Research context
This book has its origins in the research project The Making of the ‘Precariat’: Unemployment, Insecurity and Work-Poor Young Adults in Harsh Economic Conditions (ESRC ES/K003755/1) that was funded via wave 1 of the Secondary Data Analysis Initiative (SDAI) of the Economic and Social Research Council (ESRC). The particular conditions of this initiative sought to prioritise the analysis of existing data resources, data that had already been created through past research funded by the ESRC itself, governments or other bodies. The scheme resonated directly with the previous work of Furlong (see, for example, Furlong 1993) and Goodwin and O’Connor (2015), where data re-use, secondary analysis and the repurposing of existing data were central to the analyses being offered. Regarding our own analytical and epistemological proclivities, the SDAI appealed to us directly for a number of reasons. First, as we have argued elsewhere (see O’Connor and Goodwin 2013), for us the secondary analysis and re-use of existing and/or legacy data is undergirded with an ethical imperative for social researchers to make full use of data previously collected. As we have argued,
In these times of reduced social science research funding, is it good, ethically sound research practice to continue to create unique one-off cross-sectional studies, when data already exists on young people − respondents about which we already know so much? Indeed, should not more use be made of the existing data already collected on youth to avoid intrusion and duplication?
(O’Connor and Goodwin 2013: 204)
Where data already exist, where information in already known and where research and respondents have already invested heavily in the co-creation of data, as social scientists we are duty bound to make the best and fullest use of that data.
Second, despite the apparent ‘age’ of some archived data, existing data retain significant analytical potential and can be re-considered via a more ‘contemporary conceptual lens’. Put simply, when previous researchers have collected data they have done so with their own analytical and theoretical concerns at the very heart of the data collection process. This means that they would inevitably prioritise certain themes, emphasise particular correlations or highlight particular findings key to their analytical narratives. Such analysis would also usually be offered in line with the dominant intellectual trends and discourses of that time. For example, in the analysis of classic school-to-work transitions studies, past scholars would prioritise structural determinants of school-to-work transitions, according to dominant class narratives, rather than explore the individual complexities of the transition process itself (see Goodwin and O’Connor 2005). Likewise, the practice of conceptualising labour market positions as a dichotomy between employment and unemployment, reflecting traditional, hegemonic notions of work rather than exploring the complex, ever-changing sites of youth employment or the ‘shades of grey’ (neither employed or unemployed) roles increasingly occupied by young people. Revisiting past and existing datasets, with different analytical concerns to the originators of the data, affords us an opportunity to develop different insights not previously considered and to develop a greater understanding without the need to return to the field to collect new data.
Third, we have been variously associated, as students, colleagues and co-researchers, with the some of the originators of past classic school-to-work transition studies and, as such, the opportunity to revisit some of this data as part of the project was an opportunity too good to miss. It was a chance to revisit some of the intellectual antecedents of our own approaches to understanding school-to-work transitions and would enable us to re-examine some of our own assumptions around, and understandings of, these past studies.
The making of the ‘precariat’
In responding to the SDAI call, we developed an innovative programme of research that combined an important contemporary dataset, Understanding Society, with two ‘historical’ or legacy datasets from the 1980s, exploring both the contemporary and legacy data with the same research questions. The overall objective was to undertake a secondary analysis of this contemporary and legacy data to answer the research question, ‘In what ways have the experiences of unemployed, insecure and vulnerable 18−25-year-olds changed between two key periods of economic instability in the UK?’ In answering this question, we also aimed to
In so doing, we hoped to gain a better understanding of the early labour market experiences of young people in difficult economic circumstances and help pave the way for more effective policies in the future. However, it is important to note that in terms of research design, only the secondary analysis of existing data would enable us to answer these questions we had set. We could have committed to undertake new field research and interviewed those who entered work, further education (FE) or the various youth training schemes that existed in the 1980s, but in so doing we would have had to rely heavily on the individual respondents’ abilities to recall and reconstruct their experiences from over 30 years ago. Having access to school-to-work data, contemporaneous with the timeframe of concern, offered a level of analytical potential and accuracy (or so we hoped) that could not be achieved via a new retrospective focused cross sectional study. We used three datasets for the analysis:
(i) Young Adults in the Labour Market (Legacy data). The first study, partial data from which was archived in the UK Data Archive (reference UKDA Study 2664), is the Young Adults in the Labour Market (YA) from 1983. This research was led by Professor David Ashton and involved face-to-face, semi-structured interviews with a sample 18- to 25-year-olds carried out in four contrasting labour markets in 1982−3. The four areas, Leicester, Sunderland, St Albans and Stafford, were selected to represent a range of employment conditions (see Ashton et al. 1982; 1986) and resulted in an achieved sample of 1,786 young people.
