The mission of ALMP

An active Labour Market Programme (as an ingredient of labour market programmes) can be defined as a set of measures applied to the unemployed or people at high risk of unemployment to assist them to get a job, training placement or an appropriate insurance against risk of unemployment (see Reference Björklund, Edin and HolmlundBjörklund et al., 2006). According to Reference Calmfors, Forslund and HemströmCalmfors et al. (2002), ALMP’s ultimate goal is high employment and low unemployment, facilitating labour productivity and economic growth. The stated social-political goal of an ALMP programme is to promote higher levels of income, well-being and anticipated welfare resulting from an increased employment probability, or (in the previous Swedish system) a higher anticipated income in the future for eligible programme participants.

The objective defined by the Swedish government is to contribute to a well-functioning labour market (Arbetsmarknadsstyrelsen, 2008: 8). For the given period, the following priorities were set by government directives: (1) effective matching of the unemployed and vacancies, (2) an optimal adjustment of ALMP measures to the labour market demand and (3) shortening unemployment spells of the most vulnerable groups of the unemployed. ALMP should also help fight against the moral hazard consisting in the unemployment insurance effects that might result in exaggerated demands and an increase in the number of UB claimants. The political basis of the ALMP scheme was established by the Alliance of Sweden formed by the liberal-conservative Moderate Party which won the 2006 elections. The coalition’s priority was to increase employment, enhancing the efficiency of the publicly funded job-matching programme. A just social policy has always informed the development of the renowned Swedish welfare state.

Active labour market policies, as a component of the flexicurity concept, originated as part of the European Employment Strategy (EES) introduced by the 1992 Treaty of the European Union, aiming at creating new and better jobs. ALMP measures envisage effective targeting in order to enhance the social guarantee of transition to new jobs. Other European Union (EU) flexicurity approach components (institutions) are the employment protection legislation (EPL) and the net replacement rate (NRR); see, for example, Reference KeuneKeune (2008).^{Footnote 1} The index of EPL (EPLI) expresses flexible and secure contractual arrangements, NRR representing a social security model facilitating mobility in the labour market. In the article, a more flexible labour market is demonstrated by changes in the above-mentioned institutions.

Flexicurity has been applied since the 1990s as a tool to combat globalisation-induced unemployment. The 2000 Nice European Council approved the European Social Agenda with a priority to reach full employment and implement Active Labour Market programmes while improving job quality and security. The flexicurity concept (described in the section below) was a cornerstone of the EES – part of the Europe 2020 ambitious growth strategy – which covers EU employment guidelines and member states’ National Reform Programmes. One of the EES target issues was employment, its rate being specified and life-cycle approach to skills innovation emphasised.

The notion of the flexicurity concept is attributed to the Dutch sociologist Hans Adriaansens who is supposed to have used the term for the first time in the mid-1990s (see Reference Lehwess-LitzmaanLehwess-Litzmann, 2012; Reference Wilthagen and TrosWilthagen and Tros, 2004). There are also some discussions about its possible Danish origin (cf. Reference Viebrock and ClasenViebrock and Clasen, 2009).

Sweden is a representative of the European Nordic social model. The roots of the welfare state can be traced back to the 1930s when organised labour staged protests as a major social force. In the interwar period, democracy came under threat from economic and ensuing political crisis, including Scandinavia with its short-lived coalitions and social unrest. Later agreements between political antagonists allowed for a system of centralised collective bargaining and improved labour relations. During the 1930s, unemployment insurance schemes were introduced. After World War II, the collective sphere expanded, high taxation enabling the state to redistribute revenues and provide lifelong care for all citizens. The Social Democratic party committed to a general social contract with the people which guaranteed the functioning of the welfare model. The Cold War aftermath brought about mass immigration as an accompanying phenomenon of globalisation. Looking for its own answers to new challenges, the Nordic social model seeks ways to reform the tax system to boost the labour supply and mobility in particular, flexicurity being one of the instruments. (For details, see the report The Nordic model – challenged but capable of reform published as part of ‘Sustainable Nordic Welfare’ programme launched by the Nordic Council of Ministers, 2014.)

