May 16, 2005

Which Numbers Matter? The NCTC Report Revisited, Recoded and Reanalyzed

Marc Danziger (by way of reader T.M. Lutas) recently emailed an Excel file containing the narratives from the report released by the National Counterterrorism Center entitled A Chronology of Significant International Terrorism for 2004, describing the 651 terrorist attacks it identified.  (Although there has been a debate with Larry Johnson over the significance of these numbers I've no argument with the idea that we ought to be concerned.  After all, I'm concerned or I wouldn't be posting this re-analysis. However, I object to the notion that these numbers are some sort of signal that we're not winning the war.)  At any rate, I've been mulling over these data since Marc dropped them to me about two weeks ago, doing a lot of what data geeks do with this sort of delicacy, and thought I'd post a little of what I've found. 

First of all, Marc's post on Winds of Change describes his preliminary findings, according to which it's clear that two countries, India and Iraq, account for over 3/4ths of the terrorist attacks that took place in 2004. (If all attacks listed in the report are included these two countries account for 76.2%).  After looking more carefully at the narratives in the glossy report the first thing that one finds is that there are about 70 events on the list of terrorist attacks that may not belong.  That's because the target of those attacks was clearly military, rather than civilian.  These amount to about 11% of the total number of attacks listed, and more than 80% of these (57 of 70) occurred in Kashmir.

Now, it's true that the folks who did these dirty deeds probably were not very concerned about "collateral damage" to civilians, so I don't have reservations about including them, provided some civilians were either wounded or killed, but in 36 of the 70 cases where the objective was clearly military there were no civilian casualties at all.  I'm not sure why these events even appear in a terrorist attack database, but their inclusion tends to put more emphasis on terrorist activity in the troubled Indian state than might otherwise be the case.  This is not so much because the events were included for Kashmir, but more because similar events were not generally included elsewhere.  But whatever the justification, as one deletes some of these questionable "terrorist" attacks from consideration the percentage of attacks in Iraq rises from roughly 31% to 34% while those in India/Kashmir fall from about 45% down to 41%.  So, while the proportion of "terrorist" attacks rises for Iraq and falls for India after you do a bit of circumspective culling, those two nations still account for approximately 3/4ths of all terrorist attacks in the world.  That fact doesn't change.  And in sheer frequency of attacks, India/Kashmir is still more active than Iraq, or anywhere else.  But that's not the whole story, by a long shot.

Dimensions of Evil

Another dimension that hasn't been touched upon in the blogosphere or elsewhere in any rigorous way is the lethality of the attacks, by country.  And here Iraq is clearly in the lead.  Moreover, India and Iraq account for "only" slightly over half the terrorist-related deaths in the world, while Russia rather than India, is second in gross lethality.  Moreover, four nations--Iraq, Russia, India and Spain--account for 85% of the terrorist-caused deaths in the world.  Here's a brief table that ranks the most to least deadly:

Country Deaths Percent
Iraq 552 30.41
Russia 436 24.02
India 364 20.06
Spain 191 10.52
Israel 60 3.31
Afghanistan 47 2.59
Saudi Arabia 46 2.53
Egypt 36 1.98
Palestine 23 1.27
Indonesia 10 0.55
Angola 9 0.50
Sudan 9 0.50
Pakistan 4 0.22
Thailand 4 0.22
Uzbekistan 4 0.22
Bangladesh 3 0.17
Philippines 3 0.17
Venezuela 3 0.17
Serbia 2 0.11
Sri Lanka 2 0.11
Turkey 2 0.11
Argentina 1 0.06
Congo 1 0.06
Nepal 1 0.06
Somalia 1 0.06
Ukraine 1 0.06
Total 1815 100.00

[No military deaths are counted in this ranking, although it does count police, counterterrorism personnel and officials, as well as paramilitary forces.  I identified a total of 95 military deaths that were included in the report, though I'm not sure what rationale was used.  80 of the 95, which were excluded from the tally above, occurred in Kashmir.  I can understand why one might use a different yardstick there, but to be strictly comparable to terrorism elsewhere these probably ought to be omitted from the tally.]

As mentioned previously India is third in lethality, after Iraq and Russia.  Furthermore, there are 13 lucky countries that had significant terrorist events according to the report, but suffered no civilian deaths. These countries don't appear on the table above, at all.

If we count both dead and wounded, which some would argue is reasonable (even though the over 1,000 injured in the first World Trade Center attack in 1993 elicited little more than a large collective yawn from both the citizenry and their political representatives) Spain moves to the top of the list!

