Social Stratification of Cultural Participation:
Theatre and Cinema, the Visual Arts and
Reading∗
Tak Wing Chan
Department of Sociology
University of Oxford
John H. Goldthorpe
Nuffield College
University of Oxford
May 2, 2005
Abstract
We analyse survey data on cultural consumption in the fields of
(1) theatre and cinema, (2) the visual arts and (3) reading. Using latent class models and regression models, we show that the omnivore–
univore argument applies in these fields as in music. However, distinct subtypes of cultural omnivores in reading and the visual arts are
also identified. We report cross-domain similarities and differences
in the association between cultural consumption and various sociodemographic and stratificaton variables. We further show that, consistent with the Weberian perspective of social stratification, it is status,
rather than class, which predicts pattern of cultural consumption.
1
Introduction
In the current sociological literature that treats the relationship between
social stratification and cultural taste and consumption, it is possible to
Paper prepared for the Oslo meeting of the ISA Research Committee 28 on Social
Stratification and Mobility in May 2005. Please do not cite or quote without permission.
We are grateful to Arts Council England, especially Adrienne Skelton and Ann Bridgwood, for access to the detailed occupational codes of the Arts Council data set. The
views expressed in this paper are entirely our own, and not necessarily those of the Arts
Council. Our research is supported by a ESRC/AHRC research grant under their Cultures
of Consumption Research Programme Phase II, award number: RES–154–25–0006.
∗
1
identify three main—and rival—lines of argument. In their essentials, these
arguments can be stated as follows, although each has variant forms.
(1) The homology argument: This claims that social stratification and
cultural stratification map onto each other very closely. Individuals in higher
social strata are those who prefer and predominantly consume ‘high’ or ‘elite’
culture, and individuals in lower social strata are those who prefer and predominantly consume ‘popular’ or ‘mass’ culture—with, usually, various intermediate situations being also recognised. In some versions of the argument
(see e.g. Bourdieu, 1984) distinction in cultural taste is actively used by members of the ‘dominant’ social class as a means of symbolically demonstrating
and confirming their superiority.
(2) The individualisation argument. This seeks in effect to relegate the
homology argument to the past in claiming that in modern, relatively affluent and highly commercialised societies, differences in cultural taste and
consumption are losing all grounding in social stratification. Age, gender,
ethnicity or sexuality may be seen as alternative social bases of cultural
differentiation. But, at least in more radical forms of the argument (e.g.
Featherstone, 1987; Bauman, 1988), the emphasis is on the growing ability of individuals to free themselves from all social influences and to choose
and form their own distinctive identities and lifestyles—patterns of cultural
consumption included.
(3) The omnivore–univore argument. This in effect challenges both the
homology and individualisation arguments (see esp. Peterson and Simkus,
1992; Peterson and Kern, 1996). As against the latter, it sees cultural differentiation as still mapping closely onto social stratification; but, as against the
former, it does not see this mapping as being on ‘elite-to-mass’ lines. Rather,
it claims that the cultural consumption of individuals in higher social strata
differs from that of individuals in lower social strata in that it is greater and
much wider in its range. It comprises not only more ‘high-brow’ culture but
more ‘middle-brow’ and ‘low-brow’ culture as well, while the consumption
of individuals in lower social strata tends to be largely restricted to more
popular cultural forms. Thus, the crucial distinction is not between elite and
mass but rather between cultural omnivores and cultural univores.
In a previous paper on the relationship between musical consumption and
social stratification (Chan and Goldthorpe, 2005), we report results which
support the omnivore–univore thesis rather than either the homology or the
individualisation thesis. But some qualification of the omnivore–univore thesis is required. In particular, two kinds of omnivore emerge: true omnivores
(O) and omnivore-listeners (OL). Os differ from OLs in being more likely to
be women, older and having higher educational attainment and status.
The question we address in the present paper is as follows. Does a similar
2
or a different pattern of consumption in relation to social stratification occur in other cultural domains? Research in cultural stratification has so far
focussed on just one cultural domain, namely music (see e.g. Peterson and
Simkus, 1992; Peterson and Kern, 1996; Bryson, 1996; Bryson, 1997; van
Eijck, 2001; Coulangeon, 2003). While music might be a good test case for
cultural stratification,1 its pattern might turn out to be rather specific. As
the wider applicability of the omnivore–univore argument in other cultural
domains remains largely untested (for a rare exception, see van Rees, Vermunt and Verboord, 1999), there is a need to move beyond music and examine
a broad range of cultural practices. In this paper, we shall consider cultural
stratification in three domains, namely theatre and cinema, the visual arts
and reading.
2
Data and Analytical Strategy
We use the same data set as our previous paper on musical consumption,
which come from an inquiry carried out in England in 2001 by the UK Office for National Statistics on behalf of Arts Council England. In the Arts
in England Survey face-to-face interviews were carried out with a stratified
probability sample of individuals aged 16 or above and living in private households. Interviews were completed with 6,042 respondents, giving a response
rate of 64 per cent (for details, see Skelton et al. 2002). In line with our previous papers, we restrict our analysis in this paper to respondents aged 20 to
64 (N = 4, 249). After deleting cases with missing values on the covariates,
N drops moderately to 3,819.
The Arts in England survey contains the following measures of cultural
participation. Respondents were asked whether in the last 12 months they
have seen (1) a film at a cinema or other venue, (2) a play/drama, (3) a
pantomime, (4) a musical, (5) a ballet, and (6) other dance events, including
African People’s dance, contemporary dance, and South Asian dance. We
take the binary response to these six items as indicators of cultural participation in the domain of theatre and cinema.2
As for the visual arts, respondents were asked whether in the past 12
months, they have visited (1) a museum/art gallery, (2) an exhibition or
collection of art, photography or sculpture, (3) an event including video or
1
Bourdieu (1984, p.18) claims that ‘nothing more clearly affirms one’s “class”, nothing
more infallibly classifies, than tastes in music’.
2
Respondents were asked ‘to include things like community events but exclude any
events that you attended as part of your job, or events produced by a school or 6th form
college’.
3
electronic art, (4) a craft exhibition (not craft markets), and (5) a cultural
festival. The binary response to these five items are indicators of cultural
participation in the domain of visual arts.
Finally, there are seven binary indicators for reading: whether in the past
12 months the respondent have (1) visited a public library, (2) attended an
event connected with books or writing, (3) bought a novel, or book of stories,
poetry or plays for himself or herself, (4) read work of fiction, play, novel or
story, (5) read poetry, (6) read biography, (7) read non-fiction/factual or
other types of book.3
We report the overall rates of cultural participation in Table 1. There is,
as expected, considerable variation in the popularity of the various activities.
For example, over a 12 month period, almost two third of the respondents
have visited a cinema, but only 2% have been to the ballet. Similarly, while
39% of the respondents have visited a museum or art gallery, only 8% have
been to a, perhaps more avant-garde, event of video or electronic art. In the
domain of reading, 61% of the respondents have read a novel, but only 8%
have read poetry.
Table 1: Percentage of respondents who have taken part in various cultural
activities in the past 12 months.
