4. Administrative data sources for sequences of active labour market policies and public works programmes evaluation
This chapter presents an overview of the data that have been compiled for use across the analysis of sequences of active labour market policies and the evaluations of Tús and Community Employment, two public works programmes in Ireland. It outlines all the different datasets that are used in the analysis and presents information on what information they contain. The chapter then goes on to report descriptive statistics for the Tús and Community Employment schemes, to contextualise the counterfactual evaluation that is conducted later in the report. It concludes by reviewing how the data for this evaluation have been compiled and what improvements may facilitate further data analysis.
This report makes use of rich linked administrative data that enable the construction of a detailed picture on jobseekers’ individual characteristics, their labour market history and what kind of support they have received from the Department of Social Protection (DSP). More specifically, DSP data on unemployment, broader benefit receipt and detailed participation data on Tús and Community Employment (CE) are merged to Office of the Revenue Commissioners (Revenue) data on earnings and weeks of employment.
The data allow the report to compare the histories of individuals prior to unemployment and look at employment-related (and broader) outcomes subsequent to it. Individuals’ journeys over time are constructed, so that different sequences of employment, unemployment and associated labour market services for individuals can be observed. The data also enable the report to consider what impact participation in both Tús and CE have on individuals’ subsequent journeys through the labour market.
Substantial efforts were required by DSP to extract and compile the data necessary for this evaluation, as a previously collated internal dataset – the Jobseekers Longitudinal Dataset – was no longer available. These efforts have laid the groundwork for a possible re-implementation of a core analytical database, which would facilitate timely and recurrent counterfactual policy analyses. Other OECD countries, such as Canada, have adopted this kind of approach in order to reduce the costs of evaluation and to ensure reliable and repeatable analyses are easy to implement (OECD, 2022[1]). More details on the administrative data used and the methodology and issues in compiling them are provided in the accompanying technical report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[2]).
This chapter documents the data that underpin the report, to provide insight on how they support all of the analysis conducted. The chapter starts by describing the administrative data that are used for the different evaluations conducted in this report. Later sections provide some more detailed statistics for each of the separate analytical topics of the report. The final section reviews how the administrative data were compiled for this report and potential innovations that would further enhance future data analysis.
This section provides information on the different data that are used to conduct the analysis and presents descriptive statistics on the participants of both CE and Tús schemes. It starts by detailing the different sources for the data in 4.2.2, and then provides an explanation of the different cohorts of participants that are chosen for each of the analytical sections in the report.
4.2.1. A Data Protection Impact Assessment (DPIA) ensured data security was at the heart of the project
In preparing the data for transfer and use in the analysis, the DSP completed a data protection impact assessment (DPIA). The assessment detailed the reasons for which the data was required, how it would be transferred and stored and persons requiring access. The DPIA ensured that the pseudonymisation, data minimisation and simplification steps proposed were sufficient to minimise and mitigate any risks identified in transferring the data.
The project model was considered in the context of seven privacy principles which are set out by Article 5 of the GDPR. These principles incorporate the data protection and privacy requirements within EU and Irish legislation. The DPIA identified any risks to data subject privacy resulting from carrying out the evaluation and outlined mitigation and security procedures taken by the DSP to ensure secure processing of the data. Aggregation and simplification steps include grouping location, age, and other data to less granular levels, such as dates of birth being reduced to only the month and year of birth. The Local Office code, while known and used only within DSP, was replaced with a pseudonymised string. A pseudonymised ID was provided to link data.
4.2.2. Combinations of DSP register data and Revenue data build rich compiled datasets
The administrative data used for the report are sourced from the DSP (Table 4.1). The population of individuals for which information is drawn is all individuals (primarily long-term unemployed people) who were eligible for CE and Tús between 2011 and 2018. The data include information on the Tús and CE schemes, providing a rich picture of participants’ activities on the schemes and the categorisation of the placement as “activation” or “social inclusion” in the case of CE. In addition, a payments dataset captures the weekly social welfare payments made to individuals over the relevant period. Data on the receipt of benefits provide details of the type of benefit and claim duration (i.e. how long a benefit was paid) of both participants and eligible non-participants, as well as information on several other programmes that individuals participated in. All these data are underpinned by demographic information on age, sex and nationality. In addition, the Probability of Exit (PEX) dataset contains the responses to the PEX questionnaire and the associated PEX value and date which it was recorded for a subset of the individuals in the data.
DSP also holds information on social insurance contributions and earnings, collected by Revenue. These large data on claims of benefits, payments and earnings are typical examples of register data, whereas PEX is closer to a survey dataset, albeit a survey where the target is the population of newly registered unemployed people, rather than a sample.
One additional dataset, the Jobseekers Longitudinal Dataset (JLD) was also used to construct condensed summaries of individuals’ labour market histories. Although no longer updated, it is an example of a dataset explicitly designed for analytical purposes, constructing episodic longitudinal views from weekly benefits and employment data sources.
The analytical teams of the DSP, the Joint Research Centre of the European Commission (JRC) and the OECD worked together to define the scope of analysis and outline the data specification required. DSP then applied concerted resources to compile, quality assure and integrate disparate administrative data sources for this report. The descriptions of the data sources below leverage these pre-compiled datasets rather than referring to the original underlying data. That is, the descriptions abstract from any cleaning and compilation that has occurred within DSP. Additional information on the original data can be found in the accompanying technical report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[2]). The technical report also presents further descriptive statistics on the number of observations and individuals and the underlying characteristics of each of the compiled dataset.
CE scheme data provide detailed insights into CE programmes and participants
The CE scheme dataset held by DSP provides information on jobseeker placements on CE and details of CE programmes. Each observation in this dataset relates to one placement on CE, i.e. individuals may have multiple records if they have participated in more than one CE scheme. The dataset includes information on the start and end date for the placement in question, the job description (from which retrospective categorisations into social inclusion or activation categories can be derived), whether the CE scheme operates as part of the dedicated Childcare or Health and Social Care strands, as well as the administrative type, which details the underlying institutional structure of the CE provider. The latter allows to investigate whether scheme characteristics are associated with different outcomes for participants.
