1Trends in skill have had a central place in sociological accounts of the changing nature of work. They have also been one of the most contentious foci of debate, with quite diverse scenarios emerging about the long-term direction of change in skills, their implications for the quality of working life and their connection to technological change. This lack of consensus can be partly attributed to the very difficult issues involved in the conceptualization and measurement of skill, and relatedly to the absence of good indicators that are both representative of the workforce as a whole and comparable across time. This paper seeks to contribute to these long-standing debates by reporting on some recent British research that provides a more reliable basis for assessing trends in skill.
2Broadly speaking, with respect to trends in skill, three major perspectives have featured in the literature. There are those who have taken the view that there is a long-term trend for skill levels to increase in advanced capitalist societies (Blauner, 1964 ; Kerr et al., 1960 ; Piore and Sabel, 1984 ; Zuboff, 1988); those that consider that there is an inherent dynamic for employers to seek to undermine skill levels in the interests of easier control and greater substitutability of labour (Braverman, 1974 ; Edwards 1979 ; Friedmann, 1946) and finally those who point towards an increased polarisation between a sector of the workforce that benefits from processes of upskilling and a sector whose skills are either stagnant or in decline (Berger and Piore, 1980; Gordon et al., 1982).
3Most sociological accounts see a close relationship between skill level and autonomy or task discretion. This has important implications for the wider quality of working life. A wide range of research has underlined the importance of task discretion not only for employees’ job satisfaction, but also for their capacity to handle work stress and their longer-term health (Karesek and Theorell, 1990 ; Theorell, 1998 ; Theorell and Karesek, 1996). Common to most perspectives is the view that higher levels of skill are associated with greater scope for employees to exercise discretion over the way that work is carried out. To the extent that relevant knowledge is with the employee, detailed control by management is thought to be largely counter-productive. However, this led to diametrically opposed visions of the long-term scenario for worker task control. Those who have pointed to a trend for upskilling have also emphasised the decentralisation of decision-making, while those who believe the main trend is for skills to be undermined link this to more intense monitoring and control of work performance.
4These different perspectives on long-term skill trends are partly rooted in different assumptions about the implications of developments in technology. While there has been sharp debate about the primary factors that affect skill patterns, diverse sociological accounts give a central role to technological change. In successive periods, there has been extended debate on the implications of the spread of automation and of computer technologies which are variously seen either as creating the conditions for reversing earlier trends towards an increased fragmentation and deskilling of work tasks (Blauner, 1964) or as displacing skills and providing employers with the means for much more pervasive forms of control (Naville, 1963). While in its early versions, the concern with automation focused primarily on its implications for manual workers, its effects on non-manual work have become increasingly central to discussions in recent decades (Crompton and Jones, 1984).
5Arguably one of the reasons why there has been so much disagreement about the broad trends in work skills and their determinants has been the lack of suitable evidence. In particular, there has been a lack of representative data and a lack of data that was truly comparable over time.  This paper sets out to describe a cumulative programme of research carried out in Britain since the mid-1980s that has sought to provide a stronger basis of evidence.  The programme has been elaborated in successive studies carried out in 1986, 1992, 1997 and 2001. It has seen the attempt to formulate a number of underlying indicators of skill complexity or level that can be used to monitor trends across time both for the workforce as a whole and for its major subcategories.
The meaning and measurement of skill level
6Methods of assessing skill trends must clearly depend on the particular purpose in hand. If the focus is upon change in particular occupations within specific industrial settings, then fine-grained ethnographic accounts of the nature of work tasks are irreplaceable (Darrah, 1996; Attewell, 1990). A particularly fine example of this tradition is the work of the French “school” associated with Georges Friedmann, whose accounts of the changing nature of work were rooted in in-depth analyses of specific occupations and workplaces.  Such studies have a depth of descriptive detail that provided a uniquely authoritative picture of the milieux studied. It is a tradition that has still a great deal to contribute to our understanding of how particular types of work are evolving, as shown for instance in the work of Piotet and her colleagues (Piotet, 2002).
7Friedmann himself was concerned to build upon such intensive case studies to develop long-term scenarios of the changing nature of skills in work. In the immediate post-war world, with its simpler manufacturing based economy, this was perhaps a viable strategy. It was possible to view certain types of manufacturing enterprise as probable models for the future development of work more generally. In particular, great attention was paid to developments in the automobile and steel industries, with their vast scale and their leading role in the development of mass production techniques. It was a research strategy that was partly premised on an underlying assumption that work would become more homogeneous over time. However, economic development in subsequent decades has made this type of extrapolation from exemplar industries increasingly dangerous. Rather than becoming more homogeneous, advanced economies experienced a rapid diversification of types of work, as they shifted from a predominantly manufacturing to a predominantly service base. One sign of this was the expansion of occupational titles over successive censuses.
