Finland is among the highest spenders on training programmes for unemployed people in the OECD. These programmes help jobseekers acquire and augment the skills they need to prosper in the labour market. This chapter provides details on the two main training programmes that are available to jobseekers and sets them in context of one other. It then goes on to describe the methods and data used in the following chapters of this report to evaluate the impact of these programmes on individuals’ subsequent outcomes in the labour market. In particular, it describes how an occupational index is constructed to aid insight into how jobseekers move between occupations following training.
Evaluation of Active Labour Market Policies in Finland
Connecting People with Jobs
4. Training for jobseekers and impact evaluation methodology
Abstract
4.1. Introduction
Finland has a well-funded suite of training programmes to support jobseekers. Spending on training programmes for jobseekers, at 0.36% of GDP in 2020, was the second highest in the OECD. A comprehensive set of programmes to enhance the skills of jobseekers and improve their ability to find good jobs is an essential component to a well-functioning labour market. This is particularly important for Finland, which has a relatively high rate of job displacements and lower re‑employment probabilities for older and young workers, blue‑collar workers and the less well educated (OECD, 2016[1]). Training policies to match skills demand and supply are crucial in this context. Knowing which policies work and for whom can help the government to target policies when necessary and ensure they are cost-effective.
The first section of this chapter provides a description of the training programmes available to jobseekers in Finland, statistics on their frequency, information on how they are delivered and on the characteristics of participants. The following section then outlines requirements for conducting an impact evaluation of these programmes and the methodology that is used in this report (Chapters 5, 6 and 7) to estimate the impact of selected training programmes on employment, earnings and occupational mobility. It includes a discussion of the outcomes analysed and the sample time period that is chosen. The final section describes the linked administrative data that are used for the impact evaluation.
4.2. Training for jobseekers
The two primary training offers for jobseekers in Finland are labour market training (LMT) and self-motivated training (SMT). LMT has been a core part of the training offer for unemployed jobseekers for many years, providing jobseekers with vocationally focused courses to improve employment opportunities. LMT offers short programmes which are primarily organised via local employment offices and their organising bodies. The introduction of SMT for jobseekers in 2010 allowed individuals to engage long-format studies to up degree level (enrolment in the formal education system) without any job search requirements and to retain their unemployment benefit for up to two years. Because these courses existed previously and were provided by the Ministry of Education and Culture (OKM), when speaking about SMT, it is less about provision of the courses themselves and more about the financial incentives for jobseekers to participate in existing courses. In this sense SMT has opened up a more financially advantageous route to undertake longer-format education without job search requirements. It has coincided with a fall in LMT participation, as jobseekers take advantage of the new funding available to them (Figure 4.1). While LMT also allows participation in training without job search requirements, but in short format training courses, SMT offered a complementary alternative for longer format education by similarly paying unemployment benefit and removing requirements to look for work (work search requirements are discussed in Chapter 2). However, as of May 2022, legislation introduced work search requirements on SMT participants. Participants are now required to apply for three jobs every quarter. Legislation is currently in parliament to increase that number to four from January 2023, for individuals who already hold tertiary level qualifications. In practice, this removes the possibility of using SMT as a means to complete education whilst remaining inactive in the labour market.
The total number of individuals participating in LMT reached a peak in 2010, at just under 33 000 participants, the year that SMT commenced. Since 2016 LMT participants have plateaued at around 20 000 per year. In the same period, jobseekers participating in SMT have averaged 35 000 per annum. In part this reflects the longer duration of the latter, which means that for the same number of individuals starting training, participation at any point in time will be higher as fewer have completed their training.
Figure 4.1. Labour market training and self-motivated training form the core learning provision for jobseekers
Note: Yearly totals are simple averages of monthly participant volumes. Year 2022 refers to the average from January to August.
Source: Statistics Finland’s free‑of-charge statistical databases, https://statfin.stat.fi/PXWeb/pxweb/en/StatFin/.
4.2.1. Self-motivated training lasts longer than labour market training
The introduction of SMT augmented the ecosystem of training in which it is possible for jobseekers to participate in. SMT supports self-directed courses sought by jobseekers which are organised as part of the general education system provided by OKM. These studies are more directly aimed at finishing longer-format education, whereas LMT offers a route to shorter courses which are more vocational in nature. Although there can be some crossover between the two, they are largely complementary in the types of training that they offer.
Table 4.1 details the underlying composition of the duration of training, to demonstrate how LMT and SMT differ. SMT has a median duration of 341 days, meaning that half of participants spend around a year or longer on their studies and one‑quarter spend 21 months (648 days) or longer. By contrast, LMT is comprised of much shorter training spells. The median duration for programmes starting between 2014 and 2018 is 43 days and only the top 5% of individuals spend a year or longer participating in a training course. A large number of these LMT courses are very short in duration. The duration of an LMT course for the bottom 5% of individuals, when ordered by duration, is only one day in length. The bottom 25% of individuals have a course length of one working week (five days) or fewer. These very short courses may relate to short vocational courses or capture the sitting of an exam relating to a previous longer spell of training. The introduction of SMT does not seem to have altered the composition of LMT courses, from a durational perspective, with the average duration of the latter remaining fairly constant over time.
