Abstract
This study analyses data obtained from the UK Labour Force Survey (LFS) 2002–2013 to examine the ethno-religious differences in the gross hourly pay within the British salariat occupational class. It explores the extent to which these differences are associated with ethnicity, religion or both. The findings suggest that substantial between-group differences do exist, but these differences cannot be attributed to a pure religious or ethnic discrimination. Although two of the Muslim groups experience a greater penalty than many of the other groups, there was no evidence for an overarching ‘Muslim penalty’. There also was no evidence for an overarching ‘Black penalty’. It is possible that within the salariat class, mechanisms other than pure colour and cultural racism are at work.
Keywords
Introduction
Ethnic inequalities in the UK labour market have been well documented over the past couple of decades (Blackaby et al., 1998; Brynin and Güveli, 2012; Clark and Drinkwater, 2007; Heath and Li, 2008; Heath and McMahon, 1997; Leslie et al., 1998; Modood, 2003; Modood et al., 1997). Some recent studies have suggested that religious penalties, especially in relation to Muslims, might also be at work here (Brown, 2000; Heath and Martin, 2013; Lindley, 2002).
However, while there is little debate over the evidence in relation to ethnic penalties, the evidence in relation to the impact of religion and particularly the penalties against Muslims is inconclusive. For example, Joppke (2009) has rejected the claim that Islam is the reason for the employment disadvantages of Pakistanis and Bangladeshis in the UK. More recently, Longhi et al. (2013: 488) have further raised doubt about overarching Muslim penalties in the British labour market (see also Macey and Carling, 2010).
This study contributes to this growing literature on ethno-religious penalties in the UK by examining the extent and dynamics of the pay penalty in relation to 12 ethno-religious groups including the majority Christian White-British. The study utilises four strategies of enquiry to account better for the variation across individuals: (1) it uses a wide range of relatively more homogeneous groups (Longhi et al., 2013) based on their ethnic background and religious affiliation (Johnston et al., 2010; Khattab, 2009; Khattab and Johnston, 2013); (2) it analyses a recent large-sample data set obtained from the UK Labour Force Survey (2002–2013). This allows for within-group heterogeneity to be better decomposed and analysed; (3) it restricts the analysis to men and women aged 19–65, born in the UK or arrived by the age of six – second generation or above – (Heath et al., 2008), holding academic degrees and working in semi-professional, professional and managerial jobs that can be defined as the salariat class (Erikson and Goldthorpe, 1992). These restrictions will allow us to eliminate potential impacts of language issues, overseas qualifications and familiarity with labour market regulations; (4) it deploys two stages of multivariate analysis (Carmichael and Woods, 2000) for a better understanding of the ethno-religious pay gap.
The results of this study show that some groups experience a pay penalty within the salariat class, even after controlling for human capital and individual circumstances. While Christian Black-Africans, Bangladeshi and Pakistani-Muslims suffer the largest wage gaps, Indians, regardless of their religious background, fare comparatively well. Moreover, White-Muslims and Christian Black-Caribbean do not suffer a greater wage penalty than Christian White-British. Assessing the performance of all Muslim groups reveals that an overarching Muslim penalty is highly dubious. In conclusion I argue that the pay inequality within the salariat occupational class cannot be attributed to a pure religious or racial (ethnic) discrimination. There are some indications for a larger penalty against some Muslim groups and one of the Black groups, but it seems that mechanisms other than ethno-religious discrimination are at work here. For example, the disadvantage of Christian Black Africans might not be a result of their blackness per se, but perhaps an unfortunate outcome of their migration history and/or their differential processes of racialisation and experience within the British society (Daley, 1998; Ramamurthy, 2003). In the next section I discuss the literature followed by a discussion of the data and methods. Section three presents the results from the descriptive and multivariate analysis and will be followed by some concluding remarks.
Ethnicity, Religion and Earnings
The literature on the ethno-religious impact on earnings and income inequality reveals an inconsistent picture. Among the first to study this relationship was Max Weber (1905), who argued that earnings and Protestant beliefs are positively related. However, a study by Tomes (1984) found that religious and denominational backgrounds in Catholics and Protestants do not have an impact on earnings. Contrary to this, Meng and Sentance (1984) have shown some significant differences between Jews, Protestants and Catholics in Canada. They found that Jews have substantially higher returns from schooling than the other two groups. This result has been also supported by more recent studies in the USA (Burstein, 2007; Chiswick and Huang, 2008; Steen, 1996).
In the UK, Lindley (2002), in one of the first studies on the religious pay gap, examined whether religious divisions have a greater impact on employment and earnings than being a member of a particular ethnic group in Britain. The study found that even after controlling for religion, substantial ethnic labour market disadvantage still exists. In terms of religious differences, the study found an evidence for a substantial ‘Muslim-penalty’.