Three of the local labour markets were to be Leicester, Sunderland and St Albans, due to their contrasting industrial and occupational structures and levels of unemployment, and the knowledge the researchers had of these areas through their earlier work. The fourth locality (Stafford) had to be a readily identifiable local labour market with a high proportion of white collar workers.
(Ashton et al. 1982: 13)
(ii) Changing Structure of Youth Labour Markets (Legacy data). The second study is the Changing Structure of Youth Labour Markets (CSYLM) study. This was led by Professor Ken Roberts and involved face-to-face, semi-structured interviews with 854 individuals aged 17–18 in 1985 in three contrasting labour markets: Liverpool, Walsall and Chelmsford (see Roberts et al. 1986).
The aim of this research was to assess the interactive effects of changes in demand and supply, plus state interventions, for the operation of Britain’s youth labour markets in the 1980s. The fieldwork was conducted in Chelmsford, Liverpool and Walsall. These areas were deliberately selected for insights in to the main variations behind the national picture.
(Roberts et al. 1986: 1)
Both of the legacy projects were funded by the Department of Employment in the early 1980s and the results published initially in the Department for Employment Research Paper series (paper numbers 55 and 59, respectively). As such, these studies correspond to the last major recession, during which youth unemployment reached levels unseen since the 1930s. Both of these studies covered severely depressed labour markets (e.g. Liverpool and Sunderland) as well as areas that remained fairly buoyant (e.g. St Albans and Chelmsford) during the broader economic decline of the 1980s. Both studies focused on young people with limited post-compulsory educational experiences (YA excluded those in full-time education post-18 and those with degrees, CSYLM focused on minimum-aged school leavers). Neither the YA nor the CSYLM studies have been subject to secondary analysis since the data were collected in the mid-1980s. For example, a review of the records relating to YA from the UK Data Service reveals that data from this study have not been used since Ashton deposited them in the mid-1980s, nor are there any subsequent publications listed as using this data beyond those written by the original research team. Although some of the data from this research were archived, all of the original interview schedules, including detailed job summary sheets for each respondent, were archived in a storage room at University of Leicester’s Centre for Labour Market Studies. These remained largely undisturbed for over 30 years. The data from the CSYLM study were never deposited in the UK Data Archive but were instead donated to, and also came to be housed at, the Centre for Labour Market Studies. They was originally donated to a researcher in the centre with the intention of the data being used for a PhD project. Although they were not used for that purpose, this inadvertently, but fortuitously for us, meant that all the original interview schedules were retained – something quite unusual in contemporary research practice.
(iii) Understanding Society (Contemporary data). The contemporary dataset to be used is Understanding Society, the enhanced replacement to the British Household Panel Survey. Described as ‘the largest single investment in academic social research resources ever launched in the UK’, Understanding Society follows approximately 100,000 individuals in 40,000 households on an annual basis. In this study, we draw on waves 1 and 2, filtered to include only those individuals in the age range 18−25 (up to 6,500 individuals in wave 1, in excess of 3,000 in wave 2). These individuals, who were interviewed in 2009−10 (wave 1) and 2010−11 (wave 2), were asked a variety of questions relating to educational experiences and patterns of labour market engagement as well as covering subjective aspects of experiences in education and the labour market and plans for the future. The datasets contain detailed information on home circumstances and parental position and there are linkages to administrative data for school attainment. While designed as a longitudinal study, our analysis plans were largely confined to cross-sectional analysis, with the two waves being used in order to be able to draw on a broader range of questions (some questions are not asked in both waves). However, some longitudinal analysis was necessary to explore routes out of precarious positions.
