Is Care Affordable for Older People?

Annex A. Data sources and methodology

Stylised cases of needs to enable cross-country comparisons

There is no internationally agreed-upon needs assessment for older people requiring long-term care (LTC). To facilitate meaningful national and international comparisons of the level of financial protection provided by LTC systems, a set of typical cases of LTC needs was developed. These typical cases address the challenge of comparing generosity and affordability using administrative data, as differences in eligibility, coverage and generosity can confound such comparisons. This approach allows for the determination of the level of public support in different countries for a defined level of LTC needs. The term “public support” is used throughout the report and refers to the public social protection – in-kind or in cash – that individuals receive to help them cover the cost of the LTC they need. Differences between countries and regions can then be identified, analysed and form the basis for policy recommendations.

The typical cases describe several fictional older persons in terms of the types and severity of their LTC needs, and the professional services they would require (Muir, 2017[1]). The cases are based on the number of hours needed for assistance with ADLs, IADLs, and social activities (see Oliveira Hashiguchi and Llena-Nozal (2020[2]) for detailed descriptions). The typical cases are categorised into three different levels of severity of needs (low, moderate, and severe) and different ways in which these needs can be met (professional home care, informal home care, institutional care and combinations). A total of eight cases are presented in Table A.1. Some countries offer support for individuals with severe needs that encompass 24‑hour care, a more generous scenario than the severe cases modelled in this exercise. The OECD has collaborated with countries and subnational areas to map their needs assessments to the different typical cases. This process involved providing detailed descriptions of the abilities and limitations of the person in question, the services they require, and any other relevant assumptions.

Table A.1. Typical cases of long-term care needs defined for this report

Typical case	Needs	How needs are met
Typical case 1	Low	6.5 hours of professional home care per week
Typical case 2	Moderate	22.5 hours of professional home care per week
Typical case 4a	Moderate	22.5 hours of informal care (spouse) per week
Typical case 4b	Moderate	22.5 hours of mixed professional/informal care (spouse) per week
Typical case 4c	Moderate	22.5 hours of informal care (adult child) per week
Typical case 4d	Moderate	22.5 hours of mixed professional/informal care (adult child) per week
Typical case 3	Severe	41.25 hours of professional home care per week
Typical case 5	Severe	41.25 hours of institutional care

Microdata used in this report

The developed models of the social protection systems in each country codify the rules that determine the levels and eligibility for social protection for LTC. These models are used to determine the cost of LTC, the generosity of public social protection systems, and the effects of public social protection on people’s out-of-pocket costs and poverty among LTC recipients. An important assumption is that people seek care from the modelled social protection system. The models can also estimate total government spending at different levels of coverage. Analyses can focus on any subgroup of interest (e.g. women, economically vulnerable, asset-poor), depending on the availability of survey responses. The analysis covers OECD countries, including Austria (Vienna), Belgium (Flanders), Canada (Ontario), Croatia, Czechia, Denmark, Estonia (Tallinn), Finland, France, Germany, Greece, Hungary, Iceland (Reykjavik), Ireland, Israel, Italy (South Tyrol), Japan, Korea, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, the Slovak Republic, Slovenia, Spain, Sweden, the United Kingdom (England) and the United States (California and Illinois).

The OECD uses information on the actual distributions of income and wealth from the OECD Income Distribution Database (OECD, 2024[3]) and the OECD Wealth Distribution Database (OECD, 2024[4]) to determine the cost, public support, and out-of-pocket expenses associated with LTC. To assess the impact of LTC systems on poverty reduction, the report uses data from surveys on ageing and retirement.

A survey-based methodology is used with acknowledgement of its limitations like self-reporting and other technical challenges (e.g. sampling issues). The selection of datasets for each country focuses on the inclusion of ADLs and IADLs, alongside other important factors such as income, net worth (the sum of fixed and liquid net assets), and household structure, which are crucial for evaluating LTC needs. Table A.2 presents details on the datasets employed in this report.

In addition to determining poverty reduction due to social protection for LTC, microdata described in Table A.2 are used to estimate the prevalence and intensity of LTC needs in populations aged 65 years and older, stratified by socio-demographic and economic factors. Different approaches to match survey responses to the typical cases are tested, relying on the intensity of reported difficulties, followed by a comparison of the outcomes. Various validity, sensitivity, and uncertainty analyses are conducted to enhance confidence in the robustness of the results. Concurrently, the OECD collected information from representatives in the OECD countries (or subnational areas, where applicable) on the LTC needs assessments and instruments in use. This provides additional context to evaluate the representativeness of the typical cases and the appropriateness of different matching methods.

