ethnic identity and young people's outcomes - University of Sheffield [PDF]

ETHNIC IDENTITY AND YOUNG PEOPLE’S OUTCOMES

Emily McDool a, Gurleen Popli a* and Anita Ratcliffe a

February 2017

PRILIMIARY AND INCOMPLETE – DO NOT CIRCULATE

Abstract: The paper examines the impact of the ethnic identity of young people (16 to 21 years) on their well-being and their decision to continue in post-compulsory education. Our paper makes two key contributions to the literature. First, we focus on the impact of the young person’s own ethnic identity on their wellbeing and education decisions. Second, we focus on both the majority and minority groups; much of the literature has so far focused on migrants and ethnic minorities only. For our analysis we use data from Understanding Society, the UK Household Longitudinal Study. The key findings of the paper suggests that for young people, having strong ethnic identity is associated positively with their subjective well-being; however, when we allow for endogeneity this association becomes insignificant.

For

participation in further and higher post-compulsory education, we find a negative effect of strong ethnic identity especially for men.

Key Words: Identity, Young People, Post-compulsory education, subjective well-being

a. Department of Economics, University of Sheffield, UK * Corresponding author: [email protected]

Acknowledgements We acknowledge financial support from The University of Sheffield Economics Departmental Research Investment Project fund. We are grateful to the Data Archive at the University of Essex for supplying Understanding Society waves 1 to 5. interpretation of these data lies with the authors.

All responsibility for the analysis and

1. Introduction At the end of adolescence, young people (YP) begin to make the transition from childhood dependency to adulthood independence (Smith et al., 2015: Chapter 19). This key stage in life involves making numerous key decisions which may have long run implications for the ‘later life success’ of YP such as their educational attainment, labour market performance, health outcomes, relationship formation and overall well-being (Oreopoulos and Salvanes, 2011). YP must consider their school to work transition, with one important decision being whether to continue in post compulsory education.1 Those who continue in education at this age are more likely to attain higher educational qualifications, are more likely to be employed and have higher earnings (Gregg, 2001; Gregg and Tominey, 2005). While health during childhood and adolescence determines future health outcomes, both physical and mental health also play a role in influencing a range of economic and social outcomes during adulthood (Case et al., 2005; Richard and Abbott, 2008; Currie et al., 2010). Other widely accepted determinants of later life success include the ability of YP, socioeconomic status (SES) of their families, and the aspirations of both the young person and their parents (Duckworth and Schoon, 2012).

Another dimension which is being increasingly

recognised as having an impact on later life success is ‘identity’. In this paper we focus on young people (16 to 21 years) and examine the impact of ethnic identity on their well-being and their decision to continue in education. Our paper makes two key contributions to the literature; first, we focus on the impact of young person’s own ethnic identity on their wellbeing and education decisions. While a number of existing studies examine the impact of parental identity on the educational decisions of their children, very few studies look at the impact of YP’s own identity on their outcomes. During adolescence, YP start to explore and form their own identities, the notion of their selfefficacy and a sense of personal agency (Zimmermann and Cleary, 2006). There is a recognition that in early childhood years it is the parental investments in the child (Cunha and Heckman, 2008; Todd and Wolpin, 2007), and parental identity that are important (Campbell et al., 2017). As children get older,

1

The average age of end of compulsory education in most of the European and OECD countries is 16 years (OECD,

2016) 1

especially during adolescence, YP start to take more responsibility of their own actions and it is their own actions, beliefs, and identity which start to determine their outcomes (Del Boca et al., 2016). Identity may therefore shape educational choices alongside well-being, with each having a key role in influencing later life success. The second contribution of the paper is its focus on both the ethnic majority and minority groups; much of the literature has so far focused on (first or second generation) migrants and ethnic minorities only. The sense of belonging to a particular ethnic or cultural group may not be limited to ethnic minorities or immigrants alone. Though limited evidence suggests that a strong ethnic identity is observed less often amongst individuals from a ‘majority’ group (Nandi and Platt, 2014b), the ethnic identity of individuals from the majority population is not homogenous and is argued to be formed inconsistently (Nandi and Platt, 2014a); the effect of this identity upon outcomes such as health and education remains unexplored despite the important role of health and education in a young person’s life course. In our analysis we use data from Understanding Society, the UK Household Longitudinal Study, which started in 2009 and covers all four countries on the UK. The key findings of the paper suggests that for YP, having strong ethnic identity is associated positively with their subjective well-being; however when we allow for endogeneity this association becomes insignificant. For participation in further and higher post-compulsory education, we find a negative effect of strong ethnic identity especially for men. The structure of the paper is as follows: section 2 provides the background literature; section 3 of the paper discusses the data and methodology; section 4 presents the results; and section 5 provides the concluding discussion.

2. Background Identity and educational outcomes Akerlof and Kranton (2000) first introduced identity (a concept well studied by sociologists and psychologists) as a motivator into an individual utility function, where they defined identity as an individual’s self-image both as an individual and as a part of a group. They went on to incorporate the