Analytical approach
As outlined in the previous section, the legacy studies focused on specific local labour markets, while Understanding Society offers data relating to a national sample. However, it was never our intention, via the secondary analysis, to use the regional data from the legacy studies as a ‘proxy’ for the national situation in the 1980s. Nor was it ever our intention to draw out specific regions from the Understanding Society data for direct comparison with the historic data. Indeed, the numbers would not sustain this level of analysis in any meaningful way. Instead, we focused on typologies of employment. In both the CSYLM study and YA study datasets, the areas covered by the survey were selected to represent labour market typologies; as highlighted already, Ashton indicated in his report that the labour markets were deliberately chosen to represent different local labour market conditions. At a broad level, these labour markets could be categorised as chronically depressed, declining, economically stable or prosperous, and our initial analysis took these typologies as a starting points. To be clear, even if we had access to a contemporary dataset that provided good coverage of the same towns as those used by Ashton and Roberts, this was not an approach we wanted to take. As such, our approach in analysing Understanding Society was to construct the same typologies: typologies that could be constructed using a number of areas that meet the core criteria. In this way, we were able to utilise the numerical power of a national dataset without sacrificing sections of the country and losing the power of the data. For example, the legacy studies offer a combined workable sample of 2,640 individuals. Filtering those individuals in the age range 18−25 from Understanding Society gives us 9,500 individuals from waves 1 and 2. This provided a significant amount of data to achieve our stated research aims of capturing the changes in this landscape of precarious working and highlighting the experiences of young people contained within it in two periods of instability.
The types of questions asked in the CSYLM and YA studies broadly mirrored the questions asked in Understanding Society. This facilitated the exploration of key themes around precarity and the labour market experiences of young people in two key periods of labour market and economic instability. Illustrative themes can be summarised as
Legacy study data ‘re-use’: Practical issues and concerns
‘Practical’ concerns
In the previous section we outlined the broader intellectual case for secondary analysis and data re-use. However, while these broad aims are what we ‘aspired’ to, there is a ‘lived reality’, both practical and analytic, in repurposing data from legacy research projects. We have offered a fuller discussion of this elsewhere (see, for example, Hadfield et al. 2015), but it is useful here, and instructional to others, to review three areas of concern – these were organisational, definitional and interpretational issues. First, organisational issues. The interview schedules from the legacy projects had been stored in cardboard packing boxes for 30 years in a secure ‘outhouse’. While the interview booklets were largely in good order, save for the occasional spotting of damp or mould, the schedules were jumbled up and, in some cases, the job summary forms had become detached. Some of the packing boxes also contained interview schedules from both the CSYLM and YA studies. As such, we began the large undertaking of reassembling the interview schedules, arranging them by project and sorting them by sequence number or unique identification number derived from the administrative data inside the questionnaire. These were then sorted into the relevant local labour market. Identifying the questionnaires by research study for the majority could be done on appearance, as one of the studies had a distinct booklet design. Once this process was complete, the data were entered into a bespoke database built in FileMaker Pro dataset. While this all may seem like ‘humdrum administrative’ research, it is important to note the organisation issues for two reasons. First, it took a considerable amount of time to organise the data, indeed far more time than we had expected. Second, without committing time to organising the booklets, sorting the schedules and entering the data into a database, we would not have had such an immense data resource.
‘Legacy’ datasets |
Understanding Society |
Personal and family characteristics: age at interview, gender, ethnicity, domestic circumstances, parental occupation and education, receipt of statebenefits, income |
Personal and family characteristics: age at interview, gender, ethnicity, domestic circumstances parental occupations and education, receipt of state benefits (jobseeker, incapacity, tax credits), income |
Education: age at leaving school; highest qualification, qualifications obtained, post-school education; attitude to school/leaving, intentions and experiences of FE/higher education (HE) |
Education: age at leaving school, highest qualification, vocational qualification, qualifications obtained, intentions and experiences of FE/HE |
Employment: employment history, occupational aspirations, jobs held, hours worked and overtime, SOC for first job and current job, occupation at first and current job, permanent/temporary, self-employment, job tasks, size of organisation, job satisfaction, managerial/supervisory roles |
Employment: employment history, jobs held (length of time), hours worked and overtime, SOC for first job and current job, occupation at first and current job, permanent/temporary, self-employment, job tasks, size of organisation, job satisfaction, managerial/supervisory roles |
Unemployment: periods of unemployment, reasons for unemployment, benefits and support, job seeking behaviour and method, problems encountered |
Unemployment: periods of unemployment, reasons for unemployment, job seeking behaviour and methods, experience of discrimination |
Training: access to training attitudes to training, types of training received, number of training periods, training providers |
Training: access to training, discrimination in access to training, vocational qualifications, training providers |
Precarious work forms: full-time/part-time, non-standard working, pay and rewards, pensions, trade union membership, job stability, fixed-term employment, agency work, casual employment, low working hours, participation in government schemes, degree of autonomy, responsibility for others, engaging work |
Precarious work forms: job security, full-time/part-time, non-standard working, pay and rewards, pensions, fixed-term employment, agency work, seasonal and casual employment, low working hours, participation in government schemes, degree of autonomy, responsibility for others, engaging work, receipt of state benefits |
Second, definitional issues. The interviews were undertaken during the early to mid-1980s and so, inevitably, the schedules contained definitions and general terminology specific to work and employment and to the youth labour markets of that time. Furthermore, the interview schedules included specific social, political and cultural references that would make little sense to anyone born post-1986. For example, it is relatively easy to find a definition for YTS (Youth Training Scheme) but what is much more complex is that YTS is used to describe a scheme that had multiple variations and pathways throughout. Without extensive scouring of past texts and policy literature, it would have been impossible to define appropriately some of the key terms used in the research. However, more problematic were the words used that no longer reflect acceptable research practice or the use of racist or sexist terminology. Although relatively rare in the data, reading respondents being defined by those in the field as ‘coloured’ was both troubling and analytically problematic, given that such a meaningless term obscures the ethnic identity of the respondent totally.
