This chapter discusses the effectiveness of public systems in protecting older people from poverty risks associated with long-term care. It does so by comparing poverty risks due to out-of-pocket costs among older people with and without long-term care support. The analysis shows that public support reduces poverty risks associated with long-term care costs, but not sufficiently. It also highlights equity challenges in public support, particularly by sex and age. Finally, the analysis shows how the effectiveness of social protection varies across different features of long-term care systems.
Is Care Affordable for Older People?
OECD Health Policy Studies
Abstract
Social protection systems are at the heart of boosting human capital and empowering people. If they are well designed, these systems can contribute to reducing inequalities, building resilience, and ending inter-generational cycles of poverty. Public social spending across OECD countries is high with more than 20% of GDP spent on social services. It is widely acknowledged that everyone should have access to social protection that ensures an adequate standard of living, particularly in case of shocks related to unemployment, sickness, disability, widowhood, or old age. At the same time, recent OECD work has identified that there are gaps in coverage for social benefits (OECD, 2024[1]). Recently, the International Labour Conference (ILC) called on Member States and the International Labour Organization (ILO) to consider long-term care (LTC) as an integral part of social protection systems (ILO, 2024[2]).
The OECD has defined social protection systems as effective when they are adequate, equitable and efficient (OECD, 2018[3]). A system is adequate when those who need LTC can both access and afford it. Public social protection systems should provide coverage for the entire population at risk (to ensure access) and provide sufficient financial support to limit out-of-pocket spending (to ensure affordability). Social protection is equitable when it contributes to reducing the risks of poverty and addressing inequities across socio‑economic groups. Older people who are poor, or vulnerable to poverty, are more likely to need care and least likely to be able to afford it. As such, an assessment of the effectiveness of social protection for LTC must consider the distribution of costs and benefits. Finally, public social protection is efficient when gains in well-being and reductions in poverty and vulnerability are achieved at minimum cost to the public purse.
Previous OECD work highlighted the gaps in social protection for LTC in adequacy, equity, and effectiveness (Cravo Oliveira Hashiguchi and Llena-Nozal, 2020[4]). This chapter builds on previous work by presenting more countries and using a unique methodology to match cases of needs with survey data to generate population-level estimates for social protection (see Annex A). Building on the findings on the adequacy of the LTC systems presented in Chapter 3, this analysis goes one step further by exploring the effect of social protection on poverty in the older populations. To do so, it estimates older people’s poverty levels depending on whether they have LTC needs, how severe these needs are and whether they receive public support or not. Comparing these different poverty levels gives insight in whether the systems in place adequately protect older people from poverty risks associated with LTC costs. The chapter also provides an overview of the extent to which people deplete their wealth to pay for LTC. Finally, estimates of the efficiency of social protection for LTC systems are measured as the relation between public LTC spending and increases in poverty risks on the one hand and between spending and the generosity of public benefits on the other hand.
Without social protection, most people with long-term care (LTC) needs would be in poverty. Across countries, the share of people at risk of poverty due to the out-of-pocket costs of LTC would be between 90 to 100% for individuals with severe needs in 24 countries and subnational regions, were it not for public social protection. In the remaining countries, that share would still be at least 80%.
Public support generally reduces the share of people at risk of poverty due to LTC costs, but not sufficiently in most countries. The share of people at risk of poverty due to high out-of-pocket spending even is 30 percentage points lower across countries for people with severe needs after receiving public support. In 14 countries, the reduction in poverty risk is lower than 10 percentage points while in seven countries, the reduction will be at least 60 percentage points.
Poverty risks for people with LTC needs remain substantially higher than for older people in general. For individuals with severe needs, the share of people at risk of poverty after public support is 50 percentage points higher than for all older people. Only in four countries, the share of people in poverty is the same for those with needs after public support and for all older people. In 11 countries, the share of people at risk of poverty with social protection is at least 70 percentage points higher than for older people on average.
Public support might not adequately protect some vulnerable groups, such as women, and older people. Women and individuals who are at least 80 years old might compound vulnerabilities with a higher probability of more severe needs and lower incomes. For that reason, poverty risks for women and the 80+ tend to be higher than for men and younger groups, even after receiving public support: poverty risks for women are up to 28 percentage points higher than for men and those 80+ have poverty risks that are 27 percentage points higher than those 65‑79 years old.
