Droning on about civilian casualty ratios

I have been busy with statistics on the UAV strikes in Pakistan and have come up with the following: between 2004 and 2011, the ballpark figure for the estimated civilian-combatant* casualty ratio is somewhere between 1:4 and 1:5. Torturing the New America Foundation’s figures to breaking point results in an outside ratio of about 1:3.

I am not a historian by trade, but I am struggling to come up with a war in which a “better” average occurred (from the civilian point of view) in the last hundred years. I can think of Israeli claims of a 1:28 ratio at points of its targeted killing campaign in Gaza, but that’s about it.

My question to KoW readers: do you know of one?

*Taken to be people estimated in datasets as militia/Taliban/AQ etc regardless of IHL targeting, and all ‘unknowns’ as civilians.

 

Standard

18 thoughts on “Droning on about civilian casualty ratios

  1. Deskibel says:

    Probably telling you what you already know by saying that figures are extremely inaccurate, particularly the further back you look (i.e. pre 1990).

    However these are some articles that I came across when researching New Wars that deal with the ratio of military-civilian casualties over time:
    http://www.tandfonline.com/doi/abs/10.1080/00396338.2010.494880
    http://www.eui.eu/Documents/DepartmentsCentres/SPS/Seminars/SeminarsF09/PVSEMF08/LacinaGleditschMonitoringTrendsInGlobalCombatEJP2005.pdf
    http://ejt.sagepub.com/content/15/3/505.short

    I’m afraid I don’t know of any specific examples, but the ratio that you come up with does seem very low when placed in a broader context.

    (I also suspect that you are focusing on targeting killings rather than war in general, although you do not state explicitly – anyway you might find them useful).

  2. Joe says:

    How are you counting it?I ask because I dont see how you can be accurate as to whether those killed were civilians or taliban/al qaeda. I would believe ISAF would say more often than not that those they kill in drone strikes are combatants, even if it were impossible for them to substantiate such a claim. But is there any real way to know?

    • It’s an analysis of published datasets all of which derive their data from media analysis. There’s a wide degree of variance between figures, but 1:4-1:5 appears to be the average ratio, and at the outer limits (eg manipulating the figures to produce the worst possible result in a way that would cause statisticians to have a heart attack) 1:3 can be produced. It’s also restricted to Pakistan, so ISAF don’t get a say.

      Either way, it has been pointed out to me separately that I have egregiously overlooked the Falklands, on a rather important anniversary, where 3 civilians died out of about 900 casualties.

    • Joe says:

      Its very interesting and surprising. I would have estimated a much higher proportion of civilian casualties, perhaps even an inversion of the ratios above. I guess that comes from the news…

  3. Francis Grice says:

    Well, I apologise for the Wikipedia cite, but this seems to have a fairly strong list of conflicts with civilian to military ratios, including the Drone Strikes. You may have already seen it though:

    http://en.wikipedia.org/wiki/Civilian_casualty_ratio

    (Interestingly, it too misses out the Falklands conflict. Curious that it is so easily overlooked.)

  4. Dear Jack,
    I empathise with your desire to nut down an accurate ratio, nevertheless I suspect two issues militate against any treatment of this result as anything other than propaganda. Firstly, even supposedly ‘objective’ journalists (such as myself) have to take at face value claims that the individuals targeted and killed were, in fact, ‘insurgents’ or whatever the designated term is. In other words, when people aren’t wearing uniforms or actually engaged in the business of fighting, how can you possibly trust the categories deaths are assigned to. This is a very different situation to the Falkland’s, for example. After spending time in both Afghanistan and the Pakistan Tribal Areas I’d hesitate to make any such judgements . . . loyalties switch too often. Secondly, it really doesn’t matter if the kill ratio is terrific – hit one wedding party and you’ve lost the propaganda war at a blow. Even if you do manage to eliminate some notoriously bad characters while you’re at it.

