Via Trombly x Exum, an interesting paper on mobile telephony and the Iraq war is here. I was impressed by the fact the authors know there are multiple antenna sectors per cell-site, and that they bothered to find out roughly what an emerging market GSM operator’s roll-out process is like. In fact, if you wonder, the paper is actually quite a good high-level brief on how you go about picking the sites for a mobile phone network and some of the operational considerations, in Iraq or anywhere.
But that’s not the point. The authors set out to answer whether GSM helped the insurgents or the counterinsurgents in Iraq. They reasoned that it might help the insurgents by letting them co-ordinate their activities and by letting them set off bombs remotely, and the counterinsurgents mostly by making it easier for civilians to inform on the insurgents. They also took note that it could contribute to economic development, which is presumed to help the government side. (They don’t discuss the impact of either counterinsurgent ELINT spying on GSM traffic, or insurgent spying, probably as it’s very hard to get any information.)
They pulled a hell of a lot of data, helpfully provided by Zain (i.e. ex-Celtel), and concluded after a creditable effort to understand it that there was a small, but formally statistically significant, effect in favour of the counterinsurgent. Importantly, this was strongest in the areas that were classified as mixed Shia/Sunni in 2003, i.e. the battlefield, and during early 2007, i.e. the decisive peak of the battle.
That said, the R^2s aren’t huge (highest is 0.28). There’s also a problem in that they took quite a lot of care to identify antenna sectors that got service for the first time, in order to compare them with ones that already had it and ones that never did. This is scientific, but it does pose the problem that Zain weren’t putting so many BTSs out there in the spring of 2007 in the worst of the civil war, because their people didn’t want to die, and therefore you wonder about the sample size and the representativeness of some data points that are quite crucial.
Actually, you learn a lot here about the history of Zain Iraq’s network; in the months before the fateful mosque attacks of 2006, they were building out like maniacs, true to their reputation back then as Mo Ibrahim’s pioneers covering the most troubled hellholes of Africa. Then, it got bad enough and suddenly that they shut down most investment and stayed that way until the end of 2007. The authors say that “month-to-month variance in violence” didn’t lead to “major design change”, but the point wouldn’t be design change but timing, and they also include a chart that suggests they basically stopped in the worst of the war.
Growth isn’t necessarily stabilising, and I suspect that Zain engineers would tell you that straight off. In a very low-trust society, more resources are quite likely to lead to more conflict. I think it’s a safe assumption that GSM coverage helps people to rat to the government more than it helps them to rat to insurgents, because insurgents need to be in the network of strong personal ties to survive. I’m not completely convinced, though. There were examples of insurgent intelligence exploitation of GSM back then (see the 2005 and 2006 archives) and the weak/strong tie distinction is subjective.
Also, I would like to know what role anti-social networking played; to what extent did better communications, that were mobile, two-way, and also personal, help to spread propaganda, hate-speech, and paranoia? Radio Mille Collines was a broadcast system, but there’s no obvious reason why incitement to genocide can’t be participatory via SMS or (better) voice.
That said, the most responsive measurement of violence was the percentage of IEDs found and cleared, which suggests something was going on.
But one of the most interesting things in the paper is this:
Most importantly, the teams would typically enter into long-term contracts with local community members and organizations to pay for site rental, generator fueling, site security, and training of local engineers to provide these services. Where possible, they would engage with local elites to identify the personnel who could be entrusted with these jobs.
This strategy of establishing close connection to local elites meant that once marketing had identified an area for network expansion, teams were mostly able to move in effectively even in areas with high violence.
So perhaps we’re seeing the emergence of people who can navigate the emerging low-trust society. On the one hand, the interhierarchical leaders of instant tradition, who would later be the key interlocutors with the Americans in the counterinsurgent phase, on the other, the pre-merger Celtel team who were already familiar with working in this kind of environment.