Poverty reduction is associated with different characteristics of LTC systems across countries. First, countries that spend more on LTC tend to have a greater reduction in the share of people experiencing poverty risks due to the out-of-pockets costs for LTC. Second, countries with in-kind benefits and only income‑testing also fare better in offsetting the poverty risks of LTC.
Well-developed LTC systems generally provide better protection from poverty risks. Countries that perform better across other dimensions of LTC systems such as funding, governance and quality also fare better in poverty reduction. In some countries there are trade‑offs between poverty reduction and access or availability.
In almost all countries, individuals with low incomes would have to use parts of their wealth to pay for care. To cover the remaining out-of-pocket costs that are not paid for by public social protection, individuals with low incomes would need to use their accumulated wealth in all but six countries. In the other countries, the amount of wealth which is used varies: in some countries, individuals would have to use only small parts of their wealth (e.g. in Ireland or Germany) and, in others, their entire wealth (e.g. in Latvia and Czechia).
If there were no public social protection for LTC in old age, the majority of older people would not be able to pay the out-of-pocket costs of care from their incomes alone without being pushed into poverty1 (see Chapter 3). To prevent this, all countries included in this report provide some public support with LTC costs to older people. However, as suggested in the previous chapter, this support is insufficient in many cases. The following sections provide a more in-depth analysis of the effectiveness of public social protection for LTC in old age and the risk of poverty for people with different levels of needs. The chapter uses a methodology to compare countries across levels of needs (Chapter 1, Box 1.1).
LTC needs increase poverty risks despite public support
Without any social protection, most older people with LTC needs would be at risk of poverty. Individuals with severe needs face the highest total LTC costs and therefore are at highest risk of poverty without public support.2 Between 80% and 100% of individuals with severe needs would be at risk of poverty, with Greece and Portugal having the lowest risks due to lower unit costs of care (Figure 4.1, Panel B, “without public support”). For individuals with moderate needs the risk of poverty without any social protection is slightly lower due to lower total costs: between 60% and 100% (Figure 4.1, Panel A). The most significant share of people at risk of poverty due to the use of LTC is in Denmark, the Netherlands and Luxembourg. In these countries care costs are particularly high compared to people’s incomes. More moderate shares of people at risk of poverty due to LTC can be observed in the United States, Greece and Portugal, but the risk remains high.
Differences between countries are large concerning the impact of social protection on the poverty risk. On average, the reduction in the share of people at risk of poverty is 30 percentage points for both severe and moderate needs. In Luxembourg, Denmark and Finland public support leads to a reduction in poverty risks of at least 80 percentage points while in the United States, Greece and Ireland the reduction is zero or nearly zero. In 14 countries, benefits do not seem to have a big impact on the poverty risk caused by LTC needs (less than 10 percentage points). In some of these countries the generosity of the systems is more limited, while in others the benefits may be generous and still insufficient to reduce poverty risks due to the high out-of-pocket costs.
All LTC systems help to reduce the risk of poverty, but do not fully offset the risk caused by LTC needs in most countries. Figure 4.1 includes the baseline poverty level, defined as the poverty level among all individuals who are 65 years and older. Only in Finland, Denmark and Hungary public support fully reduces the poverty risks for older people with severe needs to bring it to the level of individuals in the age group 65+. On average, the share of people at risk of poverty after public support remains 50 percentage points higher than for older people in general in the case of severe needs and 45 percentage points higher for those with moderate needs. In 11 countries, for individuals with severe needs (and in five countries for those with moderate needs) the share of people at risk of poverty is more than 70 percentage points higher than for older people in general.
Share of older people with moderate (Panel A) and severe (Panel B) LTC needs at risk of poverty with and without public support compared to the baseline poverty rate for all older people
Note: Estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Moderate and severe needs correspond to 22.5 and 41.25 hours of care per week, respectively. Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. The poverty level is equal to 50% of country median income.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
The risk of full wealth depletion is overall small but high in some countries
While income‑testing is used widely, wealth-testing is less widely used but many people might still need to use their wealth to pay for care. Eight countries use wealth -testing: they make the receipt or level of public support dependent on a person’s wealth. At the same time, in countries where wealth -testing is not built into the benefit schemes individuals may still have to use their wealth to pay for care if the combination of their income and public support is insufficient to cover the cost of the care they need. Chapter 1 highlighted in most countries, at least 50% of the population may have to rely on their wealth to pay for LTC even after receiving public support. To further complement this picture, the analysis of this section looks to what extent people will need to use all of their wealth or assets.
