Research report:Equality group inequalities in education,employment and earnings: A research review and analysis of trends over time
Yaojun Li, Fiona Devine, Anthony Heath
Universities of Manchester and Oxford
This report analyses the relationship between education, employment,income and social class to identify trends in group-based inequalities relatingto gender, ethnicity, disability and sexual orientation.
留學(xué)生dissertation網(wǎng)What is already know n on this to pic:
• Women and some ethnic minority groups are increasingly likely to obtaingood educational qualifications, jobs and income.
• N evertheless, women on average continue to earn considerably less thanmen, while people from most ethnic minority groups remain less qualifiedand are less likely to secure good jobs than white people.
• There is greater variation among ethnic minority groups than betweenethnic minority groups as a whole and white people.
What this report adds:
• This is the first time that patterns and trends in the educational and work-lifeexperiences of several equality groups are analysed in a single study.
• E ducation protects ethnic minority groups, women and disabled peopleagainst disadvantage in employment and income. However, they do notenjoy the returns to education that might be expected.
• Men from some ethnic minority groups report high rates of job refusals andpromotion blockages, while women from all ethnic minority groups report
unfavourable treatment.
EQUALITY GROUP INEQUALITIES IN
EDUCATION, EMPLOYMENT AND EARNINGS:
A research review and analysis of trends over time
Yaojun Li1, Fiona Devine1, Anthony Heath2
1University of Manchester, 2University of Oxford
© Equality and Human Rights Commission 2008
First published Autumn 2008
ISBN 978 1 84206 080 3
EHRC RESEARCH REPORT SERIES
The Equality and Human Rights Commission (EHRC) Research Report Series
publishes research carried out for the EHRC by commissioned researchers.
The views expressed in this report are those of the authors and do not necessarily
represent the views of the Commission. The Commission is publishing the report as
a contribution to discussion and debate.
Please contact the Research Team for further information about other EHRC
research reports, or visit our website:
Research Team
Equality and Human Rights Commission
Arndale House
Arndale Centre
Manchester
M4 3AQ
Email: [email protected]
Telephone: 0161 829 8500
Website: www.equalityhumanrights.com
You can download a copy of this report as a PDF from our website:
http://www.equalityhumanrights.com/researchreports
CONTENTS Page
TABLES AND FIGURES i
EXECUTIVE SUMMARY iii
1. INTRODUCTION 1
1.1 Aims of the research 1#p#分頁標(biāo)題#e#
1.2 Overview of existing research 2
2. DATA AND METHODS 14
2.1 Data 14
2.2 Methods 15
3. EDUCATION, EMPLOYMENT, INCOME AND CLASS 19
3.1 Education and group-based inequalities 19
3.2 Employment and group-based inequalities 21
3.3 Income and group-based inequalities 22
3.4 Class and group-based inequalities 23
3.5 Summary 25
4. STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS 27
4.1 Logit models of employment 30
4.2 OLS models of earnings (weekly pay) 36
4.3 Predicted values from employment and income models 40
4.4 Predicted values of employment and income by ethnicity and by
education 43
4.5 Predicted values on employment and income by disability and
same-sex 46
4.6 Summary 47
5. JOB REFUSAL AND PROMOTION BLOCKAGE 50
5.1 Descriptive analysis of job refusal / promotion blockage 52
5.2 Statistical modelling on unfair treatment 54
5.3 Predicted values on unfair treatment 57
5.4 Summary 58
6. SUBJECTIVE PERCEPTION OF QUALITY OF LIFE 60
6.1 Descriptive analysis of quality of life 60
6.2 Statistical modelling on quality of life 61
6.3 Predictions of quality of life 63
6.4 Summary 63
7. CONCLUSIONS 65
7.1 Key findings 65
7.2 Future research challenges and policy implications 67
REFERENCES 130
APPENDIX 144
i
TABLES AND FIGURES
Page
TABLES
1a Educational qualifications by ethnicity, disability and same-sex orientationand by gender in 2004/5 72
1b Educational qualifications by ethnicity, disability and same-sex orientationand by gender in 1996/7 73
2a Employment situation by ethnicity, disability and same-sex orientation andby gender in 2004/5 74
2b Employment situation by ethnicity, disability and same-sex orientation andby gender in 1996/7 75
3a Weekly earnings (£) by ethnicity, disability and same-sex orientation andby gender in 2004/5 76
3b Weekly earnings (£) by ethnicity, disability and same-sex orientation andby gender in 1996/7 77
4a Class by ethnicity, disability and same-sex orientation and by gender in2004/5 78
4b Class by ethnicity, disability and same-sex orientation and by gender in1996/7 79
5a Logit regression coefficients on male employment in 1996/7, 2004/5 andcomparison over the decade 80
5b Logit regression coefficients on female employment in 1996/7, 2004/5 andcomparison over the decade 82
6a Regression coefficients on male gross weekly earnings in 1996/7, 2004/5and comparison over the decade, using education as one of predictors 84
6b Regression coefficients on female gross weekly earnings in 1996/7, 2004/5
and comparison over the decade, using education as one of predictors 86
6c Regression coefficients on male gross weekly earnings in 1996/7, 2004/5
and comparison over the decade, using class as one of predictors 88
6d Regression coefficients on female gross weekly earnings in 1996/7, 2004/5#p#分頁標(biāo)題#e#
and comparison over the decade, using class as one of predictors 90
7 Job refusal and promotional blockage by ethnicity, religion, disability and
gender (% answering ‘yes’) 92
ii
8 Logit regression coefficients on job refusal / promotion blockage by sex 93
9 Mean scores on work, social life and life overall by ethnicity, religion,
disability and sex 95
10 OLS regression on satisfaction with work, social life and life overall 96
11 Is there fair employment? 98
12 Has there been progress towards fair employment over the period
1996/7-2004/5? 100
FIGURES
1 Predicted values of male employment 101
2 Predicted values of female employment 103
3 Predicted values of male gross weekly earnings 105
4 Predicted values of female gross weekly earnings 107
5a Predicted values of male employment and gross weekly earnings by
ethnicity and by educational qualifications in 2004/5 109
5b Predicted values of female employment and gross weekly earnings by
ethnicity and by educational qualifications in 2004/5 111
6 Predicted values of gross weekly earnings by ethnicity and by class in
2004/5 113
7 Predicted values of male employment and gross weekly earnings by
disability and by educational qualifications in 2004/5 115
8 Predicted values of female employment and gross weekly earnings by
disability and by educational qualifications in 2004/5 117
9 Predicted values of male employment and gross weekly earnings by
same-sex and by educational qualifications in 2004/5 119
10 Predicted values of female employment and gross weekly earnings by
same-sex status and educational qualifications in 2004/5 121
11 Predicted values of male perception of discrimination (job refusal /
promotion blockage in the last five years) 123
12 Predicted values of female perception of discrimination (job refusal /
promotion blockage in the last five years) 125
13 Predicted values of subjective perception of satisfaction with work, social
life and life overall by ethnicity and education 127
EXECUTIVE SUMMARY
EXECUTIVE SUMMARY
This report presents a review of data on the relationship between education,
employment, income, social class and group-based inequalities relating to gender,
ethnicity, disability and sexual orientation. The aim of this review was to establish
whether higher levels of education, employment, income or socio-economic class
protect against group-based inequalities. In addition to a review of existing research,
the study analysed data from the General Household Survey (GHS 1996/7, 2004/5),
Labour Force Survey (LFS 1996/7, 2004/5), Home Office Citizenship Survey (HOCS
2003, 2005) and the British Household Panel Survey (BHPS 2005). Pooled data are
drawn from the GHS and the LFS to show trends over time. This is the first time that#p#分頁標(biāo)題#e#
gender, ethnicity, disability and same-sex status have been explored together in a
single study, along with the relationships between them. The analysis uses
descriptive and bivariate analysis as well as more complex statistical modelling for
multivariate analysis.
Key findings1
• Education protects against disadvantage in employment and earnings.
However, this is a question of degree: many people from ethnic minoritygroups with higher levels of education, experience poorer employment ratesand lower incomes than White people.
• In 2004/5, Chinese men with middle or higher levels of education had thelowest levels of employment and earnings relative to their education. At the
middle educational level, they were just over half as likely (53 per cent) to beemployed as similarly qualified White men. This rose to just three-quarters (78
per cent) for those at the higher educational level (Figure 5a). Their earningsprofiles were similarly disadvantaged (Figure 5a).
• In 2004/5, Black Caribbean men with higher qualifications were more likely tobe employed than those with lower qualifications. However, even the highlyeducated were still disadvantaged when compared with similarly qualifiedWhite men. At the lower and middle educational levels, they were only 80 and1 In this summary, sources referring to tables are raw data from frequencies or crosstabulations.
Sources referring to figures are predicted probabilities controlling forpeople’s socio-demographic characteristics and household circumstances, derived
from full models. Please note that the sample sizes for people in same-sexrelationships are small.
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
91 per cent as likely to be employed as similarly qualified White men (Figure
5a).
• In 2004/5, the earnings of Pakistani and Bangladeshi men at the low andmiddle levels of education are only two-thirds of those of similarly qualifiedWhite men (64 and 65 per cent respectively) (Figure5a).
• Among Pakistani or Bangladeshi women, it is those who are highly educatedwho find it easier to gain access to employment, higher incomes and a higherclass position. Indeed, higher education tends to protect these women to amuch greater extent than it protects White women or women from other ethnicgroups in a relative sense. For instance, highly qualified Pakistani or
Bangladeshi women were only slightly less likely to be employed than theirWhite peers (83 per cent) whereas poorly qualified Pakistani or Bangladeshiwomen were mostly jobless (18 per cent of White women’s employment rate)(Figure 5b).
• Disabled people with higher educational levels are more likely than otherdisabled people to gain access to employment (twice as likely in the case ofdisabled men) compared to those with low educational levels. However, thedata do not permit us to say whether they were already employed (or whatincome they were earning) before they became disabled.#p#分頁標(biāo)題#e#
• Education protects against lower employment rates and earnings levels onlyto a certain degree, and some disadvantaged groups do not enjoy the returns
to education that might be expected from their investment. This is clearlyseen in the reported rates of job refusals and promotion blockages. At each
level of education (in both 2003 and 2005), Black African men reported two tothree times the incidence of job refusals and promotion blockages, with thenext highest rate being among Black Caribbean men (Table 7). For women,Black Africans at each level of education also reported the highest incidenceof unfair treatment (Figure 12). It is notable that all ethnic minority womenperceived injustice in both survey years and that this perception was growing
or the highly qualified (Figure 12).
• With the exception of those of Indian origin, ethnic minority groups expressed
the least satisfaction with their work life. This was most notable among the
highly educated.
iv
EXECUTIVE SUMMARY
Education
• In 1996/7, men had higher rates of degree-level qualifications than women (21
per cent and 19 per cent respectively) (Table 1b). By 2004/5, the two groups
had the same rate (26 per cent each) (Table 1a). Thus, there was a major
improvement in women's qualifications over the period.
• Black Caribbean men and Pakistani / Bangladeshi men and women were the
least qualified: 16 per cent, 17 per cent and 11 per cent had degree-level
qualifications in 2004/5 (Table 1a). Moreover, the increase in qualifications
gained by Black Caribbean men and Pakistani and Bangladeshi women from
1996/7 to 2004/5, were the least of all ethnic groups (Tables 1a, 1b).
• Although disabled people's educational qualifications improved slightly over
the period, they remain considerably less than those of non-disabled people
(15 per cent of men and 17 per cent of women achieved degree-level
qualifications in 2004/5, compared with 28 per cent of non-disabled men and
women) (Tables 1a, 1b).
• People who reported being in same-sex relationships were more likely than
those in non-same-sex relationships to have degrees (48 per cent of men and
51 per cent of women in 2004/5) (Table 1a).
Employment
• 78 per cent of men and 67 per cent of women were in employment in 2004/5,
compared to 76 per cent and 64 per cent in 1996/7 (Tables 2a, 2b). Holding
other factors constant, the proportion of women in employment increased over
this period (Table 5b).
• Ethnic minority groups had significantly lower rates of paid employment than
White people at both time periods, with the lowest rates among Chinese men
(58 per cent in 2004/5) and Pakistani / Bangladeshi women (23 per cent). If
we take into account the changes in the education and other characteristics of
the groups, there was no real progress over the period (Table 5a).#p#分頁標(biāo)題#e#
• Disabled people were just over half as likely as non-disabled people to be
employed in 2004/5, even though their labour market participation had
improved slightly over the period. The proportionate increase in participation
over the period was higher than for non-disabled people (Tables 2a, 2b, 5a).
v
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
• People who reported being in same-sex relationships were more likely than
people in non-same-sex relationships to be employed (87 per cent of men and
84 per cent of women in 2004/5) (Table 2a). However, once educational
qualifications were taken into account, there was no significant difference from
people who did not report such relationships (Tables 5a, 5b).
Income
• Gender differences in gross weekly earnings reduced over the period, with
women's average earnings increasing from 54 per cent of men's in 1996/7 to
61 per cent in 2004/5. However, given that women's educational levels
increased more than men's, their earnings levels became relatively worse over
the period and they did not see the same returns to their education (Tables 3a,
3b, Figure 5b).
• All men from ethnic minority groups (other than those of Indian origin) earned
significantly less than White men in 2004/05 (Table 3a, Figure 3). The gross
weekly earnings of Pakistani and Bangladeshi men were 64 per cent of the
earnings of White men: this difference was the same as in 1996/7 (Tables 3a,
3b, Figure 3). Among women, Black Caribbean and Indian women earned
significantly more than White women. Pakistani and Bangladeshi women
earned significantly less, at 71 per cent of White women's earnings in 1996/7
and 76 per cent in 2004/5 (Tables 3a, 3b, Figure 4).
• Disabled men's gross weekly earnings reduced slightly over the period, from
83 per cent of non-disabled men's earnings in 1996/7 to 82 per cent in 2004/5
(Figure 3). Disabled women's earnings reduced from 87 per cent to 84 per
cent of non-disabled women's earnings (Figure 3).
• The gross weekly earnings of men in same-sex relationships remained higher
than men in non-same-sex relationships during the period, but the difference
became non-significant when education and other factors are taken into
account. Holding constant all other factors in the models, there was no change
over time. Men in same-sex relationships at higher and lower education levels
earned less than those in non-same-sex relationships. The earnings of women
in same-sex relationships were 1.5 times higher in 2004/05 than those in nonsame-
sex relationships and were also higher when controlling for education
levels (Figures 3, 4, 9, 10).
vi
EXECUTIVE SUMMARY
Social class
• In 2004/5, 40 per cent of men and 37 per cent of women were in professional,#p#分頁標(biāo)題#e#
higher administrative and managerial occupations (the salariat). This
represented an increase of three per cent for men and five per cent for women
over the 1996/7 period (Tables 4a, 4b). The overall gap between men and
women thus reduced over the period.
• In 2004/5, Indian men and women and Black Caribbean women were
significantly more likely than White men and women respectively, to be in the
salariat – this represented an improvement for Indian women over the 1996/7
period (Tables 4a, 4b).
• Pakistani and Bangladeshi men and women and Black Caribbean men were
significantly less likely to be in the salariat in 2004/5; with a relatively slight
improvement for Pakistani and Bangladeshi men and women, and
deterioration for Black Caribbean men over this period (Tables 4a, 4b). When
other factors were held constant, there was no real progress for any ethnic
minority groups – other than people of Indian origin.
• People of Black African and Chinese origin were educationally highly qualified
but this was not effectively translated into occupational success. Men of Indian
origin had a lead of 11 per cent in degree-level qualifications, compared to
White men, but this was only reflected in a 7 per cent lead in access to the
salariat (Tables 1a, 1b, 4a, 4b).
• The proportions of disabled men and women in the salariat reduced slightly
over the period, by 0.7 and 2.0 per cent (Tables 4a, 4b).
• Although men and women in same-sex relationships were more likely than
those in non-same-sex relationships to be in the salariat in 2004/5, this
reflected a decrease over the period for men and an increase for women. The
data also showed a decrease for men in comparison with other men (down
from 81 per cent higher to 48 per cent higher), while women in same-sex
relationships retained their position in relation to other women (62 per cent
higher) (Tables 4a, 4b).
