[My reflections on Britain since the Seventies](https://crookedtimber.org/2013/04/10/britain-since-the-seventies-impressionistic-thoughts/) the other day partly depended on a narrative about social mobility that has become part of the political culture, repeated by the likes of Tony Blair and Gordon Brown and recycled by journalists and commentators. In brief: it is the conventional wisdom. That story is basically that Britain enjoyed a lot of social mobility between the Second World War and the 1970s, but that this has closed down since. It is an orthodoxy that can, and has, been put in the service of both left and right. The left can claim that neoliberalism results in a less fluid society than the postwar welfare state did; the right can go on about how the left, by abolishing the grammar schools, have locked the talented poor out of the elite. And New Labour, with its mantra of education, education, education, argued that more spending on schools and wider access to higher education could unfreeze the barriers to mobility. (Senior university administrators, hungry for funds, have also been keen to promote the notion that higher education is a social solvent.)
Well, it turns out this may be wrong. I’m grateful to a friendly educational sociologist for pointing me a recent paper by the eminent British sociologist John Goldthorpe, [“Understanding – and Misunderstanding – Social Mobility in Britain: The Entry
of the Economists, the Confusion of Politicians and the Limits of Educational Policy’](http://www.spi.ox.ac.uk/fileadmin/documents/pdf/Goldthorpe_Social_Mob_paper.pdf) Goldthorpe tries to debunk this orthodoxy. It rests, he claims, largely on one, methodologically suspect piece of research by economists at the Centre for Economic Performance at the London School of Economics (most prominently Jo Blanden and Stephen Machin). This single study, promoted by the Sutton Trust (a charity aimed at securing wider access to higher education) was seized on by politicians because it suited their agenda. But the narrative of fluidity followed by sclerosis is, according to Goldthorpe, both weakly supported by the data and (at least in its popular form) systematically confuses two different concepts of social mobilty: absolute and relative. It is also worth noting that the economists’ research focuses on income mobility levels (comparing patchy samples from the 1958 and 1970s British birth cohort studies) rather than social mobility in the sense of mobility within a system of class stratification.
Absolute social mobility has to do with the chance that someone has of moving from the social class of their parents’ to a new one; relative social mobility measures the degree of social fluidity within a hierarchical set of social positions. It is easy to see that we can have quite a lot of social mobility in the first sense without getting very much (or even any) of it in the second. This is because absolute social mobility can be achieved by a change in the occupational structure of society such that social classes that were formerly large diminish, and ones that were formerly comparatively small, grow. So a diminution of the proportion of manual industrial workers in a population and their replacement over time with a larger white-collar salariat can cause a large increase in absolute social mobility but leave the newly-white-collared as far down the social pecking order as their parents were.
According to Goldthorpe, the increased social mobility of the postwar era was largely of the absolute kind, and has diminished, at least for men, since the 1970s, but relative social mobility has remained pretty constant througout. The case for education has been sold to the electorate on the basis of a systematic confusion between the two concepts: the idea that a better social position can be within the grasp of somone from a poor background if they have access to educational resources depends on the idea of relative social mobility, but since that concept is zero-sum, any such advance brings a corresponding loss for other people; the kind of social mobility that lifts all boats, as it were, depends on a change in the occupational structure, on the demand for labour, and it seems unlikely that this will be changed in a big way simply by directing more education towards the labour supply. Instead, it just results in intensified competitition among people who are objectively overqualified for the posts on offer, a competitition involving (at least from this limited perspective) a massive waste in the social resources and expenditure. Oh dear.
There is worse news to come, especially for those of us who work in higher education. Although you might think that widening educational opportunity would give disadvantaged kids a shot at the best social positions, and that this is a good thing in itself (and it is!) it turns out that it has much less impact on relative social mobility than we might hope. This is because families from the upper echelons of the social hierarchy have an impressive range of tools with which to defend the comparative position of their offspring in relation to both educational and labour-market opportunity. They can get their children into better schools than the poor can (by paying or moving), they can thereby swamp the admissions processes of leading universities, and they can secure second chances if their kids fail academically. (And in extremis, they can get a friend-of-a-friend to fix up a job directly.) So the expansion of higher education and the selling of the new UK fees regime to students on the basis of the labour-market advantage it secures for them is essentially a con: the new students, taking out loans to meet their £9k fees (and more for their living expenses) will flow disproportionately into lower-status universities and will end up no better positioned in the social hierarchy than their parents were. Nothing for something, as it were.
So what does this do for the story I told the other day? Essentially it casts doubt on the social narrative whilst reinforcing the ideological one. It turns out that Britain may have been pretty sclerotic in terms of relative social mobility all along. But the ideology of social mobility, the claim that everybody has a chance — that we’re all the same — has indeed been the defining cultural message of the Thatcher and New Labour years. We have an ideology of meritocracy — that the successful owe their success to their own efforts and that the poor get what they deserve — but a reality of entrenched hierarchy.
And what about higher education? If we can’t (honestly) sell it to the students or the taxpayers on the basis of the labour-market advantages it will bring to them and their families, what can we do? Essentially, we have to persuade people that being better educated is more of a good in itself even if not an effective means to an end. We may have some trouble with that.
{ 87 comments }
Neville Morley 04.14.13 at 3:32 pm
Not to mention the further advantage after graduation that some families can afford to support the further training and/or unpaid internships that govern entry to higher professions, and others clearly can’t.
It would be interesting – and not just because I’m currently in such a university – to see comparative statistics for higher education systems in the European mainland which don’t operate closed numbers and hence competitive entry for the majority of courses, and where there is much less of a tradition (at least at first degree level) of leaving home to go to a prestigious university rather than your local one, and so the university hierarchy is a lot flatter.
Alex K. 04.14.13 at 3:43 pm
A point not related to UK, but relevant to your post:
The unemployment in the US for college graduates is about 4%. The unemployment for high-school only graduates is over 8% — it’s worse for those without high-school degrees. The difference is so high that it does not matter that those unemployment statistics may be constructed in flawed ways: without a higher education degree you’re much more likely to get screwed.
I can only see those trends amplifying. It may be controversial to say that automation will eliminate a majority of low-skilled jobs in ten or twenty years. It is completely uncontroversial that they will be eliminated in say, a hundred years.
Maybe a more educated population will still have large income disparities and maybe not. But a population with a sizable uneducated part will certainly have a large underclass — their skills will largely not be needed.
krippendorf 04.14.13 at 4:00 pm
This confusion between absolute and relative mobility (which sociologists have written extensively about for close to 50 years, even if economists and journalists choose to ignore them) is even greater in the US. Here, the best evidence we have — not great, because of data constraints — is that both absolute and relative mobility are declining. This off a base that, in the case of relative mobility, was pretty much the same as any other advanced industrialized country. Yet the meme of American exceptionalism lives on, bolstered by an ideology of individualism that all too easily attributes the declining labor market fortunes of people born into less advantaged positions to personal rather than systemic failures.
Random Lurker 04.14.13 at 4:08 pm
@Alex K.
“The unemployment in the US for college graduates is about 4%. The unemployment for high-school only graduates is over 8%”
This might happen simply because more people, and thus people without are pushed to the bottom.
Since to some degree education is a positional good, the fact that everyone else is investing in education forces you too to invest in it.
William Timberman 04.14.13 at 4:15 pm
Wage slavery amongst the well-educated. Oh dear! indeed. It’s not as though we needed yet another crisis of capitalism to dissect at great length, is it? To me, this seems just another sign that, like it or not, the air is finally being let out of the blithe certainties of the post-war period. At some point, I hope, we’ll have done with the recriminations, and try something fundamentally different.
No one invested in the sort of nonsense that equates a college degree with a move up the class ladder has any pressing reason yet to take shelter from the consequences of their nonsense, but as the OP demonstrates, those in line to pay are already looking a bit nervous. In U.S. terms, when $100,000 worth of debt nets you a job paying $24,000 a year, it becomes crystal clear that whatever the intrinsic value of a college degree, snagging one by hook or crook is highly unlikely to improve your finances.
