Transnational terrorism 1968-2000: thresholds, persistence, and forecasts.

• Author: Enders, Walter

JEL Classification: D74, C32, H56

1. Introduction

   The death, destruction, and economic impacts resulting from the four hijackings on September 11, 2001 (henceforth 9/11), made policy makers and the public acutely aware of the threat posed by transnational terrorism to liberal democracies worldwide. (1) The sheer magnitude of 9/11 led to repercussions that have touched most countries through economic spillovers, collateral damage, or security expenditures. As the public learned about the global reach of the al-Qaida network, the true threat of transnational terrorism became better understood. To assist in this understanding, social scientists must apply theoretical and empirical tools to analyze the actions of terrorists and identify the most appropriate response of governments. In particular, economic methods and models can enlighten policy makers on transnational terrorism (Sandier and Enders 2004). Consumer-choice models can be tailored to display terrorists' choices when constrained by their resources and the policies of the authorities (Landes 1978; Sandler, Tschirhart, and Cauley 1983). Empirically, time-series techniques can derive policy implications and forecasts (Brophy-Baermann and Conybeare 1994; Enders and Sandier 1993, 1995).

   Terrorism is the premeditated use or threat of use of extranormal violence or brutality by subnational groups or individuals to obtain a political objective through intimidation or fear directed at a large audience. Terrorist acts are purposely brutal to create an atmosphere of fear while publicizing the terrorists' cause. As the public becomes numb to their acts of violence, terrorists ratchet up the carnage to maintain media interest. An essential ingredient in defining terrorism is the presence of a political objective, which can involve getting a country out of an international organization, the formation of an Islamic state, the United States ending its support of Israel, or other goals.

   A crucial factor of modern-day terrorism is the transnational implications of many terrorist attacks. When a terrorist incident in one country involves victims, perpetrators, or audiences in two or more countries, terrorism takes on a transnational character. A terrorist act may be transnational owing to its impact, its planning and execution, its perpetrators (if known), or its targets and resulting damage. A skyjacking of a plane with passengers from multiple countries is a transnational terrorist event, as is a skyjacking of a domestic flight originating in one country but terminating in another country. The skyjackings on 9/11 were transnational terrorist acts, as were the near-simultaneous bombings of the U.S. embassies in Nairobi, Kenya, and Dares Salaam, Tanzania, on August 7, 1998. We focus on transnational terrorism because of its importance in the post-9/11 world and data availability.

   Although transnational terrorist attacks are down in the post-Cold War period, the lethality of these attacks has increased, with each incident almost 17 percentage points more likely to result in death or injuries (Enders and Sandler 2000). The casualties time series, involving one or more deaths or injuries, displays predictable factors unlike the noncasualties series, which is largely random following detrending. In a subsequent study, Enders and Sandler (2002) use a threshold autoregression (TAR) model applied to just the death series to identity nonlinearities following shocks. When terrorism is in a low-intensity regime with attacks below a data-determined threshold, shocks that raise the level of events are sustainable. In contrast, shocks during a high-intensity regime of terrorism (i.e., attacks above a data-determined threshold) are episodic with series returning rapidly to the mean number of events.

   The primary purpose of the current study is to apply TAR analysis to additional transnational terrorist time series including important subcomponents of the overall all-incident series. In particular, we analyze the death and casualties time series for nonlinearities as well as the component series of bombings with one or more deaths. Additionally, we apply a TAR model to series involving assassinations, hostage-taking, and threats and hoaxes. A wide variety of nonlinearities are uncovered for component series (e.g., shocks to the assassinations series display some persistence in both the low- and high-intensity regime; shocks to the threats and hoaxes series result in persistence in only the high-intensity regime). An important finding is that the response to an external shock depends on current activity levels, which may vary among different kinds of terrorist events: therefore, policy prescriptions must be tailored to the kind of event. A secondary purpose is to engineer a forecasting technique based on the TAR representation of the component time series.

