43
Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak, Heike Wirth Mannheim, MZES and ZUMA ESeC Validation Conference 18 – 20 January 2006 Portuguese Statistical Office, Lisbon, Portugal

Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

Embed Size (px)

Citation preview

Page 1: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

Validation of ESeC:The Effect of Coding Procedures and Occupational

Aggregation Level

Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak, Heike Wirth

Mannheim, MZES and ZUMA

ESeC Validation Conference

18 – 20 January 2006

Portuguese Statistical Office, Lisbon, Portugal

Page 2: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 2

Overview

1 Criterion Validation of ESeC matrix (4-digit)

2 Aggregation of ESeC: 2-/3digit matrices

3 A Comparison between ESeC and EGP

4 Class Effects on Risk of Unemployment

Page 3: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 3

1 Criterion Validation: Operational Issues

• Database: BIBB/IAB (1998/1999) and GSOEP (2001)

• Employment Status

– Problem: Concept of Managers/Supervisors not well established

– Different distinction: „Position with Employer“ (PwE)

– Self-Employed

– Managers: ISCO 11-, 12, and 13 2-digit codes

– Supervisors: standard (direct measurement) vs. proxy (PwE)

• Employment Relation Indicators

– Monitoring problems

– Asset Specificity

– Long-Term Employment

– Career Prospects

Page 4: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 4

1 Criterion Validation: Approach

1. Calculation of mean ER-score for each ESeC class

2. Calculation of mean ER-score for each combination of ISCO code and employment status (OUG)

3. Comparison of the mean ER-score of ESeC classes with the mean ER-score for each OUG

4. Reallocation of class codes if evidently suggested by the ER-indicators

e.g. Religious Professionals

changes for 36 ISCO-categories, 10.3% respondents reallocated

5. Test whether the revised matrix performs better: increase of within-class homogeneity and between-class heterogeneity

ISCO(4) SE 10+

SE <10

SE NO

MAN>10

MAN<10

SUP EMP

2460 1 2 2 Illicit Illicit 2 2

Page 5: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 5

1 Criterion Validation: Approach

1. Calculation of mean ER-score for each ESeC class

2. Calculation of mean ER-score for each combination of ISCO code and employment status (OUG)

3. Comparison of the mean ER-score of ESeC classes with the mean ER score for each OUG

4. Reallocation of class codes if evidently suggested by the ER-indicators

e.g. Religious Professionals

changes for 36 ISCO-categories, 10.3% respondents reallocated

5. Test whether the revised matrix performs better: increase of within-class homogeneity and between-class heterogeneity

ISCO(4) SE 10+

SE <10

SE NO

MAN>10

MAN<10

SUP EMP

2460 1 1 1 Illicit Illicit 1 1

Page 6: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 6

Overview

1 Criterion Validation of ESeC V3 matrix (4-digit)

2 Aggregation of ESeC: 2-/3digit matrices

3 A Comparison between ESeC and EGP

4 Class Effects on Risk of Unemployment

Page 7: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 7

2. Aggregation of ESeC 2-/3-digit: Approach

• Starting base for generating the 3-digit and 2-digit matrices:

– revised German 4-digit matrix

– German modal values based on BIBB/IAB

Stock/Production/Transport ClerksISCO(4) SE

10+SE

<10 SE

NO MAN>10

MAN<10

SUP EMP

4131 1 4 4 Illicit Illicit 6 (68) 7 (132)

4132 1 4 4 Illicit Illicit 2 (75) 3 (179)

4133 1 4 4 Illicit Illicit 2 (62) 3 (55)

ISCO(3) SE 10+

SE <10

SE NO

MAN>10

MAN<10

SUP EMP

413 1 4 4 Illicit Illicit 2 3

Page 8: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 8

2. Aggregation of ESeC 2-/3-digit: Findings

Table 2.4: Proportion of Correspondence Between Different Aggregation Levels of ISCO in %.

• Effects of the aggregation for the performance of ESeC with respect to the ER-indicators?