Bruce Wilder 04.14.13 at 4:19 pm
Or forget the “re” and get serious about the criminations. Guillotines!
Alex K. 04.14.13 at 4:33 pm
“Since to some degree education is a positional good, the fact that everyone else is investing in education forces you too to invest in it.”
Be that as it may, it is still unlikely that unskilled labor will see its prospects increase once jobs like truck driving and burger flipping get automated.
JSeydl 04.14.13 at 4:52 pm
krippendorf is right. There’s a lot of evidence that relative mobility has fallen in the US — primarily since the 1990s. Which means the British finding is somewhat interesting. I would have expected declining relative mobility in both countries due to globalization and Reaganism/Thatcherism.
Enzo Rossi 04.14.13 at 5:21 pm
Enzo Rossi 04.14.13 at 5:25 pm
This is the quote I was referring to specifically:
So the expansion of higher education and the selling of the new UK fees regime to students on the basis of the labour-market advantage it secures for them is essentially a con: the new students, taking out loans to meet their £9k fees (and more for their living expenses) will flow disproportionately into lower-status universities and will end up no better positioned in the social hierarchy than their parents were. Nothing for something, as it were.
John Quiggin 04.14.13 at 7:02 pm
Income mobility refers to changes in position in the income distribution. Since the set of positions is fixed, there’s no distinction between absolute and relative mobility.
If, as stated in the OP, Blanden and Machin looked at income mobility (as economists normally do), it’s impossible for them to have confused absolute and relative mobility.
Chris Bertram 04.14.13 at 7:10 pm
Thanks John. The claim isn’t that Blanden and Machin confused the two, but that their research was the springboard for a popular narrative endorsed by politicians and pundits that confused the two.
John Quiggin 04.14.13 at 7:16 pm
Obviously, thinking like an economist here, but why should anyone care about class mobility in the sense Goldthorpe is using the term (I assume something like the ABCDE classification)? It’s not relevant either to income, like the kind of mobility economists look at, or to power, like the marxian-derived concepts of class we usually talk about here.
John Quiggin 04.14.13 at 7:23 pm
Coming back to the labor market advantages of education, I’d say that the primary labor market position students are concerned about is the prospect of better wages and more secure employment, not the question of relative mobility in the C20 social hierarchy. On this basis, we can say that university education has historically proved to be a very good choice on average (in the jargon of the human capital literature, the private rate of return has been very high).
Big increases in tuition make the choice quite a bit harder, especially given the risk that you will fail, but be lumbered with a large debt, as commonly happens in the US. On the other hand, growing inequality of incomes makes the income mobility associated with education even more attractive, and the downward mobility associated with not getting an education even scarier. That’s why the US Ivies are flooded with applications despite sky-high tuition.
Chris Bertram 04.14.13 at 7:26 pm
That’s a fair point, John. I guess the ideal respondent would be a sociologist who works on stratification, such as Goldthorpe himself. I guess the claim would have to be that class membership is associated with goods and opportunities that a focus on income fails to capture. If you look at the paper, that’s something he pretty much explicitly says about the transmission of intergenerational advantage. Additionally (not a fundamental point) it looks as if the focus on snapshot income data (an a very incomplete cohort sample at that) may not give a very meaningful picture. If there were more complete income data available that might change.
Adam Swift 04.14.13 at 7:28 pm
John, the class scheme Goldthorpe uses is not like the ABCDE classification. In the paper, which really is worth a read, and needs to be read if this discussion is going to be as productive and important as I think it could be, he claims (with cites) that his measure of ‘social class’ is a better measure of economic status than current income.
Chris Bertram 04.14.13 at 7:31 pm
On this basis, we can say that university education has historically proved to be a very good choice on average
True, but that was in the era before a massive expansion in participation rates. It isn’t clear that generalizations from that era will continue to hold.
Random Lurker 04.14.13 at 7:41 pm
@John Quiggin 13
” It’s not relevant either to income, like the kind of mobility economists look at, or to power, like the marxian-derived concepts of class we usually talk about here.”
I would agree to this, however you evil english speakers often refer to “middle class” as to “someone who has a degree” (I learnt this in this blog).
Plus, more people with a degree can cause some sort of “false consciousness” because people with a degree might (a) enjoy some social status psychologic bonus and (b) believe they have a better deal than they actually have (at the beginning of their careers).
John Quiggin 04.14.13 at 7:43 pm
OK, I’ll try to find some time to read Goldthorpe and return to the discussion.
Phil 04.14.13 at 7:54 pm
We have an ideology of meritocracy — that the successful owe their success to their own efforts and that the poor get what they deserve — but a reality of entrenched hierarchy.
When Michael Young coined (or at least popularised) the word ‘meritocracy’, wasn’t he actually using it to mean precisely that combination?
John Quiggin 04.14.13 at 7:58 pm
@Chris Latest evidence is that OECD private rates of return to education are still rising. That’s consistent with everything I’ve seen on wage inequality: college/university premium still rising. Expansion in higher ed hasn’t kept pace with decline in unskilled and semi-skilled jobs
http://ec.europa.eu/education/higher-education/doc/funding/vol3_en.pdf
Mike Otsuka 04.14.13 at 7:58 pm
Thanks, Chris, for this illuminating post and for the paper by Goldthorpe to which you link. Having just read the paper, I’d encourage others to as well: it isn’t of the boring, jargon-filled, academic journal sort, but reads more like a good, long New York Review of Books piece.
This probably reflects my ignorance of sociology more than anything else, but I’m having a bit of difficulty figuring out what Goldthorpe means by ‘relative mobility’ and how it differs from ‘absolute mobility’.
When he introduces the first term, Goldthorpe writes that “relative rates [of mobility] compare the chances of individuals of differing class origins arriving at different class destinations and thus indicate the extent of social fluidity.â€
I don’t follow the ‘thus.’ As far as I can tell, how these chances compare – how similar or how dissimilar they are – has nothing to do with the extent of social fluidity. To illustrate: the chances of movement of people from different classes might be perfectly equal because they are zero across the board.
A bit farther down, Goldthorpe writes: “Relative rates of intergenerational class mobility, as measured by odds ratios, showed a basic constancy over most of the twentieth century, or at all events no sustained directional change, with the possible exception of some recent slight increase in fluidity among women. In other words, the strength of the association between the class positions of children and their parents, considered net of class structural effects, appeared remarkably robust.â€
I think that, by ‘odds ratios’, he means how strongly the destination class of children correlates with the class of their parents. That sounds like a measure of social fluidity. But I find it hard to see the difference between this and ‘absolute mobility’, which Goldthorpe defines as “the actual proportions of individuals of given class origins who are mobile to different class destinationsâ€.
I also don’t see how the sentence beginning “In other words, the strength…†captures the preceding sentence in different words.
My guess is that maybe ‘absolute mobility’ is meant to be a measure of the odds of moving from one class to another, irrespective of whether this is partially explained by changes in the sizes of different classes (i.e., ‘class structural effects’), whereas ‘relative mobility’ is meant to measure these odds, except insofar as they’re explained by changes in the sizes of different classes.
If Adam or any other sociologist out there has the time and patience to clear this up for non-sociologists like me, I’d be grateful.
Phil 04.14.13 at 8:02 pm
Essentially, we have to persuade people that being better educated is more of a good in itself even if not an effective means to an end. We may have some trouble with that.
There’s a certain kind of tweedy response to the questions we get asked about preparing students for the world of work – Employability? Surely the most employable students are those who can think for themselves! Transferable skills? I can’t think of a more important skill to transfer to students than the ability to think for yourself!… and so on. I’ve used this kind of rhetoric myself – and meant it, too, although it does sometimes feel like a bit of a dodge. It’ll be ironic if that’s what we end up marketing.
