Attitudes to work and time spent unemployed across 30 years

Empirical studies have not previously related people's work attitudes to their chances of being unemployed for a substantial proportion of their lives. This study uses 1958 and 1970 British birth cohorts to show that responses to attitude questions offering a choice between an unattractive or disliked job and joblessness rival established unemployment risk variables as predictors of time spent unemployed between ages 16 and 46. The findings may help justify the provisions of the Welfare Reform Act 2012 for compelling unemployed benefit claimants to apply for, and subsequently retain, jobs they would otherwise have ruled out.

reapply for Universal Credit (s. 49).Heralding the Act, its architect Iain Duncan Smith, then Secretary of State for Work and Pensions, suggested that significant numbers of unemployed benefit claimants were too selective in the jobs they were willing to apply for.He referred to a television documentary in which some would not get "on a bus" to a nearby city to widen their job search (BBC, 2010), and later claimed that UK companies had been "unable to get British people to fill" some job vacancies, so "workers from overseas stepped in" (Duncan Smith, 2012).
Yet in keeping with a long tradition of social policy academic writing on unemployment (see, for example, Walker & Howard, 2000), leading commentators continue to challenge government claims that tougher policies are needed by presenting evidence that unemployed benefit claimants are strongly committed to employment.A joint article by researchers from two major qualitative longitudinal projects argued that government attempts to ensure claimants try to enter employment were "experienced as unnecessary" (Wright & Patrick, 2019, p. 597) by the claimants they studied, as "their existing job seeking behaviour is well matched with that objective" (2019, p. 603).Likewise, drawing upon a body of empirical work that includes her own award-winning qualitative research with unemployed benefit claimants, Shildrick (2018, p. 67) asserted that "where people can possibly find workeven that of very poor quality and paythey will choose paid work over 'welfare'".
This article is the first to focus on people's work attitudes and the amount of time they have spent unemployed over several decades.Two studies conducted by the Centre for Longitudinal Studies at University College London, namely the 1958 National Child Development Study (sometimes called the NCDS58, as respondents were born in 1958) 1 and the 1970 British Cohort Study (or BCS70, born in 1970), 2 are well suited to the topic; both have included three attitude questions which all offer a choice between a disliked or unattractive job and being jobless, and both provide month-by-month data on respondents' employment status from age 16 to 46.The main research presented here uses logistic regression to predict the likelihood of people with negative employment attitudes being unemployed for more than 2.5 years and more than 5 years, and spending more than 20 per cent of their total time either employed or unemployed.
The following section examines existing literature on unemployment and attitudes to lowstatus paid work; it identifies sociological and ideological contexts to debates about British people's choices between disliked or unattractive jobs and being unemployed, and assesses our current state of knowledge.The empirical findings are then presented.A conclusion follows.

| UNEMPLOYMENT AND ATTITUDES TO LOW-STATUS WORK: EXISTING ACADEMIC WRITING
Twentieth-century social science helped counter the possible view that large numbers prefer unemployment to a job.Notable landmarks include an in-depth study of the Austrian town of Marienthal by Jahoda, Zeisel, and Lazersfeld (1972), which took place shortly after the small community's main business folded in 1930.This social psychological research established that unemployment can be more than just economically punishing, as those affected suffered through losing social status, social contacts outside the family, compulsory activity, collective purpose, and time structure to their day.In the USA, the highly influential sociologist C. Wright Mills (1959, p. 9) cited unemployment as an example of a public issue that was often misunderstood as a "personal trouble" by those who lacked what he called a "sociological imagination".Authors in Britain's fledgling 'social administration' academic subject, notably Sinfield (1968), under the influence of its founding father Richard Titmuss, soon made more detailed arguments about unemployment that, echoing Mills, took full account of the problem's social and economic context (see also Welshman, 2012).