4-digit vs. 3-digit 93.8

4-digit vs. 2-digit 84.6

Page 9: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 9

2 Aggregation of ESeC 2-/3-digit: ER-Indicators (Figure 2.1)

Work Autonomy: Means of Factor Scores

Asset Specificity: adjusted highest degree: % requiring foreman education/college

4-digit 3-digit 2-digit 4-digit 3-digit 2-digit R² 0.142 0.136 0.127 Pseu R 0.251 0.232 0.234

Page 10: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

Asset Specificity: adjusted highest degree: % requiring any degree

Asset Specificity: introduction to the job insufficient

4-digit 3-digit 2-digit 4-digit 3-digit 2-digit Pseu R 0.170 0.162 0.149 Pseu R 0.098 0.094 0.091 Long-Term-Employment: Tenure Residuals

Career Prospects: % having obtained further education

4-digit 3-digit 2-digit 4-digit 3-digit 2-digit R² 0.054 0.049 0.042 Pseu R 0.118 0.113 0.101

Page 11: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 11

2 Aggregation of ESeC: Implications of Modal Rule I

• Problem: UK modal value does not correspond to German modal value

Stock/Production/Transport Clerks

ISCO(4) SE 10+

SE <10

SE NO

MAN>10

MAN<10

SUP EMP

4131 1 4 4 Illicit Illicit 6 7

4132 1 4 4 Illicit Illicit 2 3

4133 1 4 4 Illicit Illicit 2 3

ISCO(3) SUP EMP

Germany 413 2 3

UK 413 6 7

Page 12: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 12

2 Aggregation of ESeC: Implications of Modal Rule II

for a harmonized 3-digit matrix, country specific modal values should be taken into account

Table 2.5: Proportion of Discrepant Class Codes Depending onthe Country-Specific Modal Value. BIBB/IAB, 1999.

3-digit 7.5%

2-digit 10.4%

Page 13: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 13

Overview

1 Criterion Validation of ESeC V3 matrix (4-digit)

2 Aggregation of ESeC: 2-/3digit matrices

3 A Comparison between ESeC and EGP

4 Class Effects on Risk of Unemployment

Page 14: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 14

3 A comparison of ESeC and EGP

• Primary focus:

– Quality of the crosswalk from the German National Occupational Classification (KldB) to ISCO88

– KldB (Klassifikation der Berufe)

• standard used by official statistics in Germany

• some official data (e.g. LFS) also include ISCO (generated by a mapping procedure KldB => ISCO)

– Researchers either use ISCO or KldB, seldom both

• no empirical knowledge which kind of occupations might pose crosswalk problems

– ESeC in the LFS will be based on the crosswalk KldB=>ISCO

=> Implications for ESeC ?

Page 15: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 15

3 A comparison of ESeC and EGP

• EGP: intermediary tool to examine crosswalk problems

– well examined coding routine to generate EGP directly from the KldB

Step 1:

– EGP_ISCO: identical coding routine, but KldB codes were replaced by ISCO_3d codes

– Comparison of EGP_KldB and EGP_ISCO

=> where are movements between classes and why ?

Step 2:

– Comparison of ESeC and EGP_ISCO

=> differences in the allocation of occupations to EGP and ESeC classes:

• crosswalk problems, conceptual and other reasons ?

Step 3:

– Employment relation indicators: ESeC and EGP compared

Page 16: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 16

Step 1: Correspondence between EGP_KldB and EGP_ISCO

• Main findings: – marginal distributions of EGP_KldB and EGP_ISCO nearly identical

– agreement rate diagonal cells: 92%

– mismatches mostly concentrated on neighbour classes (e.g. I => II; IIIa=> V)

– more ‚irritating‘ mismatches (e.g. II => V) are restricted to a few cases

• Further checking on the data shows that mismatches are

– mainly due to the loss of information because of different aggregational levels

• KldB_3d (369 categories) to ISCO_3d (116 categories)– only to a very small extent due to (german specific) crosswalk problems

• e.g. upper secondary teachers/Gymnasiallehrer can not be distingished from other secondary teachers

=> high correspondence rate between EGP_KldB and EGP_ISCO – crosswalk is not a problem it is a problem of the aggregational level