Conclusion
• Even though some signs of progress are visible, the data show continuing
inequalities in relation to employment rates, earnings, job-seeking and
vii
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
viii
treatment at work. There is also evidence of labour market barriers, possibly
including discrimination and prejudice, and of some groups feeling they
experience difficulties more than others.
INTRODUCTION
1. INTRODUCTION
1.1 Aims of the research
The central aim of this research was to examine the relationship between education,
employment, income (labour market earnings), social class and group-based
inequalities relating to gender, ethnicity, disability and sexual orientation.
The report begins with an overview of existing research. It then draws together key
findings from new analysis carried out for this review, which:#p#分頁標(biāo)題#e#
1. Collects evidence as to whether group-based inequalities are greater for
people with lower educational attainment, lower levels of employment, lower
incomes and lower socio-economic positions
2. Establishes whether higher education, higher levels of employment, higher
income and higher socio-economic class protect against the worse impact of
group-based inequalities
3. Clarifies the nature and extent of any relationships between inequalities.
In seeking to address these questions, the key objective was to establish whether
there are interactions between various group-based inequalities and education,
employment, income and class. Of course, the extent to which: low education, low
levels of employment, low income and low class positions expose people to the worst
aspects of group-based inequalities; while high education, high levels of employment,
high income and high class positions protect against them, are two sides of the same
coin.
The report uses data from a variety of sources including the General Household
Survey (GHS 1996, 2004, 2005), the Labour Force Survey (LFS 1996, 1997, 2004,
2005), the Home Office Citizenship Survey (HOCS 2003, 2005), and the British
Household Panel Survey (BHPS 2005) to examine the relationship between
education, employment, income, social class and group-based inequalities relating to
gender, ethnicity, disability, and sexual orientation – wherever data are available.
1
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
1.2 Overview of existing research
This section provides a brief overview of existing research on the relationship
between equality groups and education, employment, income and social class. It
draws on both quantitative and qualitative research to ascertain, as far as possible,
some of the underlying causes of the associations and interactions between them,
and the persistence of group-based inequalities over time.
Gender
White boys and girls have remarkably similar rates of educational success as a result
of the considerable improvement in girls’ performance over the last twenty years
(Arnot & Mac an Ghaill, 2006). They have similar rates of participation in Further
Education (FE) and Higher Education (HE) and young women are now as likely to
have high-level qualifications (degrees) as young men (Elias et al. 2000). However,
despite these changes, subject choices for the General Certificate of Secondary
Education (GCSE) remain heavily gendered with young men studying the sciences
and young women specialising in the arts. This pattern can be seen in the choice of
A levels and degrees. New longitudinal research on graduates in the labour market is
showing that subject choices have major implications for employment trajectories,
income levels and class position (Purcell & Elias 2004, 2005). Young women have#p#分頁標(biāo)題#e#
higher educational credentials than in the past, although the returns to education will
not be as high as those for young men. This trend will require further monitoring.
Educational sociologists have cast a skeptical eye on public debate about boys’
underachievement. At the very least, attention should focus specifically on White
working-class boys' underachievement (Epstein, 1998). Of course, why ‘workingclass
kids get working-class jobs’ (Willis, 1977) is a very old question still in need of
an answer. Working-class boys (and girls) tend to go to under-performing schools in
their local areas, which contribute to low levels of attainment (Gewitz et al. 1995). In
addition, recent qualitative research (Evans, 2006; McDowell, 2003) has emphasised
the clash of cultures in the home and the school as White boys (seeking to defend
their masculinity and pride) do not value learning in school. Instead, they are eager to
leave the education system as early as possible. Thus, White working-class boys do
not get sufficient education to allow them to enjoy returns in terms of employment,
income and class. The same is true of working-class girls. Class inequalities in FE
2
INTRODUCTION
and HE also remain stark (Bynner et al. 1997; Devine, 2004; Ferri, 2003; Machin &
Vignoles, 2004; McKnight, 2005; Power et al. 2003).
A vast amount of literature has charted the rise of women’s employment over the last
forty and more years. Indeed, employment rates are inching closer to men’s
employment rates all the time. However, significant differences remain. For example,
it is still women rather than men who take time out of paid work when children arrive
and who then suffer the ‘parenthood penalty’ on their return to the labour market
(Scott et al. 2008). Highly-educated women are more likely to return to full-time paid
work earlier (often to the same employer) and do not suffer the parenthood penalty
as much as less-educated women – because they are behaving more like men.
Thus, it is these particular women who enjoy the returns to education, and education
seems to protect them against the disadvantages of being a mother. Of course, work
/ family balance issues (Dex & Smith, 2002; Dex, 2003) arise, and then go some way
to explaining why women do not want or do not enjoy later career progression into
top jobs (Scott et al. 2008). More research is required on this issue.
Less-educated women suffer the parenthood penalty, as a result of having longer
gaps before returning to the labour market, and returning to paid work on a part-time
basis (often with a different employer) in a narrow range of occupations in the service
sector (Scott et al. 2008). Lack of childcare options is still a considerable barrier to
women with children under 11, especially lone mothers with young children who#p#分頁標(biāo)題#e#
return to work earlier or work full-time. Employers remain wary of employing mothers
and may discriminate against them. On returning to employment, these women
experience downward mobility and the evidence suggests this penalty is actually
growing. Therefore, part-time employment is a trap where training opportunities are
limited, career progression is almost non-existent and, in effect, there are only a few
bridges to facilitate upward occupational mobility (Tam, 1997; Warren 2000, 2004).
This growing divide between highly educated and less educated women needs
further investigation.
The persistent gender pay gap suggests that women still do not enjoy the same
returns to education as men, in terms of income. Even highly-educated women
graduates do not enjoy the same rates of pay on entering the labour market and
indeed, the gender pay gap grows over time (Purcell & Elias 2004, 2005). For single
3
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
women, the parenthood penalty does not apply. Rather, it is the concentration of
highly-educated women in the public sector that is the source of the problem – nearly
half of young women graduates are found there (where pay is 10 per cent less than
in the private sector) compared to a third of men. In part, this is the result of women’s
preferences for: socially useful work in the caring professions; the desire to work in
organisations with family friendly policies; and in jobs that provide a good work-life
balance in anticipation of combining work and family in the future. In contrast, young
men graduates are geared towards highly-paid occupations and are willing to live
with the long hours work culture – which becomes the norm in such work
environments (McDowell, 1997).
While interesting changes are happening at the top-end among highly educated
women, only a quarter of women workers are graduates and the pay of nongraduates
is poorer relative to men and in need of further investigation (Rubery et al.
1997). While the Minimum Wage and Family Tax Credits have greatly helped lowearning
women, the pay gap still remains. Part-time employment is a major part of
the issue although persistent gender segregation in the labour market has to be
considered too (Bradley et al. 2000; Walby, 1997). Men work in jobs that are better
paid and many of these jobs – for example, engineering (Glover, 2000) – still bar
women (indirectly rather than directly) in terms of how they view and treat women in
everyday working practices. Women are concentrated into a narrow range of care
related occupations for example, often care assistant work in the public sector, or
retail / hairdressing work in the private sector. Why these jobs continue to be
undervalued in terms of pay needs to be explored at all levels (Joshi & Pac, 2001).#p#分頁標(biāo)題#e#
While early research on class tended to exclude women (Crompton, 1980), this
oversight has long been rectified. Research in the 1980s for example, showed that
while men dominated the top of the class structure, women dominated the middle
and bottom (Marshall et al. 1988). This was the result of gender segregation (Hakim,
1979) and the sex typing of jobs (Bradley 1989, 1996) in the labour market, which
confined women to routine non-manual jobs or manual employment. These findings
implied that the daughters of middle-class fathers often experienced downward
mobility into, for example, clerical or secretarial work or, at best, semi-professional
employment in the supposedly gender appropriate jobs of teaching (Machin &
4
INTRODUCTION
Vignoles, 2005) and nursing (Davies, 1995) and that working-class women did not
enjoy long-range mobility into professional and managerial jobs like men.
The relatively recent entry of women into the professions and management suggests
that women’s downward mobility has declined. Crompton (Crompton 1980, 1999,
2006; Crompton & Sanderson, 1990; Crompton & Harris, 1998) argues that women
have pulled the ‘qualifications lever’ which has allowed them to enter professions
such as medicine, law, accountancy as well as graduate level jobs in management
(Bolton & Muzio, 2007; Halford et al. 1997; Witz, 1992; Witz & Savage, 1992). That
said, it is predominately women of middle-class origins who are now retaining middleclass
positions rather than experiencing downward mobility into intermediate
positions. Working-class girls now have similar prospects as working-class boys for
upward mobility (Goldthorpe & Jackson, 2007) although when compared with middleclass
girls, they are the ones who continue to fill low-level gendered jobs, such as in
childcare (Gregson & Lowe, 1994; Skeggs, 1997).
Ethnicity
The ‘ethnic minority drive for qualifications’ (Modood et al. 1997) continues as levels
of educational achievement have increased for some ethnic minority groups –
notably Indians and Chinese (Dustmann & Theodoropoulos, 2006; Heath &
McMahon, 1999). Moreover, the educational achievements and aspirations of young
Black Caribbean, Pakistani and Bangladeshi women are improving too (Bhavani,
2006; Equal Opportunities Commission, 2006; Dale et al. 2002). Bagguley and
Hussain (2007) (see also Hussain & Bagguley, 2007) have looked at the increasing
number of South Asian women going to university. They found that parents play a
very major role – arguably stronger than in White families – in deciding subject
choices and choice of local university. Universities close to home are often preferred,
especially as ethnic minority women still experience prejudice and discrimination as#p#分頁標(biāo)題#e#
part of university life. Thus, expectations about HE, employment, marriage and
children are changing although continuities remain.
However, the position of some ethnic minority groups with regards to education is not
improving (Heath & Brinbaum, 2007). Black Caribbean, Mixed White / Black
Caribbean, Black African and Pakistani / Bangladeshi boys (in particular) are not
doing well especially in secondary school. For example, they are more likely to be
5
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
excluded from school for ‘behavioural issues’ and less likely to gain five GCSEs or
more at age 16. If they go to university, they are often concentrated in the lower
status post 1992 universities (Connor et al. 1996). Why the ‘visible’ minorities are not
succeeding is puzzling (Cheung & Heath, 2007). Class is an interrelated issue,
although discrimination and prejudice in school settings is considered part of the
problem (Gilborn, 2008; Mac an Ghaill, 1999). Surprisingly, there are no recent
ethnographies of underachieving ethnic minority groups which might update Mac an
Ghaill’s classic study (1988) – although Haynes’ (2008) study on Black Caribbean’s
and Aim Higher is noteworthy.
While employment rates among all men are high, it has long been known that ethnic
minority men have lower rates of employment than White men (Cheung & Heath,
2007; Li & Heath, 2008b). In particular, Pakistani and Bangladeshi men have lower
rates of economic activity, while Black Caribbean and Black African men have higher
rates of unemployment (Li & Heath, 2007; Heath & Li 2007, 2008, forthcoming). The
difficult position of these men has been well captured by in-depth qualitative work
such as Kalra’s (2000) study of Pakistani men in Oldham who experienced
redundancy and unemployment as the textile industry collapsed in the 1980s and
employment opportunities were limited to jobs such as taxi-driving. Mac an Ghaill and
Haywood’s (2005) research on Bangladeshi young men and women in Newcastle,
paints a similar picture of exclusion. In effect, many ethnic minority men are excluded
from the labour market, which implies they do not enjoy the returns to education in
terms of economic activity. Their continued exclusion (Radcliffe, 2004; Solomos,
2003) requires ongoing research.
Employment activity among ethnic minority women is similar. White women have the
highest levels of employment and ethnic minority women have lower levels, although
there are variations between them. Black Caribbean women have high rates of
employment as they are concentrated in nursing in the NHS – to which they were
directly recruited since the 1950s (Mason, 1995). Again, the evidence shows that it is
Pakistani and Bangladeshi women who have the lowest rates of employment and#p#分頁標(biāo)題#e#
highest levels of inactivity. Of course, many of these women are mothers at home,
not least because they have more children and cultural traditions place a high value
on motherhood. Even so, there are barriers to employment due to a lack of fluency in
the English language (Modood et al. 1997) and discrimination (Bradley, 2007). Thus,
6
INTRODUCTION
ethnic minority women have very similar patterns of employment as White women in
the 1950s. However, young women are now beginning to acquire educational
credentials and this increases their chances of employment (Lindley, Dale & Dex,
2006; Dale, 2005).
White men earn more than ethnic minority men although there are differences
between ethnic minority men: Indian and Chinese men enjoy higher earnings than
Pakistani / Bangladeshi and Black Caribbean men. That said, where ethnic minority
men (notably of the second generation) hold similar qualifications and class position
to White men, they earn similar amounts (Cheung & Heath, 2007). Once ethnic
minority men are employed in the labour market, they experience the same
processes of stratification as White men – although, prejudice and discrimination are
still found among employers (and employees). Highly-qualified ethnic minority men
however, are treated the same in the recruitment process into high-level professional
and managerial positions (Heath & Yu, 2004; Hoque & Noon, 1999). Educational
success increases the probability of employment and occupational success for ethnic
minority men as it does for White men.
A very similar story can be told in relation to ethnic minority women and income. As
previously mentioned, the picture is a little more complicated as Black Caribbean
women have secured relatively good incomes through nursing careers in the NHS.
With a greater tendency to head-up single-parent households than White women,
they are more likely to work full-time than part-time and have experienced less of the
penalty associated with motherhood (Dex, 2003). Be that as it may, ethnic minority
women (especially of the second generation) earn similar amounts to White women
where they hold similar qualifications and class position (Cheung & Heath, 2007).
Like White women, highly-qualified ethnic minority women have pulled the
‘qualifications lever’ (Crompton & Sanderson, 1990) which has facilitated entry into
high-level occupations in the professions – medicine, law and accountancy – which
command high and rising salaries. Therefore, education makes a difference to
employment income.
With regards to the class position of ethnic minority men, White men have a
substantial presence in the salariat (namely high-level professional and managerial
jobs), but so too do Indian men. Pakistani / Bangladeshi and Chinese men are over-#p#分頁標(biāo)題#e#
7
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
represented among the self-employed petty bourgeoisie – often in restaurants and
take-aways – as any stroll along a British high street will confirm (Phillips, 1995; Ram
et al. 2005; Li, 2007b). Once more, Black Caribbean men, along with Pakistani and
Bangladeshi men, are clustered into working-class jobs in skilled, but more usually,
semi and unskilled manual work. Why second-generation Black Caribbean men in
particular, are still concentrated at the bottom of the class structure, requires further
research (Cheung & Heath, 2007). It is evident that similarities and differences
between White and ethnic minority men are reproduced across the spheres of
education, employment, income and class.
A similar story can be told with regard to white and ethnic minority women (Cheung &
Heath, 2007; Heath & Yu, 2004). The evidence to date shows that White women are
in the salariat, although so too are Black Caribbean women (given their full-time NHS
nursing careers) and Indian women (whose educational performance has increased
in recent decades and facilitated their entry into professional occupations). Once
more, Pakistani and Bangladeshi women who are employed have a limited presence
in middle-class positions and, because of their low-level qualifications, dominate in
working-class skilled and unskilled manual jobs (Bradley et al. 2007; Dale, 2005).
Thus, the low class position of Pakistani and Bangladeshi women remains stark and
is related to their low levels of educational credentials (and levels of employment). As
education levels improve (and so too does employment as noted above (Lindley,
Dale & Dex, 2006; Dale 2005)), the class position of Pakistani and Bangladeshi
women can be expected to improve over time – although whether it will be over too
long a time is a moot point.
