Digested Read – Spirit Level 4: Other Social Problems

Written By: - Date published: 12:22 pm, September 25th, 2010 - 40 comments
Categories: equality - Tags:

Digested Read Digested – Equality: better education and social mobility. Inequality: more teen pregnancies

Education is the future – it will be what creates or destroys the success of our society when the next generation has its turn to lead.  So what produces the best outcomes?  Good teachers in innovative classrooms, to be sure; but much more than that parental support.  Parents with higher incomes and more education themselves have children who do better – but direct parental involvement is even more important.  In turn children who do better at school will not just earn more, they will be more satisfied with their jobs, are far less likely to end up in prison, and more likely to be healthy and vote.  How do we get this outcome for more of our children?

Although New Zealand does quite well at education, in general more equal societies do better.  And they do better across the board.  New Zealand and the UK’s reading scores are very high for a few who pull the average up – Scandinavian countries are much more consistent.  Indeed in some research there’s a suspicion of under-representation of lower socio-economic groups in NZ and UK helping raise the average.

There are good reasons for more equal societies doing better.  Parents will do better with more support – and more equal societies tend to provide that.  Maternity leave is just the most obvious example: in the US & Australia there is no paid maternity leave at all, and in the US a woman may take only 12 weeks unpaid.  In Sweden 18 months of parental leave at 80% of salary may be taken by either parent (or split between them).

And the results of that support and the difference in income can be stark: a UK study showed that 3 year-olds from disadvantaged homes were already 1 year behind in their development compared to those from privileged households.

The psychological effects matter too – in blind studies lower socio-economic children do better (and the wealthy worse), than in studies where status is made clear before children take the tests.  Also interesting from a psychological point of view: although their educational results were better than less equal countries, more 15 year-olds aspire to less skilled work in more equal countries (eg 50% in Japan vs 15% in the US).  The result being that those who cannot achieve university education are much happier with their lot as those jobs are not so stigmatised.

The American Dream says that anyone can grow up to do anything.  But in fact social mobility is incredibly low in the US – if your father wasn’t wealthy, it’s highly unlikely you will be.

The number of countries with data here is low, which leads us to be cautious; but there has also been a large decrease in social mobility in the UK and US since 1980 – the same period that inequality has massively increased in those countries.

So far from inequality creating ‘incentives’ for people to move up, it instead creates sinkholes from which they cannot climb out.

The main driver of social mobility seems to be education – the connection to inequality of that is seen above – and in particular, publicly-funded education.  The level of public-funding of education is in turn tightly correlated to inequality.

Another result of inequality and lack of social mobility has been that those ‘sinkholes’ have become geographical.  As the wealthy move into gated communities, the poor end up left in ghettoes.  And those economically disadvantaged end up doubly so when they are surrounded by people in the same boat.  With the whole community lacking in resources, schools suffer, education outcomes suffer and social immobility is further enforced.  Crime and violence are often left as the only way forward.

I mentioned the strong link over time for teen pregnancy rates vs inequality in the US in my previous post, but here’s the graph for between countries.  It’s teen pregnancy births which is slightly affected by abortion rates, but a similar graph of the US states for conceptions shows a roughly parallel correlation.  Indeed the biggest difference between conceptions and abortions is for the wealthier half of society – they are far more likely to abort and keep their own future prospects improved, where the poor are likely not to expect any good prospects anyway.

Even within the statistics hides a worse picture for unequal countries: in Japan, Italy and Greece, more than half of the teen pregnancies are within marriage (86% in Japan), where in NZ, US and UK that figure is less than a quarter.  Another interesting statistic is that overall birth rates appear unaffected by inequality – suggesting fewer older mums to balance out the greater number of younger ones.

So why do less equal societies have more unmarried young mums – ‘babies having babies’?  Like the young male violence it seems this is the only way these young women know how to gain status and adulthood.  In turn they likely ruin both their own life chances (of education, work, social mobility, wealth…) and their children’s – those born to a teenage mother are much more likely to become one.