Tús scheme data are structured similarly to CE scheme data
The Tús scheme dataset held by DSP provides information on jobseeker placements on Tús. Each observation in this dataset relates to one placement on Tús, i.e. individuals may have multiple records if they have participated in more than once in Tús. The dataset has information on start and end date of each placement, the reason for exit from the placement, a description of the job that individuals are engaged with and a company identifier to indicate which provider has organised the placement. Information is also held that identifies those individuals that have self-referred to the programme. One important feature to note for this dataset is that the job descriptions for Tús are different to those for CE. The analysis in this report aligns the two by re-categorising Tús job descriptions so that they align to those in CE. In future, it would facilitate better comparison and evaluation of the two programmes jointly, if the administrative definitions for job types in the two schemes were aligned.
The datasets identify all jobseekers who qualify for Tús or CE
Individuals qualify for participation in Tús or CE if they have received a qualifying social welfare payment for at least 12 months and comply with all other eligibility criteria, e.g. in terms of age and previous participation. Qualifying payments include jobseeker’s benefits, jobseeker’s allowance, jobseeker’s transitional payment and a few other types of payments. Exact eligibility criteria changed over the timeframe of the study. For the majority of the eligible populations, eligibility stems from being long-term unemployed (as defined by an unemployment register episode of over one year with a qualifying payment). For Tús, 87% of the sample qualify for eligibility through receipt of Jobseekers’ Allowance (JA) and 12% through Jobseekers’ Benefit (JB), with a further 1% via the receipt of a qualifying One-Parent Family Payment. CE-eligible individuals are similarly weighted towards the long-term unemployed (69% JA and 16% JB), though a greater proportion of individuals qualify through One-Parent Family Payments (15%), which can be paid irrespective of a person’s labour market status.
There are two separately compiled eligibility datasets, which contain all qualifying episodes for individual entitlement for Tús or CE. The datasets condense information on both unemployment and broader social benefits, for all of which DSP has responsibility. Individuals can have more than one record in this dataset if they have more than one distinct eligibility spell, e.g. in case they have two different episodes in their life where they were in receipt of unemployment benefits or other qualifying benefits for more than 12 months. There is a large degree of crossover in the two separate datasets, as very often individuals are eligible for both Tús and CE at the same time. However, there remain several important distinctions between the two datasets, where crossover does not occur. For example, an individual that had recently participated in Tús may be eligible for CE but not Tús again. Similarly, there are some DSP benefits that give rise to CE eligibility but not Tús eligibility (or vice versa).
The eligibility dataset provides information on the start and end dates of all eligibility periods, thereby allowing the construction of a panel dataset to compare individuals that start CE or Tús at a given point in time, to those who do not start but are eligible to start at the same time, which enables counterfactual comparisons. In addition to these data, information such as the benefit payment amounts for the individual, their month of birth, nationality and marital status are also available. Another variable allows identifying casual claims (i.e. jobseeker benefits received while in employment).
Benefits data contain information on all benefits that are administered by DSP
Alongside the eligibility dataset, a benefits dataset which contains all periods of DSP benefit receipt is utilised in the analysis, including benefit periods that do not lead to CE or Tús eligibility. It is populated for all individuals who are identified in the Tús or CE eligibility data and contains one observation for every episode of benefit receipt. For example, an individual who received JA for a period of time and Carer’s Allowance later on, would have two distinct observations in this dataset, which may chronologically overlap if the benefits are not mutually exclusive. Each observation contains a start and end date, a claim status stipulating whether the benefit claim is ongoing, stopped or closed, and a description of the benefit in question. The datasets cover benefits that DSP is responsible for administering, namely Carer’s Allowance, Disability Allowance, One-parent Family Payment, Maternity and Paternity Benefits and Rent Supplement for all benefit claims between 2008 and 2021. It also reports information on contributory and non-contributory pensions. The coverage both in terms of benefits and time period allows to generate a comprehensive picture of each individuals’ benefit history.
Social welfare payments data contain information on all payments related to employment supports; family and children; illness, disability, caring; jobseeker income supports
The social welfare payments dataset covers information on different types of payments made in respect of the benefits data outlined above. These relate to employment supports (e.g. Back to Education Allowance, JobBridge), family and children (Back to Work Family Dividend, Child Benefit), illness, disability and caring (Carer’s Allowance, Disability Allowance, Illness Benefit), jobseeker’s income supports (JA and JB), but also other types of income support such as maternity and paternity benefits, and contributory pensions. Data are provided as annual records, covering the period from 2010 to 2021. Each observation records an annual amount and the number of weeks it relates to, for each individual payment. An individual can therefore have multiple observations per year, in case of receipt of multiple payments. For the purposes of the analysis presented in this report, this dataset represents the main source of information for receipt of Back to Education Allowance, Back to Work Enterprise Allowance, Back to Work Family Dividend, Part-Time Job Incentive, and JobBridge.
Historic Benefit Episode data complement eligibility data and benefits data
A separate dataset provides individuals’ cumulated benefit history prior to the start of their eligibility period. This dataset contains one record per episode in the Tús and CE eligibility datasets and provides information on Back to Education Allowance, Live Register, One-parent Family Payment, Tús or CE, and “other” benefits counts and sums. While the count variable reflects the number of separate episodes for each type of benefit, the sum variable contains the cumulative weeks of receipts across all individual spells. Importantly, this dataset also includes information on instances of JA or JB receipt that led to spells on the unemployment register (the “Live Register”) of less than one year, which would not otherwise be captured, neither in the eligibility datasets, which only record episodes above one year, nor in the benefit dataset, which does not capture JA or JB.
Demographic data provide individual background characteristics
The demographic dataset contains data on individuals in the Tús and CE eligibility datasets and can contain multiple observations per individual in case demographic information changes over time. It provides information on gender, county and region of residence, nationality group, marital status and month of birth. Most changes between records for an individual represent changes to marital status but can also detail where claimants have moved residence. Personal identifiers were removed from the dataset prior to data transfer.
Earnings data record income from employment, self-employment, CE, Tús and other income sources
Annual earnings data are available for all individuals on the Tús and CE eligibility datasets and distinguish between income from employment, self-employment, CE, Tús and “other” earnings sources, with “other” as a default category for earnings that do not fall in one of the other categories. The annual records run from 2008 to 2021, and each observation records an annual amount for each individual payment source. An individual can have multiple observations per year if they have earnings from more than one payment source, such as an individual earning income from employment and self-employment in the same year. Each observation records the total nominal earnings amount and the number of weeks it relates to.
