This article explores patterns of social mobility. Using linked ONS Longitudinal Study (LS) records, measurements are made of population-level transitions between the different analytic National Statistics Socio-Economic Classification (NS-SEC) class designations. Three study populations are defined and movement across the NS-SEC designations reported. Transitions between classes are considered, separately for men and women, for two 10 year periods, 1981-1991 and 1991-2001, and an extended 20 year period 1981-2001. In addition, the probabilities of individuals’ favourable, less favourable or stable transitions between NS-SEC classes are presented and compared over the study periods.
We are grateful to referees Paul Norman, Mel Bartley and Jim Newman, who reviewed an earlier draft of this article, for their helpful comments and suggestions.
Social mobility is often highlighted in the literature as an important goal for social policy, emphasising its significance for economic and social well-being. Many studies have further shown social mobility as a mechanism to move towards more positive health outcomes. Using the ONS Longitudinal Study (LS), this article contributes to the evidence by examining patterns in social mobility between 1981-2001, using the National Statistics Socio-Economic Classification (NS-SEC).
The ONS Longitudinal Study (LS) is used to construct a study population of individuals for this analysis, which contains linked census and vital event data (such as births, deaths and cancer registrations) for 1 per cent of the population of England and Wales. This enables us to observe changes in members’ socioeconomic position over time.
Three study periods were selected to examine patterns of social mobility. An extended 20 year period from 1981-2001 is considered, in addition to two ten year periods, 1981-1991 and 1991-2001. Within each study period, the analysis is restricted to those present at each census to enable observation of socioeconomic position at each census year. Only those of working age across specified study periods were selected, to enable the possibility of social transition during these periods under consideration.
NS-SEC is used to measure socioeconomic position. The seven class analytic version of NS-SEC is applied, which is based upon members’ employment status and occupation assigned at each census using NS-SEC Derivation Matrices.
A descriptive analysis is presented that compares the percentage in each NS-SEC over time. Transitions by NS-SEC origin (that is, by NS-SEC class at the start of the study period) are also presented. Men and women are considered separately throughout. Results are standardised to the earliest time point in each study period to control for the effects of cohort ageing and economic and social changes.
Logistic regression analysis is used to model the probability of mobility into a favourable or less favourable NS-SEC class, or no transition between NS-SEC class (stability) for each study period. Separate models are estimated for men, women and mobility direction. Predicted probabilities of movement and stability are presented for the general population and by NS-SEC origin.
This analysis shows a greater magnitude of favourable social mobility trajectories throughout the study period 1981-2001 than of unfavourable trajectories. A higher percentage of study members classified as Higher or Lower Managerial and Professional (NS-SEC 1 and 2) is evident in 2001 compared with 1981. Conversely, a decline in the percentage classified as Semi-Routine and Routine (NS-SEC 6 and 7) is observed.
Small Employers and Own Account Workers (NS-SEC 4) are found to be the most stable class; however, where movements are observed, these workers were predominantly moving into NS-SEC analytic classes 1, 2 and 3, particularly so among men.
Fewer women than men are reported in NS-SEC 1, although the percentage of women assigned to this class increased over the study period. The probability of men and women in NS-SEC 2 moving to a more favourable class is lower than the general population, as expected given that there is only one destination class (NS-SEC 1) to move ‘up’ to. However, NS-SEC 2 women have a lower probability of moving into NS-SEC 1 than men.
The proportion of people assigned to the Intermediate class (NS-SEC 3) diminishes over the study period, partly because those assigned to NS-SEC 3 are least likely to remain in the same class, and partly because few workers enter this class. As more women than men are classified to NS-SEC 3, the transitions out are more numerous among women. However, the direction of destination is different; while NS-SEC3 men are moving to NS-SEC 1 and 2, women are more likely to move to less favourable classes. The decline in NS-SEC 3 assignment is consistent with evidence elsewhere that the labour market is ‘hollowing out’, which refers to the decline in jobs in the middle income distribution.
The patterns of social mobility observed in this study are consistent with evidence provided elsewhere and consistent with the economic and social background of the study period.
The coalition governments’ strategy for improving social mobility, Opening Doors, Breaking Barriers, identifies improving social mobility as the principal goal of social policy; the document stresses the importance of creating conditions which provide individuals with opportunities to succeed, thereby building a highly skilled workforce to drive economic growth. There are also equality benefits, as many studies provide evidence that favourable social mobility contributes to health improvement and social cohesion (Acheson Report 1998; Wilkinson and Marmot 2003). However, over recent years, opportunities for favourable social mobility are considered inferior compared with those available to people during the late 1950’s to the early 1970’s (see for instance Blanden et al. 2006). In conclusion, indicators of social mobility are of increasing interest to policy makers.
This article illustrates how the National Statistics Socio-Economic Classification (NS-SEC) analytic class designations can be used to measure population level social transitions, and provides a basis for future research on the relationship between social mobility and health outcomes. As NS-SEC is not a hierarchical measure, trajectory is discussed, in terms of what constitutes favourable and unfavourable movements. A descriptive statistical analysis of social mobility is first presented to examine NS-SEC movement over an extended period. The propensity for class movement is then presented for the general population and for each NS-SEC origin. The methods applied are explained in detail, including the use of linked ONS Longitudinal Study (LS) records.
The Office for National Statistics (ONS) and its predecessors have provided evidence of the persistence of a social gradient in mortality since the 19th Century. Government commissioned reports such as the Black Report (1980) and the Acheson Report (1998) identified a lack of improvement in the health experience of people working in disadvantaged occupations, and proposed a number of initiatives and strategies designed to tackle the poorer life chances of the socially disadvantaged. More recently, the Marmot Review (2010) provided a strategic review of health inequalities, which affirmed the persistence of a social gradient in health.
Since 2001, NS-SEC has been used in official statistics to classify individuals by socio-economic position. This classification is based on occupation and employment status (reduced version), as well as, in some cases, size of employer (full version). Although NS-SEC was introduced to official statistics in 2001, a matrix was constructed to enable Census 1991 records to be coded to this classification. Further, Johnson (2011) developed a matrix to enable reduced NS-SEC to be assigned to Census 1981 records. Johnson (2011) used the ONS Longitudinal Study (LS) to assign NS-SEC to sample members for each of the three censuses between 1981 and 2001 to calculate life expectancy by NS-SEC.
