Jump to ContentJump to Main Navigation
Youth unemployment and social exclusion in EuropeA comparative study$

Torild Hammer

Print publication date: 2003

Print ISBN-13: 9781861343680

Published to Policy Press Scholarship Online: March 2012

DOI: 10.1332/policypress/9781861343680.001.0001

Show Summary Details

(p.212) (p.213) Appendix Samples and attrition

(p.212) (p.213) Appendix Samples and attrition

Source:
Youth unemployment and social exclusion in Europe
Publisher:
Policy Press

Finland

Sample

The criteria for the sampling procedure were the same as in the research design. The statistical representativeness was controlled on the basis of region, unemployment level, unemployment duration, gender and education. The sample was drawn from national unemployment registers, and included young people who were receiving either flat-rate benefits or unemployment insurance payments. The following criteria must be fulfilled in order to be included in the register:

  • 17–64 years of age;

  • capable of work;

  • a jobseeker at the employment office;

  • looking for full-time work.

Attrition analysis

The Finnish register material comprises information concerning the age, residence, education, employment and unemployment of the young people in the study, as well as the municipal unemployment level. It also gives information on spells of unemployment and employment of the young people from 1992 to 1995.

The Finnish sample consisted of 2,386 people. A total of 1,736 young people responded to the questionnaire, which gives a response rate of 73%. Seven age classes were included, the oldest subjects being born in 1970 and therefore being 24 years old at the time of the sample. The youngest were born in 1976 and were therefore 18 years old at the time of the sample. The 19- and 20-year-olds were the groups with proportionately the highest response rate – nearly 80% – whereas the response rate was 70% for the 18- and 23-year-olds. The response rate was lowest among the 24-year-olds: 62%. The overall response rate was 78% for women and 69% for men; women were, therefore, somewhat over-represented in the data. As regards variables such as education and duration of unemployment, the analysis showed that there were no significant differences between the entire sample and the respondents. The attrition analysis on the local unemployment level showed that young people from average unemployment areas (17–22%) were slightly under-represented, whereas those from high unemployment areas (〉22%) were slightly over-represented (see Julkunen and Malmberg-Heimonen, 1998).

(p.214) Iceland

Sample

The sample was selected from the same age groups as in the other Nordic countries, but there were some different sampling procedures. The questionnaire was sent to those who were registered as unemployed at the 20 different unemployment offices throughout Iceland. The sample consisted of young unemployed people who had undergone at least two months of unemployment during the first half of 1995. The total number of unemployed young people is relatively small in Iceland, particularly in comparison with the other Nordic countries (the population of Iceland is approximately 270,000, which corresponds to the population of Bergen). Due to the small numbers of unemployed, and particularly long-term unemployed young people in Iceland, the research was conducted as a total study in which virtually every person who was unemployed at the time of the sampling was included. This strategy created some problems in the statistical analysis, which assumes random sampling. The following rules are applicable for being registered as unemployed in Iceland:

  • The person must be a wage-earner who has worked at least 425 hours during the previous 12 months before becoming unemployed (and also has the right to receive compensation from the unemployment insurance fund).

  • The person must be looking for work, older than 16 years but younger than 70 years, and live in Iceland or in another EEC country.

  • The person must be able to provide certification from an employment office that he or she been unemployed for at least three whole days at the beginning of the period for which the unemployment compensation is applied.

Private entrepreneurs have the same rights as wage-earners, provided that they fulfil all the basic criteria and can prove that their business has been closed down. The questionnaire was answered by 1,290 persons, which gave a response rate of 60%. The sample was 53% male and 47% female, with 70% of the women and 61% of the men answering the questionnaire. Thus, women are slightly over-represented in the material.

