the number of recipients of LTC benefits based on the demographic trends and assumed dependency rates; (2) calculating the projected number of dependent persons in-home care and nursing home care for each age group, to obtain actual utilisation of services (controlled by age and gender). People in-home care could then be split into those who choose cash and those who prefer in-kind benefits; (3) indexing the average allowance by type of services based on projected macro developments. The projected average allowances are multiplied by the number of dependents to obtain the overall expenditure. The model focuses on formal care financed by the public sector. It does not cover formal care that is entirely privately funded, nor informal care, which relatives or friends provide. In 2022, public expenditure on LTC was estimated at 70% of all formal expenditures in Israel (Taub Center, 2024[38]).

Figure 2.23. Schematic overview of the long-term care projection methodology

Note: The average allowance is a function of Unit Cost (UC) and Income (I).

Source: OECD elaborations.

Projecting the number of recipients of LTC benefits based on CBS’s demographic projections, age-specific prevalence rates of dependency and changes in these rates based on life expectancy developments. The share of individuals who received LTC benefits (without considering dependency levels) in 2022 increases significantly at 80, when more than half of the relevant population receives at least some government support (Figure 2.24). As discussed in the health section, these shares could be expected to decline as health in older ages improves and life expectancy at age 65 increases (healthy ageing effect). This is modelled by shifting the age-specific prevalence rates of dependency downward in proportion to the projected gains in life expectancy. In the baseline scenario, this effect is assumed to be 0.5, meaning that for each additional year of life after 65, the time spent free of disability/dependency increases by 0.5 years. Lastly, the expected prevalence rates are multiplied by the projected population by age and gender, according to CBS’s population projections, to obtain the number of recipients in each period.

Figure 2.24. More than 50% of the 80+ population receives LTC, but healthy ageing could help

Source: OECD calculations based on National Insurance Institute data (2023).

Dividing the number of dependent persons according to the type of care: in-home and in cash, in-home and in-kind and in nursing home care. For the sake of simplicity, the share in each group was assumed to remain constant throughout the projection period. The separation does not affect the results in the baseline, as a similar indexing formula was chosen for cash and in-kind services. At the same time, the share of individuals receiving nursing home care in Israel is low and was projected as part of the health expenditures. Nonetheless, these assumptions could be replaced in case of a policy shift.

Indexing the average allowance according to macro developments: the average LTC allowance was NIS 4 000 in 2023, 72% of the minimum wage. Its expansion will be affected by unit cost developments (supply-side effects) and the increase in demand due to income growth. Separating these drivers is not straightforward, especially in a system where public insurance pays for the services. In the baseline, unit costs are assumed to be linked to labour productivity growth, given that the LTC sector is highly labour-intensive and productivity gains due to technological improvements are difficult to achieve. Constraints on the supply of formal care workers could further increase unit costs. Then, the actual change in allowances (and, therefore, expenditures) is determined by the price elasticity of demand. If an increase in care price is offset by a decline in quantity (elastic demand), expenditures will remain subdued. In contrast, in the case of inelastic demand, the unit cost will increase along with labour productivity. Following the Bank of Israel (Finkelstein, 2019[34]), the price elasticity of demand is assumed at 0.5. In addition, the income elasticity of demand is assumed to be 0.73, which is aligned with the assumption for health expenditures. This implies that an increase of 10% in GDP per capita induces a rise of 7.3% in the average allowance.

Non-NII old-age spending

The most important old-age expenditures not covered by the NII are budgetary old-age pensions (פנסיה תקציבית). In 2023, total expenditures for central and local government budgetary pensions amounted to 1.8% of GDP, comparable in scale to NII old age pensions. Budgetary pensions refer to unfunded, defined-benefit pensions paid directly from government budgets to former civil servants and public employees who entered service before 2002 (see Box 1.5 in Chapter 1). Recipients include former central and local government employees, police officers, teachers in the state education system and certain other civil servant occupations (while conceptually similar, budgetary pensions for former career military personnel are recorded in the defence function under COFOG). Budgetary pensions are provided in addition to the old-age pensions from the NII without any reductions if recipients meet the necessary National Insurance coverage criteria.

While new entries into the budgetary pension scheme ceased in the early 2000s, current eligible active workers are expected to continue retiring under the system through the 2030s. According to Israel’s Accountant General, annual payments in absolute terms are expected to peak around 2040. However, since no new beneficiaries are entering the system, the relative expenditure as a percentage of GDP is already declining, having peaked at 2.2% in the last decade. Based on the General Accountant's forecast in absolute terms and the OECD model's GDP projections, spending on budgetary pensions is expected to decline from 1.8% of GDP in 2023 to approximately 0.9% in 2040 and to 0.2% by 2065.

Another relevant budgetary expenditure consists of government capital transfers to Israel's old defined-benefit pension funds, which covered private-sector employees and were closed to new members in 1995. These transfers were introduced alongside other rebalancing measures, such as raising contributions and reducing pension rights for fund members (OECD, 2016[39]). The adjustments became necessary due to the significant actuarial deficits accumulated by the old pension funds. In 2023, government transfers still accounted for around 0.4% of GDP, but they are projected to decline gradually as the number of new retirees and overall beneficiary population decreases. Although these funds primarily covered private-sector employees, the OECD projections assume that expenditures on the capital transfers will decrease in parallel with public-sector budgetary pensions over time, reflecting the similar closed-entry profiles.

The old-age function also includes several smaller expenditure items that are not directly related to pensions or long-term care spending. These include rent assistance, support for residential care in institutions, housing provision, and various services for the elderly. Together, these NII-excluded expenditures accounted for 0.15% of GDP in 2023. Due to their small size, they are projected using indexation. Taken together, old-age-related spending not covered by the NII is projected to decline from 2.0% of potential GDP in 2025 to 0.5% by 2065.

2.4.5. Survivors (COFOG function 10.3)

Expenditures on survivors' benefits account for a relatively small and declining share of social protection spending. In 2024, they amounted to 0.46% of potential GDP, 25% lower relative to GDP than a decade earlier. Survivors' expenditures are divided into two categories: pensions and related expenditures paid by the NII and pensions paid by the central government. NII-covered survivors' expenditures accounted for two-thirds of total spending in this sub-function (Box 2.10). The remaining third of total survivors spending in COFOG is attributed to survivors' pensions paid by the central government.

