Women, labour market outcomes and religion: evidence from the British labour market

ABSTRACT

We analyse a large-scale UK quarterly Labour Force Survey data covering a period of 16 years (2002–2017) to compare labour market outcomes for women of different religious and ethnic denominations. The paper contributes to the existing literature in analysing three labour outcomes: economic activity (participation), unemployment and occupational choice. While the religious penalties facing Muslim women in unemployment were notoriously high and persistent, the results suggest that their participation is significantly lower. Finally, the occupational distribution of Muslim women shows they are similar to white Christian women in higher occupations but have a much higher probability of belonging to lower occupations.

KEYWORDS:

• Muslim women
• UK labour market
• ethnic penalty
• religious penalty
• unemployment
• occupational attainment

JEL:

• Codes: J15
• A14
• J24

Acknowledgments

This research was made possible with a generous grant from the School of Social Sciences and Humanities (SOSH) as well as School of Public Administration and Development Economics (SPADE) at the Doha Institute for Graduate Studies. We thank our colleagues who supported us financially and otherwise. We also thank the anonymous referees of this journal for their useful comments to improve this publication. We take responsibility for any remaining errors.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Although the literature reports differences among Christians in labour market outcomes, we were not able to differentiate sub-groups within the Christian faith for two reasons: first, our data span a long time period, which did not have a unified coding for all the groups during that time span. Second, our main focus is to address religious differences among a fewer set of categories to highlight the main differences. For example, Dilmaghani and Dean (2016) report that protestant females face a high penalty in participation in the Canadian labour market. For the United States, Lehrer (2004) reviews the literature of the role of religion on a host of socio-economic variables. Another study on protestant/catholic unemployment differential reports that the gap has been, although it has not totally disappeared (Rowland et al., 2018).

2 The study of wages is also as important as these three outcomes and deserves a paper on its own right with Oaxaca gap decomposition; thus, we decided to deal with this labour market outcome in a separate paper.

3 A third difference is the reporting of the results employed here, in addition to the log odds reported in the table; a graphical disposition of the predict probabilities is a lot easier to interpret and understand.

4 It is typically found that when the gender dimension is included in the analysis, females’ penalty in the labour market outcomes is more severe than males’. For example, Khattab (2012) reports that married women are less likely than men to be in any occupational category relative to not working, suggesting that they play the role of the house care takers. In another recent study (Di Stasio et al., 2019) that addresses both gender and religion, the authors found that disclosing being Muslim on job applications results in a more severe penalty for women than men, particularly in Norway. On the other hand, Khattab (2016) addressed differences among Muslims of various ethnicities in their wages controlling for gender; he finds that female wages are lower on average than their male counterparts.

5 Country of birth reported in the analysis refers to a dummy variable equal to one if the UK born and zero otherwise (see Table 1S).

6 The UK Labour force surveys have data with rotating panels of 80%, that is each cohort enters the sample for 5 consecutive quarters, thus each quarter one new cohort enters and an existing one leaves. The data can be obtained in panel form; however, our data set does not permit tracking individuals across panels and the analyses were restricted to females between the ages 16 and 64 for the first time entrants only.

7 For more information on this issue visit: https://www.ons.gov.uk/employmentandlabourmarket/peopleinwork/employmentandemployeetypes/articles/monitoringthequalityoflabourforcesurveyresults/2017-05-22.

8 See Cameron and Trivedi (2005).

9 It is common practice to report descriptive statistics on the same sample used in the regression. Here we report the descriptive statistics on the entire sample as we have multiple regressions and subsamples. In this section, we report the time dimension when we observe particular dynamics; overall, our analyses are pooled. The variable ethnicity is also missing for 2000 and 2011 first quarter.

10 The scales of ethnicity in the sample were harmonized across the years to a common set of categories. There are nine categories as follows: 1 (White British), 2 (Mixed), 3 (Indian), 4 (Pakistani), 5 (Bangladeshi), 6 (Chinese), 7 (Other), 8 (Black African), and 9 (Black Caribbean).

11 Religion was recoded such that ‘Christian’ or ‘White Christian’ is the first category, ‘Muslim’ is the second and all ‘Other’ is the third. The ‘Other’ category includes the following: No Religion, Buddhist, Hindu, Jewish, Sikh and any other religion. Although people of the Jewish faith are more likely to be similar to Christian (vis-a-vis labour market outcomes), their inclusion with the Other group was based on Jews being a minority.

