Born in Britain: a reflection on government aspirations a decade on

Abstract

In light of substantial disadvantage faced by ethnic minorities, the UK government stated a decade ago that in 10 years’ time, ethnic minorities should no longer face disproportionate barriers to labour market achievement. From the investigation of the stock of native born ethnics conducted here, it is evident that such aspirations have not been realized.

I. Introduction

As in the USA, there is a long history of research into the employment and earnings positions of minority ethnic groups in Britain.¹ Unlike in the USA, though, the evolution of British ethnic minority groups is a more recent phenomenon, as many of their families come as immigrants but not until the 1950s. This has meant that it has not always been easy to isolate the effects of labour market disadvantage from the effects attributable to issues such as culture, religion and a poor command of the English language. To this end, Blackaby et al. (2005)  –  hereafter BLMO  –  analysed the position of British ethnic minority groups (relative to the White majority) who were born in Britain and would thus be less affected by cultural and language barriers as there would be less need for their assimilation in the same way as there had been for their (potentially) immigrant forbearers. Nevertheless, benchmark results by BLMO found that over the course of the 1990s, such native-born ethnic minorities still faced substantial unexplained labour market disadvantage and the study concluded with reference to a Government report (Cabinet Office, 2003) that endorsed the words of the then British Prime Minister stating ‘that in ten years time, ethnic minority groups should no longer face disproportionate barriers to accessing and realising opportunities for achievement in the labour market’. And so, a decade after these thoughts would have been formulated, have these aspirations been realized?

II. Data and Methodology

To investigate the current position of the stock of male native-born ethnic minorities, data are taken from the UK Labour Force Survey pooled over 60 consecutive quarters from 1997 to 2011, with the starting date representing the beginning of an unprecedented period of continuous economic growth in the economy. Using this data, employment and earnings disadvantage can be identified within the following framework:

The formulation of Equations 1 and 2 represents the Oaxaca and Ransom (1994) extension of the standard Blinder–Oaxaca framework, with the superscripts W and E denoting individuals from the White majority or ethnic minority, respectively. For the employment analysis represented by Equation 1, the terms represent mean levels of characteristics known to affect employment status,² γˆ terms the estimated regression coefficients, the Oaxaca–Ransom nondiscriminatory employment structure and the average incidence of employment (defined in terms of whether employed or ILO unemployed).³ For the earnings analysis represented by Equation 2, terms represent mean levels of earnings-enhancing characteristics,⁴ βˆ terms the estimated returns to these characteristics, and the mean level of (the logarithm of) gross hourly earnings.⁵ Thus, the difference in average labour market outcomes between Whites and ethnic minorities can be decomposed into a component that is due to differences in average characteristics (the first right-hand side term in square braces of both equations) and a component due to the unequal treatment of Whites and ethnic minorities (the second right-hand side term in square braces).

There are substantial differences, though, in the raw data between the White and ethnic minority populations in Britain. Non-Whites are on average substantially younger than Whites and, reflecting immigration patterns of earlier cohorts, have a regional concentration around large urban conurbations and in London, in particular, where the cost of living and wages is substantially higher than in other areas of the country. Given such differences, a matched sample of Whites is constructed to compare against ethnic minorities via a propensity score approach (Rosenbaum and Rubin, 1983) that satisfies the balancing tests described by Becker and Ichino (2002).

III. Results

The employment analysis is given in Table 1 and that for earnings in Table 2. In all instances, the top row shows earlier results taken from BLMO over the period 1994 to 2000 and the second row shows the aggregate situation between 1997 and 2011. Subsequent rows split this latter period into a number of (overlapping) sub-samples. While not only allowing us to track any potential movement over time, this allows the robustness and generality of the aggregate 1997 to 2011 results to be assessed.

