Last week’s Economist talks about the virtues of using randomized trials to answer macroeconomic policy questions. Randomized trials are similar to drug trials, where you select a random group of people to test an idea along with a control group on which you test something else. They talk about an experiment done in western Kenya, in the distribution of bednets going towards malaria eradication:
…researchers looked at what happened in 20 antenatal clinics in western Kenya when some gave away insecticide-treated bednets, an anti-malaria therapy, and others sold them for different prices. Their conclusion was that free distribution is far more effective in getting people to use bednets than charging even a nominal sum would be. [link]
That argument for charging people for this is simply that if people pay for something they will value it more. Of course, nothing is as it seems, and it turns out that in the part of the country that they were given away people already owned millions of bednets, so the value was already known. Furthermore, most of the recipients were pregnant women – which is good – but if you want to eradicate malaria then you want them to be universally distributed. So we are back to Square One. The experiment worked in this part of the country, but is it worthwhile to distribute bednets across the nation? Unknown.
This is of course a problem of random selection. Listening to Terry Gross’ Fresh Air on NPR the other day, I heard an interview with Elizabeth Pisani, an epidemiologist who is battling AIDS across the world. She was telling the story of a survey done by her group in Southeast Asia, to find out the average number of clients that prostitutes got per week. They were surprised by how low the numbers were – something like three per day (if I remember correctly).
Speaking later with another prostitute, she found out that the women she interviewed were so-called “dogs”. The very fact that they were available to be interviewed by her team showed that they were not in high demand. This skewed the numbers downwards. When they randomized it a little more, by asking different women at different times and so forth, they got much higher numbers. Here’s the show, if you want to listen to more.
Perhaps if there is enough research done into a particular culture/country/region to ensure that the trials will be truly randomized then one might be able to claim that macroeconomists can also run experiments, just like microeconomists. Then again, who was it that said that all macro is micro, anyway?