The Nordic model has appropriate prerequisites for applying politics of the flexicurity concept (cf. Reference RogowskiRogowski, 2008). It relies on high levels of employment with an emphasis on work skills enhancement, supporting both school education and lifelong learning. Countries applying the model are characterised by the highest level of social protection expenditure, universal welfare provision, extensive financial intervention in the labour market’s active policy measures and strong labour unions ensuring highly pressed wage structures (Reference Sapir, Aghion and BertolaSapir et al., 2004: 375).

Typology of ALMP measures

Based on the Swedish typology of Reference Calmfors, Forslund and HemströmCalmfors et al. (2002), ALMP activities are generally divided into three categories: (1) job-broking and counselling services (aimed at effective matching of job seekers with vacancies), (2) employment training (occupationally oriented courses having been transformed into programmes run by specialised centres^{Footnote 2}) and (3) subsidised employment (aimed at creating new jobs). The traditional form of the latter ALMP measure – public relief works – was replaced by the work experience scheme in the 1990s, recruitment incentives and start-up grants becoming alternative forms of subsidised employment. Resource jobs were introduced involving grants for temporary hiring of unemployed persons. Since then, work experience subsidies have been paid to specific groups of the unemployed, such as youth, academics or immigrants. In the early noughties, Activity Guarantee and Job and Development Programmes were introduced for the long-term unemployed to engage in full-time activities, thus qualifying for an activity allowance.^{Footnote 3}

Table 1 classifies job-broking activities as a type of ALMP measures supposed to increase the matching efficiency. The table is based on Reference Brown and KoettlBrown and Koettl (2015) and is adjusted to Swedish conditions according to Reference Calmfors, Forslund and HemströmCalmfors et al. (2002). It breaks down job-broking services by categories and intended effects. It is obvious from the table that most of the activities listed enhanced the supply side of the labour market.

Table 1. Classification of job-broking activities.

ALMP and other flexicurity institutions’ effects

The flexicurity concept and ALMP’s compatibility with other institutions

In this article, a more flexible labour market is identified as being characterised by changes in the indicators of the three flexicurity institutions discussed above. To demonstrate flexicurity as a multidimensional concept, the Wilthagen matrix was applied (cf. EMCO Working Group on Flexicurity, 2006, or Reference Wilthagen and TrosWilthagen and Tros, 2004); see Table 2. The matrix takes into account labour market flexibility and job security (mostly – not necessarily – domains of employers and employees, respectively), distinguishing four forms of both the former and the latter which interact with each other.

Table 2. Examples of instrumental variables for flexicurity policy (Wilthagen matrix).

Source: EMCO Working Group on Flexicurity (2006: 5), authors’ adaptation.

ALMP: Active Labour Market Policy.

Only the forms of flexibility are discussed in the article since it rests on the hypothesis of a more flexible labour market. External numerical flexibility (demonstrated by the EPLI) indicates the degree of difficulty experienced by employers in hiring and firing their staff. Internal numerical flexibility expresses an employer’s ability to adjust to the changes in the company’s volume of work. Functional flexibility represents an employer’s organisational flexibility and employees’ ability to respond to the changes in society. Wage flexibility consists in the possibility of adjusting labour costs to continuously changing conditions.

Some interactions of selected flexicurity and security components can be considered as trade-offs (Reference Wilthagen and TrosWilthagen and Tros, 2004). A higher level of job security, for instance, implies lower numerical flexibility and vice versa. The flexicurity concept, its practical use and policy implications vary between countries. To efficiently implement flexicurity measures, it is vital to consider the economic conditions. A given business cycle phase, in this case the recession, is relevant to flexicurity measures’ application and effects. Compared to other countries, Sweden (2009) had a more favourable position due to its policy of increasing employment and strong public finances, a large surplus allowing for the use of automatic stabilisers.

Main effectiveness criteria

A higher matching efficiency means that more job placements are made with a given number of vacancies and job seekers (Reference Calmfors, Forslund and HemströmCalmfors et al., 2002). If a given number of job seekers requires a lower number of vacancies allowing for the expected number of matches to be made, there is an increased probability of filling a vacant job position. Reference Bleakley and FuhrerBleakley and Fuhrer (1997) also argue that the number of new matches made with a given stock of vacancies and the unemployed is crucial. As a consequence of higher matching effectiveness, job seekers find employment faster, being addressed by employers without delay, unemployment spells and the duration of recruitment^{Footnote 4} getting shorter. Periods of unemployment and recruitment vary in length with the changing labour market tightness over the business cycle. A tight labour market signals that employers experience more difficulties in finding labour force, a loose market, on the contrary, fosters the preconditions for a shorter recruitment process, while making it more difficult to maintain short unemployment spells. As labour demand increases, competition among employers intensifies, recruitment becomes longer and the unemployment spell shorter. (A detailed description of labour market operations is presented in annual reports of the Swedish PES.^{Footnote 5})

In the following paragraphs, Swedish yearly labour market trends between 2007 and 2012 are tackled while taking into account the development of unemployment spell and recruitment duration with respect to the tightness of the labour market.