– a better job could be done, if ISCO_4d were available in the data

=>

Page 17: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 17

Step 2: Correspondence between ESeC_3d and EGP_ISCO

ESeC row %

I II IIIa V IIIb VI VIIab IVab IVc %service 1 86.4 5.1 0.6 4.4 3.5 10.5contract 2 21.8 47.9 17.4 8.9 1.6 1.6 0.7 25.2mixed 3 0.5 27.8 62.3 6.4 3.0 16.2contract 6 0.1 8.0 3.0 40.4 6.2 26.9 15.5 9.1labour 7 1.6 14.4 5.7 52.4 13.9 12.1 7.6contract 8 0.2 12.6 63.9 23.2 12.3

9 0.3 2.4 4.7 6.0 19.0 67.6 11.5self 4 6.0 4.7 89.3 7.3employed 5 100 0.2

15.1 18.4 16.1 8.5 6.7 14.4 13.0 7.2 0.6 100

Same contract types (service, mixed, labour) are indicated by shaded cells

EGP ISCO_3d

Self employed

 %n = 33,605. Agreement rate (diagonal cells;n = 20,606 ) : 61,3%; Spearman's rho: 0.85Source: BIBB/IAB 98/99

ISCO(3_d) Service contract Mixed contract Labour Contract

=>

Page 18: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 18

Step 2: Correspondence of ESeC_3d between EGP_ISCO

• Further analysis show that the deviations between ESeC and EGP are

– mainly due to effects of the employment status variable used:

• EGP class coding is based on the ‚status within employment‘ variable – differentiates between blue- and white-collar workers, civil servants and

management positions, and

– within these groups there is a further distinction between different hierachical levels

• ESeC is based on the ‚supervisory‘ concept– manager, supervisors, employees

=> ‚status within employment‘ variable thus enables a finer assignment of occupations to classes

=> =>

Page 19: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 19

Step 3: Employment Relation Indicators: ESeC – EGP

ER Indicators Expl. Revised ESeC EGP Variance 2-digit 3-digit 4-digit ISCO KldB Work Autonomy R² 0.127 0.136 0.142 0.145 0.151 Asset Specificity

1. adjusted highest degree

Pseudo R 0.234 0.232 0.251 0.288 0.294

2. requiring any degree Pseudo R 0.149 0.162 0.170 0.206 0.214 3. introduction to job

not sufficient Pseudo R 0.091 0.094 0.098 0.139 0.143

Long Term Employm. R² 0.042 0.049 0.054 0.047 0.050 Career Prospects Pseudo R 0.101 0.113 0.118 0.122 0.123

=> Model fit for EGP for nearly all ER-indicators higher than for ESeC versions

• ‚Status within employment‘ variable leads to greater homogeneity within the classes and a larger variance between the classes than the supervisory variable

=>

Page 20: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 20

Overview

1. Criterion Validation of ESeC V3 matrix (4-digit)

2. Aggregation of ESeC: 2-/3digit matrices

3. A Comparison between ESeC and EGP

4. Class Effects on Risk of Unemployment

Page 21: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 21

4 Class effects on risks of unemployment

• Construct validation analysis: class effects on risks of unemployment

1. comparison of the prototype and the revised German matrix

2. effects of different levels of aggregation (ISCO 4,3,2 digit)

3. ESeC compared to EGP

– Dependent variable

• have been unemployed in the past at least once (dummy coded)

– Control variables

• age, gender, Geman citizenship, part-time, East-West Germany; educational level (CASMIN)

=>

Page 22: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 22

Risk of unemployment : Main Findings I

• Model fit

– Pseudo-R2 nearly identical for the different versions

• EGP (Kldb; ISCO) slightly better than all ESeC versions

• Class effects

– clear hierarchy regarding the risk of unemployment

– ranking of classes nearly constant for all versions

=>

Page 23: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 23

Risk of unemployment : Main Findings II

• Effect size:

– larger for EGP than for ESeC, and

• Unemployment risk of EGP farmers much lower than for ESeC farmer

• EGP : unemployment risk for class IIIb (ESeC 7) higher than for class VI (ESeC 8)