Disability
It is well known that disabled people have much lower levels of educational
qualifications than non-disabled people, as a result of their (past) exclusion from the
education system in Britain (Barnes et al. 2002; Beckett, 2006). The education of
disabled young people has now moved away from segregated educational
institutions, which offered only a limited curriculum and promoted low expectations
among disabled pupils. Even so, their inclusion in mainstream education has not yet
been fully achieved: nursery provision for disabled children is poor; they do not do as
well as non-disabled children in Key Stage tests; they are less likely to be involved in
education, training and employment at 16; or participate in FE and HE. While
8
INTRODUCTION
progress has been made, the inclusion of disabled pupils into mainstream education
has not been straightforward (French & Swain, 2004; Swain et al. 2004).#p#分頁標(biāo)題#e#
Much of the research agenda is still devoted to understanding and explaining how
barriers limit the chances of disabled young people acquiring educational
qualifications. It appears that disabled young people’s inclusion in mainstream
schools has often been done with limited funding or inadequate support. Teachers’
expectations and aspirations in respect of disabled children’s educational capabilities
and potential remain limited and limiting. Beckett’s (2006) Economic and Social
Research Council (ESRC) funded work on disability equality in English primary
schools and the difficulties of getting access to scarce resources, is a case in point.
Disability rights activists see education as key to: opening up employment
opportunities; the chance to live independently on a reasonable income; and enjoying
a good quality of life. It is a major influence on life-chances (Priestley 2001, 2003).
Differences in educational performance among disabled people, in terms of gender,
race and sexuality, remain unknown and under-researched.
The low level of employment among disabled people is well known in disability
studies. Poor educational qualifications are a factor, although disabled people are 30
per cent more likely to be out of work than non-disabled people with the same
qualification (Mercer, 2005; Roulstone, 1998; Roulstone et al. 2003; Roulstone &
Barnes, 2005). Thus, disabled people do not get the same returns on educational
credentials in the labour market as non-disabled people. Government legislation has
sought to outlaw discrimination and improve the employment opportunities of
disabled people. Nevertheless, one of the major foci has been the continued
difficulties of securing employment. Even in employment, most disabled people are in
(public sector) lower-level jobs with low incomes, and high flyers who command high
incomes are a minority (Shah, 2005). This reality in turn, affects the occupational
aspirations and choices of disabled young people (Shah, 2008 forthcoming).
Research suggests that Government commitment to removing barriers does not
easily translate into employers paying for training, specialist advice, and making
reasonable adjustments to the workplace. Discrimination, prejudice, fear and
misapprehensions also make the workplace an uncomfortable environment in which
to work (Woodhams & Danieli, 2000). Accordingly, research continues on the barriers
9
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
to employment itself and to good employment, including the reasons why employers
(notably those in the private sector) are not prepared to make additional efforts to
accommodate the special needs of disabled workers. This research also considers
what further incentives may be required to facilitate the entry of disabled people into#p#分頁標(biāo)題#e#
employment so that employment levels increase. Differences in employment rates
among disabled people, in terms of gender, race and sexuality, and differential
returns to education for different groups, have yet to be established.
Given that disabled people often work in low-level jobs, frequently work part-time or
move between jobs and in-and-out of the labour market, their income levels are low.
Again, the evidence suggests that disabled people have lower incomes than nondisabled
people with the same qualification (Mercer, 2005; Roulstone, 1998;
Roulstone et al. 2003; Roulstone & Barnes, 2005). They do not get the returns to
education in terms of income, in the same way as non-disabled people. Moreover,
the exclusion of many disabled people from the labour market has meant they are
forced to live on welfare benefits which constitute a very low income. Therefore,
poverty remains a major issue (Townsend, 1979; Barnes & Mercer, 2002; Swain et
al. 2004). Differential poverty rates among disabled people, in terms of gender, race
and sexuality, and the extent to which education might protect different groups from
poverty, also have yet to be researched.
There has been almost no research on disability and class to date. Certainly within
class analysis, there has been neither an examination of the position of disabled
people in the class structure, nor an exploration of patterns and trends in social
mobility. Even if there had been, the low levels of employment among disabled
people would have led to their exclusion from statistical analysis because occupation
is invariably used as a proxy indicator of class. Other economically inactive groups,
like mothers at home or the unemployed, have usually been included in class
analysis on the basis of details of their previous employment (Dex, 1987; Gallie et al.
1998). This solution would not be possible for many disabled people with no
employment histories. Class position would have to be established via class
background, namely parent’s class, which is not very satisfactory in ascertaining the
current class position of disabled people.
10
INTRODUCTION
What we have learned from disability studies, is that disabled people who are
employed are likely to cluster in low level working-class jobs or, at best, intermediate
positions. Limited employment opportunities have confined disabled people to ‘jobs’
rather than ‘careers’, excluding them from high-level professional and managerial
middle-class positions. The possibilities of upward social mobility have been limited.
Shah’s (2005) recent work on the career success of disabled high-flyers has led her
to call for more research on the way in which class origins influence educational
success and occupational destinations among disabled people. If class origins are so#p#分頁標(biāo)題#e#
crucial for the life-chances of non-disabled people, their importance should be
considered for disabled people and how their adult lives unfold. Again, differences
among disabled people in terms of gender, race and sexuality, need further
exploration too.
Sexuality
There is no research which has directly considered differential rates of educational
success by sexuality. For the most part, work on sexual orientation and education
has focused on the seemingly growing problem of homophobia – forms of bullying
and abuse – which appears to be almost endemic in schools (Hunt & Jensen, 2007).
It has been argued that schools, where the ‘naturalness of heterosexuality’ is
dominant, create a hostile atmosphere for young people to understand their
emerging sexualities. This is especially true for gay men and lesbians who want to
express their homosexuality, when homophobic insults are banded about by some
young people. Moreover, some teachers and youth workers hold prejudicial attitudes
and this shapes discriminatory behaviour towards young gay men and lesbians.
Sexuality in this respect is strictly governed, and homosexuality is frequently
oppressed in the school system and through education policies (Epstein, 1994;
Epstein et al. 2003).
This research has focused on the processes by which sexualities are ‘manufactured’
in schools, colleges and universities (Richardson, 2000; Richardson & Seidman,
2002; Weeks, 2001). To repeat, we do not know whether discrimination and
prejudice affect educational performance and consequent outcomes. There are no
data on differentials in educational qualifications among young people according to
their sexual orientation. It could be surmised for example, that gay and lesbian young
people experience school as a hostile environment in which they might under-
11
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
perform. Yet, it may be that young gay men and lesbian women have survival
strategies which allow them to prosper in the education system despite the difficult
environment. More research is required on sexual orientation and educational
outcomes and any variations among gay men and lesbian women by class, gender,
ethnicity and disability.
Similarly, there is no definitive map of the position of gay men and lesbian women in
the labour market and the extent to which they enjoy appropriate returns to
education. In the sociology of work and employment, attention has focused on
instances of homophobic abuse, harassment and violence. The spotlight has been
on the suppression of sexualities in the workplace, which are often dominated by
heterosexual men. Echoing academic thinking in the field of education, research has
focused on the nature of heterosexualised cultures at work and how the workplace is#p#分頁標(biāo)題#e#
a site where the construction of masculinity, femininity and heterosexuality and
homosexuality takes place (Richardson, 1996; Wajcman, 1999). Sexuality is central
to the way in which work organisations operate (Adkins, 1994; Wolkowitz, 2006).
Links are often made between sexism and racism in work organisations (Edwards &
Wajcman, 2005; Hearn & Parkin, 2001).
The evidence suggests that the suppression of homosexuality in the face of
discrimination and hostility in the workplace thwarts career progression. Ward and
Winstanley’s (2006) research into sexual minorities working as fire fighters in
London, found that homophobia was prevalent and those who came out or were
‘outed’ in the workplace were often shunned by colleagues and worse still,
sometimes lost their jobs. This is not to say that all workplaces and heterosexual
workers are hostile to gay men and lesbian women and indeed, there are industries
like the media where gay men for example, are ‘a significant part of the employee
base’ (Ward & Winstanley, 2006). They may have all-important ‘strategies for
organisational survival’ (Thompson & McHugh, 2001). More needs to be known
about: patterns of employment among men and women in same-sex and non-samesex
couples; the returns to education; and any differences by class, gender, race and
disability.
In the UK, academic research on sexuality and income is very thin on the ground. For
example, there is no research that compares patterns and trends in income between
men and women in same-sex and non-same-sex couples. Consequently, we also
12
INTRODUCTION
13
know very little about the returns to education and any income differences between
men in same-sex couples, women in same-sex couples and variations in terms of
ethnicity and disability. There are US websites which report on the gay market, such
as the pink pound and consumer surveys (www.communitymarkinginc.com,
www.edgeboston.com). They indicate that gay men have higher incomes than
heterosexual men. They also have higher incomes than lesbian women who in turn,
have higher incomes than heterosexual women because they do not suffer from the
penalties of motherhood. These patterns have yet to be established in the UK,
although it is highly likely that such income patterns are similar.
Finally, there is little research on class and sexual orientation in terms of establishing
basic details about the position of men and women in same-sex couples in the class
structure. Nor has there been any research on patterns and trends in social mobility
for gay men and lesbians in conventional class analysis. There is a quite different
body of work, looking at cultural representations of class and the relationship of
sexuality and class (Healy, 1996; Moran & Skeggs, 2003; Munt, 2000; Skeggs,#p#分頁標(biāo)題#e#
2004). Issues of interest here include the ‘homosexual eroticization of class’ such as
the way in which the working-class ‘chav’ label has been used to sell sexual products
and services (Johnson 2006, 2008). These issues aside, it is readily apparent that
there is a paucity of quantitative data and qualitative material on: sexuality and class;
the returns to education; and differences by gender of same-sex couples, ethnicity
and disability.
Summary
Overall, this brief review has shown that a good deal of research has been done on
gender and ethnicity in relation to education, employment, income and class
(although more research could still be done of course). However, much less has
been done on disability and sexuality with regards to these issues. Within-group
differences across this range of equality groups have not been so extensively
researched and most importantly, the extent to which education protects groups from
disadvantages, has not been examined. Also, much of the research on equality
groups has been done separately. None of the existing quantitative research has
looked at gender, ethnicity, disability and sexuality and differences between them in
relation to education, employment, income, class, discrimination and life satisfaction,
simultaneously. It is these issues which are the focus of the present study.
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
2. DATA AND METHODS
2.1 Data
The data used in this report are drawn from the most authoritative Government and
academic surveys: the General Household Survey (GHS); the Labour Force Survey
(LFS); the Home Office Citizenship Survey (HOCS); and the British Household Panel
Survey (BHPS). We used: the pooled GHS2 and LFS for 1996/7 as the earlier period
and contrast it with 2004/5 as the later period; the HOCS 2003 and 2005; and the
BHPS 2005. Throughout the analysis, we focused on men aged 16-64 and women
aged 16-63, resident in Great Britain at the time of interview – except for the HOCS
data which are restricted to England and Wales only.
14
In this research wherever data are available, we used gender, ethnicity, religion,
disability (including people with limiting long-term illness) and same-sex status to
identify potentially disadvantaged groups. The last group is particularly hard to find in
quantitative analysis as the data are either not collected or only exist in very small
numbers. However, we managed to find sufficient numbers for statistical analysis.3
We appreciate that many people with same-sex orientations (couples or otherwise)
may prefer not to declare their sexual orientation to an interviewer, thus leading to an
underestimation of the true extent of the number of gay men and lesbian women and
discrimination suffered by them at the societal level.4 It is also noted here that owing#p#分頁標(biāo)題#e#
to the ambiguity of definition between disability and limiting long-term illness in the
GHS / LFS files, we cannot precisely differentiate between disability and limiting longterm
illness. Thus, we code all incidences of disability and / or limiting long-term
illness as the same attribute and simply use the term ‘disability’ in the following
discussions. We understand that the Office for National Statistics (ONS) is planning
to collect more data on sexuality in the coming years, but the data are unlikely to be
2 The GHS did not collect data in 1997, hence only the 1996 data are available.
3 The pooled the GHS / LFS from 1996/7 to 2004/5 has 1,680 respondents who are
in same-sex couples (within the age-geography limits imposed), which is nearly 4.5
times as many as available in the Household Samples of Anonymised Records
(SAR) from the 2001 Census of Population www.ccsr.ac.uk/sars/2001/hholdcams/
codebook/camrelations.pdf Yet as the data for the intervening years (1998-
2003) are not used in this report, the numbers are smaller.
4 It is possible that some of the same-sex people are not in couples, as ‘couple’ in
that sense is hard to define. Given the small sample sizes involved and the lack of
clear definition in the dataset, we cannot further differentiate same-sex people as
individuals or as couples.
DATA AND METHODS
available in the near future. Therefore, in spite of shortcomings with sample sizes for
same-sex couples and the ambiguity concerning disability, our data are currently the
best available for the research in question. Moreover, our data have the added
advantage of having information on income (earnings from the labour market) and
many other socio-economic variables that adequately meet the research needs of
this review.
2.2 Methods
We focus on the protective role of education: that is, the degree to which education
protects disadvantaged groups in their labour market position (participation and
earnings) and other aspects of socio-economic life such as discrimination in the
labour market and satisfaction with work and non-work life. For analysis on
employment and earnings, we also explore the protective role of class. The analytical
framework is shown in Diagram 1.
15
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
Diagram 1 Analytical framework for the report
Descriptive analysis: for men and women separately
Main variables of interest
Ethnicity
White (ref)abc
Black Caribbeanabc
Black Africanabc
Pakistani / Bangladeshiabc
Indianabc
Chineseabc
Otherabc
Religion (especially Muslim)b
Disabled / long-term illnessabc
Same-sex couplesab
Educationabc
Employmentabc
Incomeabc
Classabc
Life experience
Refused a job / denied#p#分頁標(biāo)題#e#
promotion in last five yearsb
Dissatisfied with workc
Dissatisfied with social lifec
Dissatisfied with life overallc
Statistical modelling
Main variables of interest
Ethnicity
White (ref)abc
Black Caribbeanabc
Black Africanabc
Pakistani / Bangladeshiabc
Indianabc
Chineseabc
Otherabc
Religion (especially Muslim)b
Disabled / long-term illnessabc
Same-sex couplesab
Control variablesd
Sex, age, education, class,
marital status, youngest
dependent child aged 0-5,
number of dependent
children under 16, country of
residence
Interactionsd
Employmentabc
Earnings from labour
marketabc
Life experience
Discrimination
Refused a jobb
Denied promotion in last five
yearsb
Life satisfaction
Satisfaction with workc
Satisfaction with social lifec
Notes:
a Available in GHS / LFS (1996/7-2004/5)
b Available in HOCS 2003 and 2005.
c Available in BHPS (2005).
d Most of the analysis will be conducted for men and women separately; other control
variables such as age, marital status, etc. will be included in the modelling as
appropriate; education and dependent children will be used in interactions in
employment and earnings; and education, class and dependent children in earnings.
16
DATA AND METHODS
We shall first analyse the situation of education, employment, income (that is,
earnings from the labour market, hereafter used interchangeably in this report) and
class to see the patterns and trends of disadvantage faced by the key groups in the
GHS / LFS. We shall then look at life experiences in terms of discrimination in the
labour market (job refusal and promotion blockage) – using the pooled data from the
HOCS for 2003 and 2005 – and of subjective perception of quality of life (satisfaction
with work, with social life and with life overall) using the BHPS for 2005. Moreover,
and key to the project, we shall conduct a series of analyses testing interaction
effects between education and ethnicity, education and disability, and education and
same-sex, to see whether, and to what extent, education protects these groups from
discrimination and disadvantage. As noted earlier, we shall also analyse the
protective role of class in employment status and earnings, with class defined as
current or last main employment based on the Goldthorpe class schema (Goldthorpe,
1987).
We code ethnicity using the categories in the 1991 Census: White, Black Caribbean,
Black African, Indian, Pakistani / Bangladeshi, Chinese and Other. People of
Pakistani / Bangladeshi origins are coded together because of the relatively small
sample sizes of their respective groups and the largely similar socio-economic
disadvantages shared by the two groups (NEP, 2006) (for a discussion of the#p#分頁標(biāo)題#e#
differences in socio-political participation between the two groups, see Li & Marsh,
2008; Li, 2008). The ‘Other’ group also includes various ‘mixed’ groupings. To avoid
repetition, we shall simply refer to them as ‘Other’ rather than ‘Other / Mixed’ in the
following discussion. With regard to education, we code it as a three-way variable:
lower level (primary or no education), intermediate level (O-A Levels or equivalent),
and higher level (first degree or above, or equivalent), which can be used both as
categorical and continuous variables. National vocational qualifications (NVQs) are
included in the appropriate levels. The coding of class will be explained in Chapter 3.