The social mobility literature covers many disciplines including sociology and economics. From a theoretical perspective, researchers make the distinction between absolute and relative mobility. Absolute mobility, also referred to as structural mobility, refers to the process of adjustment in response to changes in the occupational structure of the economy. For instance, increases in professional employment would result in social mobility without the requirement for other workers to move into less favourable employment. In contrast, relative mobility refers to individuals’ opportunities for progression within the social hierarchy, with one worker’s favourable movement resulting in a less favourable movement of another. This article is primarily concerned with absolute mobility; results are considered in the context of changing economic conditions during the study period.
Previously, many researchers have analysed ONS data to investigate the relationship between social mobility and health inequalities. Blane et al. (1999) used the ONS LS (1971-1981) to examine the relationship between class of origin and destination on risk of death using the Registrar General’s Social Class (RGSC) measure. The authors provide evidence of ‘gradient constraint’; social mobility is found to moderate rather than widen the social class differential in mortality, thereby placing doubt on the importance of health selection explanations for the gradient. Following this, Bartley and Plewis (2007), also using the ONS LS (1991-2001), examined the relationship between NS-SEC based social mobility and Limiting Long Term Illness: They too found that social mobility constrained health inequality rather than increased it. Boyle et al. (2009), however, demonstrate that gradient constraint is not always a consequence of social mobility. They examined the impact upon health inequalities of both social class mobility and area deprivation mobility, concluding that area deprivation mobility widened the health gradient between 1971 and 1991. While there was some evidence of social class mobility widening the health gradient, this was not as great as for area deprivation. This illustrates the uncertainty surrounding the directional effect of social mobility on health inequality and the importance of investigating it further using more contemporary data.
This article does not investigate the relationship between social mobility and health outcome, but it does expand the analysis period to twenty years, capturing social transitions between 1981 and 2001, on the basis of LS members NS-SEC designations. Using the classification model developed in Johnson (2011) which is based upon employment status and occupation, LS members are assigned to an NS-SEC category at each Census during the study period (1981-2001); class of origin and destination are used as the basis to ascribe a case of social mobility. The potential for further utilising the method developed by Johnson (2011) is also explored, to contribute to improving our understanding of the complex relationship between social mobility and health. Previous studies have focused on social mobility over typically a 10 year period. This article presents measures over a 20 year period, allowing a consideration of persistence over time, using fuller career histories.
This article is likely to be of interest to social policy makers and researchers, since the governments’ strategy for improving social mobility, Opening Doors, Breaking Barriers (April 2011) states that improving social mobility is a principal goal of social policy.
NS-SEC replaced the Registrar General’s Social Class (RGSC) measure in 2001 for the reporting of official statistics by socio-economic position. The RGSC was a hierarchical schema of occupational skill and social standing, and reflected the divide between manual and non-manual occupations. Table 1 shows the RGSC categories and example occupations.
|RGSC Class Description||Occupation Examples|
|I Professional||Doctors; Chartered Accountants; Professionally Qualified Engineers|
|II Managerial and Technical/Intermediate||Managers ; Journalists; School Teachers|
|IIIN Skilled Non-Manual||Clerks; Cashiers; Retail Staff|
|IIM Skilled Manual||Supervisors of Manual Workers; Plumbers; Electricians; Good Vehicle Drivers|
|IV Partly Skilled||Warehousemen; Security Guards; Machine Tool Operators; Care Assistants; Waiters and Waitresses|
|V Unskilled||Labourers; Cleaners and Messengers|
A new classification measure was developed for many reasons, including the changes in economic structure, particularly the decline in manufacturing, and the problems associated with disproportionate class sizes for statistical inference. In addition, RGSC had been criticised for not having a universally agreed underlying concept, constraining the ability to interpret the meaning of differences between classes (Rose and Pevalin 2003). Furthermore, the decreasing population size of Unskilled Manual (Social Class V) since 1981 was limiting the ability for robust estimation of health measures for this class.
The conceptual basis for NS-SEC is the structure of employment relations operating in modern developed economies (Rose and Pevalin 2003). Occupations are differentiated in terms of reward mechanisms, promotion prospects, notice periods and job security. Those occupations exhibiting a high level of such characteristics are said to be operating on a ‘service contract’. Those with the least of these attributes are said to be operating under a ‘labour contract’. While not designed as a hierarchy, there are differences in social advantage across the classes. The most advantaged NS-SEC classes (managerial and professional occupations), typically exhibit personalised reward structures, have good opportunities for advancement, relatively high levels of autonomy within the job, and have relatively secure employment contracts. These attributes are reversed for the most disadvantaged class (Routine occupations).
In this article, results are presented by the seven class NS-SEC schema known as Analytic Classes.1 Table 2 shows the breakdowns and examples of occupations in each. It should be emphasised that NS-SEC should not be regarded as an ordinal scale, as discussed in Rose and Pevalin (2003). Bartley and Plewis (2007) also highlight that NS-SEC 4 ‘Small Employers and Account Workers’ are regarded as a separate class, with this group distinctive in their life chances and behaviour. The approach utilised by Bartley and Plewis (2007) is adopted, in classifying transition as towards ‘more favourable’, ‘stable’ and ‘less favourable’ employment relations. Their approach in categorising a movement from ‘Small Employers and Own Account Workers’ (NS-SEC 4) to ‘Higher Managerial and Professional (NS-SEC 1) and ‘Lower Managerial and Professional’ (NS-SEC 2) as ‘more favourable’, is also adopted. This is because these movements result in an improvement in job security and authority. Conversely, movements into ‘Lower Supervisory’ (NS-SEC 5), ‘Semi-Routine’ (NS-SEC 6) and ‘Routine’ (NS-SEC 7) classes are categorised as ‘Less Favourable’. It is however, unclear how to classify movement from ‘Small Employees and Own Account Workers’ (NS-SEC 4) to ‘Intermediate’ (NS-SEC 3) and so the authors advise caution when interpreting such movements.2
|NS-SEC Analytic Classes||Occupation Examples|
|1 Higher Managerial and Professional||Directors ; Doctors; Dentists; Lawyers|
|2 Lower Managerial and Professional||Teachers; Nurses; Journalists|
|3 Intermediate||Police Officers; Secretaries; Clerical Officers|
|4 Small Employers and Own Account Workers||Shopkeepers ; Hairdresser and Garage Proprietors|
|5 Lower Supervisory and Technical||Electricians; Train Drivers; Chefs|
|6 Semi-Routine||Dental Nurses; Fitness Instructors|
|7 Routine||Bus Drivers; Waiters; Cleaners; Hairdressers|
The data source used in this analysis was the ONS Longitudinal Study (LS), which contains linked census and vital event data (such as births, deaths and cancer registrations) for 1 per cent of the population of England and Wales. Information from the 1971, 1981, 1991 and 2001 censuses has been linked, along with information on vital events. The unique advantage of this data source is that sample members can be grouped according to their census characteristics, often at working age, and then followed up to measure death and survival rates for each group over time.