In 2023, survivors above retirement age accounted for nearly 85% of the total survivors' pension stock. Generally, individuals who obtain survivor status due to the death of a relative and have accumulated enough insurance contributions to qualify for an old-age pension receive half a survivors' pension as a complement. Conversely, those who lack sufficient insurance coverage to qualify for a full old-age pension receive a full survivors' pension. Today, more than 70% of all old-age survivor recipients receive half a survivors' pension. According to NII administrative data, at the end of 2023, these accounted for 13.3% of all senior citizen benefit recipients. 93% of the people receiving both allowances are women. According to the NII annual report, this can be explained by several factors, including differences in insurance coverage and eligibility conditions, as well as higher female life expectancy.

Box 2.10. NII support to survivors

Eligibility criteria

The NII provides survivors’ pensions to spouses and dependent children of deceased insured individuals who met minimum insurance contribution requirements. Eligibility depends on age, family situation, and the deceased’s insurance history. Survivors who have reached the retirement age can combine their survivors’ entitlement with their own old-age pension. If they independently qualify for the senior citizen pension, they will receive their full old-age pension along with half of the survivors’ pension. They receive a full survivors' pension if they do not independently qualify for an old-age pension (for instance, due to insufficient contribution history).

Support levels and income supplement

Survivors’ pensions are calculated as flat-rate amounts, with additional supplements for dependent children and increments reflecting the deceased’s insurance seniority. Beyond pensions, the NII may also cover funeral expenses and survivors' rehabilitation. By law, benefits are indexed annually to CPI. Survivors with limited or no income beyond their NII pension are eligible for an income supplement, depending on their total income, household composition, and age.

Source: National Insurance Institute, 2025.

NII survivors spending is projected in two steps. First, recipiency rates for pensions and eligibility for income supplements are estimated relative to relevant reference groups, based on NII administrative data, and projected in line with current trends. Second, allowance levels are forecasted for the different recipient groups, starting from an estimate of current levels.

Recipiency rates are projected for the five agent groups described above (Table 2.11). Overall, recipiency rates, except for orphans, are expected to decline over time, continuing the negative trends observed in past decades. Several factors contribute to this decline. Marriage rates have been falling across all Israeli population groups, while age differences between spouses have narrowed. Additionally, lower mortality rates reduce the share of survivors below retirement age. Among senior citizen recipients, the share also receiving a survivors' pension is projected to gradually decline by one percentage point between 2023 and 2065. This decline is relatively slow due to increasing survivors' insurance coverage among old-age individuals, leading to shifts from full survivors' pensions to half-survivors' pensions. The share of full survivors above retirement age relative to the old-age population is projected to fall by 1.5 percentage points by 2065. This trend is expected to be steeper until 2032 due to the rising retirement age for women, which pushes women under 65 into the below-retirement age group. Accordingly, the negative trend for survivors below retirement age is projected to be slower before 2032.

The eligibility for the means-tested income supplement is also projected separately for the five recipient groups (Table 2.11). NII administrative data suggest that eligibility for income supplements among survivors below retirement age has been declining over the past two decades, a trend expected to continue. For recipients of a full pension above retirement age, eligibility for the supplement has remained mostly constant in the past but increased following the pandemic. It is projected to remain stable at 39%, the average over the past five years. For recipients of half a survivor's benefit, eligibility for the income supplement is projected to align with that of the overall senior citizen pension stock.

Table 2.11. Survivors benefit recipiency rates are projected to decline slowly

Recipients of survivors’ pension, % of reference group (of which are eligible to a means-tested income supplement)

Agent group	Reference group	2023	2032	2065
Above retirement age, receiving half a survivors pension (eligible for senior citizen benefit)	Senior citizen recipients	13.3% (13.9%)	13.0% (12.3%)	12.3% (9.8%)
Above retirement age, receiving a full survivors pension	Population above retirement age	4.2% (39%)	3.6% (39%)	2.7% (39%)
Below retirement age, no dependent children	Population aged 21 to retirement age	0.36% (20%)	0.36% (16%)	0.29% (9.8%)
Below retirement age, with dependent children	Population aged 21 to retirement age	0.23% (12%)	0.23% (8%)	0.16% (1.8%)
Orphans	Population aged 0-20	0.17% (1%)	0.17% (1%)	0.17% (1%)

Source: OECD calculations based on NII administrative data and CBS population projections (2023).

Average allowances for each group are estimated based on NII 2024 allowance rates. For groups with dependent children, a weighted average is calculated using NII data on dependent composition, which is assumed to remain constant in the future. The impact of this simplifying assumption is limited since the share of recipients with dependents is small. The base allowance is assumed to be indexed 75% to wage growth and 25% to CPI. Income supplement amounts have grown significantly faster than the average wage over the past two decades. Therefore, indexing income supplements to wages is chosen as a conservative assumption (see Figure 2.20 above).

2.4.6. Family and children (COFOG function 10.4)

Spending on family and children is the third-largest category within the social protection function, following old age and sickness and disability. It peaked at 1.4% of potential GDP in 2017, due to a strong increase in the child allowance, but has declined to 1.1% in 2024. Within this sub-function, NII-covered spending, including the child allowance, along with voluntary parent contribution to the Savings Plan for Each Child and the study grant, public spending on the Savings Plan for Each Child, as well as the maternity allowance, birth grant and paid maternity leave, accounted for over 75% of expenditures in 2024 (Box 2.11). The largest non-NII component comprised child and family welfare services.

The detailed OECD modelling of the family and children vector focuses on three key blocks of the NII spending:

• Child allowance and related grants: This includes spending on the child allowance, the voluntary parental contribution to the Savings Plan for Each Child, and the study grant. Projections are based on demographic trends and assumptions regarding allowance levels depending on the family composition;

• Savings Plan for Each Child: This covers public spending on the Savings Plan for Each Child programme, taking into account the 2017 eligibility cut-off;

• Parental benefits, hospitalisation and birth grants: Projections driven by wage growth, the gender pay gap and policy variables.

Box 2.11. NII support to children and parents

Child allowance

The NII pays a monthly child allowance for all resident children up to 18 years old, regardless of parental income. The amount depends on household size – as of 2025 it is about NIS 169 for the first child, NIS 214 for each of the second, third and fourth, and NIS 169 for the fifth and each additional child. Parents may opt to contribute NIS 57 from the monthly child allowance to the Savings Plan for Each Child (see below). In 2017 there was a notable surge in child benefit payments due to a government boost in child allowances, with a previously approved increase (deferred from 2015–2016) implemented alongside the new child savings scheme from that year. The child allowance is indexed to CPI.

