5. Sequences of support schemes in Ireland
This chapter sets the scene for the evaluation of the impact of participation in two public works programmes in Ireland, CE and Tús, that will be presented in the next chapters. It presents an overview of the main patterns in participation in active labour market policies (ALMPs) observed in the data for those eligible for CE and Tús, mostly long-term unemployed.
It shows the typical individual trajectories across selected ALMPs and labour market states, the time spent in three highly relevant conditions – namely CE, Tús, and eligibility to the two schemes – and the probability of moving to other states, taking into account the role played by individual characteristics like gender, age, nationality, and marital status.
The depiction of trajectories can offer a more comprehensive narrative on how specific patterns can lead to better or worse outcomes over time and provide a good introduction to the analysis of more specific events – e.g. participation in CE or Tús – that will be covered in the next chapters.
This chapter provides an empirical framework for the evaluation of the impact of participation in Community Employment (CE) and Tús, presented in Chapters 6 and 7. Instead of looking at the impact of participation vs. non-participation in a single programme, it takes a broader perspective and analyses the typical sequence of programmes in which individuals participate, therefore considering different types of schemes, and the specific timing of each of them in a sequence. The present chapter, therefore, treats CE and Tús simply as one of the many events that may occur within an individual’s life, making clear which are the most common prequels and outcomes, and assessing their relevance, in terms of the number of people involved and for how long, within the overarching context of the Irish social protection system.
Coherently with the analysis in the rest of the report, the reference population for the analysis presented in this chapter is composed of all people who have been eligible, either for CE or Tús, for at least one day between 2012 and 2018. This is mostly long-term unemployed – i.e. individuals who have been unemployed for at least a year – but also includes a small share of people (around 10%) who qualify to CE because they are in receipt of One-Parent Family Payment (OPFP). In the rest of the chapter, for the sake of clarity, eligible individuals might be called “long-term unemployed” or “jobseekers”, even though a minority of this group is not technically in this category. It should be noted that the reference population for the analysis includes also individuals who are observed participating in CE or Tús, despite not being formally eligible through the long-term unemployment or OPFP paths; these individuals likely qualify through alternative paths, not observable in the data available. Eligible individuals and participants, hence, do not fully overlap.1 Of this target population, a “snapshot” is created that portrays each individual for the 96 months from January 2012 to December 2019, on a monthly basis. Starting from this reference population, each analysis described in the rest of the chapter will further sample individuals from this dataset, according to specific aims and methodologies of the analysis in question.2
This chapter defines a limited number of possible states in which individuals can find themselves during the course of their personal trajectories through the labour market and associated support services, and categorises peoples according to these states. The possible states are eight, and were selected on the basis of relevance, as well as data availability and reliability: CE, Tús, employment with support (EWS), employment without support (EWoS), Back to Education Allowance (BTEA), JobBridge, JobPath, and a residual state capturing no participation in employment nor in any of the supporting schemes mentioned (see Table 5.1 for a summary).
The first two states considered in the analysis are participation in CE and Tús, which will be the focus of analysis in Chapters 6 and 7. The other categories capture engagement of eligible individuals in other relevant schemes. The first one of these alternative states is defined as “employment with support” (EWS). This category covers the scenario where participants are employed in the open labour market, but still receive some sort of support, either in terms of income support to complement earnings from employment, or incentives for employers hiring jobseekers, in order to be able to access or maintain this employment. While this state still captures participation in an ALMP, EWS can also be seen as a step out: the experience gathered while supported, indeed, can provide the worker with the skills needed to stay employed anyway. The schemes included under this heading are the following:
JobsPlus: this is a subsidy paid to the employers which encourages and rewards those who offer employment opportunities to individuals who are unemployed. It provides employers with two levels of payment: EUR 7 500 or EUR 10 000 over two years, with the level of payment depending on factors such as the age of the jobseeker and the length of time in receipt of a qualifying payment.
Back to Work Enterprise Allowance (BTWEA): this scheme encourages people getting certain social welfare payments to become self-employed. It allows people to develop a business while retaining a percentage of their social welfare payment for up to two years.
Back to Work Family Dividend (BTWFD): this is a weekly payment to help people with children move from social welfare into work. It gives financial support to people with qualified children who are in or take up employment or self-employment and stop claiming Jobseeker’s Allowance (JA) or Benefit (JB), One-Parent Family Payment or Jobseeker’s Transitional Payment (JST, a social welfare payment for people who are parenting alone, whose youngest child is aged between 7 and 14).
Part-Time Job Incentive (PTJI): this scheme allows certain people getting Jobseeker’s Allowance to take up part-time work and get a special weekly allowance instead of their jobseeker’s payment. It is intended to be a stepping stone to full-time work.
Casual jobseeker claims: these are unemployment claims under which jobseekers can work for up to three in seven days, and receive an unemployment payment for the remaining days.
This list demonstrates that EWS category encompasses very different situations, from strictly speaking ALMPs such as JobsPlus, to schemes facilitating the take-up of part-time work as a stepping stone to full-time work, and income supports making employment economically more viable. It should be noted that this category only includes types of support where employment is part of the design, i.e. where employment is required for receiving a certain support; this rule leads to the exclusion of some forms of income support that are designed to encourage employment but where employment is not a prerequisite of receipt, e.g. One-Parent Family Payment.
Other schemes considered as separate states are:
Back to Education Allowance (BTEA): this scheme helps people who are unemployed, are getting a One-Parent Family Payment or have a disability, to attend approved second- or third-level education courses;
JobBridge: this was a national internship scheme, in place between 2011 and 2016, offering six or nine-month placements in organisations in the private, public and community and voluntary sectors;
Referral to other Employment Services – JobPath. JobPath is a contracted employment services programme to help people who are long-term unemployed find employment.
The last substantive state considered is employment without any type of support. This captures a condition in which an individual is found with employment-related contribution weeks, without being at the same time in receipt of “accompanying” benefits or under subsidised schemes like the ones mentioned above.
A residual state, capturing the situation in which individuals in the reference sample are not observed in any of these states in the period under analysis, is added to the list of states in the sequence analysis. For simplicity, this will be labelled as “Other”, which can also be considered as “none of the above”. One can imagine this category encompasses a number of possible states. These individuals might still be registered as unemployed, possibly participating in schemes which are not captured in the data available; for example, participating in training schemes, or assigned to alternative employment services. They could have moved abroad, or simply be inactive. As will be explained in the next section, however, this state should largely exclude inactivity due to receipt of state or widow(er)’s pension, maternity or paternity leave, disability or caring for others, as all individuals in receipt of related benefits are excluded from the analysis. Deceased individuals are also excluded from the analysis.
Before explaining how the dataset for the analyses was built, and presenting the results of the analyses, it is worth highlighting two points to be considered in the study:
Each of the schemes included under the various states have specific target groups and eligibility rules; these are not taken into account in the analysis, which only considers as target group those eligible for CE/Tús. Participation in other schemes is only considered as an alternative state observed in the data for this group.
The list of possible states is not exhaustive. In the preparation phase, many other schemes were explored as options to include in the analysis. However, data imitations made it impossible to cover more states than those in the list above. More information on these limitations and options explored will be provided in the next section, as well as in the accompanying technical report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[1]).
