Luxembourg: Selected Issues


Luxembourg: Selected Issues

Promoting Employment of Vulnearble Groups to make Growth more Inclusive1

Luxembourg’s unemployment rate is low by European standards, but it has risen since the global financial crisis. The youth and lower skilled workers are particularly at risk, in spite of innovative active labor market policies. Substantial employment disparities also remain across age and skill groups, and refugees as well as non-EU migrants are less integrated in the labor market. While public spending on education is high, it does not translate into higher students tests scores compared to other countries mainly due to Luxembourg’s multilingual curriculum. To guide the design of policy reforms to promote employment of vulnerable groups and make growth more inclusive, this chapter first provides a comparative analysis of the recent performance of the labor market. Second, it identifies the role of individual socio-economic characteristics in determining labor market outcomes, using a standard probit regression model estimated on microeconomic data from the European Union Labor Force Survey. Third, it assesses the efficiency of active labor market policies and their interactions with the social benefits system. Fourth, it appraises the performance of the education system. Finally, it discusses policy options to ensure that growth benefits reach everyone in the society.

A. Recent Labor Market Dynamics

1. Luxembourg’s unemployment rate is historically low by European standards, but it has steadily increased since 2010. At below 3 percent over 2000–03, the unemployment rate in Luxembourg was far below unemployment rates in neighboring countries. But, it was on an upward trend even before the onset of the financial crisis. Between 2010 and 2015, it increased by more than half and it currently exceeds the unemployment rate in Germany, despite a decrease from 2015 and onwards. In addition, an increasing share of new jobs accrue to cross-border workers who now represent more than 45 percent of the employed, reflecting skills mismatches.


Trend in Unemployment Rate

(Percent of population 15-64)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: OECD and IMF Staff calculations.

Luxembourg Labor Force Structure, 2016Q2

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: Statec and IMF Staff calculations

2. In Luxembourg, the unemployment rate is highest for low-skilled (below upper secondary) and young (15–24 years) workers (Figure 1). For instance, the share of people who did not finish secondary school accounts for 43 percent of the unemployed, more than double their share in the labor force. At the same time, comparison with the neighboring countries suggests that the unemployment rate for high skilled workers (graduated from tertiary education) is the second highest in Luxembourg after France. The unemployment rate of young workers remains higher than for older ones in Luxembourg as in neighboring countries. Overall, the unemployment rate is higher than in Germany, and lower than in France and Belgium for all age groups. In addition, the employment rate for older workers is lower in Luxembourg than in neighboring countries and the OECD average, mainly reflecting a lower participation rate.

Figure 1.
Figure 1.

Labor Market Dynamics

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: OECD, Eurostat, and IMF staff calculations.

3. Moreover, (un)employment disparities are substantial across birth origin subgroups within prime age in Luxembourg. Despite a slightly higher participation rate, EU migrants are less likely to be employed than natives. As a result, the unemployment rate of prime-age EU migrants is almost double that of natives. With relatively lower participation, non EU migrants end up with an unemployment rate which is four times that of natives. The difference in employment rates between natives and non EU migrants is almost twice the difference in participation rate, making non EU migrants potentially the most vulnerable group in the market.


Labor Market Status by Birth Origin, 2015

(In percent, ages 25-54)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: Eurostat and IMF Staff calculations.

4. Reforms to improve the activity rate of natives could help increase growth potential by increasing employment. Indeed, cross-country comparison suggests that the overall participation rate of natives is the lowest among most European countries, mainly due to low activity rates among young (15–24 years) and older (55–64 years) workers. In addition, the difference in the participation of natives, relative to EU 28 migrants is higher in Luxembourg than in most European countries. Together with the lower ratio of per capita GNI to per capita GDP than in neighboring countries, this suggests that there is a room to increase the participation of residents, including natives.


Activity Rate of Native and EU28 Migrants, 2014

(Percent of population 15-64)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: Eurostat and IMF Staff calculations.

