Is Europe pulling apart? Findings from the analysis of social distance

Paul Lambert, University of Stirling Dave Griffiths, University of Stirling Erik Bihagen, University of Stockholm Richard Zijdeman, University of Utrecht

Paper presented to the ISA RC28 Spring Meeting,

Budapest, 8-10 May 2014

1

Part of work on the ERSC Secondary Data Analysis Initiative Phase 1 project ‘Is Britain pulling apart? Analysis of generational change in social distances’ http://www.camsis.stir.ac.uk/pullingapart http://www.twitter.com/pullingapart http://pullingapartproject.wordpress.com/

Social structure & social distance

(b) Social structure as defined by social distance is revealing Interaction structure not identical to other structures and of theoretical interest (?the trace of social reproduction)

May be particular connections of interest (e.g. bridging ties)

2

0 1 2 3 4

Other

Plant/machine op.

Sales

Services

Craft

Clerical/sec.

Assoc. prof./tech.

Professional

Manag./Admin

Source: Analysis of married males in BHPS. Scores mean standardised plus 2.

SID score (spouses job)

Income score

(a) Study associations around a distribution of positions, for connected individuals Homphily/homogamy trends are

important, but are fairly stable, and contrast with popular perceptions of fast growth in separation (e.g. Murray 2012)

Some evidence on homophily/homogamy trends

UK, spouses (Brynin et al. 2008)

X-nat, spouses (e.g. Kalmijn 1998)

US, confidants (Smith et al. 2014)

Educ. + 0, + & - +

Occup. 0 0, + & - 0

Other + (attitudes) - (ethnicity) 0 (age, ethnicity) - (gender)

Finding new evidence on trends?

• Use/compare methods that both estimate and evaluate the social distance structure

– Based on prevelance of social interactions

– Previous work on occupations (e.g. www.camsis.stir.ac.uk) generalised to other forms of social distance

• Take advantage of rich new data resources

– Vast scale of husband-wife records from resources such as IPUMS-I

– Current and previous household sharers from longitudinal household survey data

– {Proxy questions on alters, e.g. on friends}

– {Administrative data (e.g. shared workplace)}

3

Methods used • Summary stats on ego-alter associations and their trends

a) Absolute association (e.g. Cramer’s V)

b) Dimensional orientation ({1st}dimensional structure & ego-alt cor.)

c) Net association (e.g. Smith et al. 2014’s case control approach – the

level of contact relative to chance)

d) Network summaries of ego-alter relations by cohort or time

4 Papers are in rank order of average CAMSIS score of readers

Ego-alter newpapers, BHPS 1991,2,6,7,2004 (5k non-diagonals, excluding 36k diag)

Daily Mail

Financial Times

Other

Daily Mirror/Record

Independent

Daily Telegraph

Sun

Guardian

Times

Daily Star

Daily Express

Regional

Dim

en

sio

n 2

(13

%)

Dimension 1 (14%)

y x

FT T 1 0

FT M 0 0

S S 1 1

S R 0 0

Husband’s paper

Wife’s paper

Random wife’s paper

…(1) trends in time don’t seem to be altered by different permutations in the number of categories measured, or the way that time is used…

http://www.camsis.stir.ac.uk/pullingapart 5

0.2

.4.6

0 1 2 3 4 5 6 7

Cases

Year (4-cat) Age (4) YoB (4) YoB, 40s (4)

Year (14-cat) Age (14) YoB (14) YoB, 40s (14)

Cramer's V

Source: Pooled GHS time series, 1974-2004. Horizontal axis refers to different time metrics by line. Metrics refer to: Years since 1970/5; age in decades-1; birth cohort (year of birth since 1900). Lines show statisics when education is coded into 4 or 14-category versions, and for different measures of time (year, age, year of birth, and year of birth for adults in their 40s).

Educational homogamy in the UK

0.5

0 2 4 6 8

Cramer's V

05

0 2 4 6 8

High-Low 'Social distance'(Gap in score in 1st dimension)

02

0 2 4 6 8

Over-representation ratio(high-low obs./exp.)

.4.5

.6

0 2 4 6 8

H-W correl. in 1st dim0

.2.4

0 2 4 6 8

Case-control model r2

Source: Pooled GHS time series, 1974-2004. Horizontal axis refers to different time metrics by line. Metrics refer to: Years since 1970/5; age in decades-1; birth cohort (year of birth since 1900). Lines show statisics when education is coded into 4 or 14-category versions, and for different measures of time (year, age, year of birth, and year of birth for adults in their 40s). Lines smoothed with local linear smoothing (lowess)

Educational homogamy in the UK

Year (4-cat) Age (4) YoB (4) YoB + 40s (4)

Year (14-cat) Age (14) YoB (14) YoB, 40s (14)

6

..(2) Different summary statistics can point in different directions

..(3) Social distance trends don’t particularly match up to stories of pronounced social change

• Have you noticed how there’s more and more talk about social change these days?