To analyse sequences, a dataset is constructed in which every individual, at each moment in time, is uniquely attributed to a state, according to the rules described above. The dataset is constructed into a “grid” with person timelines and variables on the columns. Table 5.2 provides an artificial example of such structure in which individual identified by identifier (“ID”) 50 is observed in March 2016 under state “Other”. In June of the same year, he moves to a job with no support (EWoS), a state persisting until December 2016. In January 2017, he moves again to “Other”, and in April he enters JobPath (even though the variables on gender, age and marital status are observed in the data, the ones reported in the table are invented values to ensure data protection). The choice of this person-month structure for the data is mainly dictated by the flexibility it provides with duration analysis (see section 5.6); hence, it was decided to preserve this structure for all the analyses for the sake of comparability.
The database used for the analysis builds upon nine different individual sources of raw data, which are visually represented in Figure 5.1. Since raw data provide complete information for eight consecutive years from 2012-19, each of the 466 152 individuals included in the final dataset is observed for 96 consecutive months. In this sense, the data constructed is what is known as a balanced panel.
The starting point for the analysis is the identification of individuals who were eligible for either CE or Tús at any point in the period between 2012-18, which represents the reference population in this chapter. This information is retrieved from the eligibility dataset provided by the Department of Social Protection (DSP), as explained in Chapter 4. Once this group is selected, the next step is defining the main state in which each individual is found in each month of the panel described above.
Information on the possible states is retrieved from several different data sources: CE and Tús participants (scheme) data, DSP benefits data, social welfare payments data, earnings data (Revenue), and JobPath history data. These datasets can be broadly grouped in two categories, based on whether they contain records relative to spells – with start and end dates for each single episode – or annual records.
CE and Tús participants data, DSP benefits data and JobPath history data belong to the first category. They all contain detailed information on exact start and end dates of CE and Tús schemes, DSP benefits and JobPath referrals, making it easy to assign the relevant states to specific points in time during each year. CE and Tús scheme data were used to define periods when individuals were participating in these schemes; episodes were merged when the gap between them was shorter than 30 days.
Source: Author’s representation.
The DSP benefits dataset is the main source of information for JobsPlus participation, i.e. for one of the schemes that falls under the supported employment status. It also provides information on individuals who, during the period of analysis, have been in receipt of maternity/paternity benefits, carer’s allowance, disability allowance, contributory pensions; as mentioned below, this information will be used to select individuals for analysis. JobPath history data also show exact periods when individuals were referred to this employment service. In addition, CE eligibility data provide information on precise periods of casual claims.
The earnings and contributions and the social welfare payments datasets posed a different set of challenges. The former is the source for information on employment without support. For the purposes of the analysis presented in this chapter, the social welfare payments dataset represents the main source for data on receipt of Back to Education Allowance, three of the supported employment types (Back to Work Enterprise Allowance, Back to Work Family Dividend, Part-Time Job Incentive), JobBridge. These two datasets are only available on an annual basis, with no information on start and end dates for benefit/employment spells. In order to reconcile this with the monthly structure of the dataset built for the analysis of sequences, some rules of thumb had to be implemented to allocate weeks of benefits/employment to specific weeks of the year. This required to make hypotheses on start and end dates of EWoS, BTWEA, BTWFD, PTJI, JobBridge and BTEA, that needed to be co-ordinated among them and with (known) end and start dates of JobPath and JobPlus; the allocation rules are presented in detail in the accompanying technical report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[1]).
The main operations performed on each of the raw data sources, with an emphasis on the loss of observations implied by the main steps, are described in the Technical Report, while the final merging procedure is summarised below.
5.3.1. Data are merged together into a unified dataset defining spells
Taking these separate source datasets and combining them together in the grid structure previously described involves several steps. First, each cleansed feeder is turned into the person-month structure described above in Table 5.2, so that merging can be performed using a unique identifying key which combines an individual identifier, the year and the month. Second, information is merged on individuals eligible to CE or Tús, to build the core of the reference population. There are 364 831 people who appear both as eligible to CE and to Tús, during at least a month within the target timeframe and not necessarily in the same moment; another 215 464 are eligible to CE only, while 59 648 to Tús only. CE (Tús) episodes are then added to the pool of eligible persons. For 16 491 (1 050) individuals with CE (Tús) episodes, there is no corresponding eligible spell even partially falling within the period from January 2012 to December 2019. Although no eligibility period potentially related to these CE or Tús episodes is observed, these episodes are anyway retained in the data. Of the resulting dataset, 1 167 persons have missing demographic information: these persons are not dropped, unless the corresponding piece of demographic information is needed for the analysis (e.g. breakdowns by gender where gender is missing).
Individuals that may have a reduced or unstable labour supply or availability to participate in full to the interventions are dropped from the analysis. Individuals are dropped in this order:
People observed to receive maternity or paternity benefits: 57 288;
People observed to receive a contributory state pension or a widow(er)’s pension: 43 485;
Finally, to the remaining individuals, a number of data are added which allow the calculation of the different labour market states. These data are the computed CE and Tús caps; all cumulated benefit durations and counts until date of Tús or CE eligibility (including time spent in the Live Register, i.e. on registered unemployment); Back to Education Allowance (38 015 individuals involved during at least a month within the relevant period); JobBridge (18 101); employment with (52 251) and without support (377 982); JobPath (102 889); JobsPlus (10 739).
After this processing, 466 152 individuals are left in the data (see Figure 5.2 for a visual representation), for a total of 44 750 592 person-month observations (i.e. lines in the “grid” mentioned above). As a final step, since in the same month a person can be observed in two different – although compatible – states, rules are applied to select the prevailing one. The following lexicographic ordering (which gives preference to one state over another if they both appear in the same month) across states is applied, which gives precedence to CE and Tús, as they are the focus of this report:
8. Other (residual state, applying whenever an individual is not employed nor engaged in any of the supporting schemes covered in the analysis)
Source: Author’s representation.
What has been described so far is the dataset that represents the starting point for the analyses in this chapter. For each part of the analysis, however, individuals are then selected based on year of eligibility start, so that individuals starting eligibility e.g. in 2012 can be considered as the “2012 entry cohort”. Once this selection is done, the starting time point of the analysis becomes the month of eligibility start. This becomes time t, and all individuals histories are aligned starting from this moment. Months observed in the dataset before this moment are dropped, and the subsequent points in time considered in the analysis are chosen starting from the entry month.
5.3.2. The choices made to build the data for the analysis point towards a few possible recommendations to improve the current data availability
Before moving to the results of the analysis, it is worth pointing out a few data-related caveats to bear in mind, together with possible recommendations for improving the current situation.
In building the dataset for the analysis, information was frequently available on the same benefit or scheme participation in more than one data source. For example, information on JobsPlus was available both in the benefits and in the social welfare payment datasets. Information on JobPath assignments was provided in a specific JobPath history dataset but was also available as part of a dataset on referrals. Data on receipt of BTEA was extensively available in the social welfare payments dataset, but also sparsely appeared in the benefits and referral datasets. Piecing together these different sources of information was never easy, as very rarely the two sources were providing coherent information. The working solution to this issue was to identify what could be considered the “main” source of information on each scheme or benefit, in terms of coverage and likely reliability of data. This leaves however an open question on comparability between different datasets, reliability of the information provided, and in general, an impression that availability of more comprehensive meta-data would help interpreting the information found when exploring the data sources. As the source datasets are built for operational purposes, developing comprehensive meta-data documentation will assist in re-using them for analytical purposes. Furthermore, the development of a longitudinal database that can consider a variety of data points and use them to come to some determination of labour market status, would alleviate many of the data construction challenges outlined in this section.
