Are the winners winning in the hourglass labour market? Craig Holmes and Ken Mayhew Paper submitted for 30th International Labour Process Conference, University of Stockholm, March 27th-29th 2012 1 Introduction The occupational composition of many national labour markets has changed significantly over the past thirty years. The sort of jobs firms require and create depends on a number of factors, but one of the key ones is the available level of technology. As the general level of technology improves, firms are able to invest in more, better and cheaper capital, particularly computerised machinery. Some jobs may be complementary to the introduction of new technology, while others may be substitutable. For example, if an engineer is more productive because she now works with an upgraded computer system or software design package, then we would expect the total demand for engineers to increase as firms seek to capitalise on their increased profitability. In contrast, an automated production process may replace part (or indeed all) of a manned production line. These different changes can appear within a single firm – the same automated production process that reduces employment of production line workers may require firms to employ more highly skilled production supervisors to monitor or program it. One viewpoint is that jobs which are most complementary to technological improvements are more highly skilled, while jobs which are substitutable with these advances are low skilled. This is commonly referred to as skill-biased technical change. An alternative viewpoint is that technology is related to particular tasks, rather than particular skills. Jobs which predominantly performed routine tasks were replaced by new technology, while those which performed non-routine tasks were not. Indeed in many cases, non-routine work is complementary with new technology, though this is not always the case. Goos and Manning (2007) showed that these non-routine jobs tended to be either high-wage jobs or low-wage, jobs, while routine occupations were generally closer to the middle of the earnings distribution. They characterised the UK labour market as increasingly polarised, with a hollowing out of the middle and the development of an “hourglass” labour market. This has been taken by policymakers as supportive of the view that the UK labour market continues to provide an increasing number of good jobs. The UK government's recent strategy document on improving social mobility states: There is…evidence that the demand for skilled workers is currently outstripping supply, which suggests there is 'room at the top' for highly qualified graduates from all backgrounds. (HM Government, 2011, p. 11) This, in turn, creates “winners” in the labour market by allowing for greater upward mobility for existing workers and improving the labour market outcomes of new, more qualified entrants. A number of potential problems arise from this premise, which this paper explores. One problem is that non-routine jobs at the bottom of the labour market have also been growing, there is an increased likelihood that some of these existing workers or new, well-qualified entrants will find themselves in lower wage work – that there is ‘room at the bottom’ as well. This issue is not the main focus of this paper, although we briefly discuss our earlier research on downward occupational mobility as a result of a decline in routine jobs. The main focus of this paper is the earnings of the “winners”. In particular, it asks whether the growing number of workers found in good, non-routine jobs have all benefitted in terms of higher earnings. We present some descriptive statistics from earnings and wage distributions in these jobs, and show where the growth has occurred within the pay spectrum. Secondly, we look at issues related to wage mobility associated with the presumed ‘room at the top’. Using data from the 1958 National Child Development Study (NCDS) we look at whether there are improvements in earnings associated with transitions from routine occupations to “good” non-routine occupations. Thirdly, we look at the labour market outcomes for those entering the labour market after many routine jobs had disappeared. In the absence of this change in the occupational structure, many of these new entrants would have gone into routine jobs. Using data from the 1970 British Cohort Study (BCS), we look at whether these workers were able to find work in “good” non-routine jobs. 2 Room at the bottom, too? In some of our recent work on the mobility of routine workers, we find that the change in the occupational structure has also increased mobility towards these jobs, regardless of educational attainment (Holmes and Mayhew, 2011b). In light of this, we would argue that policymakers do not spend enough time considering the implications of the ‘room at the bottom’. For example, in a paper that provides a part of the conceptual framework for the above strategy