Perverse Fluidity?—Differential Impacts of Family Resources on Educational and Occupational Attainment for Young Adults from White and Ethnic Minority Heritages in England

Abstract

This study examines the intergenerational transmission of family resources (class, education and income) on people’s educational and occupational attainment in their early career life. It asks whether parental resources remain effective or fall into insignificance. It also asks whether the resources operate in a similar way for the ethnic minorities as for the majority. Drawing on data from the Longitudinal Study of Young Persons in England, the study focuses on resource transmission in degree attainment, access to elite class position, unemployment rates, labour market earnings, and continuous income. In each aspect, we test not only the net effects of parental resources, but also the differential transmission between the majority and ethnic minority groups. The analysis shows strong effects of parental resources on educational and occupational attainment for whites but rather weak effects for the ethnic minorities. Ethnic minority children tend to grow up in poor families, yet even those whose parents manage to achieve socio-economic parity with whites do not enjoy similar benefits. Reducing inequality in family socio-economic conditions and inequality in labour market opportunities is key to achieving social justice.

1. Introduction

The aim of this paper is to study the intergenerational transmission of family resources (parental class, education and income) on children’s educational and occupational attainment in their early career life, the relative strength of association between each dimension of parental resources and each domain of socio-economic attainment, and whether the transmission operates in a comparable way between whites and ethnic minority groups in England. Students of social mobility are keen to understand the generative mechanisms of social inequality, be such inequality manifested in terms of parental class, income, education or other ascriptive features such as gender or ethnicity. A great deal of research has been conducted in the specific areas and, in the more recent period, sustained efforts have been made to synthesize the different research traditions. Yet to our knowledge, no research is currently available that directly and simultaneously compares the effects of different dimensions of family resources on the different domains of people’s educational and labour market outcomes, nor between the majority and the ethnic minority groups in England. This paper seeks to make a contribution in this aspect.

The paper is structured as follows. In the next section, we give a brief account of the sociological analyses on mobility processes, extending from the earlier studies focusing on family class to the more recent multidimensional approach incorporating parental class, income, and educational resources, especially in the analysis of primary and secondary effects in educational attainment and transition processes, and parallel efforts in the ethnic studies tradition that accommodates insights from the social mobility research tradition. After that, we present data and analyses. The paper will conclude with some discussion.

2. Literature Review

Social mobility research seeks to understand the generative processes of social inequality and to promote social justice. Social inequalities associated with ascriptive factors such as family origin, gender, ethnicity and disability are deep-rooted. Social scientists have conducted voluminous research on patterns and trends of socio-economic inequality. In Britain, for instance, researchers have shown a high level of absolute mobility co-existing with a low level of relative mobility. Using surveys across three decades from the early 1970s to the early 2000s, Goldthorpe and Mills find that around 35–40 percent of men in Britain have better jobs than did their fathers and around 80 percent of men are in different class positions from their fathers’ (Goldthorpe and Mills 2008, p. 90). This suggests that absolute levels of social mobility were not in decline in Britain as were frequently claimed in the media or political circles (Payne 2017). Yet, when we look at relative mobility, or the chances of people from different origins trying to obtain advantaged and to avoid disadvantaged social positions, the picture is rather gloomy. Relative mobility rates are low, with odds ratios in the competitions between Classes 1 and 7 typically exhibiting themselves in the range of 10–20 (Bukodi and Goldthorpe 2019, p. 81) and that with little change over time and across space (Goldthorpe and Mills 2008, p. 92; Goldthorpe 1987; Erikson and Goldthorpe 1992, 2002). It is also of interest to note that the social structure of Britain and other developed countries is not particularly rigid, but more fluid than in some developing countries such as China (Li et al. 2015). Summarising the research findings, Bukodi and Goldthorpe (2019) suggest that mobility may have reached a plateau, with forces for and against mobility operating in a simultaneous way. Efforts to promote mobility require a scrutiny of the processes of social reproduction not only in terms of class as conventional analysis has done but also in terms of other, potentially more important, drivers such as parental economic and cultural resources. Another recent development in mobility research is the recognition that changes in the composition of the population has rendered the conventional mobility research rather inadequate, as the proportion of ethnic minority groups in the population increased from around 3 percent in the 1950s to around 20 percent in England in 2011. Combining insights from both perspectives, we need a comprehensive analysis to see whether processes of social reproduction operate in an equally effective way on different domains of socio-economic attainment, and whether such processes operate in a similar way for the different ethnic groups.

How do we explain the fact that we can simultaneously observe high rates of absolute mobility and low rates of relative mobility in both developed and developing countries? The answer is that the two phenomena are not necessarily in contradiction: the former are a consequence of rapid socio-economic development of a particular society such as the postwar development in Britain which saw the creation of large numbers of professional-managerial positions and which enabled many people to obtain higher positions than held by their parents, or the postreform development in China where around 280 million peasants left farm-work and migrated to cities in search of higher wages and a better life for themselves and their family. The reforms created an unprecedented upgrading of the country’s occupational structure and moved hundreds of millions of peasants on an upward journey, to become factory workers, technicians, shop assistants, entrepreneurs, and even professionals and managers (Li 2021b, 2022). Absolute mobility rates may vary from country to country, depending on economic conditions, historical circumstances, and social-cultural traditions. Relative mobility, or social fluidity, refers however to levels of class competition where people from different origin classes try to gain access to more advantaged positions in avoidance of more disadvantaged positions, a competition which may be ‘beyond political reach’ in societies characterised by a market economy, a nuclear family system and a liberal-democratic institution (Bukodi and Goldthorpe 2019, p. 221). Activities constitutive of family life may generate social inequalities even for people with similar levels of incomes. Typical examples include different parenting styles such as ‘concerted cultivation’ versus ‘natural accomplishment’ (Lareau 2003). This suggests that there are forms of family resources beyond income that can generate mobility differences.