8As case studies multiplied, the extent of disagreement between the conclusions even among studies concerned with similar types of work suggested that patterns of work organisation were highly conditional upon specific managerial approaches and there was no easy way of assessing which patterns had greatest prevalence.  They certainly could not be simply “read off” from specific technological conditions. At the same time the developing pattern of employment made the research practicality of seeking to generalise from myriad case studies ever more daunting. The growth of the service sector favoured the survival if not the expansion of small dispersed work units and an increasing proportion of the workforce was involved in activities that were geographically highly mobile rather than placed in a limited number of large-scale central locations. There are evident problems in hoping to use semi-anthropological techniques to provide a picture of trends in an ever more diverse and geographically mobile workforce.
9In short, there were strong reasons to try to develop a different approach to the study of skill trends and in particular to try to use the potential of large-scale random surveys to provide a representative picture of developments. However, where the objective is to establish the broader nature of trends through a common survey instrument, it is necessary to search for indicators that are applicable to a wide range of work situations. It requires trying to define some common dimensions that underlie the highly varied specific skills of particular jobs.
10While there might be underlying agreement that “skill level” refers to the “substantive complexity” of job tasks, in terms of the level, scope and integration of mental, manipulative and interpersonal tasks, there has been little consensus in the literature about what best indicates this (Spenner, 1979, 1983 ; Form, 1987 ; Spenner, 1990). An early approach was to rely on the evolution of the type of broad occupational classes that have been used for census classifications. However quite apart from the severe difficulties of establishing comparability of data given major ruptures in occupational classification schemes, this type of approach had to confront the objection that it was improbable that occupational classes represented a skill hierarchy that had stability over time. Rather, it was possible they reflected different bundles of tasks from one period to another as technological or organisational change led to the restructuring of work (Capelli, 1993). This was well analysed in the case of clerical employees: a substantial research literature showed that the marked expansion of clerical employment in the first half of the twentieth century was accompanied by a simplification of the work that was involved (Lockwood, 1958). This pointed to the need to look for indicators of task complexity that were independent of occupational titles. Given the constant change in the nature of jobs, adequate measures for the study of broad trends needed to be independent of specific types of job activity.
11The research programmes that are the focus of this paper started from the assumption that, at a very general level, the idea of relative skill level or task complexity is best proxied by the learning requirements necessary to develop the knowledge required to be able to do the job.  However, even those who have adopted this position have differed in their approach to the way in which it should be operationalised.
12In the economics literature the commonest indicator for skill is that of the person’s own education, in particular the number of years of general education. This has the considerable advantage of both simplicity and of the availability of representative data allowing some comparability over time. It has long been recognised that this is a far from ideal measure of individual human capital. But, it is a particularly problematic way of measuring job skill. It is an indicator that refers to the individual’s capacities rather than to the demands of the job as such. It ignores the problem of mismatch, namely that people with a given level of education might well be overqualified or underqualified for the specific position that they occupy. It is also a measure that focuses primarily on initial general knowledge acquisition, rather than the requirements for more specific occupational knowledge that might be acquired through training at later periods of time.  Some scenarios of the changing nature of work have suggested that skill requirements are increasingly becoming organisationally or technologically specific, resulting in the growing importance of knowledge that can only be acquired in the course of the work itself (Eraut, 2000; Eraut et al., 2000). At least three distinct types of knowledge acquisition have then to be taken into account: general education, vocationally specific training and knowledge acquired on the job. Any search for a single summary measure of skill level then is unlikely to be successful, as the types of relevant knowledge and the sources for acquiring them are too diverse.
13A significant strand of the sociological literature has also argued for the need to take autonomy, or task discretion, as an independent dimension of skill (Spenner 1983; Vallas 1990). This was particularly the case for scholars from a Marxian perspective who regarded the deskilling process as primarily involving the separation of the conceptual element of work from its execution, and the removal of the former from manual workers to management. However, it has the drawback as an indicator of skill that information about the freedom to make decisions tells us little in itself about the complexity of the decisions that are made. In principle, it is perfectly possible for people in relatively simple jobs to have high levels of autonomy. While the importance of autonomy for employee well-being is very well supported in the research literature, it does not necessarily capture well the qualitative dimension that underlies the notion of skill level. This debate is far from finished and researchers involved in the British research programmes discussed in this paper have adopted rather different positions on the issue. It is clear however that it would be unwise to seek to develop any synthetic measure of skill that combined both indicators of task complexity and indicators of task discretion. Rather these have to be carefully distinguished if only to assess how far they follow the same trends and how they inter-relate.
14The first study that developed the broad skills measures that informed successive programmes was the Social Change and Economic Life Initiative Survey carried out in 1986 (Gallie 1994). It was a comparative study of six local labour markets chosen to represent contrasting past and current economic conditions. However, subsequent analysis showed that, when aggregated, these approximated very closely to a national sample when compared with Labour Force Survey data. It has then been taken as the reference point for the national skill structure of the mid-1980s. The measures were subsequently included in three major representative national surveys. The first—the Employment in Britain survey—was carried out in 1992 (Gallie et al., 1998), and the second and third were the Skills Surveys of 1997 and 2001 (Ashton et al., 1999; Felstead et al., 2002). In each case, they were large-scale high-quality representative national surveys. The two “skills surveys” led to a considerable expansion of the scope of the research programme into skill development. Building upon job analysis methods, they were also concerned to assess the importance of specific types of skill in work and they provide a rich data base for assessing trends in generic skills.  For reasons of space, the present paper however is restricted to a discussion of some of the core findings that have emerged about trends in broad skills.