While, on average, full wealth depletion is not significant for most people across the OECD, it can be very high in some countries. On average, older people with a high income (80th percentile) would have to deplete 8% of their wealth to pay for care, those with a median income 16% of their wealth and those with a low income would deplete 26% of their wealth at the end of an average period of 8.6 years lived with needs3 (Figure 4.2). Individuals with high and median incomes do not have to deplete any wealth in half of the analysed countries as they can pay for their care using their income and public support. However, in the other half of the countries, even individuals with median and high income would have to deplete large parts of their wealth and up to 100% for those living in Czechia and Estonia. Older people with a low income would need to rely much more on their wealth in most countries and use even half of their wealth or more in Czechia, Latvia, Illinois, Italy, Poland, Croatia, Lithuania and Korea. In Czechia and Latvia, they would have fully depleted all their wealth by the end of the 8.6 years period.
For older people with low, median, and high income, assuming covering at least basic cost of living
Note: Low, median, and high incomes refer to the upper boundary of 20th, 50th and 80th percentile of the income distribution among older people, respectively. Estimates computed using the assumption that average duration of low needs is 6.1 years, moderate needs is 1.25 years, and severe needs 1.25 years. The estimates assume older people with LTC needs would seek formal home care. Low, moderate, and severe needs correspond to 6.5, 22.5 and 41.25 hours of care per week, respectively. Detailed descriptions of care recipients’ needs are available in Annex A. The basic cost of living representing food and other expenses is equal to the poverty level (50% of country median income).
Source: OECD analyses based on the Long-Term Care Social Protection questionnaire, the OECD Income Distribution Database, and the OECD Wealth Distribution Database.
Equitable social protection systems for LTC should aim to reduce disparities in the financial risks associated with developing care needs. Public benefits designed to promote equity should be targeted towards older individuals who face greater financial vulnerability, providing them with proportionate support. This is particularly important since people with LTC needs belong predominantly to more vulnerable groups (see Chapter 2). Several studies consistently highlight that individuals with lower incomes not only have a greater chance of requiring LTC compared to their wealthier counterparts but are also more likely to be unable to afford LTC services (Beltz et al., 2022[5]; Kekäläinen, Luchetti and Terracciano, 2022[6]; del Pozo-Rubio et al., 2019[7]). Other factors such as sex and housing situation also play significant roles in the loss of autonomy and increased financial risk. The findings from this section point to the lack of targeted interventions that address the multiple vulnerabilities tied to the poverty risks that arise from facing LTC costs.
Women, the oldest individuals and those living alone face higher risks of poverty
Across all analysed countries, women experience a higher risk of poverty after paying for LTC even with public support, compared to the baseline rate among all older adults. Figure 4.3 illustrates the different impact social protection has on poverty risks for men compared to women who receive care at home. In countries such as Luxembourg, the Netherlands and Denmark, where public support generally covers greater parts of the LTC costs, gender differences in poverty rates are relatively low. Differences are also small in Czechia and Italy, where public support is generally least effective at reducing poverty levels compared to the other countries. The largest differences by gender can be observed in Hungary (28 percentage points), Croatia (17 percentage points), and Spain and Austria (both 15 percentage points). In countries like Hungary the differences can partly be explained by the large income disparities between men and women. Note that, on the other hand, in a handful of countries men are more likely to be at risk of poverty than women, e.g. in Czechia and the Netherlands.
Older women and men with LTC needs, receiving care at home
Note: Estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. The poverty level is equal to 50% of country median income.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
Poverty rates for individuals with LTC needs, even after receiving support, are more pronounced among those of 80 years old or older, compared to those aged between 65 and 79 years. Figure 4.4 shows that this effect is more pronounced in some countries than in others. In Belgium and England, social protection has a relatively uniform effect across age groups. In contrast, in countries such as Ireland and Spain, the oldest age groups remain at a much higher risk of poverty after receiving benefits (27 percentage points higher than the baseline rate). In ten countries, poverty rates for older people with LTC needs after receiving social protection are higher among the younger age group (aged 65 to 79 years) than among the oldest age group (aged 80 years or older).