Table 1.  Employment decompositions

	All ethnic minorities			Indian			Pakistani/Bangladeshi			Black
	Diff.	Exp.	Unexp.	Diff.	Exp.	Unexp.	Diff.	Exp.	Unexp.	Diff.	Exp.	Unexp.
1994 to 2000	0.071	0.009	0.062	0.034	−0.020	0.054	0.116	0.013	0.103	0.097	0.034	0.063
		(0.002)	(0.008)		(0.002)	(0.014)		(0.005)	(0.023)		(0.002)	(0.012)
1997 to 2011	0.070	0.015	0.054	0.031	−0.006	0.037	0.089	0.017	0.072	0.092	0.038	0.054
		(0.001)	(0.004)		(0.002)	(0.006)		(0.002)	(0.009)		(0.001)	(0.007)
1997 to 2003	0.073	0.012	0.061	0.034	−0.008	0.042	0.100	0.012	0.088	0.089	0.034	0.054
		(0.001)	(0.007)		(0.002)	(0.011)		(0.003)	(0.016)		(0.002)	(0.010)
2000 to 2006	0.067	0.011	0.055	0.029	−0.009	0.038	0.102	0.007	0.095	0.089	0.036	0.053
		(0.001)	(0.006)		(0.002)	(0.009)		(0.003)	(0.014)		(0.002)	(0.010)
2003 to 2009	0.058	0.014	0.044	0.028	−0.009	0.037	0.081	0.013	0.067	0.082	0.032	0.050
		(0.001)	(0.005)		(0.002)	(0.009)		(0.002)	(0.012)		(0.002)	(0.010)
2006 to 2011	0.062	0.014	0.048	0.033	−0.011	0.044	0.077	0.018	0.059	0.101	0.038	0.063
		(0.001)	(0.006)		(0.002)	(0.010)		(0.003)	(0.012)		(0.002)	(0.012)

Note: SEs are given within parentheses.

Table 2.  Earnings decompositions

	All ethnic minorities			Indian			Pakistani/Bangladeshi			Black
	Diff.	Exp.	Unexp.	Diff.	Exp.	Unexp.	Diff.	Exp.	Unexp.	Diff.	Exp.	Unexp.
1994 to 2000	0.077	0.024	0.053	−0.005	−0.044	0.039	0.052	−0.042	0.093	0.160	0.080	0.081
		(0.003)	(0.014)		(0.008)	(0.027)		(0.013)	(0.040)		(0.006)	(0.019)
1997 to 2011	0.045	−0.010	0.056	−0.033	−0.064	0.031	0.056	−0.018	0.074	0.174	0.073	0.101
		(0.002)	(0.009)		(0.003)	(0.015)		(0.002)	(0.023)		(0.003)	(0.016)
1997 to 2003	0.082	0.015	0.067	−0.040	−0.056	0.016	0.089	−0.000	0.089	0.200	0.085	0.115
		(0.002)	(0.013)		(0.003)	(0.023)		(0.006)	(0.038)		(0.004)	(0.021)
2000 to 2006	0.063	0.006	0.056	−0.022	−0.056	0.034	0.020	−0.015	0.035	0.207	0.081	0.125
		(0.002)	(0.013)		(0.003)	(0.023)		(0.003)	(0.033)		(0.004)	(0.024)
2003 to 2009	0.051	−0.004	0.054	−0.028	−0.047	0.018	0.049	−0.002	0.051	0.193	0.073	0.119
		(0.002)	(0.012)		(0.004)	(0.022)		(0.003)	(0.028)		(0.004)	(0.024)
2006 to 2011	0.047	−0.010	0.058	−0.006	−0.039	0.033	0.052	−0.014	0.066	0.150	0.072	0.077
		(0.002)	(0.013)		(0.004)	(0.023)		(0.003)	(0.030)		(0.005)	(0.027)

Note: SEs are given within parentheses.

