Last night, as I am occasionally wont to do, I went down to a certain bar downtown where I ran into an English guy I know named Simon. Very nice fellow, well-educated and a decent sort. Also a lefty, as almost all the expats in Barcelona are. He's quite indignant, and with some justification, about the situation in Iraq.

Anyway, he asked me, "How many Iraqis have been killed in the war?" I thought for a second and said "I guess one to five thousand", figuring he meant enemy combatants. He said no, he meant civilians too, and I thought for another second and then said, "OK, five to ten thousand, one's way too low." Simon vehemently disagreed and claimed a much higher figure, but we agreed like gentlemen to actually go look it up and get as close as possible to the truth, and I promised him I'd post what I found here.

He told me to go to a website called "Iraq Body Count", so I did. According to them, there have been a minimum of 9137 and a maximum of 10,994 "Iraqi civilians reported killed" as of today, which actually isn't too far from my high-end estimate.

Red Flag Number One: This is a partisan site. It is not neutral. It has political objectives in mind.

One problem with these high numbers is that they are quite openly "reported": that is, based on adding up all the numbers reported in the various media of communication. So, if some villager tells a BBC reporter that seven people were killed in his village, it goes into the count apparently without being checked.

On Iraq Body Count's front page, they say that 692 people with "names and surnames", as they say here, are confirmed dead. My guess is that this more conservative number of identifiable people killed is more likely to be accurate than their higher estmated numbers based on adding up numbers from the media.

From that front page, you can click on "Database" where they give you a list of the various incidents reported that they've used for their "maximum" and "minimum" body counts.

The nine most recent cases are (the "k" numbers are the code numbers assigned to each case, then come the date, time, place, target, method, maximum dead, minimum dead, and the sources):

k135, 29 Apr 2004, 9:50 AM, Baquba, US convoy, bomb, 1, 1, AP 29 Apr AFP 29 Apr

Looks like an enemy attack on the Americans in which a civilian was killed in the explosion.

k119, 25 Apr 2004, Mosul, rockets, 4, 4, AFP 25 Apr, AP 26 Apr

Looks like an enemy terrorist attack on civilians.

k120, 24 Apr 2004, Haswa, near Iskandariya, bus carrying 21, roadside bomb, 13, 14, Tel 25 Apr, AFP 26 Apr, AP 25 Apr

Looks like an enemy terrorist attack on civilians.

k121, 24 Apr 2004, PM, Sadr City, Baghdad, possibly offices of Badr group, rocket or mortar, 1, 1, AP 24 Apr, REU 24 Apr

Looks like internal faction fighting.

k133, 24 Apr 2004, AM, market, Sadr City, Baghdad, rockets, 13, 14, WP 25 Apr, AFP 24 Apr, NYT 24 Apr

Looks like an enemy terrorist attack on civilians.

k136, 21 Apr 2004, 7:15 AM and 8:15 AM, Basra and Zubair, police stations and police academy, car bombs, 74, 74, AP 23 Apr, AFP 23 Apr

Looks like an enemy terrorist attack.

k132, 20 Apr 2004, Abu Ghraib prison, Baghdad, (Baghdad Correction Facility), mortars, 22, 22, AP 01 May, AFP 21 Apr, REU 20 Apr

Looks like an enemy terrorist attack.

k131, 19 Apr 2004, Near Samarra, al-Iraqiya employees, gunfire, 2, 2, AP 19 Apr, REU 19 Apr

Doesn't say who killed them. I imagine enemy terrorists.

k130, 18 Apr 2004, Suss, near Kirkuk, shepherdess, daughter of tribal chief, gunfire, 1, 1, SMH 19 Apr, AFP 19 Apr

I bet the US Army didn't have anything to do with this one.

You see where they're getting these numbers from? They're counting anyone some journo says (agreed, they have at least a double source on each one) was killed, and they're including all civilians killed in the fighting, no matter who killed them. What it looks to me like is that 95% of the civilian victims are being caused by the enemy. If their maximum estimated reported number of 10,994 dead is true, then I figure the Americans probably killed about five hundred of them. The rest of the estimated reported deaths are due to enemy action.

Here is a Sydney Morning Herald summary of the charges made by the atrocity mongers. Most respectable sources are not quoting Iraq Body Count's numbers; among those who do are the Guardian.