These differences in terms of length and the types of skill that the two programmes equip jobseekers with may or may not lead to different labour market outcomes, following training participation. The direct impact of duration is likely to manifest itself in a larger lock-in period for SMT, as participants are kept out of active job search for longer. Shorter LMT courses which are more vocational in nature, may have better links to immediate skills demands by employers. Whereas the increased focus of SMT on courses of an academic nature may equip jobseekers with a more in-depth set of academic skills that permits more scope for career changes, either across sector or profession and allowing for greater increases to salaries.
It is important to ensure that both sets of training provision are focussed so that the skills augmented and improved by jobseekers are those that are demanded in the labour market. As outlined in Chapter 2, the Centres for Economic Development, Transport and the Environment support the TE offices to plan and procure LMT courses. This is completed via their forecasting function and anticipation of future local development needs. For SMT, no such co‑ordination exists. Whilst counsellors in TE offices have to agree that the courses proposed by individuals will help to improve their labour market opportunities, it is up to the individual to identify and propose the course and no aggregate skills needs assessment takes place to determine how the needs of employers in the labour market can be met by participants in SMT. The national Skills Assessment and Anticipation exercise (OSKA) in Estonia provides a good example on how to integrate systematic information on occupational demand and supply to ensure well targeted training courses. OSKA is used systematically by the Estonian Unemployment Insurance Fund (EUIF) to guide the provision of training programmes to prevent unemployment. OSKA is also used, alongside the EUIF’s Occupational Barometer, which measures supply and demand of different occupations (and is itself based upon Finland’s Occupational Barometer), to provide guidance to employment counsellors when they refer jobseekers to training programmes (OECD, 2021[2]). Ensuring similar use of Finland’s occupational barometer, when ALMPs are divested to municipalities in 2024, will help to ensure that local training provision is appropriately targeted.
In order to consider LMT programmes that may give rise to sufficient amounts of skill accumulation to affect transitions in the labour market, the main impact evaluation in Chapter 6 utilises courses longer than three months to assess their impacts. This is to avoid incorporation of the very short duration courses in the first instance. Table 4.1 also shows the distribution of duration for the courses longer than 90 days. The mean length of these courses is 219 days (around seven months) in length. Sensitivity analysis is then conducted by relaxing this constraint and reviewing all LMT spells, to determine what difference to results this makes.
Table 4.1. Self-motivated training is much longer than labour market training
Duration of programme participation in days
Note: Average values for all programmes starting in the years 2014‑18, excluding those programmes with missing end dates. Distribution over 90 days taken as an approximation for courses longer than three months in length.
Source: OECD calculations on TEM data accessed via Statistics Finland. Data on jobseekers participating in self-motivated training or labour market training.
4.2.2. Self-motivated training
In 2010, a change in the law enabled jobseekers to retain their unemployment benefit for a maximum period of 24 months while enrolling in SMT (although it is possible for the studies to continue for longer than this period, without the associated unemployment benefit continuing). This opened an educational pathway for jobseekers to continue formal studies in the education system that may have been previously interrupted. The broader question for the introduction of SMT relates to how it interacts with other adult education provision. In particular, adults who wish to participate in further education were already able to access funding from KELA, via the study subsidy (SS) grants. These grants are available to individuals continuing in secondary or higher education. The amount of study grant available is typically lower than the monetary benefits of SMT. For a single individual aged over 18 living alone, the study grant provides a payment of EUR 268, compared to a payment of EUR 742 of labour market subsidy that an individual might receive in an equivalent typical month (21.5 x EUR 34.5 daily rate) if they were registered as a jobseeker. The introduction of SMT then creates a route to education attainment that is more financially advantageous to individuals. This comes with the requirement that the PES considers the studies to be of benefit to an individual’s employment prospects, which acts as a form of targeting mechanism. However, it is important to know whether this introduction of SMT is of benefit to both government and wider society. This will be reviewed in Chapter 7. In addition to the study grants, adults may also be eligible for the Adult Education Allowance which subsidises continuing studies, though as this is dependent on at least eight years of employment history, it has stricter eligibility requirements than the study grants.
The formulation of the SMT training provision means that it is dependent upon the jobseeker to find a suitable education curriculum and successfully apply to it, which is then subsequently reviewed and approved by the TE Office. If the course is deemed suitable for the jobseeker’s development, they can continue receiving their unemployment benefit for a period of 24 months whilst studying. These duration restrictions were relaxed during the COVID‑19 pandemic, for those studies whose maximum periods would be reached between 1 July and 31 December 2020. There is a requirement for monitoring of progress on the studies to continue receiving unemployment benefit, which is conducted by KELA or the individual unemployment funds (which confirm with the participant that they are still undertaking their studies). This is to ensure jobseekers are in the process of acquiring education, because during this period they no longer have to fulfil commitments on job search, as they are deemed to be studying full-time. Jobseekers must be 25 years or older to participate in this training in the mainstream for SMT, though this restriction is relaxed for those participating under the Integration Act (such as migrants accessing basic education).
The requirement for full-time study means that in practice education has to consist of certain features. It must either be to complete a bachelor’s or master’s degree (PhD courses are excluded from the policy), a vocational upper secondary qualification, or preparatory training towards it, or a general upper secondary education for young people. If it is not studying towards a bachelor’s or master’s degree, the training must be for a minimum of 25 hours per week (or the credit equivalent of this time, five credits in the current system). If jobseekers are returning to prior studies, they may only receive unemployment benefit for self-motivated study if at least one year has passed since they last participated in that education.