Older people with LTC needs, receiving care at home
Note: Estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. The poverty level is equal to 50% of country median income.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
In the majority of the analysed countries, older individuals living alone are more prone to poverty caused by the costs of LTC compared to those not living alone, even after accounting for public support (see Figure 4.5). Although this observation aligns with expectations based on the use of household equivalised income to determine poverty levels, the marked disparities suggest additional influencing factors, including sex, geographical location, and age, among others. Some countries demonstrate notable differences in LTC social protection outcomes between older residents living alone and those cohabiting. Especially in Japan, the poverty risk for older people living alone is significantly higher than for those living with someone (21 percentage point difference), followed by France (17 percentage points) and the Slovak Republic (16 percentage points). In contrast, in the other half of the countries older individuals living alone are less prone to poverty after accounting for public support, compared to those living alone. In some countries, like Belgium, Lithuania or the United States (California), there is (almost) no difference in the risk of poverty, regardless of their living situations.
Older people with LTC needs, receiving care at home
Note: Estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. The poverty level is equal to 50% of country median income. Ireland is removed for “Living alone”, as the survey data only contain coupled households.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
The effectiveness of social protection decreases when vulnerability compounds across the different factors such as age, sex, and living conditions. Table 4.1 shows that women over 80 years old who live alone are at the highest risk of poverty when confronted with LTC costs, even with public support. The average risk of poverty among them, across the OECD, is 71%; that is five times higher than the average baseline poverty rate for all older people (65+), which is 14%. Age is a significant risk factor that intensifies vulnerability to poverty after LTC expenses across living status and sex. The differences in the effectiveness of social protection between men and women are accentuated by living status; women living alone, regardless of their age, face significantly higher poverty rates than men of the same age. However, as previous sections have illustrated, the rates after accounting for LTC costs and public support, increase notably, underscoring potential shortcomings in current social protection.
Older people with low, moderate, or severe needs, receiving care at home
Note: Estimates represent the unweighted average of all analysed countries. Estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. The poverty level is equal to 50% of country median income.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
Public social protection is efficient when gains in well-being and reductions in poverty and economic vulnerability are achieved at minimum cost to the public purse. The objective should be to maintain a balance between effectiveness and efficiency. One way to analyse efficiency is to compare the costs of public social protection (the inputs) with poverty reductions (the effects). In the case of LTC this could be, for example, comparing poverty reductions due to public benefits with public LTC spending, providing a proxy measure of efficiency. As mentioned before, this comparison should be interpreted with caution as poverty reductions are estimates that do not consider real-world access and utilisation, and public LTC spending statistics include expenses that go sometimes beyond the components of LTC included in this report.
In countries where public LTC spending is higher, increases in poverty risks associated with LTC costs are generally lower (Figure 4.6). These risks are calculated as the difference between the baseline poverty risks among all individuals who are 65 years and older and the poverty risks faced by older people with moderate needs receiving public support. There is significant variation in public spending with similar outcomes in terms of poverty risks: for instance, Finland, Denmark and Germany spend significantly less than the Netherlands but manage to limit poverty risks to a similar degree. At the other end of the scale, France spends more than e.g. Poland but with a similarly small impact on poverty risks.
Public spending on LTC as a share of GDP and percentage point differences between baseline poverty rate among the older people and poverty rate among the older people with moderate needs using home care after receipt of benefits
Note: Poverty risk estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. Moderate needs correspond to 22.5 hours of care per week, respectively. The poverty level is equal to 50% of country median income.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire, OECD Health Expenditure database and responses to surveys listed in Annex A.
Another way to analyse the efficiency of public social protection is to look at a person’s out-of-pocket costs after receiving public support. The more these costs can be reduced at a minimum investment of public expenditure the more efficient the benefits are. Figure 4.7 shows countries’ public LTC spending as a share of GDP, and the share of the LTC costs that would be covered by public benefits for an older person with moderate needs receiving care at home. The estimates suggest that countries with higher LTC expenditure cover a higher portion of the costs of care faced by older people with moderate needs, on average, than countries with lower LTC expenditure. The relation is similar for low and severe needs (results not shown). The comparison of Figures 4.6 and 4.7 shows that some countries may cover a high share of the LTC costs with public support (high ranking in Figure 4.7), but the remaining out-of-pocket costs still put the person at a significant risk of poverty (lower ranking in Figure 4.6).
Public spending on LTC as a share of GDP and share of total LTC costs that would be covered by public support for older people with moderate needs receiving care at home
Note: Poverty risk estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. Moderate needs correspond to 22.5 hours of care per week, respectively. The poverty level is equal to 50% of country median income.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire, OECD Health Expenditure database and responses to surveys listed in Annex A.