Norway

Sample

A total of 97,934 young people aged 18–24 years were registered as unemployed at some time during the first half of 1995, and 19% of these fit the definition of long-term unemployed (〉3 months). The sample was selected from among those who had had at least three months of continuous unemployment during the period 1 January to 30 June 1995, and who were looking for full-time work. The group consisted of 39,020 persons, of whom 17,909 were unemployed at the time of the sample. From this population 2,000 people (p.215) were selected. The sample seems to be representative of the population (39,020) in terms of key characteristics that can be controlled through the register. When it comes to age there was a slight over-representation of the older age groups in the sample.

Attrition analysis

Altogether 1,106 people answered the questionnaire, yielding a response rate of 56%. The register data were coupled with information about those young people in our study who had given their permission (85%; n = 944). It is possible to use register information of the whole sample (n = 2,000) to analyse eventual skewness that may affect the possibility of making generalisations from the sample.

In the attrition analysis we compared the sample with the respondents on an extensive set of register information. There were no differences between the sample and the respondents with regard to proportion who had received benefits, length of the unemployment period, total experience of unemployment, dropout from compulsory schooling, previous work experience, proportion without any relevant work experience or education, and place of residence. The only difference was that 22.9% of the sample had only compulsory school education or had no qualifications, compared with 16.1% among the respondents. The difference is statistically significant (z = 4.47). However, a larger proportion of the respondents had only one or two years of vocational education, compared with the overall sample, in which a greater proportion had completed a full vocational education. There were no differences with regard to other educational categories. Contrary to expectations and despite a low response rate, the attrition cannot be considered skewed.

Sweden

Sample

The criteria applied to the sample were the same as in the research as a whole. The sample was selected through the AMS (the Labour Market Board), which created a random sample among people registered as unemployed in HÄNDEL, the database on people actively seeking work. There are no formal limitations to being included in the register, other than attending an employment office to report looking for a job. The next phase consisted of coding the different categories, for example, student, working, unemployed. The sample therefore consisted of people registered as unemployed (which excludes full-time students and retired persons).

Attrition analysis

There was a total of 801,093 young people aged 18–24 in Sweden in 1995. During the first half of the year there was an average of 82,000 unemployed (p.216) people per month aged 16–24 (Labour Force Survey), yielding a sample of 1.2% of the unemployed young people during the sample period.

The sample comprised 4,000 people (two people were excluded because of technical problems, and the net sample was therefore 3,998). There were altogether 1,853 women (46%) and 2,147 men (54%). The questionnaire was answered by 2,534 persons (1,247 women and 1,287 men), yielding a response rate of 63%. The questionnaire was answered by 49% of women and 51% men. The response rate among the women was 67% and 60% among the men.

The attrition consists of 1,084 persons. An attrition analysis was carried out on the basis of information from the register material. Due to technical problems, however, a comparison between the sample and the respondents could not be completed for all cases. Information on 191 respondents (97 women and 94 men) was not included in the attrition analysis.

The attrition analysis showed that there were no statistical differences in citizenship, work handicaps and place of residence. Men turned out to have been unemployed for an average of 22.8 weeks compared with 20.9 weeks among the respondents. Unemployment among women averaged 8.9 weeks in the attrition group and 9.7 weeks among the respondents.

Denmark

Sample

The sample was randomly drawn from the Central Unemployment Register (CRAM), in which all unemployed are currently registered. Most unemployed people (about 85%) and the main part of the labour force (about 80%) are members of an unemployment insurance fund. Both insured and non-insured are registered. The non-insured unemployed are those receiving social assistance from the local authorities. However, many people receiving social assistance are not registered as unemployed in the Central Unemployment Register. This means that some non-insured young people without employment are not registered as unemployed. Therefore, and because of the limited size of sample, it was decided only to include insured young unemployed people in the study.

The population from which the Danish sample was drawn was defined in the following way:

  • Insured people of age 19–24 years (at 1 January 1995) with a total of less than three months’ unemployment in the second half of 1994, and with more than 13 weeks of unemployment in the last 26 weeks before weeks 1–26 in 1995.