In terms of employment incidence, White males were 7.1 percentage points more likely to be in employment than their ethnic minority counterparts over the earlier period of 1994 to 2000 covered by BLMO (Table 1, column 1, 0.071 log points). Only a small fraction of this was attributable to characteristic differences with the majority (0.062 log points) due to measured unequal treatment of similar individuals, a part of which may be ascribed to discrimination. Beyond this, the position of the stock of ethnic minorities has remained remarkably stable in the aggregate 1997 to 2011 period. Even though there is slight variation over more narrowly defined time windows and an apparent downward trend in the unexplained component, such movements are minor and statistically insignificant from the figure of 0.062 in 1994 to 2000. The conclusion to be drawn is that the unfavourable employment position identified by BLMO has not been obviously moderated in the decade since and nor has there been a significant fall in the unexplained component.

Disaggregating across the largest ethnic minority groups there are clearly substantial differences, although a common feature is the unexplained component consistently emerging as the dominant factor in accounting for employment differentials. Indians have an employment rate closest to that of Whites across all time reference windows, with Whites 3.4 (3.1) percentage points more likely to be employed between 1994 and 2000 (1997 to 2011). While the unexplained component of this differential drops from 0.054 to 0.037 log points, this fall is statistically insignificant. Much greater disparities are found between Whites and men of Pakistani/Bangladeshi or Black origins. In the benchmark period of 1994 to 2000, the employment advantage of Whites is 11.6 and 9.7 percentage points over Pakistanis/Bangladeshis and Blacks respectively. As was the case with Indians, there is some suggestion that the raw differential for Pakistanis/Bangladeshis decreases over time (being some four percentage points lower in 2006 to 2011), and the unexplained component exhibits a downward trend. As such, it has fallen from 0.103 log points over the 1994 to 2000 period to 0.059 log points in 2006 to 2011, a significant drop at the 10% level. In contrast, the unexplained component for Blacks exhibits no significant trend but remains substantial.

A not altogether different picture emerges with respect to earnings.⁶ Whites had hourly earnings on average 7.7 percent higher than their ethnic minority counterparts over the period 1994 to 2000 (Table 2, column 1, 0.077 log points), and this advantage had fallen to 4.7% by 2006 to 2011. The unexplained component, though, remained largely unchanged at 0.053 and 0.058 log points respectively.

The earnings position of Indians stands out from that of other ethnic groups in that their average earnings are higher than those of Whites. Indeed, in the two aggregate periods of 1994 to 2000 and 1997 to 2011, the results indicate that we would expect average Indian earnings to be 0.044 and 0.064 log points higher, respectively, based upon the characteristics possessed by Indian men. The main driver of this finding was identified as superior education levels of Indians by BLMO for the period 1994 to 2000 and the same conclusion is drawn for 1997 to 2011. Based upon this consideration alone we would expect to see average Indian earnings 0.100 log points higher than those of Whites. There is also some evidence of a discriminatory element here, with the 0.031 unexplained estimate over 1997 to 2011 being statistically significant. While point estimates over other time periods are of a similar magnitude, none are statistically significant.

In complete contrast is the position of Blacks. An earnings advantage of 0.160 log points for Whites in 1994 to 2000 is comparable to the 0.174 figure for the 1997 to 2011 period and White advantage remains substantial in all sub-periods analysed. A sizeable part of such raw differentials is explained by the inferior characteristics of Black men relative to their White counterparts and once again it is educational attainment that is the dominant component in these decompositions.⁷ There is also a sizeable unexplained component to these decompositions that remains significant across time, estimated at 0.101 log points in the aggregate 1997 to 2011 period, but exhibiting no statistically significant trend.

Finally, the earnings position of the Pakistanis/Bangladeshis may be viewed as a half-way house between the other two minority groups. A raw earnings advantage of 0.052 log points for Whites in 1994 to 2000 compares with a 0.056 log point advantage in 1997 to 2011, although admittedly this does show some volatility in some sub-periods. The same is true for the unexplained component, which is nonetheless a statistically significant 0.074 log points in the aggregate 1997 to 2011 period. Such figures paint a very similar picture to that presented by BLMO for 1994 to 2000 and all absolute differences between the two aggregate time periods are statistically insignificant. Characteristic differences also have a muted role to play in explaining the average earning disparity between Whites and Pakistanis/Bangladeshis.

