Why is father-son intergenerational mobility stable?

Using data from four surveys over a 40 year period, Andrew Leigh’s latest paper (pdf) argues that father-son intergenerational mobiliity is stable in Australia.

In his conclusion, as reported by The Australian, he says:

“On one view, the absence of any significant rise in inter-generational mobility might be regarded as surprising,” Dr Leigh says in the study report.

“Increases in healthcare coverage, the banning of racial discrimination, the abolition of up-front university tuition fees and an increase in the number of university places are among the policy reforms that might have been expected to increase inter-generational mobility.

“Yet there were also trends in the opposite direction.” These included rising unemployment during the study period and the removal of inheritance taxes in 1979. Dr Leigh said a rise in inequality had been well-documented with the distance between income groups greater in the early 2000s than in the mid-1960s.

I know conclusions are where they let academics off the referee’s leash, but some of this seems a bit odd to me. I’m not sure why improved healthcare coverage would make much difference; in any case claimed ‘disability’ is much higher than it used to be. Nor are bans on racial discrimination likely to have influenced the figures much; such laws are easily evaded and many migrant groups were doing fine long before discrimination was outlawed.

Inheritance taxes, as Andrew L’s own research shows (pdf), took only about 30% of the largest estates – something unlikely to affect the salary data he’s using (though it may affect investment income). And rising inequality is consistent with high mobility (if the poor and the rich swapped places each generation, ie complete mobility, inequality would be unchanged because it is a static, snapshot-in-time measure).

As we have discussed recently, the Whitlam university reforms problem had little or no overall effect on mobility (despite the cherished anecdotes of some commenters and the PM). But I am pretty sure that, contrary to what Andrew L suggests, the later increase in places did have a positive impact. The data show a significant increase in the proportion of young people from working class backgrounds going to university, and good number of them have surely made it into much better jobs than their parents had.

So why isn’t it showing in Andrew L’s figures? I think it is at least partly because he is looking at sons, when most of the expansion in university places has gone to daughters. Women have been a majority of university students since 1987, and the consequences are showing up in the workforce. The 2006 census found that over the previous decade the number of female professionals increased by 44%, while male professionals increased by only 28%; for managers women increased their employment more than twice as quickly as men: 25% to 11%.

Many of the jobs working class guys took to advance themselves in the 1960s would now be taken by women who did better at school and have superior social skills. Compounding the problem, jobs requiring physical strength have declined relative to total employment, so some of these guys cannot even maintain their father’s occupational status. They end up unemployed or ‘disabled’. If all these factors are at play, stable father-son mobility is an excellent result.

29 thoughts on “Why is father-son intergenerational mobility stable?

  1. I think you are spot on, Andrew. The ‘problem’ is that if intelligence is largely genetically-determined and personality is heavily influenced by the familial environment, the sweeping away of barriers to merit-based advancement is likely to lead to lower rather than greater mobility. Perhaps this will ultimately require the left to choose honestly between equality of opportunity and equality of outcomes.


  2. Rajat — I’m not sure why you think intellingence is “largely” genetically determined — do you mean the softer large-part? In addition, the correlation between intelligence and income isn’t thrilling (r = .3 if I remember).
    Andrew — I’m not sure why women taking men’s jobs is a problem, and physical strength jobs disappearing being a compounded problem. Surely women in the workforce adds to workplace productivity, so its a win-win situation, as there are likely to be more jobs around, many of which hopefully don’t involve digging holes.


  3. Conrad – I meant the problem for male upwards social mobility. While on balance the social and economic change of the last 40 years is for the better, one thing I do agree with Bob Birrell on is that working class males have generally been the losers from this change.


  4. I’m not convinced of that argument (excluding older males — although I could be with more data). To me a lot of people that would have become working class males didn’t because of the changes. Thus whilst those who are left are certainly close to the bottom of the heap, there are lots of people that are not in it that would have been otherwise. I also believe that its most certainly a richer heap because of society being richer in general (which is why we can afford things like a high minimum wage, excellent free health care, and so on). Thus whilst that group might not have benefitted as much as other groups, they still benefited — so they aren’t losers at all.