Most studies on social mobility in educational and occupational attainment use family class (mainly father’s class) as origin resource. This is usually the only indicator of family position available in the data sources, although it has limitations. Class is the cornerstone in the employment relationship theory, serving as an indicator of economic resource closely associated with the level of current income, short-term income security and long-term income prospects (Goldthorpe and McKnight 2006). Yet, family resources are multi-dimensional, including social, economic, cultural, human and symbolic capitals (Bourdieu 1986; Coleman 1988; Putnam 2000; Lin 2001). To focus on class alone at the expense of other factors may be unduly limiting. However, to recognize the limitation is one thing, to overcome it is quite another. In most social surveys, even information on parental class is not always available, and that on parental income, education or wealth is rarer. Complicating this picture is the finding of declining parental class effects and the difficulty of complementing class with other resources. For instance, Goldthorpe and Jackson (2008: Figures 3.1 and 3.2) show that for both respondents in the 1958 National Child Development Study who turned 16 in 1974 and those in the 1985 Youth Cohort Study who turned 16 in 2001, people from professional and managerial (salariat) families have higher scores in GCSE examinations, which indicate strong primary effects as expected. Over and above the primary effects, salariat children are around 15–20 percentage points more likely to transit to A-level work than their working-class peers, demonstrating powerful secondary effects. For fear of experiencing downward mobility (Kahneman 2011), more advantaged families tend to practise a ‘strategy from above’ (Goldthorpe 2007) and use the superior resources in their possession to help their children stay longer in education. Yet, research has also shown that with the widening participation policies and growing government provision of educational opportunities, more children from all class backgrounds are staying longer in school, which will reduce the secondary effects as time goes on. Drawing on data from the Longitudinal Study of Young Persons in England where respondents turned 16 in 2006, Li (2021a) finds that, by this time, there is little secondary effect at the low and middle levels of GCSE scores on transition to A-level work. It is only at the high level of GCSE attainment that class differences emerge, and then for whites only.

Other evidence of weakening class effects on education is shown by Breen et al. (2009). Using pooled data from the General Household Survey (1973–1992) with a semi-cohort approach, the authors showed that class gaps in educational attainment were being narrowed for men from successive birth cohorts from 1908–1924 to 1955–1964. Similar findings were obtained for women (Breen et al. 2010). The authors attribute this to the reduction in class-based gaps of family resources and to the increasing government provision of educational services after the Second World War, which reduced family class effects on children’s education. It is noted here that Breen and colleagues’ findings on the overall class reduction in educational attainment is not incompatible with the findings of growing class differences at the higher levels of educational attainment as found by economists (Blanden et al. 2005; see also Li 2021a, note 1). It is the falling class effects at the lower level and the increasing class effects at the higher level of education which depict a true and comprehensive picture of educational attainment in contemporary British society, testifying to the validity of the maximum and effectively maintained inequalities theses as Raftery and Hout (1993) and Lucas (2001) have proposed.

The powerful persistency of class effects in explaining relative mobility over time and across space and the weakening secondary effects at transitioning stages pose a challenge to mobility research: what socio-economic resources possessed by advantaged families can more effectively help them to preserve intergenerational stability in class position? Do the different types of resources have equally persisting effects or do some types become less powerful when children become older? Do the resources have differential impacts in different domains of children’s educational and occupational attainment and for different social groups? As noted earlier, due to the lack of data on parental income or education, parental class is used ‘as a good proxy for permanent income’ (Erikson and Goldthorpe 2002, p. 34). When appropriate data become available, it would be advisable to measure family resources in a more systematic manner and apply them to empirical research. This is what Bukodi et al. (2021) did in their study of primary and secondary effects. In the paper, the authors measured four dimensions of family resources: class, income, education and status, and used them in analysis of five educational transitions: continuing from GCSE to A-level studies, applying to university, entering university, gaining access to an elite (Russell Group) university, and going on to postgraduate studies. On the basis of the Longitudinal Study of Young Persons in England (LSYPE), the authors found that whilst class and income played an important part in primary effects, it was education and, to some extent, status that were of greater importance in the secondary effects, especially at the higher levels of education. For instance, in the ‘cumulative total’ effects on transition to A-level studies, the secondary effect of family class was only around 5 percent and there was no effect of parental incomes. In the four other transitions, the secondary effects of class and income all disappeared. While status played limited roles in applying to and attending university, it was parental education that had notable effects in all five transitions. The authors concluded that the highly educated professional parents’ fear of downward mobility may be the greatest driver of intergenerational transmission of resources.

While the study by Bukodi et al. (2021) is the most systematic research currently available on primary and secondary effects, it did not cover ethnicity nor the labour market. This leaves us wondering whether parental class or income would still play notable roles in children’s occupational attainment once parental education is taken into account. Similarly, one may wonder whether the stratification system would be ethnic blind. In other words, do ethnic minorities with similar family resources enjoy the same kind of protection as do whites? Or will the system make ethnic minority children ‘equally disadvantaged’ regardless of their parental resources as suggested by the thesis of ‘ethnic penalty’ (Heath and McMahon 1997, p. 91)? Earlier US research found that African Americans were poor not because they were born poor but because they were born black. Even those from affluent professional families were still disadvantaged (Duncan 1968). Is this happening in England?

It is generally the case that ethnic minorities face disadvantages due to their weaker socio-economic-cultural resources. Yet, among the ethnic minority groupings there are marked differences, with some doing better than others, depending on the timing of arrival, warmth of reception upon arrival and the settlement processes, human-socio-economic capital possessed, and direct, indirect and ‘statistical’ discrimination in society (Borjas 1992; Portes and Zhou 1993; Heath and Cheung 2007). A litmus test of social and ethnic equality is whether the social mobility mechanism would operate in a similar way for the minority groups as for the majority. If those among the minority groups whose parents have managed to obtain resources similar to the majority group are persistently found to fall behind, resulting in a markedly more ‘fluid’ mobility profile, that ‘equality’ would not only be an injustice to the minorities but also a direct testimony to racial discrimination at the societal level. In that case, a more rigid stratification order, to a similar level of whites, would benefit the minority groupings more. This suggests that while social equality is a desirable goal to pursue from the social justice perspective, it cannot be achieved by placing injustice upon a particular social group, depriving their children of equal opportunities for educational and occupational benefits. That situation would, if found true, be akin to the US case earlier alluded to. The greater fluidity among the African American men was termed ‘perverse openness’ (Hout 1984, p. 1393). To the best of our knowledge, no similar study has been conducted for the ethnic minority groups in England.