    • Hi Nic,

      My intent isn’t actually to nut down a specific figure, since that would clearly be impossible. The section is on the limits of the available evidence in relation to a number of critiques of the campaign. So, for example, claims that these operations were indiscriminate, or at least as indiscriminate as most conventional wars (not conducted on islands in the south Atlantic) would require all observers to be ‘off’ by a considerable margin.

      The importance (for me at least) is the difference between subjective and objective interpretations of battlefield deaths. You’re entirely right in saying that kill ratios don’t mean a jot in the course of a campaign, but that’s because our understanding of civilian casualties includes subjective judgements of strategic communications etc. Lawyers and international law affecionados still have an objective view of the battlefield, in which high rates of civilian casualties (a sign of indiscriminate means of war) constitutes a war crime.

    • Thanks, I see. Australian forces aren’t officially involved in any drone targeting in Pakistan, but I know they take very great care with the Special Operations Task Group (Special Forces, inc SAS) missions inside Afghanistan to make sure only known and identified ‘insurgents’ are targeted. Nevertheless, even here a couple of disastrous blunders (not initially admitted to by the military) have led directly to PR ‘disasters’ for the forces, as you suggest. The burden of accountability for mistakes has climbed significantly as the war has dragged on. Although (virtually) no one would suggest for a moment that war crimes were being deliberately committed, I suspect consequent political pressure on the government has led to severely circumscribing the activities of the contingent. This has led to a changing tolerance for killings over time amongst people located a long way away from the conflict.

  5. First off, with New America Foundation you’re interrogating a problematic dataset which publicly does not measure civilian casualties. NAF lump ‘civilian’ and non-defined elements into an ‘other’ category in its data.

    A recent article in CNN placed NAF’s civilian death count for 2011 at 7%. Yet that figure is problematic, given (a) they don’t measure (publicly at least) civilian deaths and (b) there’s no explanation of how they reach that finite figure, given they employ a casualty range.

    http://edition.cnn.com/2012/03/27/opinion/bergen-drone-decline/index.html

    The Bureau of Investigative Journalism in London runs its own dataset (tbij.com – I declare an interest) which overtly records all credible reports of civilian casualties – and has followed up extensively in the field. It indicates a civilian death toll for 2011 of 9-27%. The range is as wide as it is because of multiple variables.

    However, follow-up field studies by ourselves and Associated Press show a strong correlation between original media claims of civilian deaths (which are made far less frequently than one might think) and research findings.

    TBIJ’s 2004-2012 range of civilian deaths is 479-811, of a total casualty list of 2,425 – 3,093.

    You also call the Pakistan strikes a ‘war’. I’m not sure I’d agree. Attacks often occur after many hours (even days) of observation. Targets which might be acceptable in a ‘hot’ war are often rejected here for fear of collateral damage, we’re told.

    The CIA has carried out around 300 Pakistan strikes in 8 years. In comparison NATO carried out almost as many drone strikes in as many months in Libya last year. The NYT, in a follow-up field study, found that some 40-70 civilians had died in those attacks.

    http://www.nytimes.com/2011/12/18/world/africa/scores-of-unintended-casualties-in-nato-war-in-libya.html?_r=1&pagewanted=all

    • I’m comparing the NAF, TBIJ, LWJ, Pakistan Body Count and Jamestown datasets. I should probably add – to the end of 2011, not 2012, and the Jamestown dataset is only used to 2009, therefore as a comparative measure. The PBC set doesn’t discriminate between civilians and combatants, as you know, and can therefore be used as a comparison for scale only.

      As noted previously, my methodology isn’t to discern the ‘true’ value, rather to use the worst possible interpretations of observed data against critics and supporters of the programme alike. By your own estimation, at worst, and as noted, a roughly 1:3 civilian combatant kill-ratio is the outcome of this. Therefore, I’d be using these outer limit interpretations of data as a rule of thumb against those who consider the strikes to be proportionate, and the opposite limit to consider the criticism of those who consider the strikes indiscriminate. After all, if an argument stands to the worst possible interpretation of data collected, and then some, while not ‘certain’ by any means, it is more plausible than the answer of ‘we just don’t know’ cop out which supporters of targeted killings use to defend the programme.