Savings Plan for Each Child and study grant

Since 2017, every child eligible for the allowance also benefits from the Savings for Each Child programme. The NII opens a savings account for each child and deposits about NIS 57 per month into it (plus the potential voluntary contribution from parents, doubling the savings to roughly NIS 114 a month). The accumulated savings become accessible to the child at age 18 (with an extra NIS 500 grant added by the government), or at 21 if they choose to wait – in the latter case an additional NIS 500 bonus is granted. Additionally, an annual study grant is paid at the start of each school year to help with education expenses for low-income families. This grant (around NIS 1,176 per school-age child) is provided to single-parent households and to families with four or more children who receive income support or similar NII benefits.

Maternity allowance and birth grant

Employed mothers are entitled to 15 weeks of paid maternity leave (birth and parenthood leave) funded by the NII, provided they have contributed to National Insurance for at least 10 of the 14 months or 15 of the 22 months prior to stopping work. Those with a shorter contribution record (minimum 6 months) receive a partial paid leave of 8 weeks. Maternity leave is compensated at about 100% of the mother’s average wage (based on the highest average of the last 3 or 6 months) up to a capped daily amount. New mothers also receive a one-time birth grant for each child, paid automatically after delivery. The standard grant is roughly NIS 2,054 for a first-born, NIS 924 for a second, NIS 616 for a third or later child and substantially higher sums for multiple births (e.g. over NIS 10,000 for twins).

Source: National Insurance Institute.

The child allowance and related expenditures are projected based on the number of minors and an estimated average allowance level. The share of children in the total population is expected to decline by 8 percentage points between 2025 and 2065, which might also lead to a lower average number of children per family. However, for simplicity, the model assumes a constant family composition over time: 41% first-born, 29% second-born, 16% third-born, and 7% each for fourth and fifth children (with alternative scenarios tested in the sensitivity analysis). Using these assumptions and the per-child allowance rates, an average weighted allowance is estimated. Annual allowance growth is indexed to 50% of the projected increase in the Consumer Price Index (CPI) and 50% to projected wage growth, even though, by law, the allowance should be indexed exclusively to consumer prices (see Box 2.11 and Figure 2.20). The model then uses the growth rate of the average allowance and the growth rate of the number of children to project total expenditure based on 2025 levels, incorporating the voluntary parental contribution to the Savings Plan for Each Child, the study grant and other related benefits.

Public spending on the Savings Plan for Each Child is modelled considering the basic sums, bonuses (at age 3,12/13, 18 and 21) and administrative expenses. The model projects the basic sum and bonus levels according to the indexation formula based on CPI and wage growth. It then estimates future spending based on the number of children in each relevant age group. The bonuses are added to the funds to all children while they get to the age of 3 and 12/13 in case these children were born after 2017. For those born before 2017, the same amount is given when the child reaches 18. It means that in the following years, expenditures are a bit higher as the government pays bonuses both to those born before 2017 and to those born after 2017 (at the age of 3, and since 2030 also at the age of 13). In addition, there is a bonus to those who do not withdraw the money before their 21st birthday. The share of children maintains the money is calculated based on current data and assumed to remain constant in the future. Admin expenses are assumed to increase according to the other components of the programme.

The maternity leave benefit is projected based on the average employment rate of women aged 25–44, an assumption about the recipiency rate among mothers and the average allowance. The reference employment rate is expected to increase by 4.9 percentage points between 2025 and 2065, leading to a proportional rise in the recipiency rate from 79.5% in 2023 to 85.3% in 2065. The average allowance is projected to grow in line with average wage growth, assuming a constant generosity level of 86% of the average wage. The latest actual expenditure is indexed to both the growth rate of the recipiency rate and the increase in the allowance. Expenditures on pregnancy bed rest are indexed based on the average allowance, whereas administrative expenditures are indexed to the total maternity benefit expenditure.

The birth grant is projected based on the total number of births per year and the average allowance per child. The average allowance is determined by assumptions about the composition of receiving families, which is assumed to remain constant (as with the child allowance), and the different allowance levels (described in Box 2.11). The per-child allowance is projected to increase in line with a 50/50 mix of CPI and wage growth, based on the historic development (despite being legally linked to CPI). All other NII expenditures in the vector, making up only a very small residual in 2024, are projected based on the number of total births per year.

2.4.7. Unemployment (COFOG function 10.5)

NII expenditures on the unemployment insurance accounted for 0.4% of potential GDP in 2024. Except for the COVID-19 pandemic, when the unemployment rate temporarily surged, unemployment spending has been relatively stable around 0.3-0.4% of potential GDP for over a decade. Going forward, unemployment benefits are projected in three steps. First, the number of unemployed individuals is calculated. Second, eligibility and generosity are estimated for the forecasted recipient profiles. Third, total spending is projected based on an assumption of allowance growth.

The number of unemployed individuals by age and gender is calculated based on demographic projections, the unemployment rate in a steady state, and changes in labour force participation. The unemployment rate is projected for gender-specific five-year age brackets. It is assumed to stay the same over the projection period as in the last projected year in the latest OECD Economic Outlook, implying that employment is close to its potential. This assumption seems plausible for Israel, where both the current and potential unemployment rates are low. Changes in labour force participation reflect shifts in employment as indicated by the OECD LTM model.

To assess the eligibility of the unemployed individuals for the benefit and the respective generosity levels, the variations in eligibility by gender and age, as well as demographic trends, are considered. In general, unemployment benefits in Israel are provided as short-term benefits to eligible employees who have worked the required period in the previous 18 months. Benefit calculations use prior wages and vary depending on the applicants’ characteristics. Benefit duration ranges from a few months for younger applicants without dependents to potentially 10 months (300 days) for older female workers close to retirement with multiple dependents. However, most recipients receive only a few months of coverage before shifting to the basic social safety net. The maximum duration of benefits is relatively short in comparison to many OECD countries. Once unemployment insurance expires or if an individual fails to meet the contribution requirements, they may apply for the means-tested guaranteed income allowance (see the model under other social protection expenditures below).

A simplified generosity index is calculated to determine the unemployment insurance for each gender-age subgroup. This measure is determined by the maximum number of days an unemployed person from each subgroup can remain eligible for benefits, as well as the benefit level for those who previously earned the average wage. For simplicity, all recipients are assumed to have two or fewer dependents (Table 2.12). Finally, the generosity measure is multiplied by the number of unemployed individuals in each subgroup to derive an aggregate generosity measure.

To project total spending, the latest actual spending figure is multiplied by the annual growth rate of the aggregate generosity measure as well as annual wage growth.