There are studies comparing the intergenerational social mobility patterns between the majority and ethnic minority groups in Britain (Platt 2005; Heath and McMahon 2005; Li and Heath 2016). Yet, in these studies, family position was limited to class alone as the data used did not contain information on parental education or income. Another deficiency in the studies is that ethnicity is used in main-effects rather than interaction models. They show evidence of ethnic penalty but not that of differential protection or perverse fluidity, namely, whether a particular ethnic minority group would fare better or worse than the majority group at the same level of parental resources in a particular kind of competition. A third deficiency was that the analyses were usually confined to one outcome variable—respondents’ class destination, masking the possible differential impacts of different family resources in the different domains of outcome for different groups. The present study wishes to make good these deficiencies. In so doing, it seeks to address the two main questions:

• Do parental class, income and educational resources continue to play notable roles in the educational and occupational attainments in people’s careers as young adults, or have some of the resources paled into insignificance?
• Do patterns of intergenerational mobility exhibit themselves in similar ways for ethnic minority groups as for whites for each dimension of parental resources and in each domain of educational and occupational attainment? In other words, is there a highly stratified order for whites but a relative flat pattern or ‘perverse fluidity’ for the ethnic minorities?

3. Data and Methods

To address these questions, this study used data from the Longitudinal Study of Young People in England (LSYPE). The survey represents all the young people aged 14 and resident in England who were attending maintained schools, independent schools and pupil referral units (PRU) in February 2004. It adopted a multistage stratified sampling design, with the first stage at the school level and the second stage at the pupil level. Some 838 maintained schools, 52 independent schools and two PRUs were sampled. Within each school, an average of 33 students were sampled. The response rates were 73 and 74 percent at the school and student levels, respectively. The survey followed the young persons’ lives through seven waves annually until 2010, and then again when they were aged 25, in 2015. The initial sample size was 15,770 but at wave 4, a boost sample of 352 respondents was added, mainly from ethnic minority backgrounds, with a total size of 16,122. As with other panel studies, the survey has suffered sample attritions, with only 7707 respondents found in Wave 8.

As the present analysis focuses on the ethnic differences in intergenerational social mobility, we compare the strengths of association between origins and destinations for the majority and the main ethnic minority groups. We used data on parental resources (class, education and income) from waves 1, 2 and 4, ethnicity from waves 1 and 4, and outcome variables (education and labour market positions) from wave 8. The parental resource variables used were similar to those in Bukodi et al. (2021), except for parental status which was not used in the present study. Status has a high correlation with parental class and education, at 0.71 and 0.57, respectively (Zhao and Bukodi 2020). Status in Bukodi et al. (2021, p. 629) is used as an ‘indicator of socio-cultural resources’ such as ‘forms of social participation’, but there is evidence that class is highly connected with the range, position, reachability and diversity of social ties in Britain (Li et al. 2008). If one wishes to find an effective indicator of social connection and social-cultural participation, it might be more efficient to use class directly. Class is also closely related to formal civic engagement, and bonding and bridging social capital (Hall 1999; Putnam 2000; Li et al. 2005). A methodological note is also in order here. As the present study is focused on differences in parental resource effects between the main ethnic minority groups and whites, the sample sizes for the ethnic minority groups posed a constraint on the number of independent variables and of categories of independent variables that could be used in statistical models, especially when we wish to see the impacts both within and between models. This requires a careful balancing between substantive and methodological considerations.

With regard to parental class, we used the National Statistics Socio-economic classification (NS-SEC) for the young persons’ fathers and mothers. In the first wave, 14,257 mothers and 10,166 fathers reported their NS-SEC class. We used the ‘dominance approach’ (Erikson 1984) in assigning parental class position. For missing information on parental class at wave 1, we supplemented it from information at wave 2 using a similar approach. Furthermore, parental class in wave 4′s boost sample was added. Finally, 15,638 respondents had valid parental class information combining data from waves 1, 2 and 4. We differentiated four parental classes: higher-grade professional-managerial salariat (Class 1), lower-grade professional-managerial salariat (Class 2), the intermediate class (Classes 3–5) and the working class (Classes 6–8). For parental education, the same approach was adopted, using the higher of either parent’s education from wave 1, supplemented from wave 2 if wave 1 information was missing, and then further supplemented for the wave 4 boost sample. A total of 15,741 respondents had valid parental education information in waves 1, 2 and 4. We coded a four-way variable on parental education: first degree or above; sub-degree (teaching, nursing, etc., below degree); A-level or equivalent; GCSE or below.

Parental income data were collected on main parent and partner’s total incomes from work, benefits and other sources with 32 bands for the lower income families and a further 60 bands for the higher income families. A mid-point value was taken in constructing the variable, supplemented by the income of those who were self-employed. For families who did not report valid income data in wave 1, their income information from wave 2 was used. This information was then complemented by wave 4′s boost sample, resulting in 13,791 cases with valid information on parental income. A quartile approach was adopted for analysis in the present paper. As ethnic minority groups, particularly Pakistani and Bangladeshi groups tend to have larger family sizes (the mean number of family members being 5.9 and 6.2 for the two groups, respectively, relative to 4.2 for whites), we take this into account by constructing a standardised mean household income by dividing the total household income by the number of household members.