document, Crawford et al. (2011) raise the issue of the polarised (or hourglass) labour market, but the implications they draw from it remains focused on the 'room at the top'. For those already in these low-wage jobs, the authors acknowledge that “attempting to move individuals from the bottom to the middle of the skill/income distribution may be harder, as there are fewer jobs in the middle” (pg. 9). The apparent solution here is upskilling: in the longer term increasing the supply of intermediate skilled workers may encourage firms to up-skill [low-wage] jobs in order to employ such workers (Crawford et al., 2011, pg 9) Such policies have rarely been successful in the UK – over-education and underemployment are much more commonly reported. 3 Wages at the top Earnings is good, non-routine jobs should reflect both the increased demand generated by technological progress, and increased supply resulting from an increasingly well-educated workforce able to perform these jobs. The effects of these two changes are contradictory – increased supply will lower wages, while increased demand raises them. It has been commonplace for policymakers, in some cases supported by academic research, to point to evidence of non-declining returns on higher education to argue that demand and supply for workers in these jobs are moving together (e.g. BIS, 2010; Walker and Zhue, 2008; Machin and Van Reenen, 2007), and that the growth of higher-skill jobs has benefitted those in those jobs or moving into those jobs. Recently, there has been concern that the polarisation effect identified by Goos and Manning when looking at occupational titles is less apparent when looking at wage distributions (Holmes and Mayhew, 2011a). This has been the result of a small number of high-skill jobs experiencing very rapid wage growth over the past two decades, whilst many other upper-end jobs have not. The result of this is that many of these good non-routine jobs earn wages much closer to the middle than the top. Hence, as well as looking at average wages between different groups of occupations, it is also important to look at the distribution of these earnings. Holmes and Mayhew (2012) decompose changes in the overall UK wage distribution between 1987 and 2001 into effects caused by changes in the composition of a range of factors that are related to earnings across the workforce (such as qualification levels, occupational structure and union membership) and the way different worker characteristics correspond to wages – what economists call rates of returns or wage premia. The wage differentials between different occupations have changed, but these changes are not the same across the distribution. The relative earnings of managerial workers, for example, have grown particularly around the 75th percentile of the distribution, but much less elsewhere. Rates of return to a degree also appear to have fluctuated at different points in the wage distribution. The analysis shows that the graduate premium has grown sharply for the top 15% of earners, but has remained relatively constant (or even declined) for everyone else. This suggests that many of these apparently good jobs are far less appealing than their job titles may suggest, as the types of jobs captured within an occupational definition widens. This seems to reflect both the changing composition of the workforce and changes in the way certain characteristics are linked to earnings. However, it is not a simple exercise to untangle all of the potential interactions between composition and wage premia at the aggregate level – for example an increase in demand for a range of high skill non-routine jobs alongside an increase in supply of graduates from an increasingly heterogenous university sector is unlikely to have a easy-to-quantify effect on total earnings distributions. Table 3.1: Occupational employment shares, 1995-2008 Managerial Professional Intermediate 1995 13.9% 10.8% 9.6% 2002 13.3% 12.0% 14.0% 2008 15.6% 13.2% 14.4% Source: LFS, own calculations To illustrate some of these effects, this section presents data from the Labour Force Survey on the distribution of weekly earnings in occupational categories associated with the idea of high skill non-routine work – SOC major groups 1 (managerial occupations), 2 (professional occupations) and 3 (intermediate and associate professional occupations)1. We focus on two We use both SOC90 for the 1995 data, and SOC2000 for 2002 and 2008 data – we acknowledge the differences between the two classifications, however, we believe these differences are less relevant for the top occupations than those lower down, and where issues may arise the biases which result should only strengthen our conclusions. For example the overclassification of supervisory positions as managerial occupations in 1 periods: 1995-2002 and 2002 to 2008 (prior to the onset of the recession). Table 3.1 shows the changing employment share of workers in these occupations. The LFS