  5. Rajat

    Like Conrad, I think that you need to be careful about saying that intelligence is largely determined by genetics. For a (what seems to me) good review of the literature see Neiser et al “Intelligence: Knowns and Unknowns”

    Click to access Neisser_et_al_1996_.pdf

    A general review of intergenerational mobiliy in OECD countries is at http://www.oecd.org/dataoecd/27/28/38335410.pdf

    This drew on an earlier version of Andrew Leigh’s paper for some results for Australia

    On the inheitance of IQ, one of the points made in the OECD paper is that ” heritability estimates are specific to populations and environments for which they are measured – e.g. socioeconomic status affects the heritability coefficient (Heckman and Carneiro, 2003). Turkheimer et al. (2003) documents that the share of IQ variance attributable to genes and environment vary non-linearly with SES and that, while in poor families 60% of the variance in IQ is accounted for by the shared environment (and the contribution of genes is close to zero), in affluent families, the result is almost exactly the reverse. The same authors also argue that richer families help their children to eliminate the effect of their defective genes and to enhance their productive genes; for example, richer parents may offer a treatment to their child if he is aggressive; conversely, low-income parents might not be able to do so.”

    I think that it is also worth noting that income mobility varies widely acoss developed countries, and it is hard for me to see how this can be largely explained by genetic factors Some countries with high income mobility are ethnically homogeneous – Sweden and Denmark, but Canada and Australia also have relatively high mobility.

    Also when you say ” the sweeping away of barriers to merit-based advancement is likely to lead to lower rather than greater mobility”, I’m afraid that I don’t follow your argument – why?


  6. “Also when you say ” the sweeping away of barriers to merit-based advancement is likely to lead to lower rather than greater mobility”, I’m afraid that I don’t follow your argument – why?”

    I presume because most of the advantages would go to the families able to meet the criteria of merit, ie the middle and upper classes, and particularly over the last few decades women. So greater gender equality would work against economic equality (which with assortative mating has almost certainly been true).


  7. Conrad, Peter, I’m certainly not an expert on IQ, hence the “if” in my comment. But Wiki suggests that 70-80% of adult IQ variability can be explained by genetic factors. It also discusses correlations between IQ and income, job performance, etc. While these correlations might not be huge, IQ still seems to be one of the largest single explanatory factors. And yes, if educational opportunities increase, IQ may become an even more important explanatory variable over time.


  8. Andrew

    But gender equality and economic equality are not the same as mobility. Higher immobility implies that people tend to be more likely to stay in the same economic position as their parents, irrespective of their efforts or random effects or even their innate capacities, unless by innate capacities you mean what is purely a genetic inheritance. If unused capacities or potentials are greater in the bottom part of the income distribution and over-rewarded capacities are greater in the top half, then sweeping away barrier should increase mobility.


  9. Peter – Sorry, I was blurring issues (I made a similar point about equality and mobility to you in the original post). But if the number of better-paid jobs is limited, I think Rajat’s point was that a merit basis of distributing jobs, rather than one that was biased against women (as in the past), will disadvantage men who do not do so well on the usual merit criteria.


  10. I recently read a study by UK psychologist Daniel Nettle (available here) that used data from the National Child Development Study, which looks at all people born in a particular week in 1958. It finds a general ability test conducted at age 11 was a strong predictor of class mobility.

    There is nothing on the IQ of parents, but the research I have read suggest heritability of not lower than 30%. Given that there have been significant opportunities for mobility for a significant period in Australia and other Western countries, and the strong relationship between intelligence and mobility, this would suggest that we would see relatively low downwards mobility in the top group and relatively low upwards mobility in the bottom group – with the top group have a disproportionate number of high-intelligence persons and the bottom group a disproportionate share of low-intelligence persons.