Table A.2. Sources of microdata used in this report

Dataset	Countries	Wave	Year	Source	Comment
Survey of Health, Ageing and Retirement in Europe (SHARE)1	Austria, Belgium, Croatia (non-OECD), Czechia, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Israel, Italy, Latvia, Lithuania, Luxembourg, Malta (non-OECD), Netherlands, Poland, Portugal, Slovak Republic, Slovenia, Spain, Sweden	8	2019 or 2020	SHARE‑ERIC (2024[5]). Survey of Health, Ageing and Retirement in Europe (SHARE) Wave 8. Release version: 9.0.0. SHARE‑ERIC. Data set
Survey of Health, Ageing and Retirement in Europe (SHARE)	Portugal	6	2015	SHARE‑ERIC (2024[6]). Survey of Health, Ageing and Retirement in Europe (SHARE) Wave 6. Release version: 9.0.0. SHARE‑ERIC. Data set.	The latest available wave for Portugal
Longitudinal Study on Ageing (TILDA)	Ireland	3	2014 or 2015	ISSDA (2019[7]), The Irish Longitudinal study on Aging (TILDA) Wave 3, 2014‑15. Version 3.3.	Do not sample single households and people living in the institutions
Japanese Aging and Health Dynamics Study (JAHEAD)	Japan	9	2017	Tokyo Metropolitan Institute of Gerontology, Institute of Gerontology (University of Tokyo) and University of Michigan (2024[8]), Japanese Aging and Health Dynamics Study (JAHEAD), Wave 9, 2017.	Do not sample people living the institutions
Korean Longitudinal Study of Aging (KLoSA)	Korea	7	2018	Korean Employment Information Service (2024[9]), Korean Longitudinal Study of Aging (KLoSA), Wave 7, 2006‑18, Version D.2.	Do not sample people living the institutions
English Longitudinal Study of Ageing (ELSA)	United Kingdom	9	2018 or 2019	Banks et al. (2024[10]) English Longitudinal Study of Ageing: Waves 0‑10, 1998‑2023. [data collection]. 40th Edition. UK Data Service. SN: 5 050
Health and Retirement Study (HRS)2	United States of America	14	2018	Health and Retirement Study (2024[11]), RAND HRS Products public use dataset. Produced and distributed by the University of Michigan with funding from the National Institute on Aging (grant number NIA U01AG009 740). Ann Arbor, MI..

1. See Börsch-Supan et al. (2013[12]) for methodological details. The SHARE data collection has been funded by the European Commission through FP5 (QLK6‑CT‑2001‑00360), FP6 (SHARE‑I3: RII-CT‑2006‑062193, COMPARE: CIT5‑CT‑2005‑028857, SHARELIFE: CIT4‑CT‑2006‑028812), FP7 (SHARE‑PREP: GA N°211909, SHARE‑LEAP: GA N°227822, SHARE M4: GA N°261982) and Horizon 2020 (SHARE‑DEV3: GA N°676536, SERISS: GA N°654221) and by DG Employment, Social Affairs & Inclusion. Additional funding from the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the U.S. National Institute on Aging (U01_AG09740‑13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1‑AG‑4553‑01, IAG_BSR06‑11, OGHA_04‑064, HHSN271201300071C) and from various national funding sources is gratefully acknowledged (see www.share-project.org).

2. The HRS (Health and Retirement Study) is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and is conducted by the University of Michigan.

Note: This report uses data or information from the Harmonised SHARE, ELSA, HRS, and KLoSA datasets and Codebook Version of June 2023 developed by the Gateway to Global Aging Data. The development of the harmonised datasets was funded by the National Institute on Aging (R01 AG030153, RC2 AG036619, 1R03AG043052). For more information, please refer to www.g2aging.org.

Matching methods to assign typical cases

To align typical cases with data on self-reported ADL and IADL limitations, the OECD uses a weighted average across two methodologies.1 Table A.3 lists the specific limitations considered for identifying LTC needs, chosen for their comparability across various countries and datasets. This includes five ADL limitations and four IADL limitations, with the second limitation (walking across the room and getting in/out of bed) aggregating two separate ones to facilitate cross-country comparisons. Identification of a limitation is based on self-reported information, disregarding severity levels to avoid subjective bias.