Country Casualties Percent
Spain 2095 25.72
India 1708 20.97
Iraq 1550 19.03
Russia 1387 17.03
Israel 380 4.67
Saudi Arabia 248 3.04
Indonesia 192 2.36
Palestine 113 1.39
Afghanistan 104 1.28
Bangladesh 103 1.26
Thailand 52 0.64
Egypt 48 0.59
Pakistan 26 0.32
Turkey 22 0.27
Sri Lanka 21 0.26
Angola 19 0.23
Sudan 19 0.23
Ukraine 14 0.17
France 10 0.12
Venezuela 6 0.07
Serbia 5 0.06
Congo 4 0.05
Uzbekistan 4 0.05
Bolivia 3 0.04
Philippines 3 0.04
Argentina 2 0.02
Germany 2 0.02
Malaysia 2 0.02
Somalia 2 0.02
Nepal 1 0.01
Total 8145 100.00

[Again, these are civilian casualties only.  The tally doesn't include 166 military wounded or 95 killed, the vast majority of whom were in Kashmir.]

The same four countries still account for 83% of terrorist activity in the world.  However, counting injured as equivalent to those who lost their lives obviously distorts the picture, and Spain moves to the top of the list only because of the March 11 Madrid attack, which killed 191 and injured almost 2,000.  There was only one other terrorist attack recorded for Spain in the report, perpetrated by the Basque Separatist ETA, which killed or injured no one.  Moreover, 9 countries that appear in the report were lucky enough to have had no civilian casualties at all!


The Quality of the Enemy

So, counting injured in this way leaves something to be desired in terms of method, and we need a better metric for "terrorist activity." This was the real reason I embarked on the current project to reassess the NCTC Report.  Raphael Perl, at the Congressional Research Service, and others emphasize the need for some sort of measure that guages the "quality" of terrorist attacks.  Although Perl stresses the importance of public opinion, I maintain that it's hard to distinguish between public opinion created by media bias and that attributable to the terrorists deft manipulation.  Clearly the media is a critical dimension in the war, but the proper way to account for it isn't necessarily to treat it as a reflection of enemy quality. 

I have to assume that a public well-informed about the nature and risks involved would be able to resist media manipulation, whether engineered by a crafty enemy or by a pliant media.  So until I can figure out how to assess that dimension better I've decided that the best approach is to concentrate on the quality of the information and analysis that I present, thereby diminishing the vulnerability of the public as much as possible.  Public opinion would then better reflect real conditions, and would begin to approximate the "wisdom of the people" that presents the toughest possible obstacle to an enemy, even one presenting in the guise of a friend.

There isn't a lot of information contained in the short descriptions in the NCTC Report that would enable an extensive assessment of quality, but I figured that even a crude measure might provide better insight than any of the quantity measures alone.  Although the scale I came up with for quality is pretty crude and limited, I think it's at least useful. 

The quality measure is composed of two elements that are summed and that are then multiplied by casualties to produce an overall number that's a composite of both quality and quantity.  I call this composite "gain," both for the sake of simplicity and because I'm just not very imaginative.  The two elements of quality I've used are "Success/Failure" and "Skill/Coordination."  The judgments for making the determination of success/failure are simply based on whether, and how well, the objectives of the attack seem to have been achieved.  Obviously I had to make some assumptions about the objectives to assign a score to this element, but those assumptions seem reasonable. An armed attack on a police barracks that results in no deaths and few injuries, except for the attackers, has to be viewed as an attempt that did not meet its objective.  The assessment of success/failure ranges from -1 to 2, where a negative score indicates that an attack was conducted that was either foiled outright, or that resulted in significant negative consequences for the perpetrators (such as an attack that was carried out with no casualties where all the perpetrators were immediately killed or captured).  In other words the attack backfired, in a big way.  There weren't many of these events, unfortunately (only 7, or 1%), so they don't account for any large swings in quality.  The distribution was: -1(1%), 0(9%), 1(26%), 2(64%).  Unfortunately, almost 2/3rds of the attacks were very successful.

The second element of quality is the "Skill/Coordination" aspired to by the terrorists.  This is basically a measure of whether multiple attacks were coordinated, or whether an individual attack was planned to be concurrent with political events so as to influence outcomes.  The metric also involves the prominence of authority figures who are targeted.  It's a measure of the strategic element, in other words.  Skill/coordination ranged from 0 to 3.  Here the news is a bit more promising.  The frequency distribution is: 0(72%), 1(22%) 2(4%) and 3(1%).  Well over 2/3rds of attacks did not exhibit any appreciable skill or coordination.  They weren't attempting to play a symphony, they were just making noise.  By this measure the terrorists are either not very ambitious, or their organization has been seriously disrupted, or I'm simply being too tough on them.  I can certainly see an argument for a much more refined measure of skill/coordination, but such an assessment would really demand much better data than are available to me.  Besides, although I have a good strategic sense and am a methodologist by trade, I'm not an expert in counterterrorism.