Phil 04.14.13 at 8:12 pm
An odds ratio is the ratio of two sets of odds, which in turn are defined as the relationship between the likelihoods of two (or more?) outcomes: if outcome A has a 1 in 5 probability, the odds of outcome A are 1:4 (1 A outcome to 4 not-A). It’s not entirely clear from my skim-read of the paper, but the authors may be treating the likelihoods of five outcomes (five social destinations) as a set of odds & then computing the ratio between two such sets (parent and child). So if odds P5 (parent, social class 5) were 0:5:40:40:15 and odds C5 were 0:5:41:34:20, there would be high absolute mobility from class 5 to classes 4 and 3, but low intergenerational mobility.
Mike Otsuka 04.14.13 at 8:16 pm
A comment on this remark of Chris’s:
“So the expansion of higher education and the selling of the new UK fees regime to students on the basis of the labour-market advantage it secures for them is essentially a con: the new students, taking out loans to meet their £9k fees (and more for their living expenses) will flow disproportionately into lower-status universities and will end up no better positioned in the social hierarchy than their parents were.â€
I don’t think, however, that this claim in particular is justified by Goldthorpe’s article, given that he writes the following:
“Of course, still in this period, as at other times, an association can be observed between education and mobility chances; but it is important to recognise that this association relates to the individual rather than to the population level. Education has an effect on who is mobile, or immobile, rather than on the overall rate of mobility.â€
I grant that, if Goldthorpe is right, there is no social justification for tuition fees on grounds of increase in social mobility. But you can’t infer from his paper that particular individuals who go to university are being conned by the prospect of advancement up the social hierarchy.
Phil 04.14.13 at 8:19 pm
Enzo: Unless the lower-status universities lower their fees by stopping doing research and scholarship and only provide narrow vocational training for lower-middle-class jobs — which is what the government wants them to do.
Bizarrely, few universities are taking this particular bait. I’ve come across a few that are majoring (as you might say) on getting kids into work, but more seem to be focusing on making students feel like they’re getting a good deal, a.k.a. “the student experience”. Personally I think the most valuable student experience is the experience of learning to think for your… OK, I’ll shut up now.
Alex 04.14.13 at 8:36 pm
I’m never sure if social mobility is a good thing in and of itself. The opposite, a totally ossified society, is clearly dreadful, but that is only a little more than a strawman. Absolute mobility isn’t necessarily good – the story in the OP is about climbing into a supposedly higher class with no more money, bigger mortgages, and no unions – and anyway I can’t see how it can be a long-term reality.
And relative mobility? Without greater equality, it’s just more gnawing fear. Social mobility isn’t “bigger cages, longer chains”, it’s “higher voltage, bigger drop”.
Mao Cheng Ji 04.14.13 at 9:12 pm
Social climbers are not perceived sympathetically. Bel Ami, and more recently this: http://en.wikipedia.org/wiki/New_Russian
Who needs this mobility and why? As long as classes do exist, let them be.
John Quiggin 04.14.13 at 9:17 pm
I’ve read the Goldthorpe piece, though not as carefully as I might have. Some observations
@Adam I don’t see how Goldthorpe’s I-VI classification differs significantly from the ABCDE one I recall, nor does he make much of a case for the claim that class, in this sense, is important. Can you spell this out.
I agree with Goldthorpe (if I read him right) that a more equal income distribution is a bigger factor in mobility than is education per se. On the other hand, without expanded access to education, inequality has to rise as the demand for educated labor increases (see US). This needs a lot more work.
Finally, I don’t find Goldthorpe’s critique of the economic work at all impressive. Sure, the data is limited. But the class stuff only avoids this by using a concept that is so imprecise as to overwhelm any problems with the actual data. Is it really true that there is something called a “Class I” occupation that means the same thing now as 30 years ago (as compared to “an income in the top quintile of the income distribution” which, while socially constructed, at least corresponds to something I can understand).
Chris Bertram 04.14.13 at 9:46 pm
I should say that I don’t have a lot invested in whether Goldthorpe is right or not, though I’m inclined to believe that he is at this point. When my previous post appeared I was alerted to the Goldthorpe piece by a sociologist friend, and this post is my attempt to give a brief account of his claims. I certainly don’t claim competence to adjudicate turf wars between economists and sociologists on the topic, but I am keen to learn, so I’d welcome more comments by sociologists in reply to JQ here.
Omega Centauri 04.15.13 at 1:14 am
I don’t think that more education (on average) having little effect on mobility implies there is no societal value in doing so. That is I think education is more than a zero-sum game. However the net gain to society of X more education is less than the gain accrued to an individual who gained X more education him(her)self.
The increasing fees for higher ed. and the loans that go with it, are just another way for capital to capture an increasing share of the pie. Those who buck their lower intial economic position by acquiring more education pay back a not insignificant portion of their gains to the owners of student debt. The fact that one is pretty much screwed if he doesn’t acquire the education positional good, just makes the whole situation worse.
Jason Weidner 04.15.13 at 3:01 am
The NPR program Here and Now had a discussion the other day about continued and persistent job inequality between whites and blacks in the US. Nancy Ditomaso, an organization management professor at Rutgers discussed her research that suggests that somewhere around 50% or more of jobs in the US that require at least a college degree are obtained through social networks, and that the persistent racial divide has more to do with whites’ preferential treatment of members of their own social networks than by overt racial discrimination.
Not having read the paper (yet) I wonder if Goldthorpe finds any evidence for the importance of social networks in maintaining low relative mobility, which would verify the old saw, “it’s not what you know; it’s who you know.”
Jason Weidner 04.15.13 at 3:02 am
Ditomaso’s recent book on this is “The American Non-Dilemma: Racial Inequality Without Racism”
lupita 04.15.13 at 4:41 am
Neoliberals valued education more as a way to increase a nation’s competitiveness than as a means to greater social mobility. For example, neoliberals in the US had this notion that they could be in charge of new ideas, research, innovations, and financial wizardry in exchange for manufactures goods, bananas, nannies, etc. if only they could tackle the math. Lehman Brothers pretty much put an end to that plan.
Peter T 04.15.13 at 5:46 am
re JQ @ 29 on using class rather than income, the paper states:
“Where class is treated, as in the research referred to above, on the basis of the EGP schema Erikson, Goldthorpe and Portocarero, 1979; Goldthorpe, 2007, vol. 2: ch. 5) or of the National
Statistics Socio-economic Classification (Rose, Pevalin and O’Reilly, 2005) – i.e. in terms of
individuals’ positions in employment relations – it is not only associated, and increasingly
strongly (McGovern et al., 2007: 87-93; Williams, 2011), with level of current earnings but,
further, with earnings security, short-term earnings stability and longer-term earnings
prospects (Goldthorpe and McKnight, 2006; Chan and Goldthorpe, 2007) and also with a
range of fringe benefits gained from employment (McGovern et al., 2007: 91-3).”
In essence, they are saying the research supports class as a better measure of lifetime economic prospects than current income.
Rings true to me from my time living in the UK.
John Quiggin 04.15.13 at 6:28 am
Admittedly, I don’t know the UK, and haven’t read the research described here. I also haven’t seen statistical evidence on this for Australia, but my impression is that the relative status and security of various occupational groups has changed quite markedly over the past 30 years.
Google produces this link, suggesting something v similar for the Netherlands
http://asj.sagepub.com/content/50/2/129.short?patientinform-links=yes&legid=spasj;50/2/129
and similarly I find Oesch (Swiss Journal of Sociology) saying “the conceptual bases of the schema have primarily been developed to reflect the mployment structure up to the mid-1970s, typical of high industrialism, ‘and not to predict the future’ (Erikson and Goldthorpe 1992: 237).”
So, if you are going to use this kind of setup, it seems to me necessary to consider whether entire occupational groupings are upwardly or downwardly mobile, and whether this kind of mobility has changed over time.
Like Chris, I’d be happy to hear from sociologists on this.