– a similiar pattern is found only for the revised ESeC matrix (4_d)

=> further analysis needed to examine whether these findings

– an effect of the ‚status within employment‘ variable, or

– an effect of differences in the allocation of specific occupations to classes

=>

Page 24: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 24

Conclusions

• ESeC will be an important improvement of international comparative research, but there are a few issues left where a further discussion is needed:

1. Supervisory concept concept should be clarified and an international agreed procedure should be established

2. Aggregation Level of ISCO– the higher the aggregation level of ISCO used for ESeC the lower the within-class

homogeneity A future ESeC (at least in LFS) should be based on ISCO-4digit

– model fit of the aggregated matrices are slightly better when country specific modal values ared, but also comes along with a higher between-countries heterogeneity

=> what is the lesser evil?

3. Procedural or Interpretative equivalence ?

– In sum the revised German matrix reveals only a slightly better model fit than the prototype but there are some occupations where we would prefer a change in the ESeC prototype version

equivalence of procedure: identical matrix for all countries?

=> equivalence of meanings: allowing a certain amount of variation in the matrix between countries?

Page 25: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 25

Correspondence between EGP_KldB and EGP_ISCO

EGP EGPKldB_3d I II IIIa IIIb IVab IVc V VI VIIa VIIb %I 92,7 6,7 0 0 0,4 0 0,2 0 0 0 15,8II 1,5 91,9 2,3 0,1 0,3 0 3,9 0 0 0 18,0IIIa 0,3 2,2 93,3 1,1 0 0 2,6 0,4 0,1 0 16,1IIIb 0 0 2,8 75,6 0 0 0,3 9,3 9,5 2,5 7,9IVab 0,9 2,4 0 0 95,5 1,1 0 0 0 0 7,3IVc 0 0,6 0 0 0 99,4 0 0 0 0 0,5V 0,2 2,6 2,1 0 0 0 93,8 1,1 0,3 0 7,7VI 0 0 2,7 0,8 0 0 0,3 95,4 0,9 0 14,1VIIa 0 0 0,1 3,7 0 0 0,1 1,4 94,7 0 11,7VIIb 0 0,3 0,3 0 0 0 0 0 0,3 99,0 0,9% 15,0 18,3 16,2 6,7 7,1 0,6 8,4 14,5 12,0 1,1 100

ISCO(3_d)

n = 34,203. Agreement rate (diagonal cells) : 92.3% (n = 31,403)Spearman's rho: 0.95Source: BIBB/IAB 98/99

<=

Page 26: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 26

Correspondence between ESeC_3d and EGP_ISCO (col %)

col %

I II IIIa V IIIb VI VIIab IVab IVc %service 1 60.0 2.9 0.7 6.4 61.2 10.5contract 2 36.5 65.8 27.3 26.3 6 2.8 0.1 2.5 25.2mixed 3 0.5 24.4 62.5 15.4 3.4 0.1 16.2contract 6 0.1 4 1.7 43.3 8.4 17 10.9 9.1labour 7 0.6 6.8 5.1 59.9 7.3 7.1 7.6contract 8 0.1 18.2 54.3 21.9 12.3

9 0.2 1.7 6.3 10.3 15.2 60 11.5self 4 2.9 1.9 91.0 7.3employed 5 38.8 0.2

15.1 18.4 16.1 8.5 6.7 14.4 13.0 7.2 0.6 100

Same contract types (service; mixed, labour) are indicated by shaded cells.

n = 33,605. Agreement rate (diagonal cells) : 61,3% (n = 20,606)Spearman's rho: 0.85Source: BIBB/IAB 98/99 (bibbes_neu_234_avs_egp.sav; class_im; esec3neu)

Self employedISCO(3_d)