Theatre & cinema
Ballet
1.9
Other dance 12.7
Pantomime
14.6
Musical
25.4
Play/drama 29.0
Cinema
62.7
Visual arts
Video or electronic art
Cultural festival
Craft exhibition
Exhibition
Museum/art gallery
7.7
11.0
18.5
21.0
38.7
Reading
Read poetry
Book event
Read biography
Read non-fiction
Public library
Buy book
Read fiction
8.0
9.2
25.8
41.6
45.3
54.3
60.7
In the analysis that follows, we shall first explore the structure of cultural
participation in the three domains using latent class models. We then study
the social stratification of latent class membership with bivariate and multivariate tools. Our analysis of the three domains will proceed in a separate
but parallel fashion, which in fact mirrors our analysis of musical consumption. This, we hope, will bring out the similarities and the differences in
cultural stratification across domains.
3
Regarding the four items on types of book, the question wordings are as follows: ‘Can
you tell me what sort of things you have read for pleasure over the last 12 months . . . ’
4
3
Results
3.1
Latent class measurement models
The indicators of each domain can be thought of as forming a n-way contingency table with 2n cells, where n is the number of indicators of that domain.
To understand the structure of cultural consumption, we fit a series of latent
class models to the three contingency tables separately.4
Table 2: Latent class measurement models fitted to data on cultural participation in the domains of theatre and cinema, the visual arts and reading.
model
# classes
G2
Theatre &
cinema
1
2
3
1
2
2a
1583.64
268.16
53.22
Visual arts
1
2
3
1
2
3
Reading
1
2
3
4
5
1
2
3
4
4b
p
BIC
57
50
49
0.000
0.000
0.315
1113.52
-144.22
-350.91
1826.45
121.45
21.52
26
20
14
0.000
0.000
0.089
1612.01
-43.50
-93.95
3511.38
753.65
297.24
159.84
119.62
120
112
104
96
95
0.000
0.000
0.000
0.000
0.045
2521.65
-170.10
-560.52
-631.95
-663.92
df
Note: a A local dependence term is included in this model to account for the association
between ballet and other dance event. b A local dependence term is included to account
for the association between visiting a public library and attending a book event.
The results of our latent class analysis are reported in Table 2. It can be
seen that, using the conventional criterion of 5% type I error, model 3 achieves
a satisfactory fit with the data on theatre and cinema. This model postulates
two latent classes, but it also includes a local dependence term to account for
the residual association between the two indicators of dance events: ballet
and other dance. As can be seen from Table 2, this one parameter reduces
the deviance of model 2 by 214.94. Coming to the visual arts, the threeclass model achieves a satisfactory fit with the data. In the case of reading,
however, even model 4 which postulates four latent classes fails to fit the data.
4
See Chan and Goldthorpe (2005) for more detailed discussion of the analytical techniques we use in this paper.
5
Inspection of the bivariate residuals of this model suggests that the lack of fit
is primarily due to an especially strong association between visiting public
libraries and attending book events (perhaps because many book events take
place in public libraries). Indeed, when a local dependence term is added to
account for the residual association between these two indicators, the four
latent class model (model 5) fits the data just about satisfactorily.
What are the characteristics of the latent classes that we identified? In
Table 3 we report the relative size of the latent classes and the conditional
probabilities of cultural participation given membership in each latent class.5
In the domain of theatre and cinema, the first latent class accounts for just
under two third of the sample. Its members are moderately enthusiastic
cinema-goers (p = .48), but they very rarely take part in any other cultural
activities in this domain. In other words, they are a class of theatre Univores
(U). By comparison, members of the other latent class (38% of the sample)
are much more likely to attend all activities, and we refer to them as Threatre
Omnivores (O).6
In the domain of visual arts, members of the largest latent class, which
account for 59% of the respondents, rarely take part in any of the five cultural
activities that we distinguish. We refer to them as the Inactives (I). Members
of the second latent class, which constitute about a third (34%) of the sample,
are quite likely to visit museums or art galleries (p = .81). The probability of
them visiting exhibitions of art, photography or sculpture is modest (p = .42),
and they are much less likely to attend other cultural activities. This group
could be labelled Cultural Paucivores (P).7 Finally, the third latent class is
a class of Omnivores (O). They amount to only 7% of the sample, but they
are much more likely than members of the other two classes to participate
in all five activities in the visual arts. The relevant probabilities range from
p = .48 for craft exhibition to p = .97 for museum/art gallery.
Coming to the domain of reading, we see that members of the first latent class, which makes up 45% of the sample, are moderate users of public
libraries (p = .52) and quite keen book-buyers (p = .78). But they very
rarely go to book events (p = .08), and they restrict their reading largely
to fictions (p = .90). In other words, they are a class of reading paucivores,
which we shall refer to as Fiction-readers (F). Latent class 2 accounts for
almost one third of the sample (31%). Its members rarely go to libraries,
attend book events or buy books. Furthermore, their probability of reading
5
We also report in Table 3, in parentheses, the relative size of the latent classes after
modal class assignment, see discussion in Section 3.2.
6
The probability of theatre Omnivores going to the ballet is low at p = .05, but this
reflects the general ‘exclusiveness’ of ballet (see Table 1).
7
We thank our colleague Paolo Crivelli for suggesting the term ‘Paucivores’ to us.
6
Table 3: Estimated size of the latent classes and the conditional probabilities
of taking part in various cultural activities under our preferred models.
Theatre &
cinema
relative size
post-assignmenta
Ballet
Other dance
Pantomime
Musical
Play/drama
Cinema
1
2
0.625 0.375
(0.642) (0.358)
0.001 0.050
0.056 0.246
0.067 0.279
0.069 0.562
0.061 0.671
0.480 0.871
Visual arts
relative size
post-assignment
Video or electronic art
Cultural festival
Craft exhibition
Exhibition
Museum/art gallery
1
2
3
0.586 0.344 0.070
(0.582) (0.371) (0.047)
0.092 0.252 0.632
0.040 0.120 0.644
0.035 0.067 0.478
0.004 0.416 0.922
0.071 0.809 0.966
Reading
relative size
post-assignment
Read poetry
Book event
Read biography
Read non-fiction
Public library
Buy book
Read fiction
1
2
3
4
0.446 0.307 0.155 0.092
(0.420) (0.326) (0.167) (0.087)
0.028 0.003 0.413 0.027
0.082 0.007 0.282 0.105
0.256 0.037 0.746 0.187
0.337 0.090 0.951 0.984
0.523 0.197 0.740 0.481
0.781 0.008 0.944 0.497
0.897 0.164 0.938 0.121
Note: a Relative size of the latent classes after modal latent class assignment, see
Section 3.3.
7
any types of book is either the lowest or, in one case, the second lowest of
all latent classes identified. This is a class of Non-readers (NR). By contrast,
members of the third latent class (16% of the sample) are Ominvore-readers
(O). Of all four classes of readers, they have the highest probabilities of visiting public libraries, attending book events, and buying books. They are
also avid readers of all types of books. Finally, the fourth latent class (9% of
the sample) resembles the first latent class in their probabilities of visiting
public libraries, attending book events and buying books. But in terms of
reading, their preference is for non-fictions. So they constitute another class
of reading paucivores: the Non-fiction-readers (NF).