2

concept of identity into economic analysis of work incentives (Akerlof and Kranton, 2005) and educational outcomes (Akerlof and Kranton, 2002). Much of the empirical literature on identity and educational outcomes comes from the US with a focus on racial identity and educational outcomes. Fryer and Torelli (2010) look at how academic achievements impact the social status2 of Blacks and Hispanics; their findings suggests that high academic achievements increase the social status of Black and Hispanic students up to a point, after which higher achievements lowers their social status as they are labelled as ‘acting white’. The empirical findings of the Fryer and Torelli (2010) corroborate the theoretical predictions of the Akerlof and Kranton (2002): conflict between academic achievement (which is good for labour market outcomes) and peer-rejection due to high academic achievement can lead some minorities to underinvesting in their education. Increasingly, literature is also focusing on the ethnic identity of migrants and their educational outcomes. Nekby et al. (2009) analyse the impact of identity on the completion of higher education by first generation immigrants who moved to Sweden before aged 16 and second generation migrants. The study, which identifies conditional correlations rather than causal estimates, finds that integrated males i.e. men with a strong affinity to both Swedish culture and their original background culture, are associated with significantly higher levels of completed tertiary education than assimilated and marginalised men3. Similarly, Brown and Chu (2012) investigate the determinants of educational outcomes amongst Latino immigrant children in predominantly white European American communities. The study observes first, second and third generation migrants aged between 8 and 11 and finds that a strong positive ethnic identity is associated with more positive academic attitudes. Thum (2013) examines the impact of a strong German identity upon the educational outcomes of young, first generation immigrants and their children. Ethnic identity is measured by the respondent’s ‘Germanness’, based upon day-to-day self-identification whilst educational outcomes are measured by the

2

Where social status is measured as a combination of number of same-race friends and the social status of each

friend. 3

Where assimilation is defined as having strong affinity to Swedish culture but weak affinity to home culture; and

marginalisation is defined as weak affinity to both Swedish culture and home culture. 3

International Standard Classification of Education (ISCED) score that indicates Low / medium / high level of education completed. To overcome the possible endogeneity issue, ethnic identity is measured in 1999, prior to educational outcomes which are observed in 2007. The findings indicate that the educational outcomes of migrants and their children are positively impacted by identification with the culture of the host country. A strong identification with the German culture can assist in closing the gap in educational outcomes between immigrants and second generation migrants. Schuller (2015) analyses the impact of parental ethnic identity on the educational attainment of second generation migrants in Germany, using the German Socio-Economic Panel (SOEP). The study focuses on transition from primary to secondary schools, thus the sample consists of pupils aged 10-14. The study finds that the father’s minority identity and mothers majority identity positively influences the child’s probability of enrolment in intermediate/upper secondary school. While mother’s majority identity is found to work via mother’s language proficiency, father’s minority identity is linked with self-esteem.

Identity and well-being While examining life satisfaction to capture subjective wellbeing, Angelini et al. (2015) similarly adopt SOEP data to analyse the effect of German identity for adult migrants. In a similar manner to Casey and Dustman (2010), a measure of cultural assimilation is constructed, consisting of measures of migrants’ closeness with the German culture alongside German language proficiency. The results indicate a positive relationship between identification with the host country culture (cultural assimilation) and life satisfaction. This relationship is only found to exist for ‘established’ immigrants who migrated more than 10 years ago and second generation migrants. Conversely, for ‘recent’ immigrants, who migrated to Germany within the past 10 years, assimilation with the German culture has a negative relationship with life satisfaction.

3. Data Understanding Society is a representative sample of over 40,000 households across England, Northern Ireland, Scotland and Wales. The survey is the successor of the British Household Panel Survey (BHPS) which ran from 1991 until 2008. Understanding Society is panel in nature, allowing individuals to be 4

tracked over time; it provides data on various aspects of the respondents’ lives including social and economic circumstances, attitudes and beliefs. This is a suitable data source for analyses across ethnic groups within the UK since providing data on individuals from ethnic minorities in the main sample but also in the ethnic minority boost sample of over 4,000 households. Extensive information is provided on the ethnic background of respondents and their family, such as parents and grandparents. Data collection for Understanding Society began in 2009 with respondent interviews conducted between 2009 and 2011 for wave 1 which provided data on over 50,000 individuals. Data collection is undertaken through a variety of questionnaires within each responding household; while data on the household as a whole is collected through the household survey, youths aged 10-15 respond to the youth questionnaire while individuals aged over 16 respond to the adult questionnaire. Six waves of data are currently available; we utilise data from the adult survey of Understanding society from waves 3 to 5 which correspond with the collection years 2011-2013, 2012-2014 and 2013-2015 respectively. We observe individuals aged 16-21 in each wave, who were born before 19974. Over 8,000 individual observation years are provided, excluding proxy respondents5.

Identity The main control variable of interest is ethnic identity which measures the importance of ethnic or racial background to the individual’s sense of who they are. Respondents aged 16-21 are asked about ethnic identity in waves 3 and 4 whilst in wave 5 all adults, aged 16 and over, are asked. In waves 3 and 4 respondents aged 16-21 are asked: ‘How important is your ethnic or racial background to your sense of who you are?’. 6 Respondents are asked to indicate the importance of their ethnic or racial background based upon a five-point scale ranging from very important to my sense of who I am to not at all important

4

We observe only individuals born pre-1997 due to small sample sizes of later birth cohorts

5

This includes individuals who were not present in a specific wave and gave their permission for information to be

provided by another individual, known as a proxy respondent. 6

In wave 5, all adults are asked: We'd like to know how important various things are to your sense of who you are.

Please think about each of the following and tell us whether you think it is very important, fairly important, not very important or not at all important to your sense of who you are: Your ethnic or racial background? 5

to my sense of who I am with a fifth category for don’t know / doesn’t apply responses. We exclude from our analysis the individuals who responded doesn’t apply / don’t know as these responses provide little information on the strength of ethnic identity. 7 The distribution of the ethnic identity variable is reported in Table 1. The distribution across the four categories is very similar with almost 21% of YP saying their ethnic identity was not at all important for them, and a similar number, 23%, saying that their ethnic identity is very important for them. We collapse the four categories into two categories to measure ethnic identity with a binary variable indicating whether the individual considers ethnic identity as being very important or fairly important to their sense of who they are relative to considering ethnic identity as being not at all or not very important to their sense of who they are.8 Table 2 gives the response rate for the ethnic identity binary variable by different ethnic groups in the sample. The reported ethnic groups are collapsed into a small number (five) of categories: whites, Indians, Pakistani and Bangladeshi, black, and mixed and other. In line with the analysis of Nandi and Platt (2014b), the descriptive statistics provide evidence that ethnic minorities are more likely to report having a strong ethnic identity relative to whites. Individuals from Pakistani / Bangladeshi background are the most likely to report a strong ethnic identity, with almost 87% of them reporting a strong ethnic identity; while only 40% of the whites report having strong ethnic identity. In subsequent analysis we look at all ethnic minorities together.