Table A.3. Limitations used in the matching methods

Type	Description
ADLs	1. Dressing
	2. Walking across the room & getting in/out of bed
	3. Bathing/Showering
	4. Eating
	5. Using the toilet
IADLs	6. Preparing a hot meal
	7. Shopping groceries
	8. Telephone calls
	9. Managing money

Note: Korean data (KLoSA) contains only information on getting in/out of bed, the information on walking across the room is missing. Japanese data (JAHEAD) is missing two IADLs which are substituted with “Difficulties in doing small chores such as dusting off furniture, throwing away rubbish”.

Source: Limitations as described in SHARE (Wave 8).

The two methods determine the level and intensity of needs based on the same set of reported limitations in performing ADLs and IADLs but apply different thresholds to determine severity. Table A.4 describes the criteria that each method follows to assign individuals to one of three levels of need: low, moderate, and severe. Both methods rely solely on difficulty scores, where a higher score indicates more limitations than a lower one. The definition of the thresholds between low, moderate, and severe needs influences the distribution of older people across these categories.

Table A.4. Description of matching approaches

Matching methods	Typical cases of LTC needs
	Low needs	Moderate needs	Severe needs
Matching method X	Diff. Score: 1‑2	Diff. Score: 3‑6	Diff. Score: 7‑10
Matching method Y	Diff. Score: 1‑2	Diff. Score: 3‑7	Diff. Score: 8‑10

Note: Each responded is assigned only to one level of severity or has no needs.

In eight countries (Austria, Belgium, Canada, Estonia, Iceland, Italy, the United Kingdom, and the United States), LTC costs and benefits are decentralised, and data availability at the national level is limited. Consequently, their models have been set up at the municipal, regional, and state levels. Determining how representative these models for subnational areas are of the national public LTC systems is challenging. One approach to approximate representativeness is to compare key indicators such as total population, population aged 65 years and older, and GDP, as illustrated in Table A.5. These simple comparisons, while not exhaustive, provide an initial perspective on the extent of data and information gaps at the national level.

Table A.5. Overview of key indicators in subnational areas

Country	Subnational area	Share of country’s population	Share of country’s 65+ population	Share of country’s GDP per capita
Austria	Vienna	34%	34%	106%
Belgium	Flandres	58%	61%	101%
Canada	Ontario	39%	38%	99%
Estonia	Tallinn	29%	26%	144%
Iceland	Reykjavik	40%	N/A	N/A
Italy	South Tyrol	1%	1%	161%
United Kingdom	England	85%	84%	97%
United States	California	12%	11%	117%
United States	Illinois	4%	4%	109%

Note: All data is for 2018.

Source: OECD Regional Database and OECD Metropolitan Database.

For all population-level estimates, it is assumed that the subnational LTC benefits and schemes apply nationally. All chapters in this report, including the figures and tables, use only the country names for better readability,2 e.g. “Estonia” rather than “Estonia (Tallinn)”. For better readability, this information is not repeated in the chapters.

Assumptions for the estimation of public support and out-of-pocket costs of long-term care

Survey data – like those used in this report – are not exhaustive, contain missing responses, and have limited information on important indicators. This is the case, for example, with data on access to and effective use of public support for formal care. There is also limited information on preferences for formal and informal care. Due to these and other data gaps, assumptions are made to assess the effects of public social protection for LTC in old age in OECD countries.

A key assumption is that older people with LTC needs use formal LTC services and support from public social protection systems. Based on this assumption, it is possible to estimate the public support they would be entitled to, the out-of-pocket costs they would face, their net disposable income after paying for the LTC costs, and consequently, whether they are at risk of poverty.

The graphs and estimates in this report should thus not be seen as representing the current situation in countries today. For example, the poverty rates associated with LTC shown in Chapter 4 are not current poverty rates, as many older people with LTC needs do not receive formal care in reality. Rather, the estimates presented in this report are intended to illustrate what different indicators would look like if everyone who is estimated to have LTC needs sought formal care.

Determining current coverage of public social protection for LTC

Determining the current availability and coverage of public social protection for LTC in old age in OECD countries is challenging due to limited detailed data on the existence, geographical location, and capacity of public social protection offices to handle support requests (e.g. conduct needs assessments). Even with assumed full availability of public social protection for LTC, effective use also hinges on the availability and distribution of providers, which is difficult to establish across OECD countries. Moreover, older people with LTC needs may opt not to use public sector services, further complicating the assessment of coverage.