IV. Conclusions

This article updates earlier research on how ethnic minorities were faring in the British labour market in the 1990s. Previously, substantial earnings and employment differentials between Whites and British-born ethnic minorities were found, with Black and Pakistani employment and earnings disadvantage being particularly severe. In this article, we find that little has changed from that time and this, in spite of goals set by the Prime Minister. The stock of Blacks and Pakistanis still faces particular disadvantage and new initiatives may be needed to support these groups. Such sentiments accord with the findings of Borgas (1992), who stated when analysing the US labour market, where many ethnic groups are longer established, that ‘ethnicity matters and it matters for a very long time’.

Acknowledgements

Materials from the Labour Force Survey are made available by the Office of National Statistics through the Data Archive. Neither bears any responsibility for the analysis or interpretation of the data reported here.

Notes

¹Heckman et al. (2000) is a good example of US work over an extended period. Leslie (1998) provides a comprehensive coverage of issues in the UK and Elliott and Lindley (2008) provides a nice counter-balance to our own work by focussing specifically upon immigrants.

²In the employment analysis, employment status (i.e. whether ILO unemployed or not) is regressed against age (and its square), a dummy for ill health, a dummy for pre-school children in household, marital status (3 dummies), region (6), highest educational qualification (6), housing tenure (4), number of children (4) and year (15).

³While some form of limited dependent choice model would be the more usual way of modelling the dichotomous choice between employment and nonemployment, a linear probability model is used here because it provides a more tractable way of approximating standard errors for the decomposition components. As it turns out, the nature of the results and the conclusions drawn are unaffected by the estimation strategy employed. Results are available on request.

⁴In the earnings analysis, log real hourly earnings are regressed against potential labour market experience (and its square), a dummy for ill health, a public sector dummy, marital status (3 dummies), region (6), highest educational qualification (6), job tenure (3), establishment size (6), industry (12) and year (15).

⁵For both analyses, model specifications are consistent with those of BLMO. Full details of the variables used and their definitions are given in Table A1.

⁶Consistent with BLMO, the analysis is for full-time employees.

⁷Such results are not reported here but are available on request.

Appendix

Table A1.  Variable definitions for analysis

Variable	Variable description
lnghearn	The natural logarithm of gross hourly earnings from employment in January 2011 prices. Gross hourly earnings are defined as actual gross weekly earnings deflated by usual weekly hours worked excluding unpaid overtime.
employed	Dummy variable indicating employment status: ; 0=ILO unemployed.
age	Age of respondent in years entered in linear and quadratic form.
exprn	The number of years of potential work experience entered in linear and quadratic form.
ill	Dummy variable indicating a health problem or disability that limits the kind of paid work that can be undertaken.
marstat	Dummy variables (3) indicating marital status.
	marstat1: married or cohabitating; marstat2: single; marstat3: widowed, divorced or separated.
jobyrs	Dummy variables (3) indicating tenure with current employer.
	jobyrs1: 1 year or less; jobyrs2: between 2 and 5 years inclusive; jobyrs3: greater than 5 years.
hqual	Dummy variables (6) indicating highest educational qualification.
	hqual1: university degree (postgraduate or undergraduate); hqual2: other degree equivalent qualification; hqual3: A-level or equivalent (typical requirement for degree-level study); hqual4: O-level or equivalent (typical qualification at end of compulsory schooling); hqual5: Other; hqual6: none.
psize	Dummy variables (6) indicating the size of establishment.
	psize1: 1–10 employees; psize2: 11–19 employees; psize3: 20–24 employees; psize4: do not know, but <25 employees; psize5: 25–49 employees or do not know, but >24 employees; psize6: 50+ employees.
Ind	Dummy variables (12) indicating industry of employment.
	ind1: mining, quarrying, energy and water supply; ind2: manufacturing; ind3: construction; ind4: wholesale, retail and motor trade; ind5: hotels and restaurants; ind6: transport, storage and communications; ind7: financial intermediation; ind8: real estate, renting and business activity; ind9: public administration and defence; ind10: education; ind11: health and social work; ind12: other services.
public	Dummy variable indicating public sector employment.
reg	Dummy variables (6) indicating region of residence.
	reg1: Northern, Yorkshire and Humberside, South West, Wales; reg2: East Midlands, East Anglia, South East; reg3: Inner London; reg4: Outer London; reg5: West Midlands; reg6: North West, Scotland.
yngkids	Dummy variable indicating the presence of pre-school dependent children.
kids	Dummy variables (4) indicating the number of dependent children.
	kids0: no children; kids1: one child; kids2: two children; kids3: three or more children.
Hsetype	Dummy variables (4) indicating type of housing tenure.
	hsetype1: owner occupier with mortgage; hsetype2: property owned outright; hsetype3: rented accommodation with subsidy element; hsetype4: other rented.