To date, there has only been one partial impact evaluation of SMT, which was inconclusive in its findings. This means that an important gap exists in explaining how SMT helps jobseekers. Whilst there have been several prior studies on LMT, which demonstrate its use, the same evidence‑building is necessary for SMT.
4.2.3. Labour market training
LMT is open to all jobseekers, though in practice applicants under 20 are unlikely to be selected. The application process for jobseekers is primarily online. Indeed, on the Job Market Finland website (the public employment service’s online vacancy matching platform, see Chapter 3), the application portal for vocational LMT sits directly alongside the main portal for registration of e‑services, demonstrating the primacy of the offering within the public employment services. To be eligible for LMT, people should be either unemployed, or about to become unemployed (although it is possible for working individuals to be admitted). The individual’s need for training is reviewed by the TE Offices.
Labour market training is organised into different tranches, dependent on its function and purpose. One element of LMT is integration training for immigrants, particularly to help them with aspects relating to language and culture. This is not the main focus of this report. The other branch is vocational training to help directly with employment. Within this branch there are four categories. Training aimed directly at obtaining qualifications are financed and steered by OKM. This comprises around one‑third of total LMT offered (Alasalmi et al., 2022[3]). Training not aiming for qualifications, including short-term training for licences or professional authorisations are paid for by the Ministry of Economic Affairs and Employment (TEM) and also accounts for around one‑third of training. Entrepreneurship training is also funded by TEM and is aimed at helping individuals to start new businesses and accounts for around 10% of training participants. The final strand is training that is jointly acquired directly with employers and is co-funded by them. This strand is aimed to integrating training measures directly in business, particularly those smaller and medium enterprises for whom there may be less in the way of corporate human resource functions to assume these roles. This training accounts for around 20% of the total vocational training spending in LMT.
The application process for starting on a LMT course is directed towards empowering jobseekers to make the application themselves. To apply for LMT, individuals use the ‘E‑services’ section of the TE online services. Guidance is given online on how to apply and the participation criteria of each of the courses. A link to LMT training has been added also on the Job Market Finland website, taking individuals to a dedicated website run by TE services, which has over 1 000 courses available across Finland. Courses are found using a search facility. Search can be narrowed using user-defined key words, by professional group, location or publication date. For each course, a description on the course content, eligibility and any additional information is provided. Information is also provided on the location, the start and end dates, the training provider and the number of study places available. The current application process may be subject to some change, as transfer of responsibilities to municipalities is undertaken. Those individuals living in municipalities that are in one of the pilot areas receive guidance on application from their local municipality. The local TE Office still discusses the training needs with the jobseeker to decide on the training referral.
To determine an individual’s suitability for a specific LMT course after an application has been received, the TE Office reviews several criteria. These include the skills and characteristics required in the field of study, student interests and motivation, and prior education, training and work experience. These are used to assess whether the training will address a gap in candidates’ skills, will complement existing experience, and how it will augment the possibilities of finding a job. Selection for course participants is completed using a mixture of the information provided in the application, interviews and aptitude tests. Training can also be preceded by an initial period, to determine participants suitability for continuing in the training.
A team consisting of experts from the employment and economic development services and a representative of the training provider makes selection decisions. An employer representative may also participate in the selection if they are part of planning and funding the training in question. For studies that lead to a higher education qualification and for individually procured training places, the training and education bodies first make an offer of admission for the student, which the TE Office then reviews and makes a final decision on participation. The TE Office responsible for the student selections informs the candidates about its participation decision by letter, typically within about a month after the application period for the training has closed.
4.2.4. Who does training cater to?
This section reviews the underlying characteristics of participants in training, to understand who utilises these training courses. Across a number of characteristics LMT and SMT cater to similar individuals. However, there are some important differences. SMT participants are 25% more likely to be female than LMT. They are 12% less likely to have had either a lower level or manual profession prior to the training spell and 14% more likely to live in an urban location (Table 4.2). Both LMT and SMT cater to individuals that are less likely to be Finnish citizens, or to be a native Finnish speaker, than the average jobseeker.
In addition, Table 4.2 shows that SMT participants have more children on average than LMT participants but also a lower level of labour market attachment. In particular mothers, who are either married or in a partnership, are more likely than the average to participate in the programme. In the calendar year prior to the commencement of training, they spent fewer months in both unemployment and employment, and they earned less, suggesting more time spent out of the labour market altogether.
In addition to LMT and SMT participants, Table 4.2 also provides characteristics of jobseekers who have been paid SS in the year of their unemployment spell. These individuals are much younger, less likely to have children, more likely to be Finnish and have had less labour market attachment in the year prior to their unemployment spell, than their peers participating in LMT and SMT.
Table 4.2. Participants in self-motivated training and labour market training share some similarities and differences
Training participants characteristics
Note: All values are averages of monthly data over the five years 2014 to 2018 inclusive. Variables are extracted from FOLK basic data and relate to the end of the year before the start of the training/unemployment spell. Study subsidy participants identified by looking at unemployment spells where an individual has a study subsidy paid in the same year. Lower worker defined using variable “sose”, defined as the categories “lower level” workers and “manual” workers, with values between 40 and 60. Upper secondary education defined using the variable ututu_aste recording highest education level, using values “3” for upper secondary and “9” for unknown. Rural defined using variable maka, values “M4”, “M6” and “M7”. Single defined using variable sivs, value “unmarried”, so excludes divorced or widowed individuals.