However, as Deacon (2002, pp. 22-6) noted, by the 1980s, amid mass unemployment and Marxism's growing influence on the academic discipline now known as 'social policy', understandings of social and economic structures had often extended into what he called a "denial of agency"; those who mentioned any role played by individual or subcultural dispositions when discussing social problems such as unemployment could be castigated for 'blaming the victim' (see, for example, Walker, 1990;Cook, 1997, p. 112).Sociologist William Julius Wilson (1987) wrote about the same tendency in the USA, where William Ryan's book Blaming the Victim (1971) had been particularly influential.Wilson observed that rigidly structural accounts of inner-city problems that were popular in social science could not explain why social conditions had worsened despite the most comprehensive anti-discrimination legislation and most expensive social security programmes in US history.His ethnographic work exposed widespread avoidance of the low-status jobs which were often the only jobs those he called the 'truly disadvantaged' could obtain.
Whether or not one considers specific labour market behaviours acceptable inevitably reflects one's ideological position.This can be illustrated by reference to the 'unemployment trap', a popular concept in economics that refers to the inability of unemployed benefit claimants to increase their net income via entry into badly paid jobs (see, for example, Gebauer & Vobruba, 2003).While unemployment trap implies an external constraint on individuals, those who choose unemployment over paid work when confronted with the dilemma it highlights were described by one of former UK Prime Minister Margaret Thatcher's economic advisers as displaying a "'why work?' syndrome" (Walters, 1994, p.46; emphasis added).As Lawrence Mead (1988, p. 48) noted, conservatives tend to believe that unemployed benefit claimants should be willing to accept "any legal job", whereas those on the left (which predominates in UK social policy; see also Dunn, 2014;Welshman, 2012;) tend to believe that jobs must meet certain social-justice criteria before we insist someone undertakes them.Stuart White (2003) extended Mead's discussion by identifying various 'philosophies of economic citizenship', which each imply a preferred extent of material inequality and a stance on what (if any) behavioural obligations should be imposed on the unemployed.
Thus, different ideological yardsticks are applied to the same body of evidence.Some leftwing authors (examples include Jensen, 2014;Patrick, 2015;Shildrick, 2018) have criticised Conservative government ministers' speeches (in particular, Osborne, 2012) for groundlessly portraying society as split between 'workers' and 'shirkers'; yet even evidence gathered by some of these authors can, from the conservative perspective Mead described, be regarded as validating the ministers' portrayal.For example, Ruth Patrick (2017, p. 92) juxtaposed two working-age benefit claimants.One said: "I'd do'owt, cleaning the toilets or anything as long as I knew we had enough to pay the rent", and the other: "You don't want to be stuck in a job that you don't like for the rest of your life.You want to be doing something you're enjoying."Likewise, other recent UK qualitative studies of benefit claimants have distinguished between those who see employment as imperative, and those with a tangibly stronger tendency to avoid jobs they consider unenjoyable in favour of living on unemployment benefits (Rahim, Graham, Kiss, & Davies, 2017, p. 30;Roberts & Price, 2014, pp. 13-14).Findings of this kind might be what the Conservative ministers' 'worker/shirker' dualism mainly refers to; after all, it is recognised across the board, including in DWP publications (for example, Rahim et al., 2017), that being unemployed is usually a negative experience, and that an overwhelming majority of unemployed benefit claimants want a job of some kind.
What is of most relevance here is that no evidence has been presented that indicates people with work attitudes resembling those of the first respondent Patrick (2017, p. 92) quoted are, all things held equal, as likely as the second respondent to spend a substantial proportion of their working-age lives unemployed.Dunn's (2014) in-depth interviews investigated individuals' attitudes to various jobs, their employment histories, and their rationales for past labour market decisions.Respondents differed considerably on the jobs they were willing to undertake to avoid claimant unemployment.However, employability had a greater influence than job choosiness on determining their employment status; the more educationally qualified usually secured satisfactory employment even if they expressed more negative attitudes towards employment in general.The only socio-demographic category singled out repeatedly in UK research for exhibiting a distinct set of attitudes concerning unemployment and the less attractive jobs is migrants.Numerous studies have found that employers perceive migrants as keener than the UK-born to undertake low-status jobs (see Fitzgerald & Smoczynski, 2017).
