Concepts and Measures: Empirical Evidence on the

Download Report

Transcript Concepts and Measures: Empirical Evidence on the

Concepts and Measures in
occupation-based social
classifications
Presentation to: ‘Interpreting results from statistical modelling –
a seminar for social scientists’ , Imperial College, 29th April 2008
Dr Paul Lambert and Dr Vernon Gayle
University of Stirling
A seminar for the ESRC National Centre for Research Methods, Lancaster-Warwick
Node on ‘Developing Statistical Modelling in the Social Sciences’
ESRC - NCRM - Apr 2008
1
Part 1: Data on occupations
• In the social sciences, occupation is seen as one
of the most important things to know about a
person
 Direct indicator of economic circumstances
 Proxy Indicator of ‘social class’ or ‘stratification’
• GEODE and DAMES
– how social scientists use data on occupations
– www.geode.stir.ac.uk / www.dames.org.uk
ESRC - NCRM - Apr 2008
2
Handling occupational data
[e.g. Lambert et al 2007, International Journal of Digital Curation]
Model is:
1) Record and preserve ‘source’ occupational data (i.e OUG)
2) Use a transparent translation code to derive occupation-based
social classifications
..Many people recommend this [cf. Bechhofer 1969; Rose and
Pevalin 2003] but not all applications do this..
Challenges include:
– Locating occupational information resources
http://home.fsw.vu.nl/~ganzeboom/pisa/
http://www.iser.essex.ac.uk/esec/consort/matrices/
– Large volumes of data (country; time; updates)
– Detail on occupational index units (OUGs)
– Gaps in working practices (software; NSI’s v’s academics)
ESRC - NCRM - Apr 2008
3
Stage 1 - Collecting Occupational Data
Example 1: BHPS
Occ description
Employment status
SOC-2000
EMPST
Miner (coal)
Employee
8122
7
Police officer (Serg.)
Supervisor
3312
6
Electrical engineer
Employee
2123
7
Retail dealer (cars)
Self-employed w/e
1234
2
Example 2: European Social Survey, parent’s data
Occ description
SOC-2000
EMPST
Miner
?8122
?6/7
Police officer
?3312
?6/7
Engineer
??
??
Self employed businessman
??
?1/2
www.geode.stir.ac.uk/ougs.html
ESRC - NCRM - Apr 2008
5
GEODE provides services to help social scientists
1) Disseminate, and access other, Occupational
Information Resources
2) Link together their (secure) micro-data with OIR’s
External user
(micro-social data)
Occ info (index file)
(aggregate)
id
oug
sex
.
oug
CS-M CS-F EGP
1
110
1
.
110
60
58
2
320
1
.
320
69
3
320
2
.
874
39
4
874
1
5
874
2
User’s output
(micro-social data)
id
oug
CS
I
1
110
60
.
71
II
2
320
69
.
51
VIIa
3
320
71
.
.
4
874
39
.
.
5
874
51
.
ESRC - NCRM - Apr 2008
6
Occupational information resources: small
electronic files about OUGs…
Index units # distinct files
Updates?
(average size kb)
200 (100)
y
www.camsis.stir.ac.uk
Local
OUG*(e.s.)
CAMSIS value labels
Local OUG
50 (50)
n
Int. OUG
20 (50)
y
20 (200)
n
www.iser.essex.ac.uk/esec
Int.
OUG*(e.s.)