– Rise of precariousness and neo-liberalism (e.g. Ross 2009)

– Growing nepotism, reproduction, exclusionary individualism (e.g.

Dorling 2013, Murray 2012, Winlow & Hall 2013)

– Unerring drive towards modernisation (e.g. Marks 2014)

– Diffident stability (e.g. Erikson & Goldthorpe 2010)

Social distance trends ought to be an important marker of any of these changes

To what extent do social distance trends fit these trajectories?

7

http://www.camsis.stir.ac.uk/pullingapart 8

0.5

1995 2000 2005 2010 2015

Cramer's V

0.2

.4.6

1995 2000 2005 2010 2015

H-W correl. in strat. score 0

.2.4

.6

1995 2000 2005 2010 2015

H-W correl. in 1st dim

01

23

1995 2000 2005 2010 2015

High-Low distance

0.2

.4.6

1995 2000 2005 2010 2015

Over-representation(high-low obs./exp.) 0

.5

1995 2000 2005 2010 2015

Case-control r2

Source: Pooled LFS, 1997-2013, cohabiting couples. Horizontal axis refers to time point of observation. 'Lowess' lines plotted (local linear smooth) Colours indicate age cohort within time period (age of husband). N ~= 5k couples per time period.

Homogamy in the UK

Education Occupation Age 25-35 Age 50-60

02

04

06

08

0

Lin

ea

r pre

dic

tion

20 40 60 80 100

USA, 1960-2010

02

04

06

08

01

00

Lin

ea

r pre

dic

tion

20 40 60 80 100

Mexico, 1970-2010

02

04

06

08

01

00

Lin

ea

r pre

dic

tion

20 40 60 80 100

France, 1962-2006

02

04

06

08

0

Lin

ea

r pre

dic

tion

20 30 40 50 60 70

Greece, 1971-20010

20

40

60

80

Lin

ea

r pre

dic

tion

20 40 60 80 100

Hungary, 1970-2001

20

40

60

80

100

Lin

ea

r pre

dic

tion

20 40 60 80 100

Spain, 1991-2001

05

01

00

150

200

Lin

ea

r pre

dic

tion

0 50 100 150 200

Switzerland, 1970-2000

02

04

06

08

0

Lin

ea

r pre

dic

tion

20 40 60 80

UK, 1997-2013 (occ) and 1972-2013 (educ)

20

40

60

80

100

Lin

ea

r pre

dic

tion

20 40 60 80 100

Sweden, 1970-2007

20

30

40

50

60

70

Lin

ea

r pre

dic

tion

20 30 40 50 60 70occisco score(1) in symmetric norm.

Netherlands, 2000-2011

Data for cohabiting m-f couples from IPUMS-I, or from national Labour Force Surveys (UK, NL) or census (SE)

Occupation Education Year specific lines

…the underlying orders change slightly…

(dimension scores for the same units over time)

9

USA1960-2010

Mexico1970-2010

France1962-2006

Greece1971-2001

Hungary1970-2001

0.2

.4.6

.8

Spain1991-2001

Switz.1970-2000

UK1975-2013

Sweden1970-2007

Netherlands2000-2011

0.2

.4.6

.8

Analysis based on husband-wife associations from IPUMS-I or LFS data. Statistics are ego-alter Cramer's V, Ego-Alt dim1 association, or case-control R.

International trends in social distance

Occupation Education

Cramer's V H-W dim1 assoc. Case-control R

10

• It might be more consistent to compare patterns against an anticipated (a priori) trend line?

Either flatline, or linear change by 1 sd each decade, or quadratic by (sd/dec^2)…

http://www.camsis.stir.ac.uk/pullingapart 11

Cramer’s V trend with time for education, GHS. The observed patterns fit somewhat with ‘linear increase’ but of the options, ‘no change’ is best Unconstrained, a more moderate linear increase fits best

-2-1

01

2

0 2 4 6 8timeunit

Data points No change (.875)

Linear increase (1.849) Linear decrease (10.4)

Quadratic increase (276.826) Quadratic decrease (336.548)

Statistics are a mean value for the squared error expressed as a proportion of the variance

Pulling apart!

Tearing apart!

No change

Pushing together!

Crashing together!