Social security classes A8/A9 contributions allow to identify previous CE spells
In addition to the main dataset on earnings, a second DSP dataset on social insurance subclasses A8/A9 details contributions made in a particular year in the social security subclasses A8 or A9, which relate to either payments for Community Employment, Tús or the Rural Social Scheme. It contains information on all individuals in the CE eligibility data and covers the years 2000-21. Individuals may have more than one observation per year. The dataset is important because it has a longer time horizon than the CE programme data, which allows for the computation of a cumulated CE history from 2000 and to account for lifetime limits on the number of years on CE in the CE eligibility dataset.
The JobPath history dataset provides records of all JobPath episodes
The JobPath history dataset allows the identification of individuals that are referred to JobPath for specific periods. It provides information on start and end dates of JobPath episodes for all individuals who are identified in the CE and Tús eligibility datasets. It contains one record for each episode of JobPath, therefore allowing for multiple episodes per person. It covers the time period 2015-23.
The combination of all DSP and Revenue datasets allows tracking an individual through time
The combination of all DSP and Revenue datasets allows the tracking of an individual through time regarding their labour market status and DSP benefits. This makes it possible to build a picture of the extent to which individuals experience spells of employment and unemployment over time, how much they earn, whether they might have children, care for someone, have a long-term limiting disability or receive help with their housing payment. With this information at hand, it is possible to accurately estimate the effects of participation in CE and Tús, as a very varied group of participants can be compared to other individuals with similar traits and labour market and benefit histories.
4.2.3. The analysis focuses on ALMP participation between 2011 and 2018
The analysis in the report uses a range of cohorts depending on the precise analytical question, starting in the early 2010s. As the analytical needs are different for each of the analytical sections, the precise cohorts used change slightly.
Tús participants are analysed from 2011, which means that individuals are counted from the inception of the scheme. As slightly fewer individuals start Tús per year, relative to CE, it also allows the analysis to utilise sufficient participants to facilitate more precise estimates.
By contrast, the sequence analysis analyses cohorts from 2012 onwards. This choice is made so that a full year of information is available on Tús cohorts, as the policy was introduced only halfway through 2011. This is to ensure Tús participants can be captured on a consistent basis with the other states that are analysed in the sequence analysis on an annual basis.
The CE analysis starts analysing cohorts from 2013 onwards. The larger pool of available participants meant that it was not necessary to go all the way back to 2011, whilst analysing individuals from 2013 onwards still provided a long time series stretching over a number of years, which could provide information on CE over time and at different points in the labour market cycle. For these cohorts, supporting administrative data are available from 2008, i.e. the analysis can make use of five years of pre-eligibility data for all CE participants and eligible non-participants. These data are necessary to ensure that participants are compared to non-participants with similar characteristics (and are arguably more important for the CE analysis, which does not have random selection like Tús). Using ALMP participation cohorts that start in 2013 provides a good balance between large enough samples and the practical consideration on data.
Participants on Tús and CE are included until the end of 2018 only, and not beyond, to allow for a sufficiently long series of post-participation outcome data. As annual outcome data are available until 2021, participants that joined at the latest date that is considered (December 2018) will have two years of annual post-participation outcome data, given that Tús and CE schemes run for one year. Shorter durations of outcome data would not provide any insights into their medium and long-term effects.
As an example, putting together these start and end dates for cohorts, allows the construction of a CE participation dataset between 2013-18 that observes around 28 000 CE participants (Table 4.8). This number is large enough, and the time span is long enough, to allow for sub-group analysis, enabling the report to disentangle the impact of CE on different groups of jobseekers, compare impacts of CE at different stages of post-participation and determine whether there are diverging effects depending on the economic cycle. Similarly, this is also possible for Tús.
Finally, CE and Tús participants potentially include members of sensitive groups, such as people with a history of drug misuse, people who have been convicted of crimes or members of groups that face discrimination such as members of the Traveller and Roma Communities. People on CE schemes where participation indicates a particular racial or ethnic origin, or where it reveals information about a person’s health, were not included in the evaluation. No data associated with participation in schemes aimed at these particular cohorts were analysed and the individuals were not included as potential comparison group candidates if they later became eligible (or had previously been eligible) through other routes (e.g. as a jobseeker). This aimed to eliminate the risks associated with sensitive data, such as data relating to racial or ethnic origin, or data concerning health.
The following section describes the sample used in the Tús analysis including the nature of the referral data (4.3.1), and the process of assembling the sample (4.3.2). The section also includes a description of the quarterly analysis approach and a breakdown of the number of individuals beginning a Tús episode in each quarter and the respective number of eligible candidates in that quarter (4.3.3). Section 4.3.4 details the differing characteristics of the participants and non-participants and 4.3.5 illustrates the categories of work Tús participants are involved in while on the scheme.
4.3.1. The recording and collation of selection and referral data to Tús vary by region
The recording and collation of data relating to Tús referrals vary widely across DSP divisions (operational units aggregated into geographic areas). Each division is responsible for the random selection process, which is not automated or coded, potentially allowing for some variation in how the selection process is run, even before considering how referral and follow-up is handled.
A more consistent set of practices relating to data entry and processes (perhaps hosted centrally rather than within each division) would result in a more consistent dataset for the management and analysis of the scheme (Table 4.2). Better recording of processes at the referral stage could also inform analysis of the relationship between referral to Tús and subsequent claim closures.
4.3.2. The quarterly analysis approach reflects the dynamic nature of enrolment and eligibility
This section presents a brief outline of how the administrative data were compiled into an analytical dataset for Tús that is used for the counterfactual impact evaluation in Chapter 7.
Eligibility is determined by registered unemployment and is assessed on a quarterly basis
To qualify for Tús the individual must be registered as unemployed and meet all eligibility criteria (in receipt of a payment where the relevant claim is counted on the Live Register). Individuals receiving credited contributions or people on casual claims (i.e. receiving benefits while employed) are not eligible for selection to Tús. Eligibility begins when the individual has been in receipt of the qualifying benefit for at least one year.
In assembling the eligible population for Tús only individuals meeting the above requirements were included. Using the Jobseekers Longitudinal Dataset (JLD), 469 725 individuals met the criteria outlined above for the analysis period 2011-18 (Table 4.3). The size of the population was further reduced when some individuals who were missing essential demographic data were removed from the sample. In constructing the eligible population for CE, individuals were excluded from the analysis based on their qualifying route or membership of a minority group (as described in section 4.2.3 above). The same individuals were excluded from the Tús analysis.
Selection and referral to Tús occurs on an ongoing basis and therefore there is no enrolment period (as there may be with education courses for example). The analytical approach is to assess eligibility for the treatment (Tús participants) and eligible population on a quarterly basis. The approach captures the dynamic nature of the evaluation where people move in and out of eligibility throughout the analysis period.