More recent studies have affirmed the relative disadvantage of British-Muslims and the advantage of other religious groups such as Jews and Hindus. For example, Longhi and Platt (2008) found that compared to Christian men, only Jewish men and women and Hindu men earn higher hourly pay on average. Muslim men earned less, while men who declared no religious affiliation are the most similar to Christian men. Moreover, Metcalf (2009) asserted the beneficial pay gap for Jews and the adverse pay gap for Muslim men suggesting that there might be a religious discrimination or at least a religious penalty (against Muslims) that needs to be uncovered.
Longhi and Platt (2008) have compared the pay of White-British Christian men with the pay of White-British Christian women, White-British Jewish, Indian-Hindu, Indian-Sikh, Pakistani-Muslim, Bangladeshi-Muslim, Black-Caribbean Christian, Black-African Christian and those declaring no religion. Substantial penalties are experienced by women of all ethno-religious groups, with no substantial variation among the different groups of women. Pakistani-Muslim women are more disadvantaged than other women and White-British Jewish women are less so. Among men, low rates are clear for Bangladeshi-Muslim men, with a penalty reaching that of women. Black-African Christian, Chinese and Pakistani-Muslim experience pay penalties. The authors conclude that it is possible that the pay penalty for Pakistani and Bangladeshi individuals is a ‘Muslim’ penalty rather than an ‘ethnic’ penalty. Although the pay penalty for Black-African Christian men might be an ethnic penalty, it is possible that the penalty is a result of different migration histories and differential experiences of integration within the British society.
Similarly, Longhi et al. (2013) combine the analysis of religious groups with ethnic groups to explore the wage gap. They analyse the differences in the wage gap for selected ethno-religious groups in Great Britain focusing on Indian-Hindu, Indian-Muslim and Pakistani-Muslim men, using White-British Christian men as a reference group. Using a large sample, which was obtained by pooling together 31 quarters of the LFS for the years 2002–2009, they have illustrated that Indian-Hindus have the highest wages, and Pakistani-Muslims the lowest. Indian-Muslims fare better than Pakistani-Muslims but worse than Hindu-Indians. The fact that the experience of Indian-Muslims was different from the experience of Pakistani-Muslims ‘challenges accounts of an overarching Muslim experience of disadvantage’ (Longhi et al., 2013: 488). In their study they have not included Jews, Blacks or other Whites of different religions, so their conclusion cannot be extended beyond the three aforementioned groups.
In an earlier study, Longhi et al. (2009) have ruled out the possibility of religious penalties arguing that pay gaps are a result of the over-concentration of minorities in particular occupations or types of occupations. This argument has recently been further confirmed by Brynin and Güveli (2012). They found that most of the ethnic pay gap was explained by occupational segregation, whereas the pay gap within occupations was less substantial. One of the contributions of the studies of Brynin and Güveli (2012) and Longhi et al. (2009) is highlighting the importance of occupational concentration or segregation as one of the major explanations for the ethnic (or the ethno-religious) pay gap in Britain. In addition to the impact of occupational segregation, Longhi et al. (2009) have argued that the gap, especially at the bottom of the distribution, can be accounted for by differences in personal and job characteristics. For example, the concentration of ethnic-religious groups in part-time (or full-time) work, which explains part of the mean pay differences. A portion of the pay advantage for Indian-Hindus is explained by their over-representation in some highly paid occupations such as professional jobs (especially health professionals) and under-representation in part-time work; whereas the pay gap for both Indian-Muslims and Pakistani-Muslims is explained partially by their concentration in low paid occupations.
Data and Methods
In order to analyse the wage gaps, the study employs data from the UK Labour Force Survey (LFS) over 12 years from 2002 to 2013 giving a sample of 805,848 men and women aged 19–65. Since the analysis has been restricted to people working in professional, semi-professional and managerial jobs with a known (reported) income, the final working sample was 18,761. The LFS is a quarterly sample survey of households living at private addresses in Great Britain. Its main purpose is to provide detailed labour market information but it also collects background data on respondents such as ethnicity, religion, date of arrival in the UK and nationality which makes it ideal for labour market studies.
Dependent Variables
Earnings within the Salariat
The earning (pay) variable has been measured using the ‘gross hourly pay’ variable as it has been derived by the Office of National Statistics. In the regression model we have used the natural logarithm transformation to fit the normal distribution requirement (Oaxaca, 1973). The salariat class category has been defined and measured using the highest two class occupational classes of higher and lower grade professional occupations of the National Statistics Socio-Economic Classification (NS-SEC). This classification is based on Goldthorpe’s class scheme (Erikson and Goldthorpe, 1992). Occupations falling within the salariat are regulated through a service relationship such as lawyers, scientists, higher education teaching professionals and professional engineers (higher salariat) or schoolteachers, social workers, nurses (lower salariat).