Table 2.12. Generosity index to determine unemployment insurance for each gender-age subgroup

Gender-age group	Maximum number of days (up to 2 dependents)	Average benefit level	Generosity measure: Maximum days * the average benefit level (up to two dependents)
20-24	50	0.49	24
25-27	67		33
28-34	100	0.64	64
35-44	138		88
45-56	175		112
Men 57-66	175		112
Women 57-66	300		191

Note: For beneficiaries under 28, benefit levels are a weighted average of 60% for those earning up to 0.5 × the average wage, 40% for those earning 0.5–0.75 × the average wage, 35% for those earning 0.75–1.0 × the average wage, and 25% for those earning above the average wage.

For beneficiaries 28 or older, the respective percentages are 80% (up to 0.5 × average wage), 50% (0.5–0.75 × average wage), 45% (0.75–1.0 × average wage), and 30% (above the average wage).

Source: OECD calculations.

2.4.8. Other social protection expenditures (COFOG functions 10.6-10.9)

There are various other social protection expenditures that account for a relatively small share of overall spending. These include housing (COFOG 10.6), social exclusion n.e.c. (10.7), and social protection n.e.c. (10.9). Together, these items amounted to 1.3% of potential GDP in 2024, or about 11% of total social protection spending. The OECD projections treat these expenditures differently.

Spending on housing increased by over 340% as a share of potential GDP between 2022 and 2024, primarily due to self-evacuation and accommodation costs related to Hamas’ attacks and the conflicts in Gaza and Lebanon. These expenditures are projected to gradually phase out by the end of 2026. Other housing expenditures, including in-kind benefits (such as public housing) and capital transfers for dwelling purchases, are indexed under the constant service-level assumption.

Social exclusion spending covers the income support benefit (NII), cash benefits for needy families and new immigrants, amounting to 0.4% of potential GDP in 2024 – a 13% decrease compared to ten years earlier. In 2023, the income support benefit accounted for the largest sub-item (0.1% of potential GDP). Its future evolution starts from the latest actual spending and is projected based on working-age population growth, the unemployment rate, family-type composition (assumed constant), and benefit indexation (assumed to follow CPI in the baseline). Changes in unemployment affect the number of beneficiaries according to the correlation between the log unemployment rate and the log number of beneficiaries from 2010 to 2024. For new immigrants, an exogenous projection of immigration to Israel is multiplied by a constant expenditure per immigrant, set at 39% of GDP per capita in the year of arrival. All other items in this function, including spending for needy families, follow the constant service-level assumption.

Finally, spending on social protection n.e.c., which amounted to 0.3% of potential GDP in 2024 and remained largely constant over the past decade, is indexed based on the constant service-level assumption.

2.4.9. Social protection spending throughout 2065

Total social protection spending is projected to fall from 11.1% of potential GDP in 2025 to 9.0% by 2065. The main reason for the negative trend is the 2.3 percentage point decrease in non-NII spending (Figure 2.25 right panel), reflecting the phase-out of budgetary pensions and capital transfers to old DB pension funds.

Meanwhile, NII spending is projected to increase slightly as a share of potential GDP, growing from 6.5% in 2025 to 6.8% by 2065 (Figure 2.25, left panel). Despite the moderate increase, there are notable compositional changes. Old-age spending, including pay-as-you-go pensions and long-term care benefits, is set to rise by 1.1% of potential GDP. During the retirement age increase period (until 2032), old-age spending remains temporarily stable as relatively fewer women enter retirement. Afterwards, spending is projected to enter an upward trajectory, driven by demographic developments and increasing pension increments.

All other NII expenditures will shrink by 0.8%, primarily driven by a nearly 50% reduction in family and children expenditures and the gradual normalisation of sickness and disability payments. This shift is driven in part by demographic changes – namely, a growing elderly population, fewer children, and a relatively stable working-age cohort – but the indexation formulas also play a significant role. Based on historic data, family and children allowances are assumed to track a combination of CPI and wage growth (see Figure 2.20 and Table 2.9 above) while old-age allowances are projected to continue growing at the pace of wages. The divergence in allowance indexation will cause family and children expenditures to become relatively marginal and fall behind overall GDP growth in the long term.

Figure 2.25. Budgetary spending will decline, while NII spending is projected to remain stable

Social protection spending (% of potential GDP)

Source: OECD calculations.

The difference between the projected increase in NII expenditures of 0.2% of potential GDP over the coming decades and the Bank of Israel (BOI) study published in 2019, which projected a 0.8 percentage-point increase in total spending between 2018 and 2065 under a policy-continuity scenario (Finkelstein, 2019[34]) can be explained by several factors. A key difference lies in sickness and disability spending: the BOI forecasted additional spending amounting to 0.5 percentage points of GDP, reflecting a prolonged rise in the share of the population qualifying for disability benefits. Going forward, a crucial question is whether the recent surge in disability benefit recipiency rates will persist or gradually stabilise. Moreover, spending on family and children support, which the OECD projects to decline significantly, was forecasted by the BOI to remain almost stable relative to GDP. This deviation can be explained for the most part by changes to the underlying demographic scenario and differing assumptions about allowance indexation. Another factor is the starting point difference between 2018 (BOI) and 2025 (OECD).

2.4.10. Contingent liabilities in the social protection system

Israel’s legally mandated expenditure ceiling (see Chapter 1) does not capture all deficits generated in Israel’s public institutions. In the case of social protection, deficits are absorbed by the National Insurance Fund, although, as explained in Section 1.4.5 in Chapter 1, it represents a contingent liability for the central government and impacts central government interest payments. Thus, a simulation of the NII deficit and the volume of the National Insurance Fund is conducted.

The starting value of the Fund was NIS 252 billion in fair value terms in 2023. The volume of the Fund is forecasted by subtracting the projected annual balance of the NII up to 2065. Figure 2.26 shows that the volume of the Fund as a share of potential GDP will decline over time. The NII is projected to generate growing deficits as old-age expenditures rise. The Fund is projected to be depleted by 2037. From that point onwards, the NII deficit will entirely fall under the budget ceiling as social protection spending. This result is in line with the projection published in the NII’s 2023 Actuarial Report, which expects Fund resources to be exhausted already by 2036, despite different assumptions13. Increasing the retirement age of women to that of men and linking the future statutory retirement age to changes in life expectancy could postpone the depletion of the NII’s fund only by a few years, as the gains from higher retirement ages mostly materialise once the fund is already exhausted.

Figure 2.26. The NII balance will deteriorate in the long term

National Insurance Fund volume (left axis) and NII balance (right axis), % of potential GDP

Source: OECD calculations.

2.4.11. Scenario analysis

The sensitivity of NII expenditures is tested across a range of scenarios (Table 2.13). Assuming that allowances increase only in line with their legislated basis, rather than historic trends, significant savings of up to 1.2% of potential GDP could be generated by 2065, relative to the baseline. Most of these reductions would stem from old-age benefits, where allowances are assumed to grow well above the legally foreseen CPI-based adjustments in the baseline scenario.