Chris Bertram 04.15.13 at 7:14 am
pages 11-12 of the PDF (the section beginning “An Alternative View” seem crucial here John (and that’s where Peter T’s quote is from). I see the worry in the penultimate para of your last comment, but the focus on class seems (a) to be motivated by substantial concerns about things like long-term income security and (b) to be part of a substantial body of data. By contrast, the income figures underlying the consensus view seem to be (on Goldthorpe’s account) so worryingly incomplete that it would be rash to draw any conclusions at all from them, let alone use them to underpin serious policy choices.
Lindsey Macmillan 04.15.13 at 7:43 am
Uk economist here, very much involved in this debate. You may be interested in some further work that has just been published that examines the claim that this finding is data driven. We show that it can be the case that class mobility remains constant and income mobility declines over time, driven by an increase in persistence across generations in within class permanent income. We reject the hypothesis that the income mobility findings are driven by data issues. Blanden, J., Gregg, P., and Macmillan, L. “Intergenerational persistence in income and social class: the effect of within-group inequality”, Journel of the Royal Statistical Society: Volume A. Vol 176 (2). WP version: http://ftp.iza.org/dp6202.pdf
Alex 04.15.13 at 9:15 am
It is probably true that education, specifically, contributes to better standards of living economy-wide, but that’s a different question. The UK improved labour productivity at a 2.8% annual clip between 1997 and 2010, and the best guess is that education had a lot to do with that. However, real wages (and especially those for the 99%) didn’t do anything like that.
Tim Worstall 04.15.13 at 10:15 am
@22. “@Chris Latest evidence is that OECD private rates of return to education are still rising. ”
Could be in general. But I’m sure I’ve seen the claim that the NPV of an arts degree for a man in the UK is now negative. STEM strongly positive.
john goldthorpe 04.15.13 at 10:38 am
I’ve only just seen this – a few brief comments.
1. If economists like John Quiggin want to make any useful contribution, they have to abandon their typical monocular vision do some homework in literature outside their field. First, on how sociologists conceptualise and measure social class (and no, it’s not like the rather useless MRS categories) – starting with the voluminous documentation on NS-SEC (what concept used by economists has been made operational with a similar amount of work on validation?). Second, on how the distinction between absolute and relative mobility is implemented through loglinear and logmultiplicative modelling – a distinction and techniques on which economists are only just beginning to catch up.
2. I know the article by Lindsay Macmillan and colleagues and am not much impressed. My colleague, Robert Erikson, and I raised questions with them about the assumptions and analyses from which they start out when we saw an earlier version and go no adequate response.
3. However, the main concern of my paper (forthcoming in revised form in the Journal of Social Policy) was to get away from the debate over the economists’ analysis of data from just two birth cohorts only 12 years apart, which can say rather little about long-term or, by now, current trends in mobility, and focus on the very different message that comes from sociologists’ work extending over decades – and which my colleagues and I are right now extending into the 21st century. The economists won’t like these new results either!
Leo 04.15.13 at 10:59 am
There are lots more papers, many with Goldthorpe, on social status by his Oxford colleague (and my old tutor) Tak Wing Chan here:
http://users.ox.ac.uk/~sfos0006/
The crucial one for this debate is this paper by Chan and Goldthorpe on the basis of their concepts of class and status:
http://users.ox.ac.uk/~sfos0006/papers/asr2007.pdf
I’d also recommend book for a discussion of attempts to empirically measure the degree to which various countries, the UK included, are meritocratic, along with a superb discussion of difficulties in pinning down the concept of merit.
Leo 04.15.13 at 10:59 am
well that link went wrong…
John Quiggin 04.15.13 at 12:06 pm
@John Goldthorpe: My first suggestion would be to dial down the disciplinary rhetoric. I’m happy to learn from sociologists, but I doubt that it’s helpful to claim that your class categorization is superior to any operational concept achieved by the entire economics profession (CPI, National Income accounts, elasticity of demand come to mind, to name a few off the top of my head). Similarly, I doubt that your claimed superiority is due to the fact that economists can’t understand log-linear and log-multiplicative models.
Some more specific points
First, contrary to your claim that economists will hate your results, I think lots of economists (though not the ones I like) will read you as asserting that the welfare state was no better than Thatcherism, and will (contrary to your suggestion) be very happy with your results. If that’s not what you mean to say, it’s far from clear so far.
Second, while I’m willing to be convinced that your categorization is radically different from, and superior to, the ABCDE version (did you mean NRS not MRS?), a bit of explanation would help. They look very similar to me.
Third, my immediate intuition (more that of a non-specialist observer than a monocular economist) is that the relationship between class (however interpreted) and occupation has changed a lot between generations. If you disagree, could you explain why. If not, what’s wrong with my observation that your approach requires consideration of group as well as individual mobility?
It’s v late in Oz, so I’ll sign off for now.
Enzo Rossi 04.15.13 at 1:26 pm
@ Phil
Bizarrely, few universities are taking this particular bait. I’ve come across a few that are majoring (as you might say) on getting kids into work, but more seem to be focusing on making students feel like they’re getting a good deal, a.k.a. “the student experienceâ€. Personally I think the most valuable student experience is the experience of learning to think for your… OK, I’ll shut up now.
Agreed, there’s a combination of investment in entertainment and promises of employment as return on investment. It will take a few years for the fees to settle and properly reflect stratification, no doubt.
Mike Otsuka 04.15.13 at 1:26 pm
I’ve been trying to get clear the conceptualization of ‘relative’ versus ‘absolute’ social mobility. I’ve come across the following conceptualization by Goldthorpe’s former Nuffield colleague Richard Breen that seems roughly along the lines of Goldthorpe’s and also reinforces my earlier puzzlement. Breen writes (with my bold square brackets added):
“Sociologists use two kinds of measures of intergenerational mobility – absolute and relative. [a] The relative measures … tell us how strongly parental social class predicts the social class in which a child will be located when he or she is an adult. Sociologists sometimes call this social fluidity and a situation in which child’s social class does not depend at all on parent’s class is called perfect mobility. … [b] Measures of absolute mobility, on the other hand, capture whether people are higher up or lower down in the class hierarchy than their parents. This involves a simple comparison between a person’s social class and the one he or she was born into, [c] whereas relative measures ask whether a person’s position relative to other people, is better or worse than her parents position was relative to other families.â€
Then Breen usefully (for me) illustrates the distinction between [b] and [c] as follows:
“Suppose that my father was a clerk and that I am a manager: then, in absolute terms I have been upwardly mobile. But suppose that, in my father’s generation, being a clerk gave him a class position that was better than half of the population, whereas, in my generation, being a manager puts me in a position which is better than, say, 40 per cent of the population. Then, in relative terms I have been downwardly mobile because my rank is worse than my father’s: half of the population were in a better position than him whereas 60 per cent are in a better class position than me, and 8 this is so even though I have an objectively better class position than he had. … When there has been a large upward shift in the distribution of class positions such a discrepancy will be common.â€
Okay, so now I think I understand the distinction between [b] and [c], and also why the former is regarded as an absolute measure and the latter a relative measure of social mobility.
But what I don’t understand is the identification of [a] with [c] and the contrast drawn between [a] and [b]. As far as I can tell, [a] provides an accurate general description of the absolute measure [b] as well as of the relative measure [c]. In other words, [b] and [c] appear to me to be two different ways – one absolute and the other relative – of specifying [a].
Phil 04.15.13 at 1:48 pm
[a] is a derived measure: you start with “how many people born into class n live their adult lives in class n+1?” (which is [b]), then calculate how that proportion compares with the respective proportions for people born into classes n+1, n-1 etc. Absolute social mobility is “50% of the children of typists and secretaries grow up to do clerical jobs”; relative social mobility is “…however, 80% of the children of people in clerical jobs grow up to order people like their parents around, because those jobs now have the social status of typists and secretaries”.