Before conducting the quarterly analysis there were 462 815 individuals in the population (including participants). In order to be included in the eligible non-participant group for any given quarter the individual must remain eligible for the duration of the quarter.
Once individuals have been assigned to quarterly groups, they are compared to JobPath participation records for the same quarter. If someone who is otherwise considered eligible in that quarter is simultaneously engaged with JobPath providers, they are subsequently removed from the control population in that quarter. This restriction only applies to the comparison groups from Q3 2015 to Q2 2018. When JobPath was first introduced it was not possible for someone to be engaged with JobPath providers while also participate on Tús. This rule was changed on 1 June 2018 to allow simultaneous participation on both. Accordingly, the restriction no longer applies in quarters three and four of 2018.
At the end of the quarterly analysis, the sample comprises 331 643 individuals who are either treated for at least 30 days or eligible for a full quarter (and not ineligible through JobPath participation).
Individuals with at least 30 days of participation are counted as having undertaken Tús
At the outset the treatment group (i.e. those participating in Tús between 2011-18) comprises 48 449 individuals. Participants who self-refer to Tús (i.e. who were not randomly selected but decided to apply for Tús on their own initiative) are excluded, reducing the treatment group to 46 601 individuals. The minimum treatment threshold was set at 30 days and any Tús episode below that threshold was removed from the analysis leaving 46 063 people in the treatment group. However, as evidenced in Figure 4.1, the majority of Tús episodes are one year long. The sample was restricted to participants whose qualifying claim ended within 90 days of the Tús start date. Allowing for some degree of data entry error, the sample was restricted to participants whose benefits ceased at most 30 days after the Tús start date. After removing participants without a “relevant” qualifying episode there are 42 426 individuals who start at least one episode of Tús between 2011-18.
Source: Calculations based on Department of Social Protection (DSP) data.
4.3.3. A quarterly Tús sample is compiled between 2011-18
The evaluation covers 328 879 individuals who are either in the eligible population or commence Tús between 2011-18. Of this, the number of Tús participants beginning an episode between 2011-18 (where the episode lasts at least 30 days) is 42 426. The analysis is conducted on a quarter-by-quarter basis. For each quarter, the eligible population is identified as the number of people who remain eligible throughout the quarter. The number of individuals commencing Tús in each quarter is presented in Table 4.4, as well as the number of those eligible in each quarter and, of these people, the number who, in subsequent quarters, commence Tús.
The number of people who commence Tús each quarter ranges from 664 when the scheme first became operational in Q3 2011 before climbing to a high of 2 318 in Q3 2014. It declines over the period after that and by 2018 there were 970 people beginning Tús in Q4. The consistent distribution of participants on Tús across the analysis period shows no indication of seasonal variation in commencements on the scheme. However, the number of Tús participants as a share of individuals eligible for Tús increased markedly, from 0.6% in Q3 2011 to 5.1% in Q2 2018, highlighting that the number of Tús participants is not directly linked to labour market conditions, such as the number as long-term unemployed people.
As represented in Table 4.4, Tús participants make up a small proportion of the eligible population in each quarter with just 1-5% of the eligible population in each quarter starting an episode of Tús. The number of people eligible for Tús in any quarter over the analysis period peaked at 128 254 in Q4 2012.
The DSP introduced the contracted employment service JobPath in Q3 2015 for long-term unemployed individuals. Until June 2018 anyone who was engaged with JobPath could not simultaneously participate on Tús. For this reason, an individual’s JobPath participation is factored into their eligibility status for Tús. As illustrated in Figure 4.2, the number of people eligible for Tús decreases significantly following the introduction of JobPath. Consequently, the number of people eligible for Tús was at its lowest after JobPath was introduced, most notably in the second half of 2017 and the first half of 2018, when it remains below 31 000 in each quarter.
Note: From Q3 2015-Q2 2018 it was not possible to simultaneous participate on Tús while engaged with JobPath providers. From Q3 2018 it was allowed.
Source: Calculations based on Department of Social Protection (DSP) data.
4.3.4. Tús participants and non-participants differ across key characteristics
Having established how the analytical sample was introduced, this section turns to look at the degrees of similarity and dis-similarity between Tús participants and eligible non-participants. This comparison prepares the grounds for the analysis in Chapter 7, which relies on a weighting procedure that corrects for observed differences between both groups. The section first looks at the characteristics of Tús participants as a group and then goes on to present some comparisons to non-participants.
Tús participants are male, prime-aged and Irish
Table 4.5 outlines some basic demographics for Tús participants. It shows that across the analysis period, the modal age group is those in prime age (ages 30-50). However, this average disguises some cohort ageing over time – at the scheme’s inception in 2011 just under a third (28%) of participants were aged under 30, however by the end of the period, this had decreased to one in five (20%). On the contrary, those aged over 50, whilst comprising only 20% of the 2011 cohort, accounted for 30% of the 2018 cohort.
Similarly for gender, there has been a drift towards lower male participation over the analysis period, though men still make up the majority of participants. In 2011 almost three-quarters (73%) of participants were male, but by 2018 this had fallen to two-thirds (67%).
Irish individuals make up the majority of Tús participants (85%) and they have remained a fairly constant proportion of the participants across the evaluation time horizon. Across time, the proportion of participants from new EU member states has increased from around 7% to 9-10%. Whilst this increase being relatively small, compared to the group of Tús participants as a whole, it represents a larger proportional change to individuals from EU member states themselves, indicating that the Tús participant make-up has become slightly more inclusive from a nationality perspective over time.
The median age for Tús participants and non-participants is similar but varies over time
The register of unemployed people (the Live Register) counts jobseekers between the ages of 18-65 years. As eligibility for Tús is based on the set of Live Register claims where duration is at least one year, there is no other upper or lower age limit for participants on Tús other than that for registered unemployment upper and lower limits.
Over the analysis period (2011-18), while participants and non-participants are of a similar age when it is averaged over the quarters, subtle differences in the age distribution emerge when each quarter is examined separately. In the beginning, the median age of the control group is older than the participants and the reverse is true at the end. At the top of the distribution however, the control group is always older (Figure 4.3).
The median qualifying period for Tús is of similar duration for both participants and non-participants
Figure 4.4 shows both the median and 90th percentile values of the unemployment duration spell that made jobseekers eligible for Tús. While both groups appear to have similar median durations throughout the analysis period, as with age, unemployment duration values are always higher for non-participants at the top of the distribution.