Table A2.  Variable means for analysis: Labour force survey 1997Q1 to 2011Q4

	Employment equation				Wage equation
	White^a	Indian	Pak/Bang	Black	White^a	Indian	Pak/Bang	Black
employed	0.932	0.902	0.829	0.848	–	–	–	–
lnghearn	–	–	–	–	7.148	7.155	6.967	7.086
age	31.846	29.931	28.158	34.412	–	–	–	–
exprn	–	–	–	–	15.006	11.343	9.722	17.529
ill	0.043	0.036	0.042	0.049	0.036	0.023	0.033	0.026
marstat1	0.566	0.512	0.526	0.474	0.657	0.606	0.615	0.575
2	0.401	0.455	0.442	0.478	0.304	0.359	0.366	0.365
3	0.033	0.033	0.031	0.048	0.039	0.035	0.019	0.059
jobyrs1	–	–	–	–	0.273	0.308	0.342	0.277
2	–	–	–	–	0.340	0.362	0.392	0.337
3	–	–	–	–	0.387	0.331	0.266	0.386
hqual1	0.221	0.414	0.276	0.189	0.264	0.458	0.395	0.205
2	0.118	0.113	0.099	0.122	0.134	0.127	0.124	0.153
3	0.270	0.179	0.183	0.281	0.274	0.163	0.181	0.289
4	0.176	0.149	0.174	0.165	0.169	0.133	0.138	0.162
5	0.112	0.073	0.122	0.135	0.097	0.064	0.083	0.119
6	0.103	0.072	0.147	0.107	0.062	0.054	0.081	0.073
psize1	–	–	–	–	0.146	0.130	0.154	0.122
2	–	–	–	–	0.073	0.061	0.071	0.070
3	–	–	–	–	0.038	0.042	0.031	0.048
4	–	–	–	–	0.016	0.012	0.019	0.020
public	–	–	–	–	0.182	0.177	0.176	0.244
reg1	0.141	0.096	0.258	0.085	0.154	0.099	0.276	0.113
2	0.237	0.278	0.165	0.208	0.262	0.292	0.166	0.239
3	0.122	0.041	0.087	0.230	0.115	0.048	0.071	0.176
4	0.218	0.255	0.102	0.280	0.211	0.242	0.140	0.278
5	0.163	0.235	0.192	0.136	0.151	0.245	0.169	0.134
6	0.118	0.095	0.195	0.061	0.107	0.075	0.178	0.059
yngkids	0.195	0.231	0.324	0.198	–	–	–	–
kids0	0.601	0.559	0.416	0.567	–	–	–	–
1	0.173	0.186	0.211	0.186	–	–	–	–
2	0.163	0.181	0.178	0.171	–	–	–	–
3	0.064	0.075	0.195	0.077	–	–	–	–
hsetype1	0.605	0.603	0.481	0.512	–	–	–	–
2	0.123	0.298	0307	0.072	–	–	–	–
3	0.135	0.041	0.116	0.298	–	–	–	–
4	0.138	0.057	0.095	0.118	–	–	–	–
Sample size	23 952	2105	1571	2307	15 861	832	421	909

^aSignifies that the sample has been matched to a combined group of all ethnic minorities (matches to individual ethnic groups are available on request); 15 year of interview controls and 12 controls for industry of current employment are not presented due to space constraints (but are available on request).