Source: OECD analysis of Statistics Finland databases, using FOLK and TEM datasets.
4.2.5. More detailed course information is available for SMT than LMT
Course information is available for SMT that permits insights into both the level and area of study that education relates to. To do this, it is necessary to link data from TEM to the annual education data from OKM, to look at what courses SMT are related to. However, this method is not completely accurate. Around 12% of TEM SMT records do not match to a corresponding record in the same year. A further 6% of courses which are matched, relate to courses that directly continue on the same course from the previous year (excluding vocational courses, where this is possible), which should not be possible on SMT, unless the participant has been in receipt of a work-related benefit payment. A very small minority (0.2%) of studies are identified as doctoral studies, which are also precluded on SMT. However, even with these caveats, it is possible to build up a broad picture of the type of education pursued by participants.
Figure 4.2 shows that the largest proportion of SMT education is vocational studies at the upper secondary level. In 2018 this education accounted for 64% of all studies and its share has risen over time, from 44% in 2010, at the inception of SMT. Bachelors degrees made up a further 16% of education, completed either at a polytechnic university (13%) or university (3%), whilst Masters degrees accounted for 4%. The proportion studying either Bachelors or Masters has fallen over time as vocational studies have increased.
In terms of the field of study, health and welfare makes up the largest share at around 30% of total course enrolments in 2018, a share that has remained constant from its level in 2010. This focus on the health care sector aligns well with needs of the Finnish economy, where it is frequently cited as the occupation with the highest number of skills shortages (for example, as of the Occupational Barometer published in Autumn 2022, the health care sector dominated the top shortage occupations (TEM, 2022[4]), remaining largely consistent with similar shortages in 2019, see (OECD, 2020[5])). However, without knowing how many of these individuals are re‑training from different sectors, it is difficult to be precise about how much of the labour shortage this education addresses. That said, ensuring that individuals are equipped with the up-to-date skills relevant for the occupation, should still have a relieving effect on skill needs of the bottle‑neck sector.
Services and engineering, manufacturing and construction account for just over 18% respectively, whilst business, administration and law accounts for 13%. Together, these four sectors account for around 80% of all courses, a level which has persisted over the period 2014‑18, rising from 71% when SMT was introduced. This growth has come largely from the services and engineering, manufacturing and construction courses, replacing courses in ICT, arts and humanities, and agriculture, forestry, fisheries and veterinary.
By contrast, there is a lack of information on the content of LMT training available in the Statistics Finland datasets. Figure 4.3 presents information on the distribution of the type of LMT course undertaken by jobseekers. The biggest category of LMT course is “missing” where no information exists on the type of course, this represents around 40% of LMT course participants This is followed by the category “Other”, which presumably covers all courses which are not categorised into either Primary, Secondary, Post-Secondary or General types. Because of the lack of documentation on precisely what each category means (see (OECD, 2023[6]) for more details on metadata), it is challenging to provide more insight into what these courses contain and the precise level of study they pertain to. However, notwithstanding this point, the fact remains that around 70% of courses have either missing information or are included within the “catch-all” category of Other. Furthermore, it is difficult to determine precisely how the courses relate to the different sources of LMT training outlined in Section 4.2.3. An attempt to replicate the methodology of Alasalmi et al. (2022[3]), to incorporate entrepreneurship training type, met with little success, as only 1 500 such courses could be identified (despite attempts to use the same variable to classify it), out of a dataset with 3 million entries. The result of this means it is difficult to conduct anything more than an aggregate analysis of LMT courses.
Figure 4.2. SMT has a large share of vocational upper secondary training and is orientated towards health sector
Note: Data relates to first recorded record of SMT per individual, all subsequent records dropped. 12% of these records cannot be matched with a corresponding entry in the annual education dataset, so are dropped from the analysis. In Panel A, “Other” encompasses: Specialist qualifications, General upper secondary education and Doctoral or equivalent.
Source: OECD analysis of TEM and EDUCATION datasets accessed via Statistics Finland.
Figure 4.3. There is a large proportion of missing values on LMT course type in the dataset of Statistics Finland
Note: Course type defined using the variable “klaji.”
Source: OECD analysis of TEM administrative data accessed via Statistics Finland.
4.3. Training impact analysis technique and data sample
This section describes the requirements for impact evaluation of both SMT and LMT. It then discusses the methodology used to perform impact analysis. The evaluation of the impact of both programmes uses quasi‑experimental techniques alongside linked administrative data, for reasons which will be outlined. The section outlines the different outcomes that are evaluated and the time period used for the analysis before providing detail on how the analysis provides new information on how occupational status is affected by training.
4.3.1. Requirements for impact evaluation of LMT and SMT
In order to evaluate the training programmes, there are a number of factors to consider across the two primary dimensions of data and methodology. The methodology chosen to evaluate the programme also has an influence over what data may be required. If a trial which randomly picks participants has been used to test a programme, then it means that researchers can be sure that the only difference between individuals is the enrolment in that particular programme. In this case, the only data that are needed are those on the outcomes which the research wants to address. For example, to study impacts on earnings, then earnings data are required for participants and non-participants. However, more often it is the case that programmes are implemented first and then evaluated afterwards. This is true for both LMT and SMT. In this instance rich data on individuals are necessary to ensure that participants in a programme are compared only to non-participants that are similar to them. Usually in studies of ALMP, this includes detailed socio‑economic data and previous labour market and unemployment histories. Without these data, there is a risk that any impact evaluation conducted actually just reflects innate differences between individuals, rather than any differences driven directly by the programme in question. Fortunately, Finland has a rich seam of such information to draw upon, which is detailed below in Section 4.4.