Notwithstanding these challenges, estimating current coverage of public social protection for LTC in old age is essential for evaluating different policy scenarios in OECD countries (see Chapter 5). This can be achieved using self-reported survey data and/or administrative data from OECD countries. Since administrative data are scarce, this report estimates coverage using survey data only. For EU countries, coverage estimates were produced using EHIS data, based on the Eurostat definition of needs and OECD typical cases, then averaged. For non-EU countries and Greece, coverage estimates use data from surveys listed in Table A.2. Coverage is defined as the percentage of older people (aged 65+) with any level of needs who report receiving formal LTC services (help with ADLs, IADLs, meals on wheels, other types of care). For countries where survey data do not cover people living in institutional settings, the coverage was adjusted using administrative information on the split between institutional and home care.

Estimates for older people with severe LTC needs receiving formal care, either in institutions or at home, are primarily based on data from the Ageing Report (2021), supplemented by survey data listed in Table A.2 and OECD Health Statistics 2023 for non-EU countries. Initial assessments excluded the OECD Questionnaire on Social Protection for Long-Term Care in Old Age, survey data listed in Table A.2, and OECD Health Statistics 2023 as primary data sources due to limitations in data points, sampling, and comparability. This report follows Ageing Report (2021) and defines coverage of institutional care among those with severe needs as the ratio of recipients of institutional care to the total number receiving care both in institutions and at home.

The estimated coverage of formal LTC and the split between institutional and home care varies significantly across countries (see Table A.6). The average OECD coverage is 27%, with the lowest estimates in Estonia, the Slovak Republic, Lithuania and Latvia (below 15%), and the highest in Israel, France and Belgium (over 50%). The share of people receiving care in institutions among all formal care recipients is, on average, 36%, with the lowest values in Greece and Israel (below 10%) and the highest in Portugal and Croatia, where more than 60% of formal care recipients receive institutional care.

Older people with LTC needs may have differential coverage of public support for formal care due to varying preferences for formal versus informal care, and public versus private provision. On average, across OECD countries, around 27% of older people with needs receive formal care services. However, which older people access public support, and which do not? In the absence of empirical evidence, an assumption must be made to estimate coverage for specific individuals. Broadly, two possible ways to account for differential coverage of public formal care services include: 1) Assuming that older people with more severe needs and lower means (income and net wealth) have higher probabilities of receiving formal care than those with lower needs. 2) Assuming the probability of receiving formal care is the same for all older people with LTC needs.

This report adopts the first option: the probability of receiving formal care among older people with LTC needs is higher for those with more severe needs and lower means. This assumption is viewed as a best-case scenario where those older people who most need public social protection are the ones who are first in line to take it. While this best-case scenario is unlikely, a random allocation of older people with needs is also not realistic.

A similar issue arises when it comes to the division of patients between institutional and home care. As institutional care tends to be more intensive, this report assumes that only people with severe needs receive institutional care. Additionally, people with lower income and wealth are more likely to receive institutional care than those with higher income and wealth. The logic of this assumption aligns with the coverage assumption: it is a best-case scenario where care is received by those who most need public social protection.

Table A.6. Estimated coverage of formal and institutional care among older people with long-term care needs

Country	Share of older people with LTC needs receiving any formal care	Share of older people receiving care in the institution among older people with LTC needs receiving formal care
Austria	31%	43%
Belgium	53%	20%
Croatia	25%	65%
Czechia	15%	54%
Denmark	49%	23%
England	23%	38%
Estonia	11%	31%
Finland	28%	12%
France	51%	47%
Germany	37%	53%
Greece	30%	4%
Hungary	21%	54%
Ireland	36%	27%
Israel	64%	7%
Italy	28%	47%
Japan	28%	19%
Korea	26%	28%
Latvia	15%	44%
Lithuania	13%	53%
Luxembourg	31%	38%
Malta	22%	34%
Netherlands	44%	22%
Poland	18%	44%
Portugal	16%	66%
Slovak Republic	12%	52%
Slovenia	16%	50%
Spain	31%	25%
Sweden	23%	20%
United States	29%	23%
OECD	27%	36%

Note: Estimates for OECD typical cases are first calculated using adjusted survey weights separately for each matching method (X and Y) and then averaged to obtain the final estimates. The table includes an unweighted OECD average.

Source: OECD analysis based on EHIS (Wave 2, 2014; Wave 3, 2019), data survey listed in Table A.2, OECD Health Statistics, and Ageing Report (2021).