Alternative population growth scenarios would affect future spending in heterogeneous ways. Under higher population growth, NII spending would also increase by 0.2% of potential GDP by 2065, however, with a different composition than in the baseline. Family and child-related expenditures would decline less markedly, while the increase in old-age spending would be moderated due to a smaller rise in the old-age dependency ratio and slower wage growth from weaker productivity gains. Under lower population growth, a higher employment-to-population ratio would lower the number of dependent individuals but simultaneously increase spending that is sensitive to wage growth, so the overall change by 2065 is close to the baseline.

Further increases in the retirement age and measures to compress morbidity could be particularly effective in curbing the rise in old-age spending. These reforms, which would reduce pension and long-term care transfers to the elderly while increasing their labour market participation, could lower NII spending in the old-age function as well as overall NII spending by 0.4% of potential GDP by 2065.

Several function-specific scenarios are also examined. Keeping recipiency rates for disability benefits at current levels would exert modest upward pressure on sickness and disability spending, amounting to 0.2% of potential GDP by 2045 and 0.1% by 2065. Assuming a higher income elasticity of 1 for long-term care, where all income gains translate into higher spending, would raise NII old-age expenditure by an additional 0.3% of potential GDP by 2065. Raising the indexation of all NII family- and children-related benefits to wage growth to prevent their erosion relative to GDP per capita and the quality of living would raise expenditure by 0.1% of potential GDP above the baseline by 2045 and 2065.

NII spending is also sensitive to the macroeconomic effects of labour market convergence across population groups. These effects primarily operate through productivity growth, influencing wage levels and, by extension, allowance indexation. In the Melting pot scenario, with lower productivity but higher overall GDP growth, total NII social protection spending would be 0.2% lower relative to potential GDP by 2065, whereas in the Frozen rates scenario, with higher productivity and weaker overall GDP growth, NII spending would increase by 0.3%, relative to the baseline.

Table 2.13. Overview of NII spending under different scenarios

Change between 2025 and 2065 (2045), % of potential GDP

COFOG sub-function	Baseline scenario	Allowance growth scenarios	Demographic scenarios		Policy scenarios		Labour market convergence
	Legislated retirement age change, healthy ageing = 0.5, plausible allowance indexation	Allowances are indexed according to the legal basis	Faster population growth (CBS median scenario)	Slower population growth (CBS low scenario)	Increase in statutory retirement beyond legislated period and an additional compression of morbidity	Function-specific key policy change*	Frozen rates	Melting pot
10.1 Sickness and disability	-0.2 (-0.1)	-0.2 (-0.1)	-0.1 (-0.1)	-0.2 (-0.1)	-0.2 (-0.1)	-0.2 (0.1)	-0.1 (-0.1)	-0.3 (-0.2)
10.2 Old age	+1.1 (+0.5)	+0.1 (+0.0)	+0.8 (+0.5)	+1.0 (+0.4)	+0.7 (+0.3)	+1.4 (+0.6)	+1.3 (+0.6)	+1.0 (+0.4)
10.3 Survivors	-0.03 (-0.03)	-0.15 (-0.1)	-0.03 (-0.03)	-0.03 (-0.04)	-0.1 (-0.05)	+0.04 (+0.02)	-0.01 (-0.02)	-0.04 (-0.04)
10.4 Family and Children	-0.5 (-0.3)	-0.6 (-0.3)	-0.4 (-0.2)	-0.4 (-0.3)	-0.5 (-0.3)	-0.4 (-0.2)	-0.4 (-0.2)	-0.5 (-0.3)
10.5 Unemployment	0.0 (0.0)	0.0 (0.0)	0.0 (0.0)	0.0 (0.0)	0.0 (0.0)		0.0 (0.0)	0.0 (0.0)
10.6-10.9 Other spending	-0.1 (-0.1)	-0.1 (-0.1)	-0.1 (-0.1)	-0.1 (-0.1)	-0.1 (-0.1)		-0.1 (-0.1)	-0.1 (-0.1)
Total spending	+0.2 (+0.0)	-1.0 (-0.7)	+0.2 (+0.1)	+0.2 (-0.1)	-0.2 (-0.3)		+0.6 (+0.2)	0.0 (-0.2)

Note: Function-specific key policy scenarios for the different sub-vectors are considered for each vector separately. As they are regarded as ceteris paribus (all else equal) there is no aggregation for this scenario. Sickness and disability: Recipiency rates remain constant at current levels for the basic disability allowance, child disability allowance, and special services supplement, Old-age: The income elasticity of long-term care is higher (=1), Survivors: Recipiency rates remain at current levels, Family and children: Allowances are growing according to wage growth.

Source: OECD calculations.

2.5. Other spending

2.5.1. Defence spending

Military expenditure as a share of GDP steadily declined over the two decades preceding the war, reaching 4.9% of GDP in 2022. However, the 7 October 2023 terror attacks and the subsequent war prompted a significant rise in defence spending. In 2024, defence expenditure – based on COFOG definitions – rose to 8.9% of GDP, 4 percentage points higher than in 2022. According to the 2025 budget (based on Ministry of Finance definitions), defence spending is expected to decrease by about 6% relative to 2024, though actual execution may be higher due to the prolongation of the war. Roughly one-third of military expenditure is allocated to capital formation, while the remainder primarily covers compensation of employees, pension entitlements, and intermediate consumption.

It remains highly uncertain to what extent the recent rise in military spending will hold into the future. In principle, based on an analysis of government commitments reported in the medium-term budgetary plan (the Numerator publication), the spending path is expected to decline back toward pre-war levels gradually. Nonetheless, the authorities are considering permanent increases in military expenditure and reserve duties to raise the number of active combat personnel, strengthen various military branches in terms of both human and technological capabilities, and establish infrastructure to reduce the risk of infiltration. If even part of these demands is accommodated, the implications for fiscal aggregates could be significant, with potential spillovers to long-term output. The evidence on the economic effects of military spending is mixed, with significant potential for negative effects on long-term growth (OECD, 2025[11]). 

Given these concerns, the government established the Nagel Committee to assess Israel’s force-building needs and recommend a sustainable defence budget for the next decade. The Committee advised increasing military spending to approximately NIS 97 billion annually (in 2024 prices) between 2025 and 2034 (according to the MOF’s definitions for spending of the Ministry of Defence). This would represent approximately 4.5% of the projected 2026 GDP (compared with about 3.5% in 2022), with front-loaded investment following a relatively sharp decline in terms of potential GDP after 2028.