Finally, interpretational issues. Although we have all of the original interview schedules relating to both the CSYLM and YA studies, in the case of the CSYLM we did not have a codebook or data dictionary. For YA, we had the data dictionary for the archived component of the dataset, but we did not have any materials to help decode the employment component sheets that had not been archived previously. It was clear in both studies that a combination of fieldworkers, the research team and those doing the analysis had already annotated the schedules and applied specific codes to responses and outcomes. We could transcribe the ‘raw data’, but we also needed ‘decode’ the already coded data (Hadfield et al. 2015). For example, using the YA study employment component sheets to illustrate the issues we encountered, while codes such as ‘j1’, ‘j2’ may be relatively easy to interpret in the context of an employment study (job 1, job 2 and so on), other codes, such as ‘LH’, ‘C’, ‘CH’ and ‘M’, among other handwritten annotations, were less than transparent. The only way we could resolve these ambiguities with any sense of certainty was to interview members of the original research teams so that we may ‘recreate’ the codes from memory. However, even this was less than straightforward as there are multiple instances in the legacy studies where multiple and different codes, applied by different creators/users of the data, are applied to the same question and response.
Race and ethnicity in the 1980s datasets
Neither of the 1980s projects focused attention on race and ethnicity as a key variable in understanding youth employment in the 1980s. This is, in part, a reflection of sociological practice in the early 1980s, although it is important to note that in both studies there was an awareness of the potential danger of omitting such a key variable from the research design.
In the case of Ashton et al. (1986), a note is included in the report to explain how and why ethnicity was excluded from the study. They note that of the labour markets included in the research, Leicester was known to have a demographic profile that included a high percentage of young people from minority ethnic backgrounds. To account for this, the researchers explained that they
agreed that those districts with known high concentrations of ethnic minority groups, predominantly in the inner city, were to be avoided. It was felt that their inclusion would given an unnecessary bias to the sample for a project which was not focussing on race or ethnicity as central issues.
(Ashton et al. 1986: 13)
This is not to say that during the early 1980s the experience of black and Asian youth in the city was being ignored. A number of studies were carried out that focused exclusively on transition from school to work of minority ethnic young people in the city (e.g. Brah and Golding 1983), but there was little evidence of ethnicity being mainstreamed.
In common with Ashton et al. (1986), Roberts et al. (1987: 137) did not include ‘acceptable and meaningful questions’ on ethnicity as part of their project. However, they did identify that their sample included a small number of black and Asian young people in two of the selected labour markets (Liverpool and Walsall), and they provided some limited analysis of the data they had. They contrasted the experience of the small number of minority ethnic young people with the wider sample and acknowledged that the differences that emerge were of interest and certainly worthy of greater attention, suggesting, for example, that the impact of discrimination on the minority ethnic youth in their sample would likely increase ‘as the young people move from relatively sheltered educational environments into the labour market’. However, given the small sample size, little attention is paid to this issue in the main study.
We are conscious that the data we have been able to access from the legacy studies did not account for race and ethnicity as a significant variable, and this omission conflicts with current practice. However, with so little data on race and ethnicity from the 1980s available to us, we have not used this as a variable used for analysis. This is a limitation of the study that we acknowledge.