Ratio of Per capita GNI to Per capita GDP

(In percent)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources; Eurostat, World Bank and IMF Staff calculations.

B. Who Are the Vulnerable Groups in the Labor Market?

5. The design of policies to increase employment rates requires the identification of the groups at risk. To assess the factors underlying labor market performance, we explore the relative likelihood of being out or in a job conditional on belonging to a certain socioeconomic group. We use a standard probit regression model estimated on microeconomic data for Luxembourg and neighboring countries from the European Union Labor Force Survey (EU LFS).2 In contrast to summary statistics, probit regressions allow controlling for overlap between vulnerable sub-groups. Thus, we compare the impact of individual background factors like age, gender, household composition, level of education, origin and years of residency in determining labor market outcomes both pre- and post-crisis. The study has been extended to neighboring countries (France, Belgium) to allow cross-country comparisons.3 To assess the potential effects of the global financial crisis on labor market performance, we consider 2006 as the pre-crisis reference year, and compare it to 2014, which is the most recent post-crisis year for which data are available.

6. The results show that young, non-EU immigrants and low-skilled workers, are more vulnerable than other subgroups.

  • Age has a varied effect on the probability of being unemployed. Youth has the highest probability of unemployment both before and after the crisis. Indeed, an individual aged between 15 and 24 years old is 12.6 percentage points more likely to be unemployed in 2014 than an individual aged 25 to 54 years old. Cross-country comparison shows that the unemployment youth penalty is broadly in line with neighboring countries (Annex I, Table A1).

  • In 2006, the unemployment risk of an individual who did not finish upper secondary school was 2.9 percentage points higher than the probability of being unemployed for an individual who has a university degree. This penalty went up with the crisis to 3.7 percentage points in 2014. This skill unemployment penalty is considerably lower than in Belgium and France (Annex I, Table A1).

  • These results are broadly confirmed when looking at conditional employment probabilities.

  • Conditional on all other individual background factors, the unemployment risk for non-EU born migrants is more than three times that of natives.

Table 1.

Probit Regression

article image

Indicates that the result is not significant for p < 0.1

Change in probability compared to the base category unless otherwise noted.

Estimates are robust to heteroskedasticity. Full estimation results are presented in annex.

7. Cross-country comparison shows that before the crisis, Luxembourg had the highest marginal unemployment risk for females, relative to male workers. But, this gender difference has vanished after the crisis. In 2014, there was no significant difference between the risk of unemployment between male and female workers in Luxembourg, or Belgium. This finding is explained by an increase in the absolute unemployment risk of male workers after the crisis and an increase in the absolute employment probability of females after the crisis. Despite this increase in the female employment rate, male workers are 9.6 percentage points more likely to be employed in 2014, indicating a lower activity rate for females.


Marginal Risk of Unemployment of Females Relative to Males

(Percentage points)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: IMF Staff calculations.Note: Pattern fill indicates a non significant coefficient at 10 percent.

8. The unemployment risk for older workers has increased, more than proportionally to the increase in overall unemployment in Luxembourg after the crisis. In comparison to an individual aged 25 to 54 years old, the marginal probability of unemployment for older workers (55–64 years) was 2.1 percentage point lower in 2006. However, older workers in Luxembourg have lost this premium after the crisis and in 2014 there was no longer a significant difference in the risks of unemployment between these two age groups. Over the same period, this premium increased in France.


Change in Older Workers’ Marginal Unemployment Risk Relative to Ages 25-54

(Percentage points, change 2006-14)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: IMF Staff calculations.

9. The absolute unemployment risk for EU born migrants was lower in Luxembourg than in neighboring countries but they incur the highest marginal unemployment risk (Figure 2). The absolute risk of being unemployed for EU born migrants in Luxembourg was 6.7 percent in 2014 (base category probability plus marginal effect), the lowest in comparison to neighbors, partially due to the higher overall unemployment rate in the other countries. But, relatively to natives, EU born migrants have the highest marginal unemployment penalty in Luxembourg. When focusing on employment probabilities, we find that in 2014, EU migrants have the same conditional probability to be employed as natives. Together, these findings imply EU born migrants participate more the labor market but have higher risk to be out of a job. Non-EU born migrants are much more likely to be unemployed than EU born migrants and natives in Luxembourg, as in neighboring countries. Consistent with the summary statistics, these results suggest that efforts are needed to ease migrants’ integration to the labor market, and to increase labor market participation among natives.

