Stability, trended change or data
artefacts: pooling forty years of social
mobility enquiries in the UK
Paper prepared for the ISA RC-28 Social Stratification
and Mobility conference, New York 22-24th August
2003-08-20
Ken Prandy
Cardiff University, UK
Paul S. Lambert Stirling University, UK
Marge Unt
Tallin P University, Estonia
Paper download:
www.staff.stir.ac.uk/paul.lambert/downloads.html
Associated project:
www.cf.ac.uk/CAMSIS/socmob
1
Social mobility in the UK:
 “Stability of relative inequality” paradigm
 Popularly adopted and maintained
 Based primarily on cohort comparisons
 Methods likely to reject small changes
Recent challenges
 Longer term data can detect change
 Payne and Roberts: BES studies  changes
 Family history project: trend of decline throughout
19th Century
 Prandy et al 2002: decline in 20th Century data
Current project: ‘long run changes in significance
of social stratification’, collated X-section data
with occupations, background, voting, lifestyle, …
Here attempt:
 Further evaluation of mobility trends
 Role of education
 Robustness of trends to alternative modelling
2
‘Long Run’ dataset:




Harmonises ~20 cross-sectional surveys
Different sampling, coding frames, contexts
Possible use of panel data recontacts
Possible work life history records
Key variables:
1. Respondents Own Occupation
2. Father’s occupation (at age 14)
3. Education (3, 4 category summary)
4. Time of record or Year of birth
5. Age of respondent
3
Occupational mobility : controversies over
occupational indicators
1) Continuous measures
 Reference to position in a structure of differences
 Concentrating on one dimension of hierarchy
 Suited to simpler modelling techniques and
interpretations
CAMSIS measures: typical occupational
stratification location as estimated through social
interaction patterns exhibited by occupational unit
holders
 Relative position within time period and gender
 2 UK versions used, 1971 and 1991, by gender
2) Categorical schemes





More prominent in mobility research
Emphasise bounded structures
Fixed in time and by gender
Encourage qualitative assessments of positions
Less suited to simpler summarising models
4
Father-child CAMSIS correlations:
high to low, by birth cohort, study, and study period:
Men
1890
1910
1920
1930
1940
1900
1950
1960
1970
1980
74 CAM
73 Irel
92 BES
79 BES
72 Nuf
74 Scot
63 BES
74a BES
74b BES
97 BES
83 BES
96 Nirl
87 BES
91 BHPS
84 Class
86 SCELI
Women
1910
1920
1900
1930
1950
1940
1960
1890
1970
1980
1976
1971
1965
1981
1991
1996
2000
1986
97 BES
74a BES
74b BES
92 BES
84 Class
87 BES
83 BES
96 Nirl
63 BES
79 BES
86 SCELI
91 BHPS
 Some suggestions of trends, esp for men
 But much noise also
5
1976
1981
1986
1996
1991
1965
2000
Regression models to assess multiple influences
predicting occupational outcomes
Interactions between year / period and fathers
occupations of most interest
Men
Women
4a.2
4a.4
4a.6
4a.8
FCS
AGE
AGE2
YEAR
YEAR2
+++
+++
--+++
---
++
++
-+++
---
+++
+++
-----
+
+
---
ED1
ED3
EDV
FCSAGE
FCSAGE2
FCSYR
AGEYR
FCSED1
FCSED3
AGEED1
AGEED3
YRED1
YRED3
--+++
--+++
--+++
+++
++
--+++
+++
----+++
---
--+++
+++
++
R2*100
--+++
35
35
6
-+++
32
+
+
+++
+++
+
--+++
+
31
Determinants of education:
Men
FCS
AGE
YEAR
FCSAGE
FCSYEAR
AGEYEAR
5a.2
+++
--+++
+++
+++
+++
5a.3
+++
--+
+++
++
C/S R2*100
19
11
Women
5a.6
5a.7
+++
+++
----+++
+++
+++
+++
------26
22
Suggesting net patterns of:
Men:
 Direct occupational immobility 
 But educational immobility 
 Albeit with declining impact on occ outcomes
Women:
 Direct occupational immobility [unclear]
 But educational immobility 
7
Potential errors (1): appropriate occupational
measure?
Predictions of CAMSIS with father’s job
categorised variously:
 Artificial CAMSIS break (1980) has some
significance
 Most category estimates seem aligned with a
hierarchy anyway
 Some significance to CASMIN agricultural
categories
Predictions of current job in EGP or RGSC
categories:
 Broadly similar structures to dummy category
main and interaction effects, though much more
detail to the interpretation of each coefficient
 Interactions with time much more complex: don’t
get the simple conclusion of increased or stable
mobility
eg male RGSC ordered :
+ve (adv): ED3, FRG2ED1, FRG4ED1, FRG3ED1,
AGE, FRG3YR, AGEYR, AGEED3
-ve (disadv): FRG5, FRG4, ED1, FRG3, FRG2,
YEAR, YRED3, FRG5AGE, FRG3AGE, FRG4AGE,
AGEED1
8
Potential errors (2): structural influences on
mobility propensities?
 For CAMSIS outcomes, time period effect
theorised as sufficient control for structural
change, given relative average locations
 Tests of influence of ‘sector’ dummies and
‘inheritance’ effects show no significant impact
upon other variable parameter estimates
Potential errors (3): survey sampling effects?
 Large main effects – surveys influence outcome
distributions
 But negligible influence on other variables’
prediction coefficients
 Stable patterns within restricted age groups : not
just an age-period conflation
 Between country differences: NI and Scot have
larger main effect immobility but greater rate of
decline over time
9
Conclusions
 Estimates of father’s-occupation influence on
educational attainment and occupational
outcomes seem robust to methods variants,
covering:
 Survey effects
 Occupational measurement type
 Structural mobility accounts
 Interaction terms show gender differentiation
whereby:
 Men have net declining direct occupational
relations, but increasing father’s influence on
educational levels
 Women have no discernible net occupational
trend, but declining influence of fathers on
educational levels
 Use of path model formulations, SEMs, may
account for magnitudes involved better
10