Adjustments to income variables in survey data

Comparing means of deciles of equivalised couple income in the survey data with the equivalent household-level metric from the OECD Income Distribution Database (IDD) reveals consistently lower incomes (82% of deciles) reported or imputed in survey data than in the IDD (see Table A.7). While using couple‑level income would expectedly showcase these differences, the observed discrepancies cannot be fully attributed to the different unit levels. Previous research into income measurement errors suggests multiple drivers for differences across surveys and between surveys and administrative sources, leading to both overreporting and underreporting of income (Angel et al., 2019[13]; Moore, Stinson and Welniak, 2000[14]). Moore, Stinson and Welniak (2000[14]) note that survey respondents typically underreport their incomes, remarking that the “consistency with which the survey estimates fall below the benchmarks is striking.”

In Table A.7, it is noteworthy that three outliers exist within the first decile of the income distribution for the Slovak Republic, Korea and Italy. These figures indicate that the mean of the lowest decile of household equivalent income in the OECD IDD is significantly higher than those reported in survey data for these specific countries. The particularly high ratios observed for Italy and Korea arise from an abnormal surge in zero income responses in the survey, compared to the previous two waves. This issue is addressed by imputing some of the zero income responses using a two‑part model.

Firstly, the imputation of couple income involves using total household income3 and other socio-economic characteristics such as age, gender, number of children, education level, self-reported health, and country fixed effect, following a probit regression specification. Secondly, a variable containing the conditional mean for each observation is predicted. Thirdly, the position of each individual in the income distribution generated by the reported income variable and the predicted value is evaluated. If an individual’s position in the predicted income distribution is relatively similar to their position in the reported income within a maximum deviation of three deciles, and if the individual reports zero income in reported income while reporting non-zero income in total household income, the predicted value is assigned as the final income.

Table A.7. Ratio of mean household income in the survey data and in the OECD IDD, per income decile

A value over 1 indicates mean incomes in the OECD IDD are higher than in the SHARE/TILDA

Country	Ratio of income deciles among people aged 65 years and older
	1st	2nd	3th	4th	5th	6th	7th	8th	9th	10th
Austria	1.60	1.28	1.24	1.25	1.24	1.24	1.22	1.20	1.18	1.16
Belgium	1.62	1.03	1.02	0.99	0.96	0.92	0.90	0.86	0.80	0.22
Croatia	2.76	1.63	1.50	1.48	1.54	1.53	1.52	1.51	1.48	1.34
Czechia	2.00	1.29	1.23	1.20	1.18	1.16	1.16	1.18	1.24	1.20
Denmark	1.83	1.34	1.30	1.28	1.25	1.19	1.09	1.03	0.99	1.16
England	0.84	0.91	0.91	0.90	0.91	0.91	0.91	0.90	0.90	0.94
Estonia	1.40	1.15	1.13	1.18	1.27	1.27	1.30	1.45	1.50	1.46
Finland	1.36	1.06	0.94	0.93	0.93	0.93	0.91	0.90	0.92	0.84
France	1.95	1.22	1.15	1.11	1.07	1.04	1.02	0.98	1.01	1.24
Germany	1.42	1.14	1.09	1.08	1.08	1.08	1.07	1.05	1.02	1.03
Greece	2.94	1.28	1.24	1.24	1.25	1.24	1.26	1.27	1.27	1.35
Hungary	1.05	1.05	1.11	1.21	1.19	1.23	1.26	1.27	1.33	1.58
Ireland	2.60	2.03	2.28	2.19	1.69	1.97	1.81	1.92	1.75	1.63
Israel	1.27	1.11	1.13	1.08	1.11	1.12	1.09	1.04	1.02	0.89
Italy	1 286.63	7.13	2.12	1.78	1.68	1.66	1.60	1.59	1.63	1.78
Japan	1.03	NA	0.94	NA	0.92	NA	1.01	NA	1.11	1.07
Korea	NA	65.16	7.27	4.20	3.11	2.42	1.98	1.76	1.59	1.37
Latvia	1.51	1.17	1.20	1.30	1.39	1.42	1.50	1.66	1.90	2.14
Lithuania	3.81	1.38	1.34	1.39	1.41	1.40	1.47	1.54	1.63	1.92
Luxembourg	4.50	1.96	1.36	1.26	1.25	1.21	1.18	1.12	0.93	0.47
Malta	7.88	1.99	1.86	1.94	2.04	2.08	2.14	2.12	2.07	1.24
Netherlands	2.06	1.31	1.29	1.25	1.16	1.04	0.96	0.87	0.66	0.48
Poland	1.33	1.27	1.29	1.32	1.35	1.35	1.35	1.39	1.40	1.47
Portugal	7.36	1.85	1.49	1.35	1.33	1.31	1.31	1.15	1.13	1.31
Slovak Republic	15.49	2.77	1.55	1.43	1.42	1.41	1.47	1.54	1.55	1.49
Slovenia	3.61	1.74	1.63	1.54	1.48	1.48	1.47	1.49	1.51	1.64
Spain	5.71	1.69	1.48	1.54	1.61	1.58	1.58	1.58	1.55	1.55
Sweden	1.64	1.18	1.09	1.06	1.03	1.01	0.98	0.96	0.94	1.17
United States	0.92	1.04	0.99	0.94	0.89	0.84	0.80	0.75	0.65	0.44