Figure 2.
Figure 2.

Determinants of Labor Market Performance

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: Eurostat and IMF staff calculations.

10. Staying longer in Luxembourg increases migrants’ labor market participation (Figure 2). Recently immigrated workers have a better chance to get a job in Luxembourg than in the other covered countries. But, years of residency do not affect the unemployment risk of migrants in Luxembourg, while they matter for the probability of being employed. In fact, there is no statistical difference between the unemployment risk of newcomers and those who stayed for more than 4 years, but the probability of being employed, relative to newcomers, increases by 10 percentage points after 4 years of residency. These findings suggest that activity rates of migrants increase over time, pointing to scope for targeted policies to accelerate migrants’ integration to the labor market. In France, staying for 4 years or more reduces the risk of unemployment by 22 percentage points in France, compared to recently immigrated workers.

11. Joint probabilities estimates confirm the previous results that young, non-EU migrants and low skilled workers underperform compared to other groups (Annex I, Table A2). To better gauge the factors determining individual labor market performance, we compare the likelihood of being in or out of a job for sub-groups of age (young, prime-age, older), education levels (low secondary, upper secondary, and tertiary), birth origin (native, EU born and non-EU born), and gender. The results confirm that with comparable other socio-economic backgrounds, young workers underperform compared to other age groups; low skilled are more vulnerable than high skilled; non-EU migrants have less chance to be employed than natives and EU born migrants. Estimates also confirm that there is little difference between the unemployment probability of men and women with comparable socio-economic background.

12. Surprisingly, young workers with a university degree are at least twice more likely to be unemployed than prime-age workers who do not finish secondary school. For instance, the unemployment probability of high skilled young workers is 13.1 percent, more than the double of the unemployment risk of low skilled prime-age workers. This finding remains when we consider employment probabilities. Indeed, high skilled young workers have a probability of 37 percent to get a job, while low skilled prime-age workers are employed with a chance of 72.2 percent. We also find that, across sub-groups, there is no statistical difference in the unemployment risks between workers who do not complete a university degree and those who did not finish secondary school. Indeed, jobseekers who complete upper secondary school face the same unemployment risk as those who just completed lower secondary school. However, finishing secondary school still increases the probability to be employed. This result could be explained by a higher activity rate among workers who finish secondary school compared to those who did not.


Employment Probability and Skill Premium, 2014


Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: IMF Staff calculations.

13. Finally, non-EU migrants with a university degree are twice more likely to be unemployed than low skilled natives (Figure 2). Indeed, subsample results indicate that the unemployment probability of non-EU migrants with a tertiary degree is more than twice that of low skilled natives. EU migrants need at least a university degree to be as successful as low skilled natives. However, there is no statistical difference in the employment probabilities of EU migrants and natives with the same qualification whereas non-EU migrants have a slightly lower probability to be employed. Moreover, additional education attainment provides a similar premium on employment probability for both natives and migrants.

C. Work Incentives and Labor Market Policies4

14. Recent studies by IMF staff suggest that effective Active Labor Market Policies (ALMP) can boost output and employment regardless of cyclical economic conditions.5 Indeed, World Economic Outlook estimates find that discretionary increases in public spending on ALMP have a statistically significant impact on medium-term output and employment. The effects are lower, but remain positive, in bad times than in expansions. The effects are amplified when higher spending is combined with other reforms intended to increase the efficiency of ALMP. If implemented in a budget-neutral manner, the effects remain significant and do not vary substantially with the business cycle, even though they are smaller. Moreover, higher budget-neutral spending in ALMP implies net positive fiscal benefits over the medium term.6