Finally, Figure 4.8 gives insight in the efficiency of countries’ LTC systems by comparing their generosity with the poverty reduction they achieve. Poverty reduction in this context refers to the extent to which a country’s social protection system limits older people’s risk of poverty due to LTC costs and generosity is the share of the total LTC costs that is covered by public support. The comparison shows that countries like Denmark, Finland, Luxembourg and the Netherlands cover almost 100% of an individual’s cost of LTC and thereby prevent additional poverty risks that might be caused by these costs. As such, while they have a high investment, they also have very good outcomes. At the same time, countries like Sweden, Israel or Ireland have quite generous systems as well but are less efficient in keeping older people with LTC needs out of poverty. On the other hand, there are countries like Poland, that only cover around 5% of the costs of care for a person with moderate needs but keep poverty levels low to a similar extent as e.g. Latvia that covers more than 50% of these LTC costs.
Ratio of reduction in poverty risk 1) due to the social protection for a person with moderate needs, and 2) due to the total support as a share of total LTC cost (average across all people with moderate needs)
Note: Poverty risk estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. Moderate needs correspond to 22.5 hours of care per week, respectively. The poverty level is equal to 50% of country median income.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
Some characteristics of LTC systems seem to facilitate, and others to hinder the reduction of poverty for older people who need LTC services. One of these characteristics analysed in this report is the use of means-testing. In most countries, people have to use their income and sometimes wealth to pay for parts of their care costs, as these costs are higher than the benefits, they receive in almost all countries (see Chapter 3). Another characteristic is the way that a benefit is delivered – either in cash or in-kind. Most countries choose to combine both in-kind and cash benefits and only a few rely exclusively on one or the other.
Means-testing has an impact on reducing poverty risks
As there are advantages to both – means-testing and universal – options, many countries choose blending different forms of means-testing with non-means-tested benefits. As described in Chapter 1, countries use means testing to strike a balance between affordability of care for the older people and the sustainability of the LTC systems. The amount of the benefit thus depends on a person’s income and wealth, which affects the out-of-pocket payments people have to contribute themselves. If these out-of-pocket payments become too high, their risk of poverty increases. For home care, countries can be divided in three groups: i) not using any means-testing, ii) using income testing only and iii) using both, income and wealth testing (see Table 4.2 and Table 1.1 in Chapter 1 for the overall average). The table also illustrates the association between the three different types of systems and their effectiveness in reducing poverty for older people requiring LTC.
Although there are likely no silver bullets for designing the best LTC benefits, countries that use income testing only, on average, are more effective in reducing poverty associated with use of LTC. Nevertheless, Table 4.2 shows that, the difference in poverty reduction between countries with no means testing and those with income testing is small. Countries with both income and wealth testing are the least efficient in reducing the poverty. However, all three groups are quite heterogenous. Across all groups, there is at least one country that would not seem to adequately protect older people with LTC needs from the added risk of poverty. This is in line with the findings in the broader literature (Greenstein, 2022[8]; Mkandawire, 2005[9]). Beyond merely combining different types of benefits, it is important to combine them appropriately and ensure their continuity. Means-tests must be well-designed (e.g. with thresholds that target the most vulnerable) and non-means-tested benefits should cover all relevant costs (e.g. help with IADLs).
Note: Poverty risk estimates are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates (see Annex A). Estimates assume all older people with LTC needs would seek formal home care. For countries with subnational models, these are applied to national-level survey data to produce the estimates shown. Low, moderate, and severe needs correspond to 6.5, 22.5 and 41.25 hours of care per week, respectively. The estimates assume all older people with LTC needs would seek formal home care. The poverty level is equal to 50% of country median income.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
The impact of in-kind vs. cash benefits on poverty risks
Countries that offer only in-kind benefits seem to be associated with stronger reduction in poverty risk caused by LTC expenses. Most of the analysed countries combine both in-kind and cash benefits while few rely exclusively on cash benefits or purely on in-kind services. The reduction in poverty is slightly lower in the case of countries that use both in-cash and in-kind benefits. This group of countries is also the biggest and there are significant differences across them. Finally, the poverty reduction is very small or even close to zero in the countries that use only in-cash benefits. However, as in the case of means-testing, it is difficult to establish the direct relationship between the form of support provided and the reduction in poverty risk, as it seems to be more driven by the generosity of LTC system (see Table 4.2 and Table 1.2 in Chapter 1 for the overall average).