On data, the requirements are determined by the outcomes the research intends to examine and the methodology chosen to conduct that research. For LMT and SMT, the primary goal is to establish whether the training courses help individuals to progress in the labour market. This progression could take many forms, including higher earnings, better attachment to jobs, or jobs with more flexible working conditions. There are limitations to the extent that all of these data might be available to researchers. Typically, it is easier to get information which is required by authorities to administer government services, such as the receipt of information on earnings to process tax liabilities. Data on whether an individual has more flexible working conditions, or subjectively “enjoys” their job may only be possible via a survey conducted to elicit this information. The information available for Finland and its use in the analysis is detailed in Section 4.4. When conducting evaluations on ALMPs, usually individual level information is preferred to aggregated information, as it allows a much more detailed examination to be conducted. Particularly where programmes vary over time, or individuals access at different time periods and for different durations, individual level data permit much more precision in looking at these dynamics. Finland possesses a vast array of linkable administrative data in the repository held by Statistics Finland, which are available for researchers to use for analysis. Such data and the ability to link them are key to performing causal impact analysis on policies for which randomisation has not been possible (OECD, 2020[7]).The high-quality individual-level data held by Statistics Finland permit investigations into the dynamics mentioned above.
4.3.2. Quasi‑experimental techniques are used to estimate programme impact
LMT and SMT are live‑running programmes and participation in them is not determined randomly, so it is necessary to analyse their impact using a non-experimental technique. Individuals choose whether to participate in training, which means that it is possible for participants to be different to non-participants in non-trivial dimensions. It is not viable simply to compare labour market outcomes of participants to non-participants, because they may well have experienced different labour market outcomes anyway, in the absence of programme participation. For example, if a training programme for jobseekers attracted young participants and outcomes for participants were simply compared to non-participants, the impact would likely contain some of the effects of their age rather than the programme itself. Younger people tend to have less labour market experience and so earn less than their older counterparts, regardless of training. The results from such a comparison would be subject to “bias”, meaning that the estimates do not reflect the true impact of the programme.
Quasi‑experimental methods need to be employed to evaluate SMT and LMT effectiveness. There exist several different methods for estimating causal programme impacts using non-experimental data (see (OECD, 2020[8]) for a discussion). Some of these methods rest on the type of information that is available to the researcher. OECD (2023[6]) provides more discussion of the outcomes framework that underpins the analysis conducted in Chapters 5, 6 and 7 of this report.
The analysis in this report relies on using rich administrative data to compare participants to non‑participants and propensity score matching is used to select a control group of non-participants to compare to participants. Using this technique, all the observable characteristics that affect participation in a programme are summarised into a “propensity score”. This single estimate captures the likelihood of an individual participating in the programme and reflects all the known characteristics that affect this likelihood. For this process to ensure that estimated programme effects are unbiased and reflect its true impact, it must be the case that the characteristics which affect participation in the programme and also affect outcomes are used to calculate the propensity score. For example, consider the case of degree education. Suppose having a degree made someone more likely to earn more, even in the absence of a training programme. Suppose it also made it more likely for that person to participate in a training programme. Not accounting for having a degree in the propensity score would lead to an over-estimation of the effects of training on income. The estimation would attribute increased earnings to training, when in fact it simply reflected the fact that more people with degrees participated in the training. By explicitly controlling for this in the propensity score, the treatment and control groups would compare only people with similar levels of degree attainment. The discussion in Section 4.4 demonstrates the rich set of variables available in this analysis mean that there can be confidence that these types of omission do not occur. This leads to the second important assumption that needs to be met for matching to be valid. There needs to be a good balance on the propensity score between treatment and control groups. Similar individuals need to be found to compare to one another. Statistical tests can be conducted to determine that this is the case. These technical details are provided in more detail in the technical report (OECD, 2023[6]).
4.3.3. A range of outcomes are reviewed to holistically evaluate programme impacts
The analysis in Chapters 5 to 7 documents a range of labour market outcomes from SMT and LMT participation, to provide a rich understanding of how the training programmes help to connect people with jobs. It utilises data on income, wages and unemployment spells to analyse how the training programmes affect transitions in the labour market, both into the type of job, the tenure of jobs and the income in those jobs.
Specifically, the evaluation includes the following labour market outcomes:
Probability of being employed. This probability is measured using a binary outcome variable which is equal to 1 if an individual is employed on the last day of the year, and equal to 0 otherwise.
Annual earnings. Defined as all earned income in a calendar year.
Monthly wages. Defined as all earned income in a calendar year divided by months of employment.
Total annual unemployment duration. Defined as the number of months spent in unemployment in a calendar year.
Probability of changing occupation. Defined as a binary variable which is equal to 1 if the individual’s current occupation is different to their last recorded occupation. The occupation is recorded on the last day of the calendar year.
Progression in the occupational ladder. Defined as the change in the rank of the current occupation, with respect to the average wages for that occupation, compared to the rank of the last held occupation (see Section 4.3.5).