15. Higher unemployment has been accompanied with higher spending on Active Labor Market Policies (ALMP) in Luxembourg. Indeed, ALMP spending has increased from less than 0.35 percent of GDP in 2010, to more than 0.5 percent of GDP in 2014. However, average ALMP spending over the period 2010–14, in percent of GDP, was still lower in Luxembourg than in Belgium and France, but higher than in Germany. Moreover, cross-country comparison suggests that labor market policies cover a larger share of job seekers in Luxembourg than in neighboring countries. Despite these policy efforts to promotive employment, Luxembourg’s unemployment rate increased from around 4.5 percent in 2010 to almost 6 percent in 2014. While the unemployment rate rose also in France and in Belgium, it decreased by more than 1 percentage point in Germany.


Spending in ALMP vs. Unemployment Rate

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: OECD, Eurostat, and IMF Staff calculations.

Active Labor Market Policies and Outcomes

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: Eurostat, OECD, and IMF Staff calculations.

16. Evaluations of ALMPs suggest that their effectiveness depends on program types and the targeted groups.7 The OECD breaks LMP into services, supports, and activation measures (Box 1). Activation measures include training, direct job creation, and employment and start-up incentives. Most recent evaluations in other OECD countries suggest that job search assistance programs (LMP services) yield the best impacts especially in the short run. Start-ups incentives for the minority among the unemployed who have entrepreneurial skills and the motivation to survive in a competitive environment are also effective. Training programs are not particularly effective in the short run, but have more positive medium term effects (after 2 years). Programs targeting youths are significantly less likely to be effective unless they contain an appropriate mix of schooling, strengthening of occupational skills and on-the-job training, ideally in an integrated manner. Direct employment programs in the public sector are generally less successful than other types of ALMPs. In 2014, Luxembourg spent relatively more in employment incentives compared to other countries. Germany, whose ALMPs have proven to be highly successful, spent relatively more in public employment services and training. Employment subsidies should be limited and well monitored to avoid providing windfalls to employers without creating durable jobs.


ALMP Spending by Programme, 2014

(In percent of total ALMP spending)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: OECD and IMF Staff calculations.

17. Enhancing and continuously adapting the public employment agency (ADEM) policies to vulnerable groups could help improve the efficiency of ALMPs. The Youth Guarantee Scheme initiated in 2014 aims to increase employment of the young. The professionalization placement program and the professional reinsertion employment contract, launched in 2016, offer opportunities to workers above 45 years old, to highlight their professional capabilities or to improve their knowledge and professional capacities within a company for a short period of time. A new interactive platform, JobBoard, officially launched in March 2016, is intended to improve the matching between job seekers and employers. Special measures to reconcile professional and family responsibilities, in order to promote women business entrepreneurs, are intended to improve the insertion of women in the labor market.8 The ADEM also developed internal language training for job-seekers from immigrant communities. Efforts to ease integration of refugees to the labor market include the Accelerated Integration Programme initiated to enroll newcomers into language classes, schools, and other training programs as also recommended by the 2016 consultation.

OECD LMP Database: Coverage and Limits1

Definition and coverage: Labor Market Policy (LMP) data published by OECD covers public interventions in the labor market intended to improve its efficiency and to correct disequilibria. Thus, they are limited to policy interventions targeted to favor vulnerable groups in the labor market. Data include public expenditure and participants and are collected annually from administrative sources. LMP distinguishes eight main categories of labor market interventions classified by type of actions: services (category 1), activation measures (categories 2–7), and supports (categories 8–9). LMP services and activation measures are generally considered as Active Labor Market Policies (ALMP), while LMP supports are referred to as passive LMP.

  • LMP services cover all services and activities of the public employment service (PES) together with any other publicly funded services for jobseekers.

  • LMP measures cover activation measures for the unemployed and other targeted groups, and consist in training, job rotation and job sharing, employment incentives, supported employment and rehabilitation, direct job creation, and start-up incentives.

