Immobility in Britain?

READING GUIDE TO Blanden, J .Gregg, P. and Machin, S.  (2005) Intergenerational Mobility in Europe and North America, a report supported by the Sutton Trust, Centre for Economic Performance, [online]

Executive Summary

Rates of mobility in the UK are the same as in the USA, but both are lower compared with Canada and the Nordic nations.  Intergenerational mobility fell in the UK, using data based on cohorts born in 1958 and 1970, but not so in the USA.  Family income is increasingly associated with educational attainment rates between the two cohorts: additional opportunities are taken up disproportionately by the better off.  The 1970s cohort showed more mixed results in educational terms—rates of staying on after 16 at school showed more equality, but there was widening inequality in terms of access to higher education.  The expansion of higher education disproportionately benefited the better off families.  Family income seems to be a major variable affecting educational outcomes, and its influence might be increasing.  Overall, there is a ‘low mobility culture’ in Britain, and the government needs to make a new effort to equalise opportunity at all levels.  Educational opportunity is still the main key to intergenerational mobility.

It is possible to assess mobility in terms of the movements between 'income slices' [not a term the team uses], constructing a ‘transition matrix’.  In the first example, income distribution was split into quartiles [slices of income of 25%—lowest 25%, next lowest 25% and so on].  It is then possible to assess the movement between these quartiles from class of origin to current adult occupation [when people were aged 33 for the 1958 cohort and 30 for the 1970 cohort, to be precise].  In a fully mobile society, we would expect to find ¼ of children from each cohort in each new quartile.  That is, if we looked at each income quartile when people were aged 30 or 33, we would expect to find 25% of people from each of the other quartiles earning that sort of income.  In particular, the richest 25% would have equal proportion, 25% each, of people born into the other quartiles. We convert 25% to an index number—0.25.  If there had been no mobility then we would expect to find 100% of people in each quartile coming from the same quartile as their parents -- all the most affluent born into that group originally. This would give us a value of 100%, or 1.0 in the diagonal cells of the four by four table, and a value of zero in all the other cells ( compare with the beautiful 'fluidity matrix' in the Nuffield studies here] .

The actual values found by the team for the cohort born in 1970 are given in their table one (4).  To take some example cells, 37% of males from families whose fathers were located in the poorest quartile stayed in that parental quartile, and 16% of them made it into the most affluent quartile ( remember that 25% would be the figure if mobility was perfect].  For the richest 25%, 40% of their sons stayed in that quartile, and 16% managed to enter it from the lowest quartile as we just saw. [Note that this data is for males only, a controversial choice, as we know from this file, and one which is not explained in this article].

We can use these data to estimate the statistical association [partial correlation] between parental and child outcomes, which the team call a measure of intergenerational elasticity [they are economists].  If there is a strong correlation, which would get a value of 1, this would indicate a strong connection between parental income and children’s economic success.  The actual value for the 1970 cohort is 0.29.  We can address the issue of whether this is good or bad, acceptable or unacceptable by comparing this value to the corresponding value for other countries.

This is of course going to be difficult because the data are not always comparable, and statistics are gathered for different aims.  The team has chosen the best available data, but recognize that there were still limitations.  Nevertheless, table two (6) showed interesting differences between Britain and other countries.  In this table, the intergenerational elasticity for Britain comes out of a slightly different value—0.27, as a result, the authors tell us, of adjusting their data in order to ensure comparability across countries and across time.  Compare this value with a higher one for the USA, 0.29, but much lower ones for other countries, 0.17 for West Germany, and 0.15 for Finland [, the lower values indicate a weaker correlation between parental income and son’s income, just to hammer home the point].  [Note that for some countries, the data on parental earnings include combined income, not just the father’s income].

The figure for the USA is interesting, because America is often seen as a society with high mobility rates.  This is because the USA has high class mobility, largely because their class structure has changed considerably over time (7) [An interesting difference  between the two measures of mobility: Goldthorpe is going to take another view, as we shall see] .  For Britain, parental income and the link with educational attainment is probably responsible for the relative immobility.  For the USA, there seems to be a weaker link, but there are higher returns to education which might explain some of the findings.  ‘Race’ also has an impact on mobility in the USA.  [It should have been studied in Britain?].