Here's a Christian Science Monitor piece from eleven months ago with estimates of civilian deaths during the military portion of the war.

This is from Editor and Publisher on the AP body count from the military portion of the war, which is quite high. Either the AP body count of 3240 civilians killed in the Baghdad fighting or the Iraq Body Count numbers are used by almost everybody who is writing on this subject. (If you Google it, you'll see that almost every story on "civilian deaths in Iraq" is a leftist activist source. Neither source, the AP nor Iraq Body Count, says that these were people killed by Americans or Coalition forces; both counts apparently include everyone killed as a result of fighting, whether killed by the Coalition or by the enemy. This Ed and Pub story includes the damning expose that one of the people who did the body counting in Iraq for the AP was one of the atrocity mongerers who was responsible for the No Gun Ri hoax.

Regarding atrocity stories, I came across this guy named Matthew White's webpage. White seems reliable to me. He's an American lefty but by no means a crackpot, and he includes this as part of his discussion on the history of atrocity. Then click on "Historical Atlas of the 20th century". Scroll down to the bottom, under "Sources", and click on either "Detailed death tolls of the major bloodlettings of the 20th century" or "Index of Wars and Tyrants". Browse around for a while. It's interesting.

I'm stealing a large part of Matthew's writing, the Notes from his introduction. Here it is (it requires some clicking and scrolling to get here on his site, so I'm reproducing it all. Should Matthew object, he just has to say so and I'll take it down.) By the way, he mentions Rummel, the guy who wrote that book called Democide on the subject of mass killings by governments and political movement).

Notes:
[n.1]

"... numbers matter ... correct numbers."

This sentence is fraught with complications.

Firstly, the numbers only matter in a sociological, scientific sense; they certainly don't matter in any meaningful moral sense. For example, the American Revolution killed anywhere from 40,000 to 100,000 people, which is many, many orders of magnitude higher than the number of people that were dying under the British tyranny the colonials were so upset about. Was it worth 50,000 lives to create an independent United States rather than to peacefully evolve into a bigger Canada? The answer to that question, of course, has to be decided on the basis of intangible principles, rather than a simple mathematical formula of comparative body counts.

Secondly, as to the concept of "correct numbers"... where to start?

Although we all know that a butcher is a butcher whether he murders a thousand or a million, as a practical matter we are often forced to chose the lesser of two evils -- Hitler vs. Stalin, Mao vs. Chiang, Castro vs. Batista, Sandanista vs. Contra. We can argue the intangibles all day long and still not decide, so sooner or later someone is going to get the bright idea that numbers are objective, so let's just compare body counts.

Simple, scientific.

The problem is that the numbers aren't objective. As long as the moral meaning of an event is in dispute, the numbers will be in dispute. Until we agree on the interpretation of the event, we won't agree on the death toll.

For example, it was quite easy for me to find the number of soldiers killed in the First World War. The first encyclopedia I opened had all the casualty statistics right there in the W's. So did the second one I checked -- the exact same numbers. The first history of World War One I checked also had the same numbers, as did the next four sources I checked.

Why the unanimity? Probably because everyone agrees on the moral significance of the First World War -- it was a colossal, bloody blunder. Because the accepted death toll confirms that interpretation, no one has ever felt the need to go back and recalculate. On the other hand, if someday our interpretation of the war's significance changes (let's say, to "a glorious crusade against evil"), then a new generation of historians might feel that the old numbers are getting in the way of the new interpretation, and they'll take a second look.

And when they take that second look, they'll find that the statistics are a lot messier than the agreed numbers imply. This was, after all, the war that created the tomb of the unknown soldier. People were simply blown into oblivion. Hell, entire nations were blown into oblivion -- Austria-Hungary, the Ottoman Empire -- who could keep keep track of all this mayhem? There are huge gaps in the data that have to be filled by guesswork, and that guesswork is tilted by the historian's preconceptions.

Similarly, the death toll for the Congo Crisis of the 1960s is remarkably similar in most of the sources I've checked -- 100,000 -- a suspiciously round number. It's as if somebody somewhere took a wild guess at the order of magnitude, and since this is the only number available, everyone else just accepts it. Since there is, as yet, no vast body of American scholarship on the Congo, there's no dissenting opinion. So here again we see that everyone agrees on the body count because they all agree on moral significance. In this case, however, the moral interpretation of the event is "who cares?".