Quantitative research about unemployed people's work attitudes has delivered some contrasting findings.The 'lottery' or 'millionaire' question ('Would you still work if you had no financial need to?') is the most frequently posed in quantitative research about unemployment and the work ethic/attitudes to work (Gallie, 2019).Using it, Gallie and Alm (2000) found that in all 15 EU countries surveyed, including the UK, unemployed people were significantly more likely than employed people to indicate they would choose to work.Given this counterintuitive finding, it is perhaps illustrative of the ideological orientation of many sociology and social policy scholars that it was over a decade before a UK author (Dunn, 2014) referred to the possible inappropriateness of a survey question that offers unemployed respondents a choice between being a busy millionaire and an idle one.Research using a survey item giving employed and unemployed respondents a choice between 'almost any job' and 'being unemployed' found that the unemployed were significantly more likely to choose being unemployed (Dunn, Grasso, & Saunders, 2014).The rest of this article discusses a research project that built on this earlier study by matching people's work attitudes to their employment records between the ages of 16 and 46.

| METHODS
The NCDS58 and the BCS70 have large samples (around 10,000 for most waves/surveys) and have included the following three attitude items, which all offer a choice between a disliked/lowstatus job and joblessness (the shorthand descriptions used in this article are in parentheses): If I didn't like a job I would pack it in, even if there was no other job to go to.

(Pack It In)
Once you've got a job it's important to hang on to it even if you don't really like it.

(Hang On)
Having almost any job is better than being unemployed.(Almost Any Job) All three have five response options: 'strongly agree' / 'agree' / 'neither agree nor disagree' / 'disagree' / 'strongly disagree' (the few 'no response' / 'don't knows' are excluded from the analysis, as they are for all variables).The first two items refer to a disliked job, whereas the third refers to what respondents perceive to be a poor-quality job.The first item invites responses based on what people would do, whereas the second is about what people should do, as it is framed as a general piece of advice to oneself and others.Interviews found that the third is usually, but not always, interpreted by respondents as being about what they would do (see Dunn, 2014, pp. 115-25).That project identified respondents who had 'packed in' jobs they disliked, which contributed to the number of years they had spent unemployed (see, for example, Dunn, 2014, p. 111).Given that all three attitude items are about whether or not to favour a disliked/unattractive job over joblessness, they all relate to 'sanctionable' behaviours under current UK social security rules.
For the logistic regression analysis, attitudes are measured at as similar ages as possible in the two cohorts, to facilitate more meaningful comparison.These are age 30 (BCS70) /33 (NCDS58), and 42 for both cohorts; in fact, the only other times any of the three attitude questions were asked were at age 26 in the BCS70 (two were asked) and at age 50 in the NCDS58 (all three were asked).With responses coded from 1 'strongly agree' to 5 'strongly disagree', three dichotomous variables, based on mean scores from ages 30/33 and 42, separate those averaging 2 or more (i.e.averaging 'agree') from the rest in the 'Pack It In' item, and those averaging 4 or more in the case of the 'Hang On' and 'Almost Any Job' items (i.e.averaging 'disagree').The opposite scores (4, 2, and 2 respectively) are used to calculate the pro-employment attitudes expressed in particular years that are presented in Tables 1a and 1b.
'Activity histories' datasets now include information from 1974-2014 in the NCDS58 (see Hancock, 2016b) and 1986-2016 in the BCS70 (see Brown & Peters, 2019; Hancock 2017a).While the BCS70/NCDS58 measure of 'unemployed and seeking work' does not precisely match a recognised definition or measure of unemployment, it has the advantage of being derived from a 'main economic activity' variable with categories for long-term sick/disabled, student, and 'looks after home'; this helps prevent respondents being misallocated as unemployed.Cases with fewer than 300 months (out of a maximum of 360) of 'activity history' records are excluded from most analyses presented here.For tests in which knowing the time respondents spent in employment is of particular importance, total time spent either unemployed or employed is considered a more appropriate qualifying condition.Employed months are calculated as full-time equivalent (FTE), with 0.5 months allocated for part-time employment.The minimum figure for inclusion in these tests is 240 months.These 300/240 months cut-off points were both set at levels that balance the need for more complete details about individuals with the need to avoid a substantial loss of cases.