Consequently, the population consists of young insured people with more than three months of unemployment in the last 26 weeks before weeks 1–26 in the first half of 1995. Or put more simply, the Danish population consists of young insured persons having been unemployed more than three months. From this population (about 12,000 persons) a simple random sample of 1,500 persons was drawn. Of these, 19 persons had invalid person identification numbers. (p.217) Consequently, the effective sample consisted of 1,481 persons to whom the questionnaire was sent. Those who did not answer were contacted by interviews (by telephone). The questionnaire was answered by 1,171 persons, which gave a response rate of 79%.

Attrition analysis

The questionnaire was answered by 83% of the women and 78% of the men. This difference is statistically significant (p〈0.001). The response rate did not depend on age, but young people from the eastern part of Denmark (Copenhagen and the islands) answered generally to a lesser extent than young unemployed people in the western part of Denmark. There were no statistically significant differences with regard to the duration of previous unemployment.

As mentioned, the Danish sample only includes insured people. In general, it is to be expected that the insured unemployed category comprise ‘stronger groups’ than non-insured. However, there exist no recent nationwide studies on the composition of insured and non-insured unemployed people and the mobility between these groups. There is a clear need for research and statistics in this area in Denmark.

Scotland

Sample

The sample was selected from the same age groups as in the Nordic countries, but there were different sampling procedures. In the Nordic countries, the sample was collected through the unemployment registers, while in Scotland interviewers were placed in a representative range of unemployment benefit offices throughout the country. All the young people had been unemployed for a minimum of three months at time of first contact. Postal questionnaires were completed six months after sampling, at which point some young people had found jobs, entered schemes or returned to education, while others remained unemployed or were experiencing a further spell of unemployment. The questionnaire was completed by 817 respondents, which gave a response rate of 56%. The sample consisted of 65% men and 35% women, which reflects the actual proportions of men and women unemployed in this age group in Scotland.

Attrition analysis

The attrition consisted of 629 individuals. The attrition analysis could only be based upon gender, area of residence (rural or urban) and length of unemployment, as we did not have access to unemployment register data. There were slightly more men who failed to respond than women, although this was not statistically significant. The length of unemployment did not affect response rates, but there were more non-respondents living in poorer urban areas than in rural areas, although again this was not statistically significant.

(p.218) Italy

Sample

The survey was carried out from March to June 2000 on a sample of 1,421 youths (aged 18–24) registered as unemployed a year before, living in Campania and Veneto. (Note: the young teenagers, aged 14–17, included in the UN definition of young people, were excluded from the sample). Another stratification criterion was adopted, after the interview quotas had been established, based on gender and place of residence, in such a way that the final sample should be statistically representative of the underlying population in the regions considered. In 1999, the reference period of the survey, the unemployment rate was at 6.5% in central and northern regions and at more than 22% in southern regions. More dramatically, the youth unemployment rate (15–24 years) was 19% in the north and in the centre, but over 56% in the south. The long-term unemployment rate (more than one year) was 3.1% in the centre-north and 14.8% in the south. To mirror this situation, the sample has been stratified, with two thirds of the interviews being concentrated in the south. The two regions considered represent very different labour market contexts, with Campania being one of the highest and Veneto one of the lowest unemployment regions in the country. The sample was selected among individuals registered at the local unemployment office for at least three months at the time when the sampling procedure was carried out, a year before the interview. The main problem arising from this procedure is that jobseekers enrolled in placement registers do not completely overlap with all the unemployed people recorded in the official Labour Force Survey, the so-called Rilevazione trimestrale delle forze di lavoro (RTFL). Among other reasons, this depends on the fact that enrolment in the registers is only one of the various criteria requested of those actively seeking a job. In fact, first, it is possible to be registered as unemployed while being employed as a part-time (for less than 20 hours per week) or temporary (less than four months) worker. Moreover, people enrolled in the registers are not always properly qualified as jobseeking unemployed. It is possible to register even if one is not actively seeking employment, as stated in the ILO definition (search activity in the month before the interview and immediate availability to work).