Ethnicity is of key interest in this paper. A total of 16,094 respondents reported valid information on ethnicity from waves 1 and 4 but as 52.2% of the samples (N = 8415) had dropped out by wave 8, we had to rely on a sample (N = 6336) with full information of parental class, education, income and ethnicity in our analysis. In some of the analysis below, the sample sizes were further reduced such as when we analysed unemployment risks. By definition, unemployment pertains to the economically active, which leaves out those who are long-term sick, in education or looking after the home.

Our outcome variables pertain to educational and occupational attainment when our respondents were at age 25. We focused on the highest educational qualifications obtained, employment status, class position, labour market earnings and continuous (take-home) income. In each of these domains, we tried to see whether the parental resources would still be important, or whether some of the resources had lost their significance and, if they were still found as important, whether they were equally effective for whites and the ethnic minority groups.

Our modelling strategy is straightforward, as can be expressed with the following equation

Y^R = a + b₁Y^P + b₂Y^P*E^R + b₃S^R + e

where Y^R represents the respondent’s educational and occupational position at wave 8, E^R represents the respondent’s ethnic status, and Y^P represents the different dimensions of parental resources: class, education and income as discussed above. Y^P*E^R is an interaction term between ethnicity and parental resources, and S^R represents respondents’ sex which is not our main concern in the present study but is used as a control variable in the modelling exercises. B₁ is expected to be positive, reflecting the usual association between origins and destinations. Given the findings in existing research on second-generation achievement (Platt 2005; Heath and McMahon 2005), we may expect b₂ to be negative, reflecting possible societal discrimination against members of ethnic minority heritages or lower returns to their parents’ socio-economic resources, corresponding to a higher level of fluidity or a lower level of class reproduction for the ethnic minority groups than for the majority group.

The LSYPE data contains the main ‘visible’ ethnic minority groups but when it comes to wave 8, the sample sizes for some groups become rather small. For instance, the number of Chinese with valid parental information is too small at wave 8 to be included in statistical analysis (N = 18). Regrettably, they have to be dropped together with the ‘mixed’ and ‘other’ groups who are too small in number and too diverse to be meaningful. Readers interested in the educational and labour market outcomes of Chinese and some other small groups can find useful information in Crawford and Greeves (2015), Li and Heath (2008, 2016, 2018) and Li (2021a). We included white, Black Caribbean, Black African, Indian, Pakistani and Bangladeshi groups in our analysis.

4. Analysis

We started our analysis by looking at profiles of the educational and labour market positions of our respondents at age 25, focusing on degree attainment, rates of unemployment, labour market earnings for those in work, and the continuous weekly take-home pay. As can be seen in

Table 1. Selected characteristics of educational attainment and labour market position at age 25.

	White	B Car	B Afr	Indian	Pak	Bang	All
Degree+ (%)	25	27	50	47	34	29	26
Unemployment (%)	7	8	12	10	12	9	7
Gross weekly earnings (GBP)	353	280	327	361	274	348	350
Continuous weekly income (GBP)	314	229	230	246	219	213	306
Approx. N	4850	208	232	430	326	271	6317

Notes: B Car = Black Caribbean, B Afr = Black African, Pak = Pakistani, Bang = Bangladeshi. People of Chinese, mixed, other heritages are omitted. Degree+ = having first degree or above. Weighted analysis and unweighted Ns, with weights taken from wave 8 as provided by the data collectors (the same below). Source: the Longitudinal Study of Young People in England (LSYPE) (the same below).

, ethnic minorities outperformed whites in degree-level education but fared less well in terms of avoiding unemployment. One in four white respondents had degree-level qualifications but those from Black African and Indian heritages were almost twice as likely to have degree-level qualifications. Even Pakistani, Bangladeshi and Black Caribbean groups who used to lag behind have now caught up with and surpassed whites. However, when it came to unemployment and earnings, we could see that ethnic minorities were still faring worse, with those from Black African, Pakistani and Indian origins having a markedly higher likelihood of being unemployed (10–12 percent as against 7 percent for whites) in spite of their greater educational success. It may be that whites are more likely to attend elite universities than ethnic minority groups but this is not always the case. Li (2021a) shows that 9.6 percent of whites attended Russell Group (RG) universities but Black African and Pakistani groups were not far behind, at 7.1 and 8.1 percent, respectively, and those of Indian origins were actually much more likely to attend an RG university, at 12.7 percent.

Given the overall educational success and the labour market setback of the ethnic minority groups, we would wish to see their parental class, education and income situations. The data in

Table 2. Descriptive analysis of family origin characteristics (class, education and income) by ethnicity (column percentage within each panel).

	White	B Car	B Afr	Indian	Pak	Bang	All
Class
Higher salariat	19	9	10	14	7	1	18
Lower salariat	35	37	39	23	17	9	34
Intermediate	28	26	20	38	28	24	28
Working class	18	28	32	25	48	66	20
Education
Degree+	17	15	27	17	10	5	16
Sub-degree	16	22	14	9	5	2	16
A-Level	19	13	9	14	11	2	18
O-Level or below	49	50	50	60	74	90	50
Income
Top quartile	28	22	14	17	3	2	26
2nd	28	24	23	21	9	3	27
3rd	24	25	24	28	29	23	25
Bottom quartile	19	30	39	34	60	72	22

show a remarkable degree of socio-economic and cultural disadvantages of the ethnic minority groups, particularly for Pakistanis and Bangladeshis. Around two thirds of Bangladeshi respondents were from working-class families, with 90 percent having poorly educated parents and nearly three fourths coming from families situated at the bottom quartile of incomes. Taking these factors in combination, it would be fair to say that they were the most disadvantaged group, followed by Pakistanis and Black Africans. Yet their early-career outcomes as shown above suggest that they have done much better than expected.

As most mobility researchers would expect, parental class has a positive association with children’s career outcomes.

Table 3. Selected characteristics of educational attainment and labour position by family conditions.

