The Economics of Roadside Bombs

[mhanson]

EDIT: Can't seem to get the first quote to work. Oh well.

You treat all direct, non-IED attacks as the same--and argue that IED attacks are an inferior good, since insurgents would, if they had the resources, prefer to engage in direct fire attacks if they could.

I probably didn't make this clear enough in the paper, but my only strong reason for believing that IED attacks are an inferior good is because the statistical results work out that way. I do talk about the reasons why that might be the case, but that is just speculation on my part.

What if direct attacks, however, are two separate goods: sniper-type attacks which are intended to inflict harm but not capture territory, and attacks intended to overrun/control territory? The former, it seems to me, are substitutable with IEDs by your logic (from the point of view of insurgent aims, if not technical capabilities). You don't have any option, given the way the data is coded, but to treat them this way--but a footnote raising the question might be appropriate.

You might well be right about that. In fact, I think a more detailed, classified study could really nail down a lot of issues like that and actually begin to test competing hypotheses -- presumably, if sniper attacks were not also inferior, that would disprove my speculative theory.

On a related note, I suspect that the "direct attack" numbers conflate cases where the insurgents were the initiator and those where they responded with direct fire to an immediate threat from coalition or Iraqi military activity.

Probably true, but those cases would also be interesting because they would be complex attacks. Complex attacks imply some complementarity between IED and non-IED attacks -- which would suggest that if they were treated seperately, the substitution effect I find would actually be stronger.

2) You've modeled the insurgents as the consumers, responding changes in the cost of goods and income (resource availability). Would you be led to similar or different conclusions if they were treated as competitive firms responding to high and price inelastic consumer demand?

The theoretical framework I use is utility (damage) maximization by a consumer -- that is, maximization of a nonlinear function subject to a linear resource constraint. However, every constrained maximization (primal) problem has a corresponding constrained minimization (dual) problem. The dual case corresponds to treating the insurgents as cost-minimizing firms. The conclusions are exactly the same (though, as a referee for the Quarterly Journal of Economics pointed out, this may not be an example of Giffen behavior -- in the latest version of the paper, I have taken out the Giffen hypothesis).

3) You assume that all IED have similar costs. It might be argued that the cost of PP and and CW triggered IEDs was "lower" than RC, PI and CP triggers (in terms of material cost and technical expertise). Conversely, it could be argued that the cost of LPRC and CW triggered IEDs was "higher" (given the greater risk involved in deploying the trigger team closer to IED site).

I assume that radio control triggers have higher profit than other types of triggers. Otherwise, I don't see a good explanation for why the insurgents shifted over to other triggers slowly -- with all the talk about how adaptable the insurgency is, it seems implausible to me that they would figure out how much more better non-R/C triggers were and then take 2-3 years to actually switching over to other types of triggers. I think that they don't use R/C much anymore precisely because it's not effective -- but that they would have preferred to keep using it if we had let them.

4) PI and PP systems have less tactical discrimination than do command-detonated systems, and hence their utility may change depending on the nature of the fighting (maintaining a defensive perimeter in a Fallujah-type battle, compared to harassment attacks on commonly used public thoroughfares). Can "blips" be detected in their use that corresponded to the nature of combat operations at a given time? Similarly, it is my completely uninformed understanding that PI triggers have been particularly associated with EFP use, especially by Shi'ite militias (I could well be wrong), in which case fluctuations in their use might relate to broader political issues.

With quantitative data about combat operations, perhaps. However, I don't know where this data might be found (even on the classified side). In my follow-on study, I intend to use panel data (data over time and location, as opposed to just the over time data I have now) so that I can control for regional and time effects.

5) How sure can you be that insurgents measure "effectiveness" in the same way your data does? One can imagine that an IED detonation which creates a sense of insurgent presence and threat, but which fails to kill or seriously wound, might be considered a "successful" outcome.

Certainly, but the only assumption I need for my model is that they prefer an IED that produces causalties to one that does not produce casualties. They can still derive some utility from an ineffective IED attack, just not as much.

6) You might explicitly identify other competing explanations as to why IED attacks have climbed and NIED have fluctuated--and suggest why your inferior/Giffen good argument gives a better explanation of the data. (One way of testing this would be to break the data down by region: your findings should be fairly consistent across locations, while arguments focused on say, insurgent politics or operational strategy would predict significant regional variation). At first, a stronger predictor or NIED attacks is the wave of sectarian violence that followed the February 2006 al-Askari mosque bombing.

Indeed. As I mentioned above, panel data is the way to test those hypotheses.

I realize not all the above answers are totally satisfying, but hopefully they're a start. Here's a question for you -- why might IED attacks appear inferior, if it's not a question of control of territory? I certainly believe there are reasonable alternative explanations, but I can't think of any.