Social distance trends in Britain

GHS data, 72-04 Type of Stat. Best trend line LFS data, 1997-2013

Best trend line (age 50-60)

Best trend line (age 25-35)

Educ (4) by yob Cramer’s V No change (+) Educ. Pulling apart (+) No change (+)

`` HW Dim 1 cor. No change (+) `` Pulling apart (+) No change (+)

`` High-Low dist. No change (--) `` Pulling together (-) Pulling together (-)

`` H-L occurrence No change (-) `` Pulling apart (+) Pulling apart (+)

H-W strat cor. `` Pulling apart (+) No change (+)

Educ (4) by yob Cramer’s V Pulling apart (+) Occ (9) Pulling together (-) Pulling apart (+)

for age 40-50 HW Dim 1 cor. Pulling apart (+) `` Pulling together (-) Pulling apart (+)

`` High-Low dist. Pulling together (-) `` Pulling apart (+) Pulling apart (+)

`` H-L occurrence No change `` No change Pulling apart (+)

H-W strat cor. `` Pulling together (-) Pulling apart (++)

Educ(14) by yob Cramer’s V No change (++) Ethnic (11) No change (+) No change (++)

`` HW Dim 1 cor. No change (++) `` Pulling together (-) Pulling together (-)

`` High-Low dist. No change `` Pulling apart (+) Pulling apart (++)

`` H-L occurrence No change (-) `` Pulling apart (+) Pulling apart (+)

H-W strat cor. `` Pulling apart (+) No change http://www.camsis.stir.ac.uk/pullingapart 12

UKHLS 2011 - Partners reading diff. papers Ignores couples which read same paper (39% of younger couples and 73% of older couples).

Both < 50 yrs (Scotland)

..(4) Further insight from exploring, or controlling for, different elements of social distance

Males > 50 yrs, females > 45 (Scotland)

Husbands born since 1956 (1,035 cases)

Husbands born pre 1956 (754cases)

Partners’ views on family and state in Sweden (LNU 2000)

Deleting diagonals (33% of younger couples, 32% of older)

Conclusions

• Methodological results – Number of categories and time comparisons don’t change

trends; but different summary stats may do so, and the better summary stats are probably those that try to control for other changes in social structure

– Similar (absence of dramatic) trends across domains (socio-economic, demographic, cultural)

• A diffident peroration – Both socio-economic and socio-cultural trends in social

distance reject a characterisation of dramatic change and might, if anything, suggest a ‘Europe pulling together’

– Nothing very exciting happening, to be honest

15

References cited

• Chan, T. W. (Ed.). (2010). Social Status and Cultural Consumption. Cambridge: Cambridge University Press.

• Brynin, M., Longhi, S., & Martinez Perez, A. (2008). The Social Significance of Homogamy. In M. Brynin & J. Ermisch (Eds.), Changing Relationships. London: Routledge.

• Dorling, D. (2013). How social mobility got stuck. New Statesman, 16 May 2013.

• Erikson, R., & Goldthorpe, J. H. (2010). Has social mobility in Britain decreased? Reconciling divergent findings on income and class mobility. British Journal of Sociology, 61(2), 211-230.

• Kalmijn, M. (1998). Intermarriage and homogamy: Causes, patterns, trends. Annual Review of Sociology, 24, 395-421.

• Laumann, E. O. (1973). Bonds of Pluralism: The form and substance of urban social networks. New York: Wiley.

• Laumann, E. O., & Guttman, L. (1966). The relative associational contiguity of occupations in an urban setting. American Sociological Review, 31, 169-178.

• Marks, G. N. (2014). Education, Social Background and Cognitive Ability. London: Routledge.

• Murray, C. (2012). Coming Apart. New York: Crown Forum.

• Ross, A. (2009). Nice Work If You Can Get It: Life and Labour in Precarious Times. New York: New York University Press.

• Savage, M., Devine, F., Cunningham, N., Taylor, M., Li, Y., Hjellbrekke, J., Le Roux, B., Friedman, S., & Miles, A. (2013). A new model of social class: Findings from the BBC's Great British Class Survey Experiment. Sociology, 47(2), 219-250.

• Smith, J. A., McPherson, M., & Smith-Lovin, L. (2014). Social distance in the United States: Sex, Race, Religion, Age and Education homophily among confidants, 1985 to 2004. American Sociological Reivew, DOI:10.1177/0003122414531776.

• Stewart, A., Prandy, K., & Blackburn, R. M. (1980). Social Stratification and Occupations. London: MacMillan.

• Winlow, S., & Hall, S. (2013). Rethinking Social Exclusion: The End of the Social? London: Sage.

16

Older couples links mostly involve Daily Record. Younger couples show many more links in the papers they read, and more often read different papers. Evidence that younger couples are more cosmopolitan / Britain isn’t pulling apart?

UK-HLS 2011 - Partners reading diff. papers

Ignores couples which read same paper (39% of younger couples and 73% of older couples). UKHLS, 2011 (min. 2 ties).

Both < 50 yrs (Scotland) (99 couples)

Males > 50 yrs, females > 45 (Scotland) (98 couples)

Husbands born since 1956 (1,035 cases)

Husbands born pre 1956 (754cases)

Partners’ views on family and state in Sweden (LNU 2000)

Couples after deleting those cases where both partners held same left/right & traditional/liberal values (33% of younger couples, 32% of older couples)

• Left/right division stable

• Younger cohort: more ties; older cohort: more homophily

• No evidence of ‘pulling apart’. Similar % homophilous on both factors (around 1/3rd). If anything, evidence of becoming more diverse.