Contrast this with the death toll attributed to the Castro regime in Cuba. It runs from 2,000 to 97,000. Why? Because we can't agree whether Castro is an excessively severe reformer or a psychopathic tyrant. A researcher who is predisposed to being extremely anti-Communist is going to look under every rock for hidden horrors, and interpret every statistical inconsistency as a hint of some dark evil. Faced with the need to fill in gaps in the data with guesses, he will always assume the worst. Meanwhile, the less anti-Communist (no one admits to being pro-Communist nowadays) will set a higher burden of proof -- perhaps stubbornly insisting that every accusation be proven beyond a reasonable doubt, even though historians routinely make judgements based on evidence that would get tossed out at a jury trial.

Ironically, these disputes sometimes spill over and infect the estimates of unrelated atrocities. The death toll of the Duvalier regime in Haiti runs from 2,000 to 60,000, and I suspect that the number you pick depends less on your opinion of Duvalier himself (everyone agrees he was a brutal kleptocrat) and more on whether you want to label Duvalier or Castro as the bloodiest thug of the 20th Century Caribbean.

Take a look at three major histories of the Spanish Civil War and try to find which side was responsible for more political executions: Gabriel Jackson said it was the Right Wing with 200,000 killings, compared to 20,000 by the Left. Hugh Thomas agreed that it was the Right Wing, but his ratio was more balanced, 75,000 to 55,000. Stanley Payne put the heavier guilt on the Leftists: 72,000, compared to 35,000 killed by the Right. Which side should the world have supported? Which side was the lesser of two evils? Beats the heck out of me, but whichever side you prefer, I've just given you the numbers to back it up.

I sometimes wonder if the only solution to this endless bickering is either to admit that all death tolls are subjective, or else to decide that morality is not mathematical so it really doesn't matter who killed more than whom.

Each of these solutions, however, creates uncomfortable philosophical implications. The first implies that death tolls exist merely as quantum probabilities that only collapse into certainties when we agree. This means that if we, as a society, decide that a certain horror never happened, then it really, absolutely never happened. Taken a few steps further, this implies that the past has no independent, absolute existence beyond our memories and interpretations of it, and that it's all myth.

I suspect that most of us would lean towards the second solution. After all, very few of us would have a problem consigning both Adolf Hitler (15 million murders) and Idi Amin (300 thousand murders) to the same circle of Hell despite the 50:1 ratio in their death tolls. But if we're willing to ignore a 50:1 ratio to make Hitler and Amin moral equals, then we can just as easily find a moral equivalence between 300,000 deaths and 6,000. Pretty soon, we've removed the shear scale of the crimes from consideration, and because every ruler, no matter how benign, is probably responsible for at least one unjust or unnecessary death, we're claiming a moral equivalence between, say, Winston Churchill and Adolf Hitler (which -- and do I really need to say this? -- there isn't). Not only does this foul Churchill with Hitler crimes, but it also whitewashes Hitler with Churchill's virtues. After all, if two people begin as moral equals, then it doesn't take much to tilt the balance and make one of them (either of them) morally superior. Maybe even Hitler.

So this footnote has come full circle, and we still have no answer.

[n.2]

"A useful rule of thumb ..."

Mathematically, I'm talking about the median, the number that is lower than half the others, and higher than the other half. I find this to be a more useful average than the mean (the per-unit average, the sum of all the numbers divided by the count), which can be dragged off-center by one eccentric entry. If the spread runs 1,2,2,2,18, then the median is a nice reasonable 2, while the mean is 5, which is far higher than most of our numbers. Even worse than the mean is the range. By saying that our numbers range from 1 to 18 (strictly true), the impression is that the true average falls midway, at 9.5. Thus, by using the range, we are focusing on the two most eccentric numbers (1 and 18), instead of focusing on the central, most typical number (2).

A few other rules of thumb (and really boring rules of thumb at that, so you might want to escape now while you can) would be ...