Table A3.  Regression coefficients for analysis: Labour Force Survey 1997Q1 to 2011Q4

	Employment equation				Wage equation
	White^a	Indian	Pak/Bang	Black	White^a	Indian	Pak/Bang	Black
age	0.002	0.005	0.007	0.006	–	–	–	–
exprn	–	–	–	–	0.012	0.020	0.015	0.010
ill	−0.075	−0.077	−0.118	−0.150	−0.162	−0.256	−0.052	−0.168
marstat1	(E)	(E)	(E)	(E)	(E)	(E)	(E)	(E)
2	−0.049	−0.090	−0.088	−0.122	−0.199	−0.149	−0.097	−0.128
3	−0.062	−0.098	−0.056	−0.146	−0.076	−0.086	0.070	−0.188
jobyrs1	–	–	–	–	(E)	(E)	(E)	(E)
2	–	–	–	–	0.101	0.064	0.133	0.089
3	–	–	–	–	0.222	0.168	0.191	0.192
hqual1	0.082	0.064	0.183	0.152	0.784	0.804	0.782	0.774
2	0.079	0.026	0.145	0.161	0.513	0.502	0.456	0.575
3	0.069	0.039	0.134	0.132	0.332	0.454	0.316	0.375
4	0.071	0.019	0.102	0.106	0.264	0.236	0.266	0.365
5	0.036	0.014	0.048	0.078	0.116	0.162	0.110	0.262
6	(E)	(E)	(E)	(E)	(E)	(E)	(E)	(E)
psize1	–	–	–	–	(E)	(E)	(E)	(E)
2	–	–	–	–	−0.076	−0.073	−0.076	−0.032
3	–	–	–	–	−0.028	0.120	0.003	−0.100
4	–	–	–	–	−0.076	−0.042	−0.219	−0.103
public	–	–	–	–	−0.036	0.027	0.077	−0.056
reg1	(E)	(E)	(E)	(E)	(E)	(E)	(E)	(E)
2	0.014	−0.023	0.077	0.002	0.143	0.269	0.259	0.199
3	−0.004	−0.008	0.048	−0.031	0.325	0.460	0.237	0.158
4	0.007	−0.049	0.039	−0.013	0.280	0.318	0.371	0.216
5	0.006	−0.033	−0.016	−0.015	0.050	0.061	0.043	−0.010
6	−0.004	−0.037	0.030	−0.019	0.024	0.092	0.110	−0.013
yngkids	−0.004	−0.019	0.047	0.045	–	–	–	–
kids0	(E)	(E)	(E)	(E)	–	–	–	–
1	−0.002	−0.012	−0.047	−0.048	–	–	–	–
2	−0.004	−0.006	−0.048	−0.052	–	–	–	–
3	−0.019	−0.048	−0.063	−0.056	–	–	–	–
hsetype1	(E)	(E)	(E)	(E)	–	–	–	–
2	−0.016	0.008	−0.012	−0037	–	–	–	–
3	−0.127	−0.080	−0.167	−0.148	–	–	–	–
4	−0.027	−0.006	−0.108	−0.039	–	–	–	–
constant	0.833	0.785	0.617	0.674	6.343	6.191	6.153	6.294
Sample size	23 952	2105	1571	2307	15 861	832	421	909

^aSignifies that the sample has been matched to a combined group of all ethnic minorities (matches to individual ethnic groups are available on request); 15 year of interview controls and 12 controls for industry of current employment are not presented due to space constraints (but are available on request); (E) denotes an excluded dummy variable category; figures in bold are statistically significant at the 5% level or better; figures in bold and italics are statistically significant at the 10% level or better.