Absent fathers have a large part to play as well.  Girls who grow up without a father are much more likely to become a teenage mother.  But those young fathers from deprived backgrounds are unlikely to be able to offer much stability, income or support – and they have their own battles with inequality to occupy them.  So the young mums console themselves with a strong relationship with their baby – their best chance of intimacy in their chaotic lives.

Next Friday: Sustainability.

–

For more detail: Read the book. Buy it and/or support the Trust.

Right-wing trolls: r0b had a recent post with links refuting the arguments you’re about to make…

40 comments on “Digested Read – Spirit Level 4: Other Social Problems ”

  • Comments are now closed

  1. roger nome 1
    25 September 2010 at 3:01 pm

    (*Burt) “Equally you could say that warmer weather causes slower economic growth, since most countries situated close to the equator are relatively poor”.

    No Burt – there’s no way of tieing those together, and there haven’t been hundreds of neuanced studies on the topic that have been compared with each other to reach robust conclusions…

    Does that save us from Burt?

    *”Burt” is being used here only as a random name and in no way refers to a person and/or personality thaat comments regularly at The Standard.

  2. tsmithfield 2
    25 September 2010 at 3:58 pm

    The first graph plots “average maths and reading scores” over income inequality.

    This is clearly an instance where the correlation is around the wrong way. “Average maths and reading scores” could be considered a proxy for general IQ. General IQ is highly likely to predict income. For instance, in the link provided notice the high levels of poverty associated with low IQ. Those with IQ of 75 or less had by far the highest proportion living in poverty. It is also known that IQ is the best predictor of job performance. So people who are more intelligent are also likely to be more successful.

    • Vicky32 2.1
      26 September 2010 at 7:57 pm

      “General IQ is highly likely to predict income…..It is also known that IQ is the best predictor of job performance. So people who are more intelligent are also likely to be more successful.”
      If I was to point out all the exceptions to that rule, you’d probably yell ‘anecdotal evidence’, tsmithfield. But really, your arguments are utter bollices…

  3. tsmithfield 3
    25 September 2010 at 4:58 pm

    Outlier alert. Graph 2. Complete nonsense.

  4. tsmithfield 4
    25 September 2010 at 5:53 pm

    Why do you think I said any such thing?

    Here is what I said with respect to a graph from the previous article in the series:

    I actually have no problem with the outliers in the graph you mention. This is because it appears a median based regression type method has been used on this occasion rather than a mean based one, so the outlier doesn’t appear to distort the trend. Pity they hadn’t done the same in a few of their other graphs.

    All I am saying here is the analysis method is completely wrong for the data on this occasion.

    It is almost as if the authors have tried the method I have recommended above, found no correlation so thought, “shit, we’d better use a method that does give a trend”. This is one of the things that concerns me a lot. And should concern you as well. The authors seem to choose their analysis methods on the basis of what will prove their thesis.

    Another thing, why have they gone and cut out most of the countries on that graph (graph 2).

    • felix 4.1
      25 September 2010 at 6:08 pm

      Did you get a substantially different result?

      • tsmithfield 4.1.1
        25 September 2010 at 6:19 pm

        Clearly, if the same method was used as they had used in the graph I commended, there would be little or no trend at all. Besides the fact that there are far too few data points on that graph to use regression any way. You didn’t answer my question about why they excluded so many of the countries on that graph BTW. Compare that graph to the other graphs in this article for instance.

        • felix 4.1.1.1
          25 September 2010 at 6:28 pm

          I can’t answer your question because I know nothing about it – stats aren’t my area at all. I’m just trying to learn a bit more about where you’re coming from.

          • tsmithfield 4.1.1.1.1
            25 September 2010 at 6:51 pm

            Can’t blame you. Stats is the sort of area that if you haven’t been using it for awhile, its easy to forget. 🙂 So I don’t proclaim to be an expert at all.

            However, a simple way to see the problem is to compare a mean (average) with a median (middle) score.

            Consider the following series:

            15, 19, 22, 23, 24, 26, 160.

            Here, the score “160” could be considered an outlier.

            The average (mean) for the series above is 41.28. The median (middle score) is 23. Measures of central tendency should be quite similar in value in normally distributed data. But here you can see that the single score “160” is considerably inflating the average score above the median score, giving a result that is clearly out of line with the rest of the series.
            I would conduct this sort of analysis as a very basic test before determining how to treat the data in more complex analysis.