In the baseline scenario, defence spending, as defined under COFOG, is assumed to follow the trend of Ministry of Defence expenditures implied by the recommendations of the Nagel Committee up to 202914. From that point forward, defence spending is assumed to remain constant as a share of potential GDP (Figure 2.27). Thus, defence outlays are not expected to generate additional fiscal pressures – either upward or downward – over the long term. Since security threats are not directly linked to population size, this assumption errs on the conservative side.

Figure 2.27. Defence spending is assumed to decline in the medium term and remain steady afterwards

Government expenditures on defence spending, COFOG, % of potential GDP

Source: OECD calculations.

2.5.2. Transport and infrastructure investments

According to OECD National Accounts Statistics, government investment encompasses gross capital formation and net acquisitions of non-produced non-financial assets (i.e., acquisitions minus disposals). In 2023, these were below the OECD average in Israel. The main component of government investment across OECD countries, including Israel, is gross capital formation. However, in Israel, significant land privatisations are causing disposals to exceed acquisitions (meaning the government is selling off more of these assets than buying), significantly pushing government investment downwards as they appear as a negative figure in investment aggregates. When comparing only gross capital formation as a share of GDP, they seem to be higher than the OECD average by about one percentage point (Figure 2.28).

Figure 2.28. Fixed investment is higher than in most countries

Government investment and gross capital formation as a % of GDP

Note: The blue triangles show data for gross capital formation and net acquisitions of non-produced non-financial assets. Investment in the energy sector, seaports, and airports, as well as communication infrastructure, is carried out primarily by private firms, including state-owned enterprises such as the Israel Airports Authority. As such, it is classified as private investment in national accounts and is therefore excluded from this chart. The figure for Korea only includes data for 2022. The sample average is for the years 2022-23.

Source: OECD calculations.

Further breaking down government gross capital formation by COFOG function reveals significant disparities in public investment between Israel and other OECD small economies (Figure 2.29). The figure compares Israel to Austria, Denmark, Sweden, Czechia, the Netherlands, Belgium, Finland, and Switzerland, as in other publications on the subject (Aaron Institute, McKinsey). Investment in education and defence is high in Israel. This reflects Israel’s high fertility rate and the resulting larger youth population, as well as persistent security needs. On the other hand, investment in general public services (mainly basic research) is low. Importantly, investment in transport, which is reported under economic affairs, was higher in the last few years than in Denmark, Belgium, Finland, the Netherlands, and Switzerland, but lower than in Austria, Czechia, and Sweden.

Figure 2.29. Israel invests more in defence and education

Gross fixed capital formation as a share of GDP by COFOG function, 2018-2023 average

Note: The chart compares Israel to Austria, Denmark, Sweden, Czechia, the Netherlands, Belgium, Finland, and Switzerland.

Source: OECD elaborations based on COGOG database.

Previous OECD surveys, as well as publications by the Bank of Israel, the IMF, and Israeli think tanks, emphasise the need to accelerate investment in transport, mainly public transport, to close the significant infrastructure gap and prepare for strong growth (IMF, 2018[40]; Bank of Israel, 2019[41]; Sagi and Shafir, 2024[42]; OECD, 2023[10]; OECD, 2025[11]). In recent years, government investment has indeed appeared to be higher and is planned to increase further, by about ¼ of a percentage point of GDP by 2030, compared with the planned expenditures for 2025 (Ministry of Finance data). The expected rise is primarily to initiate the construction of a Metro system in the Tel Aviv metropolitan area (+0.44% of annual GDP), which will more than offset the expected decline in investment in roads and other mass transportation.

Figure 2.30. Public investment in transport has increased

Public fixed investment (gross capital formation) in transport, % of GDP

Note: The sample of small, advanced economies includes Austria, Denmark, Sweden, Czechia, the Netherlands, Belgium, Finland, and Switzerland.

Source: OECD elaborations based on COFOG database.

In the baseline scenario, the expected investment in transport infrastructures up to 2030 is estimated to follow the Ministry of Finance's estimations, based on already approved infrastructure projects. From that point onwards, it is assumed to remain constant as a share of GDP at the elevated 2030 levels. This means that public expenditures would increase by 0.3% of potential GDP from 2025 to 2030. Thus, looking ahead, Israel is projected to become one of the OECD countries with the highest public investment in transport infrastructure. Among the selected OECD small economies, only Czechia recorded a higher share in 2023. However, due to rapid population and GDP growth, the resulting increase in infrastructure capital stock as a share of GDP would be slower. Assuming the same investment efficiency and depreciation rates in all countries, and that public investment remains constant as a share of GDP as it was on average from 2019 to 2023 (excluding Israel, for which the path described above is assumed), Israel’s transport capital stock could eventually (in the steady state) exceed that of five selected countries, but still trail behind three others15.

The efficiency of investment will be key to ensuring that the growth benefits from public investment are achieved. Selecting projects with low rates of return or raising investment faster than absorptive capacity can lead to weaker growth benefits. To enhance the efficiency of public investment, Israel should ensure that project evaluation and selection are rigorous and transparent, including by ensuring consistency with a long-term infrastructure strategy and streamlining permitting processes and other bureaucratic impediments to timely project implementation (IMF, 2018[40]).

2.5.3. All other spending

To project all other expenditures, the model assumes that the government seeks to maintain a constant level of service provision per capita, following (Guillemette and Turner, 2017[32]). At the same time, it is assumed that prices of public sector services in this residual category follow wages in the rest of the economy. This implies that changes to the employment rate will majorly influence the evolution of government residual primary expenditure as a share of GDP, as demonstrated by the following equation.

Nominal government spending on residual primary expenditure ($E_t$) can be expressed in terms of the volume of services provided (𝑉𝐸𝑡) and their price index (${PE}_t$):

$E_t$ = ${PE}_t$ ∙ ${VE}_t$

Expanding this definition with total population (𝑃𝑂𝑃𝑡) and employment (𝐿𝑡):

$E_t={\mathrm{P}\mathrm{E}\mathrm{ }}_t\bullet \mathrm{ }\frac{{VE}_t}{{POP}_t}\bullet \mathrm{ }\frac{{POP}_t}{Lt}\mathrm{ }\bullet \mathrm{ }{L}_t$

Dividing both sides of the second equation by nominal GDP (𝑌𝑡 = 𝑃𝑌𝑡 ∙ 𝑉𝑌𝑡, where 𝑃𝑌𝑡 is the GDP deflator and 𝑉𝑌𝑡 is real GDP), taking logs and differentiating both sides of this expression yields the growth rate of residual primary expenditure as a share of GDP:

$\Delta \mathrm{ }log\left(\frac{\mathrm{E}\mathrm{t}\mathrm{ }}{Y_t}\right)=\mathrm{ }\Delta \mathrm{ }log\left(e_t\right)=\mathrm{ }\Delta \mathrm{ }log\left(\frac{{\mathrm{P}\mathrm{E}\mathrm{ }}_t}{{PY}_t}\right)+\mathrm{ }\Delta \mathrm{ }log\left(\frac{{VE}_t}{{POP}_t}\right)+\mathrm{ }\Delta \mathrm{ }log\left(\frac{{POP}_t}{L_t}\right)+\mathrm{ }\Delta \mathrm{ }log\left(\frac{L_t}{{VT}_t}\right)$

The first term on the right-hand side is the growth rate of real prices for government services in the residual primary expenditure category. The second term is the growth rate of the volume of services per capita. The third term is the growth rate of the population-to-employment ratio. And the last term is the growth rate of the inverse of labour productivity (i.e. negative labour productivity growth). Given that government expenditure is mostly made up of wages, and assuming both that wages in the government sector follow wages in the rest of the economy and, as is conventional, that real wages grow at the rate of labour productivity, then the first and last terms on the right-hand side of the above equation cancel each other out. Furthermore, if the government maintains a constant level of service provision per capita, then the second term on the right-hand side is zero. Thus, only the third term remains, making it clear that residual primary spending as a share of GDP will change only to the extent that the population-to-employment ratio changes, so the equation above becomes the one below. An increase in this ratio would raise public spending on all other functions as a share of GDP. On the other hand, reforms that boost employment would help to reduce the fiscal burden.

$\Delta \mathrm{ }log\left(e_t\right)=\mathrm{ }\Delta \mathrm{ }log\left(\frac{{POP}_t}{L_t}\right)$

References

[20] André, C., P. Gal and M. Schief (2024), “Enhancing productivity and growth in an ageing society: Key mechanisms and policy options”, OECD Economics Department Working Papers, No. 1807, OECD Publishing, Paris, https://doi.org/10.1787/605b0787-en.

[8] Bank of Israel (2024), Bank of Israel Annual Report 2023, https://boi.org.il/media/cktbkdrl/full-report_2023n.pdf.

[1] Bank of Israel (2019), Bank of Israel Annual Report 2019, https://www.boi.org.il/en/communication-and-publications/regular-publications/bank-of-israel-annual-report/bank-of-israel-annual-report-2019/.

[41] Bank of Israel (2019), Raising the Standard of Living in Israel by Increasing Labour Productivity, https://www.boi.org.il/20769?culture=he.

[25] Baumol, W. (1992), The Cost Disease, Yale University Press.

[2] Blass, N. (2025), The Education System in Israel 2020–2024: A Conservative System in a Dynamic Reality, https://www.taubcenter.org.il/en/research/education-system-2024/.

[12] Blass, N. and H. Bleikh (2025), The Sources of Budget Gaps in the Official Primary Education System, Taub Center.

[13] Blass, N. and H. Bleikh (2024), Expenditure per Student in High Schools in Israel, Taub Center.

[19] Böhm, K. et al. (2013), “Five types of OECD healthcare systems: Empirical results of a deductive classification”, Health Policy, Vol. 113/3, pp. 258-269, https://doi.org/10.1016/j.healthpol.2013.09.003.

[5] Cahaner, L. and G. Malach (2024), Annual Statistical Report on Ultra-Orthodox (Haredi) Society in Israel, https://www.idi.org.il/haredi/2024/?chapter=57493.

[26] Colombier, C. (2017), “Drivers of Health‐Care Expenditure: What Role Does Baumol’s Cost Disease Play?”, Social Science Quarterly, Vol. 98/5, pp. 1603-1621, https://doi.org/10.1111/ssqu.12384.

[3] European Commission (2023), “The 2024 ageing report”, in The 2024 ageing report: underlying assumptions and projection methodologies., Publications Office, https://data.europa.eu/doi/10.2765/960576.

[34] Finkelstein, A. (2019), A long-term forecast for Israel’s National Insurance system, Bank of Israel, Jerusalem, https://www.boi.org.il/en/communication-and-publications/press-releases/a-long-term-forecast-for-israel-s-national-insurance-system/.

[21] Fries, J. (2002), “Aging, natural death, and the compression of morbidity”, Bulletin of the World Health Organization, Vol. 80/3, pp. 245-250.

[17] Geva, A. (2013), Demographic trends and their consequences on public spending, 2013-2059.

[22] Gruenberg, E. (2005), “The Failures of Success”, The Milbank Quarterly, Vol. 83/4, pp. 779-800, https://doi.org/10.1111/j.1468-0009.2005.00400.x.

[16] Guillemette, Y. (2019), “Recent improvements to the public finance block of the OECD’s long-term global model”, OECD Economics Department Working Papers, No. 1581, OECD Publishing, Paris, https://doi.org/10.1787/4f07fb8d-en.

[32] Guillemette, Y. and D. Turner (2017), “The fiscal projection framework in long-term scenarios”, OECD Economics Department Working Papers, No. 1440, OECD Publishing, Paris, https://doi.org/10.1787/8eddfa18-en.

[37] Haran Rosen, M., T. Ramot-Nyska and Y. Friedmann (2018), Long Term Care in Israel, Bank of Israel, https://www.boi.org.il/en/economic-roles/research-and-publications/all-research/periodic-papers/long-term-care-in-israel/.

[27] Hartwig, J. (2011), “Can Baumol’s model of unbalanced growth contribute to explaining the secular rise in health care expenditure? An alternative test”, Applied Economics, Vol. 43/2, pp. 173-184, https://doi.org/10.1080/00036840802400470.

[43] Helland, E. and A. Tabarrok (2019), “Why Are the Prices So Damn High?”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.3392666.

[40] IMF (2018), “Israel: 2018 Article IV Consultation-Press Release and Staff Report”, IMF Staff Country Reports, Vol. 18/111, p. 1, https://doi.org/10.5089/9781484353301.002.

[6] Israel Democracy Institute (2024), 2024 Statistical Report on Ultra-Orthodox Society.

[33] Johansson, Å. et al. (2013), “Long-Term Growth Scenarios”, OECD Economics Department Working Papers, No. 1000, OECD Publishing, Paris, https://doi.org/10.1787/5k4ddxpr2fmr-en.

[24] Lindgren, B. (2016), The rise in life expectancy, health trends among the elderly, and the demand for health and social care.