Outcomes The outcomes of interest include the subjective wellbeing and participation in post-compulsory education of YP. Subjective wellbeing is measured by the General Health Questionnaire (GHQ 36) which is

7

Based on this criterion we exclude 702 individuals. Additionally, the fifth response varied between the third/ fourth

wave and the fifth wave; in waves 3 and 4, the fifth response was doesn’t apply with individuals who answered don’t know being coded as missing. In the fifth wave the fifth response was don’t know/doesn’t apply with a secondary don’t know only response being coded as zero. 8

Analysis reported below was also undertaken using four distinct categories for ethnic identity. The results do not

change qualitatively to those reported in the paper. 6

commonly used to measure psychological stress (Clark, 2003; Gardner and Oswald, 2007; Roberts et al., 2011). The GHQ 36 is a 36-point Likert scale derived from 12 questions relating to current mental health; each question has four possible responses which are scored from 0 to 3, usually relating to frequency or intensity. For interpretation ease we code the GHQ so that a higher score indicates better psychological health. The GHQ analysis includes individuals living in England, Wales and Northern Ireland but excludes those from Scotland due to data restrictions9. To analyse the effect of ethnic identity on post-compulsory education, we observe whether the individual attends post-compulsory further education (FE) and higher education (HE) when they are expected to do so, given their school cohort. FE refers to education and study which is undertaken after secondary education in the UK, that is not part of HE. FE courses include those which provide training and qualifications to specialise in a specific job, to gain skills for general employment or to continue in additional FE or progress to HE. The qualifications provided by FE therefore varies by course, ranging from basic qualifications in maths and English to Higher National Diplomas (HNDs). HE is usually undertaken at age 18 or older when students attend universities, colleges or special institutions which provide higher education qualifications such as diplomas and degrees, including bachelor, foundation and post-graduate degrees. We identify the cohort of the individual from their birth month and year which allows for the corresponding dates within which individuals would be expected to attend non-compulsory FE and HE to be calculated10. We identify an individual as in non-compulsory FE if the date of interview falls within the expected FE period and the individual reports their economic activity as full-time student. Alternatively, if the respondent is observed at a date within the non-compulsory FE start and end dates but reports an

9

IMD data, which is utilised when adopting instrumental variables, is unavailable in the correct area level format in

Scotland in the period required. 10

For instance, an individual born in January 1991 would be expected to attend FE in England between Sept 2007

and August 2009 and HE between September 2009 and August 2012. Details of how the school leaving age is identified for different cohorts across the different countries of the UK are provided in Appendix A (section A.1). 7

alternative current economic activity, the individual is identified as not participating in non-compulsory FE. We use the same approach to identify if an individual is in HE. Table 3A provides the distribution of the outcome variables for the whole sample and by subgroups of interest. The subgroups we look at are: minority status, gender, family socioeconomic status, and migrant status. The family socioeconomic status is captured by the highest qualification of the parents, with ‘parent degree’ capturing if there is at least one parent with a degree. Migrant status has been one of the key variables of interest in the literature on ethnic identity; in line with the literature in our analysis we look at YP by their migrant status; where all YP who are first or second generation migrants are classified as ‘migrants’, rest being classified as non-migrants. For GHQ there is no significant difference between whites and minorities, by gender, parental degree, or by migrant status; whereas females have lower mental health relative to males. Minorities relative to whites, females relative to males, YP with at least one parent having a degree relative to nodegree, and migrants relative to non-migrants have a higher rate of participation in post-compulsory education, both in FE and HE. In Table 3B descriptive for the outcome variables are reported by the strength of ethnic identity. Young people with strong ethnic identity tend to have better GHQ scores and higher partition in both FE and HE.

Other control variables Understanding Society provides a wealth of information that allows us to control for a range of individual and household characteristics in line with the relevant literature. Since respondents are interviewed on a household basis, information on the household composition and the characteristics of co-residents of respondents is provided. This information is of interest since allowing for home leavers, two parent households and siblings in the household to be identified and controlled for. Additionally, information on the characteristics of parents and partners is available in the dataset; for example, parental income or the income of the respondent’s partner may be observed and controlled for within our model alongside parental education which, as identified in the existing literature, is a key determinant of educational participation (Chevalier, 2004; Micklewright, 1989). 8

Table A.2 in the appendix gives the definitions of all the variables used in the analysis. Descriptive statistics of the variables are reported in Table 4. The sample largely consists of individuals from white ethnic backgrounds who make up 76% of the sample. Ethnic minorities are more likely to belong to a religion and be migrants relative to whites. Relative to ethnic minorities and males, whites and females are more likely to live with a partner (13% of the full sample) and have their own children (4% of the full sample).