15The awareness of the inadequacy of any single measure of “learning requirements” led to the adoption of three principal measures designed to tap the learning requirements of jobs. The first was the qualification level required for the job. The wording of the question was specifically focused on the job itself rather than on the individual’s own qualifications. People were asked: “If they were applying today, what qualifications, if any, would someone need to get the type of job you have now?” A list of qualifications was presented and the highest qualification given was subsequently allocated to one of six broad qualification categories.  The second measure was concerned with the length of training time the person had received for the particular type of work they were currently involved in, asking people: “Since completing full-time education, have you ever had or are you currently undertaking, training for the type of work you currently do?” If people had or were receiving training, they were then asked “how long in total did (or will) that training last”? The third measure addressed the issue of necessary knowledge acquired directly on the job by asking: “How long did it take for you after you first started doing this type of job to learn to do it well?”
General trends in skill level 1986-2001
16What picture did these different indicators give of the broad trends in skill levels? Taking the first indicator—the qualifications required to obtain the job—there was a clear rise in skill levels for the overall period between 1986 and 2001 (Table 1). The proportion of people in jobs requiring the highest category of qualification (level 4 which covers degree and other higher education qualifications) increased from 20% in 1986 to 29% in 2001. The rise was particularly striking for degrees which rose from 10% to 17% in the space of 15 years. At the same time there was a sharp reduction in the number of the jobs which people could enter without any qualifications (from 38% in 1986 to 27% in 2001), in part attributable to the growing demand for people with level 1 qualification. The pattern for intermediary qualifications (levels 2 and 3) is less clear-cut. There was less change across the overall time period and the trend was less consistent between years. Comparing the earliest and latest time periods, there was a small increase for level 3 and a decrease for level 2. The overall pattern can be summarised through the construction of a “Required Qualifications Index”, with scores ranging from 0 (No qualifications) to 4 (level 4 qualifications). The index rises strongly between 1986 and 1992, declines a little between 1992 and 1997 and then rises again substantially between 1997 and 2001.
17The second indicator—training time—also pointed to a long-term increase in skill levels. There was a decline between 1986 and 2001 in the proportion of employees who required very short or no periods of training for their type of work. However, much of this was accounted for by an expansion of intermediate training times. Taking those with very long training durations (two years or more) the pattern is very similar in the earliest and latest time periods. Again this can be summarised with an overall index.  Over the full period there was a rise in the index from 2.01 in 1986 to 2.27 in 2001. However, the increase is not linear. The index rose steadily from 1986 to 1997, after which it declined somewhat, although still remaining higher than its level in 1992.
18Finally, the picture of a rising level of skill was confirmed by the third measure of “job learning time”. Because of filter restrictions in the early surveys, the comparisons here are only available for employees. The overall learning skill index shows a progressive increase from one year to the next: it rises from 3.30 in 1986 to 3.57 in 2001. Most of the change lies in the decline in jobs requiring quite short learning times. In 1986, 27% of all employees thought that it required less than a month after they started the work to know how to do their job well, whereas by 2001 this was true of only 20%. The proportion in jobs requiring very extensive (2 years +) on-the-job learning periods rose only slightly from 24% to 26% over the same period.
19The general picture provided by the three measures is then consistent that skill levels in Britain rose from the mid-1980s to 2001. In particular each of the measures shows a progressive decline in the proportion of low-skilled jobs. At the higher end of the skill scale, there is a marked rise in the level of qualifications required but little increase in the proportion of employees requiring very extended training or on-the-job experience times.
20This raises the issue of whether the increase in qualification requirements might reflect primarily a process of credentialism, in which employers are raising the qualification requirements for jobs even though there is little change in the tasks themselves. To examine this, a question was included in three of the surveys asking: “How necessary do you think it is to possess those qualifications to do the job competently?” The changing responses over time provide an indication of whether the rise in the demand for qualifications may have reflected credentialism. In general, approximately three quarters of respondents thought that their qualifications were “essential” or “fairly necessary” to do the job. However, among employees in jobs where qualification levels 1 and 3 were required, there was a decline over the years in the proportion who thought this, in the way that would be anticipated if credentialism were a factor. For those in jobs requiring level 3 qualifications, it fell from 77% in 1986 to 70% in 2001, with the proportion regarding them as “unnecessary” rising from 4% to 10%. For those in jobs requiring level 1, the proportions regarding these qualifications as important for carrying out their work fell by 15 percentage points, while those thinking they were unnecessary rose by 9 percentage points. However, there was no evidence of credentialism for the other two qualification levels—level 2 and level 4.