Testing the associations of attitudes with employment status across three decades sits well with in-depth employment history research, which has found that respondents usually, though not always, reported that their attitudes towards being unemployed and towards a variety of jobs had remained stable over long periods (Dunn, 2014).Here, correlations between attitudes expressed at age 30/33 and those expressed at age 42 were moderately strong; across both cohorts, using reduced samples with 300 or more activity months, all six tests (for the three attitudes in the two cohorts) produced Cramér's V (φc) scores of between 0.23 and 0.28.Dunn's (2014) interviews also matched answers to the 'Almost Any Job' variable with longterm labour market orientations understood with reference to respondents' post hoc rationalisations of their choices, finding a high level of consistency between the two.
Space does not allow for detailed reflection on the possible implications of the times NCDS58/BCS70 data were gathered; perhaps the most important consideration is still that when unemployment is relatively low the category tends to be weighted more heavily in favour of the least employable (M.White, 1991).Furthermore, while both cohorts faced a recession accompanied by a spike in unemployment in their early 20san age when people are particularly vulnerable to unemployment anywaya greater proportion of the NCDS58 cohort experienced unemployment at some point in the 30 years (37.2 per cent, against 28.1 per cent of the BCS70, among those with a minimum of 300 months of activity records).
The unemployed category is weighted heavily in favour of certain socio-demographic groups (see, for example, Dunn, 2014, p. 27) which are included, where possible, as dichotomised predictor variables in the logistic regression analysis.These predictors are intended to reflect the usual (i.e.not 'point-in-time') statuses of respondents across the three decades.'Single' is derived from the two datasets' 'partnership history' data (see Hancock 2016aHancock , 2017b)), and includes those who were partnered for fewer than six years in total.The balance of factors behind single people being far more likely to be unemployed than those who live with partners is underexplored; perhaps the key consideration is that economically inactive people with working partners are less likely to be categorised as 'unemployed and seeking work'.Not having dependent children in the household when surveys took place in respondents ' 20s, 30s and 40s (i.e. BCS70 1996' 20s, 30s and 40s (i.e. BCS70 , 2000' 20s, 30s and 40s (i.e. BCS70 , 2012;;NCDS58 1981NCDS58 , 1991NCDS58 , 2000) ) is used as an indicator of whether or not, across the 30 years, a respondent had childcare responsibility covering an entire childhood.In-depth interviews have confirmed that parenthood can lead people to view securing a substantial income via employment as imperative (Dunn, 2010(Dunn, , 2014, p. 102), p. 102).
Both the NCDS58 and the BCS70 lack adequate samples of most ethnic groups (Mokhtar & Platt, 2010); given that migrants are of theoretical interest here, it is noteworthy that the two 'birth' cohorts are overwhelmingly non-migrant.The measure of low social class in jobs includes only those who had experienced 'semi-skilled' or 'unskilled' manual jobs, and who had also never been in middle-class (non-manual) employment (i.e.class IIIa or above in the traditional schema) between ages 16 and 46.With a considerable choice of measures available, 'SC91' was chosen for its clear separation of 'semi-skilled' from classes above.It was unavailable in the BCS70 2013-16 data, so 'B10LNSECAN' was chosen.With the use of this variable, the top four classes are regarded as middle class, and the bottom two ('semi-routine' and 'routine') replace classes IV and V.For social class of origin, data gathered at birth on mothers' partners' occupation is limited in sample size, so it was augmented, though not overridden, by information gathered at age 16.The variable contains only class 'V' because this class was larger in 1958 and 1970 than during the years the cohorts were in the labour market.Low/no qualifications includes those with no formal academic qualifications and low General Certificate of Secondary Education grades/Certificate of Secondary Education below grade one, and is measured in the 'middle' year, at age 30 (BCS70) or 33 (NCDS58).Thus, the educational attainment and low social class jobs dummy variables split at points below which people incur a substantively greater risk of unemployment.