4. Empirical analysis To examine the impact of young person’s identity on their outcomes we estimate the following equation: 𝑌𝑖𝑡 = 𝛽𝑜 + 𝛽1 𝐼𝑑𝑒𝑛𝑡𝑖𝑡𝑦𝑖𝑡 + 𝛽𝑥 𝑋𝑖𝑡 + 𝜀𝑖𝑡 where 𝑌𝑖𝑡 is the outcome of interest; 𝐼𝑑𝑒𝑛𝑡𝑖𝑡𝑦𝑖𝑡 is the binary variable capturing the identity of the young person, taking value 1 to reflect ‘strong ethnic identity’ and 0 reflecting ‘weak ethnic identity’; 𝑋𝑖𝑡 is a vector of control variables; 𝜀𝑖𝑡 is a random error; and 𝛽’s are the coefficients to be estimated, the key coefficient of interest being 𝛽1. We consider three separate outcomes: GHQ, to capture subjective well-being; and two measures of participation in post-compulsory education: participation in FE and participation in HE. For GHQ we estimated an ordinary least squares model (OLS), for FE and HE participation we estimate a probit model. The models are firstly estimated with no controls before a full set of control variables are added. We start by estimating a model for all individuals in our sample. We acknowledge that the effect of ethnic identity may differ between minority and majority groups and therefore additionally undertake analyses of ethnic identity for each of these groups separately. We also estimate the models separately for men and women, by parental education, and by migrant status, to explore whether ethnic identity has a differential impact on YP according to their gender, their household socio economic status, or their migrant status. Results are

9

reported in Table 5; where column (1) reports results with no controls and column (2) reports results with a full set of controls.11 The results indicate a positive and significant effect of strong ethnic identity on GHQ (measured on a 36-point scale). Relative to a weak ethnic identity, individuals with a strong ethnic identity report better mental health, ceteris paribus. When we split the sample between whites and minorities, we find a positive and significant effect of ethnic identity for both whites and minorities with a greater effect for minorities. Splitting the sample by gender suggests that the results from the main model are driven by women whose GHQ is positively and significantly influenced by strong ethnic identity; for males, this is an insignificant effect. When splitting the sample by parental education, we find a positive and significant effect of ethnic identity which is larger for individuals from low socioeconomic background as captured by parents with no degree. Looking at the migrant status, ethnic identity has a bigger effect for non-migrants, relative to migrants. Overall, OLS results indicate a positive and significant effect of strong ethnic identity on GHQ which seems to be driven by women, is greater for minorities and non-migrants, and for individuals with parents who do not have a degree, relative to those with degree educated parents. The effect of ethnic identity on GHQ is of similar magnitude to the effect of having siblings in the household. The effect of ethnic identity on FE is positive and insignificant without controls, but becomes negative and insignificant with controls. Splitting the sample by ethnicity, gender, migrant status or parental education does not change the results. Overall, we find little evidence of a significant effect of ethnic identity on FE participation. Once a range of individual and family characteristics are controlled for in the model, ethnic identity is found to have a negative and significant impact upon the probability of being in HE, only for men, at the expected age. A strong ethnic identity significantly decreases the probability of young men being in higher education at the expected age. Further investigation indicates that white males determine this result.

11

Appendix Table A.3 reports the results for the entire sample, with coefficients for all control variables. For GHQ

men are more likely to report a higher GHQ. Ethnic minorities are more likely to participate in FE and HE, relative to whites. Migrants are more likely to participate in HE, relative to non-migrants. 10

Overall the findings so far suggest that while ethnic identity has a significant and positive impact on GHQ, it has very little impact (other than the negative impact for men) on FE and HE participation. One could argue that the positive and significant effect we find for GHQ could be due to endogeneity between ethnic identity and subjective well-being as captured by GHQ. To test for this we do an IV estimation for GHQ, where ethnic identity is instrumented for using religion and IMD score. The results for the first stage are reported in Table 6 below. YP who say they belong a religion tend to have a stronger ethnic identity. Ethnic minorities also report a stronger ethnic identity. Having parents with degree or A-level education is associated with weaker ethnic identity. In Table 7 IV results for GHQ are reported. Once we instrument for ethnic identity the significant effect of ethnic identity on GHQ disappears.

5. Discussion This paper looks at the relationship between ethnic identity of the YP and their subjective well-being and decision to enter post compulsory education. Analysis focuses on both ethnic minorities and majorities. The key findings of the paper suggest that while a strong ethnic identity is positively associated with the subjective well-being of young people, the effect disappears once we allow for endogeneity. Further analysis is still needed to see if the results from the IV estimation holds for the different subgroups. We find a negative association between strong ethnic identity and young people’s participation in postcompulsory education, this result however only significant for men. These associations are however not robust to the inclusion of a range of control variables and addressing the issue of endogeneity. Evidence from sociology and psychology suggests that for minorities, a positive relationship exists between a strong ethnic ID and perceptions of discrimination and barriers alongside a higher awareness of ethnic and racial stigma; this is argued to negatively influence engagement in education (American Psychological Association, 2012). The results presented in the paper are conditional correlations and not causal, and need further investigation.

11

References 1. Akerlof, G. A., & Kranton, R. E. (2002) Identity and Schooling: Some Lessons for the Economics of Education. Journal of Economic Literature, 40(4): 1167–1201. 2. Akerlof, G. A., & Kranton, R. E. (2005) Identity and the Economics of Organizations. The Journal of Economic Perspectives, 19(1): 9-32. 3. American Psychological Association (2012) Ethnic and Racial Disparities in Education: Psychology’s Contributions to Understanding and Reducing Disparities 4. Angelini, V., Casi, L. & Corazzini, L. (2014) Life satisfaction of immigrants: does cultural assimilation matter? Journal of Population Economics 28(2):817-844. 5. Case, A., Fertig, A., & Paxson, C. (2005) The lasting impact of childhood health and circumstance. Journal of health economics, 24(2): 365-389. 6.