Adam Swift 04.15.13 at 2:08 pm
@Mike: Your penultimate paragraph gets relative mobility right. Odds ratios are supposed to be measures of association between origins and destinations that factor out changes in the shape of the class structure (i.e. changes in the sizes of the various classes between which individuals are, or aren’t, moving).
Your puzzlement is due to an ambiguity in the phrase “different class destinations” in Goldthorpe’s claim that “relative rates [of mobility] compare the chances of individuals of differing class origins arriving at different class destinations and thus indicate the extent of social fluidity.†He doesn’t mean “a different class from the one in which they started”, he means “the [various] different class positions in which they might end up”. If there were no mobility at all, people might indeed have the same chance of moving (i.e. none), but they wouldn’t have the same chance of arriving at “[the various] different class positions”. Those who started in class 1, would have a 100% chance of ending up in class 1, those who started in any other class would have a 0% chance of ending up in that class. The loglinear models that are used to analyse the data model all the movements (or non-movements) between all origins and destinations. Hope that helps.
Gordon Marshall et al (one of them me) did our best accessibly to explain odds ratios, what they measure, in an Appendix to the book – Against the Odds? – that Leo mentions. The book as a whole tries to explain why odds ratios are the right measure for those interested in the distribution of opportunities (e.g. those interested in equality of opportunity) as between those born into different classes. (It also tries to relate the data to various conceptions of equality of opportunity, especially the meritocratic conception.)
@John: I’m not expert enough in the differences between the class scheme Goldthorpe uses and the ABCDE classification to say anything useful, sorry. Maybe he will explain further, or provide a link, or somebody else who knows more of the details will chip in.
I agree that there is an important issue about whether what it means, in distributive terms, to be a member of a class has changed over time. (I know that’s not quite the same as your point about whether occupations, or groups of occupations, may have changed their class position over time, but I think it turns into the same thing.) My paper ‘Class Analysis from a Normative Perspective”, British Journal of Sociology 2000, tries to be careful about what exactly measures of relative mobility between classes are actually measuring, given that the gap between origin and destination positions may be changing over time, and emphasizes their limitations. (I would give links to this and the next paper if I knew how.)
Part of what’s interesting here is the way the media and politicians seemed to rely on a single study that quickly became the conventional wisdom. Even more interesting, to me at least, is the way that comprehensive schools became part of the story – the press was full of stories about how comprehensives (i.e. non-selective state secondary schools) had killed social mobility – even though the study in question made no mention of them. Vikki Boliver and I had a paper called “Do Comprehensive Schools Reduce Social Mobility?, British Journal of Sociology 2011, which looked at both income and class mobility and found that, for our rather limited (but apparently best available) sample of children born in the same week in 1958 (NCDS), the answer was no (though there are a few interesting details.) A user-friendly version of the paper was in Renewal 2011.
Adam Swift 04.15.13 at 2:14 pm
Mike, ours crossed, I meant the penultimate para of your first post.
I agree that Breen’s formulations are puzzling/misleading. If you really are interested in getting this right, which it seems like you are, and if what I’ve said already isn’t enough, then honestly I can’t do better than suggest you try the appendix to Against the Odds? Or you could try Marshall and Swift “On the Meaning and Measurement of Inequality”, Acta Sociologica 1999, if that’s easier for you to get hold of.
SamChevre 04.15.13 at 2:19 pm
This intuitively makes sense to me in the US context.
In 1920, 10% of the labor force was farm laborers; another 15% were farm owners and tenants. It is nearly impossible to be employed in 1970 and not be far better off than a sharecroppers, a subsistence farmer on marginal land, an unmechanized coal miner. Even if there was no relative mobility at all, the “absolutely better off than my parents” was pretty easy to achieve.
Similarly for poor immigrants. It’s not very hard to be better off than you would be in China, or India, or even Mexico.
john goldthorpe 04.15.13 at 3:42 pm
@ John Quiggin
I think it is a bit rich to talk about ‘disciplinary rhetoric’, when economists are such masters of it. What annoys sociologists is this. Those of us who research and write on social mobility do for the most part read and try to understand relevant work by economists, even if we end up being critical. But economists – with one or two notable exceptions such as Anders Bjorklund and Markus Jantti – don’t return the compliment. And if they do comment on our work, as, say, in response to criticism, they do so without making much effort to understand what we are doing – e.g. how we understand class and seek to measure it.
As for the Market Research Society (MRS) categories, I do not know how they were formed – i.e. on what conceptual basis – nor thus what it is that they are supposed to measure. Does anyone? So far as I can see, they are essentially ad hoc. NS-SEC and my earlier class schema, of which NS-SEC is an improved instantiation, have an explicit theoretical basis – actual taken from (transaction cost and new personnel ) economics! – and much work on their construct validity has been carried out. It’s asking a bit much to have to rehearse this here, but it’s all there in the NS-SEC documentation and in the relevant chapter of my book, On Sociology (vol. 2). But just on the point of changing occupations, it doesn’t matter, from the standpoint of these classifications, if miners, shipwrights, steel workers, spinners and weavers etc. have given way to shelf-stackers, van drivers, security guards, care assistants etc so long as we are still talking about wage workers paid by piece or time with few prospects of lifetime earnings or career advancement. It’s the form of employment relations that matters, not the content of the work.
Finally, the reason why I don’t think the economists – i.e. Lindsey and her colleagues – will like our new results is that they appear to confirm that there has been no decline in absolute or relative mobility in the recent past but that the degree of intergenerational transmission of economic advantage and disadvantage appears much stronger if mobility is studied on the basis of class rather than of current earnings or income.
Chris Bertram 04.15.13 at 4:48 pm
Thanks to the sociologists and economists who have commented already. I realise that members of each tribe have some strong commitments and believe that (some) members of the other team are misguided. Can I just appeal to both to bear in mind those of us who belong to neither and are trying to make our minds up on the issues with your help.
BrendanH 04.15.13 at 5:19 pm
@John Quiggin: I don’t think anyone is implying that economists can’t understand log-linear models, but I think it is the case that through unfamiliarity they don’t see what’s being done with them, and being familiar with different (and very powerful) analytical apparatuses are disinclined to explore.
Another source of mutual misunderstanding between sociologists and economists is the continuous–categorical variable distinction. Economics has lots of measures that can legitimately be treated as interval or ratio variables and consequently a large body of regression-based techniques, and a tendency to think of variables as dimensions in a real-number space. Sociology has had to deal with concepts that are less easily measured as dimensions and is more used to thinking in terms of categorical variables. I like to think of some category sets (such as the Goldthorpe class scheme) as parsimonious representations of a clustered distribution across a latent multi-dimensional space. Some phenomena lend themselves to such a categorisation, yielding powerful analysis; some don’t, yielding a mess (tabular analysis of social mobility is in the former case, I would argue, because the class scheme captures a lot of important structure) Economists often don’t like these variables, because it’s simpler to theorise in terms of clear simple quantities, and because they have great expertise with modelling continuous data (your examples of income, CPI and elasticity, for instance).
Once you decide to put up with categorical data, an odds-ratio based way of analysing the structure of inequality becomes attractive. To a good approximation odds ratios capture the inequality in a way that is independent of the marginal distribution: the inequality between origins a and b in getting outcome c versus d will tend to result in the same OR no matter the relative a/b and c/d distributions. The “latent variable” justification of the logistic regression model makes this clear in a related context.
Log-linear models of square mobility tables are representations of the origin–destination structure as a set of odds-ratios. If the same model structure fits at two different time points, though the distributions of origins and destinations have changed, we can say that relative mobility is unchanged, because for all pairs of origins the ratio between the odds of all pairs of destinations has not significantly changed. I’m not sure this is immediately obvious when statistics-literate people from another field encounter log-linear models; particularly so when you are used to dealing with single-dimension detailed-information variables like income where the distributional shifts will drop out by other means (e.g. using quantiles).
Mike Otsuka 04.15.13 at 5:47 pm
Thanks to both Phil and Adam for your explanations and also to Adam for the references!