  • LMP support covers financial assistance that aims to compensate individuals for loss of wage or salary (out-of-work income maintenance and support, i.e. mostly unemployment benefits) or which facilitates early retirement.

Caveats: The OECD data set has some features which may limit the results of empirical evidence on the impact of ALMP. First, to be included, the labor market measures should be publically financed. Second, they should be targeted to a specific group of individuals who are at risk in the labor market. Third, it excludes in work benefits such as Earned Income Tax Credit in US or the Prime de l’Emploi in France when they are not conditional on the search for work, measures targeted to all members of a vulnerable group such as wage subsidies for young people or for people in depressed regions, and measures that pay a wage subsidy for an indefinite period. Fourth, cross-country data comparability issues also arise as some countries exclude some measures that others include. For example, France and Italy include most of their public spending on apprenticeships in the database while other do not due to the targeting criteria. Luxembourg includes the Indemnite Compensatoire in employment incentives while similar scheme does not necessarily exists in other countries. Differences in definitions and programs among the countries, and continuous changing of the mix of programs also make direct comparisons difficult.


18. The labor tax wedge is lower than in neighboring countries but unemployment traps appear strong.9 The lower tax wedge implies higher incentives for employers to hire new staff. But, unemployment traps are relatively prevalent with a relatively higher participation tax rate from unemployment across a range of family situations compared to neighboring countries. For instance, in 2014, one-income earner couples with two children resuming work after unemployment at 67 percent of the average wage lose more in tax and reduced benefits than the gross income they earn. In fact, the participation tax rate exceeds 100 percent for this family situation. For a single parent with two children, it is 95 percent, against 75 percent on average in the EU. Replacement rates are also higher for both short-term and long-term unemployed than in neighboring countries across a range of family situations (Figure 3). A one-income earner unemployed couple with two children maintains 100 percent of its previous earnings in the first year and more than 89 percent in the long run.


Average Tax Wedge, 2015

(In percent)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: OECD.

Participation Tax Rate from Unemployment, 2014

(In percent, 67% of average wage)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: OECD.
Figure 3.
Figure 3.

Social Benefits and Labor Cost

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Note: OECD defines the Net Replacement Rate (NRR) as the net income of an unemployed person receiving unemployment and possibly other benefits, expressed as a share of the income earned previously in the job before becoming unemployed. It is measured at different points in time because unemployment benefits decline over the unemployment spell. The Marginal Effective Tax Rate (METR) measures the part of an increase in earnings due to increase in the number of hours worked or to a change in employment situation that is taxed away by personal income taxes and social contributions while taking into account the possible withdrawal of social benefits. The Participation Tax Rate (PTR) is the proportion of gross earnings taken in tax or reduced benefits when an unemployed (or inactive) person gets employed. It is measured by one minus the financial gains to working as proportion of gross earnings.Sources: OECD and IMF staff calculations.

19. Inactivity and low wage traps are also higher than in neighboring countries (Figure 3). Compared to neighboring countries, the participation tax rate from inactivity is the highest in Luxembourg across a range of family situations. In addition, the Marginal Effective Tax Rate (METR) exceeds 99 percent for one-income earner couples with or without children when the pay is below 67 percent of average wage. Thus, it makes more financial sense, for these family situations, to stay inactive or unemployed than to take a job.

20. To enhance the efficiency of ALMPs, further adjustment of benefits to create incentives toward participation is needed. ALMPs are not a magic bullet. The individual decision of joining the labor force or taking up a job also depends on the generosity of social benefits and on the tax system. Hence, there are potential interactions between the welfare benefits, the size and mix of ALMP, and the benefit eligibility conditions in terms of job search and employability. Further steps should be taken to closely monitor unemployment benefits, further link them to job search, and tighten eligibility requirements.