It is possible in my schema to simultaneously achieve a high score on success/failure with a low score on skill/coordination, or  visa versa.  Operationally the two elements of quality are summed, and are then multiplied by a weighted number reflecting casualties, to create a composite quality/quantity score that I've chosen to call "gain," at least until someone gives me a better term.

The casualties number doesn't account injuries equivalent to deaths, for obvious reasons.  It is a weighted sum where the number of deaths is 10 times more important than injuries.  This weighting seems reasonable in light of the public's reaction to injuries, though I'm open to suggestions.  Essentially the weighted casualties formula is:

Weighted Casualties = Deaths + 0.1*Wounded or Injured.

The formula for Quality is:
Quality = Success or Failure + Skill or Coordination Level Attempted. 

The ultimate formula is then:
Gain = Casualties * Quality. 

Rocket surgery, it's not.

As an example of how I scored these events let's look at the single most devastating terrorist attack in history, so far: 9/11/01.  That attack involved 4 separate coordinated events where the first three achieved their objective, and the fourth did not.  The first three events received a "2" for success/failure and a "3" for skill/coordination.  These are both the highest score possible for each element.  The Shankesville event also received a "3" for the skill/coordination aspired to, but received only a "0" for success/failure, largely as a result of the resourcefulness and courage of ordinary Americans.  The event wasn't a complete failure for the terrorists, because it killed 44 courageous people, but because its main objective wasn't achieved it may have actually contributed more hope than despair on that awful day.

Total "gain" for this single attack, composed of four separate events carried out on the same day (each scored separately), was approximately 15,000, or roughly 5 times the number of people killed.  (I didn't count 9/11 injuries, because I couldn't find a source for that information.)  By way of comparison, the combined gain score for all terrorist events during 2004, or at least those recorded in the NCTC Report, was a little over 8,000.  And that number includes the attack on the school in Beslan, in the Russian province of Ossetia, as well as the Madrid transit attack on 03/11.

Again, I'm open to other scoring suggestions... but the primary drawback of making a highly refined quality assessment is the paucity of detailed information contained in the short narratives.  If there are flaws in the specification some of those will wash out as a result of using the same method for all events, so there'll be a mix of over- and under-estimation of gain.  Consistency is my remedy for imprecision.  (If the distribution of error is approximately random this assumption is more valid than if it's highly skewed, obviously.) 

I felt it valid to assume that an armed attack on a police HQ, or a security bunker, that didn't kill anyone was probably something of a failure, while an attack that created lots of carnage was a "success."  I realize this is a perverse notion of success, but I think I've made reasonable decisions that optimize the potential of the data I had available.  So, without further ado, here's the ranked table for "terrorist gain"--the enemy's best (or worst) performance, depending on how you look at it:

Country Gain Percent
Iraq 2315.2 27.88
Russia 1952.2 23.51
Spain 1906.2 22.96
India 1013.9 12.21
Israel 276.6 3.33
Saudi Arabia 240.4 2.90
Egypt 186 2.24
Afghanistan 103.6 1.25
Indonesia 84.6 1.02
Palestine 54.1 0.65
Bangladesh 26 0.31
Thailand 20.4 0.25
Angola 20 0.24
Sudan 19.7 0.24
Uzbekistan 16 0.19
Turkey 13.6 0.16
Pakistan 9.3 0.11
Philippines 9 0.11
Sri Lanka 7.8 0.09
Venezuela 7.7 0.09
Argentina 4.4 0.05
Serbia 4.4 0.05
Nepal 3 0.04
Ukraine 2.3 0.03
Congo 2.2 0.03
Somalia 2.2 0.03
France 2 0.02
Malaysia 0.4 0.00
Bolivia 0.3 0.00
Total 8303.5 100.00


What Do We Know?