John Quiggin 04.15.13 at 7:16 pm
Brendan, that’s a nice explanation of some of the issues with categorical and continuous data. As it happens, this is one of the few areas of econometrics where I have some (admittedly dated) expertise. I was doing logistic regressions on categorical variables back in the early 1980s, looking at such exciting topics as exits from the Australian dairly industry. So, I’m quite comfortable with log odds ratios.
What I’m less comfortable with is the idea that there are well-defined and stable categories in employment, particularly if those categories are based on occupation. I agree entirely with John Goldthorpe that “it’s the form of employment relations that matters, not the content of the work”. But employment relations have changed a lot, mostly in ways that increase inequality. It would be great to have some explanation of how this fits in to the analysis of social mobility.
BrendanH 04.15.13 at 8:54 pm
JQ, you’re right to focus on the longitudinal stability of the classification, and this is partly an empirical question. However, I think you underestimate its effectiveness. It captures a lot of differentiation in a very systematic (and theoretically articulated) way, and in its various forms from EGP to the NS-SEC, has had a lot of use and validation. Some of the work by Rose, Pevalin and O’Reilly in proposing the NS-SEC showed that the groupings correlate well with a range of relevant characteristics (such as autonomy, pay, education; see e.g., ), and in general it can be shown that current class is a better predictor of life-chances than is current income.
The “theoretically articulated” part is important for (some) sociologists but understandably not necessarily for economists. Part of the virtue of using a class variable (strictly constructed, not ad hoc like the MRS or Registrar General’s schemes) to predict outcomes is that it ties structures of inequality directly to the principles of the structure of production, ownership and control. Aspirations, if you like, to a grand theory of everything, though with nothing like the vaulting ambition of Marx. The mark of that other DWM, Weber, is strong though.
BrendanH 04.15.13 at 9:06 pm
Forgot to link to Rose et al: pdf
John Quiggin 04.16.13 at 1:21 am
Whatever approach you take, longitudinal studies of intergenerational mobility are hard to do, especially if you want to assess whether mobility is changing between cohorts. That’s not to say they aren’t important and useful, but you’re rarely going to have solid enough conclusions to support specific changes in policy (in this case education policy). To that extent at least, I agree with Goldthorpe.
But you don’t need this to resolve the issue as it relates to educational policy. All you need is
(i) Post-school education improves life chances substantially for those who undertake it (relative to ex ante similar people who don’t)
(ii) Access to post-school education became more unequal between the 1958 and 1970 cohorts
(i) and (ii) imply that the effects of changes in education access between 1958 and 1970 cohorts were to reduce (relative) social mobility
Point (i) is established beyond any reasonable doubt from evidence over long periods in many countries (I think it’s also clear that this is primarily due to real effects of education, and not to sorting/screening, but that’s not critical to the argument here. AFAICT, point (ii) is common ground – I have no independent knowledge.
Even if point (ii) weren’t correct, the Blairites were right in saying that if relatively more students from low-income backgrounds completed university, social mobility would increase. But that doesn’t mean “education is everything”. In particular, as I’ve pointed out many times, if you accept massive inequality in outcomes in one generation, it’s going to be very hard to achieve equal educational opportunity in the next.
ajay 04.16.13 at 8:49 am
If you take half the 18-year-olds in the country out of the workforce for three years, that will reduce economic growth. That applies whether they are spending those three years in prison, on the dole, in the armed forces or at university. The question is: is the extra productivity of a worker who’s spent three years at university always enough to make up for that three-year loss – both from the point of view of the national economy, and from the point of view of their own lifetime earnings? Or has the UK actually gone too far in opening up access to higher education?
John Quiggin 04.16.13 at 10:59 am
@ajay To put it another way, is the social return to education higher than the cost of capital. Vast numbers of studies give the answer: Yes
ajay 04.16.13 at 11:04 am
Even for the marginal student?
ajay 04.16.13 at 11:05 am
I ask because there are, for example, lots of people (eg Paul Campos) pointing out that law school is a massive scam which leaves you stuck with lots of debt and no job. In that case it seems that the social return is a lot lower than the cost of capital.
John Quiggin 04.16.13 at 11:06 am
I assume you’re switching from UK to US here
Mike Otsuka 04.16.13 at 11:46 am
I’ve just read Appendix B to Against the Odds?, which Adam Swift mentions above. It struck me as an accessible, concise, and precise explanation of what odds ratios are and how a grasp of this concept makes clear the distinction between absolute and relative mobility. There are also some useful examples of social mobility that help illustrate this distinction.
For those who have access to Oxford Scholarship Online, you can download the appendix as a pdf. (The pdf, however, leaves out the titles of the subsections of the appendix. So it won’t be clear that the discussion of The Odds Ratio begins on the second paragraph of the first page of the pdf and ends halfway down the third page of the pdf.)
Mike Otsuka 04.16.13 at 11:52 am
PS: I see that the titles of the subsections of the appendix are also missing from the html of the online version of the book. OUP should fix this.
Adam Swift 04.16.13 at 2:23 pm
Thanks Mike!
NickS 04.16.13 at 3:46 pm
“Even for the marginal student?”
In addition to John Quiggin’s response I would note that the high rate of unemployment for 18-24 year olds would suggest, at a first approximation, that the market doesn’t show a large unfilled need for the labor of the people who are marginal students at university.
ajay 04.16.13 at 3:50 pm
the high rate of unemployment for 18-24 year olds would suggest, at a first approximation, that the market doesn’t show a large unfilled need for the labor of the people who are marginal students at university.
Also suggests that the market doesn’t show a large unfilled need for the labour of people who have actually been to university.
Igor Belanov 04.16.13 at 5:20 pm
@ajay
Exactly. The expansion of higher education over the past twenty years has effectively been that of students running to stay still, as far as social mobility is concerned. There are plenty of jobs these days that insist on post-18 qualifications, but which could have been entered at 16 or 18 in my parents’ generation. I think one of the influences behind the increase in student numbers has been to reduce youth employment while creating a part-time sector of casual workers among current students. The other major factor is that companies want to train their staff as little as possible, and want the costs of training to be passed on to the state or prospective workers themselves. The constant changes in student finance over the past 15 years have been provoked by the state trying to shift some of its costs to individuals as well.
I suspect that, like myself, there will be many more graduates earning below the average wage in future.
Trader Joe 04.17.13 at 11:51 am
@69
I agree with your comment in general. That said, companies generally ask for “post-18” qualification as a screening tool. As both you and ajay note, there is more than ample supply of labor with either some higher ed or a full degree. The pay for the entry level position is likely to be pretty much the same regardless of qualifications so why not try to hire a greater potential skill set for the same price?
In my experience, the main complaint of many businesses is that higher-ed does a relatively poor job of developing the skills entry level employees might require (which tend to be interpersonal skills and work-ethic related rather than writing or analysis) so while they might prefer to train as little as possible – its not a matter of passing that cost to the state, its a matter of the entry level worker has few skills to begin with and over training them is inviting them to leave for another job as soon as you’ve finished incurring the cost.
I’d fully agree with the point that the 21st century student is not only running hard to stay in place, in many cases they are digging themself a hole as they run.
ajay 04.17.13 at 1:50 pm
There are plenty of jobs these days that insist on post-18 qualifications, but which could have been entered at 16 or 18 in my parents’ generation.
Good point. And that actually suggests that I was partly wrong. There may be no good reason for companies to insist on degrees for those jobs, but, as long as they do so, there will still be an economic advantage to going to university, even for the marginal student. If UK employers all of a sudden inexplicably turned Thai, and started expecting all their new hires to have spent a couple of years as a Buddhist monk first, then spending a couple of years as a Buddhist monk would be a very economically rational thing for an 18-year-old to do, even if it had no effect on actual skills and suitability for the job at all. (Let us assume for the moment that you don’t learn anything particularly relevant to the UK workplace by being a Buddhist monk for two years.)