Now we can compare the results for the two cohorts.  The data here are based on two separate surveys—the National Child Development Survey and the British Cohort Survey—with cohorts born in 1958 and 1970 respectively.  The tables comparing the two cohorts are provided on page 8.  Generally, they indicate that more people remain in their quartiles for the 1970 cohort [for example, it looks like 42% of people in the top 25% managed to protect their own sons and keep them in the same quartile in the 1970 cohort, compared to 35% in the 1958 one.  While 17% of sons from the bottom quartile made it into the top quartile in the 1958 cohort, only 11% did so in the 1970 cohort].  Whether we use the intergenerational elasticity measure, or the correlational measure, the differences persist.  The American data show no such apparent decline in income mobility over time.

The team suspects that education is a major factor here.  It is clearly linked to mobility, but it is only one factor.  They therefore devise a way to provide figures which are ‘decomposed into that part [of mobility] which is related to education and that part which is not’ (9).  Further, the education component itself can be split—into that part which reflects family differences, and that part which affects different returns to education.  [There is no detail provided on how these calculations are developed – doubtless it is in another article—and it is hard to see the point of looking at different returns to education, unless these are assumed to vary between the different income slices.  It would be more reasonable to assume that they vary according to gender, that women’s qualifications do not secure their incomes in the same way that men’s do, because women’s occupations are affected by so many other factors, as we know.  However, this study does not examine the effects of gender!].  It is possible to combine the effects from families together with the labour market value of the qualification to provide a measure of intergenerational elasticity again.  [The actual statistic is provided by ‘the return to education multiplied by the relationship between parental income and education, plus the unexplained persistence in income that is not transmitted through education’ (9)].

The results are displayed in table six (10), with each stage attracting a measure—the return to education, then the relationship between parental income and education, both multiplied, and then a value given to the factors not related to education.  According to the team, the results showed that education has an important role to play, accounting for somewhere between 35% and 40% of the difference in intergenerational elasticity between the two cohorts.  Secondly, and in more detail, it seems that two factors are particularly important—the increasing importance of parental income and its connection with educational attainment, and a link between parent’s incomes and son’s earnings ‘which is not explained by education’ (10).  [A rather embarrassing finding, surely, given their earlier insistence on the importance of education?].  Returns to education do not seem to be significant [I still think they will be for women].

What might explain the connection between parental income and attainment of children?  The team proposes to examine two key stages in an educational career—staying on at school after the leaving age [16 in those days], and then gaining access to higher education.  The team propose to add some more data, from another survey, in order to better calculate rates of staying on and rates of entry.  What they found is that overall, the rate of staying on after 16 has increased for all groups [note that we have a slightly different income slice here, not quartiles but quintiles, slices of 20% of the population by income].  However, the rates of staying on for the richest groups rose faster than the poorest groups between 1974 and 1986.  Then, the trend was reversed, and the poorest groups stayed on more frequently, between 1986 and the early 1990s!  The team think this was the result of more opportunities and more courses at the further education level, and the development of the new GCSE exam [less socially divisive than the old O level]. [Halsey found that FE was the major vehicle of opportunity for working class groups too -- yet it is the cinderella of the post-compulsory sector!]

The issue was whether this increasing rate then continued into entry into higher education.  The higher education completion rates by cohort were compared.  Again we find an overall increase, but also an increased inequality between the lowest income and the highest income groups.  The lowest income groups increased their access to higher education by 3% between 1981 and the 1990s.  Over the same period, the highest income groups increased their access by 26%.  The team think that the decline of means-tested support for university students was partly responsible, and they worry about the effects of top up fees.  Generally, however, the point is that just increasing places available in higher education is not enough, because the richest groups tend to get more than their share for their sons [assuming a simple model of fully equal shares, each quintile should surely get 20% of HE places?]. [Also predicted by Halsey, incidentally -- there would be little working class penetration of universities, even if they expanded, until middle class demand was satiated]

The team think that education and income might be related causally.  However, it is also often argued that family culture is a key factor in educational attainment [this has been argued since the Plowden Report at least, written in 1964].  The team have developed a model to try and separate out the effects of income and culture [again it is not clear how from this particular paper].  They showed that there is a causal influence from family income [not just father’s income then?]—if family income drops from mean income by one 3rd, this affects the chances that children will get zero ‘good’ (A to C grade) GCSEs: the chances of this happening increase by 3% or 4%.  The chances of ending up with no degree increase by the same amount.  The team suspects that this has changed between the cohorts, but they cannot show it from the data.  Income does have a definite effect, but only a rather small one compared to family culture [the usual finding in studies since Plowden].  It follows that government policy to reduce child poverty alone will not have an effect on educational attainment and therefore on social mobility.  What we need instead is ‘targeted services and access to the best schools’ (13).