21Accepting that there is some evidence consistent with a process of credentialism, how far does this affect the picture of skill trends? It is possible to construct a measure of “qualifications used” by comparing between years only those who thought that their qualifications were either “essential” or “fairly necessary”. In cases where people reported that the qualifications required were not “essential” or “fairly important”, their jobs have been allocated to the next lower qualification level. This produces the pattern that can be seen in Table 2. The broad trends that were seen previously still emerge clearly, but the differences are less marked. There is still a considerable expansion of those requiring the highest qualification level and a decline in those in jobs where no qualifications are required. There was an increase in the proportion of jobs where people were required to use level 4 qualifications from 16% in 1986 to 23% in 2001, whereas the proportion in jobs without qualification requirements declined from 40% to 31% over the same period. There is a small but statistically significant rise in jobs where level 3 qualifications are required, but little change with respect to level 1 and level 2 qualifications.
22Finally, we can compare the picture that emerges from the skill level indicators with people’s reports about change in their own skills over time. Table 3 presents the results to a question: “I’d like you to compare your current job with what you were doing five years ago even though you were in [the same / a different] job... Would you say that there has been a significant increase between then and now, a significant decrease or little or no change in the level of skill you use in your job?” This provides a picture of individual’s experience of skill change that reflects both trend and life cycle effects.
23As can be seen in Table 3, a majority of people in each year reported that the level of skill that they used in their job had increased in recent years. The 1986 data are not fully comparable with that for the other years, since they are based on employees rather than all in work. But an examination of the surveys that included both employed and self-employed shows that taking the figures for employees leads to an overestimation of the proportion for the full workforce by only one percentage point. There appears to have been a particularly dramatic acceleration of upskilling between 1986 and 1992, after which there was a slight decline. In 1992, the peak year, as much as 62% of people in work reported an increase in their skills. In 2001 the figure was still as high 59%. In contrast, in no year did the proportion that experienced deskilling amount to as much as ten percent. The overall picture across the period then is of a pervasive process of skill development, although it was probably at its most rapid in the late 1980s and early 1990s.
24In summary, there is a striking consistency between the overall conclusions to be drawn from the measures based on learning requirements at different periods and people’s own perceptions of the pattern of change over time. The general trend has been clearly for skill levels to rise.
The polarisation of skills?
25While the general pattern conflicts with the most pessimistic interpretations of changes in work, which argued for pervasive deskilling, it is not necessarily incompatible with a process of polarisation in which the benefits of skill development accrue to particular categories of the workforce but exclude others. There are two main versions of the polarisation thesis. The first emphasizes the class-biased nature of skill development. Those who are in higher occupational classes tend to attract employer investment in in-career training, leading to rapid skill evolution. In contrast, there is a greater reluctance by employers to invest in skill formation for non-skilled workers, leading to skill stagnation or even skill depreciation. The second polarisation thesis focuses on contractual status. The workforce, it is argued, is increasingly divided between a core of relatively privileged employees, who have permanent full-time contracts and a periphery of relatively disadvantaged who are either on temporary or part-time contracts. Skill development, together with job security and pay, is one of the key dimensions that distinguishes core-periphery employment conditions.
Polarisation between occupational classes
26Taking the occupational classes represented by the first digit categories of the UK Standard Occupational Classification, there were four groups that stood out as particularly low-skilled in 1986 in terms of all the measures: personal service workers, sales workers, operatives and elementary workers. Between 1986 and 2001, personal service workers saw a marked rise on all skill measures, while the scores for the other three occupational groups remained relatively stable. On each of the skill measures, elementary workers had the lowest average index score of any occupational group, with a slight tendency for an absolute rise in the score over time for required qualifications (from 0.38 to 0.58) and for on-the-job learning time (from 1.86 to 1.96).
27A summary overview of the changes in class differentials over the period can be obtained by looking at the ratios of the skill index scores of those in either managerial or professional occupations to those in elementary occupations (Table 4). The measure that shows the greatest class differential is that of qualifications required for the job. For the period from 1986 to 2001 as a whole, the ratios point to a reduction rather than an increase in class differentials. But most of this change occurred in the period between 1986 and 1997, after which the differences increased again. With respect to the training time index, there is very little change overall in the class ratios. The differential declined a little with respect to managers, but widened compared to professional workers. There was some narrowing of differentials between 1992 and 1997, but they then widened again in the final period. Class differentials with respect to on-the-job learning time changed little between 1986 and 2001.
28How do the differences shown by the indices relate to the reports given by employees about their own experiences of skill change? These also showed a very clear difference between occupational classes. Taking first the proportions experiencing an increase in skills, it is notable that in each year 60% or more of managers, professionals, associate professionals/administrative and secretarial staff reported that the skill level of their work had risen in the previous five years. In contrast, this was the case for just under a quarter of elementary workers in 1986 and just over a third in the other three years. Taking the overall period 1986 to 2001 there was very little change in reported upskilling among managers and professionals (and the same was the case for operatives), but higher proportions reported a skill rise in 2001 among all other categories with a particularly sharp increase among personal service workers.
29The trend was not however linear. For six out of the nine class categories, the highest proportions reporting skill development were in 1992 and for the other three occupational classes they were in 1997. The process of skill acceleration appears then to have particularly characterised the period from the late 1980s to the mid 1990s; after which there was a decline, although in most cases the figures still remained higher than in 1986. With respect specifically to elementary workers, there was a twelve percentage point rise over the period as a whole in the proportion experiencing upskilling (from 23% in 1986 to 35% in 2001). This almost entirely occurred between 1986 and 1992, after which there was little change.