Three separate study populations of LS members were identified. Two 10 year study periods were selected as follows:
Present at 1981 and 1991 Census
Men aged 16-55 in 1981
Women aged 16-50 in 1981
Present at 1991 and 2001 Census
Men aged 16-55 in 1991
Women aged 16-50 in 1991
In addition, a 20 year period was selected as follows:
Present at 1981, 1991, 2001 Census
Men aged 16-45 in 1981
Women aged 16-40 in 1981
Age restrictions were imposed as above to ensure members were of working age throughout the study period to enable a potential transition to occur. We assign reduced NS-SEC to LS members as reported in Table 1.3 This required information on occupation and employment status, which is collected at each census. In this analysis, individuals’ own NS-SEC is examined. However, it is possible to assign NS-SEC to a member’s spouse and also, where the LS member is a child born after census day, based on their father’s and mother’s class. Future work may consider a variety of NS-SEC assignments, including household NS-SEC, to potentially enable a consideration of inter-generational social mobility.
The NS-SEC User Manual (ONS 2007) provides the reduced NS-SEC for each occupation and employment status combination for the 2001 Census. An approximation to NS-SEC using the Standard Occupational Classification 1990 (SOC 90) is also provided, and was used to assign sample members with a 1991 NS-SEC class. The matrix detailed in Johnson (2011) was then used to enable NS-SEC to be derived using the 1981 occupational classification. Johnson (2011) provides further information on the matrix and validation tests conducted.
As an alternative to Analytic Classes, NS-SEC may also be divided into 14 functional and three residual operational categories. The seven class analytic version of NS-SEC was chosen as the most appropriate classification for this analysis. An aggregated three class measure is available, but this involves combining NS-SEC 4 and 3 which would be inappropriate as NS-SEC 4 is a distinct category. It may be possible to consider the five class version in future work, which involves combining NS-SEC 1 and 2 and NS-SEC 6 and 7.
Descriptive statistics are presented to illustrate NS-SEC distribution over the three study periods. In addition, transitions by NS-SEC origin are considered in terms of the percentage in each NS-SEC that moved toward ‘favourable’, ‘less favourable’ or remained ‘stable’ in terms of employment relations.
The analysis covers an extended 20 year period, during which there were many economic and structural changes, which should be considered when interpreting results. Figure 1 presents GDP year-on-year growth rates for the UK, illustrating the buoyancy of the economy at different stages of the economic cycle in all census years analysed. The economy was recovering from recession in 1981, and entering recession in 1991. In contrast, in 2001 there had been an extended period of economic growth. We may therefore expect more opportunities for progression between 1991 and 2001.
Economic growth is one factor that will have had an impact upon occupational structure during this analysis period, which in turn will have affected NS-SEC distribution. Another is the decline in traditional heavy industries, with the UK economy moving away from manufacturing and towards the service sector. The study period examined in this article also saw an increase in female labour market participation. Hibbett and Meager (2003) report that the female employment rate increased from 58 per cent in 1984 to 77 per cent in 2004 compared with equivalent increases for men of 77 per cent to 79 per cent (p.504). This expanded the number of women classifiable by their individual NS-SEC in the later years of the study period.
As this research selects study populations for longitudinal analysis, age period and cohort effects must be accounted for. For example, for the 1981-2001 male study population, the cohort is aged 16-45 in 1981, 26-55 in 1991 and 36-65 in 2001. This will affect the observed NS-SEC distribution over time. For instance, NS-SEC 1 and 2 include occupations that require qualifications and experience; therefore a greater number are likely to be classified to these managerial and professional classes in the later years of the study period.
All descriptive results are standardised to the population at the earliest point in each study period (see Appendix 1 - Standardisation for an example). This is to control for the numerous changes in economic and social factors that will impact upon this longitudinal analysis. In addition, standardisation controls for the effect ageing will have upon NS-SEC classification of the cohort. For the study period 1981-1991 and 1981-2001, results are standardised to the 1981 population, whereas for 1991-2001 results are standardised to the 1991 population. NS-SEC populations by age distribution in the standardisation year sample band are used to standardise all results obtained to portray NS-SEC population and transitions over the study periods. The standardisation example explains the process of how the NS-SEC 1 male population in 1991 is standardised to reflect the 1981 NS-SEC 1 male population.
To consider the probability of transition in each time period from NS-SEC origin, a logistic regression model was fitted. Three separate Logistic regression models were fitted for each study period, to enable the probability of ‘favourable’ mobility, ‘less favourable’ mobility and remaining ‘stable’ to be estimated. Separate models were also estimated for men and women given that there are likely to be many differences influencing the probability of mobility over time, and also because of the differences in male and female working age.
Explanatory variables for each logistic model include centred age (defined as age minus mean sample age), and NS-SEC class origin. For ‘favourable’ and ‘less favourable’ mobility logistic models, extremes were excluded (NS-SEC 1 for ‘favourable’ and NS-SEC 7 for ‘less favourable’). The unclassified were also excluded for all models.
Following the estimation of each model, predicted probabilities were calculated. The propensities for ‘favourable’ mobility, ‘less favourable’ mobility, and remaining ‘stable’ were then presented for the general population and for each NS-SEC origin, and compared over the three study periods.
This section presents a descriptive analysis of NS-SEC assignment. NS-SEC distribution in each study period is first considered, to gain an initial understanding of the proportion of the sample assigned to each class, and to consider any changes over time. Following this, movements between classes are examined more closely.