In order to test the coherence and consistency of the data used in the analysis, several checks were done throughout the data construction phase. Moving from the rules that in theory apply to participation in different schemes and overlapping receipt of benefits, these checks were aimed at identifying the rules that could be implemented in the construction of the data to identify the relevant cases to be considered (or discarded) for analysis. This was especially useful whenever the information available was not complete – for example, for data on social welfare payments and earning data on employment, which are only available on an annual basis. Despite the carefully selected allocation rules set in place, the annual nature of part of the data limits the precision that can be used in properly allocating statuses over the different points in time. This is one of the main shortcomings to be kept in mind in carrying out the analysis of sequences. More detailed information on the specific time periods these benefits/employment records refer to would allow for more reliable analysis.
Moreover, it is worth noting that employment data not only do not contain information on exact periods, but they also do not provide details on hours worked. As a consequence, the analysis will likely overestimate the amount of employment without support in the analysis of sequences.
Finally, as already highlighted, the choice of states considered was also largely due to data availability. Most notably, it was impossible to retrieve information on support received from Local Employment Services, as well as on many other possibly relevant schemes, for which information was very sparse (e.g. for JobClub).
5.3.3. Data describe a diverse pool of individuals in the sample
Individuals in the sample vary across several characteristics, both demographic and related to the episodes that make them eligible for either CE or Tús; these characteristics are considered in the analysis to identify differences in patterns across different population groups.
As shown in Table 5.3, 41% of the individuals under analysis are women, 80% of the sample is composed of Irish nationals, 13% of other EU27 nationals, and 4% from the United Kingdom. Seventy-two percent are single when starting the eligibility period, while most of the others are married (around 28%), with very few widowed. More than one-third are below 30 years of age at the start of the first episode, while 14% are aged 50 and above.
Individuals in the sample are then further classified based on several characteristics, such as the reason why they qualify for CE or Tús (what is labelled “claim type”), and previous history in registered unemployment (on the Live Register). These characteristics are defined at the time of start of the first qualifying episode for CE or Tús in the period under consideration – i.e. the first eligibility spell starting between 2012-18. Individuals in the final sample may have more than one eligibility spell starting in this period; indeed, 73% of those in the sample have one spell start only, while 21% have two, 5% have three and 1% more than three. However, since this analysis aims to follow individuals since their first entry into eligibility, only the characteristics attached to the first eligibility spell are considered. When both a qualifying episode for CE and one for Tús start at the same time, the information linked to the CE episode prevails.
As far as claim type is concerned, 67% of individuals in the sample qualify from the Jobseeker’s path (Jobseeker’s Allowance, Jobseeker’s Benefit or Jobseeker’s Benefit Credits), with a further 19% of episodes linked to casual jobseeker claims. Some 10% qualify from family-related payments (One Parent Family Payment). Information is missing or referred to other categories for 4% of eligibility starts.
Individuals in the sample also vary by summed duration on the Live Register (LR) prior to the qualifying episode – i.e. duration of past registered unemployment. Live Register duration comprises periods with i) means-tested Jobseekers Allowance, ii) social insurance contributions-based Jobseekers Benefit, iii) credited social security contributions, and iv) receipt of part-time unemployment payments (casuals). Around 10% of individuals in the sample begin the qualifying episode with no previous time on the LR. 20% have been on the LR for up to a year before, 14% for a period between 1 and 2 years, and 19% between 2 and 5 years. Nearly 30% of episodes do not have information on previous duration on the LR. Individuals becoming eligible under family-related claims tend to have less previous duration on the LR, with around 45% of the relative sample having no past episode, against only around 10% of those qualifying from the jobseekers’ path, irrespective of whether they come from casual claims or not.
As a way to introduce the analysis of the paths walked by long-term unemployed individuals in Ireland, the sample of the 253 578 individuals who became eligible either for CE or for Tús between January 2012 and December 2015 is drawn. They are followed for four years at 12-month intervals from eligibility. In other words, the status of a person who entered eligibility, for example in March 2014, is observed also in March 2015, 2016, 2017 and 2018. Although entry into eligibility (as well as any other state) is hence observed on a monthly basis and month-by-month changes can be observed, doing so would imply the comparison of 60 points in time (12 months in each year multiplied for the five-year length or the interval we are considering) with nearly unreadable graphs. Averaging over four different entry years – and hence over four different time spans to observe the individual paths’ evolution – makes then this descriptive analysis less dependent over specific features of the business cycle.
Figure 5.3 displays entries by month, to check for the existence of any seasonality in eligibility. Entries span rather evenly across months, with peaks in January, May, June and possibly in September, and a clear fall in December.3
Source: Calculations based on Department of Social Protection (DSP) data.
Figure 5.4 provides with a visual representation of individuals’ status across time. Each vertical bar sums to the 253 578 individuals who became eligible between 2012-15 and is normalised to 100%. Further, it is split proportionally according to the status observed. The bar at the extreme left (labelled t) represents the conditions observed at entry into eligibility, while those on its right represent the states observed one (t+1) to four (t+4) years later.4
At entry, 68% of eligible individuals combine eligibility with no other social protection provision or employment state. Another relevant share (28% of entrants) either becomes employed (with or without support) in the same month in which it enters eligibility or is already so. Employment is indeed compatible with the eligibility for CE or Tús under certain conditions – e.g. in cases of casual claims for CE, or where the person self-refers under OFP or JST for Tús.
The main message carried by Figure 5.4 emerges when the five bars are compared dynamically. Indeed, the share of those who entered eligibility without any kind of support or a job (“Other”) almost halves after four years, from 68% to 35%. This evolution is paralleled by the increase in the number of eligible individuals who get a job without support: their share jumps from 8.1% to 46.6% over the same time span. Transitions to CE or Tús peak at 3.7% and 1.9% after three and one year respectively, and fall smoothly afterwards, possibly after the maximum allowed stay in the programme is reached. After four years, the other states represent 14% of the initial eligible population, with EWS at 7.5% and JobPath at 5%.
On the one hand, this picture suggests that a good deal (more than 54%) of those who become eligible to CE or Tús are able to get some form of employment (in a large majority of cases without any public support). On the other hand, transitions to the public scheme they have become eligible to (i.e. CE and Tús) represent a small fraction of the initial population. After four years since entry into eligibility, 35% of the initial population appears as “other”, i.e. either eligible or not eligible to CE or Tús, but without benefitting from any kind of supporting scheme or working activity.
Note: Each vertical bar sums to the 253 578 individuals who became eligible between 2012 and 2015 and is normalised to 100%. They are split proportionally according to the state observed. The bar at the extreme left (labelled t) represents the conditions observed at entry into eligibility, while those on its right represent the states observed one (t+1) to four (t+4) years later.
BTEA: Back to Education Allowance; CE: Community employment; EWS: Employment with support; EWoS: Employment without support. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis.