30The pattern of change with respect to deskilling was not simply the mirror image of that for upskilling. Certainly the risk of deskilling was considerably greater the further down the occupational class hierarchy. But, the fact that most occupational groups (including elementary workers) had seen an increase in the proportions experiencing upskilling did not imply that there had been a general reduction in the proportions experiencing deskilling. There had been a slight reduction at the higher end of the scale (among managers, professionals and associate professionals) and also amongst those in skilled trades. But among the least skilled groups—sales, operatives and elementary workers—the prevalence of deskilling rose between 1986 and 2001. The timing of this differed considerably between groups. Among sales workers it occurred in the mid-1990s, with the proportion rising from 12% in 1992 to 17% in 1997, a level at which it remained in 2001. Among operatives, it is noticeable mainly in the most recent period, rising from 10% in 1997 to 17% in 2001. For elementary workers, the rise mainly took place between 1986 and 1992 (from 15% to 19%) after which the proportion experiencing deskilling slightly declined again.
31Given that upskilling and deskilling trends do not simply reflect each other, it is most informative to derive a summary indicator of class differentials from a net upskilling rate, the proportion experiencing upskilling less the proportion experiencing deskilling. This shows an overall decline in class differentials between 1986 and 2001 (Table 4). Most of this decline was between the mid-1980s and the early 1990s, with a further small decline between 1992 and 1997. The pattern based on self-report skill change comes closest then to that given by the required qualification index.
Polarisation between core and periphery ?
32The second argument predicting skill polarisation derives from the literature on labour market segmentation. The key divide within the workforce is associated with differences in contractual status. The “core” consists of permanent full-time employees, who are protected by internal labour markets and have long-term relatively secure contracts. This implies both that employers will be especially demanding at the point of recruitment and that they will be willing to invest more readily in skill development through the working career. In contrast, the “peripheral” workforce consists of people employed on temporary or part-time contracts, who have limited job security and are likely to be given few offers of skill enhancement.
33Part-timers are defined here as people working less than 30 hours a week. It was certainly the case that in all of the years they had lower scores on each of the skill indices than full-timers. This was true for both male part-timers (although it must be remembered that for most years the sample numbers were very low) and for female part-timers. However, as can be seen in Table 5, if we take the ratios of the scores of full-time workers to those of part-timers as a measure of the contract type differential, there is no evidence of polarisation over time. Rather with respect to the three indicators—required qualifications, training time and on-the-job learning time—there was a steady convergence between part-time and full-time work. In each case the skill levels of full-timers rose over the period, but those of part-timers rose faster.
34Over the period there was an increase in the proportion of part-timers that were men. It could be argued that it is particularly female part-timers that are likely to be in a weak position on the labour market and the change in their position might be concealed by the changing composition of part-time work. However if we look at the ratio of the index scores for full timers to those of female part-timers, exactly the same pattern emerges. There is a steady decline in each time period in the extent of the contract-status differential for each of the skill measures.
35Turning to the subjective measure of skill change over the last five years, the pattern within each year confirms clearly the much lower opportunities that part-timers had for skill development. However for part-timers in general and for female part-timers, there was a substantial increase from the mid-1980s to 2001 in the proportion that experienced an increase in skills. The change was particularly marked in the period 1986 to 1992, but the trend continued until 1997. After that there was a decline in the proportions with upskilling between 1997 and 2001, bringing the level back to that of 1992 for all part-timers, although still remaining higher for female part-timers. Taking, as shown in Table 5, the ratios for the differential between full-timers and part-timers (and full-timers and female part-timers), the pattern based upon self-report judgements is consistent with that from the learning requirements indices from 1986 to 1997, after which they diverge with an increasing differential on the self-report measures and a continuing convergence on the learning requirements indices. But whichever measure is taken it is clear that part-timers were in a relatively better skill position in 2001 than they had been in 1986.
36A second type of contract difference that may differentiate skill formation experiences is that between workers on permanent and workers on temporary contracts. But the scenario of growing workforce polarisation is not well supported in this case either.  Whereas part-time workers were less skilled on each measure across the whole time period, this was not the case for temporary workers. It was only in the mid-1980s that permanent workers were more highly skilled on each of the three indicators. In particular, it was only in 1986 that there was a clear differential with respect to qualification requirements. Taking the trends over time, there is no consistent evidence of polarisation. As can be seen in Table 6, the differential between permanent and temporary workers declined on all three indicators between 1986 and 1992. It then stabilised for required qualifications and rose again for training and on-the-job learning times, without reaching the figure for the mid-1980s. The same u-shaped pattern is evident in the indicator derived from the self-report skill change measure. From the mid-1980s to the early 1990s, there was a marked decline in the overall differential in skill experience between permanent and temporary workers, while in the period between 1992 and 2001 there was a smaller increase in the differential.