Hakim gender seg codes
Local OUG
2 (paper)
n
CAMSIS,
www.camsis.stir.ac.uk
ISEI tools,
home.fsw.vu.nl/~ganzeboom
E-Sec matrices
(Hakim 1998)
ESRC - NCRM - Apr 2008
7
For example: ISCO-88 Skill levels classification
ESRC - NCRM - Apr 2008
8
and: UK 1980 CAMSIS scales and
CAMCOM classes
ESRC - NCRM - Apr 2008
9
GEODE Occupational Information Depository
• Collects large volumes of OIRs across countries, time
periods
• Facilitates communication between producers of
occupational information resources
 Universality
 Hitherto the dominant approach
 same occupation-based measures valid across all
countries/time periods
 Specificity
 different occupation-based measures should be used
specific to different countries / time periods
 See http://www.geode.stir.ac.uk/publications.html
ESRC - NCRM - Apr 2008
10
Part 2) Concepts and measures
[Lambert and Bihagen 2007]
Sensible taxonomies can rarely be judged true or false, only
more or less useful for a given purpose [Mills & Evans, 2003:80]
 Relevance of reviewing lots of schemes
 (1) Broad concordance of most measures
 (2) Optimum measures are ambiguous
[EGP]...has a clear theoretical basis, therefore differences between
groups in health outcomes can be attributed to the specific
employment relations that characterise each group [Shaw et al., 2007:78]
 (1) Lots of overlap in conceptual correlates
 (3) A small residual difference does reflect concepts
ESRC - NCRM - Apr 2008
11
How to interpret β’s from occupationbased social classifications…
• What the measures measure
– Criterion and construct validity
• What measures measure in multivariate
context
– Approaches to complex analysis
ESRC - NCRM - Apr 2008
12
Micro-data
• Britain 1991-2002
• Sweden 1991-2002
 BHPS 1991, 4537 adults 23- • LNU 1991, 2538 adults 2355yrs in work
55yrs in work
 2710 adults observed every • Linked to PRESO
year till 2002
administrative data until
2002 [Tomas Korpi]
Unemployment 1991-2002 (m/f; employees)
Br
Sw
28% / 23%
36% / 39%
Unemployed for >1 year 1991-2002
9% / 6%
26% / 29%
‘Incidence rate’ (time Un. / active time)
3.4 / 2.3
Cumulative rate (log of total time Un.)
1.5 / 1.2
Ever Unemployed 1991-2002
ESRC - NCRM - Apr 2008
2.3 / 2.3
13
=> 31 Occupation-based social classifications
ES5 Employment Status (5)
WR Wright (12 categories)
ES2 Employment Status (2)
WR9 Wright (9)
CM CAMSIS (male scale)
E9 ESeC (9 categories)
G11 EGP (11 categories)
CF CAMSIS (female scale)
E6 ESeC (6 categories)
G7 EGP (7 categories)
CM2 CAMSIS (male scale, S)
E5 ESeC (5 categories)
G5 EGP (5 categories)
CF2 CAMSIS (female, S)
E3 ESeC (3 categories)
G3 EGP (3 categories)
CG Chan-Goldthorpe status
E2 ESeC (2 categories)
G2 EGP (2 categories)
AWM Wage mobility score
K4 Skill (4 ISCO categories)
MN Manual / Non-M (2)
WG1 Wage score (S)
O17 Oesch work logic (17)
WG2 Wage score (S)
O8 Oesch work logic (8)
ISEI (via ISCO88)
WG3 Wage score (B)
O4 Oesch work logic (4)
SIOPS (via ISCO88)
GN Gender segregation index
ESRC - NCRM - Apr 2008
14
(2.1) Categorical - Categorical relations, Cramer's V
0
.1 .2 .3 .4 .5 .6 .7 .8 .9
1
Britain
ES2
ES5
E6
E9
E3
E5
G11
E2
G5
G2
G7
G3
WR
K4
Employemt Status
ESeC schemes
EGP schemes
Wright schemes
Oesch schemes
Manual / Non-manual
O17
WR9
o4
O8
MN
Skill classification
0
.1 .2 .3 .4 .5 .6 .7 .8 .9
1
Sweden
ES2
ES5
E6
E9
E3
E5
G11
E2
G5
G7
G2
G3
ESRC - NCRM - Apr 2008
Men
Women
WR
K4
O17
WR9
o4
O8
MN
15
