Editore"s Note
Tilting at Windmills

Email Newsletter icon, E-mail Newsletter icon, Email List icon, E-mail List icon Sign up for Free News & Updates

September 4, 2007
By: Kevin Drum

EVERYTHING IN ITS SEASON....I've gotten a couple of emails asking me for more detail about the seasonality of violence in Iraq. There are several ways to show this graphically, but all of them present difficulties because the various data series aren't available in a consistent form for the entire course of the war.

That said, here's a chart that shows the seasonality pretty well. It comes via reader Thomas J., who has graphed U.S. troop deaths per thousand soldiers, a data series that's (a) available for every month since the beginning of the occupation, (b) highly consistent, and (c) a fairly decent proxy for the overall level of violence. The "per thousand soldiers" correction helps to control for the fact that troop fatalities will naturally be higher whenever there are more soldiers deployed in Iraq. This doesn't necessarily mean the level of violence is higher, just that the population at risk is larger.

The seasonality is pretty easy to see: violence peaks in spring, then declines during summer, peaks again in fall, and drops during winter. The peaks are probably overstated slightly thanks to unusually deadly April and November months in 2004, but even without that the seasonality is fairly pronounced. Roughly speaking, July troop casualties are typically about 40% lower than their April peak, whereas this year they were only about 30% lower. In other words, violence was worse than usual. On the other hand, August looked a little better than usual this year compared to 2007's spring peak. As always, what this means is that you can't draw any dramatic conclusions based on a couple of data points. However, taken as a whole the evidence pretty strongly suggests that the surge hasn't had any effect at all on overall violence levels. It's just moving in its usual seasonal pattern.

The obvious followup would be a similar chart showing the seasonality of civilian casualties. Unfortunately, there's simply no reliable data series for civilian casualties over the course of the war, and the data for this year in particular gives every indication of being massaged to within an inch of its life (intra-Shiite violence doesn't count, car bomb fatalities don't count, al-Qaeda attacks against Sunni tribes don't count, the figures change mysteriously from one report to the next, the supposedly lower numbers for August are classified, etc. etc.) We do have figures released by various Iraqi ministries, but we don't have a consistent series of ministry numbers for the past four years, and in any case the numbers for 2007 don't show any decline at all between spring and summer. So that doesn't suggest any surge-related decrease in violence either.

Bottom line: you should be skeptical of any claims about reductions in violence unless they take seasonality into account. So far, though, I haven't seen any credible claims of reduced violence that even mention seasonality, let alone adjust for it. That should tell you something.

Kevin Drum 9:12 PM Permalink | Trackbacks | Comments (44)

Bookmark and Share
 
Comments

Beware of progressives falling for the McNamara body count strategy in reverse.

Posted by: theAmericanist on September 4, 2007 at 9:20 PM | PERMALINK

One thing different now from previous months is the emphasis on getting troops off the megabases and out into small combat outposts, a cornerstone of COIN ops. This should expose more troops to enemy fire, so reasonably the incidents per thousand should go up.

Posted by: Swaggering Jingoistic RSM Goon on September 4, 2007 at 9:28 PM | PERMALINK

Hit return too soon. The other thought is that previously Anbar was a huge source of violence. If numbers are staying constant but Anbar is quiet (it is) then somewhere else the numbers are up.

Posted by: Swaggering Jingoistic RSM Goon on September 4, 2007 at 9:31 PM | PERMALINK

You can make all sorts of precise estimates of statistical parameters to prove one hypothesis or the other, but that is beside the point.

The decider has decided and the military leaders have given him the talking points to support his decision. Nothing that anyone can do short of a mass movement to stop this madness.

Something is wrong with our democracy or at least the media if the people can be lied to and spun so blatantly for a cause that most of the citizens do not support. Perhaps this is not the definition of authoritarianism, but it is definitely not the finest hour of American democracy.

Posted by: gregor on September 4, 2007 at 9:40 PM | PERMALINK

Where is our generation's Westmoreland?

Oh yeah -- he's VP.

Posted by: Gore/Edwards 08 on September 4, 2007 at 9:49 PM | PERMALINK

If numbers are staying constant but Anbar is quiet (it is) then somewhere else the numbers are up.