The OECD developed a typology to classify LTC systems based on five dimensions and performed a comprehensive clustering analysis to classify countries (Barszczewski and Llena-Nozal, forthcoming[10]). The information and data collection were conducted along the following five comparative dimensions: governance and organisation; funding; quality; availability; and access. The data used in this analysis are taken from the existing OECD indicators of LTC, from previous OECD questionnaires on LTC and from relevant literature. The study includes a total of 30 countries.
The OECD’s typology of LTC systems
The OECD’s typology has advantages in the data collection compared with previous LTC system typologies. The dataset employed in this report is more comprehensive compared to those used in the LTC typology literature. (Kraus et al., 2010[11]) use eight indicators to cluster countries and (Ariaans, Linden and Wendt, 2021[12]) use 12 indicators. In contrast, the dataset used in this report consists of 20 indicators. Moreover, the quality of the constructed typology is affected not only by the number of indicators but also the diversity of dimensions they cover. The indicators used in this report can be grouped into five dimensions: governance, access, funding, availability and quality. While (Kraus et al., 2010[11]) built a dataset covering governance, access, funding and quality, they did not include indicators measuring availability of LTC. On the other hand, (Ariaans, Linden and Wendt, 2021[12]) include indicators covering governance, funding, and access, but lack indicators measuring the quality of provided services as well as availability and support for informal care.
Governance and organisation
Including a dimension on the governance of public LTC provision helps to understand the organisational depth and cohesion of the systems. Four variables are used to classify countries: The first variable measures whether countries have a legislation on LTC that unifies the provision of benefits and services belonging to LTC, as these might span between the responsibilities of the health and social policy ministries. The second variable in the analysis quantifies the level of decentralisation within LTC systems. The third variable captures the degree of public versus non-public provision of LTC. It considers the percentage of providers that are public versus those which are either private for profit or private not for profit. Finally, a fourth variable measures the degree of care integration between the LTC sector and the healthcare sector, particularly through the use of guidelines, care pathways and multidisciplinary teams between the two systems.
Funding
This dimension of the typology aims to capture the degree of public funding and cost-sharing among countries. It is assessed using three variables discussed in previous sections of the report. The first variable measures the share of LTC costs covered by public funds for an older person with severe needs and a median income. The second variable measures the out-of-pocket expenses for LTC of a person with severe needs as a share of a median income. Finally, the third variable measures the poverty reduction thanks to the social protection among older people with severe needs.
Quality
The third dimension uses five variables to measure different aspects of quality. The first variable looks in particular at whether staff-to-resident ratios are in place in the countries. The second variable aims to capture the quality of staffing by identifying whether there are minimum education requirements for LTC workers. The third and fourth variables within this dimension are designed to assess the level of quality assurance and regulation in the LTC sector. The third variable indicates whether there is mandatory accreditation in place for either both institutions and home care services, for only one of them, or for neither. The fourth variable indicates the presence or absence of any quality assurance framework within the LTC system. The fifth variable in the quality dimension looks at outcomes of care.
Availability
The dimension of availability aims to look first at the supply of formal care. This includes a proxy for the supply, the number of beds and the total number of LTC workers (both at home and in institutions) with respect to the number of individuals aged 65 and above using OECD data and country-specific data when it is not available in the OECD database. Secondly, this dimension also looks at reliance and support of family carers by creating a variable on familialism. It looks at the existence of leave schemes for carers and cash benefits for cares and can be coded 0 to 4 for when all four types of support exist (cash benefits for carers, cash benefits for the care recipients which can be used for carers, paid leave and unpaid leave). In addition, another variable looks at supply measured as the reliance on informal carers or more specifically the percentage of care provided by informal carers.
Access
The dimension on access focuses on the extent to which needs are covered by the current system, the degree of targeting and to what extent countries rely on in-kind services or cash benefits. The first and second variable capture the degree of targeted access based on needs and income and compare to what extent out-of-pocket expenditures are high for more disadvantaged groups. The third variable calculates a coverage rate by relating the share of individuals aged 65+ with LTC needs to the share of recipients among the 65+, based on the reporting of needs for help with ADLs and IADLs and, if possible, using the OECD measure of needs based on typical cases. The fourth variable looks at the degree to which a country provides services directly through in-kind services, relies only on cash benefits or provides a choice of both, in kind and cash benefits, which is assigned to the highest value.