But – it would still have deadweight effects on the UK economy as a whole, because you’re still taking half your workforce out of the workplace for two years so that they can jump through this effectively arbitrary hoop.
ajay 04.17.13 at 1:51 pm
the 21st century student is not only running hard to stay in place, in many cases they are digging themself a hole as they run.
Nice mental image. Vaguely Looney Tunes-ish.
Mao Cheng Ji 04.17.13 at 2:01 pm
“There may be no good reason for companies to insist on degrees for those jobs”
It’s probably not so much about any expertise/knowledge, but rather about conformity, a verifiable degree of domestication.
Tim Wilkinson 04.17.13 at 2:03 pm
The point about education:
an association can be observed between education and mobility chances; but it is important to recognise that this association relates to the individual rather than to the population level. Education has an effect on who is mobile, or immobile, rather than on the overall rate of mobility
can apply whether not a demand constraint actually bites: in the past this wasn’t the case, and everyone who got higher ed could get a ‘higher’ job – but even where the constraint bites and there aren’t enough top jobs for all graduates, there will still be an association – albeit weaker – if there are some non-grads, and if they are less likely to get the plum positions. As others have argued, supply does make its own demand in the sense that as more people get degrees, more positions – and in particular, lower status/pay jobs – will be limited to graduates only. If not enough higher status jobs exist, then a proportion of grads will end up being ill-served by their ‘investment’ in higher ed., and over time will tend to bid down the market value of those qualifications to the detriment of others in their class.
The idea of a bottleneck which means that while any one member could escape their class, not all or even very many could do so at once has a close affinity with Jerry Cohen’s arguments in The Structure of Proletarian Unfreedom.
Tim Wilkinson 04.17.13 at 2:08 pm
But I’m not sure about this:
what would be involved in redressing a situation of declining relative mobility, as claimed by Blanden et al., would of necessity be increases in upward mobility that were offset by exactly corresponding increases in downward mobility – which was not, one could suppose, what Prime Minister Blair had in mind….while the economists’ results relate to relative mobility, politicians remain primarily concerned with absolute rates – and indeed with absolute rates of upward mobility – and are unable, or perhaps in some cases unwilling, to see the mismatch…politicians, in one breath, urge that the decline in mobility should be reversed and, in the next, insist that mobility should not be seen as a ‘zero-sum game’ in which upward and downward movements have to balance out – when the only evidence they have of
a decline relates to mobility understood in just this zero-sum sense.
I think this actually involves a certain confusion of relative with absolute mobility. The “exactly corresponding increases in downward mobility – which was not, one could suppose, what Prime Minister Blair had in mind” are of course downward relative mobility – and this does not entail any downward absolute mobility. There’s no contradiction in wanting to increase relative mobility (which relates to equal opp’ty and so-called ‘meritocracy’) while aiming to avoid downward absolute mobility. At least this seems of a piece with standard NuLab conceits about the ‘politics of envy’, rising tides, people getting filthy rich etc., as well as wider neoliberal themes of positive-sum trades, Pareto-improvement, the irrelevance of relative wealth, etc.
The twin aims of increasing equality of opportunity (maximising relative mobility across the board) and avoiding making anyone worse off (avoiding downward absolute mobility) are reconcilable in theory, and that seems to be what NuLab wanted to do, or appear to do (or appear to want to do). It doesn’t seem to me to involve the mistake Goldthorne diagnoses.
Tim Wilkinson 04.17.13 at 2:20 pm
(Apologs for multiple posts – I’ve come to the thread late.)
It seems to me that one source of confusion – which may have been cleared up already – could lie in the difference between the natural language expressions “relative rates of mobility” and “rates of relative mobility”. The former would involve comparing the mobility of those from different starting classes (and might better be called ‘comparative” than “relative”, esp. if these adjectives are going to be used of probabilities). This is not something the Goldstone paper is particularly interested in – the reader is left to eyeball the contingency tables for that purpose, as well as for the purpose of ascertaining changes in fluidity (relative mobility) between the 58 and 70 stats.
In contrast, relative mobility doesn’t necessarily involve comparisons across starting classes, but instead can – for suitably ‘well-ordered’ variables, such as income, be based on normalising the ending-class categories so as to ensure their relative sizes are the same as those of the corresponding starting classes. For most purposes one would also require normalising the starting-class categories so that all have equal membership – certainly this would be the case if one wants, as G does, to treat a 5-by-20% distribution as the expected outcome of ‘perfect mobility’, i.e. equivalent to a fair lottery.
Referring to table 1 in Goldthorne’s paper: these two normalisation constraints are satisfied automatically where the respective ante/post income quintiles are used. There is no need for any calculation of odds ratios or regression models there. But The class mobility tables have been adjusted (see Mosteller, 1968) so as to have all marginal percentages at 20%, in the same way as the quintile family income/earnings tables, while preserving the underlying odds ratios that express relative mobility rates.
I’m not sure how this works: presumably the adjustment is done for the ‘ending category’ figures – assuming there is any structural difference, as the exercise supposes, it wouldn’t be possible to adjust the figures for both starting and ending categories so that both fit a 5-by-20% profile, while at the same time preserving the membership conditions for these categories and maintaining the essential aspects of the data.
Presumably a set of numbers summing to 100 across rows and down columns have been supplied which fit – to some approximation – the odds ratios found between competing pairs within the original data. If this includes ‘backwards causation’ odds ratios – i.e. intra-column comparisons, relating to the odds that a member of a certain ending class originated from a certain starting class, I can’t see how it could be done at all.
In any case, the use of odds ratios isn’t essential to the difference between relative and absolute mobility – it’s only required, as an intermediate step, in constructing ‘relative’ mobility figures based on data which uses ‘absolute’ categories. Or at least that’s what Goldthorne seems to have done in this case.
Fill 04.17.13 at 2:22 pm
@70
“the main complaint of many businesses is that higher-ed does a relatively poor job of developing the skills entry level employees might require (which tend to be interpersonal skills and work-ethic related rather than writing or analysis) so while they might prefer to train as little as possible – its not a matter of passing that cost to the state, its a matter of the entry level worker has few skills to begin with”
Lol, just because they say it or complain doesn’t mean its true.
The same companies also say they want flexible workers at the same time they say they want workers with specific skills and years of experience.
Heads I win, tails you lose.