	% Degree+	% Unemployed	GBP Gross Weekly Pay	GBP Weekly Income
Class
Higher salariat	44	3	439	352
Lower salariat	31	5	354	329
Intermediate	19	7	327	300
Working class	12	16	270	234
Education
Degree+	52	5	429	343
Sub-degree	33	5	356	333
A-Level	24	6	345	322
O-Level or below	16	10	319	280
Income
Top quartile	41	4	419	345
2nd	26	6	334	321
3rd	17	10	309	284
Bottom quartile	20	11	314	264

shows a strong and clear gradient between parental class position and each of the four outcome variables. For instance, 44 percent of our respondents from higher (professional-managerial) salariat families had obtained degree-level qualifications in sharp contrast with a mere 12 percent of those from working-class families, or a disparity ratio of 3.7. The differences in unemployment risks were equally marked, at 3 and 16 percent respectively, or a disparity ratio of 5.3. Working-class children on average had GBP 118 less in terms of continuous take-home pay than their high-salariat peers. Even if they are fortunate enough to be in paid employment, they would still be making GBP 169 less than their higher salariat peers in terms of gross weekly pay. Interestingly, when we looked at the patterns for parental education and incomes, we found clear gradients too, and with a similar salience. All this was as expected given the close association between the three dimensions of parental resources. Our next question was: when all three types of parental resources are entered into models, will they still have marked effects or will some effects weaken or disappear?

Having looked at the overall profiles of ethnic and parental resource differences in some detail, we now come to the core of our research: will all three kinds of parental resources continue to have salient effects when placed under simultaneous investigation? Will ethnic minorities share the pattern of intergenerational transmission of advantages as do whites? We analysed each domain of our outcome variables in terms of gross and net effects.

The descriptive data on educational distribution by ethnicity are shown in

Table 4. Educational distribution by ethnicity (percentage by row).

	Degree+	Sub-Degree	A-Level	O-Level	Primary/None	(N)
White	25	7	16	27	25	4860
B Caribbean	26	11	16	16	30	208
B African	50	8	15	14	12	234
Indian	47	4	26	13	10	430
Pakistani	34	4	16	23	23	326
Bangladeshi	29	5	26	24	16	272
(All)	26	7	17	26	24	6330

(data under column 2 on degree attainment are the same as in Table 1) and the modelling data in

Table 5. Average marginal effects from ordinal logit regression on educational qualifications at age 25 on parental resources and personal attributes with predicted probabilities on degree attainment.

	White	B Car	B Afr	Indian	Pak	Bang	All EM
Class
H salariat	0.208 ***	0.270 *	−0.036	0.267 *	0.298 *	0.618 ***	0.235 ***
L salariat	0.153 ***	0.160 *	−0.017	0.074	0.253 *	0.131	0.129 **
Intermediate	0.081 ***	0.142	−0.114	−0.054	0.023	0.066	0.040
Education
Degree+	0.229 ***	0.070	0.160	0.156	0.119	0.289	0.129 *
Sub-degree	0.099 ***	−0.039	0.320 **	0.032	−0.072	0.285	0.070
A-Level	0.050 ***	−0.022	−0.050	0.020	0.229 *	−0.012	0.058
Income
Top quartile	0.109 ***	0.026	−0.050	0.034	0.029	−0.093	0.001
2nd	0.043 **	0.034	0.017	0.179 *	0.385 **	−0.038	0.124 *
3rd	0.012	0.004	0.071	0.053	0.108	−0.110 *	0.038
Female	0.056 ***	0.190 ***	−0.025	0.111 *	0.066	0.156 **	0.084 **
N	4845	207	232	430	326	271	1466

Notes: Reference groups are families in working-class position, with O-level or lower qualifications, in the bottom quartile of incomes, and male. All EM refers to all the ethnic minority groups used in this analysis. Emboldened figures for ethnic minority groups indicate significant differences from the corresponding figures for white respondents. *** p < 0.001; ** p < 0.01; and * p < 0.05. The same in the following modelling tables.

. Black Caribbean, White and Pakistani respondents have poorer educational outcomes, with 30, 25 and 23 percent having only ‘primary’ level of education or no formal qualifications at all, in sharp contrast to 12 and 10 percent for Black African and Indian respondents. A comparison with data in Table 3 suggests that the poorest achievers are not from the most disadvantaged family backgrounds. Pakistanis, for instance, have better family circumstances than Bangladeshis and the two groups have similar cultural backgrounds. Socioeconomic or cultural factors do not seem to offer compelling explanations for their educational discrepancies, which may be due to geographic location. Over half (55 percent) of Bangladeshis live in London, as compared with only 15 percent of Pakistanis. Perhaps it is the better educational resources in London’s schools which enabled Bangladeshi children to reap the benefits, but it may be safer not to read too much into the relatively small amount of difference (7 percentage points).

To capture the net effects of parental resources on our respondent’s educational attainment and to see whether the resources would operate equally effectively for the different ethnic groups, we conducted an ordinal logit regress model. As the coefficients from such models are not intuitive, we present the net predicted probabilities of having degree-level education from the models. Table 5 shows the results. The data for each ethnic group are separately presented, which is essentially the same as in interaction models. The coefficients are average marginal effects. Significant differences between a certain category and the reference category of an independent variable are shown in asterisks. We also compared the coefficients across models, namely, between each of the ethnic minority groups and whites, with significant differences shown in bold. For example, the data under column 2 for whites suggest that the net effects of parental class, education and income are all highly significant, with parental class and education having similar magnitude, both of greater salience than parental income. Even after controlling for parental class and education, we still find parental income highly significant, with those from top quartiles having a lead of 11 percentage points over those from the bottom quartile to have a degree-level education. In the same vein, gender differences are significant. All things being equal, women are around 6 percentage points ahead of men in having a degree.