More funding and poverty reduction go along with better LTC governance, quality, fair access, and availability
Cluster analysis was performed for countries, resulting in five clusters (see Table 4.3 for allocation of countries to clusters). The objective of the cluster analysis is to divide a dataset into groups (or clusters) such that the data points within each group are more similar to each other than to data points in other groups across the five dimensions described above. The final clustering is the outcome of synthesising several algorithms and has the advantage of being more robust (see Box 4.1). More details on the methodology and results can be found on (Barszczewski and Llena-Nozal, forthcoming[10]). The funding dimension included the degree of poverty reduction and the generosity of the LTC system. Cluster 1 had the highest average scores across all dimensions, while Cluster 5 had the lowest.
Countries that perform better across other dimensions of LTC systems such as governance and quality also fare better in poverty reduction (included in the funding dimension) but across countries there are sometimes trade‑offs between poverty reduction and access or availability. Countries in Cluster 1 excelled in poverty reduction and also tended to score high in the remaining dimensions: governance, availability, access, and quality of LTC. There was also a monotonic decrease in the capability of public LTC systems to reduce poverty from Cluster 1 to Cluster 4. In contrast, some other country groupings showed mixed results across the different dimensions. For example, countries that showed worse outcomes in terms of poverty reduction were in Cluster 5 but they had better results in the dimension of availability including number of beds per 1 000 older people (see Figure 4.9) than those in Cluster 4. The results also indicate that developments across the different dimensions of LTC systems might be uneven, e.g. some countries perform well in terms of poverty reduction but less well in availability and vice versa. Means-testing which is used to capture the access dimension is most common among countries in Cluster 3, which have limited financial resources and attempt to increase efficiency by restricting access based on means.
Note: Cluster allocation based on syntheses results of four algorithms.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
Country clustering using the OECD methodology
Note: The OECD methodology is described in Box 4.1. Unweighted average of three measures of LTC systems: the difference between the risk of poverty among older people with severe needs before and after receiving the social support; the number of beds per 1 000 people 65+; the number of typical cases in which a person with low income (the upper bound of the 20th percentile) receives higher support than person with high income (the upper bound of the 80th percentile), ranges from 0 to 3.
Source: OECD analysis based on the Long-Term Care Social Protection questionnaire and responses to surveys listed in Annex A.
The clustering procedure uses the Principal Component Analysis (PCA) method due to the high number of variables. The PCA is used to analyse data with a high number of dimensions and increases the interpretability of the data while preserving the maximum amount of variation in the data. In a second step, each algorithm is executed to obtain country classifications. Since none of the implemented clustering methods has shown a clear advantage over the others, the final clustering results are derived from a synthesis of the outcomes from separate algorithms. Every country is finally assigned to the cluster it was assigned to in most cases.
Four clustering algorithms are used in this report. They belong to two broad categories of clustering methods: distance‑based clustering and probabilistic clustering.
Distance‑based clustering groups data points into clusters based on the similarity between them, which is calculated using a distance metric. The fundamental idea is to place data points that are close to each other in the same cluster, reflecting the proximity in the feature space. This report considers the following three type of distance‑based clustering:
K-means is one of the most widely used algorithms. It assigns data points to clusters such that the distance between a data point and cluster centroid is smallest. Initial centroids of a specified number of clusters are randomly selected. While the advantage is its simplicity, the algorithm is sensitive to the initial random selection of cluster centroids. Besides, finding the optimal number of clusters lacks a general theoretical solution.
Hierarchical clustering groups similar data points into clusters that form a hierarchical structure, reflecting the order in which clusters are merged or divided. Unlike k-means algorithms, it does not require to specify the number of clusters beforehand and is not sensitive to the initial selection of cluster centroids. Yet, it can be sensitive to outlier observations and is influenced by the choice of distance metric and linkage method.
Self-Organizing Map (SOM) is a type of artificial neural network designed to reduce the dimensionality of data while preserving the topological relationships between data points (Kohonen, 1982[13]). The algorithm is robust to noise and outliers in the data but can be sensitive to the initial configuration of neurons. Additionally, its outcome depends on the selection of parameters such as the grid size, learning rate, and neighbourhood size.
Probabilistic clustering is a clustering approach that assigns data points to clusters using probabilistic models or probability distributions. These methods assign probabilities to indicate the likelihood of data points belonging to each cluster. This approach is particularly useful when data points can potentially belong to multiple clusters or when there is uncertainty in cluster assignments.