In 2007, labour force demand was high, the number of newly announced vacancies was rising, its growth, however, being slower than in 2006. Unemployment continued to decline. This trend was accompanied by (1) a prolonged recruitment duration (shorter though, compared to the same labour market tightness in previous years) and (2) a shorter unemployment spell. As it follows from Arbetsmarknadsstyrelsen (2008), the median value of the unemployment spell increased (to an average of 87 days). This must have been caused by a lower inflow of the newly unemployed into unemployment as there was a higher transition frequency from unemployment to employment. Those with a relatively long unemployment spell remained among the unemployed which may have indicated a structural type of unemployment. Increasing tightness was accompanied by a higher vacancy rate (VR) and a shortage of job seekers. However, VR was lower compared to periods with the same labour market tightness in the past which suggests a higher matching effectiveness.

In 2008, labour demand began to decline, unemployment increasing (cf. Arbetsförmedlingen, 2009). The number of newly announced vacancies fell, the planned dismissals having risen most dramatically since the 1990s’ recession. These developments were accompanied by (1) a shorter duration of recruitment (longer, however, in comparison to the labour market tightness in the past) and (2) the lengthening of the unemployment spell, which corresponded to a lower frequency of unemployment-to-employment transition. The median of the unemployment spell decreased, the inflow of the unemployed into unemployment having grown. The 2008 matching indicator suggested a lower matching efficiency, statistically significant only at the end of the year. The indicator trends over the previous decade are shown in Figure 1.

Figure 1. Matching indicator for Sweden in 2001–2009.

Source: Reproduced with permission from the publisher, Arbetsförmedlingen (2010: 13); authors’ adaptation.

In 2009, labour force demand declined, the continuing rise in unemployment slowing towards the end of the year. The economic cycle having bottomed out, first signs of recovery appeared. The number of newly announced vacancies stopped declining. A rise in the unemployment, however, was experienced as a result of longer spells of joblessness, which was a signal of structural unemployment (like in the 1990s). The largest decline in Sweden’s GNP (gross national product) since World War II occurred (Arbetsförmedlingen, 2010). The trend was accompanied by (1) a shortened recruitment duration and (2) a lengthened unemployment spell. The number of applicants registered with the PES who found a job was growing, but the increase did not compensate for the inflow of the newly registered candidates. Moreover, the 2009 matching indicator showed a lower matching effectiveness (Arbetsförmedlingen, 2010), that is, fewer vacancies were filled compared to the estimate.

In 2010, the Swedish economy and the labour market started to recover. The decline in unemployment was associated with increased labour supply due to a rise in immigration of working-age people and the consequences of the sickness insurance reform (Arbetsförmedlingen, 2011a). Labour force demand having increased, the numbers of announced dismissals and registered PES applicants dropped. The PES report (cf. Arbetsförmedlingen, 2011b) discusses a lower matching efficiency in the first half of the year, effectiveness statistics for the full year 2010 and half of the following year, however, corresponding to the actual stage of the business cycle (Arbetsförmedlingen, 2012b).

In the first half of 2011, there was also a strong labour market with new vacancies and an increase in the employment rates. Nonetheless, the market weakened at the end of the year as a result of the Eurozone crisis. Employment growth stopped, fewer jobs having been created (cf. Arbetsförmedlingen, 2012a). Both towards the end of 2011 and in 2012, the matching indicator showed deviations from the expected values corresponding to a given phase of the economic cycle, that is, to the real labour market tightness (Arbetsförmedlingen, 2014).