PEX coverage is high amongst Tús participants
PEX (Probability of Exit) is a statistical profiling model that predicts the likelihood that a claimant will remain unemployed one year after opening a claim. A PEX “score” is generated based on answers to a questionnaire provided by the jobseeker at claim commencement. PEX information is recorded for 82% of Tús participants (Table 4.6). This is a significant level of coverage compared to the PEX information available for 57% of all candidates in the eligible population (190 493). When conducting sub-group analysis using PEX, it is important to bear in mind this difference in coverage because it may mean that the sub-sample is not more broadly representative of the population in general.
For both the participant and non-participant groups, where PEX records exist, the values are concentrated in the range 10-40 (out of a maximum of 99). Figure 4.5 illustrates the difference in median PEX values between the two groups for each quarter. For the majority of the analysis period, the median PEX value for non-participants is marginally higher than the respective participant group in that quarter. This changed in 2016 and up to the end of 2018 the median PEX value of the participants was higher than non-participants in that quarter.
Note: PEX: Probability of exit. Median values are taken in each quarter. The “control” group refers to non-participants.
Source: Calculations based on Department of Social Protection (DSP) data.
When looking at information on self-reported health, a higher proportion of Tús participants reported “Good” and “Very Good” levels of health than non-participants (Figure 4.6, Panel A). This means that is imperative that variables exist, such as previous income, receipt of health and unemployment benefits, that are able to proxy for this difference in health outcomes for the estimation on the full sample of Tús participants, for whom this data is not captured. Similarly, when look at education, Tús participants reported a higher proportion of junior and leaving certificate than non-participants, with fewer participants having primary education or less (Figure 4.6, Panel B). However, some of these differences – towards greater education for Tús participants – are negated when looking at tertiary (third level) education. Here individuals in the eligible population having slightly higher tertiary attainment.
Note: The “control” group refers to non-participants.
Source: Calculations based on Department of Social Protection (DSP) data.
4.3.5. Tús work placements fall into five broad categories
Tús programme data has information on the type of job placements, with the placements falling into the following categories: Environmental services, General community services, Heritage cultural services, Para-educational services and Caring services.
A breakdown of work placements completed by Tús participants in 2011-18 is illustrated in Figure 4.7. “General Community Services” is the largest category for participants throughout the analysis period and includes activities such as administration, community development and community media and radio.
Source: Calculations based on Department of Social Protection (DSP) data.
Spatial information on the location of Tus services can be mapped to information on the social deprivation of areas. This can provide insight on whether the delivery of services provided by Tús participants is located in more affluent or disadvantaged areas.
A deprivation index assesses social conditions in a given area using a single indicator. The Pobal HP Deprivation Index is a census-based deprivation index for the Republic of Ireland. The index is used by Departments, State and semi-State agencies, as well as voluntary and non-governmental organisations (Haase and Pratschke, 2016[3]). Pobal is a state-sponsored intermediary agency that works on behalf of the government to support communities and local agencies toward achieving policy goals in relation to social inclusion, reconciliation, and equality.
The index draws on indicators from three dimensions of affluence and disadvantage – demographic profile, social class composition and labour market situation – to assign a deprivation score along the spectrum of affluence-disadvantage in each area relative to the mean at a specific point in time. The index contains an absolute and relative score. The relative index score details how deprivation in that area relates to all other areas at that point in time. The relative scores are derived by subtracting the underlying trend from the absolute scores and rescaling them so that they have a mean of zero and a standard deviation of ten at each Census wave.
The index used for the analysis below is based on the 2016 Census, as that was recorded in the middle of the analysis period and provides the best reference point. The relative deprivation scores used in the analysis are based on electoral divisions, the smallest legally defined administrative areas in the State. Some 3 409 boundary features are defined as electoral divisions, of which 1 209 are associated with activities carried out by Tús participants in the evaluation dataset.
The majority of Tús placements that map to electoral divisions fall in areas whose relative index score is either marginally above or marginally below average. A greater proportion of Tús placements are in areas that are marginally below average on the HP Index compared to the population overall. In short, Tús placements are under-represented in affluent areas and over-represented in disadvantaged areas, thereby contributing some degree of redistribution at the level of services provided.
This section provides more details on the individuals that participate in CE schemes and information on the CE schemes that they participate in. The section starts with descriptive statistics on CE participants (4.4.1) before it discusses the types of CE positions undertaken (4.4.2) and then compares CE participants to eligible non-participants (4.4.3).
4.4.1. Linked administrative data allow a detailed picture of CE participants to be drawn
The linked administrative data provided by DSP and Revenue provides a detailed picture of the characteristics of CE participants and eligible non-participants. There is a high degree of variability among CE participants, in terms of both personal characteristics and CE scheme characteristics (Table 4.8) highlighting that there is no “typical” CE placement but that CE is a programme supporting jobseekers in many different circumstances.
CE participant numbers remain relatively steady in the face of a large fall in the number of CE eligible jobseekers over time
The number of CE participants in the evaluation sample stays roughly stable throughout the entire period with moderate increases between 2013-15 followed by a moderate decline between 2015-18. By contrast, the number jobseekers eligible for CE per year has fallen sharply over time from 184 000 in 2013 to 98 000 in 2018, reflecting a decrease in the number of long-term unemployed people amid major labour market improvements (see Chapter 2).
The difference between the trends in the number of eligible jobseekers and CE participants is due to the role of CE for communities. While the number of eligible jobseekers directly depends on the number of long-term unemployed people, thus narrowly reflecting labour market developments, the number of CE positions is influenced by local needs for community work which is not directly linked to labour market demand in the private market. Therefore, the pool of eligible jobseekers who could potentially fill a CE vacancy was much higher in 2013 when long-term unemployment was still record-high than in 2018 and it was more challenging for long-term unemployed jobseekers to secure a CE position at the beginning of the observation period than towards its end.
Most CE participants are prime-age or older jobseekers and male participants are on average older than their female peers
Turning to the characteristics of the CE participants, the majority of CE participants fall in the age group 30-50 years, with an average age of 43 (Table 4.8). The age profile over time remains fairly constant although the proportion of young participants under 30 and that of older participants over 50 increase somewhat between 2013-18, from 14% to 17% and from 29% to 35% respectively. The distinction between age groups is important as the role of CE is not the same for jobseekers of different ages. While CE for young and prime-age participants always aims at a reintegration into the primary labour market later on, older participants just below retirement age are more likely to experience CE as a pathway from the labour market into retirement.