Data were collected by means of direct interviews. Interviewers got in touch with people during a period of four months – from March to June 2000 – through the ‘chain rule’, exploiting their own direct or indirect acquaintance network and contacts with public or private institutions involved in supporting the analysed population. Interviews were anonymous, but spot checks were carried out, to test the effective submission of the questionnaire. Of those interviews, 1,421 turned out to be valid and complete after verification. They represented 447 of the 500 envisaged in Veneto and 974 of the 1,000 envisaged in Campania.

(p.219) Attrition analysis

A comparison between registered unemployed in the Labour Force Survey (LFS) and in the YUSE data can be carried out, keeping in mind that the available data refer to young people aged 15–25, living in Italy, rather than aged 18–24, living only in two regions (Campania and Veneto). The period is almost the same: October 1999 in the case of the LFS and the second semester of 1999 in the Youth Unemployment and Social Exclusion Survey (YUSE) case. Another caveat regards the type of question asked in the two questionnaires. The YUSE questionnaire contemplates the following labour market statuses: employment, unemployment, school or university attendance, training, socially useful contacts, unpaid family work, compulsory military service, sickness and other unmentioned status. The LFS contemplates the following alternatives: unemployment; employment; seeking a job, but not actively; not seeking a job, but available to work; not seeking a job and not available to work.

In the YUSE case, it was found that 18% of the sample is unemployed, 36% employed, 44% involved in high secondary or tertiary education and 3% in other activities. In the LFS case, Barbieri and Schever (2001), 35% are unemployed, 11% are employed, 27.6% are not actively seeking a job but are available to work and 26.7% are not participating to the labour market.

Considering the differences existing in the questionnaire, the disparity in the YUSE and LFS shares can be explained as follow: (a) many young people considered as employed in the former dataset are actually involved in informal or occasional activities; (b) those involved in university education are essentially unemployed jobseekers, not available to work or out of the workforce. It should be taken into account that out of 5.6 million people, 900,000 are university students, which corresponds to about 40% of the population aged under 25. Therefore, spreading the university students over the other groups and considering that the employed include workers not considered in the LFS, the distribution of individuals is very similar in the two surveys.

Spain

Population and sample

During 1998, the average number of young people registered as unemployed in Spain was 376,056. Of these, 45.83% had been unemployed for less than three months, and 54.17% were long-term unemployed. The population consisted of young people aged 18–24 who had been registered as unemployed during 1998 for at least three consecutive months. The sample was randomly selected from the database of the INEM (National Employment Institute). The INEM is an organ of the Ministry of Labour that registers additions to and deletions from the list of people making social security contributions. In Spain, unemployed people are not obliged to register with the INEM.

The Spanish sample was selected in February 1999. The data collection process began in February 2000, one year after selection of the sample, and (p.220) ended in June 2000. During this period, the questionnaire was sent to the selected young people on three occasions.

Attrition analysis

The questionnaire was sent to 5,000 young people, 3,090 (62%) women and 1,910 (38%) men. The final sample was made up of 2,523 young people, 966 men (38,3%) and 1,557 women (61.7%). The response rate obtained was therefore 50.46%. The attrition analysis was limited to gender and level of education, as it was not permitted to use the register information. The response rate did not depend on gender. The questionnaire was answered by 50.6% of men and 50.4% of women.

There were significant differences between the sample and the respondents with regard to level of education (see Table A.1). People without qualification and those with elementary education responded to a smaller extent than the other groups. The percentage of people with secondary-level education was higher among the respondents (p〈0.01). There were not significant differences between the sample and the respondents with regard to the other educational categories.