Table 3 and Figure 2 report standardised results for the 1981-2001 study period to show the percentage of men in each NS-SEC group or unclassified in 1981, 1991 and 2001. Equivalent figures for women are reported in Table 4 and Figure 3.
|Total Number of Observations||81,391|
|NS-SEC Analytic Class||1981||1991||2001|
|1 Higher Managerial and Professional||10.4||14.6||15.2|
|2 Lower Managerial and Professional||13.3||17.5||17.7|
|4 Small Employers and Own Account Workers||7.2||10||10.8|
|5 Lower Supervisory and Technical||15.1||15.7||16|
Please note figures are percentages
|NS-SEC Analytic Class||1981||1991||2001|
|1 Higher Managerial and Professional||1.7||1.9||1.5|
|2 Lower Managerial and Professional||9.8||12.1||11.9|
|4 Small Employers and Own Account Workers||1.5||2.5||2.8|
|5 Lower Supervisory and Technical||1.9||2.3||3|
|Total Number of Observations||76,267|
Please note figures are percentages
Figures refer to England and Wales
Among men, the percentage of the population in NS-SEC classes 1 ‘Higher Managerial and Professional’, 2 ‘Lower Managerial and Professional’, 4 ‘Small Employers and Own Account Workers’ and 5 ‘Lower Supervisory and Technical’ all increased over time. Conversely, the percentage assigned to NS-SEC classes 3 ‘Intermediate’, 6 ‘Semi-Routine’, 7 ‘Routine’ and ‘Unclassified’ all reduced. The sample criteria require presence at each Census as this is where occupation and employment status data are recorded. Part of the reduction in the percentage assigned to ‘Unclassified’ may be attributed to the increase in the ability to assign NS-SEC to members in 2001. As described earlier, NS-SEC matrices for earlier periods were constructed using approximations reported in the NS-SEC User Manual for 1991, and derived for 1981. Johnson (2011) discusses the limitations of using a derived NS-SEC matrix for classifying individuals based on 1981 occupations.
For women, the standardised percentage allocated to NS-SEC 1 and 2 was predominantly stable over time, although there were consistent increases throughout the period demonstrated for NS-SEC 6 and 7. A large percentage of women were unclassified throughout, but this is partly because 44 per cent were unclassified in 1981, and the standardisation inflates the percentage unclassified in the following two periods.
The unclassified group are diverse, and this analysis is primarily concerned with movement between NS-SEC classes. In addition, there are more unclassified women than men, and so including unclassified women has a greater impact upon results. The unclassified group were therefore removed from the remainder of this analysis. They are also not considered in the Bartley (2007) analysis, with the focus instead on the classified. Sample selection criteria were extended for all subsequent analyses to require members to be classifiable at each census during a particular study period.
Table 5 and Figure 4 present standardised percentages of classifiable men by NS-SEC for the study period 1981-2001. Table 6 and Figure 5 present the equivalent for women.
Although removing the unclassified population reduces the available sample of men for analysis, this did not have a disproportionate impact upon the pattern of results. However, greater falls in the percentage distribution were observed for NS-SEC 3, 5, 6 and 7, suggesting there were a sizeable number of moves between the unclassified category and these classes over the study periods.
For women, removing the unclassified category reduces the sample to 39,790; while this had the effect of increasing the percentage of women classified to NS-SEC 3 at origin, the proportion allocated to this class fell over the study period.
|1 Higher Managerial and Professional||12.2||15.6||16.5|
|2 Lower Managerial and Professional||15.5||18.5||18.9|
|4 Small Employers and Own Account Workers||8.2||10.2||11.3|
|5 Lower Supervisory and Technical||17.4||16.2||16.2|
|Total Number of Observations||69,234|
Please note figures are percentages
Figures refer to England and Wales
|1 Higher Managerial and Professional||3.1||3.4||2.5|
|2 Lower Managerial and Professional||18.3||21.3||20.2|
|4 Small Employers and Own Account Workers||2.6||3.9||4.1|
|5 Lower Supervisory and Technical||3.3||3.7||4.3|
|Total Number of Observations||39,790|
Please note figures are percentages
Percentages are presented to illustrate the proportion of the study populations that remained stable, or had a favourable or less favourable transition over the study periods. Following Bartley and Plewis (2007) caution should be adopted when describing trajectories using NS-SEC. To enable movements to be considered in detail, transitions between the 10 year study periods 1981-1991 and 1991-2001 are considered separately, in addition to presenting figures on overall movement over the extended 20 year period 1981- 2001.
Figure 6 presents male transition percentages for each time period, with NS-SEC origin illustrated by the horizontal axis. Over the study period, for most NS-SEC classes, the percentage remaining in the same class fell, while the percentages moving into favourable and less favourable classes increased. The exception to this was NS-SEC 4 where the percentage remaining in this class remained high throughout (97 per cent between 1981-1991; 94 per cent between 1991-2001). A large percentage also remained in NS-SEC 1 and 2 in each study period. For instance, 79 per cent of men remained in NS-SEC 1 between 1981-2001, and an equivalent NS-SEC 2 figure of 76 per cent. Those designated to NS-SEC 3 were the least stable, with only 16 per cent remaining in this class for the 1981-2001 period. Both 10 year study period analyses confirm this pattern, suggesting NS-SEC 3 is a declining class. A high percentage of these study members moved into the more favourable classes NS-SEC 1 and 2 (76 per cent between 1981 and 2001).
Figure 7 presents the equivalent results for women, illustrating NS-SEC 2 women were most likely to remain in the same class throughout the study periods (94 per cent for 1981-1991; 93 per cent for 1991-2001). Stability in NS-SEC 2 was greater for women than men across all study periods. Between 1981 and 2001, 93 per cent of women remained in this class compared with 76 per cent of men. As for men, the percentage remaining in NS-SEC 5, 6 and 7 fell over time. Like men, the percentage of women remaining in NS-SEC 1 fell over time, but at a faster rate. For instance, over the 10 year periods the percentage of men remaining classified as NS-SEC 1 fell from 88 per cent to 84 per cent, compared with women where the equivalent percentage fell from 76 per cent to 65 per cent.
