So in my quest to understand the gender dimensions of water supply this week, I stumbled upon a nice paper  by Florencia Devoto and coauthors. They look at the effects of providing piped water in Tangiers, Morocco. The immediately cool thing about this paper is that they got something quite hard – randomization in an infrastructure project.
How, you ask? Well they used my favorite method of random encouragement . And for them it actually worked. Before getting to the story of their results, it’s worth discussing this. What happened was that the utility they were working with was offering potential new clients an interest free loan for the price of connection but without any kind of public information campaign. So information may have been part of the story as to why initial take up was low. But then there is the application for the loan, which they characterize as involving “quite formidable administrative barriers.” Now, I haven’t worked in Morocco, but I did get my own drivers license in Ghana and I have switched cable companies in suburban Maryland. Both of these experiences made me think hard about the value of time. Devoto and co. deal with these hurdles with a “nudge:”
We simplified these procedures radically for households in the treatment group, by obtaining pre-approval from the authorities in bulk, making digital copies of their identification document at home (with cameras), and bringing a branch officer to their home to collect the payment.
Take-up skyrockets – they get 69 percent take-up in the treatment group versus 10 percent in the control. Evaluation lesson 1: Forget just leading the horse to water. Open up his mouth and insert the hose. Policy lesson 1: Effective service delivery: it’s about making things easier for people. House calls help.
So what happens when these folks get water? Let’s start with the counterfactual, because it tells us something about how we should think about external validity here. Most of these folks (60%) were using a public standpipe, with the same quality of water, for free. The rest were getting it from their neighbors. And these folks didn’t seem to be using containers that were contaminated (like >50% of the containers tested in a recent study  in Ethiopia). So the main gain here was quantity – and somewhat less distance to go to get it.
Given that the gain here is in quantity and convenience, it’s not surprising that there are no health effects of this intervention (despite what looks like a careful diarrhea and vomiting diary – a sample of which is in the paper). Water fetching time drops from 82 minutes over three days to zero. People seem to spend this additional time on leisure. There is no effect on labor force participation but, while it’s clear they collected detailed time use on some things (leisure, water collection), it’s not clear that they collected it on hours worked for pay/in a business. Also, the treatment group has less conflict with their neighbors over water. And, of course, the one result my kids would hate: the treatment group takes more baths. These connections aren’t cheap, which gives us policy lesson 2: people are willing to borrow to finance these not-necessarily monetary improvements in welfare.
One of the really interesting results in the paper, which was designed before hand and nicely uses the random encouragement and some GIS data, consists of looking a bit further down the road (18 months instead of the 6 for the effects above) at the diffusion of information about the value of the connection. They find that having a household with a connection within 20 meters (these are small plots) increases the probability of having bought your own connection by 18 percentage points. So policy lesson 3: you don’t have to force all of the horses to drink. And evaluation lesson 2: if random encouragement works really well, beware the contamination (and use it to say something interesting).