Note: Data for Japan is missing because income in the survey data is defined in intervals, resulting in gaps in some deciles. Data for Korea for the first decile is missing due to the high number of zero incomes in the dataset.

Source: OECD Income Distribution Database and survey data listed in Table A.2.

Due to lower reported incomes in survey data compared to the OECD IDD, median income in the OECD IDD is typically higher, as seen in Table A.8. The exception is the United States, where survey data show higher median income than the OECD IDD. Lower reported incomes across all deciles lead to higher calculated poverty rates using equivalised household incomes in survey data, which influences the analysis of poverty risks associated with LTC services. High baseline rates may diminish the perceived effect of public social protection systems on poverty risks.

To adjust for underreporting and ensure comparability with OECD IDD poverty risks, a stepwise approach is employed. First, deciles of household income in the survey data are determined for each country. Second, means and upper bounds are calculated for each decile. Third, ratios between survey data and OECD IDD means (or upper bounds) are calculated. Fourth, original incomes in survey data are multiplied by these ratios for each decile, creating adjusted income variables, including those adjusted by ratios of means and upper bounds. The process is repeated with winsorised incomes (at a 5% level), resulting in four new income variables: two for non-winsorised and two for winsorised incomes.

The differences between the poverty rates in OECD IDD and those in survey data, as well as those derived from the four new adjusted income variables, are then calculated. These differences are used to select the income manipulation that minimises discrepancies with OECD IDD poverty rates. Table A.8 shows that adjusted poverty rates are closer to those in OECD IDD. All analyses of public support, out-of-pocket costs, net incomes, and poverty risk in this report are based on the income variables that minimise these differences (see the last column of Table A.8).

Table A.8. Comparisons of median incomes and poverty rates in OECD IDD and survey data

Country	Median in IDD	Median in survey data	Poverty rate IDD	Poverty rate in survey data	Poverty rate in adjusted survey data
Austria	29 343	22 200	10%	19%	11%
Belgium	28 436	21 940	10%	12%	11%
Croatia	9 263	4 457	25%	53%	25%
Czechia	309 590	197 707	8%	24%	8%
Denmark	259 354	156 337	4%	28%	4%
England	20 044	18 219	15%	11%	16%
Estonia	13 829	6 585	34%	55%	32%
Finland	27 484	24 056	7%	12%	9%
France	23 840	22 350	4%	13%	5%
Germany	25 707	20 905	11%	16%	12%
Greece	9 954	7 637	7%	17%	6%
Hungary	2 358 790	1 667 358	13%	17%	14%
Ireland	25 219	11 180	6%	63%	7%
Israel	90 862	81 635	19%	25%	20%
Italy	19 252	10 993	11%	42%	11%
Japan	2 480 000	2 100 000	20%	20%	20%
Korea	28 750 000	5 939 697	43%	72%	37%
Latvia	10 113	4 582	34%	59%	33%
Lithuania	9 597	4 446	29%	53%	28%
Luxembourg	42 486	40 102	5%	20%	6%
Malta	15 381	8 866	13%	34%	10%
Netherlands	28 300	19 800	6%	25%	5%
Poland	35 732	22 766	13%	27%	12%
Portugal	9 930	6 790	11%	32%	10%
Slovak Republic	10 439	5 940	7%	37%	6%
Slovenia	16 591	9 348	13%	40%	15%
Spain	18 090	10 889	12%	35%	12%
Sweden	306 018	233 404	11%	21%	12%
United States	42 800	44 710	23%	23%	23%

Note: Median incomes are in NCUs. The poverty rate is set at 50% of the median income. Poverty rates are for 2019. Statistics from survey data are calculated using adjusted survey weights.

