December 05, 2006

Effect of troop numbers on Iraq, update

The recent UN report on Iraq violence levels gave the effect of two more months of data on the effect of U.S. force levels on civilian casualty levels, so I thought I'd update the charts from this post.

The story so far:
*there is no correlation between U.S.-led coalition force levels and indicators of violence against coalition forces.
*there was an apparent correlation between the UN figures for violence against civilians and overall troop levels, and also between the rate of change in both numbers, but only when using a statistical measure that only went back to January, 2006 (8 data points total).

Adding two more data points sort of throws that second conclusion off, a little, as one can see by the revised graphs below the fold.

Here is the Coalition force vs. civilian violence graph, with two new points provided by the September and October fatality figures:

Coalition Force levels (x) vs UNAMI Civilian Fatalities per month (y), Jan-Oct 2006

What the increasingly flat line here indicates is that the previously assessed correlation between force levels and violence against civilians had no apparent real-world effect in September and October, two months where U.S. troop levels went back up to their historical average but violence remained high (those would be the two points in the top right). If the trendline displayed were perfectly flat, that would mean that there was no effect of U.S. troop levels on violence. Also, the low R-squared value here indicates that we're basically looking at a random distribution at this point.

The comparison of the effect a change in military force size had on any change in violence level by a month later still shows a correlation, albeit a somewhat weaker one than before:

Changes in Coalition Force levels (x) vs changes in UNAMI Civilian Fatality tallies in the same and subsequent months (y), Jan-Aug 2006:

This suggests that troop increases do have an immediate calming effect, and troop decreases vice versa, but the previous graph indicates this can hardly be a permanent one.

Finally, here's the update to the graph used before blending the UNAMI civilian death estimates with the previous Brookings Institution ones, and compared to force levels. It continues to show a fairly linear progression in the level of Iraqi violence, with no lasting effect resulting from the various troop surges over the length of the occupation:

Seasonally Adjusted Iraq Civilian Fatality Variations from Mean (two methods combined), compared with troop level variations

This again indicates that Iraq civilian violence levels are relatively unswayed by changes in U.S. force levels. The two remain, in this comparison, independent variables, with any connection being a fairly transitory one. Again, this pattern might not hold if there were to be a much larger sudden increase in troop levels (50,000-plus), but smaller (10-25,000 size) troop strength increases and decreases appear to have had little lasting impact on the trend up until now.

(The idea that Iraq civilian casualtiy levels have been following an essentially unswayed linear progression would also tend to contradict, it should be noted, a hypothesis proposed here recently, that the rapid drop in U.S. force strength in January/06 had a catalytic effect on Iraqi-on-Iraqi violence in the early part of this year. If rapid troop increases can be demonstrated to have no lasting effect, neither would rapid troop decreases. However, the Brookings Institution numbers used in this graph for 2003-2005 are not so trustworthy that this can be assumed as a point of fact, either.)

Posted by BruceR at 12:05 PM