Source: Calculations based on Department of Social Protection (DSP) data.
The figures below break down the numbers set out above by demographics (gender, nationality and marital status at first observed eligibility, Figure 5.5), two broad measures of experience (namely age and duration in the Live Register at first observed eligibility, Figure 5.6) and eligibility claim type (Figure 5.7).
5.4.1. Demographic breakdowns are consistent with aggregate dynamics
The aggregate dynamics described above broadly hold in each panel, however, some differences emerge in terms of levels. Men, for instance, although entering eligibility as employed less frequently than women (26% vs. 32%), fill the gap in less than three years, and after four years display a higher share of employed persons (56% vs. 51%: Figure 5.5, Panels A and B). A similar trend is apparent also for the Irish nationals vs. the non-nationals (27% vs. 30% at entry, 55% vs. 51% four years later: Panels C and D), while the married persons display a higher share of employed both at entry (37% vs. 24%) and in the following years (60% vs. 52%: Panels E and F). In all cases these differences mainly mirror into the share of “Other” (eligible or not, but without supports or employment), while differences in terms of the other states appear less relevant, also due to their much smaller shares.
Note: Each vertical bar sums to the 253 578 individuals who became eligible between 2012 and 2015 and is normalised to 100%. They are split proportionally according to the state observed. The bar at the extreme left (labelled t) represents the conditions observed at entry into eligibility, while those on its right represent the states observed one (t+1) to four (t+4) years later. BTEA: Back to Education Allowance; CE: Community employment; EWS: Employment with support; EWoS: Employment without support. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis.
Source: Calculations based on Department of Social Protection (DSP) data.
5.4.2. The youngest move to employment at a faster pace
Figure 5.6 considers two proxies of experience (age and years accrued in the Live Register) which provide complementary views on the evolution of the pool of eligible individuals over time. The left panels suggest that the share of individuals who are employed (either with or without support) upon becoming eligible for CE or Tús grows with age, although at decreasing rates (20% for under 30, 33% for the prime-aged, and 34% for those aged 50 or more). However, the dynamics are more pronounced for the youngest. Indeed, while the share of employed for the elderly grows by around 13 percentage points in four years, it almost doubles to 59% for those aged 30-49 and grows by a factor of 2.5 for the under 30. Although potentially positively correlated with age, the number of years of presence in the Live Register (right panels) – accrued at the first observed eligibility period since 2012 – tells a partially different story. The share of those employed at entry into eligibility grows from 30% for those with no duration, to 32% for those with between one and two years. It is instead lower (27%) for those who have been in the Live Register more than five years. The dynamics mirrors the initial levels, with shares of 50%, 58% and 42% for the three groups respectively. Again, such differences mostly mirror into the “other” category.
5.4.3. Casual claimants rarely enter CE or Tús
Figure 5.75 shows that among those becoming eligible as jobseekers (Panel B), 85% (compared to 68% in the aggregate sample) are observed under the “other” state – meaning eligible without any employment relationship or support scheme – 10% appear as EWoS – possibly because they switch from eligibility to employment within the very first month of eligibility6 with the remaining 5% on Tús, BTEA, JobBridge or JobPath. By the end of the observed timeframe, these figures respectively change to 39% (“Other”) and 42% (EWoS), with a 6% on some sort of EWS. Conversely, 100% of casual claimers enters eligibility as EWS by construction (see above). Panel A shows however that such share falls to 14% in four years, with EWoS growing to 66%. Less than 1% of those who entered as casual claimants move to either CE or Tús at any time span.
This initial descriptive analysis of sequences carries three main messages. First, within all subgroups the bulk of the dynamics appears driven by transitions to EWoS. Among the non-casual claims, while the initial share of persons holding a job with no support is as low as 10%, after four years since entry it grows more than fourfold, to 42%. Second, the shares of individuals on CE or Tús appear on aggregate minor, topping 3.7% after three years since eligibility for CE, and 1.9% after one year for Tús; even when focusing on the most typical claimants, i.e. non-casual claimants, such shares only mildly grow to 4.4% and 2.3% respectively. Third, the subgroups displaying the lowest initial shares of employment, often show the most pronounced transitions to such state; this holds true for men vs. women, Irish nationals vs. non-nationals, the young and prime-aged as compared to older individuals.
The analysis portrayed above meets two major limitations. On the one hand, it does not exploit a valuable feature of the data, i.e. that unemployed people can be followed over time individually, while so far the data have been used purely as repeated cross-sections. On the other, even when the sample is broken down by subgroups, composition effects cannot be excluded, which means, for instance, that even when the analysis is run separately by gender, the observed dynamics are at least partially explained by age, because, e.g. women are on average younger than men.
The following sections will try to overcome such limitations, first through an analysis of personal trajectories, and then through duration analysis.
Note: Each vertical bar sums to the 253 578 individuals who became eligible between 2012 and 2015 and is normalised to 100%. They are split proportionally according to the state observed. The bar at the extreme left (labelled t) represents the conditions observed at entry into eligibility, while those on its right represent the states observed one (t+1) to four (t+4) years later. BTEA: Back to Education Allowance; CE: Community employment; EWS: Employment with support; EWoS: Employment without support. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis. Time spent on the Live Register captures duration of past registered unemployment.
Source: Calculations based on Department of Social Protection (DSP) data.
Note: Each vertical bar sums to the 253 578 individuals who became eligible between 2012 and 2015 and is normalised to 100%. They are split proportionally according to the state observed. The bar at the extreme left (labelled t) represents the conditions observed at entry into eligibility, while those on its right represent the states observed one (t+1) to four (t+4) years later.
BTEA: Back to Education Allowance; CE: Community employment; EWS: Employment with support; EWoS: Employment without support. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis. The category of casual claims includes individuals who became eligible through unemployment claims of a specific type, under which jobseekers can work for up to three in seven days and receive an unemployment payment for the remaining days. Jobseekers’ claims are related to individuals becoming eligible through receipt of Jobseeker’s Allowance or Benefit, but not on a casual claim. Family-related claims (i.e. claims related to OPFP) not shown here.
Source: Calculations based on Department of Social Protection (DSP) data.
This section provides an overview of personal trajectories of the individuals in the sample. It looks more in detail into individual sequences, considering the specific order of states, combinations of states over time, as well as their timing. The depiction of trajectories can offer a more comprehensive narrative on how specific patterns can lead to better or worse outcomes over time, and provide a good introduction to the analysis of more specific events – e.g. participation in CE or Tús – that will be covered in the next chapters.
As in section 5.4, to simplify interpretation, the state of the individuals under analysis is observed at yearly intervals after the moment they became eligible to either CE or Tús. Given the focus on the analysis of order and timing of different states, in this first part of analysis it is particularly relevant to be able to compare sequences of the same length. To ensure a long enough observation period, coherently with the previous section, the analysis focuses on those who became eligible either to CE or to Tús between January 2012 and December 2015. Individuals are then observed at 12, 24, 36 and 48 months after the month of start of eligibility.7 In practice, individuals are followed at yearly intervals from the moment they become eligible, up till four years after this date; they are therefore observed at five different time points.
The analysis presented in this section will provide an overview of the individual trajectories. Sequences can be described in many ways; each figure will look at them from a different perspective.
