19 Apr 2011

Getting water to the poor

This evening in Berlin, I will debate Wenonah Hauter of Food and Water Watch on whether full cost pricing is the best way to ensure good water service to the poor.

I represent the pro-side of the debate. Ms. Hauter represents the con-side that favors (presumably) government subsidies to water-service.

One of my points is that incompetent and corrupt governments cannot get water to poor people. I claimed this in my paper on human rights and water, but I looked for some data and got this:

This figure shows Aggregate governance indicators (from the World Bank) on the x-axis and Access to an improved water source (from the UN) on the y-axis for 170 countries, as of 2008. Here's my XLS.

You can see that the R^2 (goodness of fit) is 0.43. The correlation between the two variables is 0.66.

Note that this is an indirect argument, since it's quite possible for locally-managed, independent water agencies (private, public or in-between) to provide good service without suffering from the generalized corruption of the country. It's the independence that matters, as it frees the agency from distant complications so that it can concentrate on sustainable operations.

Bottom Line: Don't depend on incompetent governments for your water.


  1. You might want to delete the trendline. It isn't a good fit, the data don't look linear, the regression coefficient is lousy, and the correlation isn't all that strong either. Putting a bad trendline onto a data set like this weakens the argument, IMHO.

    Your bottom line is accurately worded. Some poor governments apparently do a good job of water delivery - there are some 100s in the left half of the chart - but many more don't.

    Might be a better way to display this data than an X-Y scatterplot, since it really doesn't look like the two variables are mathematically related. Is there maybe a way to calculate the % chance that an individual will have access to clean water, as a function of governance? You'd need population data for each country in the data set.

  2. @Tim -- "weakens the argument"... but that's what the data say, and I put those numbers there because they are the obvious ones people want to see.

    Your suggestion for a regression (chance of having water dependent on many variables) is more than I wanted to do on short notice. I've done something similar for my paper on water and human rights (http://ssrn.com/abstract=1549570), and got this: "A simple OLS regression of 2006 share of population with access to water on base year share of population with access, 2006 GDP(PPP)/c, base year GDP(PPP)/c, and a dummy for rights shows a statisticallyigni cant coeffcient for base year access but not the other variables (R2 = 0:85; Prob > F = 0:000)."


Read this first!

Make sure you copy your comment before submitting because sometimes the system will malfunction and you will lose your comment.

Spam will be deleted.

Comments on older posts must be approved (do not submit twice).

If you're having problems posting, email your comment to me