D. Performance of the Educational System

22. Public spending on education in Luxembourg, at 12.5 percent of total government outlays, represents a higher share of government expenditure than in neighboring countries and the Euro Area average. This result remains true even after controlling for school enrollment and living standards. Indeed, in 2012 public spending per student, in percent of per capita GNI, was higher in Luxembourg than in all neighboring countries, and this is the case for all levels of education. 10 In primary and upper secondary education, public spending per student in Luxembourg is double that in Germany.


Government Expenditure in Education, 2014

(In percent of total government spending)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: Eurostat and IMF Staff calculations.

23. The high level of education spending is not reflected in overall education attainment, indicating scope for efficiency orientated reforms. In 2015, in all fields (math, science and reading), students in Luxembourg had lower performance than in neighboring countries. A simple cross-country scatter plot of spending per student in secondary school and average PISA scores shows that Luxembourg spends almost the double per student—even after controlling for living standards—than its neighboring countries, but student test scores are among the lowest (Figure 4). For instance, Luxembourg spent more than 35 percent of per capita GNI per student in 2012, but test scores rank far below those of Japan which spent less than 27 percent of per capita GNI per student.11 This finding holds true even when we focus only on primary education. Indeed, the primary education system in Luxembourg is not perceived to be of better quality than in other countries while it spends far more than almost all European countries. Also, the higher unemployment rate of high skilled workers, compared to Belgium and Germany, suggests the presence of skill mismatches partly reflecting deficiencies in education and training.


Public Spending Per Student by Education Level, 2012

(Percent of per capita GNI)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: OECD, World Bank and IMF Staff calculations.

Primary Education: Spending vs. Quality

Quality of Primary Education (index, 0 to 7, 2015)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: OECD and World Economic Forum.1 Spending data for Germany refers to 2012.
Figure 4.
Figure 4.

Education Input and Outcome

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

24. Requirements to enter the teaching profession are in line with neighboring countries, but less primary teachers have the required qualification. Duration of pre-service teacher training, and teachers’ education attainment are in line with neighboring countries. Teaching practicum is required as part of pre-service training for both primary and secondary education, and teachers are required to be certified, as in neighboring countries. The student-to-teacher ratio is also relatively low in Luxembourg (Figure 4). Moreover, teachers are highly paid by European standards. Indeed, average salaries of lower secondary school teachers with minimum training and 15 years of experience is the highest in Luxembourg compared to neighboring countries, even after controlling for living standards. While higher salaries should help school systems to attract the best candidates to the teaching profession, the requirement of multilingual competency makes the recruitment of fully qualified teachers challenging. For instance, in schools attended by 15-year-olds, a lower percentage of teachers are fully certified in Luxembourg than in neighboring countries. Furthermore, less than a quarter of primary teachers in Luxembourg has the required training against 77 percent in Belgium and 100 percent in Germany and The Netherlands.


Education Attainment of Teachers, PISA 2015

(ISCED 2011 levels)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: OECD.

Teachers with Required Training, PISA 2015

(In percent of teachers, primary education)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: OECD.

25. The lower performance of the educational system can be partially explained by the trilingual curriculum, the diversity of the student population, and the difficulties encountered by socio-economically disadvantaged pupils.12 Luxembourg’s trilingual (Luxembourgish, French, and German) education system is both an asset and a challenge for its highly diverse student population. In addition, students speaking a different language at home than Luxembourgish represent 62 percent of the student population in the academic year 2014–15.13 Students with an immigrant background represent 52 percent of the student population, more than three times their share in neighboring countries.14 Difficulties with the language of instruction lead to failure in other disciplines for numerous students, especially students from families where another language than Luxembourgish is spoken, thus diminishing their chances of academic success. In Luxembourg, average PISA scores of natives and second-generation students are broadly in line with neighboring countries, while first-generation immigrant students perform better than in neighboring countries. The overall lower students’ performance is partially explained by the high share of migrants because students with an immigration background perform less well than natives. In addition, strong correlation exits between socioeconomic status and education performance, with socioeconomically disadvantaged students underperforming compared to their more advantaged peers in all fields.15 Compared to neighboring countries, differences in performances between the two groups are the highest in Luxembourg.