Casey, T. & Dustman, C. (2010) immigrants’ identity, economic outcomes and the transmission of identity across generations. The Economic Journal 120(542): 31-51

7. Campbell…. 8. Chevalier, A. (2004) Parental education and child’s education: a natural experiment. Centre for Economics of Education Discussion Paper 40. 9. Cunha, F., & Heckman, J.J. (2008) Formulating, Identifying and Estimating the Technology of Cognitive and Noncognitive Skill Formation. Journal of Human Resources, 43(4): 738-782. 10. Currie, J., Stabile, M., Manivong, P., & Roos, L. L. (2010) Child health and young adult outcomes. Journal of Human Resources, 45(3): 517-548. 11. Del Boca, D., Monfardini, C., & Nicoletti, C. (2016) Parental and child time investments and the cognitive development of adolescents. Journal of Labour Economics, 35(2): 12. Duckworth, K. and Schoon, I. (2012) Beating the Odds: Exploring the Impact of Social Risk on Young People's School-to-Work Transitions during Recession in the UK. National Institute Economic Review, 222(1): R38-R51. 13. Fryer, R. G., & Torelli, P. (2010) An empirical analysis of ‘acting white’. Journal of Public Economics,

94(5): 380-396. 14. Gregg, P. (2001) The impact of youth unemployment on adult unemployment in the NCDS. The Economic Journal, 111(475): 626-563. 15. Gregg, P. & Tominey, E. (2005) The wage scar from male youth unemployment. Labour Economics, 12(4): 487-509. 16. Micklewright, J. (1989) Choice at sixteen. Economica, 56: 25-39 17. Nandi, A. & Platt, L. (2014a) Britishness and identity assimilation among the UK’s minority and majority ethnic groups. Institute for social and economic research.

12

18. Nandi, A. & Platt, L. (2014b) A note on ethnicity and identity among the UK born population in Understanding society. Institute for social and economic research. 19. OECD (2016) Education at a Glance 2016: OECD Indicators, OECD Publishing, Paris.

DOI:

http://dx.doi.org/10.1787/eag-2016-en 20. Oreopoulos, P., & Salvanes, K. G. (2011) Priceless: The nonpecuniary benefits of schooling. The Journal of Economic Perspectives, 25(1): 159-184. 21. Richard, M. & Abbott, R. (2009) Childhood mental health and life chances in post-war Britain. Insights from three national birth cohort studies. Discussion paper. Centre for Medical Health, London. 22. Smith, P. K., Cowie, H., & Blades, M. (2015) Understanding children's development. John Wiley & Sons. 23. Todd, P.E., & Wolpin, K.I. (2007) The Production of Cognitive Achievement in Children: Home, School, and Racial Test Score Gaps. Journal of Human Capital, 1(1): 91-136. 24. Zimmermann, B.J., & Cleary, T.J. (2006) Adolescents’ development of personal agency, in: Pajares, F., Urdan, T. (Eds.), Self-Efficacy Beliefs of Adolescents. IAP, San Bernardino, pp. 45–69.

13

TABLES Table 1: Ethnic Identity Ethnic identity scale

N

Percent

Ethnic identity binary

N

Percent

4,417

50.26%

4,371

49.74%

8,788

100%

variable Not at all important

1,865

21.22

Not important (weak

Not very important

2,552

29.04

ethnic identity)

Fairly important

2,420

27.54

Important (strong ethnic

Very important

1,951

22.20

identity)

Total

8,788

100%

Note: Data source is Understanding Society waves 3, 4 and 5; young people aged 16-21 years old. In wave 3 and 4 the question asked is: ‘How important is your ethnic or racial background to your sense of who you are?’ In wave 5 the questions asked is: ‘We'd like to know how important various things are to your sense of who you are. Please think about each of the following: Your ethnic or racial background?’.

Table 2: Ethnic identity binary variable by ethnicity Ethnicity

Frequency (percent)

Proportion

reporting

ethnic identity’ White

6,755 (76.9%)

40.40%

Mixed/ Other

522 (5.9%)

71.22%

Indian

312 (3.6%)

77.92%

Pakistani / Bangladeshi

700 (8%)

86.57%

Black

499 (5.7%)

83.97%

ALL

8,788 (100%)

49.74%

Minorities

Notes: for the definition of ‘strong ethnic identity’ see Table 1.

14

‘strong

Table 3A: Mean (standard deviation) of the outcome variables OUTCOME Full Whites Minoriti Males Females VARIABLES sample es

Parent degree

GHQ36

25.324

25.348

25.257

26.438

24.339

25.175

Parent no degree 25.440

(5.743)

(5.622)

(6.101)

(5.264)

(5.987)

(5.689)

(5.782)

(6.136)

(5.632)

N

7151

5239

1912

3363

3788

3129

4022

1524

5627

In FE

0.802

0.778

0.873

0.793

0.812

0.854

0.754

0.844

0.795

(0.399)

(0.415)

(0.336)

(0.405)

(0.391)

(0.353)

(0.431)

(0.364)

(0.404)

N

1698

1266

432

821

877

823

875

256

1442

In HE

0.492

0.434

0.643

0.478

0.491

0.604

0.409

0.613

0.449

(0.500)

(0.496)

(0.479)

(0.500)

(0.500)

(0.489)

(0.492)

(0.487)

(0.497)

3880

2901

979

1790

2090

1648

2232

1006

2874

N

Migrant

Nonmigrant

25.154

25.37

Notes: GHQ = general health questionnaire 36; FE = full time education; HE = higher education. For definition of minorities refer to Table 2. For definition of Parent degree/no degree and migrant status, see Appendix Table A.2.

Table 3B: Mean (standard deviation) of the outcome variables, full sample, by identity OUTCOME VARIABLES

Strong ethnic identity

Weak ethnic identity

25.540

25.095

(5.772)

(5.704)

N

3,680

3,471

In FE

0.806

0.799

(0.395)

(0.401)

893

805

0.515

0.468

(0.500)

(0.499)

1952

1928

GHQ36

N In HE

N Notes: see notes to Table 3A.