Because we did not obtain permission to access the register information, we could not complete the attrition analysis on length of unemployment. Nevertheless, we could obtain an approximate analysis of distortion by comparing

Table A.1: Differences between the sample and the respondents (%)

Respondents

Sample

Gender

Male

38.3

38.3

Female

61.7

61.7

Education

No qualification

1.2

1.9

Primary level

30.0

36.4

Professional education (first degree)

11.8

11.0

Professional education (second degree)

13.9

13.6

Secondary level

15.7

13.3

University diploma

16.3

15.9

University degree

8.6

7.9

Samplea

Population

Length of unemployment (months)

3–6

36.5

36.9

6–12

22.3

26.3

12–24

14.5

20.9

〉24

26.0

15.9

Note: (a) Only those currently unemployed for more than three months were included.

(p.221) the respondents’ data with the official statistics for the entire population. As seen in Table A.1, there are no differences between the entire population and the respondents when comparing the percentage of people who had been unemployed for less than six months. However, people with longer periods of unemployment are under-represented among the respondents, with the only exception being of periods longer than 24 months.

Germany

The German sample population comprised young people aged 18–24 at the beginning of a spell of continuous unemployment that lasted for at least 92 days between September 1998 and September 1999 and who were registered at the unemployment registry of the German Federal Employment Services (Bundesanstalt für Arbeit).

A sample of 3,200 young people was drawn in autumn 1999 and 1,918 interviews (a response rate of around 60%) were conducted between March and November 2000.

Data were collected by 80 qualified interviewers supported by a supervisor and using computer-assisted telephone interviews (CATI) conducted by Infas (Institut für angewandte Sozialforschung GmbH), Bonn. A sample of interviews was tape-documented and controlled by the IAB (Institute for Employment Research). Interviewers received study-specific training, in which the IAB research group participated.

Interviews lasted an average of 65.1 minutes. Each participant was informed about the telephone interview by an official letter and received as an incentive a telephone card valued at €6, which had been designed specifically for the study.

The master sample contains information about age, gender, nationality and region of the unemployed. This information was used for selectivity tests. Logistic regression models were estimated, to identify selectivity effects for the group of respondents versus non-respondents. The explained variance for all the calculated models is extremely small, which indicates poor selectivity effects.

Estimates of selectivity effects for CATI participants versus the group of non-respondents (Pseudo R2 0.008) contained weak significant effects for a small subgroup with missing data and for a small group of the long-term unemployed (aged 26 years plus, meaning that the observed spell of unemployment started long before September 1998).

Estimates of selectivity for CATI participants versus the young unemployed, who intended to participate but did not in fact participate because they were not at home at the time of the interview or because they said they did not have the time to participate. This group was called latent non-participants. Again, the estimated model discriminates poorly (Pseudo R2 0.005), which means no observable selectivity effects.

(p.222) France

The French sample used for the YUSE project was extracted from ANPE (Public Employment Service) and UNEDIC (Unemployment Benefit Agency) register data files. As required by the overall design of the YUSE samples in all the countries, the French sample has the following characteristics:

  • it consists of young people between the ages of 18 and 24;

  • the young people have been unemployed for at least three months between 1 June 1998 and 1 June 1999;

  • the sample must be representative by region of residence, level of educational attainment and gender.

As a result, we obtained an initial database to sample from of 25,013 cases and the breakdown by educational attainment is given in Table A.2.

Despite the fact that both the CAP and BEP diploma appear at level V in the French classification system1, they were kept apart from the rest of level V so that there were enough respondents at all the relevant levels. Both CAP and BEP are very important diplomas in France.

The breakdown by region is given in Table A.3 and by gender in Table A.4.