provides wage data on a subset of the overall sample – the above table gives employment share for this subset. The patterns of growth are consistent with the overall sample. Figure 3.1, 3.2 and 3.3 shows the change in the percentage of workers earning below a certain gross weekly wage (given along the x axis) in these occupations. These measures show were work has shifted within the distribution – negative values indicate proportionally less employment below a certain threshold, while positive values denote proportionately more employment below this point. Figure 3.1: Changing cumulative distribution of gross weekly earnings, managerial occupations, 1995-2008 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 900 1000 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 5.0% -5.0% 1995-2002 2002-2008 -10.0% -15.0% -20.0% Threshold gross weekly earning, £ Across all three occupations, the period between 1995 and 2002 saw a large fall in the proportion of jobs earning below median wages2, suggesting that many of the jobs in these occupations, potentially including any new jobs created during this time period, were higher wage. For managerial jobs and (to a lesser extent) professional occupations, we see quite a long tail in this shift, with a noticeable increase in the employment share of jobs earning SOC2000 would, on its own, result in a rising number of lower wage managerial occupations between 1995 and 2002 – this is not found to be the case. 2 The median gross weekly wage for these three SOC categories over the three time periods, in 1995 prices, are £406.66, £412.81 and £310.87 respectively. above £800 per week. In contrast, the growth in these occupations during the past decade has been led to much less of an increase in higher wage employment. For all three occupations, there has been a modest increase in employment at the bottom end of the spectrum in these occupations, most noticeably in professional occupations. There is also a decline in employment nearer the middle of the earnings spectrum for all occupations3, which implies a growth in jobs at the highest end. Figure 3.2: Changing cumulative distribution of gross weekly earnings, professional occupations, 1995-2008 5.0% -5.0% 1995-2002 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 900 1000 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 10.0% 2002-2008 -10.0% -15.0% -20.0% Threshold gross weekly earning, £ We extend this analysis in two directions, firstly looking at differences by qualification level, and then considering differences by sector of employment. 3.1 Graduates and non-graduates Figures A.1 to A.6 break down these distribution changes into graduates – workers with a qualification equivalent to NVQ level 4 or 5 – and non-graduates to see whether opportunities to benefit from the growth in higher-wage non-routine work is particularly linked to the increase in more qualified workers. As professional work has always been heavily dominated by graduate workers, we focus on the managerial and intermediate occupations. For the 3 Increased employment below £300 per week, but decreased employment share below £400-600 per week implies a decline in employment between £300 and £600, and a growth in employment at either side of this range. former, Figures A.1 and A.2 show that between 1995 and 2002, the increase in employment at the very top end was limited – there is very little decline in employment share below around £1000 per week for non-graduates, whereas a sizeable proportion of graduates are now being found in jobs earning over £1400 per week. Figure 3.3: Changing cumulative distribution of gross weekly earnings, intermediate occupations, 1995-2008 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 900 1000 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 2.0% -2.0% -4.0% -6.0% -8.0% 1995-2002 2002-2008 -10.0% -12.0% -14.0% -16.0% Threshold gross weekly earning, £ There is much less observable difference since 2002 at the upper end – for both types of workers, there is evidence of a decline in employment share between around £500 and £1200, and growth above that. Similarly, at the bottom end, in both time period, the patterns of change are broadly the same – between 1995 and 2002, a large proportion of both types of workers have moved away from employment below median wage threshold, while after 2002, there was an increase in the proportion of all workers doing lower wage jobs. 3.2 Sectors Figures B.1 to B.4 break down the change in managerial earnings by the four largest sectors these workers are found. This shows that the pattern of change in the distribution was broadly similar between 1995 and 2002 across sectors, although there are differences in terms of increased employment at the top end of the spectrum. Due to increasingly small sample size, we do not analyse the link between educational attainment, industry and occupation. That said, the patterns of changing employment share between managers in