Thursday, August 11, 2011

A Methodological Suggestion for Econometricians Working in Africa

When I was in Bukavu this summer, I tried to figure out how to measure the secondary economic effects of the de facto mineral embargo. It was clear how the embargo was affecting the miners and others directly involved in the mineral trade. A team of economists could presumably fan out throughout the province and measure those effects down to the decimal point. But how establish the impact of the embargo on the broader economy of the region--especially given the paucity (and likely inaccuracy) of the relevant data?

We have all sorts of ways to measure how well the economy is doing in the United States. There's the inflation rate, of course, and the unemployment rate. But there's also the monthly advanced retail sales report, the manufacturer's shipments and orders report, and the personal income and spending report, to name a few. Each of these require sophisticated data gathering techniques that have been refined over decades--the sort of intellectual infrastructure that obviously doesn't exist in a place like Bukavu. The question is whether there is any shortcut, any (relatively) accessible data that might function as a quick and dirty indicator of how well the economy is doing.

One thought occurred to me: Cell phone minutes. Cell phones are pretty ubiquitous, at least in town. Most people don't have monthly plans. Instead, they buy cell phone minutes in increments of one to five dollars at a time. When people aren't doing so well, they purchase fewer of those minutes. When they're flush, they purchase more. As a proxy for economic trends, then, minutes have several advantages. They aren't a requirement of life, like rent or food. But neither are they a lagging indicator, the result of pre-existing contracts and commitments, in the way that labor costs might be. Instead, they reflect how well people feel they are doing at the very moment the minutes are purchased. They are, to use an economic term I probably have no business using, highly elastic.

The other advantage of using minutes as a proxy is that the data are relatively easy to collect. There are only four cell phone networks in active use in South Kivu. If you could convince them to give you the data each month, you'd be all set. To be sure, you'd have methodological complications: if a network sustained a long outage, for example, or was running an introductory special, that might affect overall minute use. But these don't strike me as fatal complications. (The comparisons would be good from month to month, but probably wouldn't be useful over the long term: It would be hard to compare last month's usage to, say, the number of minutes used in July 2004.)

For all I know this idea is already old hat. Or perhaps it has been tried and found wanting. I'm not an econometrician and I don't play one on TV. But a quick search of the internet and a few of the relevant scholarly journals produced nothing on point. There's a lot on the dispersion and adoption of cell phones, but nothing on using minutes as a proxy for short-term economic trends.

So for all you econometricians out there, here's the deal: if you try this and it works, please send me five dollars--or fifty minutes of air time on Vodacom, because I'm running low!


  1. Fascinating idea David!

    You might need to look directly at how much airtime is being used (which could be difficult to get from telecoms), rather than how much airtime people have on hand. The idea that people with more money have more airtime on hand assumes that the airtime purchasing behavior is motivated by convenience - if you have more money surely you'd buy bigger amounts and less frequently so that you don't need to go through the hassle of buying more all the time.

    This assumption might not hold in Malawi. In many communities, doing anything with your phone is status increasing behavior. I may have plenty of money, but if I buy airtime more frequently in smaller increments, there are more opportunities for people to see me purchasing airtime and loading it onto my phone. Purchasing airtime isn't a hassle, it's fun! The assumption could prove statistically probably, but it would need to be validated (potentially by specific country or social group) because there is a social logic for the opposite to be true.

    Alternatively you could look at number of people who have zero airtime versus those who have any amount. It's rougher and would be hard to indicate anything about quality of life (several studies have shown people buying airtime even when they can't afford food for the entire family), but you could tell whether things are trending up or down overtime. Thanks for the post.

  2. I agree--we'd be looking at overall airtime used. And yes--it probably wouldn't be easy to get that data from the telephone networks. I guess I was thinking that the World Bank or some such might have enough pull to do that.
    I was fascinated to hear about minute-buying behavior in Malawi. I wonder if something similar goes on elsewhere? I smell a research project for some enterprising masters candidate!