3.2
The distribution of latent class members by social
class and social status
To understand the social composition of the latent classes identified, we shall
examine the association between latent class membership and covariates of
interest. To do so, we first calculate, on the basis of our preferred latent class
solutions (see Table 3), the conditional probability of respondents’ membership in each of our latent classes, given their responses to the relevant indicators. All respondents with a particular response pattern are assigned to the
same latent class—that to which they have the highest, or modal, conditional
probability of belonging. This procedure of modal latent class assignment
inevitably misclassifies some respondents. But in the present application,
the level of misclassification is modest: 11%, 12% and 16% for theatre and
cinema, the visual arts and reading respectively.8
With our respondents distributed among the latent classes, we can now
examine the bivariate associations between latent class membership on the
one hand, and social class or social status on the other. By social class,
we refer to the National Statistics Socio-Economic Classification (NS-SEC),
which is in effect a new instantiation of the Goldthorpe class schema (Rose
and Pevalin, 2003). As for social status, we use a social status scale that we
developed (see Chan and Goldthorpe, 2004) on the basis of an analysis of
the occupational structure of close friendships (cf. Laumann, 1966).
Table 4 shows that in all three domains the relative size of the least
participatory latent class increases (almost) monotonically as one goes down
the class structure. Thus, the proportion of theatre Univores rises from 44%
in class 1 (higher professional and managerial occupations) to 86% in class
7 (routine occupations). For the visual Arts Inactives, the corresponding
8
This can be seen from the substantive similarity of the relative size of the latent classes
before and after modal latent class assigment, see Table 3.
8
figures are 35% and 78% respectively. Similarly, in the domain of reading,
18% of class 1 respondents are Non-readers, compared to 50% of those in
class 7. There is also a class gradient, running in the opposite direction, for
the Omnivores in the three domains. For example, in the domain of theatre
and cinema, 56% of class 1 respondents, but only 14% of class 7 respondents,
are Omnivores.
The pattern for the ‘intermediate’ latent classes is more complex. Thus,
the share of visual Arts Paucivores also falls as one goes down the class
structure, from 57% in class 1 to 21% in class 7. The same is true for Fictionreaders, although as can be seen, the share of Fiction-readers is actually
highest for class 3, which is made up of intermediate, or routine white-collar,
occupations. Finally, there is no clear class gradient for Non-fiction-readers.
Table 4: Distribution of latent class membership within social class.
social
class
1
2
3
4
5
6
7
overall
T&C
U
O
43.9 56.2
49.4 50.6
63.2 36.8
72.7 27.3
77.7 22.3
77.1 22.9
85.8 14.2
64.2 35.8
Visual arts
I
P
O
35.0 57.4 7.6
42.5 47.9 9.6
61.5 35.9 2.6
61.8 33.5 4.7
70.8 27.6 1.7
74.8 24.0 1.1
77.9 21.3 0.8
58.2 37.1 4.7
F
47.8
47.2
50.2
33.8
32.6
39.2
30.8
42.0
Reading
NR
O
18.2 25.2
20.0 25.8
26.0 15.3
44.0 10.9
46.8 10.0
44.0
9.2
49.6
8.3
32.6 16.7
NF
8.8
6.9
8.5
11.3
10.6
7.6
11.3
8.7
n
488
1023
574
275
359
620
480
3819
How does latent class membership vary by social status? In Figures 1,
2 and 3, we plot the bivariate relationship between latent class membership
and social status. These figures are organised not by cultural domains, but
rather by levels of cultural participation. It is evident from Figures 1 and
2 that in all three domains there is a very strong and, in most cases, linear
association between the status score of our 31 occupational categories and
the proportion of respondents in these categories that are found in the least
and most participatory latent classes. The associations are in the expected
directions: positive for the most participatory latent classes (the Omnivores
in the three domains), and negative for the least participatory latent classes:
theatre Univores, visual Arts Inactives, and Non-readers. Teachers and other
professionals in education are a notable outlier in Figure 2, where their shares
as Omnivores in the visual arts and reading are especially high.9
9
The category of Managers and officials, not elsewhere classified is also an outlier. But
the very small N of this category suggests that this might be due to sampling error (cf.
Table 10 in the Appendix).
9
60
40
50
percentage
70
80
90
Theatre and Cinema−−−Univores
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.6
0.4
0.6
0.4
0.6
status
50
30
40
percentage
60
70
80
Visual Arts−−−Inactives
−0.6
−0.4
−0.2
0.0
0.2
status
30
10
20
percentage
40
50
60
Reading−−−Non Readers
0
OMO
−0.6
−0.4
−0.2
0.0
0.2
status
10
Figure 1: Proportion of respondents belonging to the least participatory
latent class in the domains of theatre and cinema (top panel), visual arts
(middle panel) and reading (bottom panel) by social status.
40
10
20
30
percentage
50
60
Theatre and Cinema−−−Omnivores
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.6
status
Visual Arts−−−Omnivores
15
TPE
10
0
5
percentage
APB
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.6
status
Reading−−−Omnivore Readers
OMO
percentage
10
20
30
40
TPE
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.6
status
11
Figure 2: Proportion of respondents belonging to the most participatory
latent class in the domains of theatre and cinema (top panel), visual arts
(middle panel) and reading (bottom panel) by social status.
40
20
30
percentage
50
60
Visual Arts−−−Paucivores
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.6
0.4
0.6
status
Reading−−−Fiction Readers
40
25
30
35
percentage
45
50
55
APH
−0.6
−0.4
−0.2
0.0
0.2
status
Reading−−−Non−Fiction Readers
CCW
SMM
12
PSW
8
6
percentage
10
GMA
2
4
SDC
FRC
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.6
status
12
Figure 3: Proportion of respondents belonging to the latent classes with
intermediate level of cultural participation in the domains of visual arts (top
panel) and reading (middle and bottom panels) by social status.
As for the latent classes with intermediate levels of cultural participation,
Figure 3 shows that there is also a very strong and linear association between
social status and the probability of being a visual Arts Paucivore. In this
sense, the Paucivores and the Omnivores in the visual arts are quite similar.
However, in the domain of reading the two intermediate latent classes are
more distinctive. For Fiction-readers, the association with status is curvilinear, with the proportion of Fiction-readers reaching a peak in the middlerange of our status order (though the rise in the share of Fiction-readers on
the left of the peak is much steeper than the drop from the right). The five
occupational categories with the highest proportion of Fiction-readers are in
fact all female-dominated clerical or social service occupations of intermediate social status. In descending order, they are: Associate professionals
in health, Administrative officers and assistants, Secretaries and receptionists, Filing and record clerks, and Childcare workers (see Table 10). Because
many occupations with intermediate status scores are in NS-SEC class 3,
this corresponds well to the pattern of Table 4. Finally, the proportion of
Non-fiction-readers declines rather gently with status. But the much greater
dispersion of the data points in this plot suggests that for this latent class,
the relationship with status is weaker.
3.3
The social character of types of cultural consumer:
Multivariate analysis
We now turn to multivariate analysis. In our regression models, we shall
control for various socio-demographic variables. But in terms of stratification variables, apart from class and status, we shall also include educational
attainment and income. It is important to include multiple stratification
measures in the analysis because the underlying mechanisms that they imply
are very different. For example, an income gradient in cultural participation,
if present, might simply reflect the ability to pay. But a gradient in education, given that other stratification measures are already controlled for in the
model, might reflect the operation of an individual psychological variable—
i.e. information processing capacity—as suggested by various proponents of
‘empirical aesthetics’ (cf. Ganzeboom, 1982).10 As for class and status, from
the Weberian perspective, we would expect that status, rather than class, to
10
The argument here is that the higher individuals’ information processing capacity, the
greater must be the information content of the cultural forms in which they participate if
they are to derive satisfaction from them. Thus, the association between ‘high’ culture and
educational attainment is due to the facts (a) that ‘high’ culture has, on average, a higher
level of information content than ‘low’ culture and (b) that education is crucially involved
in, and is thus a good proxy for, the information processing capacity of individuals.
13
have greater relevance in explaining lifestyle and cultural consumption. To
test these ideas, we carry out, for each cultural domain separately, regression model with latent class membership as the dependent variable. Some
descriptive statistics of our covariates are listed in Table 5.
We report the results of our regression models in Tables 6, 7 and 8.
Comparing across these three tables, and also keeping in mind the pattern
we observed in relation to music (see Chan and Goldthorpe, 2005, Table
7), we see that marital status consistently has no association with latent
class membership in all four domains. The following covariates, however,
are associated with latent class membership in three of the four domains.
Women are generally more active than men in cultural participation. They
are less likely to be found in the least participatory latent class, except for
the visual arts where there is no gender difference. We also observe an age
gradient in cultural consumption, with older people being less likely found
as Univores or Inactives, except in the domain of theatre and cinema where
the age parameter is insignificant.11 Also, perhaps not surprisingly, having
preschoolers (less than four year old) at home restricts cultural participation,
except for music where the effect is insignificant.
Regional differences are observed for visual arts (and music), but not for
theatre and cinema or reading. Compared with Londoners, respondents living in the Midlands have a higher probability of belonging to visual Arts
Inactives rather than Paucivores or Omnivores. Those living in the North or
the South-East are also more likely members of visual Arts Inactives rather
than Paucivores. In our previous paper on music, we saw that respondents
living in the North or the Midlands are more likely musical Univores. If the
region parameters measure the availability of facilities for cultural participation, then access to museums, galleries or other facilities for the visual arts,
and access to concert halls or other music venues, in some regions should be
a matter of policy concern.
Coming to the stratification variables, we note that respondents with
higher income are more likely theatre Omnivores rather than Univores or
visual Arts Paucivores rather than Inacives. But income has no effect on
latent class membership in the domains of reading or music. What proves
to be very important in all domains is education. The parameter estimates
for education are often (nearly) monotonic.12 Compared with the reference
category of the unqualified, more education is associated with a higher prob11
Whether this gradient should be interpreted as an ageing or cohort effect is a question
for future investigation.
12
Non-monotonicity of the education gradient, if present, is often found in relation to
the category of sub-degree. Many respondents in this category had tertiary education of
a vocational rather than an academic type.
14
Table 5: Descriptive statistics of covariates.
femalea
N
2110
%
55.3
Single (reference category)
Married or cohabiting
Separated, divorced or widowed
700
2473
646
18.3
64.8
16.9
651
779
623
17.1
20.4
16.3
493
1141
1150
617
418
12.9
29.9
30.1
16.2
11.0
865
508
889
518
347
692
22.7
13.3
23.3
13.6
9.1
18.1
488
1023
574
275
359
620
480
min.
20
260
-0.598
12.8
26.8
15.0
7.2
9.4
16.2
12.6
max.
64
37700
0.564
children 0–4b
children 5–10b
children 11–15b
London (reference category)
The North
Midlands and East Anglia
South East
South West
no qualifications (reference category)
CSE, etc.
O-levels
A-levels
post-secondary qualifications
degree
Class
Class
Class
Class
Class
Class
Class
1—higher managerial & professional occupations (ref.cat.)
2—lower managerial & professional occupations
3—intermediate occupations
4—small employers and own-account workers
5—lower supervisory & technical occupations
6—semi-routine occupations
7—routine occupations
mean
s.d.
age
42.1
11.8
annual incomec
15573 10863
status
-0.001 0.365
Note:
a
Male is reference category.
b
Not having children in the respective age ranges are the reference categories.
c
The income variable in the Arts Council data set is originally coded in terms of 32
income brackets of variable width. In our analysis, we have assigned respondents to the
midpoint of the income bracket to which they belong.
15
ability of belonging to the more partcipatory latent classes, such as theatre
Omnivores rather than Univores, all other types of readers rather than Nonreaders, Omnivore-readers rather than Fiction-readers, visual Arts Paucivores or Omnivores rather than Inactives.13
In all four domains status is in general a significant predictor of latent
class membership when the least participatory latent class is used as the reference category.14 Thus, people of higher social status are less likely theatre
Univores, visual Arts Inactives, Non-readers or musical Univores. However,
when the most participatory latent class is chosen as the reference category
in the domains of reading and the visual arts, we also see that Fiction-readers
are on average of lower social status than Omnivore-readers. Similarly, in the
domain of music, Omnivore-listeners are of lower status than the true musical
Omnivores. There is, however, no significant status difference between Nonfiction-readers and Omnivore-readers, or between visual Arts Paucivores and
Omnivores.
Finally, having controlled for income, education and social status, social
class turn out to be generally non-significant. This is consistent with the
Weberian view of social stratification that we defend (Chan and Goldthorpe,
2004). It can be argued that because we use six parameters to represent
the effects of social class but only one parameter for social status, it is more
likely that social status rather than social class will turn out to be statistically
significant. This argument has some force. If we use the fivefold version of
the class schema and discretise our status order into four broad levels,15 and
re-run the regression, then the effect of class and that of status on latent class
membership are more comparable. As we show in Table 11 in the Appendix,
under such a parameterisation, the dummy for status level 4 is, in all but
one case, statistically significant, but so are the dummies for class 3, and for
13
In the domain of visual arts, people with tertiary education are also more likely to be
Omnivores rather than Paucivores.
14
There are two exceptions. In the domain of reading, the contrast between Non-fictionreaders and Non-Readers, where the parameter for status is marginally insignificant with
p = .06. The status parameter also fails to reach statistical significance in the contrast
between Omnivore-listeners and Univores in the domain of music with p = .10 (see Chan
and Goldthorpe, 2005, Table 7).
15
Although the status order is essentially hierarchical in nature, we argue that it could
be discretised into four ordered levels. The first level comprises of occupations in categories
1–7 in the ranking and are essentially non-manual in character, and occupations in the
second level, categories 8–18, only slightly less so. The third level comprises of categories
19–25 which covers occupations, falling mainly within the service sector, that tend to have
both mon-manual and manual components. Finally, occupations at level 4, categories
26–31, require the performance of predominantly manual tasks. For details, see Chan and
Goldthorpe (2004, p.389). See also Table 9 for representative occupations within each of
the 31 occupational categories that we distinguish.
16
classes 6 & 7. We note that this is not true in the domain of music. Using
the same parameterisation, the four class dummies remain statistically nonsignificant, while all three status dummies are significant for the contrast
between true Omnivores and Univores, and the dummy for status level 4 is
significant for the contrast between Omnivore-listeners and Univores. Thus,
one might argue that status effect is more clear cut in the domain of music
than in the other three cultural domains that we examine in this paper.
Table 6: Binary logit model: latent class in the domain of theatre and cinema
as the dependent variable.