In the descriptive analysis we present bivariate tables, crosstabulating gender,
ethnicity, disability and same-sex on the one hand; and by education, employment,
income and class on the other. Significance levels are presented only for key
categories in the variables, such as degree-level qualifications or access to the
salariat. In the modelling, we only use employment and income as outcome
variables, with education and other socio-demographic factors as explanatory
17
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
18
variables. In modelling income, we also use class together with other independent
variables as predictors5 and similarly for the ‘life experience’ research. It is noted
here that income (labour market earnings) does not include benefits or transfers, and
as a large proportion of respondents reported earnings but not hours of work, or
hours of work but not earnings, we did not include hours of work as an explanatory
variable in our models. To do that would have much reduced our sample sizes.6 It is
further noted here that the GHS / LFS do not ask for income data for the selfemployed;
however, the employment sectors of some respondents are not clear-cut
– some respondents may be nominally self-employed but also work for other
companies, or for their own companies as employees. There is thus some inaccuracy
in this regard. However, the overall proportion of such respondents is very small and
we do not need to be overly concerned about this. Finally, we did not consider the
impact of occupational segregation on income, which may be of considerable
significance for some ethnic groups. For a recent study on occupational segregation,
see Elliot and Lindley (2008).
5 We also carried out analyses using class as a predictor of employment status.
However, the estimates were not clear. This is because employment status (such as
unemployment or inactivity) may be better viewed as a separate state in class
analysis, rather than class having a causal relationship to employment (other than in
a prospective panel design). Using class as a predictor of income on the other hand,#p#分頁標(biāo)題#e#
is not problematic and the results are presented in this report.
6 We repeated all analyses using both weekly and hourly pay, and the results show
the same patterns. This is because we have controlled for marital status, number of
dependent children under the age of 16 and presence of children under the age of
five, in addition to a range of other variables. The results using hourly pay are not
separately presented but are available on request.
EDUCATION, EMPLOYMENT, INCOME AND CLASS
3. EDUCATION, EMPLOYMENT, INCOME AND CLASS
In this chapter we present descriptive findings on education, employment status,
income (gross weekly pay) and class, based on men aged 16-64 and women aged
16-63 resident in Great Britain at the time of interview using the GHS / LFS as earlier
described. We present the 2004/5 data first, followed by the 1996/7 data. The
discussion of the earlier data is mainly for comparison with the later period. The
analysis is followed by statistical models on employment status and income
controlling for a range of socio-cultural and demographic-geographic factors. The
modelling results and the predicted values from the models will be reported in the
next chapter.
The data in Tables 1a-4a are on education, employment, income and class for men
and women respectively in 2004/5, and the corresponding Tables 1b-4b show the
data in 1996/7. The data on education, employment status and class are
percentages and those on income are gross weekly pay in pounds. Apart from the
descriptive data, we also present results of bivariate statistical tests in the tables for
each of the other categories in a variable against the reference group, such as White,
non-disabled, and non-same-sex (that is, people not in same-sex couples or not
having same-sex orientations).7 As the analysis is conducted for men and for women
separately, we can also see the differences between men and women.
3.1 Education and group-based inequalities
Tables 1a and 1b show the patterns and trends in educational attainment by gender,
ethnicity, disability and same-sex couples in the later and the earlier period. In terms
of gender differences, the last row in the two tables shows signs of progress. In the
earlier period, men had somewhat higher rates of degree-level qualifications than
women (21 and 19 per cent respectively). In 2004/5, the two gender groups had the
same rate (26 per cent each). Thus, there has been a major improvement in
women’s level of educational qualifications in the last 10 years.
19
7 In this way, we can not only see the extent of differences between the different
groups but also whether the differences in question are statistically significant. One
could have done this for each category of the dependent variable but we have only
done so for some categories of particular interest such as being employed, access to#p#分頁標(biāo)題#e#
the professional / managerial (salariat) class, or having degree-level (or above)
educational qualifications.
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
In terms of ethnic differences in education, we find greater differences among ethnic
groups than between them and the majority group. At both time periods, it was the
Black African, Indian and Chinese men and Chinese women who had the higher
educational qualifications (Black Caribbean and Black African women also had
substantially higher rates of degree-level qualifications although some of the rates
were not significantly higher than their White peers). On the other hand, Black
Caribbean men, and Pakistani / Bangladeshi men and women were consistently
found to be least qualified.
The period covered saw a big increase in educational provision and consequently, a
substantial overall increase in the proportion of people with degree-level
qualifications – an increase of around 5 per cent for men and 7.5 per cent for women.
If we compare the figures for the ethnic groups in terms of degree-level education,
we find that Black Caribbean men and women, and Pakistani / Bangladeshi women
had less than their expected share, whereas all other ethnic minority groups had
more than their expected share – Indian, Black African and Chinese men’s rates
increased by around 12, 11 and 8 per cent respectively compared to 5 per cent
overall, and Indian and Chinese women’s rates of degree-level education increased
by 15 and 11 per cent compared to 7.5 per cent overall for women.
Disability and same-sex based differences remained highly significant in both periods
and there were signs that the differences were increasing over time. In so far as
degree-level qualifications are concerned: disabled people were in a disadvantaged
position and men and women in same-sex relationships were in a more favourable
position, compared to non-disabled people and men and women in non-same-sex
relationships. In 1996/7 (Table 1b), the gaps between non-disabled and disabled
people were 12 per cent for men and 8 per cent for women; in 2004/5 (Table 1a), the
gaps widened to 14 and 11 per cent respectively. Similarly, the gaps between men
and women in same-sex and other relationships widened from 18 per cent for men
and 22 per cent for women in 1996/7 to 22 and 25 per cent respectively in the later
period.
20
EDUCATION, EMPLOYMENT, INCOME AND CLASS
3.2 Employment and group-based inequalities
The data on employment status in 2004/5 are found in Table 2a and those for 1996/7
are found in Table 2b. The last row in Table 2a shows the overall gender difference
in employment status in 2004/5. We can see that the majority of men were in
employment (78 per cent) with just under a fifth of men being inactive (19 per cent).#p#分頁標(biāo)題#e#
The proportion of women in gainful employment was lower (67 per cent) and women
in unemployment was also slightly lower. A much higher proportion of women were
inactive (31 per cent).
With regards to ethnic differences in employment, White men and women had the
highest rates of employment (79 and 69 per cent respectively) and men and women
in all other ethnic minority groups had statistically significantly lower rates –
especially Chinese and Pakistani / Bangladeshi men (58 and 61 per cent
respectively) and most strikingly, Pakistani / Bangladeshi women (23 per cent). The
patterns for unemployment and inactivity closely mirror those for employment. In both
aspects, we find that White men and women were the least likely to be unemployed
and inactive as compared with their counterparts in all other ethnic minority groups.
With regard to unemployment, Black Caribbean, Black African, Pakistani /
Bangladeshi and ‘Other’ men had rates two to three times as high as that for White
men, and similar patterns were found for women, albeit to a smaller extent in
absolute terms. These findings confirm previous research on ethnic differences in
employment status (Lindley et al. 2006; Li, 2007b; Heath & Li 2007, 2008; NEP 2005,
2007).
With regard to differences in employment by disability and same-sex status, we find,
as expected, that disabled men and women had much lower levels of employment
(44 and 40 per cent respectively) and higher rates of inactivity (52 and 56 per cent
respectively) than non-disabled men and women. For both men and women, the
differences between non-disabled and disabled people were highly significant. These
findings match earlier research using Census data (Karn, 1997).
In terms of employment, little previous research exists on men and women of samesex
orientation. In this analysis we found that men and women in same-sex
relationships had the highest rates of employment across the total population of men
and women (87 and 84 per cent respectively) and the lowest rates of unemployment
21
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
and inactivity. The patterns here could also suggest that among those in same-sex
relationships, people in employment are more open about, and hence more likely to
report, their sexual orientation.
The patterns for employment status in 1996/7 (Table 2b) were little different from
those in 2004/5, although employment rates for both sexes were higher,
unemployment rates lower, male inactivity rates higher, and female inactivity rates
lower in the later period. The pattern of group-based differences in the earlier period
is almost exactly the same as shown above for the later period. Thus, in 1996/7, for
men and women alike: all ethnic minority groups had significantly lower employment#p#分頁標(biāo)題#e#
rates than White men and women; disabled men and women had significantly lower
rates of employment than non-disabled men and women; and people in same-sex
couples had significantly higher rates of employment. There was little if any,
noticeable change in relative terms across the groups.
3.3 Income and group-based inequalities
The data on gross weekly pay are shown in Tables 3a and 3b for 2004/5 and 1996/7
respectively. Over the period covered, gross weekly pay increased by around £130
for men and £100 for women in monetary terms. Our interest here is not concerned
with whether this is mere inflation, a real increase or both, but rather with the
between-group differences and the change within groups over time.
Firstly, with regard to gender differences, there were clear signs of progress over
time. In the earlier period, men earned around £350 per week and women earned
around £191 per week, with the former earning 84 per cent more than the latter. In
the later period, men earned £480 per week as compared with £294 for women, with
men earning 64 per cent more than women. Thus, in the 10 year period, men’s lead
in gross weekly earnings dropped by 20 per cent and women’s position improved
correspondingly.
The profile with regard to ethnicity is more complicated. For men, we find that at both
time periods most ethnic minority groups earned significantly less than White men,
with Pakistani / Bangladeshi men earning the least – although Indian men in the later
period and Chinese men in the earlier period had non-significant differences from
White men. For women: Black Caribbean women earned significantly more than their
White peers at both time periods (most probably due to their higher social positions,
22
EDUCATION, EMPLOYMENT, INCOME AND CLASS
as we shall see in the next subsection); Indian women had significantly higher
earnings in the later period; and Chinese women had significantly higher earnings in
the earlier period. Pakistani / Bangladeshi women were found to have the lowest
earnings at both time periods.
It is of interest here to compare the changes in the earning profiles. Of particular note
is the change for Indian and Chinese men and women. As shown in Table 3b, Indian
men had significantly lower weekly income in 1996/7 but they were earning
somewhat (albeit non-significantly) more than White men in the later period (Table
3a). Indian women had a similar earning profile to White women in the earlier period
but were found to have significantly higher weekly incomes in the later period. The
picture for Chinese men and women was in the opposite direction. In the earlier
period, Chinese men were found to have similar incomes to those of White men but
in the later period, they were found to have significantly lower incomes than their#p#分頁標(biāo)題#e#
White peers. Chinese women were found to have significantly higher incomes than
White women in the earlier period but this lead was lost in the later period, which may
reflect changing age profiles.
The patterns for disability and same-sex status were similar to those for education
and employment in that disabled people tend to have poorer outcomes (in terms of
labour market incomes) and those in same-sex couples tend to have higher earning
power. We also notice that the significant lead of men in same-sex relationships over
men in non-same-sex relationships in the earlier period became non-significant in the
later period (but note the small sample sizes involved).
3.4 Class and group-based inequalities
Tables 4a and 4b contain data on social class as defined by occupational positions
for 2004/5 and 1996/7 respectively. We differentiate four social classes: salariat
(professional, higher administrative and managerial occupations); routine nonmanual
(such as lower grade administration and office clerks), small employers with
or without employees (otherwise called petty bourgeoisie), and manual working class
23
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
(including agricultural labourers).8
A notable feature in Table 4a is that both men and women had an even split between
middle- (salariat) and working-class positions – around 40 per cent in each for men
and around 37 per cent for women. Women were more likely to be in routine nonmanual
positions and men in self-employment. Men were also slightly more likely to
be in working-class positions than women (41 and 38 per cent respectively).
The class pattern for ethnicity is rather different from that in employment. Here we
find more differences among ethnic minority groups, than between them and the
White majority group. In 2004/5, 40 per cent of White men and 37 per cent of White
women were in the salariat, the most advantaged social class. Indian men were
significantly more likely to be in the salariat (47 per cent) whilst Black Caribbean and
Pakistani / Bangladeshi men were significantly less likely to be in this class (28 and
23 per cent respectively). Men of other minority groups such as Black African and
Chinese were not significantly different from White men. For women, Black
Caribbean, Indian and ‘Other’ groups were significantly more likely to be in the
salariat and Pakistani / Bangladeshi groups were significantly less likely to be in this
class – with Black African and Chinese bearing no significant differences to White
women.
The differences in other class positions are also pronounced. Compared with their
White peers, Pakistani / Bangladeshi and Chinese men were around twice as likely to
be in self-employment in 2004/5 whilst Black African men were half as likely (for#p#分頁標(biāo)題#e#
more discussion on self-employment by ethnic minority groups in Britain, see Li,
2007b). Chinese women were also twice as likely to be found in self-employment
compared to their White peers. With regard to manual working-class positions, we
24
8 The ONS used the Socio-Economic Group (SEG) classification before 2000 and the
National Statistics for Socio-Economic Classification (NSSEC) after 2000. We
followed the standard practice in converting the SEG and the NSSEC into the
Goldthorpe-class schema which is used in this report (Heath & McDonald, 1987;
Rose & Pavalin, 2003). The main difference between the SEG and the NSSEC is
with regard to lower-grade routine non-manual occupations which are coded as
‘semi-routine’ in the NSSEC. The implication is stronger for women’s than for men’s
classes. Thus, we find that in 2004/5, there were much lower proportions of women
in routine non-manual and much higher proportions in working-class positions than in
1996/7. However, as our interest in this report is in access to the salariat, the impact
is less significant.
EDUCATION, EMPLOYMENT, INCOME AND CLASS
find that men of Black Caribbean, Black African and Pakistani / Bangladeshi heritage
had markedly higher rates than White men, and that women of Black African and
Pakistani / Bangladeshi origins also had much higher rates than their White peers.
A comparison with the 1996/7 data show some important changes. Apart from
patterns for disability and same-sex couples, which show the same relative patterns
as in the current period, there are some notable differences with regard to ethnic
group (Table 4b). In the earlier period, none of the ethnic minority groups were more
likely to be found in a more advantaged salariat position than Whites, and this held
for men and women alike. Of particular note here is the finding that Black Caribbean
and Pakistani / Bangladeshi men, and Indian and Pakistani / Bangladeshi women
were significantly less likely to hold salariat positions in comparison with their White
peers. The profile for self-employment and working-class positions was basically the
same as in the later period. Thus, the most notable feature is that, compared with 10
years earlier, Indian men and women, and Black Caribbean women moved from a
position where they were either significantly less than or not significantly different
from their White peers in gaining access to the salariat to a position where they were
now significantly more likely to be found in such positions.
A similar pattern to that found earlier for employment is that, for men and women
alike, disabled people are significantly and markedly less likely to be found in the
salariat than non-disabled people and that people in same-sex couples are
significantly more likely than those in non-same-sex couples to find themselves in#p#分頁標(biāo)題#e#
these positions (Table 4a). In this regard, it is noticeable that exactly the same
pattern is found in the 1996/7 data (Table 4b), although absolute rates differ
somewhat between the two time periods.
3.5 Summary
We have given a fairly detailed account of the patterns and trends of gender, ethnic,
disability and sexual differences in education, employment, income and class in the
two periods. These can be summarised as follows:
• Women’s position improved in education, employment, income and class over
time, as compared to men.
25
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
26
• All ethnic minority groups had significantly lower rates of employment than the
White majority groups at both time periods, with the lowest rates among
Pakistani / Bangladeshi men and women and Black Caribbean men.
• Pakistani / Bangladeshi men and women and Black Caribbean men also had
the lowest proportions in salariat positions and with degree-level qualifications.
• People of Black African, Indian and Chinese origin were educationally highly
qualified but only Indians were apparently able to translate their educational
capital into occupational success. Even Indians did not seem to be able to
make the fullest use of their cultural capital. For instance, Indian men had a
lead of 11 per cent in degree-level qualifications over White men, and yet their
lead in salariat position over their White peers was only 7 per cent.
• In terms of change over time, we found that Black Caribbean men and
women, and Pakistani / Bangladeshi women did not increase their share in
degree-level qualifications. In terms of income, Indian men and women made
notable progress over time, whereas Chinese men and women were moving
in the opposite direction.
• Disabled people had lower rates in each of the four aspects under discussion
in both periods than non-disabled people, hence little change in their position.
• People in same-sex couples had higher rates in each of the four aspects
under discussion in both periods than others, so there was also little change in
position.