Further male and female differences were observed for NS-SEC 3 and 4. Whereas NS-SEC 4 men were highly likely to be stable in the second period, the percentage of NS-SEC 4 women in the same class fell over time (to 61 per cent for the 20 year study period compared with an equivalent of 94 per cent for men).
For NS-SEC 3, although a larger percentage of women were assigned to this class than men, as for men the size of the class reduced, with only 56 per cent remaining designated over the 20 year study period. As for men, a greater percentage of NS-SEC 3 women moved into a more favourable category than less favourable. However, in the 1981-2001 study period 76 per cent of NS-SEC 3 men moved into favourable classes and 8 per cent moved into less favourable classes, compared with the equivalent for women where 33 per cent moved into more favourable classes and 11 per cent into less favourable. When examining the differences between the results for men and women, some caution should be exercised due to differing working ages.
This section examines more closely sample members’ NS-SEC destination classes from point of origin.
Table 7 presents NS-SEC mobility origin and destination for men for each study period. Earlier findings are reinforced, since large percentages of NS-SEC 4 males remained in this class throughout. Earlier it was reported that a high percentage of NS-SEC 3 men moved into more favourable classes, and Table 7 allows an examination of which classes these men moved to. 54 per cent of 1991 NS-SEC 3 men had moved to NS-SEC 2 by 2001. The equivalent figure for the 20 year period was 56 per cent. In terms of movement to NS-SEC 1, 8 per cent of the men classified as NS-SEC 3 in 1981 moved into NS-SEC 1 in 1991. This figure increased to 14 per cent by 2001.
NS-SEC 5, 6 and 7 also experienced an increasing percentage of favourable movement over the study periods. There was an increase in movement from NS-SEC 7 to NS-SEC 5, with the percentage of men of NS-SEC 7 origin moving to NS-SEC 5 increasing from 3 per cent in 1991 to 8 per cent in 2001. The increase in movement to NS-SEC 5 for men went from 6 per cent in 1991 to 16 per cent in 2001.
There were also increases in the percentage of men in NS-SEC 6 moving to NS-SEC 2 from 3 per cent in 1991 to 9 per cent in 2001. The percentage of men moving from NS-SEC 5 to NS-SEC 2 also increased, from 3 per cent in 1991 to 8 per cent in 2001.
It was earlier noted that movements from NS-SEC 4 to 3 may not necessarily be considered favourable social mobility. Results demonstrated these movements made up less than 1 per cent of all NS-SEC 4 moves in each study period. When we consider more favourable mobility in logistic regression models, we would not expect this movement to have a significant impact upon our results.
Table 8 (32.5 Kb Excel sheet) presents NS-SEC mobility origin and destination for women during the study period. Earlier it was reported that compared with men, a smaller percentage of NS-SEC 4 women remained in this class in the second period. Here, it is illustrated that NS-SEC 4 movements for women were fairly evenly spread between other classes, with the exception of the 1981-2001 study period, where a high percentage of women moved to NS-SEC 2 (23 per cent). In terms of movements to NS-SEC 3, this was more common for women than men, with 6 per cent of NS-SEC 4 1981 members experiencing this move by 2001. Again, caution needs to be exercised when interpreting these results as they may not represent favourable social mobility.
A high percentage of NS-SEC 2 women stayed in this class, as reported earlier. Of note, results indicate that only two per cent of 1991 NS-SEC 2 women moved to NS-SEC 1 in 2001, which is much less than the equivalent for men (14 per cent). There are high volume movements to NS-SEC 2 from other classes however, and these are not constrained to NS-SEC 1 and 3 origins, although these are the most common moves. For instance, 17 per cent of women in NS-SEC 5 in 1991 moved to NS-SEC 2 in 2001.
In terms of NS-SEC 1, it was earlier reported that an increasing percentage moved out of this class over the study period. Movement to NS-SEC 2 is the most common movement for NS-SEC 1 women, which increased from 17 per cent to 30 per cent comparing the 10 year periods.
As for men, the percentage staying in NS-SEC 5, 6, and 7 fell over the study period. For women originally in NS-SEC 5, movement to NS-SEC 6 and 7 were most common, and for those originally in NS-SEC 7, movement to NS-SEC 6 was most common. Those originally classified as NS-SEC 6, however, experienced a greater number of movements to non-adjacent classes. For instance, 14 per cent moved to NS-SEC 1 over the extended 20 year period.
Logistic regression results, used to model mobility transition, are presented in this section, using the model specification outlined in the Methodology section. Models are fitted for favourable and less favourable movement and for stability, for the three study periods, separately for men and women. Predicted probabilities for each NS-SEC origin and the general population are presented along with confidence intervals.
Table 9 presents the predicted probability of favourable mobility for men in each study period. In terms of the general population, the probability of favourable mobility increased from 32.7 per cent to 37.7 per cent over the 10 year periods. The probability of the mean age general population moving into more favourable classes between 1981 and 2001 was 44.5 per cent. Men classified at the start of each study period as NS-SEC 3, 6, and 7, had a greater probability than the general population of favourable movement in each study period. Conversely, NS-SEC 5 men had a reduced probability of favourable movement in all periods. NS-SEC 2 also had a reduced probability of favourable movement than the general population, but this was expected given that the only possible favourable destination is NS-SEC 1.
|Favourable Mobility Probability|
|General Population||0.327 (0.323 – 0.330)||0.377 (0.374 – 0.381)||0.445 (0.440 – 0.449)|
|(0.183 – 0.196)||(0.224 – 0.237)||(0.223 – 0.239)|
|(0.346 – 0.364)||(0.502 – 0.521)||(0.525 – 0.546)|
|(0.142 – 0.157)||(0.164 – 0.177)||(0.179 – 0.200)|
|(0.297 – 0.310)||(0.345 – 0.360)||(0.419 – 0.437)|
|(0.370 – 0.385)||(0.468 – 0.484)||(0.515 – 0.534)|
|(0.460 – 0.474)||(0.530 – 0.545)||(0.592 – 0.608)|
Please note figures are percentages
Results were standardised
Figures refer to England and Wales
Table 10 presents the equivalent predicted probabilities for women. As for men, NS-SEC 2 origin women had a lower probability of favourable mobility than the general population, which was expected. However, the probabilities of favourable movement for this class were smaller than the equivalent for men. For instance, 1981 NS-SEC 2 women had a 5 per cent probability of favourable movement in 1991, compared with an equivalent figure of 19 per cent for men. The situation improved for women in the 1991-2001 period; the probability of favourable movement increasing to 10 per cent, but this was still less than the equivalent for men (23 per cent). Table 8 also illustrates that NS-SEC 3 women were less likely to have a favourable movement than the general population in all periods. In comparison, NS-SEC 3 men were more likely to have a favourable movement than the general population in all periods.