real estate and business activities, and managers in retail seem similar to the overall change for graduates and non-graduates respectively, suggesting there may be selection (or self-selection) intro different types of managerial work based on educational attainment. Since 2002, as the number of managerial occupations has increased, these jobs have become increasing heterogeneous across different industries. Retail and financial intermediation have seen significant growth in lower wage managerial occupations since 2002, whereas manufacturing and real estate, renting and business activities has predominantly seen a shift away from work at the lowest end during this time period. Here there is less obvious similarity between graduate and non-graduate earnings, suggesting something other than a division based on educational attainment. 4 Room to move up? The previous section has suggested that over the past two decades, the growth in good nonroutine jobs has created higher wage employment, although these gains are becoming increasingly spread out with, and a number of these jobs are considerably lower wage than before. One issue is whether these higher wage jobs are available to those already in the workforce, which is considered in this section. Another issue is whether the growth in these jobs creates better initial employment options for increasingly well qualified new labour market entrants, which is examined in the following section. Using data from the 1958 National Child Development Study (NCDS) we look at whether there are improvements in earnings associated with transitions from routine occupations to “good” non-routine occupations. Wage data is available at four points in time in our data – 1981, 1991, 1999 and 2004. We look at gross weekly earnings in high-skill non-routine occupations at this time, comparing these earnings by the occupational group of employment five years prior to these dates (so we exclude data from 1981 as the majority of the sample were not in employment in 1976). The occupational groups we consider are professional, managerial, intermediate, routine, and two low-wage non-routine categories – non-routine manual and service (see Holmes, 2011, for more on the derivation of these categories in a previous occupational mobility study). Figures 4.1 and 4.2 shows the distribution of weekly earnings (in 1981 terms) for those in managerial occupations and intermediate occupations, comparing workers by their occupational group five years previously. This would seem to indicate that those transferring from routine occupations earn less than existing managers or intermediate occupation workers. Figure 4.1: Gross weekly earnings distribution, managerial occupations, by occupation of origin 30.0% Managerial Employment share 25.0% Routine 20.0% 15.0% 10.0% 5.0% 950 - 1000 900 - 950 850 - 900 800 - 850 750 - 800 700 - 750 650 - 700 600 - 650 550 - 600 500 - 550 450 - 500 400 - 450 350 - 400 300 - 350 250 - 300 200 - 250 150 - 200 100 - 150 0 - 50 50 - 100 0.0% Gross weekly wage Figure 4.2: Gross weekly earnings distribution, intermediate occupations, by occupation of origin 35.0% Employment share 30.0% 25.0% Intermediate Routine 20.0% 15.0% 10.0% 5.0% 950 - 1000 900 - 950 850 - 900 800 - 850 750 - 800 700 - 750 650 - 700 600 - 650 550 - 600 500 - 550 450 - 500 400 - 450 350 - 400 300 - 350 250 - 300 200 - 250 150 - 200 100 - 150 0 - 50 50 - 100 0.0% Gross weekly wage This could be explained purely by observable differences between workers who have been in managerial occupations for longer period of time and those who had previously worked in routine occupations but who have either progressed or been displaced from these jobs. For instance, workers in routine jobs who move into managerial or intermediate occupations may be, on average, less educated. Alternatively, it may be that there are unobservable (to the econometrician) differences between these two categories of workers, leading to the creation of different types of managerial occupations, with correspondingly different earnings profiles. It may also be the case that the available progression routes between routine and higher skill non-routine occupations lead to these kind of stratification, so that early employment decisions matter. Table 4.1: Ordered logit estimation of earnings distribution group GENDER NONWHITE Occupation five years before PROFESSIONAL MANAGERIAL INTERMEDIATE ROUTINE MANUAL SERVICE UNEMP NONEMP AGE Highest academic and vocational qualifications VOC NVQ 0 ACAD NVQ 0 VOC NVQ 1 ACAD NVQ 1 VOC NVQ 3 ACAD NVQ 3 VOC NVQ 4 ACAD NVQ 4 VOC NVQ 5 ACAD NVQ 5 N Managerial -1.933 *** -0.431 ** Intermediate -2.142 *** -0.025 0.246 Ref. -0.290 -0.498 -1.212 -1.240 -1.941 -1.110 0.616 0.574 Ref. -0.982 -0.586 -1.339 -1.495 -1.098 *** ** *** *** *** 0.365 *** 0.179 -0.692 0.184 -0.483 -0.076 0.889 0.175 1.111 0.780 1.531 2550 *** *** *** *** *** *** ** *** *** *** *** *** 0.394 *** 0.279 -0.164 0.158 -0.242 0.219 0.436 0.511 