15

Table 4: Sample means of the explanatory variables CONTROL VARIABLES

Main model

Whites

Minorities

Males

Females

Parent degree

Migrant

Nonmigrant

0.471

Parent no degree 0.518

Ethnic Identity

0.497

0.404

0.808

0.487

0.507

0.722

0.444

White

0.769

1.000

0.000

0.768

0.769

0.781

0.759

0.261

0.889

Mixed/ other

0.059

0.000

0.257

0.063

0.056

0.066

0.055

0.163

0.035

Indian

0.036

0.000

0.153

0.041

0.031

0.043

0.030

0.113

0.017

Pakistani / Bangladeshi

0.080

0.000

0.344

0.074

0.085

0.033

0.116

0.271

0.034

Black

0.057

0.000

0.245

0.054

0.059

0.078

0.040

0.193

0.025

Male

0.472

0.472

0.473

1.000

0.000

0.476

0.469

0.475

0.471

Cohort 1988

0.006

0.006

0.006

0.005

0.006

0.003

0.008

0.010

0.005

Cohort 1989

0.036

0.038

0.028

0.036

0.035

0.024

0.045

0.044

0.033

Cohort 1990

0.086

0.084

0.091

0.075

0.095

0.066

0.101

0.127

0.076

Cohort 1991

0.123

0.122

0.126

0.129

0.117

0.117

0.127

0.189

0.107

Cohort 1992

0.172

0.175

0.161

0.170

0.174

0.161

0.181

0.208

0.164

Cohort 1993

0.174

0.177

0.163

0.174

0.174

0.180

0.168

0.180

0.172

Cohort 1994

0.173

0.169

0.188

0.177

0.170

0.184

0.164

0.130

0.183

Cohort 1995

0.143

0.141

0.150

0.141

0.145

0.166

0.126

0.082

0.158

Cohort 1996

0.088

0.088

0.088

0.093

0.084

0.100

0.079

0.031

0.102

Cohort 1997

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

Cohort 1998

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

0.000

Mum age

26.154

26.349

25.504

26.613

25.744

29.410

23.588

25.364

26.341

Mum age sq.

742.491 756.614

695.564

764.907

722.451

857.824

651.609

706.406

751.026

Left home

0.118

0.121

0.110

0.100

0.134

0.047

0.174

0.134

0.114

Lives with partner

0.035

0.043

0.007

0.021

0.047

0.010

0.055

0.023

0.038

Two parent HH

0.604

0.613

0.572

0.630

0.581

0.738

0.498

0.575

0.611

Own children

0.033

0.036

0.024

0.010

0.054

0.008

0.053

0.035

0.033

Siblings in

0.442

0.404

0.570

0.460

0.426

0.456

0.431

0.554

0.416

16

HH Migrant

0.191

0.065

0.611

0.192

0.190

0.194

0.189

1.000

0.000

Working parent

0.743

0.777

0.629

0.768

0.721

0.890

0.627

0.642

0.767

Partner & own income

0.706

0.753

0.549

0.566

0.831

0.253

1.063

0.717

0.703

Parent income

7.014

7.053

6.885

7.172

6.874

7.853

6.353

6.736

7.080

Parent degree

0.441

0.448

0.418

0.445

0.437

1.000

0.000

0.446

0.439

Parent Alevels

0.182

0.204

0.112

0.188

0.178

0.000

0.326

0.104

0.201

England nonLondon

0.614

0.653

0.487

0.620

0.609

0.607

0.620

0.519

0.637

London

0.147

0.046

0.485

0.146

0.149

0.162

0.136

0.399

0.088

Wales

0.080

0.099

0.016

0.080

0.080

0.074

0.084

0.030

0.091

Scotland

0.086

0.108

0.011

0.085

0.086

0.095

0.079

0.031

0.099

Northern Ireland

0.073

0.094

0.001

0.070

0.076

0.062

0.081

0.021

0.085

Wave 3

0.357

0.358

0.351

0.351

0.361

0.340

0.370

0.405

0.345

Wave 4

0.323

0.320

0.334

0.323

0.323

0.327

0.320

0.323

0.323

Wave 5

0.320

0.322

0.315

0.326

0.315

0.333

0.310

0.272

0.332

Observations 8788 6755 2033 4148 4640 3873 4915 Notes: For definitions of the explanatory variables, refer to Table A.2 in the appendix.

1681

7107

(.) Report the standard errors of the continuous variables.

17

Table 5: Main results for the binary ethnic identity coefficient

ETHNIC IDENTITY Panel A ALL

GHQ OLS (1) (2) 0.446*** (0.150)

IN FE

0.519*** (0.156)

(1)

(2)

(1)

(2)

0.002 (0.018)

-0.023 (0.018)

0.040*** (0.015)

-0.018 (0.016)

7151

N

IN HE

1954

4226

Panel B: Ethnicity WHITES

0.526*** (0.171)

N Panel C: Gender WOMEN

N Panel D: Parents education DEGREE

N Panel E: Migrant MIGRANT

0.526** (0.223)

4022

0.007 (0.023)

0.795* (0.437)

-0.008 (0.027)

0.010 (0.050)

0.087*** (0.020)

-0.006 (0.020)

-0.006 (0.020) 2441

0.043 (0.051)

0.093*** (0.033)

277 0.469*** (0.165)

-0.031 (0.024) 1782

1016

1524 0.431*** (0.163)

-0.037 (0.024)

0.019 (0.027)

-0.050** (0.023) 1964

896 0.566*** (0.208)

0.016 (0.021)

0.016 (0.023)

935

3129

0.921** (0.435)

0.060*** (0.021) 2262

-0.038 (0.026)

-0.012 (0.024)

-0.006 (0.040) 1018

-0.010 (0.024)

-0.012 (0.027)

0.455* (0.233)

0.365* (0.201)

-0.009 (0.040)

1018 0.179 (0.220)

-0.015 (0.017) 3207

-0.001 (0.038)

0.015 (0.025)

3363

N NON-MIGRANT

0.789*** (0.217)

0.251 (0.208)

-0.033* (0.018)

450

3788

N NO DEGREE

0.006 (0.039)

1912 0.688*** (0.209)

-0.027 (0.020) 1501

0.796** (0.390)

0.624 (0.387)

N MEN

-0.031 (0.022)

5239

N MINORITIES

0.455*** (0.169)

0.029 (0.035) 1066

-0.025 (0.019)

-0.024 (0.018)

-0.026 (0.017)

5627 1676 3160 N NO YES NO YES NO YES CONTROLS Notes: GHQ is OLS, FE (further education) and HE (higher education) are probit marginal effects. First column for each outcome is with no controls. Second column has full set of controls. (.) give the estimated standard errors. * denotes significance at 10%, ** at 5% and *** at 1%.

