This release is categorised as a CSO Frontier Series Output. Particular care must be taken when interpreting the statistics in this release as it may use new methods which are under development and/or data sources which may be incomplete, for example new administrative data sources. For further information on the data sources, linking procedures, and limitations of this release, see the Background Notes section.
This release is based on linking administrative income data with those who indicated they were living in private households in Census 2022. When calculating household income, estimates for some households are underestimated, as not all income sources are included, such as undeclared employment income, inter-household transfers, and some foreign pensions. Throughout this release, the term administrative income will be used to describe the income sources covered.
Equivalised income allows for a more meaningful comparison of income across households by accounting for the number of adults and children living in the household, thus allowing for analysis at an individualised level. The term low-income threshold refers to households with an equivalised net administrative income of below 60% of the national median equivalised net income as derived from administrative data sources. In this release, the low-income threshold is deemed to be €16,271.
This release focuses on children aged under 18, living in private households with one or both parents.
The results presented in this release are based on the Geographical Profiles of Income in Ireland 2022 (GPII) dataset, which was a data-linking exercise of two pseudonymised Central Statistics Office data sources:
With additional insights included by data-linking with two additional pseudonymised data sources:
Two releases have previously been produced from this matched dataset:
The linkage and analysis was undertaken by the Central Statistics Office (CSO) for statistical purposes in line with the Statistics Act, 1993 and the CSO Data Protocol.
Before using personal administrative data for statistical purposes, the CSO removes all identifying personal information. This includes the Personal Public Service Number (PPSN), a unique number used by people in Ireland to access social welfare benefits, personal taxation, and other public services. A pseudonymised Protected Identifier Key (PIK) is created by the CSO when the PPSN is removed. This PIK is unique and non-identifiable and is only used by the CSO.
Using the PIK enables the CSO to link and analyse data for statistical purposes, while protecting the security and confidentiality of the individual data. All records in the matched datasets are pseudonymised and the results are in the form of statistical aggregates which do not identify any individuals. For further information see Administrative Data FAQ.
The PIR is a pseudonymised income register held internally within the CSO. It contains information on income received by individuals relating to employment, self-employment, and social transfers. It is derived from administrative holdings held by Revenue and Department of Social Protection. Therefore, the PIR provides a near complete picture on individual level income, for a calendar year.
The COPA is a pseudonymised copy of the Census of Population 2022 dataset held internally within the CSO for analysis purposes. It contains Census attribute information for individuals and households of which 95% of records have a PIK which allows them to be linked to pseudonymised administrative data sources to create new analysis.
Student Universal Support Ireland contains funding information for all higher and further education grants. SUSI offers funding to eligible students in approved full-time, third-level education in Ireland and also, in some cases, funding for students studying outside the State. It offers support to all types of students, from school leavers to mature students returning to education.
The ITForm11 contains the annual income tax returns of the self-employed. Data for a calendar year is only complete three years after the reference year, because of the nature of self-assessment, although the majority of records are available about 14 months after the reference year. The ITForm11 was linked to rental income.
The CSO’s IPEADS 2022 features demographic estimates that are based on administrative records. It is important to note that such administrative records are designed for the operational needs of Irish public bodies and not as statistical data sources.
All income data referenced in the release spans the period from 01 January 2022 to 31 December 2022.
The geographic and demographic variables were collected on Census reference day, 03 April 2022.
A private household comprises either one person living alone or a group of people (not necessarily related) living at the same address with common housekeeping arrangements - that is, sharing at least one meal a day or sharing a living room or sitting room. In order to be included in the household, a person had to be a usual resident at the time of the Census. Therefore, visitors to the household on Census Night were excluded.
Usual residents who were enumerated outside of their place of usual residence, were as far as possible moved back to their place of usual residence. Enumerated persons that could not be matched to a place of usual residence were excluded. Persons usually resident in communal establishments are out of scope. Persons absent abroad on Census night are excluded, as are households that were fully absent on Census night. Private households of individuals where the place of work was identified as outside of the Republic of Ireland were excluded from any analysis.
From an initial population of 5,084,879 persons usually resident in Ireland and present on Census night, a total of 161,519 persons were excluded from analysis in the Income Analysis using Administrative Data Sources (IAADS) 2022 release, resulting in a target population for the IAADS of 4,923,360 persons.