Average of 3 PISA Scores and Migration Background, 2015

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Sources: OECD and IMF Staff calculations.

Socio-Economic Status and Performance, PISA 2015

(Difference of PISA scores between disadvantaged and advantaged students)

Citation: IMF Staff Country Reports 2017, 114; 10.5089/9781475599480.002.A002

Source: OECD

26. Further reforms to improve the quality of education and vocational training are essential to ensure that the educational offer is in line with the skills needed in the labor market. The authorities have undertaken some measures to diversify education and training curricula through the creation of public European schools and public English primary schools, and to improve the quality of early childhood education and care. A newly created National Education Training Institute took over the initial training of teachers from the University of Luxembourg and will also be in charge of in-work training of teachers. An agreement signed between the ministry of education and the University of Luxembourg in June 2016 envisages the creation of the Luxembourg Centre for School Development, which will draft a report on the quality of the educational system and assist schools in developing curricula and creating teaching materials. But, further steps such as on-job training for teachers, and additional support for struggling and socio-economically disadvantaged students are still needed to improve education outcomes.

E. Conclusion

27. Luxembourg’s unemployment rate has been on an upward trend since the financial crisis, despite higher spending on Active Labor Market Policies. The unemployment rate remains relatively high, by historical standards. In spite of innovative measures by ADEM, many young and low skilled persons are not working, and non-European migrants and refugees are less integrated to the labor market. In addition, more than half of new jobs created go to cross border commuters, mainly due to skills mismatches, reflecting deficiencies in education and training. The generosity of the unemployment and social benefits also creates substantial unemployment and inactivity traps.

28. This paper identifies a range of policy options to make growth more inclusive and fully unleash the potential of the economy.

  • Promoting employment. To bring the growth benefits to all, the public employment agency (ADEM) should continue to increasingly target its interventions at the most vulnerable groups in the labor market, notably the young and low-skilled, as well as non-EU immigrants and refugees, including by expanding job search assistance and enhancing the apprenticeship system.

  • Easing integration of refugees. Further steps to speed up diploma recognition, and to provide language classes and other training for refugees are needed to facilitate their integration. Coordination between the Ministry of Family and ADEM is important for an efficient implementation of the Accelerated Integration Programme.

  • Reducing inactivity and unemployment traps. The social benefits system could be rationalized to promote participation and to reduce work disincentives, with the aim to increase the share of take-home income related to activity. Further linking unemployment benefits to job search is also needed to promote active job search and acceptance of available vacancies. The tax system could also be re-designed to increase the wedge between inactivity income and work income, and reduce the speed of the reduction in existing social benefits received as work income rises.

  • Improving education outcomes. To better ensure that graduates are equipped with the skills needed in the labor market, education reforms should focus on upgrading education outcomes in the context of a multi-lingual society with pupils coming from diverse backgrounds, and on improving the quality of vocational training. Enhancing teachers’ participation in professional development activities through practical training initiatives could help strengthening teachers’ knowledge base for teaching. Establishing early-warming mechanism and providing additional instructional support for struggling students, as well as providing additional support for socioeconomically disadvantaged students constitute examples of measures which could improve the efficiency of the educational system.

Annex I. Regression Analysis of Labor Force Survey Data

Data coverage and limitations. The Eurostat Labor Force Survey (LFS) contains yearly and quarterly variables, but the anonymized LFS microdata do not contain the information which would allow tracking people across cohorts because the household numbers are randomized each year. In this analysis, we focus on the yearly dataset. The database contains nearly 85,000 individual responses for 2006 and 14,000 for 2014. LFS data cover residents—Luxembourg natives and former migrants living in the country, but does not cover cross border workers. For the purposes of this study, we identify as “natives” all LFS respondents born in Luxembourg (though some of them have foreign citizenship), and as “migrants” all the respondents who moved to Luxembourg at some point in the past (though some of them have since acquired Luxembourg citizenship).