Table A.2: Initial database, by aggregated level of educational attainment

Level of education

French classification

ISCED

Frequency

%

Cumulative Frequency

Cumulative %

Missing

Missing

2

0.01

2

0.01

BEP diploma

2

5,151

20.59

5,153

20.60

CAP diploma

2

4,480

17.91

9,633

38.51

Levels I, II and III

5–7

4,938

19.74

14,571

58.25

Level IV (Bac)

3

7,074

28.28

21,645

86.54

Level VI and other V

1 and 2

3,368

13.46

25,013

100.00

Table A.3: Initial database, by region of residence

Region

Frequency

%

Cumulative frequency

Cumulative %

Missing

1,956

7.82

1,956

7.82

1. Île de France (Greater Paris)

3,527

14.10

5,483

21.92

2. Unknown

2,190

8.76

7,673

30.68

3. CEst (East)Œ

1,791

7.16

9,464

37.84

4. Loire (LoireValley)

3,156

12.62

12,620

50.45

5. Bourgogne (Burgundy)

2,032

8.12

14,652

58.58

6. Normandie (Normandy)

2,201

8.80

16,853

67.38

7. Aquitaine

2,459

9.83

19,312

77.21

8. Centre

2,609

10.43

21,921

87.64

9. PACA (Alps and Provence)

3,092

12.36

25,013

100.00

(p.223)

Table A.4: Initial database, by gender

Gender

Frequency

%

Cumulative frequency

Cumulative %

Male

11,130

44.50

11,130

44.50

Female

13,883

55.50

25,013

100.00

Given the budget, a random draw of 4,000 units was carried out. Those 4,000 raw cases yielded 2,001 completed cases, available for the analysis. The survey was carried out from March to May 2000 by a subcontractor (CSA) and conducted by telephone (CAPI). The mean duration of the questionnaire was 29 minutes. Since the survey is retrospective and is not a panel data survey, we were not faced with the usual problem of attrition. However, and for the usual reasons – not at home, moved, refused and so on – there is some kind of non-response (see Table A.5). Finally, the structure of the sample used for the analysis is given in Table A.6.

Table A.5: Respondents and reasons of non-response (%)

Completed cases

51.2

Not at home

8.8

Not at home for the entire survey period

6.3

Moved

5.0

Not applicable

19.2

Refuse to answer (several reasons)

9.5

All

100.0

Table A.6: The final sample, by level of educational attainment

Level of education

French classification

ISCED

Frequency

%

Cumulative frequency

Cumulative %

Level VI and other V (without diploma)

0–1

372

18.59

372

18.59

CAP and BEP diploma (V)

2

577

28.84

949

47.43

Level IV (Bac level)

3

595

29.74

1,544

77.16

Levels I, II and III

5–7

400

19.99

1,944

97.15

Other

Other

57

2.85

2,001

100.00

As in the questionnaire, CAP and BEP diplomas are grouped together. As a consequence, we cannot distinguish the two items in Table A.5 either. The structure by level of education is quite different from the structure of the initial database. Since all the other levels of educational attainment are represented in the same proportion in the sample as the initial database, the proportion of individuals holding a CAP or a BEP is less than expected.

(p.224) The final breakdown of the sample by gender is given in Table A.7.

As a comparison between Table A.4 and Table A.7 shows, the structure by gender is slightly modified in the sample: the proportion of women is only slightly higher in the sample but it does not impact on the quality of the sample.

Table A.7: The final sample, by gender

Gender

Frequency

%

Cumulative Frequency

Cumulative %

Male

844

42.18

844

42.18

Female

1,157

57.82

2,001

100.00

Notes:

(1) The French social classifications are as follows:

  • VI Personnel carrying out jobs that do not require any training beyond the end of compulsory schooling

  • Va Personnel holding jobs that presume a short training period of less than one year leading notably to a Vocational Education Certificate or any other certification of the same nature

  • V Personnel holding jobs that normally require a training level equivalent to that of the Vocational Studies Certificate (BEP) and the Vocational Certificate (CAP)

  • IV Personnel holding supervisory jobs or having a qualification at a level equivalent to that of technical baccalaureate

  • III Personnel holding jobs that normally require training at the level of the higher technician's certificate (Bts) or diploma from the university institutes of technology (Iut), at the end of the first cycle of higher education; including Deug

  • II and I Personnel holding jobs that require training at a level equal or superior to licence or Grandes Ecoles