For the Characteristics of Children Living in Low-Income Households (CCLLIH) release, households containing persons with an unknown relationship to one another are excluded from analysis. In the IAADS file, 71,966 persons were enumerated somewhere other than their place of usual residence and moved back to their usual residence. The relationships of these persons with other persons in these households is unknown, as such the 61,905 households they were in are excluded from the CCLLIH for the purpose of this report. Also excluded, were households not matched to administrative income and households without a parent present. Focusing on children aged under 18, living in private households with one or both parents, results in a target population for the CCLLIH of 1,071,475 children.
The COPA2022 dataset records where persons were enumerated on Census night. This may not have been their place of usual residence. For the purpose of producing household income and equivalised income statistics, persons not enumerated in their place of usual residence were moved back to their home as far as possible. It was not possible to re-locate all such persons. A combination of the Eircode PIK of the usual residence provided on the Census form and the Irish Population Estimates from Administrative Data Sources, 2022 dataset were used to return persons enumerated elsewhere to their place of usual residence. Those that could not be linked to a place of usual residence were excluded from analysis. Households fully occupied by persons that indicate their usual residence is elsewhere, are thus dropped and the residents moved to their place of usual residence where possible.
Persons who indicated that they worked outside of Ireland and had employment income on Irish administrative data totalling less than €10,000 were excluded from analysis, along with their household, as the household income records were deemed incomplete.
Persons not present in the state and enumerated on Census night are excluded from analysis.
These exclusions led to the removal of 51,684 households from the GPII/IAADS dataset. The largest concentration of these was in border areas. Particular care should be taken with border area estimates, where 27 of the 30 electoral divisions with the most dropped households were in Donegal, Monaghan, or Cavan. Also, after removing these households, many of these households were not matched to administrative income sources. Areas with high student populations are also liable to having a lower percentage of households matched with administrative income.
For further information see Tables 5.1 and 5.2 of the Geographical Profiles of Income in Ireland 2022 - Household Income - Background Notes.
For the Characteristics of Children Living in Low-Income Households (CCLLIH) release, households containing persons with an unknown relationship to one another are excluded from analysis. The removal of such households, as well as households not matched to administrative income data, and restriction to the Census definition of family units with children, resulted in the removal of 77,839 children leaving 1,071,475 children aged under 18 in the final CCLLIH file. The impact of removing these children was that the proportion of children aged under 18 under the low-income threshold of €16,271 on the IAADS file was 16.8%, while on the CCLLIH it was 16.1%.
There are records on the COPA2022 dataset which do not have a PIK and thus cannot be linked with administrative income. For the target population of the IADDS, the PIK coverage for persons aged 15+ was 95.2%.
At the household level, 4.8% of households in the IADDS dataset were not matched to administrative income. The reasons for not being matched include: records which do not have a PIK on the COPA2022 and thus cannot be linked to administrative income; households with income sources not covered by the IADDS gross income definition; or households subsisting on savings or interhousehold transfers such as students.
There were records in the PIR that could not be linked to COPA2022. The reasons for not being recorded on the COPA2022 include: absent persons identified in Census; persons living abroad but working or paid for work in Ireland; and persons who arrived in Ireland after or left Ireland before the 03 April 2022. This "missing" gross income came to a total €20.4 billion for 2022. This figure also includes income related to records which do not have a PIK on the COPA2022 dataset.
The annual gross income before deductions such as tax and social insurance which was measured in nominal terms and includes:
Excluded from the income measure are:
Income outputs are produced at household level. Households were identified using information gathered in Census 2022.
Income tax, Universal Social Charge (USC) and social insurance contributions were estimated and summed to household level and subtracted from the household gross income to calculate the annual household net income. The components of household net income are household gross income less:
Total annual payments from the Department of Social Protection (unless otherwise stated), categorised into the following:
Individuals and households who were identified as living in Ireland in the COPA but could not be linked to administrative income data in the PIR in 2022.
Households with no administrative income are excluded from estimates of means and median household income. The only tabulations in this release where they are included are in the distribution of household income graph and associated tables.