37In short, whether one takes the case of part time workers or of temporary workers, the view that the long-term trend has been one of workforce skill polarisation receives relatively little support from the empirical data. In terms of the skill level measures, there was a clear convergence over the period as a whole between the skills of full-time and part-time workers. There was also a convergence between permanent and temporary workers on two of the three skill level measures—the exception being the on-the-job learning measure. The general trend to convergence between 1986 and 2001 is supported by the self-report skill change evidence. There appears to have been some reversal of this trend from the early 1990s, but it was not sufficient to restore the extent of difference between contract groups that existed in the mid-1980s.
38As mentioned earlier, task discretion is sometimes seen as a component of skill and sometimes as an independent characteristic of the work task. However, even those who adopt the latter position have tended to see the two as very tightly associated. The argument is generally that higher levels of skill are accompanied by greater task discretion, as employers find it more difficult to monitor performance in detail and are more concerned to encourage discretionary effort. Given that the general trend has been for rising levels of skill, has this been accompanied by rising task discretion, as would be expected on the basis of much of the literature?
39The data do not provide a consistent series across the four time periods. However, the surveys for 1992, 1997 and 2001 each contain four questions in which people were asked to assess how much personal influence they thought they had over specific aspects of their work: how hard they worked, deciding what tasks they were to do, how the task was done, and the quality standards to which they worked.  These questions have the virtue, compared to some indicators found in the literature, that they are both fairly precise in reference and allow a reasonable range of response variation.
40Earlier work (albeit using less satisfactory indicators) pointed to a rise in task discretion from the mid-1980s to the early 1990s (Gallie et al., 1998). However, the evidence for the 1990s points not to a rise, but to a decline, in employee task discretion. The decline in control was sharpest with respect to work effort and quality standards. For three aspects of task control—control over work effort, which tasks to do and how to do the task—the decline was continuous between the three surveys, although control over work effort declined particularly sharply between 1997 and 2001. With respect to control over work quality, the change occurred almost entirely between 1992 and 1997, with little further change between 1997 and 2001. To provide an overall picture from these items, a summary index was constructed by giving a score ranging from 0 (no influence at all) to 3 (a great deal of influence) and then taking the average of the summed scores.  The index score for task discretion declined from 2.43 in 1992 to 2.25 in 1997 to 2.18 in 2001 (Table 7). The decline in the overall task discretion index was very similar indeed for both sexes. For men, it was reduced from 2.43 to 2.19 between 1992 and 2001, and for women from 2.44 to 2.17.
41As with skill levels, it is important to examine how far such patterns were general across the workforce or reflected diverging patterns by occupational or contractual category that might point to a polarisation of employment conditions. Job control is certainly strongly related to occupational class position. The task discretion index ranged from 2.58 among managers and administrators to 1.86 among plant and machine operatives. But the decline in job control occurred across all occupational classes. There were variations in the extent to which this was the case. Professionals, personal service and sales employees and those in elementary occupations experienced a particularly substantial decline in task discretion over this period.
42The ratios of the scores of managers on the one hand and professionals on the other to those of elementary workers can again be taken as indices of polarisation (Table 8). Taking the period 1992 to 2001 as a whole there is some support for the view that there were increasing differences between the top and bottom of the class hierarchy. For both the comparisons with managers and with professionals, differentials were increasing between 1992 and 1997. This trend continued between 1997 and 2001 for the contrast between elementary and managerial workers. But the differential between elementary workers and professionals declined in the second period, although still remaining higher than in 1992. Turning from class to contract status divisions, there was a small increase in the differential between full-time and part-time workers, and a rather more substantial rise in the differential between permanent and temporary workers.
43In contrast to the case for the skill measures, there is then rather more consistent evidence of polarisation with respect to task discretion, at least for the 1990s. With respect to the comparisons between managers and elementary workers, between full-time and part-time workers and finally between permanent and temporary workers, differentials in task discretion had become somewhat greater in 2001 than they had been in the early 1990s.
New technology, skill trends and task discretion
44How has the rapid growth of new computer-based technologies affected the trends with respect to skill and task discretion? Much of the early debate about the consequences of automation for work patterns was conceived in the context of the transformation of mass production techniques and their impact upon manual work. However, a central feature of computerisation since the 1980s has been its multi-faceted character and its pervasiveness. The range of indicators available on a continuous basis from the surveys is decidedly restricted, but it is possible to make certain limited comparisons.
45Our broadest and longest trend indicator on the use of advanced technology in jobs is a question that asks people: “Does your own job involve use of computerised or automated equisetum?” This was asked in the Social Change and Economic Life survey of 1986, the Employment in Britain Survey of 1992 and the Skills Survey of 2001.
46As can be seen in Table 9, there has been a rapid and continuing expansion of the use of computers and automated equipment in work over the period from 1986 to 2001. The fullest series of comparable figures relates to employees. In 1986, 40.3% of employees used such equipment, whereas in 2001 the proportion had risen to 73.7%. The trend for all people in work can be compared between 1992 and 2001. The proportion using computerised or automated equipment rose from just over half of all workers in 1992 to nearly three quarters in 2001. There also has been a marked convergence between men and women in the use of advanced equipment. In 1986 there was a gender gap of 13 percentage points. This fell to 5 points in 1992. By 2001 the gap had disappeared, with women now at least as likely to be using such equipment as men (74% compared with 73%).