Al-Anbar residents killed 20 US troops in July. The total US fatalities in July were 79 according to icasualties.org, and some of those were presumably from accidents, etc. So al-Anbar, despite being reduced to the stone age, managed to kill a fourth or more of all US troops killed in combat in July. Al-Anbar is roughly 1/24 of Iraq by population. So it killed six times more US troops than we would have expected based on its proportion of the Iraqi population.

That's what the Bushies are celebrating, that the deadly al-Anbar has been wrestled down to only killing a fourth of the US troops killed in a month. It used to be more.

In mid-July, There were about 100 violent attacks in a single week in al-Anbar. That's a bright spot. That's progress. Since the year before, there were 400 violent attacks in that same period.

Well, yes, that's a relative improvement. But a hundred violent attacks in a week? That's being touted as good news to be ecstatic over? There were probably on the order of 1100 attacks that week in all of Iraq. So al-Anbar generated nearly one-tenth of all attacks. But it is only 1/24 of Iraq by population, so it is more than twice as dangerous with regard to the number of attacks than you would expect from its small population.

One of the ways "calm" has been produced in the city is to simply forbid vehicular traffic. Since May, if you wanted to get somewhere in Fallujah, you have had to walk. So when the National Review tells us things are suddenly miraculously "calm" in al-Anbar, this is being produced artificially. Things would be calm in most hot spots if you could ban all forms of locomotion save walking.

. . .August 28: A suicide bomber detonated a vest packed with explosives in a Sunni Arab mosque in Fallujah yesterday, killing 10 worshipers...

Now, if ten worshippers were killed in a church just last week in a small US city of 200,000, would Congressmen be flocking there to proclaim how wonderful the security situation was?

Just a month before, a bomber killed two policemen in Fallujah and wounded 11 others.

On July 23, a female suicide bomber killed 7 policemen at a checkpoint in downtown Ramadi.

On July 8, a truck bomb killed 23 persons at a police recruiting center in Haswa, al-Anbar province.

- Juan Cole

So, so quiet.

Posted by: trex on September 4, 2007 at 9:54 PM | PERMALINK

SJRSM's points and that this examines a small subset of the violence. Even Brit deaths (Brit deaths are way up per thousand this year) are not included, let alone Iraqi (surely it's looking at where the bulk of the violence occurs that's important). While I guess it's natural for you guys to focus on your own (and the data re. Iraqi death is sketchy), this only has meaning if you make the suposition that changes in this subset reflect changes in the set at large.

Oh, and what the Americanist said.

Posted by: snicker-snack on September 4, 2007 at 9:56 PM | PERMALINK

Something that needs to be taken into account: Is this seasonality related to US troop deployments in any way? I can see violence decreasing when more troops are in country or when they're more concentrated.

Posted by: Remus Shepherd on September 4, 2007 at 9:56 PM | PERMALINK

Thomas J, thanks for the data, but the seasonality of the violence is easy to demonstrate without graphing each year on top of one another. If you can follow t (time) in a linear fashion from March 2003 and have a skinny little column for each month (heck just paint it black) from that point onwards to the present time, you can see a general upward trend in fatalities per thousand with several Golden Gate suspension bridge-like undulations that would show the seasonality. I bet some wingers will look at the 2004 peaks and see them as outliers and try to deflate the seasonal argument, when the overall UPWARD trend is the REAL argument.

Posted by: Doc at the Radar Station on September 4, 2007 at 9:57 PM | PERMALINK

I really think the best way to display US military fatalities in Iraq is to simply display them as is, with no averaging. I'm ambivalent about dividing by the troop level; on the one hand, the more troops there are in Iraq, the more that can be killed in Iraq; on the other hand, a larger and more powerful force should discourage attacks somewhat. And more important than the number of troops is what those troops are doing, which is something we can't average out.

So, on balance, I think it's best to just plot the monthly figures as given by icasualties.org. I've been doing just that for quite a while now, plotting US and coalition fatalities from March 2003 to the current month. Whatever changes there are, seasonal or otherwise, show up in these plots. The most recent is here.

Posted by: Dr. Drang on September 4, 2007 at 10:10 PM | PERMALINK

So the bloodthirsty goons tell us everything is going according to plan. Just like they told us everything was going according to plan in 2003, 2004, 2005, and 2006. In longer than it took the United States to fight WWII the military has failed to provide a stable Iraq.