A few caveats.  According to the metric that I've devised, if a kidnapping or abduction occurred, but the abductees were released without injury (or if their fate is unknown), the event contributed nothing to the "gain," because at least one of the two terms in the product (in this case casualties) was zero.  Clearly terrorists probably gained from some of these abductions, either in terms of publicity or ransom.  Plus, it doesn't seem quite accurate to account zero gain from an abduction simply because the fate of the victim isn't known. Worldwide there were 66  kidnap cases with apparently no casualties, 49 of which were in Iraq and 7 in Kashmir.  In fewer than a handful of these was the fate of the victim still in doubt, but it might be worthwhile to modify the metric to include these abductions somehow.  However, doing so would only amplify the significance of Iraq which the current metric already reveals as the front line in the War on Terror, so that modest deficiency in the composite gain measure is not a good reason to reject these findings.

Next, to make any ultimate assessment of how well we're doing in the mis-named "War on Terror" it ought to be obvious that one needs comparable indices to quantify the gains for both sides.  Ignoring the performance of the allies is like looking at the polling data and campaign contributions for only one candidate in a Presidential race, without considering how well his opponent is doing.  But having said that, it's at least possible to get a sense of the price we're paying even if it's not clear what price the enemy has paid, by simply comparing the aggregate gain score with the analogous number for 9/11.  And the price the opponents of terrorism paid in 2004, distributed mostly over a handful of countries, was a little more than half what the US paid in one single day, almost four years ago.  That, at least, gives us some perspective.

Finally, my intention in conducting this analysis is not to suggest that we ignore "the numbers" that Larry Johnson considers important: the raw frequency of attacks per country.  Clearly the frequency of attacks in India/Kashmir is important for several reasons.  For one thing, the perpetrators are almost certainly linked to influences that are well-placed in the intelligence apparatus of an important US ally (Pakistan) who is concurrently engaged in a kind of "lukewarm war" with another US ally (India).  This isn't good news.  For another, it is at least conceivable that the people responsible for many small attacks in one theater could scale up to many large attacks, a process we could call the "malignancy effect."  It is a real potential, but there are reasons why we shouldn't be alarmed about it just yet.

For one thing, the average competency of the attackers in Kashmir isn't very high.  Compared to Iraq, for instance, which has an average gain per attack of 11.5, the average gain in Kashmir is 3.4, against an enemy not nearly as militarily competent as the US.  The "insurgents" in Kashmir would not last long if transplanted to Iraq.  Almost 20% of the events in Kashmir are home invasions, while a significant number of the rest are small scale street attacks.  There are a lot of attacks directed at police in Kashmir, but many of them are unsuccessful, resulting either in no deaths and few injuries, or harm to civilians who might otherwise sympathize with the attackers.  Many of these are grenades tossed or launched that either miss their targets entirely, or do little damage.  Those terrorists engaged in Kashmir, as well as their managers and organizers, are clearly a second or third string team, the members of which would not survive long if transplanted to the primary theater in the War on Terror: Iraq.

In addition, looked at from the strategic perspective "small ball," either in Kashmir or Iraq, while impressive, is not ultimately very successful.  In both cases the totalitarians are arguably losing the fight.  Talks between India and Pakistan over the disposition of Kashmir are more promising than ever, and it's not inconceivable that one might one day be able to honeymoon in the Vale of Kashmir.  (I may actually be holding out for that, on some deep psychological level.  It's an enchanting place, which explains in part why it has always been so contested.)

And in Iraq not only have the activities of the terrorists awakened a certain revulsion in the larger Middle East, but the restraint of the Shi'a has tapped and inspired a latent Arab pride in humanitarian values, and hope for the future of self-governance that has been dormant since the beginning of the Cold War midway through the last century.  The knees of the autocrats aren't sturdy.

But, the war isn't abating.  If anything it's growing more, rather than less, intense, just as did a previous war when Grant made his fateful decision to turn south in pursuit of Lee after the Battle of the Wilderness.  And one can also expect that as "small ball" doesn't achieve their objectives the enemy may very well intend some large scale attacks to recoup, possibly in areas of the world that aren't prepared for the onslaught.  The parade of names that join those of Antietam, Gettysburg, The Bloody Angle, Chateau Thierry, The Argonne, Belleau Wood, Iwo Jima, The Bulge, 9/11 and even Hiroshima and Nagasaki, may not have reached their end.  There isn't much room either for complacency, or despair.

(Cross-posted by Demosophist to Demosophia, Anticipatory Retaliation and The Jawa Report)

Posted by Demosophist at May 16, 2005 02:30 AM | TrackBack
Comments

mnemonic overview:reading spanned saline pokes?pawns reintroduces

Posted by: at June 26, 2006 11:51 PM

nab shoal introverted damning attempt reproducers infringed .

Posted by: at June 27, 2006 11:22 AM

evenness grated push,Guadalupe chamber tottered miraculously scantiness!

Posted by: at June 27, 2006 11:23 AM