For the calculation of mean and median household income, households with no administrative income are excluded.
Mean: Sum of all household income, divided by the number of households with administrative income. e.g. The mean of the following four household incomes {€0, €18,000, €45,000, €70,000} is, after dropping the household with no administrative income, (18,000+45,000+70,000)/3 = €44,333.
Median: Midpoint of the ordered values, after removing households with no administrative income. e.g. The following four household incomes {€0, €18,000, €45,000, €70,000}, after dropping the household with no administrative income, becomes {€18,000, €45,000, €70,000}. The median of which is €45,000.
Equivalence scales are used to calculate the equivalised household size in a household. Although there are numerous scales, we focus on the national scale in this release. The national scale attributes a weight of 1 to the first adult, 0.66 to each subsequent adult (aged 14+ living in the household) and 0.33 to each child aged less than 14. The weights for each household are then summed to calculate the equivalised household size.
Net household administrative income is divided by the equivalised household size to calculate equivalised net administrative income for each person, which essentially is an approximate measure of how much of the income can be attributed to each member of the household. This equivalised income is then applied to each member of the household.
Persons with an equivalised net administrative income below 60% of the national median equivalised net administrative income are deemed to be below the low administrative income threshold.
To identify the primary structural determinants of child poverty, this study constructs a binary logistic regression model with the below/above (1/0) low-income threshold status as the dependent outcome variable.
To optimise the model development, theoretical and logical filters were manually applied to the available variable pool. Income variables were excluded from serving as independent variables as the dependent variable is directly derived from income. Granular geographic variables such as Local Electoral Area (LEA) regions were also excluded as the specific name or border of a geographic area does not inherently exert a causal influence on economic strain. Instead, spatial risk is driven by structural attributes such as labour market accessibility and localised socioeconomic disadvantage. Instead, the six-way urban/rural classification was included to capture employment proximity, alongside the Pobal HP Deprivation Index to capture concentrated neighbourhood-level deprivation.
The following variables were considered for inclusion in the logistic regression analysis:
From this pool of potential independent variables, the selection process began with univariate screening of the variables where each candidate predictor was run in isolation, fitted separately in a logistic regression model against the binary outcome to evaluate its baseline Akaike Information Criterion (AIC) score, establishing an empirical hierarchy of relative model fit.
Following this, an automated forward stepwise selection procedure was initialised from an empty baseline model, sequentially adding variables that yielded the steepest reductions in AIC score. Next, the model entered a recursive diagnostic loop to assess multicollinearity and identify potentially confounding and redundant variables. The Generalized Variance Inflation Factor (GVIF) was calculated for all parameters to check for multicollinearity. Variables contributing to multicollinearity were identified and systematically removed. This resulted in the removal of the variable indicating whether there is anyone in the household with a long-lasting condition or difficulty.
Following the reduction of multicollinearity, a final check was performed on the forward stepwise path to determine the optimal stopping point. While automated algorithms accepted minor double-digit AIC drops at the tail end of the selection matrix, a manual inspection revealed a distinct "elbow" or wall of diminishing returns. This step resulted in the removal of the following variables from the model: whether there is a parent in the household with a long-lasting condition or difficulty; Pobal HP deprivation index; sex of the child.
The final resulting logistic regression model presented in this release identifies associations between structural, household, and economic characteristics and the odds of a child falling below the low-income threshold across the population. Because this analysis is based on census-level data, the reported odds ratios (ORs) describe model-based associations in the observed population rather than estimates derived from a probability sample.
As this analysis leverages exhaustive, census-level administrative records rather than a probability sample, traditional inferential validation metrics such as p-values, standard errors, and confidence intervals are conceptually inappropriate due to the absence of sampling error. Instead, the structural soundness, reliability, and validity of the final logistic regression model were verified using three non-parametric diagnostic criteria: multicollinearity stability, probability calibration, and predictive discrimination.
Multicollinearity was assessed using Generalized Variance Inflation Factors (GVIF). This metric accounts for the multi-category nature of the categorical variables by calculating an adjusted index.
All adjusted GVIF values squared remained below 2.0, far beneath the standard conservative critical threshold of 5.0. This confirms that the major structural blocks of the model, e.g. parental education, occupational classifications, and NACE industry sectors, provide unique, non-redundant statistical information, ensuring the mathematical integrity of the reported population Odds Ratios.