Looking more closely at Table 5, we found some notable features. First, in contrast to the majority group, the parental resources for the ethnic minority groups in terms of class, education and income are rarely found to be significant and the magnitude is much weaker too. While the sample sizes might be a reason, they were not too small. In addition, the pattern applied to the last column when all ethnic minority groups under investigation were pooled together with a combined sample size of 1466. The patterns here give a strong indication that parental resources have a weaker protective role for the ethnic minorities than for the majority. Secondly, while the strength of association on parental resources is weaker for the ethnic minorities than for whites, the tests do not, in most of the cases concerned, show significant differences from whites. Hence there is no decisive evidence of perverse fluidity. Only 11 out of 60 coefficients for the ethnic minorities relative to whites are significant. Thirdly, if there is a systematic disadvantage for any ethnic minority group, it is the Black Africans whose parental classes show no significant differences, but if we compare these with whites’ parental classes, we find that the effects are all significantly weaker. For instance, for people from higher-salariat origins, whites would have a net lead of 24.8 percentage points in having a degree than their peers from Black African origins. Similarly, for those from lower salariat and intermediate class families, the net disadvantages are 17 and 19.5 percentage points. Our previous analysis (Table 2) showed that Black African parents were much more likely to have a degree than white parents (27 versus 17 percent) but less likely to be in higher salariat position (10 versus 19), or in top quartile family income (14 versus 28 percent). First-generation Black Africans are known to have high educational qualifications, many being ‘students who stayed’ (Daley 1996) but having difficulty in achieving class and income positions commensurate with their education. Here, we found that their children are much behind their white peers even when they have the same level of parental class.

Turning to labour market position, we looked at class (differentiating higher and lower salariat, intermediate and working classes) and unemployment among those who are economically active. The descriptive analysis in

Table 6. Distribution of labour market position by ethnicity for the economically active (percentage by row).

	Higher Salariat	Lower Salariat	Intermediate	Working-Class	Unemployed	(N)
White	10	29	31	23	7	4243
B Caribbean	4	25	34	29	8	181
B African	8	39	29	11	13	194
Indian	18	31	29	11	10	393
Pakistani	14	28	30	16	12	274
Bangladeshi	11	30	33	16	9	223
(All)	11	29	31	22	7	5508

shows that white respondents are least likely to be unemployed. Yet, whites are not most likely to be found in higher-salariat positions. Here Indians are most advantaged, with Black Caribbean and Black African respondents being least likely to be found to be in higher-salariat position.

In

Table 7. Average marginal effects from ordinal logit regression on labour market position at age 25 on parental resources and personal attributes with predicted probabilities on access to the higher salariat.

	White	B Car	B Afr	Indian	Pak	Bang	All EM
Class
Higher salariat	0.093 ***	0.019	0.173	0.059	0.053	0.058	0.076 *
Lower salariat	0.063 ***	0.024	0.036	−0.008	−0.007	0.119	0.025
Intermediate	0.037 ***	0.012	0.058 *	0.029	−0.017	0.008	0.030 *
Education
Degree+	0.065 ***	0.039	0.000	−0.047	0.136	0.262	0.026
Sub-degree	0.022 **	0.012	0.003	0.043	0.026	0.018	0.001
A-Level	0.007	0.004	0.084	0.036	0.106 *	0.007	0.046 *
Income
Top quartile	0.040 ***	0.019	0.055	0.130	0.146	0.441	0.057 *
2nd	0.011	0.035	−0.002	0.024	0.096	0.069	0.021
3rd	0.003	0.023	0.005	−0.003	0.042	−0.027	0.017
Female	0.013 *	0.011	0.024	0.055	0.018	0.005	0.025 *
N	4232	181	192	393	274	223	1263

, we show the predicted probabilities of gaining access to the higher salariat for different ethnic groups controlling for parental resources and gender. We conduct within and between model analyses as in Table 5. The overall patterns are similar to those for degree-level education. For the white respondents, we still find that parental class, education and income exerting significant effects, with parental class effects being more prominent than for parental education, which again are more salient than for parental income. For the ethnic minority groups, very few of the parental resource effects are significant, again attesting to the muted parental origin effects. When we looked at the results of cross-model comparisons, we only found six out of sixty tests being significant. This again suggests that ethnic minority social mobility patterns are more fluid than those for whites, but not perversely so.

Having looked at education and class positions, we then turned to labour market earnings. Data were limited to those who were in paid employment. As mentioned earlier, our respondents are relatively young and some of them are still in education whereas others may be actively involved in home-making (young mothers). Not all forms of economic inactivity represent disadvantage. In view of this, we use the Heckman selection model, estimating firstly the probability of being in paid employment (avoidance of worklessness) and, for those in paid employment, the amounts of weekly pay they receive. As the selection model needs at least one ‘identifying variable’ not included in the regression model, we used limiting long-term illness (LLTILL) as such. For ease of exposition, the logit coefficients in the selection part were turned into proportions. The results are shown in

Table 8. Average marginal effects (AME) on avoidance of worklessness (%) and gross weekly pay (GBP) conditional on employment by family situation and ethnicity.