O vs U
female
married
separated
age
child (0–4)
child (5–10)
child (11–15)
The North
Midlands
South East
South West
income
CSE/others
O-levels
A-levels
sub-degree
degree
class 2
class 3
class 4
class 5
class 6
class 7
status
constant
β̂
0.615∗∗
0.148
0.188
0.005
−0.562∗∗
0.070
0.088
−0.231
−0.207
0.083
−0.189
0.026∗∗
0.169
0.668∗∗
1.130∗∗
1.027∗∗
1.223∗∗
0.078
−0.161
−0.205
−0.134
−0.199
−0.507∗
0.631∗∗
−2.118∗∗
s.e.
(0.092)
(0.112)
(0.139)
(0.004)
(0.113)
(0.100)
(0.105)
(0.124)
(0.123)
(0.135)
(0.153)
(0.005)
(0.152)
(0.128)
(0.145)
(0.160)
(0.151)
(0.126)
(0.160)
(0.203)
(0.218)
(0.195)
(0.230)
(0.179)
(0.292)
Note: * p < 0.05, ** p < 0.01.
17
Table 7: Multinomial logit model: latent class in the domain of visual arts
as the dependent variable.
female
married
separated
age
child (0–4)
child (5–10)
child (11–15)
The North
Midlands
South East
South West
income
CSE/others
O-levels
A-levels
sub-degree
degree
class 2
class 3
class 4
class 5
class 6
class 7
status
constant
P vs I
β̂
s.e.
0.079
(0.090)
0.037
(0.111)
−0.050
(0.138)
0.022∗∗ (0.004)
−0.235∗ (0.111)
0.161
(0.100)
−0.078
(0.106)
−0.366∗∗ (0.124)
−0.390∗∗ (0.123)
−0.483∗∗ (0.138)
−0.297
(0.152)
0.010∗ (0.005)
0.525∗∗ (0.138)
0.631∗∗ (0.123)
1.068∗∗ (0.142)
1.194∗∗ (0.157)
1.652∗∗ (0.153)
0.040
(0.133)
−0.225
(0.164)
−0.059
(0.205)
−0.089
(0.217)
−0.253
(0.198)
−0.227
(0.224)
0.684∗∗ (0.180)
−1.923∗∗ (0.293)
O vs I
β̂
s.e.
0.223
(0.192)
−0.200
(0.239)
0.180
(0.295)
0.026∗∗ (0.009)
−0.639∗ (0.285)
0.260
(0.232)
0.039
(0.252)
−0.089
(0.253)
−0.880∗∗ (0.279)
−0.150
(0.270)
−0.174
(0.321)
0.006
(0.009)
1.220∗ (0.499)
1.072∗ (0.462)
1.849∗∗ (0.471)
2.219∗∗ (0.469)
3.260∗∗ (0.450)
0.613∗ (0.241)
−0.396
(0.376)
0.699
(0.411)
0.073
(0.554)
−0.480
(0.514)
−0.325
(0.646)
1.229∗∗ (0.402)
−5.461∗∗ (0.688)
Note: * p < 0.05, ** p < 0.01.
18
P vs
β̂
−0.144
0.237
−0.230
−0.003
0.404
−0.099
−0.117
−0.277
0.490
−0.334
−0.123
0.004
−0.694
−0.441
−0.782
−1.025∗
−1.608∗∗
−0.573∗
0.171
−0.759
−0.162
0.227
0.098
−0.544
3.538∗∗
O
s.e.
(0.188)
(0.234)
(0.290)
(0.009)
(0.283)
(0.229)
(0.250)
(0.245)
(0.272)
(0.262)
(0.313)
(0.009)
(0.506)
(0.467)
(0.475)
(0.470)
(0.450)
(0.229)
(0.370)
(0.402)
(0.551)
(0.511)
(0.645)
(0.397)
(0.678)
Table 8: Multinomial logit model: latent class in the domain of reading as the dependent variable.
19
female
married
separated
age
child (0–4)
child (5–10)
child (11–15)
The North
Midlands
South East
South West
income
CSE/others
O-levels
A-levels
sub-degree
degree
class 2
class 3
class 4
class 5
class 6
class 7
status
constant
F vs
β̂
0.915∗∗
0.008
−0.105
0.013∗∗
−0.365∗∗
0.163
0.036
−0.129
−0.105
0.009
−0.030
−0.003
0.594∗∗
0.830∗∗
1.121∗∗
1.118∗∗
1.671∗∗
0.002
−0.165
−0.349
−0.241
−0.308
−0.354
0.722∗∗
−1.168∗∗
NR
s.e.
(0.101)
(0.121)
(0.150)
(0.004)
(0.118)
(0.110)
(0.114)
(0.143)
(0.143)
(0.161)
(0.175)
(0.005)
(0.139)
(0.126)
(0.155)
(0.182)
(0.182)
(0.168)
(0.200)
(0.238)
(0.248)
(0.230)
(0.253)
(0.200)
(0.334)
O vs
β̂
0.939∗∗
−0.180
−0.090
0.034∗∗
−0.508∗∗
0.147
−0.239
−0.313
−0.290
−0.058
−0.051
0.008
0.938∗∗
1.404∗∗
1.992∗∗
1.971∗∗
2.828∗∗
0.310
−0.082
−0.206
0.203
−0.071
0.170
1.189∗∗
−3.880∗∗
NR
s.e.
(0.131)
(0.158)
(0.195)
(0.006)
(0.166)
(0.149)
(0.161)
(0.180)
(0.178)
(0.197)
(0.218)
(0.007)
(0.227)
(0.197)
(0.222)
(0.243)
(0.238)
(0.192)
(0.245)
(0.307)
(0.324)
(0.295)
(0.333)
(0.262)
(0.435)
NF vs NR
β̂
s.e.
0.177
(0.154)
−0.004
(0.189)
0.006
(0.230)
0.021∗∗ (0.007)
−0.059
(0.183)
0.050
(0.173)
0.053
(0.178)
−0.276
(0.208)
−0.405
(0.211)
−0.110
(0.233)
−0.388
(0.266)
−0.010
(0.008)
0.842∗∗ (0.212)
0.981∗∗ (0.197)
0.993∗∗ (0.249)
1.085∗∗ (0.283)
1.499∗∗ (0.276)
−0.167
(0.252)
−0.121
(0.304)
0.027
(0.348)
0.057
(0.372)
−0.297
(0.353)
0.179
(0.376)
0.570
(0.308)
−2.517∗∗ (0.507)
Note: * p < 0.05, ** p < 0.01.
F vs
β̂
−0.024
0.189
−0.016
−0.021∗∗
0.144
0.016
0.275
0.184
0.186
0.067
0.022
−0.011
−0.344
−0.574∗∗
−0.871∗∗
−0.853∗∗
−1.158∗∗
−0.309∗
−0.083
−0.143
−0.444
−0.237
−0.524
−0.466∗
2.711∗∗
O
s.e.
(0.116)
(0.141)
(0.176)
(0.005)
(0.154)
(0.133)
(0.146)
(0.157)
(0.154)
(0.168)
(0.190)
(0.006)
(0.225)
(0.192)
(0.210)
(0.222)
(0.210)
(0.151)
(0.205)
(0.275)
(0.293)
(0.259)
(0.302)
(0.238)
(0.386)
NF vs O
β̂
s.e.