The gender profile too has been fairly stable over time. It is weighted more towards men who comprise 57% of total participants reflecting that there are more long-term unemployed men than women in Ireland (Chapter 2). Male participants tend to be older than female participants, with 39% of male participants being aged over 50, compared to only 22% for females. Conversely, 20% of the female participants are aged under 30 whereas they are only 10% of men.
Participants are mainly located in urban areas in all parts of the country
Around two-thirds of CE participants reside in urban areas largely reflecting differences in population size between urban and rural parts of the country. More specifically population size accounts for approximately two-thirds of the variation in the number of CE participants across counties (Figure 4.8), whereas the remaining variation is driven by other factors such as local labour market conditions and the number of CE sponsors in the area.
CE is present all around the country. While most CE placements are found in Dublin, Cork, Limerick and Galway due to their large population, less populated areas also have, relatively speaking, a high CE density. For example, County Monaghan has the highest number of CE participants per 100 000 of population in the data at 3.87 but is only 17th out of the 26 counties in terms of its absolute number of participants due to its small population.
Note: County populations based on 2016 census population data (www.cso.ie/en/releasesandpublications/ep/p-rsdgi/regionalsdgsireland2017/nt/). All CE schemes starting 2013-18. Ordered by total proportion.
Source: Calculations based on Department of Social Protection (DSP) data.
Similar to Tús, CE participants are more likely than the average Irish person to be located in deprived areas. According to the Pobal HP Deprivation Index classification, 21.7% of participants live in disadvantaged areas and a further 37.2% live in areas marginally below the average (this compares to 7% and 41% respectively for all Irish citizens). A further 10% of CE participants live in affluent areas and 31.2% in areas marginally above the average (compared to 10% and 42% respectively for all Irish citizens). In this sense then, CE is supportive in a socially inclusive manner through the provision of greater support to more deprived areas.
However, the limitation of this analysis (and the corresponding analysis for Tús) is that this information is available for participants only and this dynamic may simply reflect more eligible individuals in these areas. It cannot provide insight as to whether Tús and CE are better targeted to those individuals actually eligible in deprived areas.
Most CE participants are Irish but in relative terms British and EU migrants are more likely to take up CE
In absolute terms Irish participants make up 86% of the total CE population followed by jobseekers from “recent” EU member countries (2004 or later) who account for 7% of participants, whilst individuals with British nationality comprise a further 5% of participants. Only 2% come of participants come from other countries. The distribution of CE participants by nationality has remained broadly stable over the evaluation window.
The picture appears different when employing relative metrics, i.e. comparing participant prevalence relative to overall population size using population and migration estimates (CSO, 2016[4]). British nationals residing in Ireland are twice as likely to participate in CE than the overall population, while EU nationals from post-2004 accession countries are 30% more likely and those from pre-2004 EU countries are 23% more likely. Conversely, Irish nationals are slightly less likely (98%) to participate in CE than the overall population, and “other” nationals are significantly less likely (37%). These differences may be the result of differences in labour market outcomes and eligibility across nationalities as well as other possible factors such as knowledge of the programme integration into the community, desire to participate and the availability of CE positions in the place of residence.
Most CE positions are activation placements, fall into the general CE scheme and are provided by voluntary bodies
CE positions can be distinguished according to their main objective (activation in the primary labour market or improved social inclusion), the type of work that is carried out (general CE schemes or positions in Childcare or Health and Social Care) and the type of CE provider that offers the position (voluntary bodies, public bodies or other).
Activation placements make up the majority of the places offered on CE. Around 60% of the total CE places are categorised as activation placements, whilst 22% are classified as social inclusion and for 18% information on their classification is missing. Without further knowledge on unclassified placements, the activation strand could account for anywhere between 60-78% of placements and social inclusion between 22-40%. The share of activation placements seems to be falling at the end of the observation period from 66% in 2016 to 48% in 2018. However, there is a trend towards more missing information on categorisation over time making statements on the evolution of the distribution of activation and social inclusion jobs less reliable. In 2018 for instance more than a quarter (27%) of jobs were unclassified.
Most CE participants (84%) are in general CE schemes. In addition, a significant minority of CE schemes cater to the specific Childcare (7%) or Health and Social Care (9%) programmes. While these two programme types make up 16% of total CE placements on average between 2013-18, their role is decreasing over time. In 2013 almost one-fifth (19%) of total CE placements were catered to one of these two schemes against only 9% in 2018. The prevalence of these schemes in the data is high enough to determine whether better outcomes are associated with these specific programmes. This is an important piece of evidence to review given that places for both of these schemes are generally ring-fenced and better or worse relative performance may lead to review of the extent to which these programme streams are fostered by DSP in terms of overall CE places.
The overwhelming majority (95%) of CE providers are classified as voluntary bodies. These bodies are typically charities and provide a range of health, social care and homeless services to the Irish population. The remaining 5% of CE sponsors are public bodies (4%), community groups (1%) and local authorities (<1%). In practice the sponsor type of the CE placement has a lower relevance than the exact job type.
Risk-scoring information on CE participants is limited historically but growing over time
Since the introduction of Ireland’s current public employment service (PES) Intreo in 2012, newly unemployed individuals entering registered unemployment should have completed a risk scoring questionnaire, known as the Probability of Exiting (PEX) questionnaire. This questionnaire is designed to estimate the likelihood of an individual exiting the Live Register prior to one year of unemployment. The questionnaire asks a raft of questions on a jobseeker’s background and skills that provide useful contextual information for CE participants. In particular, it provides an additional set of socio-economic questions that allow to test the sensitivity of the central methodology to omitted information. PEX information is available for an average of 58% of CE participants, although the availability is heavily weighted towards the latter end of the sample (Table 4.8). While individuals that were already unemployed in 2012 do not have PEX information at all 75% of the 2018 cohort had answered the questionnaire.
4.4.2. CE placements cover a wide range of different job roles
In addition to the high-level categorisation of CE placements into activation and social inclusion placements lower-level job descriptions are available in the data providing more detailed insight into typical CE jobs and how they provide a range of services to local communities. These job descriptions have been compiled by DSP for this project, to reduce the dimensionality of the longer job description titles held in the CE programme administrative records (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[2]).
Detailed job descriptions are available for about four-fifths (82%) of placements (Table 4.9). The CE placements without any job description are the same placements that do not allow for a higher-level categorisation into activation and social inclusion strands. The top 30 job descriptions account for the vast majority of all CE placements (91%) for which a job description is available.