	White	B Car	B Afr	Indian	Pak	Bang	All EM
Gross weekly pay
Class
Higher salariat	135.56 ***	60.73	168.09 *	−57.45	155.14 *	−155.38	62.82
Lower salariat	71.63 ***	33.21	124.25 *	16.65	76.66	16.96	47.37
Intermediate	58.08 ***	18.26	89.93	24.71	57.80	−52.12	34.93
Education
Degree+	27.59	−26.09	44.02	54.54	20.82	188.32	53.31 *
Sub-degree	−0.16	−26.65	−38.58	3.57	156.60 **	182.36 *	−5.48
A-Level	2.59	9.87	−6.67	36.13	−8.59	69.67	7.69
Income
Top quartile	67.44 ***	5.50	−63.02	148.48 **	−47.47	−217.85	7.06
2nd	−1.85	29.58	−91.43	62.62	60.23	−43.52	5.06
3rd	−4.59	36.57	−98.72 *	29.84	−17.15	−98.26 *	−19.57
Female	−99.55 ***	34.95	33.07	−3.25	−42.95	−129.24 *	−22.19
Avoiding worklessness
Class
Higher salariat	0.16 ***	0.18	0.03	0.13	0.04	0.40 ***	0.07
Lower salariat	0.15 ***	0.03	−0.15	0.07	0.02	−0.06	−0.02
Intermediate	0.11 ***	−0.04	0.02	−0.07	0.06	−0.02	−0.01
Education
Degree+	0.03	0.13	0.09	−0.09	−0.08	0.33 ***	0.02
Sub-degree	−0.00	0.17	0.13	−0.15	−0.14	0.08	0.03
A-Level	0.03	0.02	−0.07	0.02	0.09	0.13	0.04
Income
Top quartile	0.07 **	0.18	0.17	0.07	0.44 ***	−0.48 **	0.14 *
2nd	0.06 *	0.09	0.21	0.17 *	0.13	−0.14	0.13 *
3rd	0.02	0.02	0.20 *	0.05	0.17 *	−0.10	0.09 *
Female	−0.01	−0.09	0.05	0.11 *	0.03	−0.12	0.01
LLTILL	−0.16 ***	0.02	−0.08	−0.04	−0.08	−0.17	−0.05
N	4757	202	225	414	317	260	1418

Notes: LLTILL = limiting long-term illness.

.

Looking firstly in the selection part of Table 8, we could see clear parental class and, to a smaller extent, income effects. White people from salariat families are around 15–16 percentage points more likely than their counterparts from working-class families to avoid worklessness, and those from rich families are also less likely to be workless by around 6–7 percentage points than those from the bottom quartile of incomes. It is interesting to note that parental education plays no noticeable role net of class and income. The findings in this regard suggest that, at least for whites, parental class may be a very good indicator of the breadth and depth of social connections, the possession of weak and strong ties, and of bridging and linking social capital (Granovetter 1973; Lin et al. 1981) which, when supplemented by economic capital, are of indisputable importance in helping children avoid worklessness. Parental resources among the ethnic minority groups are again weak and generally non-significant.

Turning to the regression part of Table 8, we find, for those who have surmounted the barriers of worklessness and gained entry to paid employment, a similar pattern to that on avoidance of worklessness, namely, that parental resources in terms of class and income exert a strong and powerful impact on children’s weekly earnings but parental education has no salient effect. White respondents from the three higher classes of families earn GBP 136, 72 and 58 more than their working-class peers, and those from the top-quartile income families earn GBP 67 more, all showing highly significant differences. Other things being equal, white women earn GBP 100 less than white men. Turning to coefficients for ethnic minority groups, we again found generally weak associations between parental resources and their own earnings, with most parental resource categories being non-significant. Bangladeshis from higher salariat and high income families are found to earn significantly less than their white peers, at GBP 291 and 285, respectively. Our previous analysis (Table 2) showed that Bangladeshis have the poorest family conditions, being least likely to come from higher salariat and top-income families. As the sample size involved is rather small, it might be better not to speculate too much here.

As indicated earlier, not all respondents at the young age of 25 are in paid employment but everyone has to have some form of income, which is called ‘continuous income’ from all sources. It is also of interest to see the parental resource and ethnicity effects on such income, which we analyse next with findings shown in

Table 9. OLS regression of weakly take-home income (GBP) by family situation and ethnicity.

	White	B Car	B Afr	Indian	Pak	Bang	All EM
Class
Higher salariat	85.00 ***	76.08 ***	40.91 **	48.99 ***	59.89 ***	53.91 ***	59.38 ***
Lower salariat	72.97 ***	52.84 ***	24.42 *	46.75 ***	51.52 ***	63.40 ***	43.09 ***
Intermediate	55.66 ***	35.34 ***	17.95 *	36.16 ***	34.75 ***	34.18 ***	34.12 ***
Education
Degree+	18.01 ***	4.17	18.71	17.39 *	8.01	36.18 ***	11.62 *
Sub-degree	15.39 ***	2.57	13.19	18.46 **	11.80	16.39	7.85
A-Level	13.60 ***	11.06	6.71	23.32 ***	24.24 ***	18.61	18.14 ***
Income
Top quartile	33.28 ***	13.33	14.34	27.34 ***	17.16	−8.83	17.60 ***
2nd	25.03 ***	18.61 *	0.58	22.62 ***	23.74 **	7.87	15.06 ***
3rd	8.08 **	7.06	7.31	7.52	15.79 **	3.06	9.04 ***
Female	34.18 ***	26.71 ***	19.83 **	18.78 ***	15.56 ***	13.46 *	17.41 ***
Constant	213.46 ***	166.31 ***	191.75 ***	187.99 ***	177.38 ***	188.13 ***	183.06 ***
N	4850	208	232	430	326	271	1467

.

With regard to findings in Table 9, the most striking feature is the continued salience of all three types of parental resources for whites in the order of parental class, income and education. For the first time in our analyses, we see pervasively significant parental class effects on people’s continuous income (take-home pay) for all ethnic minority groups, and significantly weaker effects for minority than for the majority respondents at each level of parental classes. Apart from Indians, parental income and education are not of notable importance for children’s continuous incomes for the other ethnic minority groups. Other things being equal, Black Africans from higher-salariat families, and Black Caribbeans from highly educated and top-income families earn half or less than their white peers.

5. Discussion and Conclusions

This paper has sought to contribute to scholarship on social mobility in England in two important ways. Drawing insights from recent studies by Li and Heath (2016), Li (2021a) and Bukodi et al. (2021), the present study has combined the social mobility and the ethnic studies traditions in analysing the effects of different dimensions of parental socio-economic-cultural resources on a range of respondents’ educational and occupational outcomes and comparing the effects of each of these dimensions on each outcome domain between the majority and the ethnic minority groups. The data source used is currently the only one that contains good quality data on parental class, income and education as well as ethnicity with a range of attainment indicators for the respondents.