−0.762∗∗ (0.169)
0.176
(0.208)
0.096
(0.255)
−0.013
(0.007)
0.449∗ (0.212)
−0.097
(0.193)
0.292
(0.204)
0.038
(0.225)
−0.114
(0.225)
−0.052
(0.245)
−0.337
(0.285)
−0.017∗ (0.009)
−0.096
(0.280)
−0.423
(0.248)
−0.999∗∗ (0.289)
−0.886∗∗ (0.314)
−1.329∗∗ (0.299)
−0.478
(0.245)
−0.039
(0.311)
0.232
(0.380)
−0.147
(0.409)
−0.226
(0.379)
0.009
(0.417)
−0.618
(0.340)
1.363∗ (0.555)
3.4
Magnitude of education, income and status effects
In substative terms, how strong are the education, income and status effects?
Let us consider a hypothetical woman who is forty years old, childless and
lives in London. By setting her education, income and social status at different values, and then computing the probability of her belonging to the
various latent classes, we could gauge the substantive magnitude of these
effects.16
0.8
Theatre & cinema−−−Omnivore
0.5
0.2
0.3
0.4
probability
0.6
0.7
no qual.
O−levels
degree
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.6
status
Figure 4: Predicted probability of being an theatre Omnivore by education
and social status.
Note: Other covariates fixed as follows: Forty years old female Londoner, with income of
£25,000 and no children.
For example, the slope of the lines of Figure 4 measures the strength of
the status effect in the domain of theatre and cinema. It can be seen that,
irrespective of education, the probability that our hypothetical woman is a
theatre Omnivore would be about 24% higher if she is at the top rather than
at the bottom of our status hierarchy. The education effect is also substantial
16
The predicted probabilities are estimated under a model very similar to those reported
in Tables 6, 7 and 8, but the class and marital status dummies have been dropped. Note
also that when we vary the effects of education and status in Figures 4, 6 and 8, income
is fixed at 25,000, while education is fixed as O-levels when we vary income and status
in Figures 5, 7 and 9.
20
(see the vertical distance between the lines): at all status levels, the maximum
education effect (i.e. that between university degree and no qualifications)
is about 29 percentage points. The strength of the income effect is more
modest (see the slope of the lines of Figure 5): other things being equal and
irrespective of status level, if the annual income of our hypothetical woman is
thirty-five rather than fifteen thousand pounds, her probability as a theatre
Omnivore would increase by 14%.
0.8
Theatre & cinema−−−Omnivore
0.5
0.2
0.3
0.4
probability
0.6
0.7
PMO
MPS
HP
15
20
25
30
35
income
Figure 5: Predicted probability of being an theatre Omnivore by income and
social status.
Note: Other covariates fixed as follows: Forty years old female Londoner, with O-Levels
and no children.
We report similar predicted probabilities for the visual arts in Figures 6
and 7. Here the pattern is more complex. Figure 6 shows that the probability
that our hypothetical woman is a visual Arts Inactive declines almost linearly
with status. The decline is substantial in magnitude (20–25% across the
whole status range), and the rate of decline is roughly the same for the three
education levels considered. We also see that with increasing status our
hypothetical woman is more likely a Paucivore or an Omnivore, depending
on her education. If she is a university graduate, then her probability as an
Omnivore increases by 12% across the entire status range. The rise is only
4% if she has O-levels and 2% if she has no qualifications. Correspondingly,
the middle panel of Figure 6 shows that with increasing status, the rise
21
0.8
Visual arts−−−Inactive
0.5
0.4
0.1
0.2
0.3
probability
0.6
0.7
no qual.
O−levels
degree
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.6
0.4
0.6
0.4
0.6
status
0.7
Visual arts−−−Paucivore
0.5
0.4
0.2
0.3
probability
0.6
no qual.
O−levels
degree
−0.6
−0.4
−0.2
0.0
0.2
status
0.25
Visual arts−−−Omnivore
0.15
0.10
0.00
0.05
probability
0.20
no qual
O−levels
degree
−0.6
−0.4
−0.2
0.0
0.2
status
22
Figure 6: Predicted probabilities of latent class membership in the visual
arts by education and social status.
Note: Other covariates fixed as follows: Forty years old female Londoner, with income of
£25,000 and no children.
0.8
Visual arts−−−Inactive
0.6
0.5
0.3
0.4
probability
0.7
PMO
MPS
HP
15
20
25
30
35
income
0.4
0.3
probability
0.5
0.6
Visual arts−−−Paucivore
0.2
PMO
MPS
HP
15
20
25
30
35
income
0.03
0.01
0.02
probability
0.04
0.05
0.06
Visual arts−−−Omnivore
0.00
PMO
MPS
HP
15
20
25
30
35
income
23
Figure 7: Predicted probabilities of latent class membership in the visual
arts by status and income.
Note: Other covariates fixed as follows: Forty years old female Londoner, with O-Levels
and no children.
in the probability as a Paucivore is greater for people with O-levels or no
qualifications than for graduates.17
Figure 7 shows the magnitude of the income effect in the domain of visual arts. Overall, the income effect seems to be weaker, and affecting the
Inactives and Paucivores only. More specifically, irrespective of status, the
probability that our hypothetical woman is an Inactive would decline by a
moderate six percent if her annual income is thiry-five rather than fifteen
thousand pounds. Such a decline is offset by an increase of the same amount
in the probability of belonging to the Paucivores. Furthermore, the lines
of bottom panel of Figure 7 are very flat, suggesting that the probability of
Omnivore membership is insensitive to income difference. These patterns are
consistent with the results reported in Table 7.
Finally, coming to reading, Figure 8 shows that both education and status has substantial effect on membership in certain types of readership. It
also reveals a good deal of interaction between education and status. For
example, as we traverse across the whole status range the probability that
our hypothetical woman is an Omnivore-reader increases by about 13% if she
is a university graduate. The increase would be smaller if she is less qualified
(11% and 8% for O-levels or no qualifications respectively). Correspondingly,
we see a decline with status in the probability of being a Non-reader. But
here the decline is largest for the least qualified: 27% for no qualifications,
19% for O-levels and 10% for university degree.
If our hypothetical woman has no qualifications, the probability of her
as a Fiction-reader increases by 20% across the whole status range. The
increase would be 11% if she has O-levels, but only 1% if she is a graduate.
At all three education levels, the proportion of Non-fiction-readers is quite
low. And their share declines very modestly with status (about 4%, 2% and
1% for degree, O-levels and no qualifications respectively).
Figure 9 shows that the income effect on reading is very weak. Omnivore
readership would rise by three to four percent over the entire income range
for all three status groups considered, compensated by a decline of Fiction
readership or Non-Fiction readership of no more than 2%.
4
Summary and discussion
In this paper, we study the relationship between social stratification and
cultural consumption in three domains: (1) theatre and cinema, (2) the visual
17
Note that our regression model does not contain any interaction term. The apparent
interaction effect is due to the logit link function of multinomial logit models which relates
the linear predictor to the predicted probabilities in a non-linear fashion.
24
Reading−−−Omnivore Reader
0.5
0.6
Reading−−−Non−Reader
0.3
probability
0.2
0.3
None
O−L
Degree
0.0
0.0
0.1
0.1
0.2
probability
0.4
0.4
0.5
None
O−L
Degree
−0.4
−0.2
0.0
0.2
0.4
0.6
−0.6
−0.4
−0.2
0.0
0.2
status
status
Reading−−−Fiction Reader
Reading−−−Non−Fiction Reader
0.4
0.45
probability
0.50
0.10
0.55
None
O−L
Degree
0.35
0.05
0.40
probability
0.6
0.15
0.60
−0.6
−0.6
−0.4
−0.2
0.0
0.2
0.4
0.00
0.30
None
O−L
Degree
0.6
−0.6
status
−0.4
−0.2
0.0
0.2
0.4
0.6
status
Figure 8: Predicted probabilities of latent class membership in reading by
education and social status.