            In this sort of situation, one would normally either trim off the outlier if using a “mean” (average) based method of analysis such as least squares regression that appears to have been used on the graph. Or use a method of analysis that is not based on the “mean” (average) to determine the trend so that it is not unduly affected by the outlier.

            The problem I have with a lot of the “spirit level” stuff is that they seem to have chosen the method to suit what they want to find. For instance, there are some charts, as I pointed out the other day, where there is clearly a nice trend in the main body of data points, so they appear to have been happy to use a method that excludes the effect of obvious outliers. However, where there is no clear trend in the main body of data, they seem happy to use a method that is highly affected by that outlier.

            In graph 2 of this series, see what sort of trend line you would get if you covered up the UK and US. That is not to say these data points are unimportant. Just that the analysis method is wrong.

            I hope that helps.

            • RedLogix 4.1.1.1.1.1
              25 September 2010 at 7:26 pm

              Outliers should not always be rejected:

              In some data sets, there are values (points) called outliers. Outliers are points that are far from the least squares line. They have large “errors.” Outliers need to be examined closely. Sometimes, for some reason or another, they should not be included in the analysis of the data. It is possible that an outlier is a result of erroneous data. Other times, an outlier may hold valuable information about the population under study. The key is to carefully examine what causes a data point to be an outlier.

              http://cnx.org/content/m17094/latest/

              What look like outliers may actually be ‘influential points’ which can actually improve the reliability of the correlation.

              http://stattrek.com/AP-Statistics-1/Residual.aspx?Tutorial=Stat

              • tsmithfield …
                25 September 2010 at 7:39 pm

                Fair point. And I am aware of the need to study outliers.

                However, consider the point I have made below. I would rather show a weaker trend not relying on an outlier than include it and leave the door open to the critiicism. I like to understate rather than overstate data in a study. This makes the case much more convincing.

                If the trend is reliable it should be present regardless of outliers. In a number of the graphs I have seen from “the spirit level” the trend is entirely dependent on the inclusion of the outliers. Therefore, I still hold that it has not been justifiable to include the outliers, even on the basis of the point you have just made.

                In the case above, the authors would need to provide considerable justification as to why the outlier was included. That starts getting into the realm of the subjective. What some consider as important, others might not.

                Also, the inconsistency of how they have apparently used different analysis methods depending on whether a trend was obvious or not is something that is quite concerning.

        • RedLogix 4.1.1.2
          25 September 2010 at 6:40 pm

          You didn’t answer my question about why they excluded so many of the countries on that graph BTW.

          Unfortunately my father still has my copy so I can’t answer directly but it’s probably for the simple reason that usable data on ‘social mobility’ may only be available for those countries.

          Sure it would be nice to have perfect data for everything, but we don’t. That doesn’t mean we cannot make reasonable inferences from incomplete, imperfect information. Makes the job more interesting and fun.

        • Puddleglum 4.1.1.3
          25 September 2010 at 6:42 pm

          Hi TS, to quote the authors:

          “Comparable international data on inter-generational social mobility are available for only a few or our rich countries [i.e., those developed countries chosen for clear reasons expounded by the authors in their book and on the Equality Trust website]. We take our figures from a study by economist Jo Blanden and colleagues at the London School of Economics. Using large, representative longitudinal studies for eight countries, these researchers were able to calculate social mobility as the correlation between fathers’ incomes when their sons were born and sons’ incomes at age thirty. Despite having data for only eight countries, the relationship between inter-generational social mobility and income inequality is very strong.” (p. 159)

          So, in answer to your question: There’s only eight countries because there’s no data for others. Also, the relationship was found by economists from the LSE. While it’s no guarantee, you’d expect economists from the LSE would get basic stats right in a publication from the Centre for Economic Performance at the LSE – wouldn’t you?

          • tsmithfield 4.1.1.3.1
            25 September 2010 at 6:53 pm

            I don’t know, Puddlegum. I’ve seen some pretty bad stuff in published “peer reviewed” articles from time to time. See my reply to Felix above.