Figure 5.8 shows the sequence index plot of the sequences observed in the data. Sequence index plots use line segments to show how individuals move between states over time. Changes of colour represent changes in state.
Note: BTEA: Back to Education Allowance; CE: Community employment. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis.
Sequence index plot of the sequences observed in the data for the 253 578 individuals becoming eligible between 2012 and 2015. Sequence index plots use line segments to show how individuals move between states over time. Changes of colour represent changes in state. The Y axis shows one line per individual. The X axis represents time points from the moment individuals become eligible; t represents the conditions observed when becoming eligible, while the other points represent the states observed one (t+1) to four (t+4) years later.
Source: Calculations based on Department of Social Protection (DSP) data.
Several interesting patterns emerge from the individual trajectories in Figure 5.8. First, a large share – 17%, corresponding to the homogenously red rectangle at the bottom of the graph – of individuals are never observed participating in any of the schemes considered, nor in employment. This implies that the targeting of measures could be improved to be more inclusive. Second, a fairly large share of individuals (16%) become employed without support without ever participating in any support scheme, even after prolonged unemployment spells. These are shown in the chart by rows starting with a red spell (i.e. the residual state of non-participation in any scheme), and then become yellow (EWoS). Third, 19% of those becoming eligible are already employed with some form of support – most likely, under casual claims; more than half of these individuals then move onto being employed without any support. Finally, around 60% of individuals are employed without any type of support at some point over the time span considered.
A different perspective is provided by Figure 5.9, which shows how many times individuals are found in the eight different states over the five years considered. The figure clearly shows that a vast majority of individuals never participate in any of the schemes considered in the analysis.
Some 93% of individuals never participated in CE; 3% had a one-year participation. The same share of individuals never participated in Tús; 6% participated once, while a small share – less than 1% – had two Tús episodes. Only 4% of the individuals in these entry cohorts have been recipients of BTEA, and only 2% of JobBridge while 12% have been at some point assigned to JobPath.
The situation is different when considering employment with support: around one-third of individuals shows participation in these schemes. Despite a lack of involvement in the schemes taken into account, most individuals in the cohort register spells of employment without any support over the time period considered. Some 4% appear as always in employment and 41% are never employed. The others are relatively equally distributed across number of years of employment.
Note: BTEA: Back to Education Allowance; CE: Community employment; EWS: Employment with support; EWoS: Employment without support. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis.
Source: Calculations based on Department of Social Protection (DSP) data.
Figure 5.10 goes into more detail on participation in the two main schemes analysed in this report. As mentioned above, 7% of individuals participated in CE; 2.5% participated in the scheme for at least three years in a row. Overall, only 1% of the sample participated in both Tús and CE at some point over the period considered. Only 0.7% of the sample – or less than 1 700 individuals in the 2012/15 cohort – went directly from Tús to CE. This represented 10% of Tús participants, while 9% of CE participants were on Tús before starting CE.
Note: CE: Community Employment. The figure shows: (i) participation in CE, (ii) participation in Tús, (iii) participation in both, and (iv) participation in the specific Tús-to-CE sequence. The latter is shown both overall, and as a share of those who participate in Tús – and then move to CE – and those who are in CE – and were in Tús right before.
Source: Calculations based on Department of Social Protection (DSP) data.
5.5.1. Individual trajectories vary significantly by age
Personal trajectories vary a lot depending on age. As shown in Figure 5.11 and Figure 5.12, younger people tend to participate less in CE, with only 4% of individuals below 30 having been on a CE scheme in the period considered. On the other hand, as expected, a relatively higher share has been in receipt of BTEA (7%) and JobBridge (3%). Those in their prime working life, aged 30-49, are more likely to be in some form of employment with support (38%) during the observed timeframe, while the share is the lowest among young people. This is partly linked to the higher presence of casual claims among those above 30 (around 24% of individuals, against the 13% among the youth). The highest share of participants in CE schemes is found among those aged 50 or above: 15% of them is indeed observed at least once in CE over the period under scrutiny. Similarly, LTU individuals aged 50 or more are also slightly more likely (9%) to be observed on Tús placement.
The highest frequency of employment without support is found among those aged 30-49 (62%), followed by younger people (59%), while only 50% of those aged 50 or above registered unsupported employment.
Note: BTEA: Back to Education Allowance; CE: Community employment. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis.
Sequence index plot of the sequences observed in the data for the 253 578 individuals becoming eligible between 2012 and 2015, and for whom information on age upon becoming eligible is available. Sequence index plots use line segments to show how individuals move between states over time. Changes of colour represent changes in state. The Y axis shows one line per individual; while the number of individuals varies across age groups, the Y axis is rescaled to show percentages, to facilitate comparison across groups. The X axis represents time points from the moment individuals become eligible; t represents the conditions observed when becoming eligible, while the other points represent the states observed one (t+1) to four (t+4) years later.
Source: Calculations based on Department of Social Protection (DSP) data.
Note: BTEA: Back to Education Allowance; CE: Community Employment; EWS: Employment with support; EWoS: Employment without support.
Source: Calculations based on Department of Social Protection (DSP) data.
5.5.2. Individual trajectories also vary depending on the path through which individuals became eligible
The typical sequences observed are also very different depending on the type of claim for the first qualifying episode, and on whether they were casual claims or not. Indeed, Figure 5.13 shows that individuals becoming eligible through a casual claim have a strong attachment to the labour market; while by definition they are all found in employment with support when starting to be eligible, 78% of them experience employment without support over the period considered, showing they successfully manage to progress towards more stable forms of employment. Of those who qualify through other types of claims, only around 55% register unsupported employment.
On the other hand, Figure 5.14 suggests that overall, individuals entering eligibility through casual claims show very low participation in every scheme considered in the analysis apart from employment with support (1% in CE, Tús and BTEA, and nobody in JobBridge), while those from jobseekers’ path (i.e. qualifying through receipt of Jobseeker’s Allowance of Benefit) show much higher rates. For the latter, Figure 5.13 also shows a relatively quick progression of jobseekers to some form of activation; within a year from eligibility, almost half of the jobseekers are either employed or participating in one of the support schemes considered, including in CE and Tús.
Finally, those qualifying through receipt of OPFP show very low participation in the support schemes considered; 10% are at some point over the period employed with support, while 4% participate in CE, and 3% in BTEA. This can be explained by several reasons: OFP recipients can access student maintenance and fee grants while receiving OFP, hence are less in need to come off the payment. Also, after 2012 a person in receipt of OFP was not allowed anymore to get CE payments, while they were until 2012. Finally, this category is not composed of jobseekers strictly speaking, and is therefore less likely to engage in support schemes due to family circumstances and caring responsibilities. Nevertheless, more than half of these individuals are employed without support within four years of becoming eligible.