Variables definition. We are interested in analyzing how individual labor market performance depends on individual characteristics. For example, we want to access how education attainment or age affect the probability to be in or out of a job. We do this by estimating the probability of being un(employed) using a probit regression model as function of individual backgrounds. Estimating both unemployment and employment probabilities helps to assess the effects of labor force participation. Our probit model can be expressed as:



  • yi* is a latent variable that is not observed but determines an outcome. What we actually observe is a labor market outcome yi. In this study, we consider two dummy variables. To estimate unemployment probability, yi is a dummy variable which takes value 1 if the individual “i” is unemployed, and 0 if the individual “i” is employed. For employment probability estimates, yi is a dummy variable which takes value 1 if the individual “i” is employed, and 0 if the individual “i” is unemployed or inactive.

  • X is the set of individual characteristics. We focus on age, gender, education attainment, migration status, years of residency in the country, and household composition.

  • μi is an error term, and β are the coefficients to be estimated.

Estimation. First, we estimate the probability of being (un)employed in 2006 and 2014 for Luxembourg and neighboring countries, except for Germany which is not covered by the Eurostat LFS database and present the results in Table A1. Second, we estimate the likelihood of being in or out of a job across sub-groups of age (15–24, 25–54, 55–64), education levels (lower secondary, upper secondary, tertiary), migration status (native, EU born and non EU born), and gender (male, female). This step allows to assess the joint effect of combining two individual characteristics on the probability of being un(employed). The results are presented in Table A2.

Results and interpretation. When presenting the results, the absolute probability of being (un)employed is shown for the base category in bold, and marginal effects are shown for other categories. This means that the interpretation of the model is relatively easy. The marginal effect is the change in the probability of being (un)employed compared to the base category. For example, Table A1 shows that an individual aged between 25 and 54 years old (the base category) has a probability of 3.6 percent to be unemployed in 2006, and an individual aged between 15 and 24 years old is 13.9 percentage points more likely to be unemployed in 2006 than an individual aged between 25 and 54 years old. So, the total probability of being unemployed for someone aged between 15 and 24 years old is 17.5 percent (3.6 +13.9 percent).

Table A1.

Effects of Individual Characteristic1

article image
article image

Indicates that the result is not significant for p < 0.1

Coefficients represent the change in probability com pared to the base category unless otherwise stated. Estimates are robust to heteroskedasticity.

Table A2.

Joint Effects of Two Individual Characteristics1

article image
article image

Indicates that the result is not significant for p < 0.1

Coefficients represent the change in probability compared to the base category unless otherwise stated. Estimates are robust to heteroskedasticity.


Prepared by William Gbohoui (EUR).


For further details on data and econometric specification, please refer to Annex I.


Germany is not included to the LFS data set.


The analysis in this section is based on the OECD’s tax and benefits model. Definition of work incentives measures are provided in the note at the bottom of Figure 2.


International Monetary Fund, World Economic Outlook, April 2016, Chapter 3.


Fiscal cost and gains from structural reforms, forthcoming IMF SDN.


Grubb et al. (2001), Kluve (2010), Card et al. (2010), Martin (2014) provide a review of the most recent generations of ALMPs in OECD. Kluve (2010) focused on ALMP evaluations in Europe.


Luxembourg 2020, National plan for smart, sustainable and inclusive growth, 2016.


For further details on the social benefits system, see “Addressing Disincentives to Work” in the Selected Issues for the 2014 Article IV Consultation, pages 14–17.


While 2013 is the most recent data available from OECD, Germany data are not available for 2013.


In absolute terms, Luxembourg spent more than 22,000 USD per student in 2012 while Japan spent 10,000 USD.


PISA 2015 defines an index of Economic, Social and Cultural Status (ESCS) using several variables related to students’ family background: parents' education, parents' occupations, proxies for material wealth, and the number of books and other educational resources available in the home. Students are considered socio-economically advantaged (disadvantaged) if their ESCE values are among the top (bottom) 25% students within their country.

Luxembourg: Selected Issues
Author: International Monetary Fund. European Dept.