Overall, Britain has relatively higher stakes of immobility, on a par with the USA.  Educational expansion alone has not increased equality, but ‘has tended to benefit children from affluent families more, and thus reinforced immobility across the generations’ (14).  Income and educational attainment are causally related, but a redistribution of income by itself will not equalise educational attainments.  What is needed instead is much more direct intervention in educational terms, at all levels, ‘early years, better schools, more financial support for postcompulsory education’ (14).  Government policies such as Sure Start, Excellence in Cities, and the provision of Educational Maintenance Allowances are on the right lines, although they must be evaluated, and they have any insufficient impact at the moment.

READING GUIDE TO: Goldthorpe, J.  & Jackson, M.  (2007) ‘Intergenerational class mobility in contemporary Britain: political concerns and empirical findings’, in British Journal of Sociology, 58 (4): 525 – 46.

Social mobility has reemerged as a test of New Labour policy, but we need good empirical data.  Like the Blanden study, this analysis uses the same data from the NDCS and BCS, and makes a link to some later General Household Survey data as well.  Unlike the Blanden study, this team found no changes in absolute mobility, but there is a change in the balance of upward and downward mobility for men.  Relative mobility stayed the same for men and women.  Overall, there is less long range mobility, though.  The policy implications are that there are no ways to go back to the higher rates of absolute mobility of the 1980s.  Instead, relative mobility rates need to be changed, but this will produce downward mobility as well as upward mobility, which will be uncongenial to politicians (it will upset the middle classes)

Social mobility became a political issue after New Labour’s change of policy emphasis from the quality of condition to equality of opportunity.  The idea was that high rates of mobility would mitigate the social divisions arising from large inequalities.  This would also be a just and fair system [classic functionalist meritocratic arguments in New Labour].  However, there is less adequate data on social mobility than ever before.  There is now no data on social origin in the General Household Survey, not since 1993.  The data limits are not sufficiently acknowledged by politicians, nor have they been by those who argue that social mobility rates are falling, like Blanden.  The Blanden findings have been cited by Tony Blair to reinforce his particular policies, but they are about income mobility, not social positions.  And the data are from two snapshots—the cohorts born in 1958 and 1970.  At best this gives us data for only 12 years of social mobility.  [It is a puzzle to me why Goldthorpe does not attempt to match these findings back on to the famous Nuffield studies of 1980.  It is possible to make a rough comparison, for example by comparing the fluidity matrices—see appendix below].

The Blanden study found that earnings at age 30 in the 1970 cohort were still connected to class of origin, more so than for the 1958 group.  But this cannot be generalised to the population as a whole, nor to other forms of social mobility.  To demonstrate this, Goldthorpe and Jackson reworked the same data, acknowledging its limits, and connected it to some GHS data for a later period, using measures of intergenerational mobility in the usual sense: class mobility, calculated by applying the classic 7 category class scale.  This is not necessarily the same as income mobility: for example income inequalities are widening within classes.  If the issue is really about income security and longer term prospects, economic life chances in general, class is a better predictor than income slices gathered as snapshots.

There are some technical issues too.  The data are more complete than for income, with fewer missing items, and it becomes possible to include the self employed, who were omitted by Blanden because of the unreliability of their figures for income.  We can examine absolute and relative mobility, structural and fluidity issues rather than looking at quartile [or quintile] incomes, which are already partly relativized.

The data shows class origins and class destinations as usual.  Data on origins was gathered using the parental socio economic group at age 11 [note that this study has data for men and women, unlike Blanden again.  Curiously, though the data for origins records paternal occupation only—female occupations in 1958 would be misleading?].  Socio economic groups had to be recoded to fit the Goldthorpe class schema, and adjusted, since the NCDS used quite large categories.  Female occupations were coded differently—the routine non-manual category was split, and lower levels  (III b) were recorded as non skilled manual [a working response to some of the earlier criticisms and ambiguities about female occupations voiced by Stanworth and others].  Data on destinations was gathered for both men and women at age 33.  Their present employment was recorded [not the full Hope-Goldthorpe classification then --unavailable data?] .  Further adjustments to coding combined IIIb with VI, and classes I, II and IVa [self employed professionals] were considered together as a ‘salariate’.  The BCS data refers to occupation at age 30 and this was also recoded, this time through the SOC90 categories.  The sample provided nearly equal numbers of men and women.  There is a slight over representation of the salariate, doubtless due to different attrition rates over time.  This over-representation would minimise immobility at the bottom, but the same error affects both databases so comparisons between them are unaffected.