In 2012, surprisingly, the labour market trend was favourable, unlike the weakened growth in the Eurozone. The number of people who found a job was slightly higher than in the previous year, the number of new vacancies being relatively large, smaller though compared with 2011. The situation worsened in the last quarter when temporary contracts were restricted (Arbetsförmedlingen, 2013). Total sectoral employment, however, increased. The unemployment rate was gradually rising, the number of PES registered people being higher than in the previous year but lower than in 2010. The matching conditions deteriorated. Overall, labour force demand declined, the number of people searching for a job increasing.

Swedish national Active Labour Market Programmes in 2007–2012

A particular test of ALMP effectiveness occurred in Sweden during the global crisis (especially in 2008–2009 and 2011–2012). The government took reform measures to improve the job intermediation services and matching efficiency. Additional measures were also applied, responding to labour market weakness in the then ongoing economic crisis. The introduction of the Job and Development Guarantee programme in 2007 made the year critical in terms of activation and motivation measures. In the following years, cost-saving provisions^{Footnote 6} and measures enabling a broader range of jobless people to reap UBs were adopted (cf. Reference SibbmarkSibbmark, 2008, Reference Sibbmark2009, Reference Sibbmark2010) thus alleviating the economic problems. In 2009, for example, a requirement for certain time worked as a precondition for membership in the unemployment insurance agency^{Footnote 7} was cancelled. During the recession, job-broking services were also developed by mobilising PES human resources, intensifying its operation, including opening and running new job centres in the regions most affected by unemployment and allocating substantial funds for work experience schemes. PES intensified the delivery of job coaching services, employing both internal coaches and staff of kompletterande aktörer (complementary actors).^{Footnote 8} Job coaching services and work experience schemes became the cornerstones of the labour market programme.

Such a focus on job-brokering and matching was not only typical for the period 2007–2012 marked by economic downturn. The issue had been given due attention even before – within the reform programme launched in order to, among other things, moderate the labour market effects of new social realities in Europe (cf., for example, European Commission, 2006). The reform having been instituted by the Swedish government in 2006, the PES has been operating in a more organised and efficient mode since 2008 (Arbetsförmedlingen, 2009). To make job mediation services more effective for employers, the PES concluded framework agreements on active job-broking services to supply employers with competent candidates. As for the job seekers themselves, PES provided them with the services of professional job brokers, runs an online self-service portal and offered other applicant-friendly tools such as Jobbnätet (a job-networking programme), thus affecting job seekers’ behaviour and shortening the unemployment spell.

In 2010, sickness insurance and establishment reforms were initiated aiming at the inclusion of the long-term sick and new immigrants into the labour market, affecting ALMP. A focus was also placed on work experience, practical skills development and coaching including the provision of kompletterande aktörer. This initiative was in line with the argument of Reference SianesiSianesi (2008) that a higher probability of drawing UBs while under a work experience scheme is connected with lower employment rates.

Better functioning of the labour market made it possible to allocate the funds to the most vulnerable people at the expense of anti-crisis measures. Effective matching along with job brokerage improvements remained a high ALMP priority. According to Arbetsförmedlingen (2011a), agreements with employers on active job-broking services contributed to shorter unemployment spells. The PES started to use social media, namely, Facebook and YouTube, as an information channel, focusing on announced vacancy searches. To facilitate the matching process, job brokers were trained in the knowledge of the most demanded jobs and skills, self-service being still promoted. The development strategy aiming at providing customers throughout the whole country with the same services was consistently pursued.

In 2011, PES established contracts with important employers. Counselling services and new vacant positions started to be offered via telephone, which helped reduce recruitment and unemployment spell duration. Since that year, PES has been waging a permanent information campaign.

In 2012, the previous 5 years of PES reform activities resulted in implementing novel approaches to pro-active ALMP agenda. Encompassing cooperation with employers materialised in contracts with businesses and the national job-matching project conducted across the country; it aimed at preventing long-term unemployment with an emphasis on an individual, case-by-case approach. The budget and the number of job brokers having been increased, more interviews with the candidates and contacts with employers could be made. The establishment of a special PES division for national clients facilitated this process (cf. Arbetsförmedlingen, 2013).

Statistical evidence of effectiveness

Trends in flexicurity indicators in 2007–2012

This section provides evidence of the increased flexibility of the Swedish labour market between 2007 and 2012. Having remained constant for several years, the overall EPLI value decreased in 2008 as a consequence of the loosened temporary contract regulations. Table 3 indicates a sharp decline by 0.625 points in the value of temporary EPLI in 2008, which stayed the same in 2009–2012, suggesting a trend towards less strict regulations. In general, weaker EPL implies a higher market mobility, that is, greater numerical flexibility.