0.457 0.771 0.877 * * *** *** *** *** *** 2313 To test this, we create a variable representing groups of earnings in Figures 4.1 and 4.2, with Y = 1 indicating earnings between up to £50 per week, Y = 2 indicating earnings between £50 and £100 per week, and so on up to Y = 20. We then estimate the following ordered logit for managerial and intermediate occupations: Pr (Y = y) = ologit(age, gender, ethnicity, occupational group-5y, highest academic qualification, highest vocational qualification) (1) Table 4.1 shows the results of this estimation. The main result shown in this table is that even controlling for observable demographic and educational differences between workers, those who transition into these jobs from routine work (as well as from service occupations and from unemployment or non-employment) tend to experience less high wages in those occupations. This can not be purely attributed to lower specific occupational skills developed by doing either job for a longer period of time – the results show that those who move into managerial or intermediate occupations from other higher-skill non-routine occupations do not experience lower wages. There are two additional aspects that are beyond the scope of this paper, but which need to be addressed. Firstly, there is a question of whether these progression paths lead to earnings gains – so far, we have showed that these transitions lead to lower than average earnings relative to existing workers, but not whether these transitions are associated with upward wage mobility. Secondly, we would like to know whether different sorts of transition are associated with different outcomes. In previous work on occupational mobility using the same NCDS data (Holmes 2011), the causes of transitions from routine occupations to nonroutine occupations was decomposed into career progression and displacement caused by a change in the occupational structure. That research showed that some mobility in this cohort was associated with the decline in the total number of routine jobs. As we have less observations of earnings in these data, it is not possible to associate wage mobility with changes in the occupational structure in the same way. However, one possible extension is to use data on the causes of previous job exits to distinguish between more voluntary career progression transitions and those caused by a decline in the total number of routine jobs. 5 A better start? We now consider the change in the initial employment opportunities resulting from the increase in non-routine work. In the absence of this change in the occupational structure, many of these new entrants would have gone into routine jobs. Comparing data from the NCDS with that from the 1970 British Cohort Study (BCS), we look at whether new entrants into the labour market benefitted by being able to find work in better non-routine jobs, and what role the increased educational attainment of the later cohort played in this. Table 5.1 shows the results of a logit regression on occupation of employment at the beginning of the first period of the sample for each cohort (i.e age 23 for the NCDS cohort, and 25 for the BCS cohort). To include interaction between cohort and education (given that the precise role of different qualifications as mechanisms for entry into the labour market may have changed over time), we reduce the latter down to two dummies – one for academic qualifications at level 3 and above, and another for vocational qualifications at the same levels. This reduces the number of interaction terms to a more reasonable level4. Table 5.1: Logit regression on occupational group of initial employment GENDER NON WHITE BCS Routine -0.138 *** -0.101 0.028 0.455 *** -0.109 ** 0.730 *** 0.253 ** 0.495 *** 1.011 *** -1.201 *** -0.074 BCS * ACAD LVL3-5 0.012 BCS * VOC LVL3-5 -0.072 ACAD LVL3-5 * VOC LVL3-5 -0.226 BCS*ACAD LVL3-5 * VOC LVL3-5 0.466 0.018 -0.813 *** -0.446 *** 0.444 0.098 0.254 ** 0.368 *** 0.160 CONST -2.493 *** -0.406 *** Highest ACAD LVL3-5 qualifications VOC LVL 3-5 Interaction terms Managerial Intermediate -0.451 *** 0.053 -0.255 ** -0.149 -2.238 *** The estimation shows that, controlling for educational differences, workers in the younger BCS cohort are less likely to start work in routine occupations. This is expected, as there are far fewer routine occupations in 1995 than there are in 1981. Again controlling for educational differences, the table shows that younger BCS cohort is more likely to work in intermediate occupations, but not any more likely to work in managerial occupations early in the working life. 4 We follow the best practice set out in Brambor, Clark and Golder (2006) and include all two-way and threeway interaction terms in the model, even where no significant effects are expected. Keeping a large number of educational qualification dummies would necessitate fourth, fifth and higher interaction terms, which would overcomplicate the analysis. The younger cohort