That's because a) this isn't a war, it's an occupation, b) the goal has never been a stable Iraq - there was already a stable Iraq before George W. Bush and the idiots who put killing people before the national security of the United States assaulted the Iraqi people.

All Hail the murderous King George and his malevolent enablers.

Posted by: heavy on September 4, 2007 at 10:13 PM | PERMALINK

So, if the previous seasonality trend continues, there'd be a sharp increase in deaths between July and August this year, right? Was there an increase?

If not, then perhaps averaging last month is with previous Augusts is masking what's really happening. The problem with any statistical analysis like this is that history can't predict when the trend breaks and a new one begins.

It will be interesting to see how many deaths there are in September (it's way to early to tell, but so far the number is down sharply). Give it two or three more months and you might have enough data to tell read the real trend since the Surge rather than what happened before it.

Posted by: Strick on September 4, 2007 at 10:21 PM | PERMALINK

kevin drum -

this is really nice work you have done, here and in the previous post.

work that (relatively) few others can, or would, do.

thanks.


in politics,

graphs really ARE worth a thousand words.

i have a feeling (ok - a hope) that these graphs are going to get a workout

Posted by: orionATL on September 4, 2007 at 10:25 PM | PERMALINK

Ramadan begins around September 13th, when in the past, attacks have tended to increase. Couple that with the upcoming Petraeus evaluation, and it would be a major surprise if there were NO uptick in violence in the next month or so.


[In the last twenty-four hours you have posted 26 times under no less than nine different handles.

elmendorf
rnc
monkeybone
harry (benson)
bart
ein
maryjane
Al "Trelane" Gore
Dingleberry

Choose a handle and cut out the trolling or be banned again -- mod.]

Posted by: harry on September 4, 2007 at 10:28 PM | PERMALINK

and thanks, too

for putting down a marker

that can, and should,

be used to challenge teevee swish-swish-spit about iraq.

Posted by: orionATL on September 4, 2007 at 10:29 PM | PERMALINK

Nice graph, colorful. Killing seems highly correlated with blue.

Posted by: Matt on September 4, 2007 at 10:38 PM | PERMALINK

Dr. Drang, thanks for providing that graph. Yep, if you took it and mirrored it in X I would say yeah, things are getting better...

Posted by: Doc at the Radar Station on September 4, 2007 at 10:40 PM | PERMALINK

I heard a commentator on NPR today make the best observation about the brouhaha over this American casualty count nonsense - "If you torture these statistics enough, you can make them say about anything".

These are not charts, graphs, lines or numbers, Kevin. These are dead human beings with families. Don't let these asshole conservatives distract you from that fact.

Posted by: The Conservative Deflator on September 4, 2007 at 11:01 PM | PERMALINK

Consider the headlines after General Petraeus informs the Congress that the American counterinsurgency effort in Iraq needs to go on for at least 10 more years in order to be successful. Thus American troops fighting and dying in Iraq at least until 2017.

Of course, this assumes that the Congress presses him on the point and that the media is attentive. His point, by the way.

Posted by: headliner on September 4, 2007 at 11:09 PM | PERMALINK

What happened to Year 3?

I'm not all sure that the trend is significant. It seems to be driven by a couple of spikes rather than a uniform trend. Maybe if you switch to the Lunar Calendar?

Posted by: ogmb on September 4, 2007 at 11:11 PM | PERMALINK

Kevin,

I love that you post your own Excel graphs, but you shouldn't use the default gray background. You'll note that this reader's graph looks a lot sharper because of its white background.

Posted by: Steve Sailer on September 4, 2007 at 11:49 PM | PERMALINK

Are the raw data behind this chart available? Having recently absorbed Tufte, I'd like to try my hand at producing a better figure.

Posted by: dob on September 4, 2007 at 11:54 PM | PERMALINK

I think 5 superimposed line graphs (one for each year) would actually be more informative and interpretable (you can easily see whether/to what extent they move up and down in sync). I can almost guarantee there's a statistical test of periodicity or some such out there that would also tell you if the apparent seasonality is significant, though I'm too lazy to hunt it and the dataset down to confirm this.

Posted by: gary1 on September 4, 2007 at 11:58 PM | PERMALINK

Looks like coyote/rabbit graphs.

http://www.tiem.utk.edu/~gross/bioed/bealsmodules/pred-prey.gph2.gif

Posted by: Luther on September 5, 2007 at 12:14 AM | PERMALINK

Perhaps it is more than significant that there is no mention of civilian casualties outside Kevin's post. Might have missed it as I skimmed through. Correct me if I'm wrong.

Of course there's nothing unfortunate about it. It is absolutely deliberate. From the first day the US has declined to offer security, legal process, or the barest acknowledgement to dying Iraqis. They have walked away frm all their obligations as an occuppying power.

It is an absolute disgrace and an abrogation of international law. Again!

Enough! String 'em up!

I don't mean the Iraqis!

Posted by: notthere on September 5, 2007 at 12:37 AM | PERMALINK

That's the worst graph I've ever seen! What's wrong with a continuous line to plot a single variable vs. another single variable? Instead, we have a histogram AND a line, and on top of that the graph has been folded over five times so the series wraps around on itself.

Somebody is in serious need of a copy of Tufte.

Posted by: Paul J. Camp on September 5, 2007 at 12:39 AM | PERMALINK

I'm not sure that deaths per 1000 soldiers is necessarily a better metric. After all, the purpose of adding more troops is to reduce the level of violence and increase their safety. If increasing troops results in more deaths, then obviously the surge strategy is wrong.

Posted by: Joe on September 5, 2007 at 12:40 AM | PERMALINK

Beware of progressives falling for the McNamara body count strategy in reverse.

Posted by: theAmericanist on September 4, 2007 at 9:20 PM | PERMALINK

I'm trying hard to understand this.

The US Army counted bodies as the pathway to victory. It proved not only distracting but wrong. As other armies have found in the past and as it is in Iraq.

But counting bodies as to how we are failing? How could that have the same effect? As an occupier, the greater the domestic kill how could we not deduce the greater the failure?

All yours, theAmercanist.

Oh, and Americanist: White Euro, Native American or American Indian, Latino, African, some mix?

I'm promoting USian, unless you want to apply a sub-culture, for citizens of the US to differentiate from all the others the USians don't seem to think exist in the American continents.

Time to end the colonial attitude.

Posted by: notthere on September 5, 2007 at 12:55 AM | PERMALINK

...If increasing troops results in more deaths, then obviously the surge strategy is wrong.

Posted by: Joe on September 5, 2007 at 12:40 AM | PERMALINK

Yeah. I forgot this thought. Tactics have varied and we reduced our interdiction to lower casualties. We now, supposedly (I wonder if it's still true after the first 4 months of the surge), put more targets out there. What's the cost/benefit from those increased deaths?

There should be an answer.

Of course correlated to the "increased" contribution of Iraqi police and security.

Right!

Posted by: notthere on September 5, 2007 at 1:09 AM | PERMALINK

Year five levels are higher than year four levels every single month so far. That's pretty easy to read. That means there has been no drop in violence since the surge.

Year five levels are also higher than in any year so far except in January, April and August. This is not complicated.

Posted by: expatjourno on September 5, 2007 at 6:38 AM | PERMALINK

Arguments against "the surge" completely miss the point (so do the arguments FOR it), Not.

Since you asked, this is why it's like McNamara's body count metrics.

Oversimplified, American strategy in Vietnam under McNamara and Westmoreland was simply to kill more of them than they could theoretically stand, cuz then they would give up, both within South Vietnam and as an invasion from the North. By that theory, the body count measured something useful -- the more we killed, the closer we were getting to that magic #: 'more than they could stand'.

But there was never any evidence that there WAS a # that was 'more than they could stand', so body counts weren't measuring anything useful. (My Lai was about a body count.) The theory was wrong.

Hint: we CHANGED our strategy.

It is a sign that progressives aren't real interested in the actual history of the Vietnam War, which is why (however ill-advisedly) Bush raised it: Petraeus doesn't want to be Westmoreland, he wants to be Abrams. Under THAT, the latter American strategy, we weren't focused on body counts, but on clearing and holding territory in order to make POLITICAL progress.

So even in theory, the #s of American troops killed, just like the #s of Iraqis killed, is at best an imperfect measure of what's going on.

The fact that Sunni tribes in Anbar who had been a major opponent of American goals (cuz they are against a Shi'ite dominated central government), are now a major ally for a different set of American goals (cuz they are against foreign fighters) shows that using the same factor to measure wildly different things is misconceived.

Put it this way: if Sunni and Shi'a militias succeed in utterly segregating all of Iraq into heavily armed camps by sect and tribe, the place might just become stable, in a scary 'nasty brutish and short' sorta way. But first it would be a damned bloody place, like the partition of India in a brutal, highly localized patchwork manner. That DOES work (in theory, and grimly, it could work in practice), and the casualty counts would be more like D-Day than 9-11.

Don't get me wrong: more American casualties are bad, fewer American casualities are good -- that's an absolute. But it's not a measure of SUCCESS -- we lost a couple thousand guys on D-Day, and that was a success.

We lost a couple thousand on 9-11, which was not.

Finally, cuz folks really are myopic: arguments over the surge miss the friggin' point.

By next spring or so, about a fifth of our troops will leave Iraq. There are no troops to replace 'em. So we WILL be making a substantial withdrawal. We're just looking for a rationalization.

It seems quite likely to be that along about then, there will ALSO be a new Iraqi ruling coalition -- dominated by Shi'ites, naturally, but perhaps with some other mix of Sunni and Kurd interests.

So I predict that this new coalition will pointedly ask Bush to remove, say, a fifth of US troops. And Bush will comply, with a stirring explanation of how his strategy worked, despite the skeptics: President Chanticleer.

The rest is just so much bullshit, folks.

Posted by: theAmericanist on September 5, 2007 at 8:06 AM | PERMALINK

By next spring or so, about a fifth of our troops will leave Iraq. There are no troops to replace 'em. So we WILL be making a substantial withdrawal. We're just looking for a rationalization.

The withdrawal of the Brits from Basra is illuminating. The Brits declare it is part of a strategic realignment. The Shia thugs (Mahdi, etc.) declare they were driven out. Perception is reality, so this coming Spring expect to see a great deal of posturing and maneuvering on all fronts as each tries to create perceptions in their favor.

Posted by: Swaggering Jingoistic RSM goon on September 5, 2007 at 9:17 AM | PERMALINK

Petraeus doesn't want to be Westmoreland, he wants to be Abrams

I don't want to be Thersites, I want to be Achilles. With divine ponies, and a chariot.

The point is, Patraeus is perhaps a smarter general than his predecessor. Maybe if Patraeus had been in charge from the start things would be going better now. But BushCo has screwed this up beyond repair.

The fact that Sunni tribes ... are now a major ally... shows that anyone will act friendly if you give him weapons.

Posted by: thersites on September 5, 2007 at 10:09 AM | PERMALINK

But I do agree with the Americanist that the emphasis on body counts is almost beside the point. And I agree with the person who commented last week that the emphasis on American casualties is somewhat narcissistic. Sure, it's important to count American losses, since that's what most Americans pay attention to. But it would be good to see similar attention paid to losses to other Coalition forces. Not to mention Iraqi deaths.
But the real point is this surge was sold as a way to buy time for the political process to work, and that's not going so well, is it?

Posted by: thersites on September 5, 2007 at 10:18 AM | PERMALINK

Perception is reality, so this coming Spring expect to see a great deal of posturing and maneuvering on all fronts as each tries to create perceptions in their favor.

Yes, Red State Mike, we're well aware of the Doschstosslegende you dishonest Bush Cultists -- but I repeat myself -- have been working on.

Posted by: Gregory on September 5, 2007 at 11:21 AM | PERMALINK

Thanks for the feedback, folks...

-The data is available from icasualties.org and brookings.edu/iraqindex.

-I used death rate because it is as even-handed as I could come up with.

-Correcting for seasons isn't hard, I just didn't do it (I banged out the graph between bouts of actual work this weekend). It is graphically evident, to me, though.

-Correcting for the change in tactics represented by the surge isn't really possible, but I would guess the relatively high death rates since the beginning of the surge would be expected when you abandon a force-protection posture.

-You can't say from this data if the surge is working. Just isn't possible. However, it is data! Yay.

-Guess I'll have to look at this Tufte fellow. Sorry if the format made things unclear.

Cheers

Posted by: ThomasJ on September 5, 2007 at 11:46 AM | PERMALINK

"the real point is this surge was sold as a way to buy time for the political process to work..."

Yeah, and the war was sold as a way to avoid a smoking gun in the shape of a mushroom cloud. Oy, let's pay ATTENTION, shall we?

Since nobody else will, let ME ask the really hard question: since we're going to take out a fifth of our troops early next year, is it better for America if that is described (with some at least plausible reasoning) as something like a victory, or as a defeat?

Note that I'm not asking about BUSH. In the end, I care less about that guy than I do about our country.

To put it starkly, there are more or less two ways it can go: the Aiken solution, in which we declare that the new coalition government in Iraq (with all those Sunni tribes shooting foreign fighters) clearly shows that our often-changed mission has finally succeeded, so we're getting out of Dodge.

OR we can draw down our forces while bitching that we could be doing something else (like what?), and generally act like what WE think matters the most.

It's the same set of facts: we pull out a fifth of our troops, with a new coalition government. If we call it victory, we are that much closer to getting the rest out.

If we call it defeat, we MIGHT get the rest out just as fast (though I doubt it). But it won't change the result, no matter what we call it.

So, which is better for AMERICA: victory or defeat?

Posted by: theAmericanist on September 5, 2007 at 12:02 PM | PERMALINK

Just curious--when do you think you banned me before?

Posted by: harry on September 5, 2007 at 12:47 PM | PERMALINK

Yes, Red State Mike, we're well aware of the Doschstosslegende you dishonest Bush Cultists -- but I repeat myself -- have been working on.
Posted by: Gregory

It's nothing that more Scott Beauchamps and 1960's memos typed in MS Word 2003 can't take care of.

Posted by: SJRSMG on September 5, 2007 at 12:57 PM | PERMALINK

The "unusually deadly April and November months in 2004" are due to the two assaults on Fallujah.

Posted by: croatoan on September 5, 2007 at 12:57 PM | PERMALINK

the Americanist: "... is it better for America if that (withdrawal of 1/5 of American troops) is described (with some at least plausible reasoning) as something like a victory, or as a defeat?"

Good luck coming up with plausible reasoning. The place is an utter cesspool, with cholera outbreaks occurring simultaneously in different parts of the country; there's no infrastructure to speak of; the government is at best dysfunctional, and at worst nonexistent; and we've ensured that whatever happens after we leave, it's going to be even bloodier than it would have been before we came up with the brilliant idea of arming Sunni factions.

Go ahead and put lipstick on that pig.

Posted by: junebug on September 5, 2007 at 2:20 PM | PERMALINK
...So, which is better for AMERICA: victory or defeat? theAmericanist at 12:02 PM
Losers can always CLAIM victory; FOOLS can always claim knowledge, but the words will not conceal the TRUTH. The US lost its WAR on Iraq the DAY it invaded. If PEOPLE accept the facts, they are less likely to CONDONE similar war mongering in the FUTURE, but from HISTORICAL experience, selling FEAR AND WAR generally works...for a LITTLE while. Posted by: Mike on September 5, 2007 at 2:39 PM | PERMALINK

Great post Kevin: you bring a tear to the eyes of statistical geeks everywhere.


Economists deal with seasonal problems all the time: headline figures for inflation, unemployment and GDP are invariably seasonally adjusted.

Surge discussions that don't mention seasonals are bunk. Just as budgetary discussions that don't adjust for inflation are made by those who don't know or those who don't care.

Every reporter should adopt a checklist to distinguish between serious and bogus analysis. Scientists should be asked about their control group. Surveys should be spot-checked for biased or nonrandom samples. And figures should be adjusted for seasonality and inflation where appropriate. It's not that hard.

Posted by: Measure for Measure on September 5, 2007 at 3:15 PM | PERMALINK

So, which is better for AMERICA: victory or defeat?

Well excuse my prejudices but as a non-Yank, I'm more interested in this question:

"So, which is better for THE WORLD: American victory or defeat?"

On balance, I'd have to say I'm coming down on the 'd' side of things. Out-of-control American empire is far more threatening than any rag-tag terrorists could ever be. You get your democracy back and your leaders frightened of the people my thinking could change.


Posted by: snicker-snack (the non-Americanist) on September 5, 2007 at 8:39 PM | PERMALINK




 

 

Read Jonathan Rowe remembrance and articles
Email Newsletter icon, E-mail Newsletter icon, Email List icon, E-mail List icon Sign up for Free News & Updates

Advertise in WM



buy from Amazon and
support the Monthly