Table 3. Seasonality test and unit root test of the seasonal adjusted time series.

Source: Own calculations.

OIE: outflow into employment; VR: vacancy rate; RUE_M_u20: percentage ratio of unemployed 15- to 19-year-old men to labour force; RUE_W_u20: percentage ratio of unemployed 15- to 19-year-old women to labour force; RUE_MW_o55: percentage ratio of unemployed men and women aged over 55 to labour force; R_MW_o55: percentage ratio of men and women aged over 55 to labour force; RUE_o6M: percentage ratio of those unemployed for over 6 months; ADF: augmented Reference Dickey and FullerDickey–Fuller (1979).

The NRRs fell by 5 to 9 percentage points both for households with and without children following a trend that began in the early 2000s.^{Footnote 9} NRRs for households without children were the same as for single persons and one-earner married couples, while a lone parent’s NRR in a household with two children was higher than that of a one-earner married couple with two children. According to mainstream economic theory, lower UBs tend to increase labour market flexibility.

During the monitored period, the Swedish government took reform steps to improve job-broking services aiming to increase the matching efficiency (see detailed analysis in the section above). The same effect of increasing labour market flexibility should have been achieved by adopting intensified job brokerage measures. Nevertheless, the intensified job-broking did not seem to contribute to an increase in the matching effectiveness during greater part of the monitored period.

Reasons for reduced matching efficiency

To explore the reasons for a decline in the matching efficiency, trends in labour market supply and demand were examined using regression models. Trends in demand were reflected in the development of the VR and the outflow into employment, defined as the number of new job matches. Vulnerable groups of the unemployed, namely, those of 15- to 19-year-old men and women, respectively, as well as unemployed 55- to 65-year-olds were adopted as the key indicators of trends in labour supply.

Demand and supply developments were empirically examined by analysing the relationships between their time series. It was explored how the time series of new job matches (i.e. outflows into employment) and VRs depend on the percentage ratio of

• Unemployed 15- to 19-year-old men to the labour force (RUE_M_u20)

• Unemployed 15- to 19-year-old women to the labour force (RUE_W_u20)

• Unemployed 55- to 64-year-old men and women to the labour force (RUE_MW_o55)

• All 55- to 64-year-olds to the labour force (R_MW_o55)

• Those unemployed for over 6 months to all openly unemployed persons (RUE_o6M).

Monthly time series from January 2007 to December 2012 having been available, research data were drawn from the national government agencies – the Swedish PES (Arbetsförmedlingen) and Statistics Sweden.

Prior to such a relationship analysis, it is necessary to test whether the time series contain a seasonal component. If so, they are to be seasonally adjusted using the X-13ARIMA-SEATS method (cf. Census Bureau, 2017). The prerequisite for statistically meaningful comparisons between time series is that they are of the same integration type, that is, both are either stationary I(0) or non-stationary I(1) (Reference Engle and GrangerEngle and Granger, 1987, or Reference Granger and NewboldGranger and Newbold, 1974). Therefore, in the second step, unit root tests were performed.

Seasonal adjustment was performed for all the time series since the F-test identified a statistically significant seasonal component in each of them (cf. Reference McDonald-Johnson, Monsell and FescinaMcDonald-Johnson et al., 2006). Table 3 presents the augmented Dickey–Fuller (ADF) and seasonality test outcomes, the former test proving that the non-stationarity (I(1)) hypothesis was not rejected at the 5% significance level for any time series.

Spurious regression relationships having been excluded, cointegration test (Reference Engle and GrangerEngle and Granger, 1987) was conducted. As Table 4 indicates, at the 5% significance level, spurious relationships were rejected only between the time series of new job matches and the ratios of unemployed teenage men and women to the workforce, and of the long-term unemployed to all openly unemployed people. The other time series were not analysed since in their case (as well as in all VR relations) only a spurious relationship occurred.

Table 4. Test of cointegration.

Source: Own calculations.