12 In the first quarter of 2011, ‘white Christian’ could not be calculated due to missing data on ethnicity; therefore, ‘Christian’ was used instead.

13 The occupations are detailed in Table 1S.

14 Variables were tracked across the quarters; their names were value labels and were unified maintaining the highest level of detail possible. This was not possible in the case of place of birth because of the high variance of countries reported. As such, a single dummy was created taking the value of 1 if the place of birth was the UK. Missing values were dropped across all variables, this will not affect the statistical power due to the very large sample size and our expectations the missing values are random.

15 This selection was based on a decreasing order of hourly pay; although this may seem ordered, the distance between adjacent categories is not uniform and in one case may overlap. Thus, we estimated the occupational choice by multinomial logit.

16 As with unemployment, we ran model 1 of Table 2 with country of birth as the only independent variable. The coefficient increased from 0.18 to 0.42; this is expected as it carries other effects of omitted variables. It still remained highly significant.

17 The reported coefficients are those of log-odds, for example, the age coefficient is −0.127 which produces an odds ratio of 0.88 < 1 which implies an increase in age lowers the probability of unemployment.

18 This category does not include currently married women who are living with spouse, it includes divorced, widows, separated, etc.

19 The model in column 1 of Table 2 was estimated with country of birth as the sole explanatory variable, the coefficient becomes −0.5 instead of −0.11 which is expected but remains highly significant.

20 Row 1 of Table 4 shows the following occupations: managers and senior officials, Professional Occupations, Associate Professional and Technical, Administrative and Secretarial, Skilled trades and Personal service.

21 It should be noted the model estimates for the results in Figure 2 are not reported. This reported coefficient in Table 4 is with reference to all occupations considered, while Figure 2 only refers to three occupational aggregations.

22 This rise in the lower and higher occupations came at the expense of medium occupations, the predicted probabilities in those categories (figure not shown) fell in 2011 and stabilized after that.

23 The coefficient of age in the unemployment regression changed signs and became insignificant for model 2 (white Christians) when using the new age specification. For LFP, the coefficient of country of birth became significant for the new age regression (23–64).

24 Two noted differences emerged in Table 5; the coefficient on other turned negative and significant for Administrative and Secretarial as opposed to positive and insignificant for ages 16–64. This means that older cohorts are less likely to be in administrative and secretarial positions relative to the lowest scale positions. The other exception was that of other religion (administrative and secretarial occupations) which remained negative but became strongly significant.

25 Table 4S in the electronic supplementary material file contains information with respect to the probability of belonging to a certain occupation as the migrants’ duration of residence in the UK increases. In some cases (administrators and managers), there is evidence of convergence, the longer they stay, the smaller is the penalty in belonging to that group relative to white UK born. As for the other occupations, the evidence is not consistent.

26 The 2008 financial crises affected the British economy severely. GDP growth was negative between April 2008 and July 2009. GDP did not rebound to its 2008 level until mid-2013 and unemployment until late 2015. For more detail please see the following link: https://www.ons.gov.uk/economy/grossdomesticproductgdp/articles/the2008recession10yearson/2018-04-30.

27 It is reported in the literature (Preston & McLafferty, 1999) that geographic as well as institutional factors may lead to discouragement from participation in the labour market; this is consistent with the job search model where the propensity to participate declines with lower probability of finding employment.

Additional information

Funding

This research was made possible with a generous grant from the School of Social Sciences and Humanities (SOSH) as well as School of Public Administration and Development Economics (SPADE) at the Doha Institute for Graduate Studies.

Notes on contributors

Yousef Daoud

Yousef Daoud is an associate professor of economics at the Department of Economics in Birzeit University. He is also a former program director of the master in development economics at the Doha Institute and former dean of the college of Business and Economics at Birzeit University. His research interests are focused on human capital, inequality, gender, and entrepreneurship.

Nabil Khattab

Nabil Khattab is a professor of sociology at the Doha Institute for Graduate Studies, he has previously held a number of academic positions including a senior lecturer at the Hebrew University of Jerusalem. His research interests centers around a number of issues including social and economic inequality, ethnicity, religion, class and migration.