Many of the findings reported above confirm some of the existing research on groupbased
inequalities in education, employment, income and class. Some important
changes have been noted. Moreover, this is the first time that systematic research
has explored gender, ethnicity, disability and same-sex relations at the same time.
This section has presented a descriptive and some bivariate analysis, which paves
the way for more systematic multivariate analysis in the following chapter.
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
4. STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
Background characteristics
Before turning to the statistical modelling, it is necessary to give a brief account of#p#分頁標(biāo)題#e#
some other characteristics of the key social groups under consideration. These
characteristics are crucial for our understanding of the labour market participation
and earnings of the groups in question. For instance, labour economists have long
argued that human capital – as indicated by education and labour market experience
(age) – plays a very important role in terms of labour market position. People with
higher educational qualifications are more likely to gain access to higher social class
positions and to make more money. Younger people and those approaching
retirement are less likely to be in employment and even when in the labour market,
are more likely to make less money. Family situation, such as number of dependent
children and personal health, is also a factor that has a decisive impact on people’s
labour market participation and earnings. Given these and other considerations, we
shall highlight some key points in terms of: mean age; mean number of dependent
children under the age of 16 in the household; presence of dependent children aged
0-5; and proportion with disability / long-term limiting illness; by different ethnic group
and by sex in the two periods. The full data are set out in the Appendix.
What we already know about these groups is as follows:
• For both men and women, ethnic minority groups are found to be younger
than the majority White group at both time periods. Pakistani / Bangladeshi
men and women were the youngest, together with Chinese men in the later
period.
• Compared to the White group, ethnic minority groups (except the Chinese)
have a greater mean number of dependent children under the age of 16 and a
higher proportion of young children under the age of five – this is most
probably owing to their younger age structure. This is particularly the case for
Pakistani / Bangladeshi men and women who were, at both time periods,
found to have larger numbers of children and to be more likely to have
dependent children under the age of five than the White group. Black African
women were, at both time points, almost twice as likely as the White group to
have dependent children.
27
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
• Black African, Indian and particularly Chinese men, were substantially and
significantly less likely to have a disability or limiting long-term illness. Also,
Black African and Chinese women had lower instances of long-term illness. In
spite of their young age profile, Pakistani / Bangladeshi men and women were
found to have significantly higher rates of disability / limiting long-term illness.
As we shall see in the next chapter, these demographic conditions will have varying
impacts on the labour market position of the various ethnic groups in their#p#分頁標(biāo)題#e#
employment status and income levels. We shall take into account these and other
available information on socio-cultural characteristics which are generally assumed,
and are frequently found, to have a significant bearing on labour market outcomes,
and which also have a considerable bearing on policy-making.
Introduction to the analysis
In the remainder of this chapter, we report findings of statistical modelling on two of
the four outcome variables discussed: employment and gross weekly pay (income).
With regard to employment, we focus on being employed and we use logistic
regression which is designed for modelling binary outcomes. We coded being
employed as 1 and unemployment / inactivity as 0. With regard to gross weekly pay,
we use ordinary least regression (OLS) which is designed for continuous outcome
variables. As the employment status and earnings profile differ a great deal between
men and women, we present results for the two gender groups separately.
For each variable, we conducted models for the earlier (1996/7) and the later
(2004/5) periods separately on the pooled data, so that we could model the changes
over time. Within each period, we conducted three models: Model 1 controls for the
three key variables of ethnicity, disability and same-sex together (we have already
seen bivariate tests for each of these variables in the previous chapter); Model 2
adds age,9 age squared, marital status, number of dependent children under the age
of 16, education / class, and country of residence; and Model 3 further adds
interaction effects: ethnicity and education / class, ethnicity and dependent children
under the age of five, ethnicity and age, disability and education, disability and age,
and ethnicity and disability. We also conducted an analysis using the pooled data
28
9 We coded age as: age divided by 10 (and similarly for age squared) to improve the
stability of patterns in the models.
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
29
lso
i &
Heath, 2007b).11
earnings are in Tables 6a for men and 6b for women, and those of the class effects
where, in addition to the variables in Model 3, we included interactions for ethnicity
and period, disability and period, and same-sex and period. When we use education
as a main predictor and in interaction effects, class is not used. Similarly, when class
is used as a main predictor and in interaction effects, education is not used. This is
because of the generally close association between education and class, and
because many respondents reported either class or education but not both. To use
both education and class in the same models would thus reduce the sample sizes
and make the estimates unstable.10 This nested modelling follows a clear
sociological rationale. For instance, human capital theories (Becker 1957, 1964)#p#分頁標(biāo)題#e#
assume that as employers in a free market are keen to maximise their profits, people
with skills that can increase productivity are highly valued in the labour market. As a
result, those with higher levels of educational qualification and greater work
experience are more likely to be in employment and to have higher earnings. There
are many theories and research findings that show that, apart from human capital
differentials, employer and societal-level discrimination against the minority groups –
women, ethnic minorities, disabled people, gay men or lesbian women – should a
be taken into account (Akerlof, 1997; Borjas, 1995; Darity & Mason, 1998; L
The estimates of the effects of education on employment status are presented in
Tables 5a for men and 5b for women. The estimates of the educational effects on
10 In addition to all the variables in the modelling tables in this chapter, we carried out
a series of models including both class and education as main effect variables and in
interaction effects. However, the patterns are not clear. Further analysis shows that it
is those who have very poor education, unstable labour market engagement or are
long-term unemployed that are most likely to have missing data on class. This
confirms existing research (Cheung & Heath, 2007). All the additional analyses were
submitted to the Equality and Human Rights Commission and we have consulted
them and obtained their approval for not using the data from the education and class
models in the final report.
11 We did not conduct interaction models for same-sex status with other sociodemographic
variables as the sample sizes for same-sex couples are too small.
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
on earnings are in Tables 6c for men and 6d for women.12
4.1 Logit models of employment
The data in Tables 5a and 5b are coefficients from the logistic regression on
employment status for men and women respectively (together with other covariates).
The results pertain to the log of odds ratios namely, a comparison of probabilities
between two groups in terms of being employed rather than unemployed. The
reference groups have their log odds set as 0. Thus with all other variables in the
models controlled for, figures (coefficients) lower than 0 would mean less favourable
situations and coefficients higher than 0 would mean more favourable situations, in
terms of gaining access to the labour market, compared to the reference group. For
example, see the figure -1.962 for disabled men in Table 5a, under the heading of
Model 1 for 1996/7. This indicates that, holding constant ethnicity and same-sex
status compared to non-disabled men, disabled men have less favourable chances
of being employed and of avoiding non-employment. The figure is in terms of logged#p#分頁標(biāo)題#e#
odds ratio. If we combine the constant and this figure, we may get the probability of
disabled men being employed at 42.0 per cent, compared to 83.7 per cent for nondisabled
men.13 This is fairly close to the 41.1 per cent for disabled and 82.9 for non-
30
12 It is noted here that as respondents in the skilled manual working class (manual
supervisors, lower grade technicians and skilled manual workers) tend to have
similar employment security and earnings profiles as routine non-manual workers or
small employers (classes IIIa, IV-VI in the Goldthorpe class schema), we group them
into the same ‘intermediate’ class in this part of the analysis, leaving the working
class as composed of semi or unskilled manual workers or lower grade routine
workers (classes VIIa, b and IIIb). It is also noted here that, as in the educational
analysis in Tables 5a and 5b, we use the continuous version of class in the table.
This is because if we used the categorical version, this would add many more
categories, particularly in the interaction effects, making the presentation of the table
difficult given the number of other variables already entered in the models. We also
carried out analysis using the categorical dummies in all the corresponding analyses
and the patterns are similar. The results for the dummies are not presented but are
available on request.
13 The expected probability is calculated as the logged odds divided by 1 + the
logged odds. In the present example, the probability of being employed for the
disabled men is e(1.638 + (-1.962)) / (1+e(1.638 + (-1.962))) = .41970123 or around 42.0 per
cent, and for the constant it is e(1.638) / (1+exp(1.638)) = .83726261 or around 83.7 per
cent. Please note that the figures from Table 2b do not control for ethnicity and
same-sex status but the predicted values here do control for the two variables. We
do not need to know the formulae for converting logged odds, odds ratios and
probabilities, as statisticians have done this for us already.
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
disabled men as shown in Table 2b, where only bivariate but not multivariate
significance tests were employed.
Of course, we do not need to turn these coefficients into proportions in order to
understand the patterns. We only need to note the sign and magnitude associated
with each coefficient in a comparative way – that is, in comparison with other
coefficients. Another thing to note is the (number of) stars (*) following the coefficient,
which indicate significance levels. One star implies significance at the 5 per cent
level, two at the 1 per cent level, and three at the 0.1 per cent level. For example,
significance at the 0.1 per cent level actually means that the chances of this kind of
difference (in terms of sign and magnitude of coefficients) being due to sampling#p#分頁標(biāo)題#e#
error are very slight indeed (less than in 1 out of 1000 samples). This further implies
that we can be fairly sure that the difference in question is an accurate estimate of
the real difference in employment between non-disabled and disabled men in the
population during that period. Later on in this chapter, we shall use predicted values
from the models which are then turned into probabilities and shown in graphs for
easy comprehension.
Logit models of male employment, with education as a predictor
The data in Table 5a show the coefficients for logistic models of male employment.
Under Model 1 for 1996/7 (the earlier period) and 2004/5 (the later period), if we
control for ethnicity, disability and same-sex status and hold constant the other
factors in the model, we find that:
• All groups of ethnic minority men were less likely to be employed than White
men in both periods, with the Black African and Pakistani / Bangladeshi men
being the least comparable with White men in the earlier period, and Chinese
and Pakistani / Bangladeshi men being the least comparable with White men
in the later period
• Disabled men were the most disadvantaged in both periods
• Men of same-sex status were more likely to be employed in both periods.
We also find some notable changes in the coefficients over time. As the reference
groups (White, non-disabled and non-same-sex) have a value of 0 in the table, a
change towards 0 from the negative signs would imply an improvement in
employment status and a change towards 0 from the positive signs would mean
31
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
otherwise. Thus, we need to note the signs and the changes in absolute terms. For
example, in the earlier period, the estimate for Black Caribbean men is -0.737 in
terms of logged odds; this became -0.619 in the later period. As the value for White
men was set as 0, Black Caribbean men moved closer (by 0.737 – 0.619 = 0.118 in
terms of logged odds) to the White men in their relative employment chances over
the period covered. Proceeding from this, and again holding constant the other
factors in the models, we find that in the 10-year period:
• The relative chances of employment improved for men of Black Caribbean,
Black African, Pakistani / Bangladeshi origin and Other ethnic groups
• The relative chances of employment improved for the disabled men
• The relative chances of employment for Indian and Chinese men compared to
White men worsened over the period, particularly for the latter (by a
magnitude of (-0.819 – -1.315) = 0.496 in log odds)
• The relative advantages in employment for the same-sex men over non-samesex
men were reduced over the period (by 0.412 log odds).
The data in Table 5a, Model 2, control for more variables measuring sociodemographic#p#分頁標(biāo)題#e#
and geographic factors. Here, we find that apart from Pakistani /
Bangladeshi men in 1996/7 and Chinese men in 2004/5, all ethnic coefficients
deteriorated in comparison with those from Model 1. This means that once we take
account of the other factors which are included in Model 2, such as education, men in
ethnic minority groups were even more disadvantaged (relative to their White peers)
in gaining access to the labour market than had appeared to be the case from Model
1. When controlling for socio-demographic and geographic factors, disabled men are
also found to be more disadvantaged. It is also noticeable that the coefficients for
same-sex men changed from highly significant to non-significant, from Model 1 to
Model 2. This implies that it was not same-sex orientation that gave the men the
distinct advantages in labour market participation, but other socio-demographic
attributes such as their higher educational qualifications – as shown in Table 1a.
Focusing on the other features in Model 2, we find that in both periods, as expected,
age had a curvilinear function for men’s employment – employment increased as
men became older but after a certain point, it began to decrease. Married men tend
to be more likely to be employed, especially in the later period. Having a large
32
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
number of dependent children depressed men’s employment status in the earlier but
not later period. However for the same number of children, having dependent
children under the age of five increased men’s employment at both time periods,
reflecting perhaps their commitment to the labour market (Chun & Lee, 2001). When
compared to English men, Scottish and Welsh men had lower chances of
employment at both time periods. Education, as expected by human capital theory,
increased men’s chances of employment, other things being equal.
The data in Models 1 and 2 refer to main effects and those in Model 3 to interaction
effects. While the coefficients associated with a category of interest in the main
effects models can be fairly straightforward when compared across models, the
comparison between main effects models and interaction effects models is less
straightforward. For example, when comparing the ethnic disadvantages in Models 1
and 2 (and holding constant age and other factors in the models), we may say that
the situation of Black Africans relative to Whites with similar attributes was even
worse than the overall picture without the controls shown in Model 1. This is most
probably due to Black Africans’ higher educational qualifications (as we saw in
Tables 1a and 1b) which had tended to mask their ‘true’ disadvantage.
However, we cannot directly compare coefficients from Models 1 to 3. Take the Black#p#分頁標(biāo)題#e#
Caribbean case in 2004/5 for example. The coefficients changed from -0.619 in
Model 1, to -0.847 in Model 2, to -2.073 in Model 3. The changes between Models 1
and 2 are slight but those between Models 1 and 3 are dramatic. One might wonder
why the same group suddenly became so much worse (over three times as
disadvantaged). The answer lies in the complementary coefficients in the interaction
effects. For example, if we look at Model 3 for the later period, we find positive
interaction effects for this group with greater education (+0.485), negative interaction
effects for having children under the age of 5 (-0.823), and again positive (but nonsignificant)
interaction effects for age (0.096). What this means is that older Black
Caribbean men with higher education, would be in a much better situation than their
younger and poorly qualified counterparts, especially those with young children. In
other words, it is the young and poorly qualified Black Caribbean men who were
(relative to their White peers) very much disadvantaged in gaining access to the
labour market. The information in Model 3 would allow us to calculate the
employment probability of, for example, a 45 year old Black Caribbean man with a
33
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
degree qualification and no dependent children under the age of five, compared to a
20 year old counterpart with no qualifications and with dependent children under the
age of five. The same reasoning works for all other groups or group comparisons.
The interaction effects in Table 5a, Model 3, were generally weak. Yet, as Black
Caribbean men tended to have poorer educational qualifications, those amongst
them with higher qualifications tended to have more favourable employment
opportunities in the later period. In this sense, higher education did indeed act as a
protection for Black Caribbean men. It is important to realise that these interaction
effects mitigate the main effects. Thus, highly educated Black Caribbean men were
less disadvantaged than their less educated minority group peers, but even the
highly educated were disadvantaged relative to their White peers. The key finding is
that the gap for the highly educated is smaller (-2.073 + 2*.485 = -1.103) than for the
low educated (-2.073).
As noted earlier, disabled men had rather poor employment profiles but those among
the disabled who had higher educational qualifications had significantly less
unfavourable employment rates (relative to their White peers) than their peers with
poorer qualifications. On the other hand, disabled men faced increasing
disadvantages in employment as they grew older, which was true in the earlier and
the later period. There are some other features concerning ethnicity and age,
education and disability in the two periods, as shown in the table.#p#分頁標(biāo)題#e#
With regard to patterns in the pooled data (1996/7 as the base), at the bottom of the
last column of Table 5a, we find that the overall employment situation for men was
more favourable in the later than in the earlier period, with a highly significant
coefficient of 0.130. Yet controlling for all other factors, there is no statistically
significant improvement for any of the ethnic minority, disabled or same-sex groups
relative to their White, non-disabled and non-same-sex peers.
Logit models of female employment using education as a predictor
Turning to estimates for women’s employment as shown in Table 5b, we find many
similar features to those for men. For instance, estimates in Model 1 show that:
women of all ethnic minority groups were less likely to be in employment than their
White peers; disabled women were less likely to be employed than non-disabled
women; and women in same-sex relationships were more likely to be employed than
34
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
those in non-same-sex relationships. All this holds true at both time periods. The
patterns in Model 2 are also similar to those for men. The exception is that, for
women, both the number of dependent children and presence of children under the
age of five had a significant and pronounced negative impact on their employment
chances.