benefits from being increasingly well educated – a greater proportion of them go into good non-routine jobs as they are more qualified. What is surprising, however, is that the better-educated workers in the younger BCS cohort are not relatively more likely to find employment in good non-routine jobs – the interaction terms between higher qualifications and cohort membership are either insignificant or negative in some cases. Given the decline of routine occupations, the growth of managerial and intermediate occupations, and the effects these patterns would be expected to have on future employment prospect, career progression and earnings, we would expect sufficiently qualified new entrants to be increasingly more likely to seek good non-routine jobs earlier in their careers5. Understanding why early employment has not behaved in this way is a necessary extension of this. One hypothesis that could be tested is that some routine jobs remain important gateways into better employment later on, and that there are non-human capital barriers to entering higher wage routine occupations for younger workers. 6 Concluding remarks This paper has questioned an assumption underlying much of the UK government’s thinking on the occupational structure of the labour market: that the growth in occupational groupings associated with higher earnings will generally create “winners”, both in opportunities to progress upwards in the labour market, and for increasingly well-qualified new entrants to move into. We raise three concerns about this optimism. Firstly, that there is evidence that earnings in these jobs has become increasingly heterogenous, with an increase in low pay in these occupations, at the same time that the wages of a small proportion of these jobs has accelerated away from the rest. Secondly, that upward mobility from declining routine occupations to higher-skill jon-routine occupations is associated with a lower distribution of earnings than those currently in these jobs, even after controlling for educational attainment. Thirdly, comparing those joining the labour market in the early 1990s with those entering in the late 1970s, newer entrants are more likely to start working lives in these growing occupations. However, the incidence of this is not as great as we might expect – being more educated in the later cohort is not associated with being more likely to go into managerial or intermediate non-routine jobs, although it is not clear exactly what barriers to such entry are causing this. 5 This motivation has been discussed in the literature, leading to the phenomena of routine jobs 'getting old' (Autor and Dorn, 2009) References Autor, D, and Dorn, D, (2009), This job is “getting old”: measuring changes in job opportunities using occupational age structure, American Economic Review, 99(2): 45-51. Autor, D, Levy, F, and Murnane, R, (2003), The skill content of recent technological change: an empirical exploration, Quarterly Journal of Economics, 118(4): 1279-1333. BIS (2010), The returns to higher education qualifications, BIS research paper no. 45, London: Department of Business, Innovation and Skills. Brambor, Clark and Golder (2006), Understanding interaction models: improving empirical analyses, Political Analysis, 14(1): 63-82. Crawford, C, Johnson, P, Machin, S, and Vignoles, A, (2011), Social mobility: a literature review, paper for the Department for Business, Innovation and Skills, available online: http://www.bis.gov.uk/assets/biscore/economics-and-statistics/docs/s/11-750-social-mobilityliterature-review.pdf. Goos, M, and Manning, A, (2007), Lousy jobs and lovely jobs: the rising polarization of work in Britain, The Review of Economics and Statistics, 89(1): 118-133. HM Government, (2011), Opening doors, breaking barriers: a strategy for social mobility, London: Cabinet Office. Holmes, C, (2011), The route out of the routine: where do the displaced routine workers go?, SKOPE Research Paper No. 100, Cardiff: SKOPE ―, and Mayhew, K, (2011a), Calling time on the hourglass labour market: the absence of polarisation in UK wage distributions, paper presented at ILR Conference on ‘Job Quality’, Cornell University, NY, USA, November 4th ―, and Mayhew, K, (2011b), The route out of the routine: mobility and the changing structure of occupations, paper presented at International Labour Process Conference, University of Leeds, April 7th ―, and Mayhew, K, (2012), The changing shape of the UK job market and the implications for the bottom half of earners, London: Resolution Foundation Machin, S, and Van Reenen, J, (2007), Changes in wage inequality, CEP Special Paper No. 18, London: CEP. Walker, I, and Y, Zhu (2008), The college wage premium and the expansion of higher education in the UK, Scandinavian Journal of Economics, 110(4) pp. 695-710. A Earning distributions: graduates vs. non-graduates Figure A.1: Changing cumulative distribution of gross weekly earnings, graduates in managerial occupations, 