(2.3) Categorical-Metric relations, Anova R
0
.1 .2 .3 .4 .5 .6 .7 .8 .9
1
Britain
ES2
ES5
E6
E9
E3
E5
G11
E2
G5
G2
G7
G3
WR
K4
O17
WR9
CAMSIS / CG Scale
ISEI
SIOPS
AWM
Income averages
Gender segregation
o4
O8
MN
0
.1 .2 .3 .4 .5 .6 .7 .8 .9
1
Sweden
ES2
ES5
E6
E9
E3
E5
G11
E2
G5
G7
G2
G3
ESRC - NCRM - Apr 2008
Men
Women
WR
K4
O17
WR9
o4
O8
16MN
1
(2.6) Associations - Employment Relations and Conditions
0
.1 .2 .3 .4 .5 .6 .7 .8 .9
Men and Women (categorical social classifications)
E6
E9
E3
E5
G11
E2
G5
G7
G2
G3
Promotion / retention
Pay - bonus / increments
Labour contract type
Subjective skill requirements
WR
K4
O17
WR9
o4
O8
MN
Hours and level of monitoring
.1 .2 .3 .4 .5 .6 .7 .8 .9
1
ES5
0
Men and Women (metric social classifications)
CF
CM
CF2
CM2
ISEI
CG
AWM
SIOP
ESRC - NCRM - Apr 2008
Britain
Sweden
WG2
WG1
GN
WG3
17
What measures measure
1) Broad concordance of schemes
•
Measures mostly measure the same thing
 Generalised concepts are better
 Occupation-based measures don’t uniquely measure
the concepts on which they are based (doh!)
•
Criterion validity is asymmetric
•
cf. Tahlin 2007: Skill or employment relations for EGP
ESRC - NCRM - Apr 2008
18
(3.4a) R-2 and BIC for predicted unemployment risk
Britain, Males
Null
ES2
E6
E3
G11
G5
G2
WR
O17
O4
CM
ES5
E9
E5
E2
G7
G3
K4
WR9
O8
MN
CF
Pseudo R-squared
CG
SIOP
ISEI AWM
WG3
GN
Increase in BIC
0
Sweden, Males
Null
ESRC - NCRM - Apr 2008
ES2
E6
E3
G11
G5
G2
WR
O17
O4
CM
CM2
ES5
E9
E5
E2
G7
G3
K4
WR9
O8
MN
CF
CF2
19
SIOP WG1
ISEI AWM WG2
GN
What measures measure
2) Construct validity is..
 also asymmetric
 conflated by level of occupational detail
3) Ambiguity of optimal schemes
 Balancing explanatory power and parsimony
 No schemes stand out as substantially stronger
 Highly collapsed versions are limited
•
(e.g. ESeC & EGP 3- and 2-class versions)
 Metrics are generally fine
ESRC - NCRM - Apr 2008
20
(4.1): Unemployment risks (British men)
(1): with additional explanatory variables
0
.025
.05
.075
(2): (1) plus industry indicator variables
E3
G7
G11
CM
K4
AWM
ISEI
0
E9
Decrease in log-like
E3
Increase in BIC
E9
CM
K4
AWM
ISEI
(4): Heckman selection, Industry = private manufacturing
0
0
.025
(3): Heckman selection, Industry = public sector services
G7
G11
E3
E9
G7
G11
ESRC
AWM - NCRM - Apr 2008
CM
K4
ISEI
E9
E3
G7
G11
21
AWM
CM
K4
ISEI
EGP cf. CAMSIS – critical individuals
Britain (males)
Better EGP predicted risk of Un. (H – rightly higher; L – rightly lower)
7121 (L) Builders (traditional)
8322 (L) Car / taxi drivers
1314 (L) Wholesale / retail managers
7141 (L) Painters
7231 (H) Motor mechanics
2411 (H) Accountants
4131 (H) Stock clerks
7124 (H) Carpenters / joiners
8324 (H) Truck / Lorry drivers
Better CAMSIS predicted risk of Un. (H – rightly higher; L – rightly lower)
5169 (L) Protective service workers
4212 (L) Tellers / counter clerks
4190 (L) Office clerks
7230 (L) Machinery mechanics/fitters
1314 (H) Wholesale / retail managers
ESRC - NCRM - Apr 2008
22
Measures in multivariate context
4) Multivariate contexts of coefficient effects
in occupations…
• ..are generally problematic – ‘everything
depends on occupations’
• Endogeneity of employment itself
• Household / career context of occupations
• Some residual differences do seem to
reflect conceptual origins [cf. Chan & Goldthorpe 2007]
ESRC - NCRM - Apr 2008
23
Conclusions
• Do measures measure concepts?