|Favourable Mobility Probability|
|General Population||0.268 (0.264 – 0.273)||0.364 (0.361 – 0.368)||0.400 (0.394 – 0.405)|
|(0.044 – 0.052)||(0.100 – 0.109)||(0.085 – 0.100)|
|(0.186 – 0.198)||(0.320 – 0.332)||(0.331 – 0.347)|
|(0.315 – 0.360)||(0.304 – 0.334)||(0.367 – 0.424)|
|(0.264 – 0.302)||(0.346 – 0.380)||(0.341 – 0.391)|
|(0.342 – 0.358)||(0.437 – 0.450)||(0.480 – 0.500)|
|(0.503 – 0.522)||(0.626 – 0.641)||(0.694 – 0.714)|
Please note figures are percentages
Results were standardised
Table 11 presents predicted probabilities for NS-SEC stability for men. The probability of staying in the same NS-SEC class fell over time for the male general population of mean age, from 53.4 per cent between 1981 and 1991 to 47 per cent between 1991 and 2001. NS-SEC 4 men were significantly more likely to stay in the same class in all periods, which is consistent with the descriptive findings. Also, NS-SEC 3 men are significantly less likely to remain in the same class in all study periods, providing further evidence that this population is declining.
|General Population||0.534 (0.530 – 0.537)||0.470 (0.467 – 0.474)||0.395 (0.391 – 0.400)|
|(0.597 – 0.615)||(0.566 – 0.582)||(0.511 – 0.533)|
|(0.543 – 0.559)||(0.507 – 0.521)||(0.454 – 0.473)|
|(0.438 – 0.457)||(0.250- 0.267)||(0.185 – 0.202)|
|(0.690 – 0.710)||(0.616 – 0.632)||(0.599 – 0.624)|
|(0.500 – 0.515)||(0.433 – 0.449)||(0.354 – 0.371)|
|(0.445 – 0.461)||(0.339 – 0.354)||(0.280 – 0.297)|
|(0.526 – 0.540)||(0.455 – 0.469)||(0.392 – 0.408)|
Please note figures are percentages
Table 12 presents the equivalent for female NS-SEC stability. In contrast to men, NS-SEC 4 women were less likely to remain in this class than the general population in the 1981-1991 and 1981-2001 periods; however, in the 1991-2001 period, the probability of those classified as NS-SEC 4 to remain in this class was greater than for the general population. This was the only increase in propensity for stability that is found for NS-SEC 4. In terms of NS-SEC 3 women, they were significantly more likely than the general population to remain in this class in the second 10 year period and extended period. The fact that this did not hold for the 1991-2001 period illustrates that as for men, this class population steadily declined over the period 1981-2001. This result was stronger for men, probably related to the earlier findings that a greater percentage of women were assigned to this category than men in all periods.
|General Population||0.555 (0.551 – 0.560)||0.459 (0.456 – 0.463)||0.399 (0.393 – 0.404)|
|(0.486 – 0.533)||(0.441 – 0.470)||(0.388 – 0.442)|
|(0.658 – 0.676)||(0.616 – 0.630)||(0.604 – 0.626)|
|(0.610 – 0.624)||(0.435 – 0.446)||(0.373 – 0.389)|
|(0.404 – 0.450)||(0.458 – 0.489)||(0.330 – 0.385)|
|(0.242 – 0.279)||(0.206 – 0.235)||(0.140 – 0.178)|
|(0.488 – 0.506)||(0.436 -0 .449)||(0.368 – 0.388)|
|(0.479 – 0.497)||(0.359 – 0.374)||(0.285 – 0.306)|
Please note figures are percentages
|Less Favourable Mobility Probability|
|General Population||0.226 (0.222 – 0.229)||0.249 (0.246 – 0.252)||0.268 (0.264 – 0.272)|
|(0.366 – 0.383)||(0.401 – 0.416)||(0.438 – 0.459)|
|(0.253 – 0.267)||(0.251 – 0.263)||(0.296 – 0.313)|
|(0.190 – 0.205)||(0.217 – 0.233)||(0.261 – 0.280)|
|(0.141 – 0.157)||(0.197 – 0.210)||(0.183 – 0.203)|
|(0.185 – 0.197)||(0.202 – 0.215)||(0.206 – 0.220)|
|(0.164 – 0.176)||(0.168 – 0.181)||(0.180 – 0.195)|
Please note figures are percentages
The predicted probabilities of less favourable mobility for men are presented in Table 11. Those classified as NS-SEC 1 and 2 had a greater propensity than the general population to have a less favourable movement in all periods. This was expected, as there are more classes in this direction for them to move to and men are more abundant in these classes. NS-SEC 3 has a lower propensity to move into a less favourable class than the general population in both 10 year periods but not the 20 year period. This confirms earlier findings that the NS-SEC 3 population is declining, with the majority of movers going to NS-SEC 1 and 2.