      What I find odd is that you’d single out the NAF, since they produce (overall) a tally that is quite similar to TBIJ (differing by a couple of percentage points), whereas the Long War Journal’s figures for civilian casualty ratios (estimated) are significantly lower. The Jamestown methodology which was criticised for pro-American ‘bias’ is also remarkably similar, given the difficulty of obtaining data. The most significant difference between the datasets is the interpretation of the 2006 Bajar strike.

      As for the ‘war’ thing, that’s a whole other five thesis chapters, and would probably cloud an otherwise interesting debate on stats. Thanks for the heads up on the NYT article, I missed that first time round.

    • I singled out NAF as it’s the only source you cited in your original post. And also because it’s a problematic one, in my view, when plotting civilian casualties. Their numbers may be similar to ours – but I’m not sure you could replicate their conclusions from available data.

      Pakistan Institute for Peace Studies and Conflict Monitoring Centre (reportedly not a million miles from the Pakistan military’s views) are also worth a look for their datasets.

      I take your point about ‘is it war?’ – a bit like Paxman droning on about ‘Is it art?’ Though it may be, given the specifics of the platform, that the only valid comparison is with other combat UAV campaigns (UK, Italy, US in Afghanistan; US in Iraq; Israel in Gaza; NATO in Libya).

  6. carl says:

    A lot hinges on whether you are talking about an entire war effort or part of the war effort. Some of the island battles in the Pacific had few if any civ deaths. Naval battles between naval forces had none or almost none whereas submarine campaigns had many or not so many depending on how you define merchant sailors manning a ship carrying 40 Sherman tanks. And I suppose the more remote the battlefield the fewer civilians available. The war in the mountains between China and India in the early 1960s was pretty far from anything so maybe not so many civs got killed.

  7. I reckon this thread is fascinating exactly because rather than contributing to your original request we’ve all been keen to debate the validity of the strategy. I suspect this really means that the key variable is not (simply) the number of civilians killed as a ratio to combatants, but rather the utility of the bombs to bring the war to a successful conclusion. In other words, as the campaign has drawn on without delivering a ‘knock-out blow’ to match the hype of the idea of painlessly eliminating enemy commanders, Western civilians have become progressively less willing to accept collateral damage.

    • The interesting bit for me is that the targeted killings debate revolves around two key incomplete areas of information: the numbers/nature of those killed, and the targeting process itself. Both sides of the debate interpret these ‘unknowns’ in different ways.

      On one side, critics state that there’s no way to know how many AQ/militants are being killed, but we do know that many civilians are being killed, and we cannot trust in the targeting (since it is secret/CIA). Therefore the campaign appears disproportionate and it is up to the govt to prove otherwise (which they won’t).

      On the other side, supporters state that we do know the campaign is killing key AQ members, and that we should trust in the targeting because the military are discriminate elsewhere – it is up to the critics to prove the programme disproportionate (which they can’t).

      The problem is that each approach comes at it with such disparate world views that it is difficult for them to even argue. One quite good example is the framing of the strike that killed Baitullah Mehsud, which killed him, his wife, his parents-in-law as well as seven ‘bodyguards’ and a ‘lieutenant’. Mary O’Connell framed this as ‘the strike killed twelve for one intended target’ (and considers it illegal anyway). Those who consider such strikes legitimate might point out that nine of those killed could be considered legitimate targets, since it is difficult to think of any method of killing a man with eight persons who would (rationally speaking from their designation) fight to protect him. In O’Connell’s view, it’s a ratio of 11 unintended targets to 1 intended target (indiscriminate), to a supporter it’s 9 legitimate targets to 3 double effect casualties (proportionate).