Looking at the interaction effects in Model 3, we find (in both periods) a substantial
and positive interaction effect of education on the employment prospects of Pakistani
/ Bangladeshi women. In other words, higher education tended to protect these
women to a much greater extent than it protects white women (or women from other
ethnic groups). We also see positive interaction effects for number of children under
age five, especially for Black African, Indian and Chinese women. This means that
these groups of women are not as disadvantaged by having young children (other
things being equal) as are White women. This may well be because these ethnic
minority women have greater access to extended family support with childcare.
Another notable feature concerning Pakistani / Bangladeshi women is that (unlike
their male counterparts who followed the White men’s employment profiles in terms
of age) their employment quickly shrank as their age increased (possibly reflecting
generational change).
We also see a positive interaction for disabled women with education: for disabled
women, too, higher education seems to have an especially protective role. However,
there is a negative interaction with age.
We now turn to some of the differences between the patterns in this table and those
for men in Table 5b. We see that differences between Wales (and Scotland to a
lesser extent) and England for women were much less pronounced than for men.
Educational qualifications had a more pronounced impact on women’s than on men’s#p#分頁標(biāo)題#e#
employment.
Looking at the patterns in the pooled data, we find an improved employment situation
for women in 2004/5 compared with the earlier period, a finding similar to men. There
are few notable changes in women’s employment situations in the period covered,
except a relative deterioration in employment by Pakistani / Bangladeshi women over
the decade, as evidenced by the significant interaction term for Pakistani /
Bangladeshi in 2004/5 at -0.264. One explanation is that, for this group, there was an
35
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
increase of 4 per cent (from 6.6 to 10.5 per cent) in degree-level qualifications
(Tables 1a and 1b), but only a 2 per cent increase in employment (Tables 2a and
2b). In this sense, their progress in educational attainment was not matched by a
commensurate increase in employment rates.
4.2 OLS models of earnings (weekly pay)
The data in Tables 6a and 6b are on gross weekly pay from the labour market for
men and for women respectively, using education as one of the predictors. The data
in Tables 6c and 6d are on weekly pay for men and women, using class as one of the
predictors. The structure of the tables is the same as for the employment models
discussed in the previous section. Note that as the dependent variable (gross weekly
pay) is measured in pounds, we only keep one decimal point in the estimates.
OLS models of male weekly pay using education as a predictor
Looking firstly at the data for men’s weekly pay in Table 6a, we find that when only
ethnicity, disability and same-sex variables are in the model (Model 1), most ethnic
minority men had significantly lower weekly earnings than their White peers at both
time periods, with the exception of Chinese and Other men in the earlier period and
Indian men in the later period. Pakistani / Bangladeshi men had the lowest earnings
(£133 and £182 less than the White men in the two periods respectively). Disabled
men earned significantly less than non-disabled men at both time periods and men in
same-sex relationships earned somewhat more in the two periods although in the
later period the difference was not significant.
Turning to the data in Model 2 where more socio-demographic factors are controlled
for, we find that, other things being equal, the disadvantages associated with ethnic
minority status and disability remain largely the same in the two periods, with the
coefficients for Black Africans being even more unfavourable in Model 2 than in
Model 1, probably because their higher education had masked their disadvantages.
The significantly higher earnings for men in same-sex couples in Model 1 in the first
period became non-significant in Model 2, suggesting that it is higher levels of
education that account for the higher earnings of men in same-sex couples. There is#p#分頁標(biāo)題#e#
no significant difference between men in same-sex couples and in non-same-sex
couples in the second period. The patterns for age, children, geography and
education in the two periods were as expected, and were in the same direction as for
36
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
employment status shown in Tables 5a and 5b. Note that with all other factors taken
into consideration, men in Scotland and particularly men in Wales had significantly
lower earnings than their counterparts in England in both periods, more so in the later
than the earlier period.
Education was a highly significant factor for men’s earnings in the labour market (b =
100.8 and 143.6 in the two periods respectively in Model 2), and further analysis
(holding constant all other factors in the pooled data) shows that the change was
significant (b = 46.3, p. = 0.000). As overall earnings and education increased, it was
those at the bottom of the educational hierarchy who were losing the most.
With regard to patterns in Model 3 in Table 6a, the interpretation of the coefficients
for the different groups is complicated by the presence of the interaction effects. We
therefore focused on the interactions themselves. These show that in both time
periods, Black Africans have much lower returns to their education than do other
groups. As we know, Black Africans tend to be rather highly educated, but we
suspect that many of them will have overseas higher qualifications which are not
evaluated favourably by British employers. We also see that Pakistani / Bangladeshi
men with children under the age of five have particularly low earnings, reinforcing
concerns that have been expressed elsewhere about poverty in these families.
With all other factors taken into consideration, men in Scotland and particularly men
in Wales, had significantly lower earnings than their counterparts in England in both
periods – more so in the later than in the earlier period.
Finally, we give a brief account of the changes over time as shown in the pooled
data. On average, men in the later period earned more than in the earlier period.
Pakistani / Bangladeshi, Chinese and ‘Other’ men (in relation to ethnicity), as well as
disabled men, seemed to fare significantly worse than their peers a decade earlier.
OLS models of female weekly pay using education as a predictor
The data in Table 6b are on women’s income from the labour market in the two
periods. Model 1 shows that Black Caribbean and Other women earned more in both
periods, as did Chinese women in the earlier, and Indian women in the later period.
Disabled women were found to earn less, and those in same-sex couples more, in
37
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
both periods. Pakistani / Bangladeshi women were earning substantially and#p#分頁標(biāo)題#e#
significantly less than their White peers.
Turning to the data in Model 2 (with the main effects of the other variables controlled
for), we find that the main patterns were as predicted by human capital theories.
Thus, women with higher educational qualifications and more labour market
experience tended to make more money in both periods. The number of dependent
children in the household had a negative association with earnings but somewhat
surprisingly, the presence of young children had a positive impact on women’s
earnings in the earlier period. This is perhaps due to chance significance, which is
likely to creep in under complex models using large-scale data sets. At any rate, the
effect is, at best, substantively small.
As in the case of men, education was a highly significant factor for women’s earnings
in the labour market (b = 82.0 and 118.7 in the two periods respectively in Model 2),
and further analysis holding constant all other factors in the pooled data shows that
the change over time was significant (b = 35.8, p. = 0.000). Thus for men as for
women, the overall improving structure in earnings and education hit the least
qualified most heavily. Other things being equal, women in Wales and Scotland
earned less than their counterparts in England in both periods, a pattern similar to
that of men.
With respect to the full models (Model 3) in Table 6b (with interaction effects also
taken into account), we find little in the way of a clear and consistent pattern. As in
the case of men, we see that Black African women had significantly lower returns to
education than did the White women. And as in the case of men, disabled women
had lower returns to their educational investments.
As for the changes over time in the pooled data, we find an overall increase of gross
weekly pay over the period and more specifically, disabled women had a more
negative profile in the later, as compared to the earlier period. Pakistani /
Bangladeshi women’s disadvantage was brought into sharper relief when all other
factors were taken into account.
38
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
OLS models of male weekly pay using class as a predictor
The data in Table 6c have the same structure as those in Table 6a except that class
is used instead of education. Here we find, again (as can be expected), that class
exerts a powerful impact on men’s earnings, and is actually stronger than education.
Thus, men in higher classes were on average (holding constant their other sociodemographic
attributes) earning £115.1 and £174.1 more than those lower in the
class hierarchy in the two periods. Further analysis for the pooled data again shows a
significant increase for the interaction effects between class and period (b = 63.6, p.
= 0.000), suggesting that (as in the case of education which is of course strongly#p#分頁標(biāo)題#e#
associated with class) it was those at the lower levels of the class hierarchy who
experienced a smaller increase in earnings.
In most other respects however, the story is very similar to that we told earlier when
we used education as the main predictor. Thus, the results for Model 2 in Table 6c
are very similar to those found in Model 2 in Table 6a, with most ethnic minorities and
disabled people earning significantly less than the members of the reference group.
However, it is perhaps worth noting that the magnitude of the disadvantages is
slightly reduced from those found earlier. This means that these minority groups had
problems in gaining access to the more favourable class situations. However, even
when they did gain access to positions in, for example, the salariat, their earnings
remained lower than those of their equally-qualified White peers. However, we
should be aware that the salariat is a rather broad grouping of occupations, and the
disadvantages shown in Table 6c may simply reflect the fact that minorities have
gained access to lower-level occupations within the salariat. It does not necessarily
mean that they receive less pay than their peers in the same occupation. More
detailed analysis would be needed to demonstrate this.
We also see that in Models 3 for both periods, the pattern of the interactions is fairly
similar to those found when education was used as the predictor, rather than class.
There is thus, no major change in the findings.
OLS models of female weekly pay using class as a predictor
Finally we look at the class effects on women’s earning profiles (Table 6d). We again
find significant class effects (b = 87.0 and 175.5) under Model 2 in the two periods,
39
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
and further analysis for the pooled data shows a significant increase of 89.2 (p. =
0.000) over time.
Holding constant the other factors, we also find in Model 2 in both periods, that some
of the ethnic minority groups, namely, Black Caribbean, Black African and Indian
women were earning more, while Pakistani / Bangladeshi women were earning less,
than their White peers. This is possibly because of differences in the extent of fulltime
and part-time working, which we have not been able to take account of in this
model. We also see that disabled women were earning less in both periods while
women in same-sex couples changed from significantly more to non-significant from
the earlier to the later period (other things being equal). Again, exactly as in the
earlier analysis when education was used as the predictor rather than class, women
in Wales and Scotland were earning less money than their peers in England in both
periods, other attributes holding constant.
4.3 Predicted values from employment and income models#p#分頁標(biāo)題#e#
We have given a fairly detailed account of the statistical modelling results for
employment and income (earnings from the labour market). In this section, we
present graphic information based on predicted values from the full models (Model 3)
in Tables 5a- 6d, hence controlling for all other socio-demographic information in the
models. Figures 1-4 show the predicted values for employment status and income by
ethnicity, disability and same-sex status for men and women in the two periods.
Figures 5a and 10 further differentiate ethnicity and education, disability and
education, and same-sex and education combinations for employment and income,
for men and women in the current (2004/5) period. For income, we also include
ethnicity and class combinations to see the protective effects of class on income. In
each figure, we set the values of the reference groups – White, non-disabled and
same-sex – respectively at 100 so that the profiles of each of the other groups can be
directly compared with the reference groups. Differences that manifest themselves
can thus be regarded as gaps in terms of per cent from (that is, above or below) the
reference groups, holding constant all other factors in the models.
40
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
Predicted values on employment by ethnicity, disability and same-sex
relationship
Figures 1 and 2 give predicted values of male and female employment respectively,
in the two periods. In terms of male employment, we find that in the first period,
Indian men’s employment rates most closely matched those of White men while all
other ethnic groups were around 20 to 30 per cent lower. Black African and Pakistani
/ Bangladeshi men fared much worse. Although still behind White men, ethnic
minority groups improved their employment prospects in the later period compared
with the earlier period, with the sole exception of Chinese men whose rates dropped
by 7 per cent, from 83 per cent of White men’s rates in the earlier period to 76 per
cent in the later period.
The data in the lower panels in Figure 1 show that employment prospects for
disabled men improved by a slight margin over time, from 49 to 52 per cent of nondisabled
men. The differences between men in same-sex couples and non-same-sex
couples widened by 9 per cent over time – from a gap of 6 per cent to one of 15 per
cent.
Data on women’s employment are shown in Figure 2, again by ethnicity, disability
and same-sex status, and by period. In both periods, we find that Black Caribbean
women had employment rates second only to White women (only seven per cent
lower), and that Pakistani / Bangladeshi women’s employment rates remained the
lowest, at around 68 per cent below White women, with little change over time.
Comparing the relative changes over time, we find that Indian, Other and Black#p#分頁標(biāo)題#e#
African women’s rates grew by six, four and three per cent respectively. Only
Chinese women’s rates dropped, by three per cent.
The pattern for disability and same-sex status for women was similar to that for men.
The relative distances between disabled and non-disabled women narrowed by five
per cent in the period covered whereas, differences between same and non-samesex
groups widened by seven per cent.
Predicted values of income by ethnicity, disability and same-sex relationship
Figures 3 and 4 show predicted values of income for men and women, with the same
structure as that for employment. Figure 3 shows that in the earlier period, Chinese
41
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
and White men had the highest earnings, followed by Other, Indian and Black men –
with Pakistani / Bangladeshi men having the poorest incomes. In the later period,
there was quite a bit of reshuffle concerning the relative position of the different
ethnic groups. Indian and White men were the highest earners and Pakistani /
Bangladeshi men were still the lowest earners. Looking at the changes, the relative
position of Indian men rose by 10 per cent, Black Caribbean, Black African and
Pakistani / Bangladeshi men rose by 3, 3 and 1 per cent respectively, but Other men
dropped by 7 per cent, and Chinese men dropped by 17 per cent.
Looking at the lower panels of Figure 3, we see that disabled men were earning 83
and 82 per cent of the earnings of non-disabled men in the two periods, with little
change in the relative situation. In contrast, the situation of same-sex men was
brought much closer to that of other men over the period covered. In the earlier
period, men in non-same-sex relationships were only earning 65 per cent of what
men in same-sex couples were earning, but in the later period, the figure was 92 per
cent, hence a big reduction of 27 per cent. However, this might be due to a greater
willingness of such men to report that they were in a same-sex relationship rather
than to a real change in earning power. In other words, we need to remember the
possibility of reporting biases. As society becomes more open, these reporting biases
may change.
Figure 4 shows the predicted values of earnings for women. With regard to ethnicity,
we find that at both time periods, women in most ethnic minority groups were earning
more money than White women with the exception of Pakistani / Bangladeshi
women. As we have emphasised earlier, it is important to recognise that this may be
because of differences in the number of hours worked, rather than actual differences
in wage rates. The rank order in the earlier period was Other, Chinese, Black
Caribbean, Indian, Black African, White and Pakistani / Bangladeshi. In the later
period, the rank order was Black Caribbean, Indian, Other, Chinese, Black African,#p#分頁標(biāo)題#e#
White and Pakistani / Bangladeshi. In terms of relative changes, we find that: Indian
and Pakistani / Bangladeshi women’s positions rose by seven and five per cent
respectively; that there was little, if any change for the Black groups; and that
Chinese and Other women’s positions fell by 13 and 15 per cent respectively.
42
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
The incomes from the labour market of disabled women did not improve over time. In
the earlier period, they were 13 per cent behind non-disabled women; in the later
period, they were 16 per cent behind. However, the differences between women in
same-sex and non-same-sex relationships were much reduced. In the earlier period,
the former had a lead of 48 per cent but this dropped to 23 per cent in the later period
– a reduction of 25 per cent.
4.4 Predicted values of employment and income by ethnicity and by
education
In this section, our interest is to see how education protects ethnic minority groups in
gaining parity, with respect to employment and income with their White peers. For
this reason, we organised the data by ethnicity-education and ethnicity-class
combinations; that is, we consider in turn each ethnic group with lower, intermediate
and higher levels of educational qualifications, and in working, intermediate and
salariat class positions. The data are still the predicted values from the full model
(Model 3) in the relevant tables (Tables 5a to 6d) but we restrict the analysis to the
current period as this is of greater relevance to the present report.
Male employment and income by ethnicity and by education
Figure 5a shows the data for men, with the employment data in the left-hand, and the
income data in the right-hand, columns. The overall impression is that there are more
disadvantages for ethnic minority men with medium-level qualifications (O / A Level
or equivalent) in employment and income (middle panels) and with high qualifications
(first degree or above) in income (bottom panel in the right column) compared to
White men with comparable levels of education. For poorly-qualified men, Indians
were doing as well as their White counterparts and Chinese men were not far behind
(by six per cent) in employment. All other groups – Black Caribbean, Black African,
Pakistani / Bangladeshi and Other men were 15 to 20 per cent behind their White
peers. In terms of income, most groups were similar with the sole exception of
Pakistani / Bangladeshi men who were earning less than two-thirds of what their
White peers were earning from the labour market.