Two main research questions guided our analysis: first, do different dimensions of parental resources (in terms of class, income and education) still play significant roles in our respondents’ educational and occupational attainment in their early careers? Second, are there significant differences between the majority and the minority groups in each link of the social mobility chains? In the latter regard, we were seeking to address the question of whether, for members of ethnic minority heritages, parents’ resources would play a much weaker role as to constitute a perverse fluidity. Put differently, we were exploring the possibility of two separate mobility scenarios in co-existence, with one having a highly stratified mobility structure where parental resources play a decisive role in determining children’s educational and occupational success, and the other where, regardless of levels of parental resources, children’s attainments in education and different domains of labour market position are depressed, creating a specious milieu of equality or ‘perverse fluidity’. In the present context, we would expect that the highly stratified mobility mechanisms would mostly work for the majority group whereas the perverse fluidity would represent the mobility regimes of the ethnic minority groups given the extensive research on disadvantages or ‘ethnic penalty’ faced by members of ethnic minority heritages (Heath and McMahon 2005; Heath and Di Stasio 2019; Platt 2005; Li 2018a, 2018b; Li and Heath 2008).

To address such questions, we used more indicators of parental resources than found in most mobility studies including parental class, income and education, all as categorical variables rather than putting the latter two into quartiles and then standardised as continuous variables as carried out by Bukodi et al. (2021). We believe that our approach is more capable of detecting the effects of parental resources, as amply demonstrated in our analysis. We also used more outcome variables as destination indicators than found in other studies. The combination of these make our study more comprehensive than hitherto available. More importantly, we tested the strengths of association between each aspect of parental resources and each domain of our outcome variables for each of the ethnic minority groups in a rigorous way to establish the presence or otherwise of ‘perverse fluidity’. All this makes our study the most systematic currently available on socio-ethnic mobility in England.

Our main findings can be summarised as follows. Firstly, at the overall level, we found (Table 1) that for our respondents at the relatively young age of 25, the whites are least likely to have degree-level of education, at 25 percent, almost half for Black Africans and Indians (at 50 and 47 percent, respectively) but they are also least likely to face unemployment, at 7 percent, which is also almost half as compared with those for Black Africans and Pakistanis (at 12 percent each). For people in paid employment, whites earn more, and they have the highest continuous (take-home) incomes from all sources. These are gross differences and do not include mobility processes or differential effects in resource transmission between different ethnic groups.

Ethnic minorities are in general disadvantaged, especially in terms of parental class and income but, apart from Pakistanis and Bangladeshis, they do not seem to be too much behind whites in terms of parental education (Table 2), although their overseas qualifications may be devalued in the British labour market. In fact, Black African parents have much higher education than, and Indian parents have similar levels of education to, white parents. We would expect people with superior parental resources to enjoy greater benefits as most mobility studies have shown, but the recent findings of parental education eclipsing class and income effects as shown in Bukodi et al. (2021) have reinstated the need to investigate the net effects of different types of origin resources. At the same time, we wish to see whether there is a differential transmission of parental resources depending on ethnicity, with whites enjoying more benefits than do similarly resourced members from ethnic minority backgrounds.

Our analysis shows that parental resources do have pronounced effects on children’s educational and occupational attainment at the gross level (without controlling for other effects) as shown in Table 3, which gives us some reassurance. Yet, these are not net effects and do not show ethnic differentials. Analyses presented in Table 5, Table 7, Table 8 and Table 9 are geared to these purposes. The overall impression is that, for whites, parental resources of all three types continue to have pronounced significance in each of the outcome domains under investigation, be they degree-level attainment, access to elite position, avoidance of worklessness and amount of weekly earnings, or continuous income from all sources, even though the magnitudes vary depending on the outcomes being assessed. Parental class and education assume similar effects in children’s education, both of which are of greater salience than parental income. For access to the higher salariat, parental class assumes the greatest importance, followed by education and then income. Parental class and, to a smaller extent, income have notable impacts on children’s avoidance of worklessness and labour market earnings but parental education has no net effects. A basically similar pattern is found for continuous incomes although, here, parental education has significant net effects.

Yet, it is also the case that in most of the analyses covered, with the sole exception of parental class impact on continuous incomes, parental resources are generally non-significant or only very weakly so for the ethnic minority groups. This suggests that parental resources have a highly stratified order for the majority but a rather limited impact for the minority groups. In other words, social mobility patterns are much more ‘fluid’ for the minority groups than for the majority. Is this fluidity so severe as to constitute an overwhelming ‘perverse fluidity’? Our detailed analyses suggest a negative answer. Out of 315 tests conducted, only 56 are significant, with 26 in Table 5, Table 7 and Table 8, and 30 in Table 9. All this suggests that ethnic parental resources play a much smaller role as compared with the majority.

Our analysis also has various limitations. Some ethnic minority groups are not included in the analysis due to small sample sizes. The data are restricted to England and we do not know whether the mobility profiles found in England would be similarly shown in the other nations within the UK or other developed countries (D’Addio 2007). Even though the initial sample was large, the attrition over the successive waves has reduced the sample to around a half at wave 8, making it hard to include some potentially important covariates in the modelling exercises. Hopefully, these and other shortcomings could be overcome with new and more powerful data.

In sum, the second-generation ethnic minority groups in England are faced with various disadvantages, with fewer parental resources and weaker protections even at similar levels of parental resources, but we also find evidence of remarkable determination, aspiration and resilience, or what might be termed ‘ethnic capital’ (Modood 2005), as exemplified by striking success rates of some of the most disadvantaged groups such as Bangladeshis who lived through dire family poverty but managed to make achievements in education, class and earnings parallel to the majority (see also Li 2018a, 2018b, 2021a; Li and Heath 2010, 2018). Social progress is in train but barriers to social equality need to be further reduced. It is incumbent upon the government, employers and wider society to work harder to eliminate all forms of socio-ethno disadvantages.