1995-2008 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 5.0% -5.0% 1995-2002 -10.0% 2002-2008 -15.0% -20.0% Threshold gross weekly earnings, £ Figure A.2: Changing cumulative distribution of gross weekly earnings, non-graduates in managerial occupations, 1995-2008 -5.0% 1995-2002 -10.0% 2002-2008 -15.0% -20.0% Threshold gross weekly earning, £ 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 5.0% Figure A.3: Changing cumulative distribution of gross weekly earnings, graduates in professional occupations, 1995-2008 5.0% -5.0% 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 10.0% 1995-2002 2002-2008 -10.0% -15.0% -20.0% Threshold gross weekly earnings, £ Figure A.4: Changing cumulative distribution of gross weekly earnings, non-graduates in professional occupations, 1995-2008 5.0% -5.0% 1995-2002 -10.0% 2002-2008 -15.0% -20.0% -25.0% Threshold gross weekly earning, £ 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 10.0% Figure A.5: Changing cumulative distribution of gross weekly earnings, graduates in intermediate occupations, 1995-2008 2.0% 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 -2.0% 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 4.0% -4.0% -6.0% 1995-2002 -8.0% 2002-2008 -10.0% -12.0% -14.0% -16.0% -18.0% Threshold gross weekly earnings, £ Figure A.6: Changing cumulative distribution of gross weekly earnings, non-graduates in intermediate occupations, 1995-2008 2.0% -4.0% -6.0% -8.0% 1995-2002 -10.0% 2002-2008 -12.0% -14.0% -16.0% -18.0% Threshold gross weekly earning, £ 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 900 800 700 600 500 400 300 1000 -2.0% 200 0.0% 100 % change in employment share below threshold wage 4.0% B Earning distributions: sectorial analysis of managerial occupations Figure B.1: Changing cumulative distribution of gross weekly earnings, managerial occupations in manufacturing, 1995-2008 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 5.0% -5.0% -10.0% 1995-2002 2002-2008 -15.0% -20.0% -25.0% Threshold gross weekly wage Figure B.2: Changing cumulative distribution of gross weekly earnings, managerial occupations in retail and wholesale, 1995-2008 10.0% 5.0% -5.0% 1995-2002 -10.0% 2002-2008 -15.0% -20.0% -25.0% Threshold gross weekly wage 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 15.0% Figure B.3: Changing cumulative distribution of gross weekly earnings, managerial occupations in real estate, renting and business activities, 1995-2008 5.0% 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 10.0% -5.0% -10.0% 1995-2002 -15.0% 2002-2008 -20.0% -25.0% -30.0% Threshold gross weekly wage Figure B.3: Changing cumulative distribution of gross weekly earnings, managerial occupations in financial intermediation, 1995-2008 -5.0% -10.0% 1995-2002 2002-2008 -15.0% -20.0% -25.0% Threshold gross weekly wage 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 0.0% 100 % change in employment share below threshold wage 5.0% C Datasets Our analysis of earning distributions comes from the Labour Force Survey. Data on employment have been collected by the LFS since 1973, on a yearly basis until 1992, then on a quarterly basis. Wage data have been available from the LFS since 1993 for a subset of the overall survey. The sample size is around 60,000 households, leading to approximately 120,000-130,000 individual observations. Our analysis of mobility uses data from the National Child Development Study (NCDS). The members of the NCDS study were all born in a single week in March 1958. Data has been collected on these members in a series of waves. The most useful waves for assessing labour market outcomes the fourth and seventh waves, taken in 1981, 1991, 1999-2000 and 2004-5 respectively. The fourth wave is the first one taken after the school leaving age (respondents were aged 23) and records early labour market experience. The seventh wave was completed in 2004-5 (respondents were aged 46-47), and has the most recent data on wages, employment and education. It is possible to construct an entire working life history over this time period using responses from all four waves, including periods of employment, unemployment, self-employment and non-participation for a number of reasons such as sickness or further education. As with all longitudinal studies, there are missing data. The sample size is around 12,000 for the fourth wave, and around 10,000 for the seventh wave. Employment data are available for a subset of these observations. To compare early employment patterns, we use the British Cohort Study (BCS). The BCS is a longitudinal study following 17,200 subjects born in England, Scotland, Wales and Northern Ireland in a particular week in April, 1970. The cohort has been surveyed in series of waves since the study's inception, and until now there have already been eight waves. The fifth wave, taken in 1996, is the first full wave taken after the school leaving age of the subjects, and has a sample size of around 9,000.