Aid Thoughts

Digesting the difficult decisions of development
📚 This is an archive of Aid Thoughts, a development economics blog that was active from 2009 to 2017. Posts and comments are preserved in their original form.

The limits of handwaving and fancy graphs

By Matt Collin · May 23, 2012

Hans Rosling is quickly becoming famous for wowing audiences with fun, dynamic graphics of demographic changes. In the video below (hat tip to A View From The Cave), the audience applauds more watching total fertility decline in colour than it does when he tries to explain why it is happening.

http://www.youtube.com/watch?v=ezVk1ahRF78

I really like Rosling's approach to making complicated data more interesting to the general public and his belief that secular trends can be incredibly revealing. I'm less happy when he draws broad conclusions from what really just looks like extended eyeballing of his charts over time. In the talk from the video, Rosling charts countries by their majority religion (as far as I can see, ignoring within-country variation) and shows that reduction in fertility rates is pretty uniform across countries. He concludes that what matters for fertility isn't religion but instead:

  • Child mortality
  • Child labour
  • Women's education and labour force participation
  • The acceptability of family planning
I don't doubt that, if you had a regression with total fertility on the left side and you controlled for these four things, religion wouldn't have as large an "impact" on total fertility. However, I think the most pertinent question is: how much does religion matter for these four things? Rosling's conclusion is that we need to attach these four things directly - a standard public health approach - but if religion (or, perhaps, religiosity, something left out of Rosling's analysis entirely) is a determinant of these things, the standard approach might be less successful than we would hope.

Categories: Development
Tags: Han Rosling fertility

2 Comments

Tom · May 23, 2012 at 04:24 PM

Glad you went a bit further with this, Matt. I alluded to it a bit with the girl effect reference, but it feels a bit like Rosling is confirming what we already know. It is great that the data he finds reaffirms the four conclusions and he presents it in a way that can reach more people. The questions that follow, which you highlight well, are how to achieve those ends within the question of the religion.

What struck me above all else is that he did not want to deal with the messy parts of religion that are not always held within state bounds. The constraints of available data probably make it hard to look at religion itself as opposed to a country level. However, I wonder to what extent the trends may change if we could examine it on a more precise level. I think that would help to answer the excellent question you pose.

Robert Swope · May 24, 2012 at 08:12 PM

Good post. Analysis and visualization of data are two areas of research I'm doing a lot of work on right now. You bring up a good point which I mention in the following post:

http://www.robertswope.com/home/2012/5/24/issues-with-presenting-and-understanding-data.html

Best, RJS