Sub-groups of participants are also examined to see whether impacts vary
The impact of training may be different for different individuals. For example, a recent study of a training programme in Lithuania found differential impacts by sub-groups and contextualised those found in the meta‑analysis by Card, Kluve and Weber (OECD, 2022[9]). In order to determine whether this feature holds true for participants of training in Finland, and to add evidence to that already gathered by other existing studies on Finland (detailed separately for SMT and LMT in Chapters 5 and 6 respectively), the impact analysis in this report is conducted separately for a number of sub-groups: gender, individuals aged under 30 or 50 and above, rural and urban participants and by level of education.
In addition to participant sub-groups, Chapter 7 of this report investigates whether the introduction of SMT changed the provision of training for jobseekers as a whole. Because SMT was introduced nationally in 2010, it brings challenges to analysing how the change to the suite of provision has affected both the composition of jobseekers engaged with training and their outcomes subsequent to that training. This is exacerbated because it entails analysing outcomes so close to the onset of the financial crisis in 2008, which means temporal employment outcomes may be significantly different as the economic cycle progresses. Nevertheless, an exploration of the types of individuals that LMT and SMT cater for and whether outcomes for cohorts as a whole were impacted by this introduction will be an important part of describing how the training ecosystem helps people to connect with jobs (see Chapter 7).
4.3.4. Analysis is conducted on data from 2010 onwards for main outcome variables
To take advantage of the long panel data held by Statistics Finland, programmes are evaluated from 2010 onwards. The use of a longer time series of data has two primary advantages. It allows outcomes (such as earnings and employment) to be analysed for a longer period after the initial training period. This is of extra importance for programmes with longer durations, such as training programmes. A longer period spent participating in an ALMP outside of the labour market increases the potential of foregone earnings, as participants delay job search. However, if this serves to increase either the probability of employment, or the earnings whilst in employment, over the longer term the programme may still be beneficial. Therefore, evaluating longer term outcomes allows an exploration into these dynamics. Secondly, it provides more available data to look at periods before programme participation. This is important when quasi‑experimental techniques are being used, which require detailed data on participant characteristics. This helps to ensure that similar individuals are being compared and relies on having individual attributes that are measured prior to participation. As an example, prior unemployment spells can help to compare individuals with similar labour market histories.
The primary reason for starting the analysis in 2010 is twofold. For SMT, it is also the first year of the implementation of SMT, so it means that the participation in this programme is fully captured using this time frame. For occupation data, 2010 is also the first year for which consistent information is available. This means that changes to occupation and moves up and down the job ladder can be captured across time on a consistent basis, starting in 2010.
For the separate analysis on LMT and SMT in Chapters 5 and 6, the cohorts selected are from the years 2012‑14. This is related to the data availability, both pre‑ and post-programme. It means that incorporation of occupational data can be done for two years prior to unemployment, and for outcomes, a full four years of outcomes are available after the initial spell of unemployment.
Chapter 7 reviews how the introduction of SMT in 2010 changed the composition of training for jobseekers and whether or not this impacted upon their outcomes. In order to conduct this analysis it is necessary to evaluate a slightly earlier time period, going back to 2009, the year prior to its introduction. This allows the comparison of cohorts immediately before and after the introduction of SMT. However, by doing this, it precludes using occupation as an outcome variable, instead focusing on the other outcome variables described in the previous section.
The sample includes unemployed jobseekers in December of each year
The analysis uses the FOLK Basic dataset to define its sample of unemployed jobseekers. This is an annual dataset that categorises individuals’ status on 31 December of that year. The advantage of this dataset is that it offers a consistent sample frame with the outcomes’ variables, which are defined on the same basis. To analyse how training impacts upon jobseekers’ occupational mobility, it is necessary to observe their occupation prior to becoming unemployed. For this reason, unemployed individuals without any work experience are excluded from the sample (see the robustness checks of the analysis when including this population in (OECD, 2023[6])). It is possible for employed jobseekers to undertake LMT, but the analysis in the report focusses on how training helps unemployed individuals to improve their labour market outcomes.
4.3.5. Looking beyond employment outcomes to analyse occupational mobility
Occupational mismatch is one of the issues at the heart of unemployment (Patterson et al., 2016[10]; Belot, Kircher and Muller, 2018[11]; Marinescu and Rathelot, 2018[12]). Jobseekers may look for jobs in occupations with few job vacancies while other occupations with relatively more vacancies may not have enough candidates. SMT and LMT can allow jobseekers to gain the skills required to transit to occupations with better employment prospects. It is therefore important to investigate whether jobseekers change occupations and if they move towards better quality occupations as a result of their participation in SMT and LMT. This report estimates indeed the effect of participation in training programmes on occupational mobility.
To address this question a tractable index (OECD[9]; Laporšek et al., 2021[13]) neatly summarise moves along the occupational ladder corresponding to how they reflect underlying income dynamics. For example, an individual may, as a result of an ALMP, move into a lower paying job having also transitioned into an occupation with higher average wages. In this sense, current earnings may not be a good guide to potential earnings, as possibilities for future promotion are unlocked. Therefore, this index provides a rich source of additional information on job transitions that can be utilised alongside more traditional outcomes such as employment rates and earnings, to provide a more insightful analysis on labour market transitions.
The occupational index is built for 122 3‑digit ISCO occupations. For each occupation, the average monthly-earned income is computed over the 2012‑18 period. The occupations are then ordered from the lowest to the highest paid and attributed a percentile rank. Occupational mobility can thus be measured in income units and in percentiles. To illustrate this measure, Table 4.3 shows the occupational index in euros and in percentiles for the ten bottom, middle and top occupations.