 %

ESeC Service contract Mixed contract Labour Contract

EGP ISCO_3d

<=

Page 27: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 27

Position within employment – Employment Status

Position within Employment -col % row

EGP Man>10 Man<10 supervisors employees n %Routine employee; semi-skilled worker 1.0 3.3 7.3 23.1 5420 18.2Employee with simple tasks; skilled worker 1.9 4.3 16.4 31.8 7954 26.8Employee with complex tasks; foreman 7.6 9.2 26.0 25.3 7378 24.8Experts; master craftsman 51.6 37.0 32.0 14.5 6028 20.3head of division 3.5 1.1 5.5 2.9 1059 3.6head of department 31.1 44.0 10.4 1.1 1387 4.7director; head official 3.4 1.1 2.5 1.3 493 1.7n 892 184 7654 20989 29719 100.0col % 3.0 0.6 25.8 70.6 100.0

Employment Status - ESeC

<==>

Page 28: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 28

Position within employment – Employment Status

Position within Employment -row % row

EGP Man>10 Man<10 supervisors employees n %Routine employee; semi-skilled worker 0.2 0.1 10.3 89.4 5420 18.2Employee with simple tasks; skilled worker 0.2 0.1 15.8 83.9 7954 26.8Employee with complex tasks; foreman 0.9 0.2 27.0 71.9 7378 24.8Experts; master craftsman 7.6 1.1 40.6 50.6 6028 20.3head of division 2.9 0.2 39.8 57.1 1059 3.6head of department 20.0 5.8 57.2 16.9 1387 4.7director; head official 6.1 0.4 38.7 54.8 493 1.7n 892 184 7654 20989 29719 100.0col % 3.0 0.6 25.8 70.6 100.0

Employment Status - ESeC

<=

Page 29: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 29

ESeC class 2 => EGP: Movers and Stayers

I II IIIa IVab V IIIb VI VIIabMw S Mb Mb Mb Mb Mb Mb total

1 Armed forces 12 30 55 0 0 0 0 0 9712 Corporate managers 307 16 0 0 175 0 0 0 49813 Managers of small enterprises 68 52 15 0 0 5 0 0 14023 Teaching professionals 732 518 0 0 0 0 0 0 125024 Other professionals 309 643 0 0 0 0 0 0 95231 Physical & engineering science associate professionals290 399 0 39 578 0 4 4 131432 Life science & health associate professionals58 1040 5 0 0 0 0 0 110333 Teaching associate professionals 25 209 40 0 0 0 0 0 27434 Other associate professionals 12 541 931 22 0 46 0 0 155241 Office clerks 32 499 308 0 0 48 0 0 88742 Customer services clerks 0 17 55 0 0 30 0 0 10251 Personal and protective services workers7 94 70 0 0 6 130 0 307

total 1852 4058 1479 61 753 135 134 4 8476% 21.8 47.9 17.4 0.7 8.9 1.6 1.6 0.0 100.0

Mw: Movers within same contract typeS: Stayers (ESeC and EGP class corresponds)Mb: Movers between contract types

ISCO Submajor groups

Stayers and Movers by ISCO submajor groupsESeC class 2 => EGP_ISCO

<=

Page 30: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 30

Step 3: Employment relation indicators: ESeC – EGP

Fig. 5.1: Work Autonomy, Means of Factor Scores, (BIBB/IAB)

Fig. 5.2: Asset Specificity: adjusted highest degree: % requiring foreman education/college (BIBB/IAB)

EGP_KldB EGP_ISCO EseC_3d EGP_KldB EGP_ISCO EseC_3d R² 0.151 0.145 0.137 Pseudo R 0.294 0.288 0.232 Fig. 5.3: Asset Specificity: adjusted highest degree: % requiring any degree (BIBB/IAB)

Fig. 5.4: Asset Specificity: only introduction to the job not sufficient (BIBB/IAB)

EGP_KldB EGP_ISCO EseC_3d EGP_KldB EGP_ISCO EseC_3d Pseudo R 0.214 0.206 0.162 Pseudo R 0.143 0.139 0.094

=><=

Page 31: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 31

Employment relation indicators : ESeC – EGP compared

Fig. 5.5: Long-Term-Employment: Tenure Residuals, BIBB/IAB

Fig. 5.6: Career Prospects % having obtained further education (BIBB/IAB)

EGP_KldB EGP_ISCO EseC_3d EGP_KldB EGP_ISCO EseC_3d R² 0.050 0.047 0.049 Pseudo R 0.123 0.122 0.113