47In every year, the use of advanced technologies varied substantially depending on a person’s occupational class. It has been most common among clerical and secretarial employees and among professionals, followed by managers and administrators and associate professionals. In contrast, even in 2001, only half of plant and machine operators used computer-based equipment and less than half of those in skilled trades, personal and protective, and elementary occupations. The growth in use over these years affected all occupational classes. However, it was generally stronger among those in higher skilled than among those in lower skilled occupational positions. The major exception was the very sharp rise among those in sales occupations.
48It is also notable that there was a very substantial difference between women in full-time and women in part-time jobs. Women in full-time jobs were substantially more likely than men to be using computerised or automated equipment, whereas the reverse was the case for women in part-time jobs. While both female full-timers and part-timers substantially increased their use of new technology, there has been little change in the “contract” differential between 1986 and 2001. Temporary workers were also less likely to work with new technologies than permanent workers, but again the difference between the two was almost identical in 1986 and 2001.
49A further question helps to explore whether computing has not only come to affect a wider range of jobs, but also has become more important to the nature of tasks. In both 1997 and 2001, people were asked how important “Using a computer”, “PC” or other types of “computerised equipment” was to their job. Taking those who said that the use of such equipment was either “essential” or “very important” as an indicator of the centrality of computer skills to the work task, there was a marked growth (9 percentage points) in work where computing activities constituted a central component of the job (Table 9, last row). In 2001 approximately 40% of all people in employment reported that the use of computing equipment was “essential” and a further 15% that it was “very important”. Women were a little more likely than men to consider it essential (41% compared with 39%), but again the much sharper divide is between women in full-time employment and women in part-time work. Among the former, 50% reported that the use of such equipment was essential to their job, whereas among the latter the proportion was only 29%.
50Our broad measure of the prevalence of the use of computerised equipment also covers a wide range of tasks of very different levels of complexity. To what extent has the growth been primarily in terms of routine types of computer use as against more advanced use? To address this issue, those who used computers were given a set of statements about possible types of use and asked which best characterised their own job. The four broad types of use given were: “Straightforward” (for example, using a computer for simple routine procedures such as printing out an invoice in a shop); “Moderate” (for example, using a computer for word-processing and/or spreadsheets or communicating with others by e-mail); “Complex” (for example, using a computer for analysing information or design, including use of computer aided design or statistical analysis packages); and “Advanced” (for example, using computer syntax and/or formulae for programming).
51As can be seen in Table 10, the most frequent type of computer use in 2001 was at a “moderate” level of complexity (46%). The next most frequent category was “simple” use (31%). This general pattern was the case for both men and women, although not for female part-timers who were predominantly in the simple use category. There were smaller proportions involved in complex (17%) or advanced (6%) types of use. But, the marked growth of use of computerised equipment between 1997 and 2001 was not accompanied by a substantial change in the relative importance of more complex and simpler types of use. In 1997, 23% of those who used such equipment were “complex” or “advanced” users, while in 2001 the proportion was 24%. At the other extreme, there had been some decline in the relative importance of “simple” use (from 38% to 31%), and some growth in use at “moderate” levels of complexity (from 39% to 46%).
52In short, whether we take a simple indicator of the use of computerised technology in the job or its importance to the task, there is a consistent picture of the increased importance of such technologies for work tasks and work experiences. There is no evidence that this growth was accompanied by a particularly marked expansion of the simplest types of such technology.
53Turning to the implications of computerised technologies for skills and discretion, Table 11 presents results from a series of regression analyses in which sex, age, employment status and occupational class are controlled. The broadest indicator of advanced technology has been used for the analysis, that is to say whether or not a person uses computerised or automated equipment in their job. A first point to note is that there is a strong positive effect for each of the skill measures in each of the three years for which there is fully comparable data. The coefficient is particularly strong with respect to required qualifications and training time, but the pattern also points the same way for on-the-job learning time. Similarly there is a strong positive effect of use of advanced technology on reported skill increase over the previous five years. However, the results with respect to task discretion are quite different. Here we find no overall effect of computerised technology in either of the two years for which there is comparable data.
54The conclusion must be that advanced technology has been consistently linked to higher skill and that the strength of the effect has been broadly similar across time. However, there is no evidence for the view that it has been a significant factor conducive to higher or lower levels of task discretion. While skill developments may be in part a response to technical imperatives, the control that people have at work is more likely to be primarily a function of managerial decisions about organisational design.
55Trends in skill have been a central component of many sociological theories of social change and they have been particularly important to theories of change in patterns of work organisation. Yet the evidence underlying many of the key claims about skill developments has been precarious, often involving generalisation from a limited numbers of specific work situations. This paper has discussed a series of British research projects that have sought to provide a stronger foundation for claims about the development of broad skill levels. It is research that is still very much in progress and the results presented here should be seen just as an early stage of a longer-term programme of analysis. They are clearly open to much finer levels of disaggregation and to detailed statistical analysis of the factors underlying skill patterns. In later studies, the scope of enquiry was broadened to include the issue of changes in specific generic skills, but the discussion here has been restricted to evidence about broad skills for which there is the longest time sequence.