            • Puddleglum 4.1.1.3.1.2
              25 September 2010 at 7:39 pm

              TS, here’s a 2009 update from Joanne Blanden on what is now known about intergenerational mobility in international comparisons.

              You’ll note, in the Conclusions section, confirmation of the general inverse relationship between inequality and mobility but also the interesting difference in different mobility measures for Germany and the US. In the US income mobility is very low but there’s some mobility in terms of education and social class between generations. The reverse in Germany.

              At a glance, her statistical proficiency looks ok to me.

              • tsmithfield …
                25 September 2010 at 7:51 pm

                Thanks Puddlegum.

                I am not trying to argue if the construct is valid or not. It is the methods used in analysis that I have a problem with.

                In the case of graph 2, rather than try and find trends in the data points, a more interesting question to me would have been why the UK and US are so different than the other countries. That might have led to a similar conclusion. However, it would have been a much more valid way of getting there.

                • Puddleglum …
                  25 September 2010 at 8:48 pm

                  Fair enough.

                  Pearson’s r is affected by outliers. So let’s follow what you suggest and ignore the fact that the correlation for the eight countries is strong (r=0.93, p<0.01).

                  The funny thing is, if most people were asked what was different about the US and UK (compared with the other countries included) they'd probably say more economic freedom, neo-liberalism, etc.. Yet, oddly, they are the lowest on mobility, so that kind of ‘freedom’ doesn’t seem to generate mobility. Mobility here is measured by ‘father-son’ incomes, not educational attainment or social status.

                  I’d add in defence of W&P that if you read pages 159-160 in their book you’ll find that they are deliberately cautious and they only include it because of a range of other observations in the literature (e.g., on social mobility changes over time within the US and other countries, spending on education, etc., etc.) which they then go on to detail. They note that it is those additional observations that “lend plausibility to the picture we see in Figure 12.1”.

                  They’re not trying to pull the wool over anyone’s eyes, TS – it’s clear when you read the book.

                  • tsmithfield …
                    25 September 2010 at 9:26 pm

                    Puddlegum, the reason the correlation is strong is due to the inclusion of the outliers. So I wouldn’t read too much into those figures. Given there are only eight data points, a very steep slope is about the only way a low p value could be achieved.

                    I accept they have qualified their inclusion of the graph. However, I think there are much better ways they could have analysed the data to make their point. What they have done in presenting the data is really quite meaningless.

                    The funny thing is, if most people were asked what was different about the US and UK (compared with the other countries included) they’d probably say more economic freedom, neo-liberalism, etc.. Yet, oddly, they are the lowest on mobility, so that kind of ‘freedom’ doesn’t seem to generate mobility. Mobility here is measured by ‘father-son’ incomes, not educational attainment or social status.

                    Maybe But given we are talking about two countries out of many, there could be other significant differences as well that could account for the affect. We simply don’t know. Since there is a reliance on data from only eight countries, I simply don’t think there was enough data to make any conclusion.

                    What they could have done was simply summarized the previous research they cite to make their point rather than try and display data in a way that is really quite shabby.

                    • Puddleglum …
                      25 September 2010 at 11:04 pm

                      I think you’re getting a bit contradictory. You say “there are much better ways they could have analysed the data to make their point” and then say “Since there is a reliance on data from only eight countries, I simply don’t think there was enough data to make any conclusion”.

                      Either their analysis and display of data was ‘shabby’ but could have been done better through alternative analyses or there’s nothing in the data. Which is it?

                      What they did was show an analysis that was ambiguous and open to interpretation but could be supported by additional observations (data). In that context, the analysis was not ‘shabby’ but, rather, was consistent with what else is known about mobility and inequality. That is, it represents another brick in quite a big evidential wall.

                      Remember that, while the correlation is not robust, that doesn’t make it wrong. Scientists are detectives and they piece together the evidence in ways that seem to make sense of as many data as possible.

                      W&P are not concluding anything from the graph in and of itself – that’s why they discuss other evidence that makes that apparent correlation seem more plausible. It’s just not true to say it’s ‘shabby’ of them to present the regression lines, given that they acknowledge its limitations and provide further evidence to suggest it may well show something real.

                      Quite genuinely, this is how science progresses, for better or worse.

                    • tsmithfield …
                      25 September 2010 at 11:22 pm

                      The other thing is that I very much doubt that with only 8 data points they would be able to satisfy the underlying assumptions of normally distributed data that is required for reliable use of regression techniques.

                      I accept that this is not all they were relying on. However, I think I still have a valid point in that including the graph weakens their argument rather than strengthens it. And I have seen a number of other examples where they have done the same thing. So its not an isolated case. All it achieves is to attract criticism toward what might otherwise be a very good study. I actually like statistical analysis, and my eyes tend to roll back into my head when I see this sort of thing. Then I tend to feel quite skeptical about anything else they say.

                      What would be better would be time-series data from one country (say the US) correlated against changes in income equality over that time-frame.

  5. tsmithfield 5
    25 September 2010 at 7:23 pm

    Further to my comments above, the general criticism I have of “the spirit level” is as follows:

    Sometimes its what is left out that strengthens an argument. If I were doing this “spirit level” study and felt I had a really strong argument, I would cut out anything that was the slightest bit dodgy and just leave in the really strong stuff.

    In this case, I would have definitely excluded graph 2. The reason is that only having data for eight countries out of the 50 odd they decided to include in their study is simply too few to be convincing. We don’t know what the trend would have looked like had their been a heap more data points, so that graph isn’t at all convincing, without even thinking about whether their analysis method is appropriate or not.

    The way they have done it has left it open for people such as myself to find fault with what they have done. When I see what appears to be quite a naive approach to data analysis, it makes me feel quite dubious about the study as a whole. If they had focused on the strong stuff I would have been more impressed.

    • Puddleglum 5.1
      25 September 2010 at 7:41 pm

      See my comment just above. There’s a bit more known now.

    • mcflock 5.2
      26 September 2010 at 3:11 am

      Sometimes its what is left out that strengthens an argument. If I were doing this “spirit level” study and felt I had a really strong argument, I would cut out anything that was the slightest bit dodgy and just leave in the really strong stuff.

      Um – just because there are weaker points beside stronger arguments does not mean that the stronger arguments are invalid. Particularly if the topic is a larger picture that covers many different areas and aspects.

      I can’t figure out your purpose, TS. Are you suggesting that because one graph in the post might have issues, then the bigger claim that inequality is associated with a number of other negatives is therefore in doubt? Are you going to put similar effort into all the other graphs?

      Or do you agree with the gist of the Spirit Level as summarised in the post, but just like getting anal about statistics in lieu of an actual point?

  6. tsmithfield 6
    26 September 2010 at 10:15 am

    “Um – just because there are weaker points beside stronger arguments does not mean that the stronger arguments are invalid. Particularly if the topic is a larger picture that covers many different areas and aspects.”

    True enough. However, the weaker points can tend to obfuscate the stronger ones; or undermine the stronger points if fallacious ones are included. So why keep them in?

    “I can’t figure out your purpose, TS. Are you suggesting that because one graph in the post might have issues, then the bigger claim that inequality is associated with a number of other negatives is therefore in doubt? Are you going to put similar effort into all the other graphs?”

    Except its not just one graph. As I pointed out above, the authors are prone to doing this sort of thing. There are a number of issues I have about the way the authors appear to have done things.

    “Or do you agree with the gist of the Spirit Level as summarised in the post, but just like getting anal about statistics in lieu of an actual point?”

    To be fair, I haven’t actually read the book itself. Hence, it is unreasonable for me to be drawing firm conclusions about the book itself. That is why discussions with the likes of Puddlegum is quite interesting; in that he has read the book and appears to have a good knowledge of scientific method etc.

    I have read a lot of scientific reports in the past based on statistics, and they are very careful about methods they use, justification for those methods etc. Because I have some knowledge in statistics, I am also very aware of how statistics can be misused. So, although it may seem a bit “anal” I like to think I do have a point. 🙂

    • Redlogix 6.1
      26 September 2010 at 1:19 pm

      However, the weaker points can tend to obfuscate the stronger ones; or undermine the stronger points if fallacious ones are included. So why keep them in?

      ummm… because if they were left out folks like you would have accused them of ‘cherry picking’?