For men with a middle-level education, we find that Chinese men were doing the
worst, being just over half as likely to be employed and making little over half as
much money as their White peers. As shown in Figure 5a, they were 47 and 43 per#p#分頁標(biāo)題#e#
43
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
cent below White peers in employment and income respectively. They were even 17
and 8 per cent behind Pakistani / Bangladeshi peers, a group generally regarded as
the most disadvantaged in the British labour market (NEP, 2007).
For the highly qualified, Black Caribbean, Indian and Other men were achieving
parity in employment with White men, and Black African and Pakistani / Bangladeshi
men were 12 per cent behind. It was Black African and Chinese men who were faring
the worst in terms of income, being 25 per cent behind White men. In both regards,
Indian men were doing well.
Data on the distribution of educational qualifications were included earlier in Table
1a. We noticed that White men were the least likely to have the poorest qualifications
and men in ethnic minority groups were 1.5 to 2 times as likely as their White peers
to be poorly qualified. In Figure 5a we see that, except for Indian men, all other men
in the lowest education bracket were disadvantaged in gaining access to the labour
market although once in the labour market, most groups (except Pakistani /
Bangladeshi) were fairly close to the White peers. Yet the real disadvantages
occurred amongst the middle and higher educational brackets, and contrary to much
myth in labour market research, it is not Pakistani / Bangladesh men but men of
Black African and Chinese origins who were least likely to find employment and,
when in employment, they were earning the least.
Female employment and income by ethnicity and by education
The profiles of women, shown in Figure 5b, are rather different from those of men. In
terms of employment, we find that the lower their educational levels, the more
disadvantaged the minority groups are compared with similarly qualified White peers.
This is most clearly shown in the gaps for Pakistani / Bangladeshi and Black African
women when compared with White women. For the three educational levels from the
lowest to the highest, the gaps are 82, 56 and 17 per cent for the former and 42, 40
and 11 per cent for the latter compared to their White peers. In terms of income, few
differences exist. The poorly qualified women from ethnic minority groups (apart from
Pakistani / Bangladeshi women) were earning non-significantly (see Table 6c) more
than their White peers, and well-qualified women from ethnic minority groups, again
with the same exception, were earning similar amounts of money to their White
peers.
44
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
Income from the labour market by ethnicity and by class
The class effects on earnings for men in different ethnic groups (Figure 6), show
quite marked class differences. Apart from the two Black groups, working-class men
from Indian, Pakistani / Bangladeshi and Chinese ethnic groups were only earning 60#p#分頁標(biāo)題#e#
to 70 per cent of what their White peers were earning. For men in the intermediate
and the salariat positions, all minority groups (with the sole exception of Indians in
the salariat) were earning less than their White peers, with the Black groups’
earnings being between 4 and 17 per cent less than those of their White peers, and
Pakistani / Bangladeshi men's earnings around two-thirds to three-quarters of those
of White men.
The patterns here, in conjunction with our previous findings, suggest two related
features: barriers to employment / pay and community structure. The South Asian
and Chinese communities are known for their niche economic activities in Britain,
such as Indian shops, Pakistani / Bangladeshi restaurants and Chinese take-aways.
These places tend to employ co-ethnic workers. Thus, many working-class
respondents in those communities would be more likely to work in such niche sectors
than men from Black groups who tend to find jobs in the mainstream sectors. The
mainstream sectors tend to be more regulated and, for working-class Black men with
jobs, earnings are similar to their White peers, whereas South Asian and Chinese
working-class men tend to work long hours with poor pay. The net disadvantages
associated with Black and Pakistani / Bangladeshi men in the salariat, may be due
more to the different occupations they occupy. The salariat is a very broad category.
It is possible that many Indian and White men in the class were working in highpaying
jobs such as doctors, lawyers, accountants, engineers and higher education
researchers – jobs where the proportions of Black and Pakistani / Bangladeshi men
are lower.
For women in similar class positions, the ethnic differences are generally small. For
working-class women, ethnic minority groups were apparently earning more,
although the details in Table 6d showed the differences were not significant. For
women in the intermediate and the salariat positions, only Pakistani / Bangladeshi
women were earning notably less money.
45
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
4.5 Predicted values on employment and income by disability and same-sex
relationship
In this section, our interest is to see how education protects disabled people and
people in same-sex relationships in gaining parity in employment and income with
their non-disabled and peers in non-same-sex relationships. For this reason, we
organised the data by disability-education and by same-sex-education combinations
– that is, each disability or same-sex group with lower, intermediate and higher levels
of educational qualifications. The data are still the predicted values from the full
model (Model 3) in the relevant tables (Tables 5a to 6d) but we again restrict the
analysis to the latest period.#p#分頁標(biāo)題#e#
Employment and income by disability and by education
Figure 7 shows the data on men’s employment and income by disability and
education. Two features manifest themselves clearly. First, educational effects are
mainly shown on access to the labour market. Thus, holding constant all other
factors, poorly educated disabled men were only half as likely to be in employment
as their highly educated peers, when compared to non-disabled men (39 and 76 per
cent respectively as shown in the left column). Second, for those in employment, the
earnings differences are much smaller: disabled men at each level of educational
qualifications earned around 86 per cent of that of their non-disabled peers. A
cautionary note is in place here, though: our findings in this respect may, or may not
be a valid indicator for labour market discrimination, as there are many other factors
associated with disability that are not controlled for in the models, and many of these
factors are unavailable in the datasets being used.
Figure 8 on female employment and income shows basically the same patterns as
those for men. The difference is that, compared to their non-disabled peers, disabled
women fared better in both employment and income, although the differences for the
highly educated are not significant.
Employment and income by same-sex status and by education
Figures 9 and 10 show the data on men’s and women’s employment and income by
same-sex status and education. Note that to be consistent with the foregoing
discussion, we used same-sex as the reference group (indexed at 100 per cent). For
46
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
men, as shown in Figure 9, education narrows the gaps between same-sex and nonsame-
sex status. Thus, poorly educated men not in same-sex relationships only had
an 84 per cent chance of being employed as compared with their peers in same-sex
relationships. Yet for the highly educated, the figure was 95 per cent. As for income
(in the right column), we find that men in both low and high educational qualification
brackets but who were not in same-sex relationships were earning more than their
peers in same-sex relationships.
The pattern for women (Figure 10) is largely the same, especially in employment.
With regard to earnings, women in same-sex relationships were still earning more in
the middle and higher educational brackets. There may be other unobserved
characteristics. It is also the case that our three-way coding on education is rather
crude. However, given the very large number of variables included in the models and
with the relatively small numbers for certain groups such as people in same-sex
relationships, it would not make sense to make much more refined differentiations.
Note also, that no observations were found for poorly educated women with valid#p#分頁標(biāo)題#e#
information on reported earnings and other characteristics used in the model, hence
no graph was produced for them.
4.6 Summary
In this chapter, we have presented a large amount of data from logistic regression of
employment and OLS regression of income. The main results from the modelling
(tables 5a – 6d) can be summarised as follows:
• Both men and women in ethnic minority groups were generally found to fare
less well than White men and women in terms of employment and to incur
‘ethnic penalties’ to varying degrees (that is, comparing people with similar
levels of human capital as indicated by educational qualifications and work
experience) with the penalty for Chinese and Black Africans appearing most
pronounced.
• For those in employment, ethnic differences in income were still remarkable in
both periods, particularly for men in most groups, yet ethnic penalties (that is,
ethnic differences, while holding constant human capital indicators) were
much less apparent than in employment, lending support to a recent study of
employment and class (Cheung and Heath, 2007). In this regard, one might
47
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
say that the labour market sets very high thresholds at entry level but once
inside, the playing field is less bumpy.
• In relation to employment, differences between disabled and non-disabled
people widened for men and narrowed for women over the decade, and
differences between people in same-sex and non-same-sex relationships
reduced. In terms of income, differences in relation to disability and same-sex
status reduced for both men and women.
Apart from the detailed modelling results, we also presented graphic information
using predicted values based on the full models (Figures 1 - 10). Here the main
results can be summarised as follows:
• Pakistani / Bangladeshi, Black African and Chinese men, and Pakistani /
Bangladeshi women were found to have the lowest relative position in
employment, and Pakistani / Bangladeshi men had the poorest earnings in
both periods. All this generally confirms the great wealth of empirical research
on ethnic differences in the British labour market.
• The relative position of Chinese men and women became worse over the
decade, while most other ethnic minority groups made relatively steady
progress. A socio-culturally distant group, the Chinese in Britain are
geographically scattered, economically segregated and enclaved (Li, 2006,
2007b), and civically and socio-politically disengaged (Li & Marsh, 2008).
Their low socio-political profile may have hampered their socio-economic
integration. Even at the same (intermediate and higher) levels of educational
qualifications, Chinese men fared worse than Pakistani / Bangladeshi men,#p#分頁標(biāo)題#e#
although Chinese women fared somewhat better than Pakistani / Bangladeshi
women in terms of both employment and income in the later period.
• Indians, both men and women, were doing well and were moving towards full
integration in the mainstream British labour market.
• Greater education does help disabled men and women to gain access to the
labour market although, once inside, its impact on income is less obvious. Yet,
our data also show that even for those who have a job and who have similar
levels of educational qualification, disabled people still fared worse than their
non-disabled counterparts.
48
STATISTICAL MODELLING ON EMPLOYMENT AND EARNINGS
49
• As same-sex people tend to be well qualified, the differences between them
and non-same-sex people tend to reduce as we move from lower to higher
educational qualifications, especially for men.
Having completed our analysis on changes in the socio-economic position in the
labour market, in the next chapter we shall turn our attention to more direct measures
of labour market disadvantage, namely discrimination in terms of whether our
respondent has been refused a job or denied an opportunity for promotion in the last
five years. In addition to ethnicity, disability and same-sex status, we shall also look
at religious differences, particularly Muslim effects on job refusal and promotion
blockage.
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
5. JOB REFUSAL AND PROMOTION BLOCKAGE
In the previous two chapters, we have looked at patterns and trends in labour market
position in terms of employment status and gross weekly pay by gender, ethnicity,
disability and same-sex status. In this chapter, we shall focus on subjective
perceptions of unfair treatment (or otherwise termed discrimination) as indicated by
job refusals and perceived promotion blockage. The analysis is based on the pooled
data from the Home Office Citizenship Survey (HOCS) in 2003 and 2005.14 The
HOCS data pertain to England and Wales only. As in the previous chapters, we shall
focus on men aged 16-65 and women aged 16-63. Wherever possible, we have used
the same explanatory variables with the same coding as in the previous chapters. In
addition, we have used religious orientation as another explanatory variable.15 We
code religion as a six-way variable: Christian, Muslim, Hindu, Sikh, Other and No
religion. As the number of respondents who believe in Buddhism and Judaism is too
small for statistical analysis, we have included them in the category ‘Other’. Existing
research based on HOCS 2001 (Li & Marsh, 2008) shows that Buddhists and Jewish
people have generally similar socio-political profiles to those of Christians. Another
point to note here is that as the sample sizes for respondents in same-sex#p#分頁標(biāo)題#e#
relationships are too small for statistical analysis, we have not included the variable
on sexual relationships in this chapter.16
The Home Office Citizenship Survey contains two important questions that enable us
to make some headway on the issue of discrimination. In both years, it asked
50
14 This is mainly for the purpose of improved stability in statistical models arising from
larger sample sizes. We control for year of interview in the models to take into
account the possible time effects.
15 In HOCS 2005, there is only one variable on religion [RELIG]: ‘What is your religion
even if you are not currently practising?’ In HOCS 2003, there are four variables on
religion: [RPASREL] ‘Thinking first of your childhood, were you raised according to
any particular religion?’ If yes, [RRELPAS] ‘What religion was that?’; [RNOWREAL]
‘Do you actively practise any religion now?’ If yes, [RRELNOW] ‘Which religion is
that?’ As people do stop practising religion or change to another religion, we need to
take that into account. And our coding also needs to be compatible with HOCS 2005.
Thus, we coded religion in HOCS 2003 as current religion for those practising and
past religion for those not practising according to the religion in which they were
raised. Thus for both years, the religion variable pertains to religious orientation.
16 There are only 22 respondents in HOCS 2003 and eight in HOCS 2005 who
reported themselves as of same-sex status.
JOB REFUSAL AND PROMOTION BLOCKAGE
respondents who were currently in work or who had had a job in the last five years or
who were looking for a job:
May I check, in the last FIVE YEARS, have you been refused or turned down for a
job?
[IF YES] Do you think you were refused the job for any of the reasons on this card?
Your gender
Your age
Your race
Your religion
Your colour
Where you live
May I check, in the last FIVE YEARS, have you been treated unfairly at work with
regard to promotion or a move to a better position?
[IF YES] Do you think you were discriminated against because of:
Your gender
Your age
Your race
Your religion
Your colour
Where you live
We cannot be certain about the validity of the responses about the reasons for job
refusals or promotion blockages. In theory, it is possible that people might rationalise
any job rejections as being a result of racial or religious discrimination, when in fact
the job rejection was perhaps due to lack of appropriate skills or experience. If this
was the case, we would expect to find the same overall rejection rates for White and
ethnic minority respondents and for Christians and non-Christians but partitioned
differently between the various reasons. On the other hand, it is also possible that#p#分頁標(biāo)題#e#
respondents underestimate how often they have been treated unfairly on racial or
51
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
religious grounds, since they may well be unaware whether their skills and
experience are superior to those of White or Christian applicants for the same job.
While the reasons given for job refusals and promotion blockages must be treated
with great caution, the overall rates of job refusal / promotion blockage will
nonetheless be of great interest. In particular, do we find that ethnic minorities are
more likely to report that they have been refused jobs or blocked promotions than
White British? To be sure, any ‘excess’ ethnic minority refusal / blockage rate might
be due not only to employers’ hiring / promotion practices but also to the applicants’
patterns of application. For example, minority applicants might apply for jobs that are
inappropriate for their levels of qualification and experience, or for their language
proficiency. Although the evidence in existing research on ethnic minority aspirations
(Heath & Li, 2007) suggested that such aspirational differences are fairly small,
turning aspirations into productive skills valued by employers, is a particularly difficult
task for most, if not all members of ethnic minority groups. It could also be argued
that employers ought to make their requirements as clear and as precise as possible
in their job advertisements so that inappropriate applications are deterred. However,
the requirements for many jobs, especially those of a non-technical kind, may defy
precise specification (Warhurst & Nickson, 2001; Jackson, 2007).
In the following section, we shall firstly describe the patterns of subjective perception
of discrimination by ethnicity, religion and disability groups for men and for women,
separately. The measures of discrimination are: the reported rates of job refusal;
promotion blockage; and the overall rates covering the incidence of either. After that,
we shall report findings of statistical modelling on the overall incidence. Finally, we
shall again use graphs to bring into sharper relief the features drawn from the
predicted values of the models of the overall incidence.
5.1 Descriptive analysis of job refusal / promotion blockage
The data in Table 7 show the proportion of male and female respondents in each of
the equality groups who reported that in the last five years, they had been turned
down for a job (‘job refusal’), or received unfair treatment in terms of having been
rejected for promotion or a move to a better position (‘promotion blockage’) or overall
incidence of either kind. The last row shows that, on the whole, men were somewhat
more likely than women to report such incidences of unfair treatment. This is#p#分頁標(biāo)題#e#
52
JOB REFUSAL AND PROMOTION BLOCKAGE
probably due to the greater propensity for labour market participation by the former
and the greater risks of unfair treatment that arise from it. For instance, 29 per cent of
men compared to 27 per cent of women, reported that they had been refused a job or
blocked for promotion in the last five years.
Looking more closely at the patterns associated with disadvantaged groups, we find
serious indications of discrimination and some differential treatment between gender
groups by employers. In terms of ethnicity, the ranking order of disadvantage in
either separate or joint incidence is Black African, Black Caribbean and South Asian
for men, while for women, most ethnic minority groups are similarly disadvantaged
with the Black African group again being the most disadvantaged. Black Caribbean
women tend to consider themselves less disadvantaged in comparison with their
other minority peers, possibly due to their higher occupational class positions. For
instance, they tend to be employed in lower-grade salariat jobs, such as nursing in
the NHS as noted in Chapter 1 (Cheung & Heath, 2007; Mason, 1995).
