Is Terrorism Falling? A critical look at the Human Security Brief for 2007

Ms. Yael Shahar and Mr. Don Radlauer

The Human Security Brief for 2007[1], released in May 2008, declared that terrorist attacks are on the wane due to declining support for al-Qaida. If true, this is good news for Western governments, which have placed great emphasis on countering al-Qaida’s public appeal in the Muslim world, so far with less than impressive success. However, the report’s optimistic conclusions are questionable on both methodological grounds and historical grounds. In particular, its use of incident statistics to prove its contention is of questionable value.

The Human Security report argues that terrorist incident data for Iraq should be removed from the overall total. The report concludes that

Absent Iraq, both START and MIPT now show net declines of more than 40 percent in fatalities from terrorism since 2001. NCTC’s fatality trend data still show an increase from 2005 to the end of 2006, but it is much less steep than when Iraq deaths are included.

In other words, the claim that the incidence of global ter­rorism is increasing is dependent upon accepting the unusual argument that violence intentionally perpetrated by non-state armed groups against civilians in Iraq should be treated as terrorism, but that similar violence elsewhere should not be counted as terrorism. If this argument is rejected, there has been a net decline in the global terrorism toll between 2001 and 2006.[2]

While apparently persuasive, this conclusion is not as optimistic as it at first appears. The report maintains that MIPT and START terrorist incident databases distort the data by counting Iraqi civil war attacks on civilians as terrorism rather than war crimes. However, this neglects the fact that these fatalities are inflicted by NGO’s rather than by governmental forces—and thus are properly considered terrorism. While a case can be made that a large percentage of the fatalities in Iraq are the result of civil war, it is just as true that terrorism—deliberate politically motivated attacks on civilians—does occur there. These attacks should be included in the total. By throwing out all Iraqi fatalities, a skewed picture emerges.

Far from decreasing the overall fatalities, one could argue that fatalities from other conflict zones, hitherto not included in the totals, might indicate an increasing number of global terrorism-related deaths. Of course, civilian fatalities in Africa may also be counted inconsistently—and the whole division between war crimes and terrorism is fairly artificial when there is no longer a clear division between what’s governmental and what’s private. Meaningful statistics can be compiled only if classification standards are well thought out and applied consistently, and data gathering is thorough and consistent both geographically and temporally.

It is true that Iraq accounts for a disproportionately high share of the terrorism fatalities in the conventional trends analysis. However, it is unclear whether removing Iraqi fatalities from the overall trend is appropriate, since it’s unclear to what extent Iraq is acting as a “lightning rod” for potential terrorists who—absent the conflict in Iraq—might well be carrying out attacks elsewhere. A blanket dismissal of all Iraqi civilian deaths would thus conceal an important part of the story.

In fact, what is currently happening in Iraq has a direct bearing on the future of terrorism; Iraq is, like previous conflicts involving Islamic militants, a training camp for tomorrow’s terrorists. It is generally understood that a large number of those currently fighting in Iraq (and who are responsible for a large percentage of the civilian fatalities in actual terrorist violence, as opposed to civil war, honour killings, etc.) will export their activities from Iraq as soon as conditions there make their activities in Iraq less profitable. When this happens, it is likely enough that the rest of the world will witness a new wave of terrorist violence.


How useful is statistical incident analysis?

However, there is a much more fundamental problem with this whole type of analysis: it is attempting to use standard “bell-curve” statistical thinking to forecast trends in data that is not really “statistical” in that sense. If we were talking about car-accident fatalities, where the statistics are based upon a large number of accidents most of which involve a small number of people, we could meaningfully talk about statistical trends. For example, a recent Jerusalem Post editorial includes the following numbers for Israeli highway fatalities:[3]

2004 428

2005 381

2006 373

2007 351

The number was as high as 500 or 550 a few years earlier. Note the fairly “smooth” declining curve. Statistics like this can indeed tell us meaningful things about road safety, even though by themselves they don’t explain why road fatalities have declined.

Terror fatalities, however, are not distributed like car-accident fatalities. Particularly as regards Jihadi terrorism in the West—but even elsewhere—terrorism consists of a smaller number of larger incidents. This means that trends are much harder to detect with any statistical reliability—and when we do detect a “trend”, it may very well be meaningless. For example, a look at Figures 1.1 and 1.2 in the Human Security Brief Press Release show that over the three years 1998 through 2000, there was a significant decline in the number of terror fatalities! How much predictive value did this “trend” have for 2001? Basically, statistics with lots of “scatter” (or a high standard deviation, to be more technical) are much less valuable for establishing meaningful trends.


Why the dog did not bark in the night

Quite apart from statistical concerns, assessing the level of “threat” by counting fatalities has some potential weaknesses. First, it takes no account of changes in the effectiveness of counter-terrorism measures. If more attacks are being thwarted but the terrorists’ motivation and capabilities remain the same, is it correct to say that the threat has declined? Second, it takes no account of the efforts terror organizations may expend to expand or rebuild their capabilities. In particular, after the loss of its facilities in Afghanistan—along with many personnel—al Qaeda likely has had to focus in recent years on rebuilding itself more than on carrying out major attacks. If this is the case, the threat (measured by al Qaeda’s ability to carry out attacks and its interest in doing so) may now be greater than it was a couple of years ago, even though fatality figures do not yet reflect this.