      As for knock out blows, my PhD is about the use of targeted killings for containment at the fringes of the state system, which throws up a whole different bunch of legal/ethical issues. Personally, I don’t think they’ll be ending AQ any time soon.

    • Gunrunner says:

      “Hi Jack,” (good thing I didn’t say that in an airport. . .)

      “On one side, critics state that there’s no way to know how many AQ/militants are being killed, but we do know that many civilians are being killed,”

      That’s the point—the critics are wrong. . .we don’t know anything of the kind.

      Are civilians inadvertently killed in battle? Yes, but they are not the object/aim of the attack, whereas, we DO KNOW that AQ/militants use civilians as shields and also target them for attack.

      http://www.dailymail.co.uk/news/article-1177516/Taliban-fighters-use-civilians-human-shields-battle-Pakistan-tears-Swat-Valley-apart.html
      http://news.bbc.co.uk/2/hi/8519507.stm
      A UN survey (Afghanistan 2010 Mid-Year Report on Protection of Civilians in Armed Conflict) released in August confirmed that the Taliban use civilians as shields in continuing their attacks on Afghan and international forces.

      “and we cannot trust in the targeting (since it is secret/CIA).”

      Critics say we can’t trust targeting because it is done in “secret”? They want real-world targeting processes and intelligence to be reveled just so it can be trusted by them. . .really?

      (By the way, I am sure everyone is aware targeting is an integrated process that spans many agencies and military departments, not just the bug-a-boo CIA).

      Targeting staffs do their job, from the basic analyst to the strategic planner. They all pool their information, synthesize it and make their best estimates. They are men trained in the art and science of targeting and schooled in LOAC. Targeteers operate under layers of legal oversight as the target list develops.

      The actual targeting process (not intelligence) is not “secret”:
      http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA360613, http://www.bits.de/NRANEU/others/jp-doctrine/jp3_60(07).pdf, https://www.hsdl.org/?view&did=3301, http://www.dtic.mil/doctrine/jel/service_pubs/afdd2_1_9.pdf, http://www.e-publishing.af.mil/shared/media/epubs/AFPAM14-210.pdf, http://www.wpafb.af.mil/shared/media/document/AFD-110516-047.pdf,. . .just to list a few that popped up in the net.

      The targeting process is under constant review and development. Just last year an RFI was issued seeking bidders to see if they could offer improvements: https://www.fbo.gov/index?s=opportunity&mode=form&id=e7ca17a024afdde24a3c241eaab0a988&tab=core&_cview=0 . Next time an RFI (or RFP)v is posted, those critics that want transparency should submit their own proposal on how they would conduct targeting. Heck, the government would pay them for it. . .Bonus!

      Of course, if those critics want true transparency, based upon level of clearance and need-to-know, they could apply for a job in targeting (www.simplyhired.com/a/jobs/list/q-joint+targeting+course).

      Thing is, targeting is not done willy-nilly or by just a few guys hanging around throwing darts at a map. It is complex, constantly changing and evolving, with several hundreds of highly intelligent people evaluating raw data and intelligence products, and thoughtfully creating COAs. (And yes, I have worked some in a targeting cell).

  8. friederbürkle says:

    No, referring to UAV strikes in Pakistan I don’t know of a better ratio, BUT some sources suggest that at the same time (well, somewhat lagged t+1) the number of insurgent’s strikes and suicide attacks in Afghanistan and Pakistan and the number of civilians killed went up (quantity & quality).

    Prof. Dr. Sebastian Harnisch from the University of Heidelberg, Germany, suggested this correlation might occur as an effect of norm erosion and disinhibition in asymetrical warfare due to civilian casualties in UAV-attacks, and the delegitimisation of the Afghan and Pakistani governments through covert US-military operations on their ground, which might also increase support for insurgents (got no citation, he talked at a summer school of the Atlantic Academy Rhineland-Pfalz). Intuitively, it makes sense that invisible attacks by air, carried out outside the law (although somewhat unclear what law or norms) by a foreign power, don’t find favour within the population, negative press and the Pakistani government denial of their approval might have contributed.