Poor health is measured by 'Malaise' scores, which are based on (usually) 24 questions on issues such as whether a respondent 'often feels miserable or depressed', 'often gets into a violent rage' and 'is easily upset or irritated'; these traits are relevant to workplace etiquette, and hence employability, and they tend to persist longer than the effects of accidents and illnesses reported in just one survey.Data are included from the same three surveys as those used for 'no dependent children'.Those reporting at least a third of the possible total Malaise points are classed as being in poor health (the BCS70 2012 included only nine questions, so scores were multiplied by 2.667 before being added to the other two years, to balance the impact of findings gleaned at particular life stages).'Region' is another variable derived from values in the same three surveys as 'Malaise' and 'No dependent children'.Risk of unemployment has been especially low in two regions covered by the datasetsthe south-west and the south-east of England (excluding London)so living in other parts of Britain in all three waves is used as a predictor.Unfortunately, the BCS70 and the NCDS58 lack local neighbourhood and urban/rural data.
Cases with missing values on any of the predictor variables are excluded from all tests.Logistic regression was chosen because the project sought to compare with other respondents those who have experienced an unusually large amount of unemployment; 2.5-year and 5-year cut-off points were chosen, to test lengthy and very lengthy time spent unemployed.A '20 per cent of employed (FTE) plus unemployed time' measure is included to take account of people's differing propensity to be available for employment.Only claims of association, not causation, are made about relationships between variables.As Zhang (2016) suggests, readers are told how predictors are selected.Using the 'entry' method, only predictors that associated with the outcome variable at the conventional level (P < 0.05) in an initial 'one-to-one' significance test are included in models.Details of how the reduced samples differ from the overall cohort samples are provided in the Appendix.Longitudinal tests use the ages 33 (NCDS58) and 30 (BCS70) attitude variables only, and compare these attitudes with employment records before and after all of the interviews in these surveys took place (for example, in the BCS70 this is before September 1999); doing this helps remove the possible effect of current unemployment on attitudes.

| RESULTS
The key findings are presented in Tables 1a and 1b, which show bivariate associations between work attitudes and time spent unemployed, and Tables 2a, 2b, and 2c, which give details of the logistic regression models.Table 3 presents findings around whether employment status predicted subsequent attitudes, or vice versa.
Coincidentally, the percentages of the two cohorts who spent long periods unemployed are the same; in both, 5.9 per cent were unemployed for at least 2.5 years, 2.6 per cent for at least 5 years, and 2.6 per cent for at least 20 per cent of their total time spent employed (FTE) or unemployed (Tables 1a and 1b).Negative employment attitudes are more strongly linked to spending a long time unemployed in the BCS70 than in the NCDS58; only one of the 18 NCDS58 percentage figures (on the left side of Table 1a) which relate to negative employment attitudes exceeds its corresponding BCS70 figure in Table 1b ('Pack It In' and 'At least 5 years').Note too that BCS70 'Almost Any Job' disagreers' percentage figures are all more than twice those of agreers, whereas in the NCDS58 none are.Nevertheless, smaller percentages of the younger (BCS70) cohort, as a whole, expressed negative attitudes towards employment (see Appendix).