Independent Variables
The main independent variable in this study is the ethno-religious backgrounds in which 12 ethno-religious groups have been defined: Christian White-Irish Hindu-Indians Muslim-Bangladeshi Christian Black-Caribbean Sikh-Indians Muslim-White Christian Black-Africans Muslim-Indians No religion – White-British Jewish White-British Muslim-Pakistani Christian White-British
The other independent variables were: age, age2, sex, marital status, number of children under 10, part-time versus full-time employment, occupational attainment (two-digit scale) working in the public sector versus the private sector, length of employment experience, region of residence and year of survey to control for periodical effects.
The Analysis
The first stage of the analysis utilises the method used by Brynin and Güveli (2012). Linear regression models for the natural logarithm of gross hourly pay will be conducted with and without controlling for occupations using the the two-digit British Standard Occupational Classification SOC90/SOC2000. Due to potential auto-correlation of wages within specific occupations, a mixed model (multilevel analysis) will be used with the two-digit occupations defined as the level-2 (Snijders and Bosker, 2002). All the other individual variables will be used within the fixed effect part of the model.
The second stage employs an approach which has been proposed by Carmichael and Woods (2000) in order to analyse further any possible ethno-religious penalties in wages. In this phase of the analysis, I will use the regression models to calculate the predicted gross hourly pay with and without the ethno-religious background in the equation. The difference in the predicted gross hourly pay between the two models will give an indication whether discrimination does exist (Carmichael and Woods, 2000: 78). The predicted gross hourly pay will be compared with the actual pay among Christian White-British.
Findings
Descriptive Analysis
Table 1 shows the results from the descriptive analysis by ethno-religious group. While the table covers most of the variables used in the multivariate analysis, I will only highlight the results for the gross hourly pay. The only variable that is not listed in the table is the two-digit Standard Classification of Occupations due to space (the variable has about 80 categories – occupations). The table shows that Jewish White-British are at the lead earning the highest hourly pay (£25.05), followed by Muslim-Indians (£22.48) and Christian White-Irish (£21.25). In the fourth place come Hindu-Indians with £20.33, whereas the majority Christian White-British come only in the fifth place (£19.19).
Descriptive statistics of the sample by ethno-religious group.
Note: aAverage.
Turning to the lower end of the scale, Table 1 shows that Muslim-Bangladeshis earn the lowest income within the salariat class (£13.35), which is only 70 per cent of the income of Christian White-British or 53 per cent of the group with the highest income (Jewish White-British) within the salariat. Christian Black-African and Muslim-Pakistanis earn on average slightly more than Muslim-Bangladeshi (£16.02 and £16.57 respectively), but their income is 17 per cent and 14 per cent less than the income among Christian White-British. All other groups, including Muslim-Whites, earn on average between £17 and £19.
These results suggest that there are some substantial ethno-religious differences. To establish whether the differences described above are statistically significant, once individual characteristics are taken into account, this section reports on a multilevel analysis evaluating the extent to which the average earnings of the ethno-religious minorities vary from those of the native Christian White-British majority.
The Multivariate Analysis
The first model in Table 2 controls only for the ethno-religious background as a one-level model. The second model controls for the two-digit standard classification of occupations to define the level-2 in the model while retaining the ethno-religious background as the only independent variable. The third model regresses the outcome variable with all the other independent variables.
Mixed (multilevel) linear model for logged gross hourly pay (N = 18,761).
p < .05; **p < .01; ^p < .10.
Compared to Christian White-British, Model 1 shows that Christian White-Irish, Jewish White-British and Hindu-Indians earn significantly more, whereas Christian Black-Africans, Muslim-Pakistanis and Muslim-Bangladeshis earn significantly less.
When the two-digit occupational classification is taken into account in Model 2, some of these coefficients turned statistically insignificant. Christian White-Irish workers still earn significantly more than the comparator group, and Muslim-Pakistanis and Christian Black-Africans earn significantly less. The coefficient for Muslim-Bangladeshis is negative and relatively large, but it is statistically insignificant.
In Model 3, where all independent variables are controlled for, the coefficients of three ethno-religious groups remain significant and negative indicating a lower income than the majority group. These groups are Christian Black-Africans, Muslim-Pakistanis and Muslim-Bangladeshis. All the other groups do not seem to have significantly different earnings than the majority Christian White-British.