    Social conditions and trends could amplify or diminish these tendencies; with academic merit being used to allocate jobs an amplifying force.


  11. Andrew

    The OECD paper I referred to does show that mobility tends to be lowest in the top and bottom of the distribution, and that most movement is in the middle 60% of the income distribution. So yes, the top income groups could be expected to have a disproportionate share of more intelligent persons and the bottom quintile of less intelligent. This will explain some of the differences in economic rewards to different groups, but the challenge is to identify how much of this difference is not amenable to change and how much can be changed.

    For example, performance in general ability tests at age 11 will partly reflect genetic endowments, but will also reflect the first five or six years of school and the family environment.

    The point is that you have to get from the inherited bit of IQ to intelligence to ability to utilisation of ability, influenced by a range of complex factors. So while inheritance of IQ is certainly an important factor number of studies cited in the OECD paper find it actually accounts for less than half of observed differences in father-son outomes. Bowles and Gintis find for example that education is more important influence than intelligence on outcomes.

    It is also worth noting that the “intergenerational income elasticity” measures the fraction of income differences between parents that are transmitted, on average, to their children . For example, a value of the elasticity of 0.5 implies that half of the relative difference in parental incomes is transmitted, on average, to the children (e.g as in the USA and the UK). In the case of Australia using Andrew Leigh’s estimates this means that only about 20-30% of the difference between parents’ incomes is transmitted to their children. So for example this would imply that if you had two sets of parent where one family had incomes twice as high as another, then on average you would expect their children to have incomes 20-30% higher than the children of the low income parents. This is a quite high level of convergence.

    A couple of other points. Most evidence suggests that market incomes and earnings have become more unequal over time in Australia (but disposable income less so, because of the tax and transfer systems). So we appear to have widening earnings inequality but apparently stable earnings mobility, so the link between the two trends is affected by other factors.

    On the contribution of wives to family earnings inequality, Australia does have a problem that at the bottom of the distribution we have a relatively high share of no-earner couples. But the rate of non-earner couples is about 4 times higher in Australia than in Denmark or Sweden. What explains this difference? I would argue that most of this difference is driven by the nature of welfare policies and not by any inherent characteristics of low income Australian couples. (The UK, New Zealand, and Ireland also have high shares of completely jobless couples.)

    Finally, Andrew I am a bit surprised that you say that the number of high paid jobs is limited. Mike Keating and the ACTU (separately!) have done studies showing that in the last 20 years the share of high paid jobs has increased in Australia.


  12. Peter – I did carefully say ‘if’ the number of better-paid jobs is limited; the ABS data clearly shows the professional workforce is expanding (the managerial workforce is expanding, but is shrinking slightly relative to the total workforce). But that does not mean that everyone who wants such a job can have one – there are consistently hundreds of thousands of graduates without professional or managerial jobs, some of whom would like one. There is a complex relationship, as you would be aware, between demand and supply in the labour market.

    And as I noted in the original post, women are doing much better in the managerial and professional labour market than men. (Probably largely reflecting their better prior educational achievement.)


  13. Conrad – A table here showing significantly declining labour force participation for low-education men compared to increasing participation for women across all educational groups. I expect the numbers would be a little better now, but the overall trend is clear.

    Birrell’s point was also that these men have relatively low rates of partnering and marriage, so they are without the two main sources of male life satisfaction.


  14. I’m not sure if Andrew Leigh discussed this in his paper, but I would be very much interested in Australian studies looking at the intergenerational mobility of immigrants. In an ACER study http://www.acer.edu.au/documents/LSAY_lsay22.pdf the researchers found that: “Year 9 achievement cannot explain the high performance of Asian students in Year 12. Their tertiary entrance performance is well above what was expected given their Year 9 achievement scores…For the ‘Asia’ group, socioeconomic differences only very partially explain their higher performance”


  15. Brendan – Andrew L does do a calculation which excludes all respondents who were born overseas or with fathers born overseas, which he says suggests that natives might have slightly higher mobility than migrants. Given what we know about migrants and education, I’d be surprised if the children of migrants had lower mobility.