Further inspection of the data shows that compared to 21 per cent of White men
having been turned down for a job in the last five years, 11 per cent having been
rejected for promotion and 28 per cent having experienced either incidence: the
disadvantages facing Black African men were two to three-fold, while Black
Caribbean men were around twice as likely to have similar experiences. The rates for
men of Indian and Pakistani / Bangladeshi origins were 36 and 41 per cent
respectively in terms of overall incidence. It is interesting to note that Chinese men
(but not women) reported lower rates of unfair treatment than their White peers. This
is probably due to a large proportion of Chinese men being in self-employment,
working in Chinese shops, restaurants and take-aways (Li, 2007b). In that regard,
their lower rates should not be interpreted as implying greater advantages than other
workers but rather as their having taken a ‘pre-emptive’ strategy against the
possibility of job refusals or promotion blockages by mainstream employers. The
disadvantages faced by Black women are less severe than their male peers but
those of South Asian women are similar to their male peers. Chinese women were
around twice as likely to report unfair treatment as their male counterparts.
The differences between religious groups are less pronounced than between ethnic
groups. For both men and women, it is Muslim, Sikh and Hindu groups who were
53
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
more likely to report incidences of unfair treatment than Christians. One interesting
feature that manifests itself here is that the magnitude of such reported incidences is#p#分頁標(biāo)題#e#
similar for the two gender groups, whereas men were reporting much higher
incidences than women in terms of ethnic differences. Sikh women reported a quite
high level of promotion blockage (23 per cent compared with 11 per cent of White
women).
With regard to disability, we find greater differences for women than for men. Rather
surprisingly (given the results reported in Chapter 4), disabled men and women do
not report statistically significantly higher incidences of job refusal than non-disabled
people; however, the rates are significantly higher for promotion blockage. Disabled
women report statistically higher rates of joint incidence than non-disabled women
but there is no significant difference between disabled and non-disabled men. One
possibility is that disabled people do not experience excess job refusals because
they avoid applying for jobs where they anticipate discrimination.
The reported rates for ethnicity, religion and disability groups indicate considerable
perceived disadvantage. In the next section, we shall look more closely at the net
effects, that is, results from statistical models controlling for a range of sociodemographic
and geographic factors. For instance, it is well-known that ethnicity and
religion are not the same. While most people of Black Caribbean origin are
Christians, as many as 16 per cent of Black Africans are Muslims. People from Indian
ethnic heritage have three main religious identities: Hindu, Sikh or Muslim (44, 28
and 16 per cent respectively). We shall therefore take these factors into account in
the modelling exercise.
5.2 Statistical modelling on unfair treatment
As the patterns for job refusal, promotion blockage and joint incidence of either kind
are fairly similar across the groups, we are going to focus on joint incidence in the
modelling. Table 8 shows the data for men and for women respectively. In this table,
the results of three models are reported: Model 1 controls for ethnicity, religion and
disability; Model 2 adds socio-demographic and geographic controls; Model 3 adds
interaction effects (ethnicity and education, Black African and Muslim, Indian and
Muslim, religion and education, and ethnicity / religion / disability in 2005 – with 2003
54
JOB REFUSAL AND PROMOTION BLOCKAGE
as the reference year to control for time effects associated with the three key
variables).17
The data in Model 1 of Table 8 show that when all three key variables are
simultaneously controlled for, ethnic effects are pronounced, disability effects are
weak, but religion effects (apart from categories of ‘Other’ and ‘None’) have largely
disappeared. This is an interesting contrast with the patterns in Table 7, where most
of the religious groups were found to be significantly disadvantaged compared to the#p#分頁標(biāo)題#e#
Christians for both men and women. In terms of the pattern of the coefficients for
ethnicity, we find a similar pattern to that in Table 7, with the rank order of Black
African, Black Caribbean, Pakistani / Bangladeshi, Indian, and Other for men; and
Black African, Indian, Other, and Black Caribbean for women. Controlling for ethnicity
and religion, patterns for disability are the same as in Table 7.
With regard to the data in Model 2 of Table 8 (where socio-demographic and
geographic attributes are also included), we find two main features. First, disability
effects for men have become significant and are almost as marked as for women.
This means that, for people with the same socio-cultural attributes, disabled men do
have a higher sense of unfair treatment than their non-disabled peers – a feature not
visible in Table 7. Second, looking at the effects of socio-cultural factors as shown in
Model 2, we find that (other things being equal) older people tend to be less likely to
report unfair treatment.18 In this case, age may serve as an indication of economic
security in the labour market, as older people tend to be more secure and less
vulnerable than younger people (Goldthorpe & McKnight, 2006). For men and
women alike, being married is associated with a lower degree of subjective
perception of unfair treatment. This is probably because (other things being equal
55
17 We also explored the class effects on discrimination. At a descriptive level, further
analysis shows that people with a job (that is, in salariat, routine non-manual, lower
supervisorial and routine occupations) were similar in reported rates of overall job
refusal / promotion blockage (at around 25-30 per cent) and it was the unemployed,
including long-term unemployed, who reported much higher rates of unfair treatment
(46-57 per cent). Yet, including the class effects in Model 2 for men and women does
not significantly improve the model fit as class categories did not show differential
effects with ethnicity. Therefore, we decided not to include the class effects in the
models and the discussion in the text.
18 Further analysis shows that for men and for women, adding age squared terms do
not yield significant results, suggesting linear, but lack of curvilinear, age effects in
unfair treatment.
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
and when compared with their non-married counterparts) married people tend to be
favoured by employers who may view marriage as a symbol of commitment and
responsibility (Chun & Lee, 2001). Women in Wales are also less likely to report
unfair treatment than their peers in England, suggesting perhaps, greater labour
market competition in England than in Wales. Further analysis shows no significant
differences between ethnic minority women in England and their counterparts in#p#分頁標(biāo)題#e#
Wales. This last finding is due perhaps, to the small sample sizes of ethnic minority
women in Wales.
Surprisingly, education is associated with a greater sense of unfair treatment. In this
regard, the effects of education should not be interpreted as having a protective role
safeguarding people from experiencing and subsequently reporting incidences of
unfair treatment, but must rather be understood from a different perspective. Apart
from teaching people technical knowledge, a more important function of education is
to make people intellectually developed and to give them a critical perspective. As
Gouldner (1979) famously says, education cultivates a ‘culture of critical discourse’.
Thus, a similar incidence of job refusal might be interpreted by the poorly educated
as simply bad luck or lack of skills but by the more highly educated (and more critical)
as unfair. We need to remember however, that this is a main effect rather than an
ethnic-specific effect. In other words, it applies to White respondents as well as to
minorities. Further analysis controlling for all other variables in the model and
including ethnicity and education interaction effects shows that for men, Blacks and
Chinese have a similar perception at each level of education but Indian and Pakistani
/ Bangladeshi groups are more likely to report unfair treatment when they have
higher levels of education. For women, it is highly educated Chinese who are more
likely to report unfair treatment than their poorly educated peers (results are not
shown in the table).
Turning to the results in Model 3 of Table 8, we find that the effects of ethnicity and
religion have all disappeared. This may well be due to the relatively small sample
sizes over the very large number of main effects and interaction effects entries in the
model. We have noted earlier the possible effects of being Black African or Indian
ethnicity and of Muslim religion. In Model 3, we find that the interaction effects are not
significant for either men or women. Another important feature is that the interaction
effects between ethnicity / religion / disability and time are generally non-significant.
56
JOB REFUSAL AND PROMOTION BLOCKAGE
Yet even with all the controls in the model, people in 2005 tend to report lower
incidences of unfair treatment across the board. A third important factor in this regard
is that, even though the interaction effects between potentially disadvantaged groups
and time are largely non-significant, we find that men of Muslim, Hindu and Sikh
religions were less likely to report unfair treatment in 2005 than in 2003 – perhaps
suggesting the very high pressures on them in the wake of the 9/11 event and the
pejorative representations of Muslims in the media (Poynting & Mason, 2007).19
Given the patterns and trends that can be discerned from Table 8, it makes sense to#p#分頁標(biāo)題#e#
base our graphic presentation on the predicted values from Model 2 for men and
women, and to do it for the two data sources separately. This we do in the next
section.
5.3 Predicted values on unfair treatment
The data in Figures 11 and 12 show the predicted values on unfair treatment based
on Model 2 in Table 8. We present data for men and for women respectively, and in
each figure show separately the results for 2003 and 2005. We differentiate three
levels of education as before, and measure the perceived level of unfair treatment of
each of the ethnic minority groups compared with the White majority. It is important to
remember here that we are comparing different ethnic groups within each level of
education rather than between different levels of educational qualifications. It would
therefore be inappropriate to compare the patterns in the graphs with those for
education in Table 8.20
The patterns in Figures 11 and 12 can be summarised as follows:
• In both 2003 and 2005, ethnic minority men were more likely to sense injustice
than their female counterparts.
• For both men and women and in both years, ethnic minority groups (especially
Black groups) are more likely to perceive unfair treatment than their White
57
19 It is likely that the effects of the 7/7 London bombing in 2005 and the subsequent
Islamophobia against the Muslim community in the media, were not fully reflected in
the 2005 survey.
20 Controlling for all other factors in the relevant models, the predicted rates of unfair
treatment for White men are estimated at: 19.8, 29.9 and 32.4 per cent for low,
medium and high qualifications in 2003; 16.8, 23.5 and 26.3 per cent in 2005. Those
for White women are: 17.1, 25.6 and 30.7 per cent in 2003; 13.6, 20.5 and 24.4 per
cent in 2005.
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
counterparts within each educational level – with poorly educated Black
African men reporting unfair treatment three times as much as their White
counterparts.
• Chinese men are the least likely to report unfair treatment at each of the
educational levels and in both years. This may partly reflect their segregated
employment within their ethnic haven as characteristic of the Chinese
community, and partly reflect the centuries-old tradition of fatalism and Taoism
in Chinese culture that may have become ingrained in their world outlook. In
contrast, highly educated Chinese women do feel strongly about, and have a
significantly higher likelihood to report unfair treatment.
5.4 Summary
In this chapter, we have reported patterns of the subjective perception of unfair
treatment in terms of job refusal and promotion blockage in 2003 and 2005 using the
Home Office Citizenship Survey for respondents resident in England and Wales. The#p#分頁標(biāo)題#e#
main findings can be summarised as follows:
• Seen in their own right, most ethno-religious and disability groups report grave
disadvantages in terms of higher rates of job refusal and promotion blockage,
with Black African (and to a lesser extent, Black Caribbean) men and women
reporting more unfair treatment, confirming Cheung and Heath (2007) and
Heath and Li (2007) on the ‘visible’ minorities experiencing the most serious
forms of disadvantage in the labour market.
• When the other socio-demographic attributes are taken into account, the
religious disadvantages tend to disappear but those associated with ethnicity
remain strong.
58
• Although recent evidence suggests there are various difficulties facing Muslim
women in accessing the labour market (Bunglawala, 2008), our evidence
shows that for Black African or Indian women, being a Muslim does not entail
added disadvantage. This is because most of them21 are out of the labour
market (Heath & Li, 2008), which may mean that the small portion who are
economically active are also a highly motivated and self-selected group.
21 Further analysis shows that amongst Muslim women, 81 per cent of Black
Africans, 68 per cent of Indians and 56 per cent of Pakistanis and Bangladeshis were
not working in 2003 and 2005, compared to 65 per cent of White Muslim women.
JOB REFUSAL AND PROMOTION BLOCKAGE
59
• The relatively low perception of unfair treatment by Chinese men may be seen
as arising from their segregated employment and / or cultural traits.
• On the face of it, Black groups (particularly Black Africans) are facing serious
disadvantages of unfair treatment in the labour market and – in the absence of
longer-term data – the available data show that their situation got worse
between 2003 and 2005. This is also true, albeit on a much smaller scale, for
some of the other ethnic minority groups.
There has been much recent discussion on ethno-religious differences in the labour
market. This is because ethnicity data were available for the first time in the 1991
Census and subsequent Government and academic surveys. In the academic
community, there was a suspicion that it might not be ethnicity but religion that was
the more important marker, and cause of disadvantage in the labour market
(Bunglawala, 2008). In this regard, our findings of persistent ethnic and relatively
unimportant religious impacts (in terms of perceived unfair treatment in the labour
market), may come as a surprise. Of course, there could be many reasons to explain
this. One is that ethnicity is a more readily visible feature than religion and is thus
likely to be a more decisive factor at selection processes. The same may be true for
promotion processes as line managers or panel members may not know what#p#分頁標(biāo)題#e#
religion, if any, is being practised by a candidate for promotion.
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYMENT AND EARNINGS
6. SUBJECTIVE PERCEPTION OF QUALITY OF LIFE
In the previous three chapters, we have looked at group-based disadvantages in
terms of access to the labour market, earnings from the labour market and
perception of discrimination in the labour market. In this last empirical chapter, we
shall turn our gaze to a broader horizon: the subjective perception of quality of life.
Following existing research (Ross & Willigen, 1997; Pevalin, 2000; Pevalin & Rose,
2003; Li, 2007a), we use three satisfaction measures as indicators of quality of life.
They are: satisfaction with work life, satisfaction with social life and satisfaction with
life overall. In order to do this, we draw data from the British Household Panel
Survey22 (BHPS) of 2005 (Wave 15), the only data source currently available with
information that can meet our research needs in this chapter.
In the BHPS, the three satisfaction variables are measured as Likert scales ranging
from 1 (not satisfied at all) to 7 (completely satisfied). To be consistent with analyses
in the previous chapters, we have confined our analysis to men aged 16-65 and
women aged 16-63 in Great Britain. As the ethnicity variable is collected the first time
the respondent is interviewed, we merged the variable from the cross-wave data set.
Religion is collected in Wave 14 and merged with the Wave 15 data. Given the
attrition in panel data, only respondents successfully interviewed in both Waves 14
and 15 are retained in the current analysis. There are no data on sexual orientation,
and hence we cannot discuss differences for same-sex relationships. Our focus is
therefore on the intersectionality of ethnicity, religion and disability. We control for all
other socio-economic and geographic variables as we did in previous empirical
chapters. As in the previous empirical chapters, probability weights (in this case
cross-sectional respondent weight) are used in all analyses in this chapter.
6.1 Descriptive analysis of quality of life
The data in Table 9 show the satisfaction scores by ethnicity, religion and disability,
and by men and women. As the scores range from one to seven with higher scores
meaning greater satisfaction, the last row shows that most people were fairly
satisfied with their lives. For both men and women, work life seemed most
satisfactory and social life seemed a little less satisfactory. Women were significantly
60
22 Details at http://www.iser.essex.ac.uk/ulsc/bhps/
SUBJECTIVE PERCEPTION OF QUALITY OF LIFE
more satisfied than men in work life but the two sexes were no different in social and
overall life satisfaction.
Looking at the ethnic differences, no significant differences emerged for men with#p#分頁標(biāo)題#e#
regard to work life (note that the BHPS was not designed for ethnicity research and
there are insufficient sample sizes for ethnic groups, hence our results here should
be regarded as tentative). In terms of social life, we find that Black African, Pakistani /
Bangladeshi and Other men expressed greater satisfaction than their White
counterparts. Pakistani / Bangladeshi men also expressed greater satisfaction with
overall life. For women, most ethnic groups were similar in their subjective
evaluations of the various facets of life satisfaction. Black Caribbean women were
less satisfied with their work life whilst Indian women were less satisfied with their
social and overall life than their White peers.
Religious differences for men were negligible except that Hindu men were somewhat
more likely to express greater satisfaction with their social life than other groups. For
women, Muslims were least satisfied with their social life whilst Hindu women were
least satisfied with their overall life. Again, the numbers are small and we would urge
caution in interpreting the results.
Disabled men and women were less satisfied in their social and overall life than were
their non-disabled peers.
The overall picture is that there are few gender differences in the three aspects of
satisfaction under consideration. The ethno-religious differences are also small, as
are differences on disability. Given this, we shall use the pooled data for men and
women in the modelling exercises below.
6.2 Statistical modelling on quality of life
The data in Table 10 show the results of statistical modelling on the three aspects of
satisfaction: work life, social life and life overall. In each aspect, we present two
models: the main effects of ethno-religious-disability variables and sociodemographic-
geographic variables in Model 1, and additional interaction effects
61
EQUALITY GROUP INEQUALITIES IN EDUCATION, EMPLOYM