2. Terrorists' Choice--Theoretic Model

   We view terrorists as rational actors who maximize their well-being by allocating their resources among alternative goals while accounting for underlying constraints. (2) Rationality is judged by the terrorists' predictable response to changes in their constraints and not by their particular goals or the means used to achieve them. Terrorist reactions to shocks in high- and low-terrorism states are dependent on available resources. For instance, the ability to shift sufficient resources between activities and intertemporally to sustain higher levels in a low-intensity period is typically greater than in an already high-intensity regime.

   At the group level, terrorists must allocate resources at three different layers: (i) between terrorist (T) and nonterrorist (N) activities, (ii) among alternative terrorist modes of attack, and (iii) among different time periods. For conceptual simplicity, we assume that each of these decisions is independent. Our interest here is to discuss briefly the intertemporal choice of terrorists, germane to the analysis of this study. Analogous to other investors, terrorists can invest resources to earn a rate of return, r, per period. When terrorists want to augment operations, they can cash in some of their invested resources. Suppose that terrorists have a two-period horizon and must decide terrorist activities today ([T.sub.0]) and tomorrow ([T.sub.1]) based on resources today ([R.sub.0]) and tomorrow ([R.sub.1]). The intertemporal budget constraint is

   (1) [T.sub.1] = [R.sub.1] + (1 + r)([R.sub.0] - [T.sub.0]),

   where tomorrow's terrorism equals tomorrow's resource endowment plus (minus) the earnings on savings (the payments on borrowings) from the initial period. Terrorists maximize an intertemporal utility function, U([T.sub.0], [T.sub.1]), subject to Equation 1 and, in so doing, decide terrorist activities over time. Thus, terrorists can react to shocks by augmenting operations not only from curbing nonterrorist activities, but also through an intertemporal substitution of resources. If high-terrorism regimes are supported by an intertemporal substitution, then shocks during such a regime can typically elevate resource-using terrorist activities for only a short time, unlike low-terrorism regimes where terrorists can draw on accumulated resources. This prediction may not characterize non-resource-using threats and hoaxes.

   Thus far we analyze the choice-theoretic decision of a single terrorist group. When groups are linked in a network, the same model applies to the network by attributing the utility and resource constraints to be those of the network. If terrorists are tied together implicitly through similar hatreds and grievances, then multiple terrorist groups may act as a unified whole and respond identically to shocks, so that our choice-theoretic representation would also characterize this conglomerate of groups. By copying one another's innovations, terrorists worldwide take on the appearance of a single group, thereby giving rise to distinct peaks and troughs in transnational terrorist events (Enders and Sandier 1999; Faria 2003).

3. Data and Terrorist Time Series

   Data on transnational terrorist incidents are drawn from International Terrorism Attributes of Terrorist Events (ITERATE), which records the incident date; type of event; casualties (i.e., deaths or injuries), if any; and other variables. By splicing together the earlier ITERATE data sets, ITERATE 5's "common" file contains 40 or so key variables common to all terrorist events from 1968 to 2000 (Mickolus et al. 2002). Thus, earlier ITERATE data for various subperiods no longer need to be consulted. Coding consistency for ITERATE and its updates has been maintained by applying the same criteria for defining transnational terrorist events and associated variables. (3) ITERATE draws its information from the world newsprint and electronic sources.

   In total, we extract six quarterly time series from ITERATE for the purposes of the article. The most inconclusive of these series is the casualties series, in which one or more persons were injured or killed in a transnational terrorist incident. The death series is a subset of the casualties series and includes only incidents where one or more persons--terrorist or victim--died. We focus on these time series, because earlier work indicates that they are more predictable than the inclusive series of all incidents (Enders and Sandler 2000). We use quarterly, rather than monthly, data to avoid periods with zero observations. Because bombing is the favorite tactic of terrorists, accounting for almost half of all events, a time series involving bombings with one or more deaths (henceforth called the bomb series) is also extracted. Bomb combines seven types of events--explosive bombings, letter bombings, incendiary bombings, missile attacks, car bombings, suicide car bombings, and mortar and grenade attacks--involving at least one death.

   The remaining three series are components of the all-incidents series. The assassinations series consists of...