[36] Llena-Nozal, A., S. Araki and K. Killmeier (2025), “Needs assessment and eligibility criteria in long-term care: How access is managed across OECD countries”, OECD Health Working Papers, No. 181, OECD Publishing, Paris, https://doi.org/10.1787/461811c4-en.

[15] Lorenzoni, L. et al. (2019), “Health Spending Projections to 2030: New results based on a revised OECD methodology”, OECD Health Working Papers, No. 110, OECD Publishing, Paris, https://doi.org/10.1787/5667f23d-en.

[23] Manton, K. (1982), “Changing concepts of morbidity and mortality in the elderly population”, The Milbank Memorial Fund Quarterly. Health and Society, pp. 183-244.

[30] Ministry of Finance (2024), Approximately NIS 300 million for pension fund savers since the start of the designated bond reform, https://www.gov.il/en/pages/press_19052024.

[31] Ministry of Finance (2016), Increasing certainty in pension savings, https://www.gov.il/he/departments/units/pension_saving_team_committee.

[14] Morgan, D. and M. Mueller (2023), “Understanding international measures of health spending: Age-adjusting expenditure on health”, OECD Health Working Papers, No. 162, OECD Publishing, Paris, https://doi.org/10.1787/043ed664-en.

[35] National Insurance Institute (2023), Annual Report.

[28] OECD (2025), Global Debt Report 2025, OECD Publishing, Paris, https://www.oecd.org/en/publications/global-debt-report-2025_8ee42b13-en.html.

[29] OECD (2025), OECD Economic Outlook, Volume 2025 Issue 1: Tackling Uncertainty, Reviving Growth, OECD Publishing, Paris, https://doi.org/10.1787/83363382-en.

[11] OECD (2025), OECD Economic Surveys: Israel 2025, OECD Publishing, Paris, https://doi.org/10.1787/d6dd02bc-en.

[7] OECD (2024), Education at a Glance 2024: OECD Indicators, OECD Publishing, Paris, https://doi.org/10.1787/c00cad36-en.

[10] OECD (2023), OECD Economic Surveys: Israel 2023, OECD Publishing, Paris, https://doi.org/10.1787/901365a6-en.

[39] OECD (2016), OECD Economic Surveys: Israel 2016, OECD Publishing, Paris, https://doi.org/10.1787/eco_surveys-isr-2016-en.

[44] Poghosyan, T. (2012), “Long-Run and Short-Run Determinants of Sovereign Bond Yields in Advanced Economies”, IMF Working Papers, Vol. 12/271, p. 1, https://doi.org/10.5089/9781475529142.001.

[18] Rosen, B. et al. (2021), “Addressing vaccine hesitancy and access barriers to achieve persistent progress in Israel’s COVID-19 vaccination program”, Israel Journal of Health Policy Research, Vol. 10/1, https://doi.org/10.1186/s13584-021-00481-x.

[42] Sagi, Z. and O. Shafir (2024), Quality Transport for 2040: Planning, Investment, and Benefits.

[4] Shoresh (2020), Shoresh institute/graphs, https://www.shoresh.institute/graphs_heb.html.

[9] Tamir, A. and N. Avraham (2021), Teacher workforce in Israel’s education system: A comparative perspective, Forum Kahalat.

[38] Taub Center (2024), Long Term Care in Israel, https://www.taubcenter.org.il/en/pr/ltc/.

Annex 2.A. Projected higher education enrolment rates

Annex Figure 2.A.1. Bachelor's degree holders by cohort, including projections

Source: OECD elaborations based on the Labour Force Survey.

Source: OECD elaborations based on the Labour Force Survey.

Notes

← 1.  Between 2021 and 2024, the number of students with special needs increased by 13% annually, with particularly significant rises in the number of children with special needs in regular schools (Taub Centre).

← 2.  The youngest age groups (under the age of 24) are not impacted by healthy ageing and are therefore not shifted.

← 3.  To understand this idea intuitively, compare a string quartet performing a 40-minute piece of music today versus 100-years-ago. It takes the same number of people the same amount of time to perform the music today as it did 100-years-ago, but the cost of producing the piece of music would be far higher because the productivity and wages of other sectors in the economy, such as manufacturing, technology, and farming, has gone up significantly. That is, the musicians must be paid significantly more to compensate them for the forgone wages that they could earn today in other sectors (Helland and Tabarrok, 2019[43]).

← 4.  The Baumol effect provides an alternative explanation for the significant increase in Israeli physician wages seen in recent years, beyond the effect of competition between the private and public healthcare providers. Increased productivity in other sectors of the Israeli economy has likely driven up the wages needed to attract and retain health sector workers, increasing overall health spending.

← 5.  The temporary increase in spending in 2021 and 2022 was mainly due to adjustments of public debt to inflation and the depreciation of the Shekel. According to a definition of interest payments based on cash expenditures, they remained stable during this period, amounting to 2.4% of GDP in 2024.

← 6.  According to the latest OECD Economic Outlook for Israel from June 2025.

← 7.  Arad bonds are government bonds linked to the consumer price index for a 15-year term. They pay a semiannual coupon of 2.4% (or 4.86% annually).

← 8.  Between 1965 and 1990, non-tradable bonds for life insurance institutions were issued. These bonds are named “Hetz” (life-indexed) and were issued for policies featuring a guaranteed yield. They are comprised of an indexed fixed interest rate of 4.0%- 6.2%. In the early 1990s, new entries to this channel were ended.

← 9.  Keeping the interest rate constant means that matured debt, which could have a different interest rate, does not affect the interest rate paid on the remaining debt.

← 10.  See a dedicated section on alternative specifications of the fiscal risk premium below.

← 11. The size of the premium is based on an IMF study on the long-run and short-run determinants of sovereign bond yields in advanced economies (Poghosyan, 2012[44]).

← 12.  Over the last few years, recipiency rates for the basic disability and child disability allowance have increased considerably. One reason is the facilitation of the recognition of disabilities, including the child disability allowance.

← 13.  A sensitivity analysis is conducted using the Median population projection of the CBS, which was also used in the NII’s analysis. The results still project depletion in 2037. This is because the demographic divergence between the baseline (new Mixed variant) and the Median variant has only a limited impact on NII-covered social protection spending in 2025-2037, while the spending difference becomes more pronounced once the fund is already depleted (see Figure 1.25, Panel A in Chapter 1).

← 14.  Since the committee’s recommendations were published, the geopolitical environment has undergone some changes, which may have affected defence needs. In practice, the defence budget will continue to be reviewed annually by the authorities.

← 15.  The so-called steady state capital stock was calculated by dividing the investment levels by a constant depreciation rate of 4% and the projected average growth between 2025 and 2065.