The most common CE jobs are caretaker, general operative, childcare worker, administrator and environmental worker. These jobs descriptions provide a hint to the extent to which CE can provide opportunities for the use of human capital accumulation to prepare for the broader labour market after CE participation. In the case of caretakers and general operatives, it is likely that CE participants help to perform routine maintenance and domestic support to service provision in most cases. Such placements offer an opportunity to sharpen general skills such as reliability, punctuality and teamwork but they may have limited ability to contribute to the accumulation of specific job-related knowledge and competences even with the embedded CE training budget.
CE positions for childcare workers or administrators arguably offer greater scope for the development of job-specific skills that could be useful for obtaining further private-sector employment via the provision of training that is tailored to the needs of employers. This is precisely the idea behind the Childcare CE programme where there is a requirement for applicants to demonstrate a commitment to engage in sector-specific certified training which would lead to the achievement of a certified Quality and Qualifications Ireland (QQI) major award . Similarly, for administrators, already small modular courses might offer the opportunity to gain skills required by employers. For example, a short course in basic excel use or database management for administrators might be enough alongside the experience offered on CE to open up job opportunities across a range of private-sector jobs. These types of courses have been shown to be effective in other OECD countries (OECD, 2023[6]).
Further jobs in the list of top 30 CE jobs include secretaries, accounts clerks and information officers. Many of these jobs similarly offer the possibility for the development of career paths based on the start of accumulation of skills and experience in these roles. In order to be effective, CE has to allow participants to acquire skills and competences in line with labour market needs and the way CE schemes operate should be informed by evidence on what works well for other active labour market policies (ALMPs).
The fieldwork for this report uncovered promising practices from CE schemes, some of which have been able to establish strong business links with leading technology firms and structure their placements with a view to equip participants, through appropriate training and experience, with the skills and competences required to join these firms. For example, in some schemes, private-sector staff from big private companies are available to provide motivational talks and help candidates with interview preparation. While this level of connection with big tech companies is not possible for all CE schemes, better links to local employers can help tailor skill development during CE participation and increase the chances of finding employment after CE.
The current application procedure for approving CE placements is mostly based on whether a placement meets certain criteria on the type of job (such as the voluntary nature of the work and its necessity, and the training provided to the individual) rather than the extent to which the placement enhances the candidate’s further job prospects. Additional efforts to prepare participants to the realities of the local labour market could be worthwhile. For example, this could take the form of some kind of learning forum for the DSP regional managers, the CE supervisors or scheme administrators to ensure that targeting takes place where it can.
The availability of training budget also distinguishes CE from Tús
Unlike Tús, CE includes an annual training budget of EUR 250 per person. The uptake of training towards a major award is also one of the conditions for repeating CE participation for younger participants. Table 4.10 presents an overview of the types of training associated with CE schemes. Over half (56%) of CE schemes have an associated training record attached to them. However, most commonly there is little information on what this training comprises with over 60% of those CE schemes with training having “other or unknown” training (34% of all CE schemes) recorded in the data. Of the training with existing underlying information on the content of the training, the most common type is for QQI minor awards. This training is featured on 12% of all CE schemes. The most common types of training within this category are for occupational first aid and training related to horticulture. Another significant training category is for non-certified training. These training courses are most likely to be associated with jobs requiring training in manual handling, safety training or health-related training.
4.4.3. CE participants and eligible non-participants are not alike
This section presents descriptive statistics on the characteristics of CE participants relative to the eligible population that did not participate in CE. It complements section 4.4.1, which provides a description of CE participants but not the eligible population as a whole.
CE participants are older and more likely to be men, Irish and have experienced worse labour market outcomes in the past than the broader eligible population (Table 4.11). The average age of CE participants when they start the scheme is 43 years old which is two years older than the average eligible non-participant. They also have slightly fewer dependent children (0.82 vs. 0.89 on average) and are more likely to be married (40% vs. 29%).
The larger proportion of male CE participants than female participants is primarily driven by the fact that there are more long-term unemployed men than women, with men accounting for 55% of the eligible population but also because they have a greater tendency to participate once eligible, raising the final proportion of male participants to 57%.
Similarly, when accounting for eligibility, people with Irish nationality are proportionally more likely to take up CE than foreigners, as Irish nationals account for only 82% of the eligible population but 86% of all CE participants. Irish nationals may have a higher tendency to participate in CE as they have better social networks including friends or family who have completed the scheme or because they are more familiar with the Irish benefit system.
In terms of labour market outcomes, CE participants have worse recent earnings and unemployment histories. In total they experienced more registered unemployment and have earned less in the labour market than their peers. In the two and three years prior to eligibility, CE participants had employment earnings of about three-quarters (74%) of that of eligible non-participants and, for those who were self-employed, less than two-thirds (63%) of their self-employment earnings. CE participants also have more weeks of receipt of unemployment benefits for both JA and JB. Two and three years prior to eligibility, they average 13% more weeks of JA receipt and 34% more of JB. They have experienced a higher number of instances in registered unemployment (Live Register) and have spent more time in registered unemployment.
There are also systematic differences between CE participants and eligible non-participants in the use of the wider ALMP and benefit system. Most strikingly, the likelihood of having been a Tús participant is much higher for CE participants than for eligible non-participants. Tús appears as a strong entry point for CE participants, with 16% of CE participants having previously completed Tús relative to only 3% in the broader eligibility population (see Chapter 7 for a detailed evaluation of how Tús helps jobseekers). In addition, CE participants have spent almost twice as long in receipt of Back to Education Allowance than eligible non-participants, suggesting that, for some CE participants, there is strong motivation and willingness to engage in ALMPs that will help to foster better links to labour market opportunities. Finally, CE participants are also less likely to have recently been recipients of Carer’s or Disability Allowance or Family Benefits.
These differences give rise to a vastly different qualitative description of CE participants relative to eligible non-participants in terms of their previous labour market characteristics, their demographics and their receipt of benefits. Any analytical strategy to determine the impact of CE will need to ensure that it is the programme driving difference in outcomes, rather than innate differences between participants and non-participants. This is an issue that Chapters 6 and 7 will address.
Despite the richness of the data on which this evaluation draws, a number of data limitations are worth highlighting. This section outlines some immediate improvements that could be made to the administrative data and highlights longer term developments.