In both cohorts, the 'Almost Any Job' and, in particular, the 'Pack It In' attitude percentages are sizeable enough to rival long established unemployment risk factors (Tables 1a  and 1b).For example, in both surveys in which highest qualification data used here was gathered (BCS70 age 30, NCDS58 age 33), those who agree to 'packing in' a disliked job are more likely to have spent a long period unemployed (on all three measures) than those whose highest academic qualification is below 'O' Level standard/GCSE grades A to C. This finding is remarkable, considering that people's highest qualification is so closely linked to their career opportunities and chances of avoiding unemployment that it has been used by some quantitative researchers as a proxy for employability (for example, Glyn & Salverda, 2000).The particularly weak results for the 'Hang On' attitude variable, across both cohorts, might be indicative of it eliciting views about what people in general should do, not what they themselves would prefer to do.Table 2b confirms that the 'Hang On' attitude variable has, at most, a weak relationship with time spent unemployed in the NCDS58 cohort; however, it delivers 'odds ratios' of around 2 in the BCS70 cohort, which means respondents expressing the attitude are around twice as likely to have been unemployed for over 2.5 years, 5 years, and 20 per cent of their time in the labour market, after other variables are controlled for.Odds ratios tend to be highest, at around 3, for the 'Pack It In' variable in both cohorts (Table 2a), and for the 'Almost Any Job' variable in the BCS70 cohort (Table 2c).
Finding out whether attitudes pre-date or post-date the experience of unemployment can shed light on likely sequences of experience, attitude formation and behaviour.In both cohorts, 'middle year' attitudes tend to associate more strongly with spending a long time unemployed prior to that time (Table 3).This indicates that work attitudes are a better guide to past experience than to future employment status.The weaker correlations for a test that excludes those unemployed in the 'early' period (right-hand column of Table 3) reflect the known importance of previous unemployment as a predictor of subsequent unemployment.

| CONCLUSION
In both cohorts, and consistent with findings from earlier in-depth research, attitudes regarding a choice between 'almost any job' and 'being unemployed', and about whether to 'pack in' a disliked job in favour of having 'no job' at all, were important predictors of whether someone had spent a substantial proportion of their time unemployed between the ages of 16 and 46.These attitudes' 'odds ratios' were comparable in size to those of variables more commonly

T''''
A B L E 1 a Proportions of NCDS58 subgroups who spent long periods unemployed between ages 16(1974)  and 46 (2004) If I didn't like a job I would pack it in, even if there was no other job to go to' Once you've got a job it's important to hang on to it even if you don't really like it' n = 6,612) 2.6 (n = 6,612) 2.6 (n = 5,322) Note: P < 0.05 = *, < 0.01 = **, < 0.001 = *** in left-hand columns only; unemployment risk variables are all significant at < 0.001.Source: NCDS58 data T A B L E 1 b Proportions of BCS70 subgroups who spent long periods unemployed between ages 16 (1986) and 46 (2016) If I didn't like a job I would pack it in, even if there was no other job to go to' Once you've got a job it's important to hang on to it even if you don't really like it' n = 6,889) 2.6 (n = 6,889) 2.6 (n = 4,673) Note: P < 0.05 = *, < 0.01 = **, < 0.001 = *** in left-hand columns only; unemployment risk variables are all significant at < 0.001 except 'Gender' and 'No dependent children' in 'at least 20%', which are both significant at < 0.05 Source: BCS70 data ratios and significance (P < 0.05 = *, < 0.01 = **, < 0.001 = ***) are presented.Standard errors are in parentheses.Sources: BCS70 and NCDS58 data T A B L E 3 Negative attitudes to employment at age 30/33 and being unemployed for at least two years before or after that time Negative employment attitude expressed at age 33 (NCDS58, in 1991) or age 30 (BCS70, in 2000) Cohort Strength of correlation (Cramér's V [φc]) between being unemployed for at least 2 years and expressing a negative work attitude in 'middle' year : Not sig.= P > 0.049; P < 0.05 = *, < 0.01 = **, < 0.001 = * ** Table includes only cases with at least 25 years of activity records; for tests including only those not unemployed in the 'early' period NCDS58 n = 4,676, BCS70 n = 5,584.Sources: NCDS58 and BCS70 data Notes