  7. Bored 7
    26 September 2010 at 11:31 am

    TS, to date you have written probably as much on this book as the original authors.As such you probably have a body of work that you might publish in the best spirit of free and open market competition with the original authors. Why dont you do this, really, no kidding. Then we can do two things: one see how many you sell by comparison which in itself might be instructive, and two have it put up on webs sites like this so that you can watch incessant critiques by bloggers of your nature. Have fun.

  8. tsmithfield 8
    26 September 2010 at 12:14 pm

    Bored, you have just committed a sort of backward logical error of appeal to authority. I would encourage you not to be blinded by ideology but to actually be able to step back and assess the quality of the work in question.

    Sure, I haven’t published papers or the like. However, I have done some papers on critical data analysis, analysis methods etc. Also, I have used reasonably complicated analysis techniques such as structural equation modelling in my thesis. So, I know enough to be able to criticise this sort of stuff. Notice above that even Puddlegum and the authors themselves concede that the graph I have been criticising in itself is very weak evidence for their argument. So, I am not just blowing hot air.

    If I was going to do a study similar to “the spirit level” I wouldn’t bother with all these regression charts. If you have ever used a technique called “multiple regression” you will realise why.

    The way I would have done the study would have been to start with a time-series study in one country, such as the US. I would then study the effect of changes in the data for income inequality with subsequent changes in social variables over a thirty year period or such. Doesn’t mean I would have to wait 30 years. Just that I would use historical data. My thesis would be that changes in income inequality would be associated with subsequent changes in social variables. I could then model for goodness of fit to confirm the direction of causation. Demonstrating that the theorised effect occurred after the theorised cause is a good way to rule out “correlation is not causation” arguments.

    If that study demonstrated a significant association, I would then test other countries to see if the effect was repeatable in other cultures. If it repeats reliably then the argument would be very strong.

    This would provide much more convincing results than what I have seen in the articles to date on “the spirit level”.

    • Bored 8.1
      26 September 2010 at 12:29 pm

      So you dont want to write a book BUT you are prepared to write the first chapter in reply to the idea……my goodness you could earn some cash if your prodigious output was saleable. And theres the rub.

    • RedLogix 8.2
      26 September 2010 at 2:20 pm

      The way I would have done the study would have been to start with a time-series study in one country, such as the US. I would then study the effect of changes in the data for income inequality with subsequent changes in social variables over a thirty year period or such. Doesn’t mean I would have to wait 30 years. Just that I would use historical data.

      The experiment had been done.

      According to Naomi Klein, in her chapter from “The Shock Doctrine” called “Bonfire of a Young Democracy” (pp. 275-309), in 1989 there were about 2 million Russians living in poverty. Under corporatism (which the press mislabeled “democracy”), by the mid-90s, 74 million Russians lived in poverty. Along the way, addiction, alcoholism, violent crime, and AIDS skyrocketted, and longevity plummetted. Bottom line, the suffering in Russia was horrific, but was passed off in the media as the “growing pains” of the transition to “democracy,” when, in fact, it was an orgy of greed and abuse.

      The result of Yeltsin throwing the nation’s assets into a vast fire sale was that around half of Russians finished up living in abject poverty, a much diminshed professional and middle class barely held their heads above water by working two or more jobs or cutting deals in various black markets…while a small elite prospered beyond all belief. The change in income inequality was extreme, and the consequences stark.

      No fancy statistics needed.

    • KJT 8.3
      26 September 2010 at 8:45 pm

      Start with New Zealand.
      Strong correlation between the rise in inequality and many indicators of social decline over the last 50 years, no matter what type of regression you use.
      The time based study that TS wants to do confirm the ideas in the Spirit level ,especially over the last 30 years in the USA.

      If you want an interesting comparison. Compare Illinois with North Dakota.

Links to post

  1. The Standard: posted on 25 September 2010 at 12:29 am

The server will be getting hardware changes this evening starting at 10pm NZDT.
The site will be off line for some hours.