None of the incident databases used by the report keep track of failed attacks or plots which were intercepted before they came to fruition. This is significant, as it may be indicative of a different reason for the (arguable) decline in terrorism-related violence: namely, that increased security has arisen to meet the terrorist threat, and that this security is beginning to pay dividends in decreasing the number of successful terrorist attacks. The report does in fact mention this as one of the factors involved:

However, there are several reasons for believing that the recent decline in Islamist terrorism does in fact mean that the threat is diminishing.

First, counterterrorism efforts, although still plagued by a multitude of problems, are more widespread, more coordi­nated, and more effective today than they were prior to 9/11. Part of the reason we are seeing fewer terrorist attacks is that a greater number are being prevented from occurring before they can even be launched. [4]

This begs the question of whether the same number of attacks are being attempted as previously. If so, then it indicates, not a decline in terrorist attacks, but merely a decline in successful terrorist attacks. This in turn would mean that the motivation to carry out attacks is still high. The Human Security Report argues that in fact, the motivation is itself on the wane:

Second, there is growing evidence of bitter doctrinal infighting within, and defections from, the now largely decentralized global Islamist network. Such developments are a classic sign of organizational crisis and incipient breakdown. Given that the Islamists have failed to achieve any of their strategic goals, and given the humiliating recent defeats experienced by al-Qaeda In Iraq, this development is not surprising.

Third, is the extraordinary drop in support for Islamist terror organizations in the Muslim world over the past five years—a decline that is driven by the increasingly popular rejection of the terrorists’ indiscriminate violence (that mostly targets fellow Muslims), their extremist ideology, and their harshly repressive policies.[5]

One interesting point here is that if the report is right in saying that al-Qaida’s support is on the wane in the Muslim world, it does not necessarily mean that support for al-Qaida’s ideology is also declining. Put simply, the decline in support could indicate that al-Qaida’s tactics are now perceived as ineffective at accomplishing the goals it purports to achieve. The surveys cited by the report, none seem to have focused on the perceived justification of al-Qaida’s Islamist goals. Without knowing the degree of support for the ideology which gave rise to al-Qaida’s terrorism, we cannot predict whether the same publics will simply champion a different organization which sets itself up as the heir to al-Qaida’s throne.

Conclusion

All this does not mean that the report is necessarily wrong in pointing out a decline in the number of terror fatalities. It is entirely possible that fatalities have indeed declined over the last few years—although the measurement problems are huge, and you’d need an awfully good dataset backed by a rigorous reporting and classification methodology to make such pronouncements with any degree of assurance. The real question is whether such a trend—if it exists—has any predictive value; only if it does can we say that the threat of terrorism has been reduced. (Otherwise we’re living in the world of September 10, when the “trend” said things were getting better!)

In order to establish a really meaningful trend, we need either strong statistics—like those we get for automotive fatalities, with relatively low “scatter”—or less-strong statistics (i.e. with substantial “scatter”) backed up by other information that might give us some assurance that the threat of terrorism was genuinely declining. The Canadian report attempts to do so by citing surveys indicating declining public support for al Qaeda terrorism; however, there is a huge missing “middle premise” here: the report implicitly assumes that a decline in mass support for Jihadi terrorism necessarily implies a reduced threat of Jihadi terror attacks. This is not necessarily the case! Even if a decline in public support is bad news for al Qaeda, the organization still has a potentially huge pool of recruits and contributors. It is even possible that al Qaeda might respond to its declining popularity by (for example) ceasing to carry out (or encourage, or authorize, or publicly praise) attacks in Moslem countries, and instead will carry out more and bloodier attacks against the United States, Israel, and other distant (and less sympathetic) targets. (Actual information indicating that al Qaeda’s leaders are now less willing or able to carry out attacks than they were in the past would, of course, make a much stronger case for a decline in the threat of terrorism.)

Al Qaeda is indeed facing a number of serious challenges today, including an ideological backlash among respected Islamist theorists. It is possible that this will result in a decline in the threat of Jihadi terrorism over some unspecified time span. It is further possible that such a trend has already begun to manifest. However, short-term fatality statistics alone are not a very powerful tool for establishing or measuring such a trend; to the extent that fatality statistics are meaningful at all in measuring the threat of terrorism, they can be interpreted only on a time scale substantially greater than the one the Canadian group is using. (a five- to ten-year scale is the minimum we should accept as even marginally meaningful; but ideally twenty- to fifty-year measurements would be preferable.) Any additional information that is adduced to help establish such a trend must be carefully and skeptically evaluated (and weighed against contrary evidence) to ensure that it has a genuine and strong influence on the phenomenon being measured.

In short, while it would certainly be nice if the threat of terrorism had really declined, it cannot be said that the Canadian study has made a convincing case for a such a decline.