<=

Page 32: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 32

4 Class effects on risks of unemployment: Findings

Logistic regression of the odds of having been unemployed in the past for at least once BIBB/IAB (N= 33.433) ESeC EGP Prototype(V3) +

UK-modal value German Matrix +

German modal value

KldB

ISCO 4-digit 3-digit 2-digit 4-digit 3-digit 2-digit 3-digit 3-digit

(...) (...) (...) (...) (...) (...) (...) (...) (...) ESeC / EGP 1 / I 0.42*** 0.44*** 0.48*** 0.39*** 0.44*** 0.43*** 0.34*** 0.36*** 2 / II 0.49*** 0.48*** 0.51*** 0.45*** 0.46*** 0.49*** 0.42*** 0.44*** 3 / IIIa 0.57*** 0.59*** 0.61*** 0.56*** 0.61*** 0.61*** 0.48*** 0.51*** 4 / IVab 0.57*** 0.57*** 0.58*** 0.55*** 0.57*** 0.58*** 0.47*** 0.48*** 5 / IVc 0.51* 0.51* 0.51* 0.50* 0.51* 0.51* 0.23*** 0.27*** 6 / V 0.61*** 0.64*** 0.60*** 0.59*** 0.63*** 0.60*** 0.46*** 0.48*** 7 / IIIb 0.68*** 0.63*** 0.62*** 0.73*** 0.68*** 0.63*** 0.61*** 0.61*** 8 / VI 0.74*** 0.76*** 0.76*** 0.69*** 0.75*** 0.76*** 0.47*** 0.52*** 9 / VIIab (rc) 1 1 1 1 1 1 1 1 Pseudo R² 0.044 0.044 0.043 0.045 0.044 0.044 0.048 0.047

<= =>

Page 33: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 33

Odds of having been unemployed – class effects

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

1 / I 2 / II 3 / IIIa 4 / IVab 5 / IVc 6 / V 7 / IIIb 8 / VI 9 / VIIabclass effects - ESeC/EGP

Prototype_4d/ 0.044 German Matrix_4d/ 0.045

EGP_KldB/ 0.048 EGP_ISCO/ 0.047

<=

Page 34: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 34

Odds of having been unemployed – class effects

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

1 / I 2 / II 3 / IIIa 4 / IVab 5 / IVc 6 / V 7 / IIIb 8 / VI 9 / VIIabclass effects - ESeC/EGP

Prototype_4d/ 0.044 Prototype_3d/ 0.044 Prototype_2d/ 0.043

EGP_KldB/ 0.048 EGP_ISCO/ 0.047

Page 35: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 35

Odds of having been unemployed – class effects

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

1 / I 2 / II 3 / IIIa 4 / IVab 5 / IVc 6 / V 7 / IIIb 8 / VI 9 / VIIabclass effects - ESeC/EGP

German Matrix_4d/ 0.045 German Matrix_3d/ 0.044 German Matrix_2d/ 0.044

EGP_KldB/ 0.048 EGP_ISCO/ 0.047

Page 36: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 36

Odds of having been unemployed – class effects

0,20

0,30

0,40

0,50

0,60

0,70

0,80

0,90

1,00

1 / I 2 / II 3 / IIIa 4 / IVab 5 / IVc 6 / V 7 / IIIb 8 / VI 9 / VIIabclass effects - ESeC/EGP

Prototype_4d/ 0.044 Prototype_3d/ 0.044 Prototype_2d/ 0.043

German Matrix_4d/ 0.045 German Matrix_3d/ 0.044 German Matrix_2d/ 0.044

EGP_KldB/ 0.048 EGP_ISCO/ 0.047

Page 37: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 37

Appendix: Criterion Validation: Main Results BIBB/IAB Prototype

standard German

standard Prototype

proxy German

proxy Work Autonomy Factor 13.4 14.2 13.6 14.3 Asset Specificity Any degree 14.8 17.0 15.4 17.6 Master, College 22.1 25.2 23.2 26.7 Career Prospects 10.6 11.8 10.2 11.2 Long-term Employment. 4.6 5.4 4.4 5.3 GSOEP Prototype

proxy German

proxy Work Autonomy Factor 10.4 10.4 Delegated Authority 10.2 9.9 Asset Specificity Only introduction to job