Absolute mobility was calculated in two ways, as a total, then in terms of outflow.  Total figures are provided in the fluidity matrix for men on page 531 [see below] which shows the usual pattern of diagonal cells indicating quite a degree of self recruitment.  There is more upward mobility than downward.  This study also acknowledges horizontal mobility as well, apparently from a recognition of the difficulty of organizing social classes into a strict hierarchy, and allowing for a certain level of similarity.  Generally, total mobility amounted to about 75% [I am averaging from both databases – as you might guess, this implies that the differences between them are rather slight. Actual figured are in the appendix below].  Upward mobility took up about 43% of all mobility, downward about 28%, horizontal 3%.[doesn't add up to 100 because these are proportions of all mobility -- about 25% of people were immobile]   There is a slight increase in upward mobility between 1958 and 1970 cohorts, but this is not statistically significant.  The figures indicate a ‘net outcome of a decrease in upward mobility together with increases in both downward and horizontal mobility’ (531), and there is some statistical significance here.

The data for women show a very similar pattern, with slightly less net upward mobility and slightly more downward and horizontal mobility.

In more detail, the figures for men show more immobility in movements from I to II than from VI to VII.  This seems to be less long range mobility from working-class groups into both service class (especially I) and intermediate class.  There seems to be relatively more short range downward mobility, from I to II and from IVa and III.  For women, there have ‘negligible’ changes over time, with a slight rise in total mobility.  A decrease in downward mobility has been offset by increases in upward mobility and horizontal mobility, although these are not statistically significant.  There has been a certain immobility within the salariate, which is increasingly noticeable, and more downward mobility from I to II +Iva.  The main feature though is less downward mobility by daughters of skilled manual groups to unskilled manual occupations.

Outflow refers to numbers from different class origins ending up in classes of destination.  When we calculate this, we have to allow for changes in the size of classes over time [just to remind you, if the service class has expanded a great deal, it simply must admit more people from lower social classes of origin, and this tells us nothing about policies designed to improve equality of opportunity].  For men, there have been substantial changes for those of class VI origin (20% of all men in 1958, 10% in 1970).  Most of these moved up to class V, some to II +IVa.  Thus seem to be fewer such men in class I in 1970.  These changes affect all men, however, so that class VI, V, and  II +IVa also had fewer men of all class origins, a definite decline in the stability of class VI especially (15% of men in 1970 as opposed to 24% of men in 1958 – so signs of considerable structural change affecting the sizes and stability of classes].  There was also a decline in mobility from VII to VI over the same period (this is the classic route from unskilled to skilled manual class following ‘getting a trade’).  There is also decline in the numbers of self employed male professionals (class IVb), from 27% of men in 1958 to 15% in 1970.

For women, there seemed to have been more positive opportunities, with a general increase in women’s access to the salariate.  In 1970, 45% of all employed women were in that category, almost the same proportion as men, but with a slight tendency to gravitate towards the lower levels.  Women are approaching men in terms of middle class distribution of occupations.  The comparison shows greater chances for women to enter the salariate whatever their class origins (535). This seems to be an effects of changes over time.

Comparing these data with those from the GHS (a slightly longer period from the 70s to 1992) reveals consistency.  For men, upward class mobility levels out from the 1970s onwards, with a slight increase in downward mobility, with less change for women overall, but with net upward mobility especially into the salariate, and decreasing downward mobility from the salariate.

Thus the comparison shows few real changes overall, although there are different patterns, especially when considering men and women: the period has been more favourable for the social mobility of women.

Turning  to relative mobility, the team persist with odds ratios [despite the objections of Saunders] they begin by comparing the number of immobile individuals in class I to the number of mobile individuals from class I, then class II, III and so on.  What you get is a measure of ‘net association…  between class origins and destinations’ (536) [I’m not sure if this means net of changes in the sizes of classes].  An odds ratio of 1 would indicate no association between origins and destinations.  Any value greater than 1 means a positive association.  We are able then to estimate how much absolute mobility occurred from figures of general mobility [which will include the effects of structural changes, I think].