  16. Andrew, a terrific post (“off the referee’s leash” is a beautiful barb), which has prompted a fascinating comments thread. It’s a pity that we have so little evidence (from any country) on the impact of policies on intergenerational mobility. I reckon universal healthcare and more university places would increase it; you say the opposite. And while we can tell persuasive stories, there’s nothing I know about to provide firm facts either way. Tis one reason that I’d like to keep working on this topic. Another is so that we can better understand women – a clear omission from my research so far.
    One minor thing: you say that “rising inequality is consistent with high mobility”. As a theoretical point, this is true. But empirically, it turns out that unequal socieities also tend to be more immobile than equal ones.


  17. AL – I do think more university places make a difference, but I think it is not showing more strongly in your data for a couple of possible reasons. One is that it is really only among people born after 1970 that we see a dramatic surge in the proportion of people from working class backgrounds attending university. The effects of this are still working their way into the data I imagine.

    The second, as noted above, is sons vs daughters. In the late teen kids I have been looking at in census data working class sons are about one-third less likely to go to university than working class daughters. So the mobility is less likely to show if you look at sons rather than daughters.

    Incidentally, young working class men are the only group I can find who become less likely to attend university as family income increases (and more likely to attend TAFE). I suspect it is because the see their fathers doing ok in a manual occupation, and don’t see the point in university. More university places won’t interest them.

    I read your inequality paper after I wrote the post; I suspect that high inequality and low mobility have common drivers, such as bad schools, rather than inequality causing low mobility.


  18. Andrew

    You say “I suspect that high inequality and low mobility have common drivers, such as bad schools, rather than inequality causing low mobility.”

    As Andrew L noted there is a negative correlation across countries between mobility and inequality – countries with low inequality tend to have high mobility (Denmark, Sweden) while countries with higher inequality tend to have lower mobility (USA, UK). The difficult-to-explain countries are Australia and Canada, which have mobility rates at the Nordic level, but higher inequality (but not the US level). Like Brendan, I suspect this is partly because of the higher mobility of migrant children in these countries. PISA shows that in Canada the children of migrants have higher attainments on average than the children of native-born, and in Australia the difference is negative but relatively very small.

    While in a general sense it may be true that poor schooling is associated with higher inequality and lower mobility, presumably this would be caused by changes in the difference in educational quality – low income children go to schools of declining quality, but higher income children go to schools where quality is either stable or improving.

    However, what we appear to see in Australia is increasing earnings inequality but stable earnings mobility. This either implies that changes in the distribution of educational quality have differential effects on mobility and inequality, or other changes modify the effect on mobility or perhaps on inequality.

    Sounds like there is an interesting research agenda to me!


  19. I just did some quick calculations on the backgrounds of university students compared to the general population. 22.8% of domestic students in Australian universities in 2006 were born overseas. The census finds that 19.9% of people in Australia aged 15-24 were born overseas. This suggests that the overseas-born are slightly ‘over’ represented at university relative to population. However, contrary to stereotype it may not be Asian-born students driving this. They make up 7.8% of the population of that age group but 7.5% of domestic students. However, this does not include Australian-born children of Asian migrant parents.


  20. AN: I suspect that high inequality and low mobility have common drivers, such as bad schools, rather than inequality causing low mobility.

    The other possibility is that low mobility causes high inequality. In political systems that block mobility (ie more socialist systems) they often had high levels of inequality.


  21. “what we appear to see in Australia is increasing earnings inequality”

    Peter – is this increasing inequality of (total) earnings or hourly earnings?

    If it is the former, not only does Australia have a higher than average proportion of no-earner families, but it has a higher proportion than nearly every other country of families with part-time workers. That could also contribute to us having higher earnings inequality (at both the individual and household levels) than most other comparable countries.