Note: Other covariates fixed as follows: Forty years old female Londoner, with income of
£25,000 and no children.
25
Reading−−−Omnivore Reader
0.30
0.40
Reading−−−Non−Reader
0.20
probability
0.25
20
25
30
35
15
20
25
30
income
income
Reading−−−Fiction Reader
Reading−−−Non−Fiction Reader
0.60
15
0.10
0.08
0.50
probability
0.12
0.55
35
PMO
MPS
HP
0.14
PMO
MPS
HP
0.40
0.06
0.45
probability
PMO
MPS
HP
0.10
0.10
0.15
0.15
0.20
probability
0.30
0.25
0.35
PMO
MPS
HP
15
20
25
30
35
15
income
20
25
30
35
income
Figure 9: Predicted probabilities of latent class membership in reading by
status and income.
Note: Other covariates fixed as follows: Forty years old female Londoner, with O-Levels
and no children.
26
arts and (3) reading. The results we obtain are very similar to those reported
in our previous paper on musical consumption (Chan and Goldthorpe, 2005).
To elaborate, a small number of distinct and quite interpretable latent classes
can be identified in each field using latent class models. Just as two third
of our respondents are musical univores, most of them are not very active in
the three cultural fields considered here. Thus, theatre univores (63%) and
visual arts inactives (58%) are the largest latent classes in their respective
field, and Non-readers is the second largest class in the domain of reading
(31%).
There are also strong similarities across cultural fields in the social characteristics of the latent classes. For example, educational attainment is invariably a very strong predictor of latent class membership. Respondents with
more education are consistently and substantially more likely members of
the most participatory latent class in all fields. As we argued above, because
we have controlled for income, social class and social status in our regression
models, this education gradient is best interpreted as reflecting the operation
of an individual level mechanism (i.e. the information processing capacity of
individuals) rather than that of social stratification.
Furthermore, we show that consistent with the Weberian view of social
stratification, it is status rather than class which is more directly relevant to
understanding patterns of cultural consumption. This result is again consistent with the findings we reported in our previous paper on music. As might
be expected, the effect of income is significant mainly for theatre and cinema
where ticket price is often quite considerable. However, in the visual arts and
reading, where admissions or participation is generally quite cheap or even
free, income is generally not an important predictor.
Coming to our geographic and demographic variables, our results are also
quite sensible. For example, we see that the regional dummies are significant in the field of visual arts, reflecting the geographical concentration of
museums and galleries in London. Because theatres, cinemas, libraries and
bookshops are much less concentrated geographically, these dummies are
generally not significant for the other two domains.
Overall, the results of this paper confirm our earlier findings, and the
main conclusions we draw in relation to music also apply in the present
case. Thus, the fact that we can identify distinct, well-defined and quite
interpretable latent classes in each of the three domains must throw doubts
on the individualisation argument. And as these latent classes are clearly
stratified by educational attainment and social status, the credibility of the
individualisation argument is further undermined. Secondly, a fairly large
group of univores or inactives can be found in each domain. While they
could be considered as the mass in cultural consumption, we see no evidence
27
that a cultural elite exists in any domain, contrary to Bourdieu’s homology
argument. Finally, our result is most congruent with the omnivore–univore
argument. But we also offer revision of that argument by pointing to groups
of Paucivores whose relation to the other latent classes or to social stratification is more complex than has so far been recognised.
References
Bauman, Z. (1988) Freedom. Milton Keynes, Open University Press.
Bourdieu, P. (1984) Distinction: A Social Critique of the Judgement of Taste.
London, Routledge & Kegan Paul.
Bryson, B. (1996) “Anything but Heavy Metal”: Symbolic exclusion and
musical dislikes, American Sociological Review, 61(5), 884–899.
Bryson, B. (1997) What about the univores? musical dislikes and groupbased identity construction among Americans with low level of education, Poetics, 25, 141–156.
Chan, T. W. and Goldthorpe, J. (2005) Social stratification and cultural
consumption: Music in England. under review.
Chan, T. W. and Goldthorpe, J. H. (2004) Is there a status order in contemporary British society? Evidence from the occupational structure of
friendship, European Sociological Review, 20(5), 383–401.
Coulangeon, P. (2003) La stratification sociale des gôuts musicaux, Revue
francaise de sociologie, 44, 3–33.
Featherstone, M. (1987) Lifestyle and consumer culture, Theory, Culture and
Society, 4(1), 55–70.
Ganzeboom, H. B. (1982) Explaining differential participation in highcultural activities: A confrontation of information-processing and
status-seeking theories, in W. Raub (ed), Theoretical Models and Empirical Analyses: Contributions to the Explanation of Individual Actions
and Collective Phenomena, E.S.–Publications, Utrecht, 186–205.
Laumann, E. O. (1966) Prestige and Association in an Urban Community.
Indianapolis, Bobbs-Merrill.
Peterson, R. A. and Kern, R. M. (1996) Changing highbrow taste: From
snob to omnivore, American Sociological Review, 61(5), 900–907.
28
Peterson, R. A. and Simkus, A. (1992) How musical tastes mark occupational
status groups, in M. Lamont and M. Fournier (eds), Cultivating Differences: Symbolic Boundaries and the Making of Inequality, University of
Chicago Press, Chicago, chapter Seven, 152–186.
Rose, D. and Pevalin, D. J. (eds) (2003) A Researcher’s Guide to the National
Statistics Socio-economic Classification. London, Sage.
Skelton, A., Bridgwood, A., Duckworth, K., Hutton, L., Fenn, C., Creaser,
C. and Babbidge, A. (2002) Arts in England: Attendance, participation and attitudes in 2001, Research report 27, Arts Council England,
London.
van Eijck, K. (2001) Social differentiation in musical taste patterns, Social
Forces, 79(3), 1163–1184.
van Rees, K., Vermunt, J. and Verboord, M. (1999) Cultural classifications
under discussion: Latent class analysis of highbrow and lowbrow reading, Poetics, 26, 349–365.
29
Table 9: The 31 occupational categories ranked by status scores and representative occupations within each category.