    The mentioned effects might even thwart the COIN- or Counterterrorism-objectives of UAV-attacks. So checking the civilians-militants ratio of isolated UAV attacks might be useful to assess their effectiveness in relation to other COIN, counterterrorism or targeted killing strategies such as sniper assaults, but it might underestimate overall and unintended effects of UAV warfare on civilian deaths. While other variables migth influence the number and quality of terrorist attacks and insurgency warfare, and causal mechanism remain to be proven, one should at least regard the assessments of UAV-effectiveness with some caution.

    Also, it seems to me that no immediate independent source or method is available to ascertain the credibility of newspaper reports and databases regarding militants-civilians ratios.

    Best,
    Frieder Bürkle

  9. Rob Dover says:

    Posted on behalf of my friend ‘Gunrunner’ who couldn’t get the site to accept the comment! So it’s his comment, not mine.

    ““Hi Jack,” (good thing I didn’t say that in an airport. . .)

    “On one side, critics state that there’s no way to know how many AQ/militants are being killed, but we do know that many civilians are being killed,”

    That’s the point—the critics are wrong. . .we don’t know anything of the kind.

    Are civilians inadvertently killed in battle? Yes, but they are not the object/aim of the attack, whereas, we DO KNOW that AQ/militants use civilians as shields and also target them for attack.

    http://www.dailymail.co.uk/news/article-1177516/Taliban-fighters-use-civilians-human-shields-battle-Pakistan-tears-Swat-Valley-apart.html
    http://news.bbc.co.uk/2/hi/8519507.stm
    A UN survey (Afghanistan 2010 Mid-Year Report on Protection of Civilians in Armed Conflict) released in August confirmed that the Taliban use civilians as shields in continuing their attacks on Afghan and international forces.

    “and we cannot trust in the targeting (since it is secret/CIA).”

    Critics say we can’t trust targeting because it is done in “secret”? They want real-world targeting processes and intelligence to be reveled just so it can be trusted by them. . .really?

    (By the way, I am sure everyone is aware targeting is an integrated process that spans many agencies and military departments, not just the bug-a-boo CIA).

    Targeting staffs do their job, from the basic analyst to the strategic planner. They all pool their information, synthesize it and make their best estimates. They are men trained in the art and science of targeting and schooled in LOAC. Targeteers operate under layers of legal oversight as the target list develops.

    The actual targeting process (not intelligence) is not “secret”:
    http://www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA360613, http://www.bits.de/NRANEU/others/jp-doctrine/jp3_60(07).pdf, https://www.hsdl.org/?view&did=3301, http://www.dtic.mil/doctrine/jel/service_pubs/afdd2_1_9.pdf, http://www.e-publishing.af.mil/shared/media/epubs/AFPAM14-210.pdf, http://www.wpafb.af.mil/shared/media/document/AFD-110516-047.pdf,. . .just to list a few that popped up in the net.

    The targeting process is under constant review and development. Just last year an RFI was issued seeking bidders to see if they could offer improvements: https://www.fbo.gov/index?s=opportunity&mode=form&id=e7ca17a024afdde24a3c241eaab0a988&tab=core&_cview=0 . Next time an RFI (or RFP)v is posted, those critics that want transparency should submit their own proposal on how they would conduct targeting. Heck, the government would pay them for it. . .Bonus!

    Of course, if those critics want true transparency, based upon level of clearance and need-to-know, they could apply for a job in targeting (www.simplyhired.com/a/jobs/list/q-joint+targeting+course).

    Thing is, targeting is not done willy-nilly or by just a few guys hanging around throwing darts at a map. It is complex, constantly changing and evolving, with several hundreds of highly intelligent people evaluating raw data and intelligence products, and thoughtfully creating COAs. (And yes, I have worked some in a targeting cell).”

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>