4.5.1. Implementing a structured analytical data framework would facilitate timely and consistent analysis
Linked administrative data form the bedrock of any analysis of ALMPs but substantial efforts had to be made to clean, link and assimilate disparate data for this report. Ensuring that such data cleaning, linking and manipulation is performed routinely in DSP would lower the costs of evaluation of different policies, by making it quicker and easier for analysts to perform counterfactual impact evaluations.
The replacement of the previously maintained Jobseekers Longitudinal Dataset, which was compiled for this reason, would help to re-establish such data for analysis. This would make any future DSP policy evaluations more nimble, more consistent and require less analytical resources to undertake. Of course, there would be an analytical cost to the establishment and maintenance of such a data architecture, alongside the concomitant need for strong supporting metadata to aid analysts. A choice would have to be made as to the appropriate scope and update frequency of such a system, taking in views from analytical and policy staff across DSP (and wider government agencies). However, once done, it would significantly enhance the ability of analysts (both internal to DSP and the Irish Government, but also to any approved external researchers) to provide policy-relevant analysis and evidence quickly and efficiently.
4.5.2. Improving data quality and data scope would also support future policy analysis
The compilation of administrative microdata for this report, also revealed several areas whereby improvements could be made to aid policy evaluation. A few examples are described in this section to provide some information on the type of error and conflict that emerged in the analysis. More details can be found in the accompanying technical report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[2]).
Data sometimes had incompatible start and end dates, where different programmes could not be undertaken simultaneously, but were reported as such in the data. For example, there were individuals with seemingly overlapping Tús and JobPath episodes, which, prior to June 2018 was not possible. Administrative efforts to clean and reconcile such inconsistencies would be helpful to any collated evaluation data, so that all analysts make the same assignment rules on which dates are correct.
Data coverage was limited on some different administrative data sources; improving this would enable more precise and detailed policy evaluation. For example, PEX data is either missing completely, or there are consistent question responses missing, even where individuals should have this information completed. There is no unified database that collates information on the receipt of the range of benefits that DSP offers. These are instead contained on disparate datasets. Having a unified and consistent record of such receipt would enable a consistent view of the totality of DSP support and would decrease analytical re-work. Lastly, historic information on DSP benefit receipt and payment amounts is sometimes limited. For example, payment information goes back to 2010 only, which make it challenging to construct individuals’ payment histories prior to this point in time.
The incorporation of administrative datasets from other ministries would also enhance the quality and scope of policy evaluation. Education data from the Department of Education would provide information about education and training could also be a useful addition to DSP’s data on labour market status and interventions. Education data provide critical information to control for individuals’ background, particularly for young people where past labour market outcomes cannot be used a proxy for this. Alongside supporting broader ALMP analysis, a reduction in the number of questions addressed to the jobseeker on the PEX questionnaire is taking place following an ESRI review. Already having education data would allow questions on education to be removed from the PEX questionnaire. Turning to outcomes, incorporating information on hours worked to the administrative employment data in Ireland could yield a significant improvement to analytical capacity. Similarly, at present there are no linked health or justice data available to examine the wider impacts of ALMPs and how the participation in an ALMP affects usage of health services or criminal acts.
A more consistent set of practices relating to data entry and processes would result in a more consistent dataset for management and analysis, as manually collected data are often quite variable at a regional level. For example, the variation in the data on how people are referred to Tús and any subsequent actions appears to hinge on data recording and data collation practices in different geographical divisions. Bringing some kind of central co-ordination and collation to these statistics, to ensure that different areas are recording and reporting information in the same manner, would bring more consistency to any monitoring and evaluation and ensure that any differences between delivery are the result of actual differences in operation rather than in reporting.
Metadata on administrative data was often limited. Therefore, the quality of variable responses, the inference on any missing data and the preference for utilisation (of the same underlying characteristic) from two (or more) separate but inconsistent sources, was indeterminable. Ensuring that data quality, data limitations and more generally data sources are fully documented across all sources will help to ensure that the right data are used for the right purposes with the right caveats.
This chapter has outlined how individual-level microdata from the DSP and Revenue have been leveraged to permit a rich analysis of how individuals eligible for Tús and CE navigate their way through the system of support offered by DSP and how those that participate in either scheme are able to enjoy the benefits from this participation. At its core, detailed programme information on entry into different ALMPs and wider DSP services, the receipt of benefits from DSP and on related labour market outcomes are linked together to enable the analysis to build detailed pictures of individuals’ journeys over time.
The data show some similarities and some differences between Tús and CE participants. Both schemes are more likely to have men participate in the sample used for the evaluation, particularly so for Tús where 70% of all participants are male. The age distribution for the Tús participants is slightly younger, where around one-quarter (26%) of participants are aged under 30 (contrasting to only 14% of CE participants). Whilst Irish nationals make up the vast majority of Tús and CE participants, in relative terms those with British or EU nationality are more likely to take up a place on the scheme. Both schemes are more likely to operate in areas with a higher-than-average level of deprivation compared to the broader Irish population, underlying their benefit to society in a redistributive sense.
Notwithstanding the good data that have enabled the analysis in this report, further efforts could be made that would better support any future policy evaluations. Better data assimilation to create and maintain datasets for analysis would lower future research costs. Expanding the provision of data across a number of areas would permit richer analysis. Supporting all of this with detailed and informative metadata is imperative to maintain consistency and improve analysis quality.
References
[4] CSO (2016), Population and Migration Estimates, April 2016, Central Statistics Office, http://www.cso.ie/en/releasesandpublications/er/pme/populationandmigrationestimatesapril2016/.
[5] DSP (2019), Community Employment Procedures Manual, Department of Social Protection, Ireland, https://www.gov.ie/en/collection/6f5e19-community-employment/.
[3] Haase, T. and J. Pratschke (2016), The 2016 Pobal HP Deprivation Index, http://trutzhaase.eu/.
[6] OECD (2023), Evaluation of Active Labour Market Policies in Finland, Connecting People with Jobs, OECD Publishing, Paris, https://doi.org/10.1787/115b186e-en.
[1] OECD (2022), Assessing Canada’s System of Impact Evaluation of Active Labour Market Policies, Connecting People with Jobs, OECD Publishing, Paris, https://doi.org/10.1787/27dfbd5f-en.
[2] OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre (2024), “Technical report: Impact Evaluation of Ireland’s Active Labour Market Policies”, OECD, Paris, http://www.oecd.org/els/emp/Ireland_ALMP_Technical_Report.pdf.