Model calibration was evaluated to check whether the estimated logistic regression probabilities accurately reflect real-world rates. The population was partitioned into risk deciles based on their predicted probability. The mean predicted probability was mapped against the observed proportion of children below the low-income threshold in each decile.
The resulting model calibration curve demonstrated very close alignment with the ideal 45-degree reference line across the entire risk spectrum. The final model showed good agreement between predicted probabilities and observed rates across the risk spectrum, showing no systematic overestimation or underestimation across the population. In the highest-risk decile, the model successfully isolated a sub-population with an average predicted risk of 68.7%, closely matching the true observed population rate of 67.3% children below the low-income threshold.
The AUC-ROC score (Area Under the Receiver Operating Characteristic curve) measures the model's discrimination power. While the calibration curve showed that the model's probabilities are accurate, the AUC score shows how effectively those probabilities separate the two groups: children who are below the low-income threshold versus children who are not.
The model achieved an AUC score of 0.887 in the observed population, showing the model has very strong discrimination power. This means there is an 88.7% probability that a randomly selected child living below the low-income threshold will be assigned a higher risk score by the model than a randomly selected child living above the threshold.
The GPII 2022, and by extension the IAADS 2022 and CCLLIH 2022, are not directly comparable to the 2016 GPII release due to methodological changes. The primary change is that the 2016 GPII was based on where persons were enumerated on Census night, whereas the 2022 GPII persons enumerated outside of their usual residence were returned to their place of usual residence as much as possible. Also the 2022 GPII includes income from HAP, which the 2016 edition did not.
SILC is the official source of household income and at risk of poverty statistics in Ireland. However, SILC is a sample survey is a sample survey and cannot go into the level of geographic detail that the IAADS can. The calculation of household income is not comparable between the two as the methodologies differ significantly. The IAADS is limited to the use of administrative income in the calculation of gross, while SILC includes additional components such as:
Data on ethnicity was taken from Census 2022. The question on ethnic group or background was updated for Census 2022 with new response categories added including Roma, Indian/Pakistani/Bangladeshi and Arab.
Question 11 on Census 2022 persons form asked “What is your ethnic group/background?”, with response options:
For further information on Census 2022 questions see Census 2022 Household Form Questions.
For household level income analysis by ethnicity, the ethnicity refers to that of the head of household.
Data on self-perceived general health was taken from Census 2022. People were asked to rate their general health on a five-point scale from very good to very bad.
Question 17 on Census 2022 asked “How is your health in general?”:
Data on long-lasting condition or disability was taken from the 2022 census. This was derived from answers to Questions 15 and 16 of the census questionnaire. Question 15 was a seven-part question that asked about the existence of the following long lasting conditions: (a) blindness or a vision impairment, (b) deafness or a hearing impairment, (c) a difficulty with basic physical activities such as walking, climbing stairs, reaching, lifting or carrying, (d) an intellectual disability, (e) a difficulty with learning, remembering or concentrating, (f) a psychological or emotional condition or a mental health issue and (g) a difficulty with pain, breathing or any other chronic illness or condition. Respondents had the option to indicate that they did not have any of these long-lasting conditions, had one or more of them to some extent, or had one or more of them to a great extent.
Question 16 was a four-part question that asked whether an individual had a difficulty doing any of the following activities: (a) dressing, bathing or getting around inside the home (self-care disability), (b) going outside the home alone to shop or visit a doctor’s surgery (going outside the home disability), (c) working at a job or business or attending school or college (employment disability) and (d) participating in other activities, such as leisure or using transport. As with Question 15, respondents could indicate that they had no difficulty with any of these activities, that they had some difficulty or that they had a lot of difficulty. Question 16 on day-to-day difficulties specifically included difficulties caused by old age.
For the release of the results, one category was created based on the response options in Q15 and Q16: Long-lasting condition or difficulty experienced to any extent.
Respondents who ticked any of the 'Yes' boxes in Q15 or Q16 were also included in the 'disability to any extent' rate. For further information see Census of Population 2022 Profile 4 - Disability, Health and Carers Background Notes.