Tim Wilkinson 04.17.13 at 2:48 pm
(One last addition)
It strikes me actually, given that I’m not sure that a clear measure of relative mobility has been presented (along the lines of GINI for income equality), that it might be worth looking at class immobility, as represented by the diagonal in normalised, relative mobility, contingency tables:
1958 . . . . . . . . . . . . . . . . . . . . . . ..Inc. 5ile 1 . . . .2 . . . .3 . . . .4 . . . .5 . . . .(Tot)Top . . . 31 . . . 22 . . . 22 . . . 15 . . . 11 . . . 1012 . . . . 22 . . . 22 . . . 21 . . . 18 . . . 17 . . . 1003 . . . . 17 . . . 17 . . . 22 . . . 21 . . . 22 . . . 994 . . . . 15 . . . 19 . . . 21 . . . 22 . . . 22 . . . 99Bottom . .14 . . . 19 . . . 14 . . . 24 . . . 28 . . . 99(Tot) . . 100 . . .101 . . .103 . . .104 . . .105..Rel. Class I . . . .II . . . IIIa+V . VI . . . IIIb+VII (Tot)I . . . . 36 . . . 28 . . . 18 . . . 9 . . . .9 . . . .100II . . . .28 . . . 27 . . . 19 . . . 16 . . . 10 . . . 100IIIa+V . .16 . . . 20 . . . 23 . . . 19 . . . 22 . . . 100VI . . . .13 . . . 17 . . . 20 . . . 26 . . . 24 . . . 100IIIb+VII .7 . . . .9 . . . .19 . . . 30 . . . 34 . . . 99(Tot) . . 100 . . .101 . . .99 . . . 100 . . .99...1970 . . . . . . . . . . . . . . . . . . . . . . ..Inc. 5ile 1 . . . .2 . . . .3 . . . .4 . . . .5 . . . .(Tot)Top . . . 37 . . . 19 . . . 22 . . . 15 . . . 8 . . . .1012 . . . . 24 . . . 25 . . . 20 . . . 17 . . . 14 . . . 1003 . . . . 15 . . . 22 . . . 22 . . . 22 . . . 18 . . . 994 . . . . 14 . . . 18 . . . 19 . . . 24 . . . 26 . . . 101Bottom . .10 . . . 16 . . . 17 . . . 22 . . . 35 . . . 100(Tot) . . 101 . . .102 . . .103 . . .104 . . .106..Rel. Class I . . . .II . . . IIIa+V . VI . . . IIIb+VII (Tot)I . . . . 39 . . . 23 . . . 19 . . . 9 . . . .10 . . . 100II . . . .26 . . . 28 . . . 19 . . . 15 . . . 12 . . . 100IIIa+V . .16 . . . 19 . . . 22 . . . 24 . . . 19 . . . 100VI . . . .11 . . . 15 . . . 21 . . . 28 . . . 25 . . . 100IIIb+VII .8 . . . .15 . . . 20 . . . 24 . . . 34 . . . 101(Tot) . . 100 . . .100 . . .101 . . .100 . . .100The values in those diagonals show the percentage probability of staying in that category from one generation to the next. The figures for the 1970 cohort are appreciably higher than those for the 1958 cohort – though I know Goldthorne has problems withe 58 income data (and I’m not sure how usable the ‘relativised’ class data is, given the statistical operations it’s been through.)
Going by Goldthorne’s benchmark of 5×20% as the ‘perfect mobility figure, a multiple can be derived, showing how much higher the immobility-probability is than would be expected under perfect mobility. I’ve used probability ratios rather than odds ratios because the latter are hard to interpret – e.g. a prob of 0.5 as against 0.25 has a prob ratio of 2, which is easy to understand, but an odds ratio of 3, which is much less intuitive.
So dividing each of the percentage point values by 20, we get:
1958 . . . . . . . . . . . . . . . . . . . . . . . .Inc. 5ile 1 . . . . 2 . . . . 3 . . . . 4 . . . . 5immob. . .1.55 . . .1.1 . . . 1.1 . . . 1.1 . . . 1.4.Rel. ClassI . . . . II . . . .IIIa+V . .VI . . . .IIIb+VIIimmob. . .1.8 . . . 1.35 . . .1.15 . . .1.3 . . . 1.7.1970 . . . . . . . . . . . . . . . . . . . . . . . .Inc. 5ile 1 . . . . 2 . . . . 3 . . . . 4 . . . . 5immob. . .1.85 . . .1.25 . . .1.1 . . . 1.2 . . . 1.75.Rel. ClassI . . . . II . . . .IIIa+V . .VI . . . .IIIb+VIIimmob. . .1.95 . . .1.4 . . . 1.1 . . . 1.4 . . . 1.7.Obviously this doesn’t provide a single figure a la GINI, and some info has been lost, but it seems like it could provide a useful summary measure of relative (or indeed absolute) mobility, without requiring a lot of manipulation of data (except whatever may have to be done to get data from absolute categories into quasi-relative form, as with Goldthorne’s adjusted data).
Random Lurker 04.17.13 at 3:01 pm
“The same companies also say they want flexible workers at the same time they say they want workers with specific skills and years of experience. ” Fill 77
They want workers with years of specific experience, and then they want them to flex.
john goldthorpe 04.17.13 at 3:02 pm
@ tim wilkinson
The adjustment to which you refer is a quite standard one. It allows you, with any mobility table, to change both marginal distributions to whatever you want while preserving the odds ratios implicit in the table – odds ratios being, as others have explained, the measure sociologists use of relative mobility. In this case, I adjusted both marginal distributions so that each of five classes accounted for 20% of cohort members and their fathers so as to match quintile distributions using the economists’ earnings/ family income data.
On your earlier point, if relative rates become more equal – i.e. odds ratios move closer to 1 which would imply ‘perfect mobility’ – then mobility both upward and downward must increase to exactly the same extent. What one might call the ‘inherent stickiness’ between parents’ and chidren’s positions, – i.e. net of all structural change – just gets weaker. So Blair was indeed confused.
engels 04.17.13 at 3:15 pm
the 21st century student is not only running hard to stay in place, in many cases they are digging themself a hole as they run
Personally, I have found it to be more like this.
Alex 04.17.13 at 3:18 pm
you don’t learn anything particularly relevant to the UK workplace by being a Buddhist monk for two years
like school, sitting down, shutting up, and taking orders. also, accepting a fundamentally absurd universe with serenity.
Trader Joe 04.17.13 at 4:10 pm
@77
My point wasn’t really the particulars of what employers complain about and whether those views are fair. Sometimes its true, sometimes not it really depends on both the specific employer and employee.
The post I was responding to asserted a general need for training and that employers didn’t train in preference to shifting costs to others. My comment was aimed at the idea that a lot of the training that is required doesn’t need formal and expensive coursework but is rather more experience driven (i.e. interpersonal skills and work ethic) and that spending money on training usually just benefits a different employer – that’s what I was trying to get at.
Tim Wilkinson 04.17.13 at 4:13 pm
JG – thanks for response. Yes, I assumed it was standard, and didn’t intend to imply that anything ‘fishy’ was going on. I didn’t have access to to the cited Mosteller ’68, and might not have got far with it anyway. If this is not a matter of a best-fit model, I would guess that only odds ratios within rows are preserved, and not those within columns? The former would be the relevant ones for the forward-looking concept of social mobility, and I’m pretty sure both couldn’t be preserved.
I’ve no problem believing that Blair was confused, or unconcerned, about the relative/absolute distinction. My point was only that the underlying position – with that distinction applied correctly – is coherent, if not very plausible. Increase relative mobility but eliminate downward absolute mobility – indeed that’s more or less what underlies Browns appeal to a situation of ever-increasing ‘room at the top’ in which – the great political attraction of the position – no need arises for [absolute – TW] mobility to become a zero-sum game – even though having raised everyone to the top absolute class, in some cases leapfrogging others on the way, ‘meritocratic’, equal-opp’ty, jockeying for relative position within that class will presumably take place.
Fill 04.17.13 at 4:22 pm
@79
If you’re going to flex afterwards, what’s the point of gaining the specific skills and years of experience in the first place?
I work in a stem industry. What you often see are a lot of highly qualified people getting pushed out of industry and then they have difficulty finding jobs afterwards b/c their skills are so specific and hard to transfer. They then have to start a whole new career usually starting from the bottom (which includes going back to school and getting re-trained)
The same companies who push workers out of industry are the same companies who cry about not being able to find trained or qualified workers.
BrendanH 04.17.13 at 8:37 pm
Tim W, (at the risk of getting really nerdy), “odds ratios within rows … not those within columns” doesn’t make sense because ORs cross both rows and columns.
Any table can be decomposed into its marginals (row and column totals) and (r-1)*(c-1) independent odds ratios. These (r-1)*(c-1) ORs can be considered a representation of the structure of association. A 2X2 table has one OR and it doesn’t matter if you view it row-wise or column-wise (one way is the inverse of the other). For a 2X2 you can apply any marginals and reproduce a table with different data and the same OR, by simple arithmetic. For larger tables it is only slightly more complicated. It can also be done by predicting values from a saturated loglinear model (which will have (r-1)*(c-1) interaction terms corresponding to the ORs, capturing the structure of association) where you manipulate the row and column parameters to get your chosen margins.
Tim Wilkinson 04.18.13 at 8:05 pm
BrendanH thanks – I now get it. I won’t test anyone’s patience further.
Comments on this entry are closed.