Social Mobility and Statistical Immobility

(Crossposted to Core Economics)

I typically find that there’s a great benefit in posting draft papers online, and getting feedback before the paper finally goes to the journal. But I’m also learning about possible disadvantages.

In 2006, I wrote what I think is the first Australian paper to have estimated the ‘intergenerational elasticity of earnings’ for fathers and sons. In other words, what impact does a 10% rise in fathers’ earnings have on sons’ earnings? (We use fathers and sons rather than mothers and daughters due to lower female participation rates in previous generations). I presented a preliminary version at a conference in Germany, and the conference organisers posted it online.

In the initial version, I had a very low estimate of the intergenerational elasticity (IGE), putting it in the range 0.14-0.19. If true, this would have made Australia one of the most socially mobile countries in the world (the higher the number, the less mobility across generations). Actually, the reason the number was low was the methodology, and at that stage, I still needed to figure out how much to adjust it upwards. The text included these qualifications, but didn’t give the reader a better answer than 0.14-0.19.

It wasn’t until my paper was published in December 2007 that I worked out the best methodology for adjusting the figure, and suggested that the true Australian IGE was probably in the range 0.2 to 0.3. In other words, a 10% increase in a father’s earnings translates into a 2-3% increase in his son’s earnings. In terms of international rankings, this suggests that Australia is more socially mobile than the US (IGE=0.4-0.6), less socially mobile than Scandinavia (IGE=about 0.2), and approximately as socially mobile as the UK. Given that’s where we fit in the inequality rankings, this feels about right to me. For anyone who’s interested in the topic, here’s the final version, which also estimates changes over time (bottom line: no evidence that Australian social mobility changed from the 1960s to the 2000s).

But by December 2007, the draft number (implying that Australia is one of the most socially mobile countries in the world) had a life of its own. In 2007, an OECD report picked up on it and included the centrepoint of the 0.14-0.19 range (0.165) in a survey article. Later that year, an Economic Roundup article repeated it. Both were inadvertent, and I emailed both authors afterwards. But that didn’t seem to help in the case of the OECD. In December 2008, a new OECD report (Growing Unequal, chapter eight) carried out extensive statistical analysis using the 0.165 number. I’m grateful for the citations, but can anyone help me get a message to Paris that if they want a point estimate, an Australian intergenerational elasticity of 0.25 would be closer to the truth?

This entry was posted in Inequality. Bookmark the permalink.

5 Responses to Social Mobility and Statistical Immobility

  1. conrad says:

    You could look on the bright side — at least they are copying the initial number correctly. In case you haven’t noticed it yet, in education (and to a lesser extent health), you often start with a number, and then it changes over time without any evidence (almost always for the worse). So you end up seeing figures stating that there is more and more dyslexia, cancer, and so on than the actual real figures suggest. In some areas, you can actually find almost any figure you want.

  2. Kevin Cox says:

    This is a problem with all personal data.

    The solution is to allow an individual to have control over all their personal data. For example if I discover that this post is wrong and I no longer wish to be associated with the opinions then I should have the right to remove it. I should definitely have the right to remove it if it was put up by someone else pretending to be me.

    What this means is that information proporting to be from me should be “signed” by me and if I no longer want it to be public then I should have the right to remove it or to say I no longer wish to be associated with the opinion or the piece.

    This does not have to apply to everything on the web but we could implement it if we have a little symbol saying “signed by and authorised by” the individual. Another “no longer agreed to by the individual”.

    This would be easy to implement and we could start with the people who write the software for blogs.

  3. stuart says:

    Interesting to see how well some of these institutions check there facts… Having read the draft report I was suprised how well Australia performed comapred to some of the nordic countries, so these updated results come as less of a surprise. Its a very interesting topic so looking forward to more posts on this in the future!

  4. Pingback: Är den amerikanska drömmen svensk? « Ekonomistas

  5. For those of you who do not read Swedish (see the previous comment), I posted earlier today a text on my Swedish blog, discussing a new report on social mobility in Sweden and elsehwere. In the report, Andrew’s preliminary and understated estimate of IG-mobility was mentioned and to help Andrew communicate the correct estimate I referred to his errata on this blog.

    Hopefully this will enlighten at least some Swedes that Australia is not fully as mobile as the Nordic countries, at least not yet (there are signs that Swedish inequality is increasing, which according to Andrews work elsewhere should indicate that also mobility will decrease eventually).

Comments are closed.