Friday, June 25, 2004

Finding the Right Troop Levels

Sorry for the dearth of posts. I haven't been lacking in ideas or motivation, only in time, as I've been writing a paper for a conference (basically done) and I had a big meeting Wednesday for my research project. But while I was thinking of work-related material yesterday I saw this Slate article entitled "The Pentagon's Fuzzy Math."

Note: I'm an engineer coming from the standpoint that if you can't quantify it, it doesn't exist. Fortunately you can quantify almost anything. The following is mildly technical, but the concepts should be understandable for smart people of any background. [Note to the Note: I may not be the best judge of who will understand this, so if you don't, it doesn't mean you're an idiot.]

The problem of determining optimal troop levels is obviously difficult due to all the uncertainties involved in what wars/peacekeeping/nation-building/insurgent-crushing/war-mongering/et-cetera will happen in the future. But there are techniques for managing uncertainty and risk in these situations that aren't that complicated. How so?

First, assume that the Pentagon has some kind of model that can determine the troop levels required for a wide-range of scenarios. If they don't, heads should roll and some quick assumptions can be made to create a rough one based on past experience and war-gaming results.

Then the intelligence people should develop a list of possible scenarios and provide a range of values capturing the probability of that scenario happening over some specified period of time. For example, suppose they say there is a 0-40% chance of war with Iran over the next five years, that would be one input.

The last step is to integrate these into a combined model that will randomly determine, based off the distributions provided by the intel folks, what scenarios will occur and calculate the required troop levels for that. This model should be run many, many times (tens of thousands, the process is called Monte Carlo simulation) and the results can be analyzed to show what the chance of any troop level being insufficient will be (based on the assumptions made by the troop-required model and the intel scenarios). This gives some idea of the risk associated with any given troop level. See, not so hard. It can be made much more complicated, but this kind of technique is not that difficult to apply, and I'd be surprised if somebody in the Pentagon hasn't done it. I would be tempted to make my own BS assumptions and troop-model for illustrative purposes but since I can't post pictures here I wouldn't be able to share it (oh yeah, and I'm lazy).

The problem is that people aren't interested in getting a real answer; this being politics they are mainly concerned with scoring points at Bush's expense and/or burnishing their tough-on-defense credentials, not that I'm cynical or anything (*cough* *cough*).

No comments: