Although I rarely seem to get around to posting on this blog (I probably should call it: "not a blog" in the vein of George RR Martin), and no one reads it, I still think it’s valuable to post information here when I think it is relevant. This way if people stumble upon it in a Google search they will, perhaps, have found something of interest.
In any case, the post today has to do with US military deaths in Iraq. I wrote a post back in 2005 where I performed an analysis that showed that military deaths were on the rise. Now that almost two years have gone by since I performed that analysis I would like to ask the question: has the death rate of US soldiers continued to rise?
I was inspired to redo this analysis because the US military stated this week that the death toll for July was the lowest in eight months and then spun it to say that the security conditions are getting better. While the factual statement is true, when put into context, this spin is at best meaningless and at worst misleading.
So what is the real picture of US deaths? To answer this question I used data from Iraq Coalition Casualty Count. I also use Iraq war “periods” to illustrate how death rates have changed based on certain well acknowledged conditions on the ground in Iraq. These periods are: 1) the invasion, March and April of 2003; 2) the beginning of the occupation and the establishment of the Coalition Authority, May 2003 to June 2004; 3) the transfer of sovereignty to the Iraqi Transitional Government, July 2004 to January 2005; 4) the election of the first Iraqi government, February 2005 to December 2005; 5) the second Parliamentary election and the ratification of the Iraqi Constitution, January 2006 to present. In the graph I plotted the average deaths per month throughout the entire period for the war, and performed several trend analyses.
Trends by period (thin solid lines): these show that with the exception of the invasion each period has gotten deadlier for US troops from the beginning of the period to the end. Also transitions between periods usually coincide with the drops in death rate, especially when the transitions involved elections. A possible explanation for these drops is the heavy security and use of strict curfews surrounding the elections. Of course, there may be other explanations for this phenomenon, but it is nevertheless thought provoking.
General trends (linear: heavy dot-dash, three factor polynomial: heavy dot): both general trends are shown for the occupation only and do not include the initial invasion. Both the linear and the three factor polynomial trends show an increasing rate of fatalities throughout the occupation period. The linear regression (R2 = 0.23, P = 0.00039; translation: the trend explains 23% of the variation in death rate and has a 99.9% probability of being a true description of the data) shows a change from 1.55 deaths per day at the beginning of the occupation to 3.02 deaths per day at the present time. I think this is an important point, because the media has focused on the fact the mean for July 2007 is consistent with averages for the years 2004 and 2005. As you can see there is a lot of variation in the death rate from month to month. A low death rate in a single month may be due to any number of factors that are not indicative of changes in the overall conditions of the war. Clearly the overall trend for the war is an increasing rate of US military deaths. The three factor polynomial is not significantly better in explaining the overall trend than the linear trend, but it does give a fairly good general description of changes in the overall death rate over time. The death rate of US soldiers rose sharply from the beginning of the occupation to June 2004 (mean 1.69/day) at which time it leveled off and remained fairly constant until August 2006 (mean 2.25/day). Since August 2006 death rate has again risen sharply (mean 3.12/day). These trends clearly show that the war has gotten increasingly deadly for US soldiers.
July trends (thin dot-dot-dash): one of the most interesting parts of this graph is that July 2007 has actually been the deadliest July on record for US troops. This may be coincidental, in that other Julys just happened to have lower death rates. It may also be caused by external factors that modify the death rate; July is the hottest month in Iraq, perhaps it’s too hot to fight? But I actually prefer a statistical explanation. Conditions on the ground (i.e. strength of insurgency, vulnerability of US troops, etc) may exist that cause an overall death rate. However, even with an overall rate there will still be an amount of variation from day to day and month to month. Conditions on the ground will also affect the variation in the death rate, just as they affect the mean. Interestingly, while the conditions that cause the mean death rate may be murky, conditions that cause variation in the death rate may be easier to spot. For example, we would expect death rates to vary substantially from the background rate when major battles occur. And indeed this occurs. The two high points for US military deaths during the occupation (the highest points in periods 2 and 3) correspond with the first and second battles of Fallujah (I should note that if these points are removed from the dataset the R2 for the linear trend rises to 0.42, in other words it explains 42% of the variation). These battles were the largest and bloodiest of the entire occupation. Since no other battles of this scale have occurred, we can make the reasonable expectation that in all other cases the variation in the death rate has remained fairly constant over time. If this is true then if we pick any two months that occurred during times with different mean death rates, we would in all likelihood find that the month with higher death rate also corresponded to the general time period with the higher death rate. In fact July 2007 has a higher death rate than 60% of all other months during the occupation, showing that it was a very deadly month indeed.
Conclusion: this analysis demonstrates 1) the strong overall trend has been for a steadily increasing death rate from the beginning of the occupation to the present, 2) the death rate for July is not out of character with this trend, and is therefore likely to not be indicative of a change in conditions on the ground. From this I can say that media hype and Pentagon optimism for July’s “lower” death rate are unfounded, and in fact may be deliberately misleading. To end, I sincerely hope that the July death rate is indicative of actual conditional changes in Iraq, and that the death rate for US soldier continues to drop. However, we cannot know whether this is true for several months or years. Until then I remain highly skeptical.