    As you pointed out, policy settings have a part to play in all of this, but to the extent that inequality is a product of the choices that women in particular make about the extent to which they supply their labour, I find it hard to see that as a bad thing.


  22. And if the increasing inequality is being driven by the fact that higher paid jobs are growing faster than low-paid jobs, as suggested by Mike Keating, how can that be a bad thing either?


  23. John

    Actually, virtually all socialist systems had very low levels of measured income inequality before 1989 in Eastern Europe and the FSU and before economic reforms in China and Vietnam, mostly even lower than in countries like Sweden or Denmark. This is because they were command economies and set wages by government fiat, and they guaranteed employment, plus in Eastern Europe and the FSU having fairly generous social insurance systems.

    Of course the cost of this was low levels of productivity and national income, and very high levels of inequality in terms of political power.


  24. Backoom Girl

    As I undertand it from research by Jeff Borland and also by Peter Saunders there is growing inequality both of wages per hour and total earnings. Both sources are a bit dated, however, and do not go up to past 2000. Hourly earnings inequality has increased both for men and for women, but earnings gaps between men and women have declined significantly over the long run.

    This is not just a matter of more high skill jobs, because hourly earnings of low paid men fell in real terms for quite a long time (but dependent on the starting point), although hourly earnings for low paid women have increased in real terms (but not as fast as for high paid women.

    So my intepretation would be that some of the driving forces are ” good” things and some are not.


  25. It’d be surprising if high inequality and low mobility were not correlated, for obvious political economy reasons.

    A highly unequal society will put the power to decide policy on schools, health, etc into the hands of the better off (after all what’s the point of being rich, and hence powerful, if you can’t use it to give your kids a head start?). In the other causal direction, an immobile society may have more people socialised to accept their position, which leads to more acceptance of inequality.

    As I’ve said elsewhere, this is why you can’t have reasonable equality of opportunity if you have extreme inequality of outcomes.


  26. You have had very interesting debate on your post.

    Without straying unduly, and at the risk of repeating the comments I made elsewhere, can I point you to Andrew Leigh’s web site which draws our attention to the latest Maxwell poll on attitudes to inequality in USA. Of particular interest to this current debate is the polling on how Americans view such issues as
    – the “fairness” of the reward system,
    – whether citizens have an equal opportunity to succeed and
    – the extent of upward mobility (how easily it is to get up the ladder)

    They show an incredible smugness among Americans about equality of opportunity. It explains why there is not much pressure on governments to do anything about rising inequality and low income mobility. Are you aware, Andrew, of comparable figures for Australia?

    In my writings I have referred to polls showing that Australians have a greater tendency to blame the poor themselves for their predicament than their counterparts in Europe or Scandinavia and that only 15% of Australians disagree with the proposition that ‘people are rewarded for their skills’ – a low figure compared with a majority of the other countries surveyed but comparable with the US (Maxwell p. 8). But many of the Australian surveys are out of date. Do you know of any recent survey findings for Australia comparable to the ones posed by Maxwell above (see pp. 8-9 of Maxwell report)?


  27. Fred – Sorry for the delay in replying. I was having some computer problems yesterday that prevented me from accessing some files, but in the end they don’t contain a match on those questions – they are about living standards.

    For what it’s worth, this question was asked in 1993, 2003 and 2005:
    “The way things are in Australia, people like me and my family have a good chance of improving our standard of living.” Pretty much identical results all three times: 46-47% agree, 28% neither agree nor disagree, and 19-23% disagree.

    In 2003 and 2005, the following question was asked: “Compared to your parents when they were the age you are now, do you think your own standard of living is better, worse or the same as theirs was?” 76% and 73% respectively said better. A similar question in 1993 asked about the respondent’s father; 77% said better.

    This all looks very stable over the 1993-2005 period at least, but is more evidence that the rising tide is indeed lifting most boats than any particular mobility.

    On a quick search of social science data archive I found a 1992 survey with questions on getting ahead. Vast majorities thought natural ability and hard work were important, a smaller majority thought well-educated parents.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s