Note: BTEA: Back to Education Allowance; CE: Community employment. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis. Sequence index plot of the sequences observed in the data for the 253 578 individuals becoming eligible between 2012 and 2015, and for whom claim type of first qualifying episode is one between Casual claims, Jobseekers’ path, One-Parent Family Payment (OPFP). The category of casual claims includes individuals who became eligible through unemployment claims of a specific type, under which jobseekers can work for up to three in seven days, and receive an unemployment payment for the remaining days. The jobseekers path covers individuals becoming eligible through receipt of Jobseeker’s Allowance or Benefit for over a year, but not on a casual claim. Sequence index plots use line segments to show how individuals move between states over time. Changes of colour represent changes in state. The Y axis shows one line per individual; while the number of individuals varies across claim type, the Y axis is rescaled to show percentages, to facilitate comparison across groups. The X axis represents time points from the moment individuals become eligible; t represents the conditions observed when becoming eligible, while the other points represent the states observed one (t+1) to four (t+4) years later.
Source: Calculations based on Department of Social Protection (DSP) data.
Figure 5.14, Panel B also shows that the share of CE and Tús participants is slightly higher among those who have been registered as unemployed for more than five years (9% for CE and 7% for Tús). On the other hand, participation in supported employment is lowest among those who have never been registered as unemployed, with 30% of individuals on such schemes. For all other groups, participation is between 32% and 37%.
Note: BTEA: Back to Education Allowance; CE: Community Employment; EWS: Employment with support; EWoS: Employment without support; OPFP: One-Parent Family Payment. The category of casual claims includes individuals who became eligible through unemployment claims of a specific type, under which jobseekers can work for up to three in seven days and receive an unemployment payment for the remaining days. The Jobseekers‘ path covers individuals becoming eligible through receipt of Jobseekers Allowance or Benefit for over a year, but not on a casual claim.
Share of individuals becoming eligible in 2012 who participate in the different schemes over the time period considered, by claim type of first qualifying episode and months on the Live Register (i.e. duration of registered unemployment) prior to the qualifying episode. Information on claim type is not available for 3% of the sample, info on duration on the Live Register for 34%.
Source: Calculations based on Department of Social Protection (DSP) data.
5.5.3. Individual trajectories of CE participants
This section provides a snapshot of trajectories for individuals who participate in CE, starting from the moment they are first observed on a CE scheme. The purpose of this part of analysis is to observe typical patterns in CE participation, and the transitions towards subsequent states in the sequences. For this reason, the analysis considers only CE episodes that start early enough in the sequence to ensure the possibility to follow individuals for at least four points in time after the first CE participation. This rule of thumb is adopted to cover a period long enough to allow for participation on a three-year CE scheme, while still observing at least two time points after participation. As in the previous section, the analysis covers cohorts that enter eligibility between 2012-15. This ensures the presence of 12 141 individuals in the sample.
Figure 5.15 shows that almost half of the individuals observed starting CE are on a CE scheme for at least three years. Around one fifth appear to be on CE for the whole five-year period considered in this part of the analysis.
One year after being first recorded on CE, 63% of the sample appears to be still on a CE scheme and 17% are employed without support. The share of individuals employed without support grows over time, reaching 45% four years after being first observed in a CE scheme. On the same year, another 6% of the sample is employed with some form of support, while around one fourth of the sample is not employed nor engaged in any other scheme.
Note: BTEA: Back to Education Allowance; CE: Community employment. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis.
Sequence index plot of the sequences for individuals who participate in CE, starting from the moment they are first observed on a CE scheme. Sequences are shown for the 12 141 individuals becoming eligible between 2012/15, who can be observed for at least 5 time points starting from the first CE participation. Sequence index plots use line segments to show how individuals move between states over time. Changes of colour represent changes in state. The Y axis shows one line per individual. The X axis represents time points from the moment individuals are first observed on a CE scheme (t), and then one (t+1) to four (t+4) years later.
Source: Calculations based on Department of Social Protection (DSP) data.
5.5.4. Individual trajectories of Tús participants
Similar to section 5.5.3, this section provides a snapshot of individual trajectories for individuals participating in a Tús scheme, starting from the moment they are first observed on Tús. As before, the analysis considers only Tús episodes that start early enough in the sequence to ensure the possibility to follow individuals for at least four points in time after the first Tús participation. While Tús schemes normally last one year, the five-year time frame is still useful, in that it allows to follow individuals potentially moving on to CE – or sequences of other schemes. As in the rest of this section, this analysis takes into account cohorts that enter eligibility between 2012-15; the final sample is composed of 9 899 individuals.
Figure 5.16 shows that 12 months after being first observed on Tús, more than half of participants are not employed nor engaged in any of the support schemes considered, while 22% are employed without any support. After two years, the share of employed goes up to 30%, reaching 37% after three years and 44% after four. Around 5% of the individuals under analysis are still observed on Tús one year after being observed on the scheme for the first time; this might be connected to scheme participations which have been suspended in between the two time points.
Note: BTEA: Back to Education Allowance; CE: Community employment. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis.
Sequence index plot of the sequences for individuals who participate in Tús, starting from the moment they are first observed on a Tús scheme. Sequences are shown for the 9 899 individuals becoming eligible between 2012/15, who can be observed for at least 5 time points starting from the first Tús participation. Sequence index plots use line segments to show how individuals move between states over time. Changes of colour represent changes in state. The Y axis shows one line per individual. The X axis represents time points from the moment individuals are first observed on Tús (t), and then one (t+1) to four (t+4) years later.
Source: Calculations based on Department of Social Protection (DSP) data.
Some 11% of those initially observed under Tús move on to a CE scheme one year later. Figure 5.17 shows that most of these individuals (around 60%) stay on CE for three years. Some 34% of those moving from Tús to CE are observed as employed without support four years later, while another 5% is employed with some form of support.
Note: BTEA: Back to Education Allowance; CE: Community employment. “Other” refers to persons who are not employed nor engaged in any of the supporting schemes covered in the analysis.
Sequence index plot of the sequences for individuals who participate in Tús and then move on to CE, starting from the moment they are first observed on a Tús scheme. Sequences are shown for the 1 127 individuals becoming eligible between 2012/15, who can be observed for at least five time points starting from the first Tús participation followed by CE. Sequence index plots use line segments to show how individuals move between states over time. Changes of colour represent changes in state. The Y axis shows one line per individual. The X axis represents time points from the moment individuals are first observed on Tús (t), and then one (t+1) to four (t+4) years later.
Source: Calculations based on Department of Social Protection (DSP) data.
Duration analysis concerns all those methods for the analysis of the length of time until the occurrence of some event. Applications in the social sciences are pervasive, including – but not limited to – economics (e.g. time spent in unemployment until one finds a job), education studies (e.g. time spent in education after completion of compulsory schooling), and demography (e.g. time to divorce since marriage).
Three reasons lie behind the decision to undergo a duration analysis. First, the duration analysis solves the problem of the so-called dynamic selection: the composition of any given sample changes over time, e.g. because best connected individuals find a job earlier – hence the “surviving” sample evolves towards one with lower quality of connections with respect to the original sample – or because students with better parental background survive longer at the university, possibly to completion. Whenever these variables are not observed – labour market connections, parental background – any analysis is at risk to return biased results, due to the fact that the sample evolves over time towards one which is on average different from the one originally chosen for the analysis. Duration analysis techniques control for dynamic selection. Second, duration analysis is informative on the so-called “duration dependence”, i.e. on the fact that in some cases it is persistence in a state per se that determines a higher or a lower probability to move to a different state. Third, through formal modelling durations and estimation techniques described in the Technical Report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[1]), composition effects due to gender, age, marital status, nationality and years of registered unemployment are excluded.