56The core methodological development in the programme was the attempt to achieve a representative picture of the distribution of skills through the use of large-scale random surveys. This inevitably involved a search for a number of key underlying indicators of skill that could be applied to people in highly diverse work situations. The ultimate choice was to focus on the learning requirements for carrying out job tasks, distinguishing in turn between the qualifications, the training time and the on-the-job experience required to do the job. Comparable data have been collected by different studies at four different time points—1986, 1992, 1997 and 2001—making possible a first assessment of longer-term trends in broad skills.
57The results provide a consistent response to the extended debate between those who have argued that the long-term trend is towards upskilling and those who have advanced a pessimistic scenario of deskilling. They point very clearly towards an overall tendency for a rise of skill levels over the period. This is confirmed by the self-report evidence on people’s experience of change in their own skills over time. Moreover, there is little support for analyses of the labour market in terms of skill polarisation—whether with respect to occupational class or to contractual status. There appear to be good grounds for the view that new technologies have been an important factor in enhancing skill levels. The pervasive spread of computer-based technologies is clearly depicted by the data and these technologies are strongly related both to indicators of skill level and to self-reported skill change.
58However, the results also produced a major puzzle, namely with respect to trends in task discretion. The literature has usually seen trends in skill level and trends in task discretion as integrally linked. This was the case whether the long-term trend was seen to be one of upskilling or of deskilling. However, in practice, the general tendency for skill levels to rise was not accompanied by increased employee task discretion but rather by a reduction in the control that people felt they had over the way they carried out their work. There are different possible interpretations of this pattern. It might be the case that individual task discretion has given way to team decision-making, although some research provides rather sceptical conclusions about the real prevalence of semi-autonomous teams. While higher skill is one of the influences on task discretion, there are clearly other quite distinct processes that influence management’s decisions about job control. These may reflect managerial organizational philosophies that vary across time periods. They may be related to the emergence of tighter national regulations and higher levels of accountability, for instance with respect to the quality of products and services or health and safety standards.
59In exploring further the factors that affect trends with respect to both task complexity and task discretion, the future research agenda must clearly be one with a strong comparative element. There is a growing literature on the dynamics of different types of skill regimes, which are thought to vary between countries as a result of distinct institutional complexes. But an assessment of these arguments depends crucially on good comparative data on skill trends. Similarly if task discretion is affected by historically formed philosophies of management or by the nature of external regulative systems, this can only be examined through rigorous comparative analysis, which requires data that provide both a representative picture within countries and comparability across time.
A point made some time ago by William Form (1987): “Firm conclusions about skill degradation must await time-series analysis of national surveys that measure components of occupational skills in different industries” (p. 29).
A parallel programme to chart skill trends over time has been carried out by Swedish sociologists, using indicators from the Swedish level of living surveys (Jonsson, 1998).
The locus classicus of Friedmann’s views is (Friedmann, 1946). Another major research programme using such an intensive case study approach was conducted by Naville and colleagues (Naville et al., 1961).
Compare for instance the sharply diverging conclusions about developments in the British oil and chemical refining industries, despite the fact that the research was carried out at the same period (Gallie, 1978; Nichols and Beynon, 1977).
Compare William Form who argues that, in the absence of objective and representative skill measures of jobs: “Probably the best indicator is the total preparation time a job requires for an average worker to attain an average level of performance” (Form, 1987). Interestingly, a careful comparison of self-report job measures and so-called objective measures collected from work sites by trained occupational analysts suggest they correspond well (Gerhardt, 1988). Spenner (1990) also concludes that overall empirical studies support the view that respondents are relatively accurate reporters.
The training element has been particularly important in the French literature. While Georges Friedmann and Pierre Naville had rather different underlying conceptions of skill (substantialist vs proceduralist) they shared a common emphasis on the length of institutionalised training time as an indicator of skill level (Piotet, 2002).
This extended the range of indicators to explicitly capture decision-making and problem-solving skills, bringing British research a little closer to some of the preoccupations of analysts of competence in France since the late 1980s, for useful overview discussions see (Dugue and Maillebouis, 1994; Stroobants, 1993).
These are broadly: no qualifications, poor lower secondary, lower secondary, upper secondary, non-degree higher education and degree-level higher education.
The index was constructed with the following scores: less than a month of training = 1; less than three months = 2; three to six months = 3; six to twelve months = 4; one to two years = 5; and over two years = 6.
In 1986 and 1992, people were asked: “Do you think that your job is considered by your employer to be: a temporary job lasting less than 12 months; a fixed term job lasting between 1 and 3 years or a permanent job with no fixed time of ending?” The first two categories are taken as temporary work for the analysis. In 1997 and 2001, people were asked: “Leaving aside your own personal intentions and circumstances, is your job: a permanent job or, is there some way that it is NOT permanent?”
The question format was: “How much influence do you personally have on... how hard you work; deciding what tasks you are to do; deciding how you are to do the task; deciding the quality standards to which you work?”
The index was statistically robust, with an overall alpha of .78.