The occupational index distribution for Finland shows that unemployed individuals are disproportionally represented in lower-ranked occupations (in the last job held prior to becoming unemployed) as compared to employed individuals (Figure 4.4). The figure demonstrates the importance of access to good quality training for jobseekers, to enhance their skills and to enable them to unlock job vacancies in occupations with better potential for progression. Figure 4.4 shows data for a cohort of individuals in 2014. This is to align the time period with the statistical analysis conducted in Chapters 5 and 6. However, there is a large degree of stability over time and this pattern continues to be present for later cohorts. On average, unemployed individuals have an occupational index that is approximately 7.7 percentage points lower compared to employed individuals and have mean monthly earnings below the median occupation, corresponding with a drop of EUR 272 on the average occupational monthly earnings. Using this index as a tool for analysis in Chapters 5 and 6 will permit a decomposition into how training programmes affect individuals’ subsequent labour market trajectories. Whether they use training primarily to unlock jobs in the same occupation as they were previously employed, or whether they use training as a means to access different occupations. Utilising this decomposition, alongside a more traditional focus on earnings, will give greater insight into mechanics of job moves following unemployment and training and can help to explain how training helps to support re‑allocation of labour in Finland.
Table 4.3. Ten bottom, middle and top occupations according to the occupational index
Source: OECD calculations based on FOLK datasets.
Figure 4.4. Unemployed individuals are overrepresented in lower-paid occupations and underrepresented in higher-paid occupations
Note: The heights of the lines indicate the relative share of individuals in occupations whose occupational index in percentiles are on the horizontal axis. The distributions are calculated for all individuals who were employed or unemployed at the end of 2014. Observations with index value above 100 are excluded from the kernel density chart. For unemployed individuals the occupation is their last employment known in the data, prior to becoming unemployed.
Source: OECD calculations based on FOLK datasets.
4.4. Linked administrative data form the basis of the data for evaluation
Statistics Finland offers a catalogue of “off-the‑shelf” administrative datasets that external researchers can access. It has a standardised application process and a fee schedule for both data access and the accompanying server capacity to execute analysis on its FIONA secure access server (Statistics Finland, 2022[14]). Researchers can also make requests for bespoke datasets, where charges are made dependent on the amount of time it requires for Statistics Finland staff to prepare the required data.
The analysis in this report uses two primary sources of information in the Statistics Finland repository.
FOLK data – These consist of several distinct datasets which draw in registry information from a wide variety of administrative sources. The sample these datasets use is usually all individuals that are permanently living in Finland on the last day of the year. These datasets cover several different dimensions including information on the education, family status including presence and age of children, cohabitation status and dates, socio‑economic background, household dynamics and rental status, assets, income, taxes paid, periods of training, unemployment and job search, employment firm and industry information. The information contained within these datasets on labour market aspects such as unemployment and training are drawn from the Ministry of Economic Affairs and Employment’s (TEM) register data.
TEM data – These data come directly from TEM and are not processed further by Statistics Finland, in contrast to the data which are processed to make the FOLK datasets. They include much more detailed information on the operational level data that are generated when the TE Offices conduct their labour market activities with jobseekers. These include data on job postings and the requirements of the advertised jobs, information on registered disabilities, data on job-search activities, actions and associated tasks agreed in individual employment plans, job offers and details about different types of work trials and training. These data provide detailed dynamics on the interaction between TE Offices and jobseekers, but less information on outcomes (in particular, income and job dynamics) and socio‑economic characteristics.
The processing of administrative data by Statistics Finland into compiled FOLK datasets brings advantages and disadvantages. Compilation of data into integrated datasets can mean that data preparation for the end user is easier and some aggregation has already been performed. For example, tax return data to tax ministries can often be vast and unwieldy. They may be weekly, fortnightly, monthly or any other frequency which employers pay their staff. This requires significant data processing to assimilate into useable records for analysis. Compiling these data into, for example, an annual income dataset can save researchers lots of data re‑work and can prevent inconsistencies that may arise from individuals using different data processing rules. In addition, meta data can be produced to help document and describe the data that are available. Statistics Finland produces meta data for its “off-the‑shelf” datasets, which detail the sample framework in the datasets, the time period for the data and the description of variables and their associated values. However, there are some datasets for which meta data have not been translated into English, which include the TEM datasets, which increases the difficulty of use for non-native researchers (see (OECD, 2023[6]) for further details).
However, there is a cautionary note on the production of compiled datasets. Where both the compiled data and the raw data used for its production are available, then it means inconsistencies may still arise, dependent on which data researchers use. Statistics Finland warehouses both the compiled (FOLK) and raw (TEM) data on a number of labour market variables. For example, the number of individuals participating in SMT or LMT appears different if you take this information from FOLK rather than from TEM (see (OECD, 2023[6]) for further details). Without detailed and comprehensive metadata on all the underlying data, it is difficult to determine from the outset which data provide researchers with the “right” answer.
The richness of the FOLK and TEM data means that it is possible to control for most conceivable factors that might bias the outcomes from the impact evaluation if they were not included. Studies have shown that detailed controls for unemployment, socio‑economic factors, pre‑treatment outcomes, geographical information and short-term labour market histories are sufficient to control for most selection bias and account for bias that might otherwise be explained by factors that are often unobservable in the administrative data. These computations were evaluated on a German training programme, similar in nature to those analysed here, providing more comfort that similar conclusions might be reached with these programmes (Lechner and Wunsch, 2013[15]).