1
2
3
Code
HP
APB
SM
4
5
TPE
GMA
6
API
7
SET
8
9
10
11
12
13
14
15
FRC
OMO
AOA
NCC
APH
SEC
OCW
BSR
16
17
18
CCW
MPS
PDM
19
SW
20
21
HW
PSW
22
23
PSP
RWS
24
25
26
27
28
CW
SDC
SMO
TO
SMC
29
30
SMM
PMO
31
GL
Representative occupations
chartered accountants, clergy, medical practitioners, solicitors
journalists, investment analysts, insurance brokers, designers
company treasurers, financial managers, computer systems managers,
personnel managers
college lecturers, education officers and inspectors, school teachers
bank and building society managers, general managers in industry, national and local government officers
computer analysts and programmers, quantity surveyors, vocational and
industrial trainers
civil and structural engineers, clinical biochemists, industrial chemists,
planning engineers, software engineers
conveyancing clerks, computer clerks, library assistants
security managers, cleaning managers
clerical officers in national and local government
accounts assistants, bank clerks
community workers, nurses, occupational therapists, youth workers
personal assistants, receptionists, secretaries, word processor operators
general assistants, commerical and clerical assistants
buyers and purchasing officers, technical sales representatives, wholesale
representatives
educational assistants, nursery nurses
catering managers, hoteliers, publicans, shopkeepers and managers
clerks of works, farm managers, maintenance managers, transport managers, works managers
cash desk and check-out operators, sales and shop assistants, window
dressers
ambulance staff, dental nurses, nursing auxiliaries
caretakers and housekeepers, hairdressers and beauticians, travel attendants, undertakers
fire service and police officers, security guards
car park attendants, cleaners, counter-hands, couriers and messengers,
hotel porters, postal workers
bar staff, chefs, cooks, waiters and waitresses
despatch and production control clerks, storekeepers
gardeners and groundsmen, printers, textile workers, woodworkers
bus and coach drivers, lorry and van drivers, taxi drivers
bricklayers, electricians, painters and decorators, plasterers, roofers,
telephone repairmen
fitters, setters, setter-operators, sheet metal workers, turners, welders
assemblers, canners, fillers
30and packers, food processors, moulders and
extruders, routine inspectors and testers
agricultural workers, factory labourers, goods porters, refuse collectors
Table 10: Distribution of latent class membership within occupational categories.
occ.
cat.
HP
APB
SM
TPE
GMA
API
SET
FRC
OMO
AOA
NCC
APH
SEC
OCW
BSR
CCW
MPS
PDM
SW
HW
PSW
PSP
RWS
CW
SDC
SMO
TO
SMC
SMM
PMO
GL
overall
status
score
0.5643
0.5337
0.5107
0.5017
0.4114
0.3116
0.3115
0.2559
0.2355
0.2274
0.2238
0.2228
0.1539
0.1443
0.1193
0.1097
-0.0453
-0.0625
-0.1151
-0.2121
-0.2261
-0.2288
-0.2974
-0.3261
-0.3353
-0.4072
-0.4114
-0.5014
-0.5121
-0.5589
-0.5979
T&C
U
O
35.9 64.1
48.5 51.5
41.8 58.2
39.5 60.5
36.8 63.2
58.2 41.8
55.9 44.1
57.1 42.9
33.3 66.7
55.1 44.9
56.8 43.2
44.1 55.9
61.8 38.2
70.5 29.5
62.1 37.9
52.8 47.2
62.4 37.7
58.1 41.9
71.4 28.6
71.3 28.7
62.0 38.0
78.5 21.5
84.1 15.9
75.0 25.0
80.0 20.0
83.3 16.7
84.4 15.6
84.5 15.5
81.0 19.0
90.3
9.7
81.0 19.0
64.2 35.8
Visual arts
I
P
O
28.1 63.3
8.6
33.9 53.8 12.3
36.3 55.5
8.2
23.4 57.5 19.2
43.4 50.0
6.6
54.5 42.7
2.7
41.9 52.2
5.9
53.6 41.1
5.4
66.7 33.3
0.0
48.0 44.9
7.1
65.1 32.5
2.4
42.1 51.3
6.6
53.5 43.3
3.2
64.2 28.4
7.4
41.4 50.0
8.6
55.1 41.6
3.4
54.1 37.1
8.8
45.3 47.7
7.0
68.7 29.8
1.5
72.0 25.0
3.0
64.1 34.8
1.1
63.3 35.4
1.3
83.7 16.3
0.0
72.1 23.5
4.4
76.0 24.0
0.0
70.3 29.0
0.7
77.1 22.0
0.9
77.6 22.4
0.0
69.4 28.9
1.6
83.6 16.4
0.0
73.6 24.8
1.7
58.2 37.1
4.7
F
51.6
45.0
46.7
41.3
51.3
42.7
44.1
51.8
44.4
54.1
49.7
58.6
52.2
47.4
50.0
51.7
42.9
44.2
50.4
38.4
44.6
36.7
39.4
44.1
24.0
24.6
23.9
28.4
24.8
24.2
28.1
42.0
Reading
NR
O
10.9 29.7
18.7 30.4
20.9 25.3
10.2 41.3
14.5 22.4
30.9 18.2
22.1 23.5
23.2 23.2
0.0 44.4
19.4 16.3
23.7 15.4
13.2 21.7
23.6 17.8
24.2 22.1
25.9 17.2
18.0 16.9
31.2 18.8
31.4 15.1
33.2
9.5
39.6 12.8
34.8
7.6
38.0 16.5
42.3 10.6
36.8
8.8
56.0 16.0
52.9 11.6
56.0
8.3
58.6
4.3
55.4
6.6
62.3
3.9
53.7
7.4
32.6 16.7
NF
7.8
5.9
7.1
7.3
11.8
8.2
10.3
1.8
11.1
10.2
11.2
6.6
6.4
6.3
6.9
13.5
7.1
9.3
6.9
9.2
13.0
8.9
7.7
10.3
4.0
10.9
11.9
8.6
13.2
9.7
10.7
8.7
Note: a For examples of occupations within each category and other details, see Chan
and Goldthorpe (2004).
31
n
128
171
182
167
76
110
136
56
9
98
169
152
157
95
58
89
170
86
262
164
92
79
208
68
25
138
109
116
121
207
121
3819
Table 11: Effects of class and status on latent class membership.
Theatre & cinema
O vs U
β̂
s.e.
class 3a
class 4
class 5
class 6 & 7
status level 2b
status level 3
status level 4
−0.247∗
−0.289
−0.222
−0.437∗∗
−0.058
−0.192
−0.571∗∗
(0.124)
(0.174)
(0.190)
(0.157)
(0.109)
(0.158)
(0.183)
Visual arts
P vs I
class 3
class 4
class 5
class 6 & 7
status level 2
status level 3
status level 4
O vs I
β̂
s.e.
β̂
s.e.
−0.257∗
−0.151
−0.188
−0.328∗
−0.119
−0.361∗
−0.493∗∗
(0.125)
(0.171)
(0.184)
(0.155)
(0.113)
(0.159)
(0.177)
−0.969∗∗
0.253
−0.229
−0.851
−0.013
−0.731
−1.478∗∗
(0.316)
(0.352)
(0.530)
(0.437)
(0.207)
(0.414)
(0.541)
Reading
O vs NR
β̂
s.e.
F vs NR
β̂
s.e.
class 3
class 4
class 5
class 6 & 7
status level 2
status level 3
status level 4
−0.188
−0.373
−0.249
−0.351∗
0.012
−0.207
−0.621∗∗
(0.150)
(0.191)
(0.201)
(0.174)
(0.141)
(0.183)
(0.198)
−0.339
−0.525∗
−0.160
−0.348
−0.171
−0.494∗
−0.880∗∗
(0.191)
(0.262)
(0.282)
(0.236)
(0.165)
(0.239)
(0.273)
NF vs NR
β̂
s.e.
0.036
0.108
0.144
0.053
−0.109
−0.370
−0.416
(0.237)
(0.282)
(0.306)
(0.272)
(0.219)
(0.287)
(0.300)
Note: * p < 0.05, ** p < 0.01, a classes 1 & 2 as reference category, b Status level 1,
comprising of occupational categories 1–7, as reference categories, see Chan and
Goldthorpe (2004).
32