In the following, this approach is applied to the study of durations of i) periods of eligibility to CE or Tús, until the eligible individual moves to another of the eight states classified in the present chapter, e.g. CE; ii) CE episodes; iii) Tús episodes. What changes with respect to the above is that instead of fixing ex ante the intervals at which the (potentially new, but not necessarily) state of an individual trajectory is observed, an individual is followed as long as the chosen state (eligibility, CE, Tús) persists, and until some event (a transition to another state) occurs. In the following, duration models are estimated assuming that transitions out of the current state occur discretely at monthly intervals, i.e. that in each month of the process under scrutiny (CE/Tús eligibility, CE, Tús) observed individuals undergo some probability to move out of the current state, and that such probability is constant within the month. In addition, the analysis distinguishes multiple destination states, the specific list of which depends on the process under scrutiny (i.e. CE/Tús eligibility, CE, Tús). So, when CE/Tús eligibility is under scrutiny, each individual in the sample is allowed to persist into the eligibility condition, move to CE, to Tús, to EWS, to EWoS or to any other state. When the duration of CE (Tús) is instead analysed, possible exit states are – beyond persistence – Tús (CE), employment (EWS or EWoS), or the state which has been labelled “other”, and that includes eligibility and non-eligibility to CE/Tús. As anticipated above, all estimated models include controls for gender, age, nationality and marital status. Models of eligibility to CE/Tús also include years accrued in the Live Register. Models of CE/Tús include the initial type of claim. Further details, in particular on how the structure of duration dependence is modelled, are provided in the related Technical Report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[1]).
5.6.1. Duration of eligibility spells is hump-shaped
To study the duration of CE/Tús eligibility spells, and consistently with the previous analyses, only the individuals who became eligible to CE or Tús between January 2012 and December 2015 are retained in the data, under the condition, however, that they were not receiving any kind of support nor were holding a job. The latter aims at focusing on a homogeneous subsample with a comparable incentive to exit from the current state of long-term unemployment. Based on the descriptive analysis above, this subsample is very close to that of jobseekers’ claims. The analysis distinguishes among terminations due to i) CE, ii) Tús, iii) EWS, iv) EWoS, v) any other state. Alternatively, when individuals are not observed moving to any of the states above, they remain into the initial state, which is CE/Tús eligibility. Control variables in the model include age, gender, marital status, nationality, years accrued in the Live Register, calendar years (to account for the business cycle) and calendar quarters (to account for seasonality). Figure 5.18 plots the probability of an individual eligible to CE/Tús to exit to different states for a benchmark profile, i.e. a male, married, Irish national with mean age (37 years old), no experience accrued in the Live Register and experiencing the event of interest (one of the exits) in the first quarter of 2012.
Note: CE: Community employment. Benchmark individual: male married Irish national with mean age (37 years old), no experience accrued in the Live Register and experiencing the event of interest in the first quarter of 2012. T1,...,T6 are four-month (i.e. quarterly) time intervals, T7 is elapsed time since 25th month onwards. Estimates are based on a discrete-time competing risks model. Dotted lines represent 95% confidence intervals.
Source: Calculations based on Department of Social Protection (DSP) data.
In Figure 5.18 the horizontal axis represents the time (measured in four-month intervals) since entry into eligibility, while the vertical axis the estimated probability to exit to CE, Tús, EWS (employment with support, Panel A) or EWoS (employment without support, Panel B).8 The reader should first focus on the shape of the four lines. Irrespective of the type of exit (except Tús, described below), the likelihood grows until the second four-month interval, and then falls, exhibiting a negative duration dependence (meaning that the longer one stays into the initial condition, the lower is the probability to find a position in CE or any kind of employment). In other words, after an initial period in which people “learn how to do” and increase their chances to exit from long-term unemployment, the probability of exiting long-term unemployment gradually falls for reasons related to the time elapsed (e.g. loss of human capital or of motivation).
For Tús the argument is partially different: the profile is flatter, grows and reaches a plateau until the sixth period (i.e. two years since entry into eligibility), and then falls. In terms of levels, the probability to exit towards some EWoS is always higher than to other states; this is consistent with the descriptive analysis above. Exits to CE and EWS are never statistically different. On the contrary, those to Tús are always lower than to CE and to EWS.9
Individual characteristics like gender, age and the other variables included in the model (see above and the Technical Report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[1])), can move the plots commented above up or down, but cannot modify its shape. For instance, a woman – instead of a man – with the same characteristics of the benchmark case assumed above, will exhibit a plot with the same shape, but shifted up or down depending on the destination state considered.10
Table 5.4 provides a summary view of these “shifting” effects. Variables appear on the rows, the different destination states in columns. While studying why different individual features differently affect the probability of exit to the four destination states goes beyond the scope of the present analysis, it is anyway worth noting that such heterogeneity emerges clearly: for instance, the second row (”Women”) of the column under “CE“ has a “+“. This means that women – other things being equal and with respect to men – enjoy a higher probability to move from the initial state (CE/Tús eligibility) to CE at any elapsed time since entry into CE/Tús eligibility. The next column (Tús), however, has a “-“, meaning that women have a lower probability with respect to men to move to Tús. Using models that allow to control for such heterogeneity is hence crucial.
5.6.2. For CE and Tús episodes, spikes at most common durations emerge
The present subsection replicates the analysis in Section 5.6.1 on CE/Tús eligibility, on CE and Tús episodes. In this case all the CE and Tús episodes started from 2012-15 are drawn from the data, irrespectively on whether the corresponding qualifying periods are observed or not. The specification described for duration of eligibility is kept as similar as possible, for the sake of comparability. Nonetheless some changes have been necessary:
Beyond persistence in the current state (CE or Tús, depending on the specific case), the exit states are reduced to three, namely the state we labelled “Other” in the construction of the data (i.e. being not employed nor engaged in any of the supporting schemes covered in the analysis), employment (with or without support), and a residual state indicating participation in one of the other schemes (not shown in the graphs), namely Back to Education Allowance, JobBridge, JobPath and Tús (when duration in CE is studied) or CE (when duration in Tús is under scrutiny).
The duration dependence structure is now specified in bimesters and not in four-month intervals. This is aimed to highlight exits in some well-defined turning points, i.e. when CE or Tús reach the maximum statutory duration for a subset of the sample under scrutiny.11
Information on previous presence in the Live Register is dropped, as its inclusion makes estimation troublesome. On the contrary, a series of dichotomous variables are included to capture the type of initial claim.
Figure 5.19 and Figure 5.20 plot the duration dependence profile as in Figure 5.17, for CE and Tús episodes respectively. As for exits to a condition of being not employed nor engaged in any of the supporting schemes (Panel A in both figures) and to employment (with or without support, Panel B), the two figures display very similar patterns. They suggest that there are “typical” durations of CE/Tús episodes, either driven by institutional features (typically, the maximum duration of the programme) or recurrent behaviours (e.g. sponsors of CE episodes typically propose job periods the duration of which is a multiple of a year). Profiles are flat before these typical durations, and spikes appear thence. At this stage of the analysis, it is hard to say whether this “lock-in effect” is more a matter of a choice, or of lack of alternatives. However, the fact that at spikes fewer of the beneficiaries move to some form of employment than to the residual state, suggests that the latter may be the right interpretation.