Source: OECD analysis based on OECD Income Distribution Database and survey data listed in Table A.2.

Adjusting weights for observations in institutional settings in the SHARE

While the SHARE sampling frame in most countries (except the Slovak Republic) includes persons living in nursing homes and residential care facilities, people living in institutions have a lower probability of being included in the survey (Bergmann, Scherpenzeel and Börsch-Supan, 2019[15]). According to De Luca, Celdoni and Trevisan (2015[16]), Belgium, Czechia and Italy excluded people living in retirement and nursing homes from their population registers for sampling purposes. The heterogeneity in the sampling frames used across countries may also lead to country-specific differences in representation (Bergmann, Scherpenzeel and Börsch-Supan, 2019[15]).

In this report, the underrepresentation of people living in residential care and nursing homes in the SHARE sample is adjusted ex-post, for the countries where it is possible, using adjusted sample weights. Based on the methodology used by Barczyk and Kredler (2019[17]), and in collaboration with researchers from the Center for Economic and Social Research at the University of Southern California, individual-level weights were constructed to align the shares of nursing home residents in the SHARE with the shares in OECD Health Statistics 2023 (OECD, 2023[18]) for the relevant years. Table A.9 shows the results of survey weights adjustment.

Table A.9. Share of population age 65+ living in a nursing home, by sample weights

Country	Using SHARE weights	Using adjusted weights	Number of respondents interviewed in nursing homes
Austria	3.9%	7.8%	40
Belgium	4.5%	5.6%	54
Croatia	1.1%	2.3%	5
Czechia	2.1%	4.3%	49
Denmark	2.5%	3.4%	32
Estonia	1.4%	4.5%	30
Finland	0.8%	4.3%	6
France	2.4%	4.1%	46
Germany	2.1%	4.2%	26
Greece	0.2%	0.3%	2
Hungary	0.2%	3.0%	2
Israel	1.9%	1.6%	13
Italy	0.4%	0.8%	4
Latvia	0.5%	0.4%	2
Lithuania	0.5%	12.6%	4
Luxembourg	4.8%	5.4%	15
Malta	3.8%	7.6%	15
Netherlands	1.9%	4.2%	21
Poland	0.2%	0.8%	2
Portugal	2.0%	1.3%	19
Slovak Republic	0.0%	0.0%	0
Slovenia	1.9%	4.6%	32
Spain	1.8%	2.2%	29
Sweden	1.6%	4.2%	30

Source: OECD analyses based on the surveys listed in Table A.2.

References

[13] Angel, S. et al. (2019), “What did you Really Earn Last Year?: Explaining Measurement Error in Survey Income Data”, Journal of the Royal Statistical Society Series A: Statistics in Society, Vol. 182/4, pp. 1411-1437, https://doi.org/10.1111/rssa.12463.

[10] Banks, J. et al. (2024), English Longitudinal Study of Ageing: Waves 0-10, 1998-2023. [data collection]. 40th Edition. UK Data Service. SN: 5050, https://doi.org/10.5255/UKDA-SN-5050-27.

[17] Barczyk, D. and M. Kredler (2019), “Long‐Term Care Across Europe and the United States: The Role of Informal and Formal Care”, Fiscal Studies, Vol. 40/3, pp. 329-373, https://doi.org/10.1111/1475-5890.12200.

[15] Bergmann, M., A. Scherpenzeel and A. Börsch-Supan (2019), “SHARE Wave 7 Methodology: Panel innovations and life histories”, Munich: Munich Center for the Economics of Aging (MEA), https://doi.org/10.6103/SHARE.w7.900.

[12] Börsch-Supan, A. et al. (2013), “Data Resource Profile: The Survey of Health, Ageing and Retirement in Europe (SHARE)”, International Journal of Epidemiology, pp. 992-1001.

[19] European Commission (2022), Proposal for a COUNCIL RECOMMENDATION on access to affordable high-quality long-term care, https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=COM%3A2022%3A441%3AFIN (accessed on 27 October 2023).