– Yes (sometimes) – criterion validity
– No (not uniquely)
• How should we choose between measures?
– Practical issues: favour widely used schemes and metrics
– Conceptual assumptions: favour generalised schemes
• What about standardisation (e.g. ESeC)?
– Few clear strengths in empirical properties
– Practical advantages if widely used
ESRC - NCRM - Apr 2008
24
References
•
Bechhofer, F. (1969). Occupations. In M. Stacey (Ed.), Comparability in Social Research (pp. 94-122). London:
Heinemann (in association with British Sociological Association / Social Science Research Council).
•
Chan, T. W., & Goldthorpe, J. H. (2007). Class and Status: The Conceptual Distinction and its Empirical
Relevance. American Sociological Review, 72, 512-532.
•
Elias, P., & McKnight, A. (2003). Earnings, Unemployment and the NS-SEC. In D. Rose & D. J. Pevalin (Eds.), A
Researcher's Guide to the National Statistics Socio-Economic Classification. London: Sage.
•
Goldthorpe, J. H., & McKnight, A. (2006). The Economic Basis of Social Class. In S. L. Morgan, D. B. Grusky & G.
S. Fields (Eds.), Mobility and Inequality. Stanford: Stanford University Press.
•
Hakim, C. (1998). Social Change and Innovation in the Labour Market : Evidence from the Census SARs on
Occupational Segregation and Labour Mobility, Part-Time work and Student Jobs, Homework and SelfEmployment. Oxford: Oxford University Press.
•
Lambert, P. S., & Bihagen, E. (2007). Concepts and Measures: Empirical evidence on the interpretation of ESeC
and other occupation-based social classifications. Paper presented at the International Sociological Association,
Research Committee 28 on Social Stratification and Mobility, Montreal (14-17 August).
•
Lambert, P. S., Tan, K. L. L., Turner, K. J., Gayle, V., Prandy, K., & Sinnott, R. O. (2007). Data Curation Standards
and Social Science Occupational Information Resources. International Journal of Digital Curation, 2(1), 73-91.
•
Mills, C., & Evans, G. (2003). Employment Relations, Employment Conditions and the NS-SEC. In D. Rose & D. J.
Pevalin (Eds.), A Researchers Guide to the National Statistics Socio-economic Classification (pp. 77-106).
London: Sage.
•
Rose, D., & Harrison, E. (2007). The European Socio-economic Classification: A New Social Class Scheme for
Comparative European Research. European Societies, 9(3), 459-490.
•
Rose, D., & Pevalin, D. J. (Eds.). (2003). A Researcher's Guide to the National Statistics Socio-economic
Classification. London: Sage.
•
Schizzerotto, A., Barone, R., & Arosio, L. (2006). Unemployment risks in four European countries: an attempt of
testing the construct validity of the ESeC scheme. Bled, Slovenia, and http://www.iser.essex.ac.uk/esec/: Paper
presented to the Workshop on the Application of ESeC within the European Union and Candidate Countries, 29-30
June 2006.
•
Shaw, M., Galobardes, B., Lawlor, D. A., Lynch, J., Wheeler, B., & Davey Smith, G. (2007). The Handbook of
Inequality and Socioeconomic Position: Concepts and Measures. Bristol: Policy Press.
•
Tahlin, M. (2007). Class Clues. European Sociological Review, 23(5)557-572.
Appendices
ESRC - NCRM - Apr 2008
26
• Picture – uploading data file
ESRC - NCRM - Apr 2008
27
ESRC - NCRM - Apr 2008
28
ESRC - NCRM - Apr 2008
29
Searching – uncurated resources
ESRC - NCRM - Apr 2008
30
Searching – curated resources
ESRC - NCRM - Apr 2008
31
Java portal
• picture
ESRC - NCRM - Apr 2008
32