Table 3 presents the predicted pay with and without the effect of the ethno-religious background (discrimination) along with the pay ratio which was derived by dividing the predicted earnings for each group by the actual earning among Christian White-British. The results confirm that Christian Black-Africans, Muslim-Pakistanis and Muslim-Bangladeshis experience the greatest pay penalty. For instance, under the assumption of discrimination, Muslim-Bangladeshis earn only 67 per cent of the actual income among the majority group creating a gap of 33 per cent. The comparable gap between Christian Black-African and Muslim-Pakistanis on the one hand and majority group on the other is 22 per cent and 20 per cent respectively. Muslim-Whites and Sikh-Indians have a slightly lower gap of 18 per cent and 15 per cent respectively. A similar gap of 15 per cent exists in relation to Christian Black-Caribbeans. However, the wage gap falls down to only 3 per cent and 4 per cent for Muslim-Indians and Hindu-Indians respectively.
Actual and predicted (mean) gross hourly pay by ethno-religious group.
The pay for each group divided by the actual pay among Christian White-British.
Examining the differences in the pay ratios in column B and C reveals an interesting finding in relation to the earnings of Bangladeshis. Given their initial large wage gap, there was an expectation that the gap would drop by more than 11 per cent. It appears that even if discrimination is removed, the wage gap between them and the majority group remains relatively large (22 per cent). This suggests that a significant proportion of the wage penalty for Muslim-Bangladeshi does not stem from pure discrimination. For other groups (mainly non-White), removing discrimination (Column C) seems to increase their earnings significantly, to the extent that it virtually eliminates these gaps particularly in relation to Muslim-Indians and Hindu-Indians.
Concluding Remarks
This study found that some groups experience a pay penalty within the salariat class, even after controlling for human capital and individual circumstances. However, this penalty does not seem to be strongly associated with religion or skin colour. No solid evidence was found in relation to the ‘Muslim penalty’ as suggested by Lindley (2002) and more recently by Heath and Martin (2013) although two out of the three groups facing on average significantly lower incomes than the majority group were Muslims (Pakistanis and Bangladeshis). Neither Muslim-Indians nor Muslim-Whites face such penalties. Additionally, a large proportion of the pay gap for Bangladeshis was not found to be associated with discrimination confirming the conclusion reached by Longhi et al. (2013: 488) that an ‘overarching Muslim experience of disadvantage’ is highly questionable.
Another important finding is that in relation to the differences between the two Christian Black groups: the pay penalty that is experienced by Christian Black-Africans was substantially greater than that among Christian Black-Caribbeans. In fact, Christian Black-Africans seem to have experienced the greatest pay penalty, even more than any of the Muslim groups. This suggests that their blackness per se might not be the reason for their disadvantage, but perhaps their relatively recent migration and settlement in the UK (Daley, 1998) and the way they have been perceived and presented in public discourses and in the media (Ramamurthy, 2003).
The source of these differences therefore, at least partially, is other than discrimination on colour and/or religious grounds. It is possible that these groups (Black-Africans and the two Muslim groups of Pakistani and Bangladeshi origins), lack some important ‘soft skills’ such as social and communication skills, attitudes towards co-workers and the ability to adapt to various organizational environments which can affect their wages as suggested by some studies in relation to the Black–White gap in the USA (Hall and Farkas, 2011; Moss and Tilly, 1996). Unfortunately, the data presented here do not allow me to examine such explanations which is a limitation of the current study, so future research can indeed expand our understanding of the ethno-religious pay penalties within the salariat class by examining the potential effect of having/lacking ‘soft skills’ on the employment opportunities and wages of the various groups.
It is worth noting that not finding solid evidence for ethno-religious penalties within the salariat class should not of course lead us to conclude that such penalties do not exist in the UK labour market. It is possible that the main barrier or penalty facing minorities in Britain is finding jobs commensurate with their qualifications (see Cheung and Heath, 2007), especially jobs within the salariat class (Khattab and Johnston, 2014). However, when they eventually find these jobs, they tend to receive the same level of returns as majority workers. Thus, the penalties facing Christian Black-African, Muslim-Pakistanis and Bangladeshis should not be automatically attributed to ethno-religious discrimination.
Footnotes
Acknowledgements
The author wishes to thank the editors of Sociology and the three referees for their helpful comments and suggestions. Data from the Labour Force Survey April-June 2002-2013 (Office for National Statistics, 2014) are Crown Copyright and are provided courtesy of the ESRC’s Economic and Social Data Service (ESDS) and distributed by the UK Data Archive. The author alone is responsible for the interpretation of the data.
Funding
This study was supported by the European Commission, Marie Curie Inter-European Fellowship for Career Development, project number 328423.