[11] Health and Retirement Study (2024), RAND HRS Products public use dataset. Produced and distributed by the University of Michigan with funding from the National Institute on Aging (grant number NIA U01AG009740). Ann Arbor, MI., https://hrsdata.isr.umich.edu/data-products/rand?_gl=1*1iljav1*_ga*Mzk4NjA5NDMzLjE3Mjc4ODAwNTE.*_ga_FF28MW3MW2*MTcyNzg4MDA1MS4xLjEuMTcyNzg4MDkyMC4wLjAuMA..

[7] ISSDA (2019), The Irish Longitudinal study on Ageing (TILDA) Wave 3, 2014-2015. Version 3.3., http://www.ucd.ie/issda/data/tilda/wave3.

[22] ISSDA (2019), TILDA Derived Variables Codebook: Wave 3, Irish Social Science Data Archive, https://www.ucd.ie/issda/t4media/0053-04_TILDA_Wave3_Derived_Variables_Codebook_v3.4.pdf.

[9] Korean Employment Information Service (KEIS) (2024), Korean Longitudinal Study of Aging (KLoSA), Wave 7, 2006-2018, Version D.2., https://survey.keis.or.kr/eng/klosa/klosa01.jsp.

[16] Malter, F. and A. Börsch-Supan (eds.) (2015), Item nonresponse and imputation strategies in SHARE Wave 5, MEA, Max Planck Institute for Social Law and Social Policy, https://share-eric.eu/fileadmin/user_upload/Methodology_Volumes/Method_vol5_31March2015.pdf.

[14] Moore, J., L. Stinson and E. Welniak (2000), “Income Measurement Error in Surveys: A Review”, Journal of Official Statistics, Vol. 16/4, pp. 331-361, https://www.census.gov/content/dam/Census/library/working-papers/1997/adrm/sm97-05.pdf.

[1] Muir, T. (2017), “Measuring social protection for long-term care”, OECD Health Working Papers, No. 93, OECD Publishing, Paris, https://doi.org/10.1787/a411500a-en.

[21] Neri, A. and R. Zizza (2010), “Income Reporting Behaviour in Sample Surveys”, SSRN Electronic Journal, https://doi.org/10.2139/ssrn.1792512.

[3] OECD (2024), Income distribution database, https://data-explorer.oecd.org/vis?df[ds]=DisseminateFinalDMZ&df[id]=DSD_WISE_IDD%40DF_IDD&df[ag]=OECD.WISE.INE&dq=.A.INC_DISP_GINI..._T.METH2012.D_CUR.&pd=2010%2C&to[TIME_PERIOD]=false (accessed on 24 July 2024).

[4] OECD (2024), Wealth Distribution Database, https://www.oecd.org/en/data/datasets/income-and-wealth-distribution-database.html (accessed on 14 July 2024).

[18] OECD (2023), OECD Health Statistics 2023, https://www.oecd.org/en/data/datasets/oecd-health-statistics.html (accessed on 24 July 2024).

[23] OECD/Eurostat/WHO (2017), A System of Health Accounts 2011: Revised edition, OECD Publishing, Paris, https://doi.org/10.1787/9789264270985-en.

[2] Oliveira Hashiguchi, T. and A. Llena-Nozal (2020), “The effectiveness of social protection for long-term care in old age: Is social protection reducing the risk of poverty associated with care needs?”, OECD Health Working Papers, No. 117, OECD Publishing, Paris, https://doi.org/10.1787/2592f06e-en.

[6] SHARE-ERIC (2024), Survey of Health, Ageing and Retirement in Europe (SHARE) Wave 6. Release version: 9.0.0. SHARE-ERIC. Data set., https://doi.org/10.6103/SHARE.w6.900.

[5] SHARE-ERIC (2024), Survey of Health, Ageing and Retirement in Europe (SHARE) Wave 8. Release version: 9.0.0. SHARE-ERIC. Data set., https://doi.org/10.6103/SHARE.w8.900.

[8] Tokyo Metropolitan Institute of Gerontology; Institute of Gerontology, The University of Tokyo; University of Michigan (2024), Japanese Aging and Health Dynamics Study (JAHEAD), https://www2.tmig.or.jp/jahead/index.html.

[20] UNECE (2011), Canberra Group Handbook on Household Income Statistics, Second Edition, http://www.unece.org/fileadmin/DAM/stats/publications/Canberra_Group_Handbook_2nd_edition.pdf (accessed on 17 June 2019).

Notes

← 1.  Initially a broader set of matching methods was undertaken and, after an assessment of their performance, two were retained.

← 2.  Except of the subnational models for the United States.

← 3.  There is no household income data for Korea, so Korean data are not used in this procedure.