This report uses the Gaussian Maximization Method (GMM) as a representative of probabilistic clustering. The underlying assumption of this method is that data points are generated from a mix of several Gaussian distributions, each corresponding to one cluster. GMM employs an Expectation-Maximization technique to estimate parameters (the mean and the covariance) for each of these distributions. It allows for the calculation of the probability of belonging to each cluster. This makes GMM a more flexible approach compared to k-means. However, it also lacks a general theoretical framework for determining the optimal number of clusters and is sensitive to initial parameter guesses.
Prior to constructing clusters, it is essential to specify a number of clusters, as most clustering algorithms require this information beforehand. There is typically no universal theoretical solution to determine the optimal number of clusters. However, the literature offers various rules of thumb to assist in selecting the appropriate number of clusters, such as the elbow method for the k-means family of algorithms. Considering these rules, as well as stability of cluster assignments across different methods, the number of clusters is set to three.
References
[12] Ariaans, M., P. Linden and C. Wendt (2021), “Worlds of long-term care: A typology of OECD countries”, Health Policy, pp. 609–617, https://doi.org/10.1016/j.healthpol.2021.02.009.
[10] Barszczewski, J. and A. Llena-Nozal (forthcoming), “A typology of long-term care systems across OECD countries”, OECD Publishing, Paris.
[5] Beltz, S. et al. (2022), “Multivariate analysis of independent determinants of ADL/IADL and quality of life in the elderly”, BMC Geriatrics, Vol. 22/1, https://doi.org/10.1186/s12877-022-03621-3.
[4] Cravo Oliveira Hashiguchi, T. and A. Llena-Nozal (2020), “The effectiveness of social protection for long-term care in old age”, OECD Health Working Papers, No. 117, OECD, Paris, https://doi.org/10.1787/2592f06e-en.
[7] del Pozo-Rubio, R. et al. (2019), “Catastrophic long-term care expenditure: associated socio-demographic and economic factors”, The European Journal of Health Economics, Vol. 20/5, pp. 691-701, https://doi.org/10.1007/s10198-019-01031-8.
[8] Greenstein, R. (2022), Targeting vs. Universalism, and Other Factors That Affect Social Programs’ Political Strength and Durability, The Brookings Institution.
[2] ILO (2024), Social Protection Spotlight, http://www.social-protection.org/gimi/Media.action;jsessionid=Pqnuco7kNnqK0XWICi_e_UHjBe4i5M2yStqSsXNvhx5FGrtmf-eL!-511164124?id=19818.
[6] Kekäläinen, T., M. Luchetti and A. Terracciano (2022), Functional Capacity and Difficulties in Activities of Daily Living From a Cross-National Perspective, https://doi.org/10.1177/08982643221128929.
[13] Kohonen, T. (1982), “Self-organized formation of topologically correct feature maps”, Biological Cybernetics, Vol. 43/1, pp. 59-69, https://doi.org/10.1007/bf00337288.
[11] Kraus, M. et al. (2010), A typology of long-term care systems in Europe, ENEPRI, https://www.researchgate.net/publication/277226669_A_Typology_of_Long-Term_Care_Systems_in_Europe_ENEPRI_Research_Report_No_91_August_2010.
[9] Mkandawire, T. (2005), Targeting and Universalism in Poverty Reduction, United Nations Research Institute for Social Development.
[1] OECD (2024), Modernising Access to Social Protection: Strategies, Technologies and Data Advances in OECD Countries, OECD Publishing, Paris, https://doi.org/10.1787/af31746d-en.
[3] OECD (2018), Social Protection System Review: A Toolkit, OECD Development Policy Tools, OECD Publishing, Paris, https://doi.org/10.1787/9789264310070-en.
← 1. Relative income poverty: Disposable net income after paying out-of-pocket costs below 50% of the population wide median equivalised income.
← 2. It should be noted that individuals with severe needs tend to be older and more likely to have less means left to pay for care while individuals closer to 65 years old are less likely to face poverty risks.
← 3. The analysis is based on two assumptions: 1) the average time an older person has to spend with LTC needs is composed of 6.1 years with low LTC needs, 1.25 years with moderate needs, and 1.25 years with severe needs, and 2) each older person has median wealth when they develop LTC needs independent of their income. It then compares the share of that median wealth that people have deplete to pay for LTC during this period of 8.6 years, depending on whether they have a low, median or high income.