Table 14 presents less favourable mobility predicted probabilities for women. Those classified as NS-SEC 1 had a greater propensity to move into less favourable classes, as expected as there are more less favourable classes to move to. However, the propensity to move into less favourable classes increased at a faster rate than the general population. Women classified as NS-SEC 5 had a higher propensity to move into less favourable classes than the general population in all periods. However, these results should be treated with caution as a small proportion of women were classified as NS-SEC 5. NS-SEC 3 women had a greater propensity to move into less favourable classes for the 1991-2001 and extended 1981-2001 periods. This was in contrast to men, where results indicated a greater propensity than the general population to move into more favourable classes.
|Less Favourable Mobility Probability|
|General Population||0.226 (0.222 – 0.231)||0.231 (0.228 – 0.235)||0.262 (0.256 – 0.267)|
|(0.467 – 0.515)||(0.529 – 0.558)||(0.555 – 0.609)|
|(0.274 – 0.290)||(0.261 – 0.274)||(0.281 – 0.302)|
|(0.186 – 0.197)||(0.228 – 0.238)||(0.272 – 0.286)|
|(0.223 – 0.264)||(0.196 – 0.222)||(0.225 – 0.276)|
|(0.440 – 0.482)||(0.399 – 0.435)||(0.451 – 0.503)|
|(0.147 – 0.159)||(0.109 – 0.117)||(0.125 – 0.139)|
Please note figures are percentages
This analysis expands upon a wealth of studies examining social mobility, following Glass and Hall (1954) who presented the first nationally representative study. Glass and Hall reported a very stable structure and identified substantial inequalities in progression. They predicted relative social mobility would improve with increasing equality of opportunity and Blanden et al. (2006) confirmed this hypothesis. Goldthorpe (1980) also reported an increasing number of professional and managerial vacancies throughout the 1970’s, resulting in absolute mobility. The research presented here, for 1981-2001, expands upon this literature and further shows that patterns of social mobility are likely to be strongly influenced by the back-drop of economic and social phenomenon of the period.
In terms of the occupational structure, these results show an increase in the percentage of the population in occupations associated with the managerial and professional classes (NS-SEC 1 and 2), and a decline in Intermediate (NS-SEC 3), Semi-Routine (NS-SEC 6) and Routine (NS-SEC 7) classes. This is consistent with findings elsewhere, suggesting this is unlikely to be due to the sample selection criteria of this work. For instance, Wilson et al. (2006) reports employment levels by major occupation group from 1984. They illustrate increases in the percentage of total employment classified as Managers and Senior Officials and Professional occupations. In contrast, they show a fall in Administrative, Clerical and Secretarial, Elementary, Machine and Transport Operatives and Skilled Trades occupations (p.70).
Throughout, differences between male and female results are highlighted, which justifies their separate consideration. Fewer women are reported in NS-SEC 1 and 2, but there is evidence that the proportion assigned to these managerial and professional classes increased over time. The Report of the Independent Commission on Social Mobility (2009) described the “dramatic rise in the participation of women in the labour market in the second half of the twentieth century” (p.14) with many of these women taking newly created employment opportunities in the professional classes. However, the research presented in this article illustrates less favourable mobility for women in some cases. For instance, NS-SEC 2 women had a lower probability of progressing to NS-SEC 1 than men, although this improved over the study period. NS-SEC 1 women also had a greater propensity to move into less favourable classes than the general population of women.
Evidence elsewhere that suggests female re-entry into the labour market after having children may constrain progression. Manning and Petringolo (2008) summarise that “there is evidence that women face downward mobility, particularly when returning to the labour market after having children” (p.14). Walby (2006) also highlight that in addition to child care problems, female progression in the labour market may be constrained by ‘gendered employment’ and occupational gender segregation. However, it should be emphasised that there has been considerable improvement in equality of opportunity over time, and the results here indicate this progress continued up to 2001. Nunn et al. (1997) emphasised that the “situation for mothers has improved over the last 30 years” but still they find evidence that the “contemporary UK labour market severely punishes any break in employment” (p.52)
This article also presents evidence of a decline in the proportion of workers assigned to Intermediate Occupations (NS-SEC 3), including Secretaries and Clerical Officers. The impact of this is greater for women as more women than men are designated in this class, and there is some evidence that while NS-SEC 3 men moved to NS-SEC 1 and 2, female NS-SEC 3 workers moved to less favourable classes over the study period. The decline in NS-SEC 3 assignment is consistent with evidence elsewhere that the labour market is “hollowing out”, which refers to the decline in jobs in the middle income distribution (Crawford et al. 2011, p.9). Goos and Manning (2003) provide further evidence that while there has been an increase in demand for professional occupations, demand for employment has declined. Further, Heath and Payne (1999) illustrated that those employed in such middle occupations were least likely to remain there over time.
In conclusion, this article presents evidence that men and women move between NS-SEC classes over time. Patterns observed are consistent with evidence provided elsewhere and are consistent with the economic and social background of the study period.
There are certain limitations to this analysis that should be considered. As highlighted, movement to and from ‘Small Employers and Account Workers’ (NS-SEC 4) are difficult to interpret in terms of whether a move results in an improvement or otherwise in employment relations. In particular, the treatment of movements from NS-SEC 4 to 3 as favourable should be interpreted with caution. Results indicate this accounted for a small percentage of total NS-SEC 4 moves, but was more common for women than men.
The unclassified population are not considered in this analysis. In doing so, instances of favourable and less favourable mobility in terms of movement in and out of the labour market are not explored. This analysis is therefore limited to the social mobility of those with complete occupational histories recorded at censuses.
The fact that this is an extended longitudinal analysis presents challenges for analysing mobility. As described, numerous economic and social changes occurred over the study period, which we would expect to impact upon our findings. We standardised our results to enable us to control for changes in the occupational classifications used to derive NS-SEC and corresponding age distribution over time. However, we were unable to control for any changes in job descriptions over time. For instance, whereas a job may be described as being characteristic of the employment relations of NS-SEC 3 in 1981, the description may change, which results in it being classified as NS-SEC 2 in 2001. The contract, however, may not have changed. This would result in an over-estimation of true mobility. Similarly, the relationship between classes in terms of the employment contract may change over time. For instance, an increase in autonomy for the NS-SEC 3 class over time may make it more reflective of NS-SEC 2 in the 1980’s. This improvement in employment relations would not be captured here. This however, would result in an under-estimation of true mobility.
This article has illustrated the distribution between NS-SEC classes for working age men and women over the period 1981-2001. The findings of earlier studies are expanded upon by considering these two periods together.
There are several potential areas for future research. Here, we have considered individual NS-SEC only. Household NS-SEC is likely to be of particular interest in aiding understanding of the influence of other household members on social mobility transitions. Future work could extend the analysis beyond individual NS-SEC assignment to look at inter-generational mobility.
There is likely to be particular interest in the movement of people in and out of the ‘unclassified’ category and the corresponding impact upon health. Further work could be carried out on this group to examine their social mobility in more detail. This has not been examined in the recent studies using the LS, probably because it is difficult to specify which unclassified group members belong to using only the employment status and occupation variables used to derive reduced NS-SEC. However, it is possible to examine economic position alongside NS-SEC. In addition, as a substantial proportion of the unclassified group may be permanently sick or disabled, it would be interesting to look at the mobility of this population over time and corresponding reported health as different social mobility patterns may be observed.