18

Table 6: Ethnic identity determinants ETHNIC IDENTITY Instrument 1: Religion Instrument 2: IMD score Mixed/ other Indian Pakistani / Bangladeshi Black Male Cohort 1988 Cohort 1989 Cohort 1990 Cohort 1991 Cohort 1992 Cohort 1994 Cohort 1995 Cohort 1996 Mum age Mum age squared Left home Lives with partner Two parent HH Own Children Siblings in HH Migrant 2nd gen Working parent

0.134*** (0.017) 0.001* (0.000) 0.261*** (0.030) 0.245*** (0.035) 0.288*** (0.028) 0.351*** (0.029) -0.015 (0.013) -0.046 (0.084) 0.064* (0.038) 0.030 (0.026) 0.025 (0.024) 0.007 (0.023) 0.018 (0.022) 0.054** (0.024) 0.019 (0.030) 0.000 (0.001) -0.000 (0.000) -0.132 (0.091) -0.032 (0.039) 0.037** (0.018) 0.006 (0.035) 0.016 (0.013) 0.026 (0.020) -0.026 (0.020) 19

Partner & own income Parental income Parent degree Parent A-levels Wave 3 Wave 4 London Wales Northern Ireland Constant N

-0.006 (0.008) -0.011 (0.011) -0.058*** (0.017) -0.047** (0.020) 0.040*** (0.013) -0.005 (0.012) 0.004 (0.019) 0.046* (0.025) 0.010 (0.025) 0.441*** (0.084) 7151

Standard errors in parentheses Note: Ethnicity base: White Parental qualifications base: Below A-levels Living status base: Living at home with parents and no partner Region base: England outside London Cohort base: 1993 Wave base: 5 * p < 0.1, ** p < 0.05, *** p < 0.01

Table 7: GHQ results with IV GHQ IV ETHNIC IDENTITY ALL N CONTROLS:

(1)

(2)

0.601 (0.542)

2.206 (1.422) 7151

NO

YES

20

APPENDIX A A.1 School Leaving Age The expected dates of non-compulsory FE may vary by country and cohort due to differences in the compulsory school leaving age (SLA) between cohorts; differences and changes in the SLA may influence the period in which education is no longer compulsory. For instance, in England the Education and skills act 2008 required school leavers from 2013 to remain in education or training until aged 17; for this cohort, FE was therefore compulsory for up to an additional year, with the non-compulsory participation decision and expected start date being delayed by a year12. Similarly, the age at which non-compulsory education begins may vary between countries due to differences in the educational systems; for example, in Wales individuals may leave education at 16. In Scotland and Northern Ireland, the month of birth determines an individual’s leaving age alongside their cohort; this varies from England and Wales where a cohort comprises of children born between September and August who all complete compulsory schooling in the same month. In Scotland, the school year consists of pupils born between March and February who, based on their month of birth, may leave in May or in December. In Northern Ireland, if a pupil is born between 1st September and 1st July, they may leave compulsory schooling in June, though if born between 2nd July and 31st August, pupils must leave school in June of the following academic year. We take account of the differences in the cohort composition and SLA across countries and adjust the expected non-compulsory FE start dates accordingly. This is indicated in Table A.1, which provides the expected FE start and completion dates for each country. Due to differences in the SLA of Scotland and Northern Ireland within cohorts, we simply observe individuals as in non-compulsory education at or after the age of 17 since at this age, all pupils in Scotland and Northern Ireland who are in education will be in non-compulsory education.

12

Since observing only individuals who were born prior to 1997, due to small sample sizes, individuals who were

affected by the 2014 raising of the compulsory schooling leaving age in England are not included in this analysis. 21

Table A.1: School Leaving Age Country

FE start

FE end

Sept following completion of secondary school (aged 16)

August aged 18

Sept one year after completion of secondary school (aged 17)

August aged 18

Wales

Cohort (*change to birth years* Pre-2013 (Pre-1997 cohort) 2013 (1997 cohort) All

Sept following completion of secondary school (aged 16)

August aged 18

Scotland

All

Month after turning 17

August aged 18

England England

22

Table A.2: Variable definitions VARIABLE NAME

DEFINITION

EXPLANATORY VARIABLES IN X Ethnic identity

Importance of ethnic identity to the sense of who you are 1=Not at all important; 2=Not very important; 3= Fairly important; 4=Very important Binary variable: 0 = Not at all import and not very important (weak ethnic identity); 1 = fairly important and very important (strong ethnic identity)

Controls for ethnicity, gender and school cohort White

Ethnicity = White

Mixed / other

Ethnicity = Mixed / other

Indian

Ethnicity = Indian

Pakistani / Bangladeshi

Ethnicity = Pakistani / Bangladeshi

Black

Ethnicity = Black

Male

1=Male; 0=Otherwise

Cohort year 1988

1=School cohort consists of individuals born as early as 1988; 0=otherwise

Cohort year 1990

1=School cohort consists of individuals born as early as 1990; 0=otherwise

Cohort year 1991

1=School cohort consists of individuals born as early as 1991; 0=otherwise

Cohort year 1992

1=School cohort consists of individuals born as early as 1992; 0=otherwise

Cohort year 1993

1=School cohort consists of individuals born as early as 1993; 0=otherwise

Cohort year 1994

1=School cohort consists of individuals born as early as 1994; 0=otherwise

Cohort year 1995

1=School cohort consists of individuals born as early as 1995; 0=otherwise

Cohort year 1996

1=School cohort consists of individuals born as early as 1996; 0=otherwise

Socio-demographic characteristics Mum age

Age of mother (may be natural or step mother reported in the household)