You're free to ignore any one estimate on each list, no questions asked. If I could only find one source, then maybe no one else is able to corroborate the body count, so you can legitimately ignore it and leave a big question mark beside the atrocity. If I could only find two estimates, then you can pick whichever one you want. On the other hand, if ignoring one estimate still leaves a half dozen others, then you're just being mule-headed if you refuse to believe the general order of magnitude.
Watch for sleight of hand, and don't be afraid to ask, "Didn't we count that already?" If different writers describe a death toll as "100,000 people starved", "100,000 war dead", or "100,000 children died", don't automatically add them all together. Although strictly speaking, these are all different categories, the various writers might be talking about the same 100,000 labeled differently. We can't tell from these descriptions how distinct each count is or how much overlap exists between them. It might have started with an estimate that "100,000 people, mostly children, died in the war, often from malnutrition," and subsequent writers interpreted and rewrote that estimate with slight, but significant, differences. Similarly, "50,000 prisoners executed" may or may not be included among the "200,000 deaths in forced labor camps".
Don't be afraid to ask, "If this [regime, dictator, massacre, whatever] was so bad, why has no one else mentioned it?"
Writers usually focus on the biggest, most impressive totals they can get their hands on, so when one says, for example, "5,000 prisoners were executed in the first year of the new regime", he is probably calling attention to the first year because he considers this to be the peak. If another historian says that "45,000 were executed in the first five years", you can't just reconcile them by saying, "OK. 5,000 were killed in the first year, and 10,000 per year after that," because, after all, why would the first writer focus on the first year alone if the killing actually intensified? Sometimes different authorities are just irreconcilable.

[n.4]

"... the best thing about Rummel ..."

The unbest thing about Rummel's numbers is that they fit his theories just a little too neatly, so you might want to approach with caution. Here are a few dangers to be aware of:

He generally goes high on the numbers killed by Totalitarian regimes. If the range of estimates for the number of deaths under a communist like Stalin run from 15 to 60 million, Rummel will usually pick a number near the top. Thus, his estimate for the total number of unnatural deaths under Communism even exceeds the number set forth in The Black Book of Communism.
At the same time, he often goes low on the numbers killed by Authoritarian regimes. For instance, his estimate for the number of democides in the Congo Free State is the lowest of eight authorities I consulted.
During eras of widespread civil war, Rummel sees a proliferation of local governments rather than an absence of central government. By calling every bandit hideout a quasi-government, he can fit killings by Chinese warlords, Lebanese militias, lynch mobs, paramilitary death squads and corporate security forces into the death-by-government pigeonhole, rather than tallying these as examples of death by the lack of government. Therefore, "Government" gets blamed coming and going.
Some of his conclusions seem rather tautological. For example, his assertion that citizens of democracies are far less likely to die at the hands of their own governments is not surprising when we remember that not killing huge numbers of your own people is already included in the definition of democracy.
Based on Rummel's calculations, it has become customary on the Internet to accuse Government of 170 million murders during the 20th Century. The small print, however, is still important:
Of Rummel's 169 million democides, 118 million (or 70%) were victims of just three regimes -- the USSR, Communist China and Nazi Germany. That means that if the world were a single village of 1000 people, we would be basing complex socio-political theories of governing on the behavior of just three guys, the last of whom died a quarter century ago.
The margin of error for these three regimes can dramatically alter the total, and more importantly, it can alter the sociological conclusions we draw from it. For instance, I estimate that these 3 nations committed 45 million murders, which by itself would reduce Rummel's total by 73M. With Rummel's original total, democide is far and away the leading cause of preventable death in the modern world. My numbers would put it at about the same level as smoking.
In table 16A.1 of Statistics of Democide, Rummel lists 218 pretty nasty regimes, but only 142 of these were sovereign states, and the median number of democides committed by these regimes is 33,000. Sure, that's a lot. It's more people than I've killed; it's almost 3 dozen Titanics, but even so, it means that the average member of this 20th Century rogue's gallery killed about the same number of people as a couple of years of drunk driving in America (32,000 alcohol-related fatalities in 1999-2000).
Rummel accuses quasi-governments of some 6,681,000 democides, which may not seem like a big slice of the overall 170M, but it actually indicates that lack-of-government might be more dangerous than government. The 24 quasi-governments on Rummel's list racked up a median death toll of 100,000, which means that, on average, quasi-governments are three times bloodier than governments.
And most importantly: Governments don't kill people; people kill people.

I think all of Matthew's points are well worth taking into consideration.

Finally, here's Josh Chafetz's criticism of Iraq Body Count's methodology from the Weekly Standard.