The Census definition of a family unit or nucleus is defined as:
Family members have to be usual residents of the relevant household. The determination of household and family composition is based on responses to the question on the census form dealing with relationships within the household.
This release focuses on four types of family unit that contain children under 18 years of age:
The working status of a parent is derived from the presence of employment income from administrative income data during the 2022 income reference year.
The Occupation classifications used in the Census are based on the UK Standard Occupational Classification (SOC), with modifications to reflect Irish labour market conditions. The primary version of SOC used in Census 2022 is SOC2010 and was first used in Census 2011.
The code to which a person’s occupation is classified is determined by the kind of work he or she performs in earning a living, irrespective of the place in which, or the purpose for which, it is performed. The nature of the industry, business or service in which the person is working has no bearing upon the classification of the occupation. For example, the occupation clerk covers clerks employed in manufacturing industries, commerce, banking, insurance, public administration, professions and other services etc.
The term industry used for Census of Population purposes is not confined to manufacturing industry. It is synonymous with the term 'sector of economic activity'. The basis of the industrial classification is, in the case of employees, the business or profession of their employer and in the case of self-employed persons, the nature of their own business or profession.
In Census 2022, industry is coded using NACE – the General Industrial Classification of Economic Activities within the European Communities. The current version, NACE Rev. 2, is a 4-digit activity classification that was first used in Census 2011 and is a revision of the version originally published by Eurostat in 1970.
The industry in which a person is engaged is determined (regardless of their occupation) by the main economic activity carried out in the local unit in which he or she works. If, however, the local unit provides an ancillary service to another unit in the business (e.g. administration, storage, etc.) then the persons in the ancillary unit are classified to the industry of the unit it services. Thus, while the occupational classification is concerned only with the particular work performed by an individual regardless of the activity carried on at the local unit, the industrial classification is concerned only with the ultimate purpose of the unit or end product regardless of the precise nature of the work performed by each individual.
A manufacturing or commercial unit may employ persons with many different occupations for the purpose of making a particular product or for giving a particular service. Conversely, there are cases in which particular occupations are largely confined to a single industry. For example, the majority of persons with agricultural occupations are in the agriculture industry and most miners are in the mining industry.
In Census 2022, persons aged 15 years and over were asked two questions. The first question asked if a person had ceased their full-time education. If they had answered Yes, they were then asked the age at which it ceased. The second question asked what the highest level of education completed (full-time or part-time) was. The levels ranged from no formal education to a tertiary doctorate degree.
The Pobal HP Deprivation Index uses data from Census 2022, analysing ten measures of an area’s levels of disadvantage. These include educational attainment, employment status and the numbers living in individual households. Almost 19,000 small areas (50-200 households) were indexed leading to the development of a detailed map of the relative affluence and disadvantage. For further information, see Pobal HP Deprivation Index.
Administrative Counties correspond with the Local Authorities, covering 31 city and county councils. For example, Cork County refers to the region administered by Cork County Council.
In census reports, the country is divided into 26 counties/administrative counties and the five cities. Outside Dublin, there are 23 administrative counties and four cities: Cork, Limerick, Waterford, and Galway. In Dublin, the four local authority areas are identified separately: Dublin City and the three administrative counties of Dún Laoghaire-Rathdown, Fingal and South Dublin. For further information, see CSO Geographical classifications - national level.
For the purposes of Local Authority elections, each county and city is divided into Local Electoral Areas (LEAs) which are constituted on the basis of Orders made under the Local Government Act, 1941. Statutory Instruments 610-638 of 2018 and 6-8, 27-28, 156-157 of 2019 state the current composition of LEAs and their subsequent amalgamation into Municipal Districts.
In general, LEAs are formed by aggregating Electoral Divisions. However, in a number of cases, Electoral Divisions are split between LEAs and in order to render them suitable for production of statistics, the CSO has amended some LEA boundaries to ensure that statistical disclosure does not occur. As a result of these amendments, Census 2022 LEAs are comprised of whole Census 2022 Electoral Divisions. For further information see Census 2022 Small Area Population Statistics.
Learn about our data and confidentiality safeguards, and the steps we take to produce statistics that can be trusted by all.