Calculations of relative mobility are based on two statistical models.  The constant social fluidity (CSF) model assumes constant odds ratios so that any change must be structural.  The UNIDIFF model assumes changing odds ratios moving towards or away from one [I don’t really understand this but I think what might be happening is a statistical prediction being made for each particular case, against which actual observed patterns can be compared.  The statistical predictions seems to be based on some notion of uniform fluidity.  Try the discussion for yourselves on page 537].  Both can be fitted to the data.  The CSF model seems to fit both male and female data, which implies that there is a strong possibility that the changes between the cohorts was the result of structural changes: CSF explains 96% of the association for men and 94% of the association for women. UNIDIFF also fits the data, and does no better than CSF [which is a test of some of the known problems with CSF discussed on page 537 – any errors cannot now be explained by CSF failures to underestimate fluidity, apparently].

It is still possible that the general patterns might conceal important local variations, which led the team to examine odds ratios in more detail.  They found two interesting anomalies.  First, in class IVb (self employed but not professionals) there is a slight shift towards fluidity between the cohorts for men (smaller for women), which shows that the levels of self recruitment to the self employed group have diminished, allowing people from more diverse origins to move in.  Second, classes I and VI seem to be becoming less fluid for men and women, a sign of less long range mobility in either direction.  This is important, because if [income ?] inequalities are reducing mobility, we would expect that to show up with long range mobility first. 

Both of these anomalies might indicate possible trends for the future, but in general, the data show that the overall picture stayed the same over the 12 year period, and the same for the period covered by the GHS for the later period.  There were fluctuations, but not trends.  Changes were mostly driven by class structural defects.

Overall, this sort of analysis shows the point of pursuing sociological work.  For politicians, absolute mobility is given more importance, probably because it can be experienced by individuals [and appears in personal stories].  Relative mobility on the other hand needs much more technical analyses of fluidity. 

To return to the debate, it seems that mobility is not falling, or at least not yet (540).  If anything, there has been a slight increase.  There is a picture of female improvement, rising upward mobility and falling downward mobility, especially since they seemed to have increased their share of positions in the salariate.  For men, there seems to have been a levelling out of upward mobility and increasing downward mobility.  So the issue really is a balance between men and women.  Longer term, the structural effects are important, especially the levelling off of service class opportunities for men and increased competition from women.  The decline in skilled manual employment also penalises men.  [A nasty cynic, not me, might argue that Goldthorpe has finally focused on female employment just in time to save his overall thesis about class mobility!].

There are difficult policy decisions ahead.  For example, recent changes seem to have benefited women but not men.  The structural changes cannot easily be reversed [although there is a hint that government policies to fully embrace the knowledge economy might increase the size of the service class again—actual data here, cited by Brown and Lauder and others, is very discouraging, though].  The only way to increase mobility is therefore to increase fluidity.  New Labour might be prepared to do this.  They seem to think that fluidity is declining in recent times, although there is no empirical backing for this: the data show a constancy in fluidity if anything.  The team agrees there is some possible decline in long range mobility, and this might need to be increased again especially for men, through educational reform.  However, if fluidity is increased, downward mobility must balance upward mobility, by definition.

So the current policies of New Labour are not really attainable.  Earlier levels of mobility arose more from structural change than educational expansion.  Downward mobility might lead to a more efficient than just society, but the social consequences have been unrecognised so far.  The team still suspects that the political focus on social mobility serves best to shift attention away from real inequalities!

The notes are, as usual fascinating:

Note 6 justifies the inclusion of I, II and IVa into the salariate

Note 7 argues the necessity of subdividing III for women

Note 19 cites a 2004 study were male and female data were taken together, and the head of household not define exclusively as a male.  The team argue that this shows a variable impact, however— a lower position of a male still has a greater importance on the female than a higher position of the female has on a man [importance in terms of social mobility?]

Note 21 says there is now increasing evidence of a weakening association between educational qualifications and class destination, together with the new importance given to ‘non cognitive factors’ such as social skills and personality.  There is a reference here to an article by Jackson, Goldthorpe and Mills (2005) ‘Education, Employers and Class Mobility’, Research in Social Stratification and Mobility, 23: 3-34

There are some superb extra references to follow up for anyone interested in this area, including, IC, a new study by Bowles, Gintis and Osborn – Groves (2005) Unequal Chances, Family Background & Economic Success, Russell Sage: New York

Appendix:  fluidity diagrams

Goldthorpe and Jackson (2007)

fluidity men


(I am slightly puzzled since these cells are normally drawn to scale, but there are two actual values from the different cohorts -- must be based on an average?)

For comparison, here is the famous social fluidity diagram from the Nuffield study of 1980 -- men only: NOte that categories have been combined in the 2007 study though.

social fluidity 1980

back to index