Table 4.4. A rich set of variables is available to control for differences between individuals
Variables used to select non-participants that are alike to training participants
Using the FOLK datasets means that there is a long run of panel data to use for analysis, this is useful to observe long-term outcomes following participation and to control for prior labour market and unemployment histories. The FOLK datasets, which contain all the variables needed to control for differences between individuals, and to review outcomes, run as far back as 1987. In principle it is also possible to derive participation data from these datasets as well, meaning a feasible analysis could go back to 1987. The TEM datasets, which contain more detailed information on interactions of jobseekers with the PES stretch back to 1998, later than the FOLK datasets, but still with a long back series of panel data on which to draw.
The following chapters now go to describe the results for the analysis which has been undertaken using the data and methodology outlined above.
4.5. Conclusion
This chapter has outlined the details of the main training programmes that are available to jobseekers in Finland. Since its introduction, SMT has allowed individuals to retain their unemployment benefits whilst studying for longer-format education up to degree level. Over time the number of individuals studying with this support has increased so that there are more individuals studying via this pathway than there are via LMT, changing the overall nature and type of training and education that jobseekers participate in. In part this reflects the relative duration of courses. LMT courses are vocational and typically much shorter than the studies undertaken using SMT. SMT education is longer-format, part of the formal education system administered by OKM and includes continuation of degree‑level education. The programmes also cater to different individuals. SMT participants are more likely to be female, to have more children and are less likely to have previously had a lower level or manual profession.
It was not possible to use a randomised trial to evaluate LMT and SMT, therefore the analysis in the report uses quasi‑experimental techniques to ensure that training participants are compared to similar non-participants. It utilises the rich data on personal characteristics that are available at Statistics Finland to make this comparison, so that detailed past labour market outcomes, socio‑economic and personal characteristics are accounted for. The rich data held at Statistics Finland also permit a range of labour market outcomes to be assessed, across type of employment and level of income. In addition to this, data held on occupations of individuals allows an occupational index to be constructed. This means training policies can be assessed on a consistent and tractable basis to determine how they influence individual’s movements across different occupations.
References
[3] Alasalmi, J. et al. (2022), Ammatillisen työvoimakoulutuksen toimivuuden ja vaikutusten arviointi (Evaluation of the effectiveness and impact of vocational education training), Prime Minister’s Office Helsinki 2022.
[11] Belot, M., P. Kircher and P. Muller (2018), “Providing Advice to Jobseekers at Low Cost: An Experimental Study on Online Advice”, The Review of Economic Studies, Vol. 86/4, pp. 1411-1447, https://doi.org/10.1093/restud/rdy059.
[13] Laporšek, S. et al. (2021), “Winners and losers after 25 years of transition: Decreasing wage inequality in Slovenia”, Economic Systems, Vol. 45/2.
[15] Lechner, M. and C. Wunsch (2013), “Sensitivity of matching-based program evaluations to the availability of control variables”, Labour Economics, Vol. 21, pp. 111-121, https://doi.org/10.1016/J.LABECO.2013.01.004.
[12] Marinescu, I. and R. Rathelot (2018), “Mismatch Unemployment and the Geography of Job Search”, American Economic Journal: Macroeconomics, Vol. 10/3, pp. 42-70, https://doi.org/10.1257/mac.20160312.
[6] OECD (2023), Technical Report: Evaluation of active labour market policies in Finland, OECD, Paris, https://www.oecd.org/els/emp/FITechnicalReport2023.pdf.
[9] OECD (2022), Impact Evaluation of Vocational Training and Employment Subsidies for the Unemployed in Lithuania, Connecting People with Jobs, OECD Publishing, Paris, https://doi.org/10.1787/c22d68b3-en.
[2] OECD (2021), Improving the Provision of Active Labour Market Policies in Estonia, OECD Publishing, Paris, https://doi.org/10.1787/31f72c5b-en.
[5] OECD (2020), Continuous Learning in Working Life in Finland, Getting Skills Right, OECD Publishing, Paris, https://doi.org/10.1787/2ffcffe6-en.
[7] OECD (2020), “Impact evaluation of labour market policies through the use of linked administrative data”, OECD, Paris, https://www.oecd.org/els/emp/Impact_evaluation_of_LMP.pdf.
[8] OECD (2020), “Impact Evaluations Framework for the Spanish Ministry of Labour and Social Economy and Ministry of Inclusion, Social Security and Migrations”, OECD, Paris.
[1] OECD (2016), Back to Work: Finland: Improving the Re-employment Prospects of Displaced Workers, OECD Publishing, Paris, https://doi.org/10.1787/9789264264717-en.
[10] Patterson, C. et al. (2016), “Working hard in the wrong place: A mismatch-based explanation to the UK productivity puzzle”, European Economic Review, Vol. 84, pp. 42-56, https://doi.org/10.1016/j.euroecorev.2015.10.013.
[14] Statistics Finland (2022), Pricing of research data and services, https://www.stat.fi/tup/hinnat/tutkimuspalvelut_en.html (accessed on 01 July 2022).
[4] TEM (2022), Autumn 2022 Occupational Barometer, https://tem.fi/en/-/occupational-barometer-increase-in-labour-shortages-has-slowed-down-but-health-and-social-services-continue-to-ac-count-for-top-shortage-occupations.