24.2 27.0

College required 28.5 33.7 Career Prospects Factor 9.4 9.4 Long-term Employment 4.1 4.6

Page 38: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 38

Appendix: Aggregation of ESeC: 2-/3-digit Matrices I

• Starting base: revised German 4-digit matrix and German modal values

Table 2.1: Distribution of ESeC 4-, 3-, and 2-digit (BIBB/IAB, N=33.607) ESeC 4-digit 3-digit 2-digit 1. Higher Salariat.Occupations 12.4 10.5 11.9 2. Lower Salariat Occupations 21.9 25.2 21.6 3. Intermediate Occupations 15.9 16.2 16.6 4. Self employed (Non-Prof.) 7.1 7.3 7.3 5. Self employed (Agriculture) 0.2 0.2 0.2 6. Lower Superv./Technician Occ. 11.0 9.1 8.9 7. Lower Services/Sales/Clerical Occ. 7.3 7.6 9.9 8. Lower Technical Occ. 11.6 12.3 12.3 9. Routine Occupations 12.6 11.5 11.3 Total 100.0 100.0 100.0

Page 39: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 39

Appendix: Aggregation of ESeC: 2-/3-digit Matrices II

Table 2.2: Correspondence between the 4-digit and 3-digit matrices (BIBB/IAB, 1999).

3-digit ESeC class 1 2 3 4 5 6 7 8 9

1 99.5 6.9 0.3 2.5 2 0.5 85.2 0.9 2.5 0.5 3 0.7 93.1 7.0 0.9 4 0.5 95.0 5 100.0 6 6.6 99.1 2.2 7 5.7 83.0 0.1 8 0.5 91.3 3.1

4-di

git

9 10.0 6.6 95.9 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Page 40: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 40

Appendix: Aggregation of ESeC

Table 2.3: Flows from 4 -digit to 2 -digit (BIBB/IAB, 1999). 2-digit ESeC

class 1 2 3 4 5 6 7 8 9 1 85.2 9.5 0.3 2.5 2 14.8 79.5 12.5 2.8 8.0 3 81.9 23.4 4 0.8 94.7 5 100.0 6 10.2 92.0 4.5 7 2.8 69.0 0.1 8 88.9 5.8

4-di

git

9 2.6 7.6 6.7 94.1 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Page 41: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 41

Appendix: Operationalization en detail I

• Supervisor function:

– Standard version: “Do you have co-workers for whom you are their direct supervisor?”

– Proxy version: based on “Position with Employer” (14 categories): foreman or master, employee with high level managerial tasks, high-level and executive level civil servants

• Work Autonomy: Factor Scores, 3 items (almost always – hardly ever)

– Work tasks are prescribed in all details

– An identical work operation recurs in all details

– A precise number of product units, a minimum work performance or the time to carry out a specific work task is prescribed

Page 42: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 42

Appendix: Operationalization en detail II

• Asset Specificity:

information on the qualification job holders in reality have combined with their assessment whether the job could be done by a worker with lower qualifications. Starting from a distinction between the following three levels of the highest (vocational) qualification declared by the respondent

– no vocational qualification

– vocational school, vocational training etc.

– foreman degree (corresponding to post-secondary non-tertiary education or (tertiary) college degree

the score for the measure is reduced by one level whenever the respondent indicates that his job could be performed with a lower education than he himself has

Page 43: Validation of ESeC: The Effect of Coding Procedures and Occupational Aggregation Level Cornelia Hausen, Jean-Marie Jungblut, Walter Müller, Reinhard Pollak,

German Team: Validation of ESeC Lisbon, 19 – 20 January 2006 43

Appendix: Operationalization en detail III

• Long-Term Employment:

– Residuals of number of years with the current employers controlled for variation in length of labour force experience (number of years with the current employer) and gender

• Career Prospective:

– extent of further education in the last five years (with the present employer)