Note: Benchmark individual: male married Irish national with mean age (46.3 years old), reaching CE as a jobseeker and experiencing the event of interest in the first quarter of 2012. B1,...,B21 are bimesters. Estimates are based on a discrete-time competing risks model. Dotted lines represent 95% confidence intervals.
Source: Calculations based on Department of Social Protection (DSP) data.
Indeed, profiles are rather flat at quite low probability levels (around 1% for CE and even lower for Tús) during most of the time spent in CE or Tús, but at yearly durations, when exits jump up. In the case of CE, exits to employment (no employment nor engagement in any of the supporting schemes) reach 3% (7%), 3% (4%) and 10% (15%) at one-, two- and three-year durations respectively. For Tús – which lasts one year for a vast majority of individuals – the spike reaches 6% and 83% after one year for exits to employment and the “other” state (no employment nor engagement in any of the supporting schemes) respectively. Afterwards, almost no one continues to receive Tús and the duration profile drops almost to zero.12 In both the cases of CE and Tús durations, the profiles of exits to Tús/CE, BTEA, JobBridge and JobPath are never statistically different from zero at the 5% confidence level, hence are not shown.
Note: Benchmark individual: male married Irish national with mean age (39.0 years old), reaching Tús as a jobseeker and experiencing the event of interest in the first quarter of 2012. Dotted lines represent 95% confidence intervals. Vertical axes have different scales to improve readability of the two panels.
Source: Calculations based on Department of Social Protection (DSP) data.
Finally, Table 5.5, which can be read in the same way as Table 5.4, shows again the relevance of taking heterogeneity of individuals into account. Broadly speaking, the Table shows that younger individuals and women stay longer both in CE and in Tús. Married people and casual claimants move to employment more frequently than singles and jobseekers respectively, and they go (back) to a state of no employment nor engagement in any of the supporting schemes less frequently. In terms of nationality, no clear pattern emerges.
This chapter draws from the wealth of information provided by DSP to study the main patterns in participation in employment and ALMPs over a five-year period for individuals eligible for CE and Tús, mostly long-term unemployed.
Combining datasets on CE and Tús episodes and eligibility with demographic data and information on DSP benefits, social welfare payments, earnings and social security contributions, the analysis showed that four years after becoming eligible to either CE or Tús, almost half of the individuals are employed without any form of support. Around 60% of individuals are observed in this state at some point over the five years considered. On the opposite side, 17% of individuals are never observed participating in any of the schemes considered, nor in employment. Four years after becoming eligible, 35% of eligible individuals are found in this state, irrespective of what happened in between the two years. Only 7.5% of eligible individuals participate in CE over the time period considered, while for Tús the share is lower at 6.6%. Individuals aged less than 30 participate relatively less in CE, and more in BTEA or JobBridge. Those aged 50 or more are more likely to participate in CE (15%) or Tús (9%) at least once. Although entries into the scheme are not that frequent, persistence in CE is high: 50% of the individuals who start a CE scheme are still observed there after three years and 20% after five years.
With this evidence, this chapter sets the scene for the evaluation of the impact of participation in two public works programmes in Ireland, CE and Tús, that will be presented in the next chapters.
Reference
[1] 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.
Notes
← 1. Out of all the CE (Tús) episodes starting since 2012, 30% (25%) begin during a non-eligibility period.
← 2. Time boundaries are dictated by raw data availability. The analysis starts from the reference period for eligibility considered for the evaluation of CE and Tús, i.e. 2012-18. Year 2019 is added to the panel as a way to have a longer period for observing states in what can be considered as a purely “post-participation” year. The decision not to expand the panel further in time is due to the fact that 2020 saw the start of the COVID-19 pandemic, which led to the introduction of different programmes; these schemes might not be fully captured by the data this analysis relies on, as all post-2018 information was collected to have “outcomes” relevant for the CE/Tús evaluation; for this reason, the period from 2020 onwards was not taken into account in this chapter. Tús episodes starting since 2019 are indeed not available, what leads to a fall in the frequency of transition towards such state. However, as will be seen below, they represent a minor share of the overall observed dynamics, with limited potential to affect the narrative of the findings.
← 3. Entries by calendar year are not reported as this exercise would be biased. Indeed, if a person enters eligibility, for instance, in 2012, then loses it and becomes eligible again before the end of 2015, s/he appears in the data as entered in 2012 only. As a consequence, a breakdown by calendar year would mechanically show a fall in the number of entries, which nonetheless would have no economic meaning.
← 4. For a right interpretation of Figure 5.4 – and of the others alike that follow – one should bear in mind that people are not followed over time individually, but only as a group. This implies that people observed in a given status in two different periods are not necessarily the same. For instance, the individuals observed as employed without support in period t+1 are not necessarily the same observed as employed without support in period t+4. Figure 5.4, hence, gives an idea of how the population of entrants into eligibility evolves over time, without implying that the aggregate pattern should be true for each single entrants as an individual.
← 5. Paths originating from family-related forms of support represent 7% (17 838) of the individuals in our sample. Their trajectories are similar to the jobseekers’, and hence aggregated to the latter. For another 3% (7 657) of the eligible individuals the type of initial claim is missing.
← 6. Another explanation is that the choices we made to allocate the employment periods across months were not always correct: see above.
← 7. The selection of yearly intervals implies the loss of information on short-term movements that might happen within the year. For the cohort becoming eligible in 2012, for example, this choice implies the loss on information on CE participation for around 8% of individuals with CE spells, and around the same share for Tús participation; by definition, all these spells would be shorter than the standard one-year scheme. While in theory it would be possible to fully exploit the monthly structure of the dataset, this would produce an explosion in the number of possible sequences observed in the data, making it very difficult to interpret the evidence produced. With the choice of yearly intervals, the analysis allows to retain a strategic view of the main patterns observed. The Technical Report (OECD/Department of Social Protection, Ireland/European Commission, Joint Research Centre, 2024[1]) shows examples of similar analysis with high-frequency intervals.
← 8. The choice to use two different panels is then dictated by the different magnitudes of the estimated exit probabilities, much higher for exits to EWoS.
← 9. The reasoning in terms of levels is related to the specific benchmark individual chosen, as the vertical position of the time profiles are affected by control variables; the Technical Annex provides more details on this. In addition, it also depends on that we look at 95% confidence intervals, whose width depends on the level of significance – the higher the level of significance, the larger the confidence intervals.
← 10. This happens by construction, as the duration model is specified with no interactions between the control variables like gender, age, etc. and the time dummies which define the spline: see the Technical Report.
← 11. For the sake of completeness, the reader should be aware that the first bimester actually covers the first three months in the data. This is done because a period of 12 months in real-world continuous time (e.g. from 15 April 2012 to 14 April 2013) spans over 13 discrete months in the data (April 2012 would be month 1, while April 2013 would be month 13). In order to allow the specification to visualise the discontinuity at yearly intervals – an institutional feature of both CE and Tús – we need to compare (discrete) months 12-13 with (discrete) months 14-15. The same holds at two- and three-year durations.
← 12. If someone had to pause the placement to take sick leave for example, Tús episodes would be longer than one year.