OIE: outflow into employment; VR: vacancy rate; RUE_M_u20: percentage ratio of unemployed 15- to 19-year-old men to labour force; RUE_W_u20: percentage ratio of unemployed 15- to 19-year-old women to labour force; RUE_MW_o55: percentage ratio of unemployed men and women aged over 55 to labour force; R_MW_o55: percentage ratio of men and women aged over 55 to labour force; RUE_o6M: percentage ratio of those unemployed for over 6 months.

It follows from the above that the relationship is only possible between new matches and time series of the ratios of unemployed 15- to 19-year-old men and women to the labour force (see Figures 2 and 3, respectively), and of the proportion of the 6-months plus unemployed to all openly unemployed persons (Figure 4). In order to present the time series in one chart, their values were normalised, that is, adjusted to the same mean and variance.

Figure 2. Relation of new matches to ratios of unemployed 15- to 19-year-old men to the labour force.

Source: Authors’ calculations.

OIE: outflow into employment; RUE_M_u20: percentage ratio of unemployed 15- to 19-year-old men to labour force.

Figure 3. Relation of new matches to ratios of unemployed 15- to 19-year-old women to the labour force.

Source: Authors’ calculations.

OIE: outflow into employment; RUE_W_u20: percentage ratio of unemployed 15- to 19-year-old women to labour force.

Figure 4. Relation of new matches to proportion of the 6-months plus unemployed to all openly unemployed persons.

OIE: outflow into employment; RUE_o6M: percentage ratio of those unemployed for over 6 months.

Table 5 presents autoregressive-distributed lag (ADL) regression models estimated (see Reference Hendry, Pagan, Sargan, Grilliches and IntrillgatorHendry et al., 1984) and diagnostic test outcomes achieved. To eliminate jumps in time series, a dummy variable D1 (D1 = 1 for 12/2007, 4/2008 and 8/2012, otherwise D1 = 0) was added to the model.

Table 5. ADL models.

Source: Own calculations.

OIE: outflow into employment; RUE_M_u20: percentage ratio of unemployed 15- to 19-year-old men to labour force; RUE_W_u20: percentage ratio of unemployed 15- to 19-year-old women to labour force; RUE_o6M: percentage ratio of those unemployed for over 6 months.

* Dummy variable D1: D1 = 1 for 12/2007, 4/2008 and 8/2012, otherwise D1 = 0; ADL: autoregressive-distributed lag.

Expressing the long-term relationship between the analysed time series, error correction models (cf. Reference Engle and GrangerEngle and Granger, 1987) were recalculated from ADL models via the following equation

$\begin{array}{c}\Delta \widehat{OIE}\tau =-0.387\Delta OI{E}_{t-1}+125.446\Delta RUE\_M\_u{20}_{t-1}-5976.35D_{1t}\\ -0.097\left(OI{E}_{t-2}-1294.361RUE\_M\_u{20}_{t-2}\right),\\ \Delta \widehat{OIE}\tau =-0.434\Delta OI{E}_{t-1}+177.015\Delta RUE\_W\_u{20}_{t-2}-6254.61D_{1t}\\ -0.123\left(OI{E}_{t-2}-1434.973RUE\_W\_u{20}_{t-2}\right)\\ \Delta \widehat{OIE}\tau =-0.509\Delta OI{E}_{t-1}+430.889\Delta RUE\_o6M_{t-2}-6212.85D_{1t}\\ -0.260\left(OI{E}_{t-2}-1655.859RUE\_o6M_{t-2}\right)\end{array}$

That is,

$\begin{array}{c}{\widehat{OIE}}_t=1294.361RUE\_M\_u{20}_t,\\ {\widehat{OIE}}_t=1434.973RUE\_W\_u{20}_t,\\ \widehat{OIE}\tau =1655.859RUE\_o6M_t\end{array}$

That is,

where OIE is outflow into employment.

In 2007–2012, the creation of new job matches in Sweden was directly proportional to the ratios of 15- to 19-year-old unemployed men and women to the labour force, respectively; the former proportional dependence being relatively weak, the latter one slightly stronger. This means that the number of newly occupied positions grew with an increase in the ratio of unemployed teenagers to the workforce. It is also obvious that the new matches directly depended on the ratio of those who are unemployed for more than 6 months to all the unemployed, the number of new jobs rising with an increase in the above ratio. It can be thus concluded that new jobs were taken by different groups than the long-term and teenage unemployed, the matchmaking being the least effective for the latter ones.