The LS data contain marital status and household composition at each Census, in addition to births to sample mothers. There is therefore potential to study the patterns of social mobility observed in women in relation to child-bearing and family responsibilities. Again, different patterns of social mobility may be observed than those presented here for the general population, and so this is likely to be of interest.
This analysis makes the assumption that more or less favourable employment circumstances as measured by NS-SEC can be regarded as a meaningful indicator of social mobility. However, the social reality may be much more complex. Future work could, for instance, redefine the order of ‘advantage’ of the NS-SEC classes in terms of factors such as average household income. There is also scope to examine changes within the NS-SEC 4 class using the operational categories that distinguish Small Employers from the Own Account Workers.
Further research is also planned to consider the effect of social mobility upon health outcomes, such as self reported general health, limiting long term illness, mortality and life expectancy. In examining mobility over two decades (1981-2001), this analysis would contribute to the knowledge-base on how social mobility influences health outcomes.
On a related point, ONS have plans to publish research examining the influence of a number of socio-demographic characteristics on poverty durations using similar logistic regression approaches applied here.
Details of the policy governing the release of new data are available by visiting www.statisticsauthority.gov.uk/assessment/code-of-practice/index.html or from the Media Relations Office email: firstname.lastname@example.org
Acheson Report (1998) Independent Inquiry into Inequalities in Health, TSO: London
Bartley, M & Plewis, I (1997) ‘The Relationship Between Social Mobility and Illness in England and Wales 1971-1991’, Journal of Health and Social Behaviour, No.38, pp.376-687
Bartley, M & Plewis, I (2007) ‘Increasing Social Mobility: an Effective Policy to Reduce Health Inequalities’, Journal of the Royal Statistical Society, No.170, pp.469-481
Black Report (1992) in Townsend P and Davidson M (eds) Inequalities in Health. The Black Report and the Health Divide. Penguin Books, London
Blanden, J, Gregg, P & MacMillan, L (2006) ‘Accounting for Intergenerational Income Persistence: Non-Cognitive Skills, Ability and Education’, Centre for Economic of Education, London: London School of Economics
Blane, D, Harding, S & Rosato, I (1999) ‘Does Social Mobility Affect the Size of the Socioeconomic Mortality Differential? Evidence from the ONS Longitudinal Study’, Journal of the Royal Statistical Society, No.162, pp.59-70
Boyle, P, Norman, P & Popham, F (2009) ‘Social Mobility: Evidence that it can Widen Health Inequalities’, Social Science and Medicine, No.78, pp.1835-1842
Crawford, C, Johnson, P, Machin, S & Vignoles, A (2011) Social Mobility: A Literature Review”, Department for Business Innovation and Skills Report
Glass, D V & Hall, J R (1954) “Social Mobility in Great Britain: A Study of Intergeneration Changes in Status”, In D V Glass (ed) Social Mobility in Britain, London: Routledge & Kegan Paul
Goldthorpe, J H with Llewellyn, C & Payne, C (1980) Social Mobility and Class Structure in Modern Britain, Oxford: Clarendon Press
Goos, M & Manning, A (2007) ‘Lousy and Lovely Jobs: the Rising Polarisation of Work in Britain’, The Review of Economics and Statistics, Vol 89(1), pp.118-133
Heath, A & Payne, C (1999) Twentieth Century Trends in Social Mobility in Britain, Centre for Research into Elections and Social Trends, Working Paper 70
Hibbett, A & Meager, N (2003) ‘Key Indicators of Women’s Position in Britain’, Labour Market Trends, vol.111, no.10, pp. 503-512
Johnson, B (2011) ‘Deriving Trends in Life Expectancy by NS-SEC Using the ONS Longitudinal Study’, Health Statistics Quarterly 45, pp.9-52
Manning, A & Petrongolo, B (2008) ‘The Part-Time Penalty for Women in Britain’, The Economic Journal, Vol.118, pp.28-51
Marmot Review (2010) Fair Society Healthy Lives, The Marmot Review, London
Office for National Statistics (2007) The National Statistics Socio-economic Classification Report from the Independent Commission on Social Mobility (2009). Available at: http://www.mnarey.co.uk/resources/Social%20Mobility%20(Narey%20Report)%20Final.pdf
Rose, D and Pevalin, D (Eds) (2003) A Researcher’s Guide to NS-SEC, SAGE Publications Ltd.
Walby, S (2006) Gender (In)Equality and the Future of Work, Manchester: Equal Opportunities Commission
Wilkinson, R and Marmot, M (2003) The Solid Facts, 2nd ed. Copenhagen, World health Organisation
Wilson, R, Homenidou, K & Dickerson, A (2006) Working Futures 2004-2014: National Report, Wath on Dearne: Sector Skills Development Agency
In the 1981 -2001 male study population we had a sample of 81,437. In 1981, 8,474 were classified as NS-SEC 1; in 1991, 12,667 were classified as NS-SEC 1.
Standardisation of the NS-SEC 1 population in 1991 to the 1981 population base
As there is a baseline sample selected in 1981, the age range of the 1981-2001 cohort increases over time.
In 1981 the sample were aged 16-45,
In 1991 they were aged 26-55,
In 2001 they were aged 36-65.
To account for this ageing effect in the standardisation, the number of study members in 1981 that had the same age range as our 1991 sample (aged 26-55) was used.
There were 67,336 cohort members in 1981 aged 26-55, and 9,830 assigned to NS-SEC 1.
So the percentage that NS-SEC 1 contributed to the 1981 sample for the age distribution of the 1991 sample was:
Using the study population of 81,437, the 1991 NS-SEC population was standardised to 1981 as:
(81,437/100)*14.6 = 11,888.9
Standardisation has the effect of reducing the number assigned to NS-SEC 1 in 1991 from 12,667 to 11,889. As the cohort is younger in 1981, the expectation is that fewer LS members would be assigned to NS-SEC 1. In addition, structural changes in the economy resulted in an increase in managerial and professional jobs over time.
Standardisation controls for these ageing and structural effects.