Left home

1=Does not live with parents 0=Lives with at least one parent

Lives with partner

1= Lives with partner (in or away from parent’s home); 0=otherwise

Two parent household

1= Individuals mother and father live in the same household (may be step parent); 0=otherwise

Natural child

1= Parent of 1+ biological child/ren living in the same household

Siblings in household

1=Siblings under 16 living in same household; 0=otherwise

Migrant

1= First or second generation migrant; 0=3rd or higher generation migrant or non-migrant

Parent qualification, work status and income (own and parents) Working parent

1=At least one parent is employed (part-time or full-time); 0=otherwise

Partner & own income

Natural logarithm of own or joint own and partner income of individuals who have left parental home

Parent income

Natural logarithm of parental income of individuals living with parents

Parent degree

1= The highest qualification of at least one parent is a degree; 0=otherwise

23

Parent A-levels

1= The highest qualification of at least one parent is A-levels; 0=otherwise

Regional and wave controls Wave 3

1= Observed in third wave of Understanding Society; 0=Otherwise

Wave 4

1= Observed in fourth wave of Understanding Society; 0=Otherwise

Wave 5 England Outside London London

1= Observed in fifth wave of Understanding Society; 0=Otherwise

Wales

1= Lives in Wales; 0=Otherwise

Northern Ireland

1= Lives in Northern Ireland; 0=Otherwise

Scotland

1= Lives in Scotland; 0=Otherwise

1= Lives in England but not in London; 0=Otherwise 1= Lives in London; 0=Otherwise

24

Table A.3: Results for ALL, with full sets of controls

Ethnic Identity Mixed/ other Indian Pakistani / Bangladeshi Black Male Cohort 1988 Cohort 1989 Cohort 1990 Cohort 1991 Cohort 1992 Cohort 1994 Cohort 1995 Cohort 1996 Mum age Mum age squared Left home Lives with partner Two parent HH Own Children Siblings in HH Migrant Working parent Partner & own income Parental income

(1) GHQ OLS 0.519*** (0.156) -0.454 (0.388) 0.207 (0.451) -0.611 (0.407) 0.231 (0.451) 1.967*** (0.166) -0.622 (1.119) -0.078 (0.464) 0.029 (0.345) 0.162 (0.316) 0.471* (0.285) 0.551* (0.289) 0.512* (0.304) 0.053 (0.384) 0.009 (0.013) -0.001 (0.000) 0.672 (1.122) -0.614 (0.525) -0.174 (0.214) 0.361 (0.469) 0.597*** (0.164) -0.167 (0.299) 0.118 (0.269) -0.058 (0.092) 0.124

(2) GHQ IV 2.206 (1.422) -0.969 (0.602) -0.349 (0.650) -1.266* (0.692) -0.486 (0.764) 2.001*** (0.168) -0.517 (1.127) -0.184 (0.473) -0.017 (0.350) 0.123 (0.320) 0.460 (0.288) 0.512* (0.292) 0.408 (0.318) -0.004 (0.390) 0.009 (0.013) -0.001 (0.000) 0.897 (1.138) -0.567 (0.525) -0.244 (0.221) 0.340 (0.468) 0.560*** (0.166) -0.238 (0.302) 0.172 (0.272) -0.044 (0.093) 0.143 25

(3) FE -0.023 (0.018) 0.065* (0.037) 0.047 (0.047) 0.158*** (0.024) 0.033 (0.048) -0.021 (0.017)

(4) HE -0.018 (0.016) 0.123*** (0.035) 0.142*** (0.045) 0.214*** (0.033) 0.152*** (0.038) -0.037** (0.014)

-0.075* (0.039) 0.103*** (0.029) 0.170*** (0.034) 0.237*** (0.036) -0.002 (0.003) 0.000** (0.000) 0.720*** (0.171) 0.040 (0.104) -0.045** (0.022) -0.177 (0.117) -0.014 (0.018) -0.001 (0.030) 0.045* (0.026) -0.054** (0.024) 0.065***

-0.173*** (0.042) -0.091*** (0.029) -0.015 (0.024) -0.039* (0.021) 0.061** (0.028) 0.079* (0.046) 0.233 (0.154) -0.002** (0.001) 0.000*** (0.000) 0.825*** (0.106) -0.268*** (0.046) -0.021 (0.020) -0.309*** (0.046) -0.010 (0.016) 0.058** (0.023) 0.008 (0.024) -0.063*** (0.010) 0.030**

(0.130) -0.341 (0.213) 0.261 (0.252) -0.027 (0.276) -0.151 (0.330) 0.700** (0.281)

(0.132) -0.246 (0.226) 0.344 (0.260) -0.047 (0.277) -0.222 (0.336) 0.581* (0.303)

Wave controls

22.807*** (1.017) Yes

Observations

7151

Parent degree Parent A-levels London Wales Northern Ireland

(0.016) 0.067*** (0.022) 0.007 (0.024) 0.077*** (0.028) 0.054* (0.028) 0.098*** (0.024) -0.226*** (0.046)

(0.012) 0.198*** (0.018) 0.075*** (0.022) 0.049** (0.024) 0.009 (0.028) 0.167*** (0.029) -0.043 (0.026)

21.979*** (1.211) Yes

Yes

Yes

7151

1954

4226

Scotland Constant

Standard errors in parentheses Note: Ethnicity base: White Parental qualifications base: Below A-levels Living status base: Living at home with parents and no partner Cohort base: 1993 Region base: England non-London Wave base: 5 * p < 0.1, ** p < 0.05, *** p < 0.01

26