HOW DOES TERRORISM RISK VARY ACROSS SPACE AND TIME? AN ANALYSIS BASED ON THE ISRAELI EXPERIENCE

Abstract

We study the spatial and temporal determinants of terrorism risk in Israel, using a geocoded database of Israeli terrorist attacks from 1949 to 2004. In selecting targets, terrorists seem to respond rationally to costs and benefits: they are more likely to hit targets more accessible from their own homebases and international borders, closer to symbolic centers of government administration, and in more heavily Jewish areas. We also examine the waiting time between attacks experienced by localities. Long periods without an attack signal lower risk for most localities, but higher risk for important areas such as regional or national capitals.

Keywords:

• Terrorism risk
• Spatial
• Temporal
• Israel

JEL Codes:

• D74
• N4

ACKNOWLEDGMENTS

We thank the editors for many helpful comments and suggestions. We benefited from the programming assistance of Michael Tseng, the research assistance of Arkadipta Ghosh, and consultations with Kim Cragin, who first proposed the importance of terrorist homebasing as a determinant of terrorism risk. We are grateful to Emily Lakdawalla for assistance with the geographic images. We are grateful to the RAND Center for Terrorism Risk Management and Policy (CTRMP) for financial support. The views in this paper are those of the authors alone and not those of the RAND Corporation or the RAND CTRMP.

Notes

¹ Tavares (2004) conducts a cross‐country analysis that concludes richer countries are more prone to terrorism than poorer ones, but that democracy often reduces the risk of terrorism. On the policy side, Barros (2003) has found that policy approaches, like economic growth or terrorism deterrence, have decidedly mixed effects on terrorist behavior and terrorism risk. Similarly, Li and Schaub (2004) find little evidence that countries with more foreign direct investment or global economic ties experience more terrorism.

² Berrebi and Klor (2004) show the potential importance of political cycles on terrorism level in Israel. Blomberg et al. (2004) have argued that, within rich, democratic countries, economic contractions are associated with upsurges in terrorist activity. Enders and Sandler (2005) note that sudden upsurges in terrorist activity tend to be more persistent when the overall level of terrorism is lower, but not very persistent otherwise.

³ This refers to British mandate Palestine.

⁴ An ‘official bureau’ includes the following: National Election Inspection Bureau, Regional Appeal Committee, Regional Licensing and Supervising Bureau, and Regional or Sub‐Regional Population Administration Bureau.

⁵ The database does not include attacks by the Israeli government against Palestinian targets; unfortunately, these data were never collected as part of the RAND Terrorism Chronology or as part of the Claude Berrebi data. However, it is important to note that previous research has suggested that Israeli attacks on Palestinian targets do not causally influence Palestinian attacks on Israel (Goldstein et al., 2001; Jaeger and Paserman, 2005).

⁶ See Berrebi (2003) for a detailed description of the data set and its sources.

⁷ Conflicting cases consisted (with a few exceptions) of incidents not recorded in the RAND database.

⁸ The following sources were consulted: the website of Jane’s (www.janes.com), the website of The Institute for Counter‐Terrorism (www.ict.org.il), The MIPT RAND Terrorism Chronology, the website of the Israeli Ministry of Foreign Affairs (www.mfa.gov.il), works by Edward Mickolus and collaborators (Mickolus, 1980, 1993; Mickolus and Flemming, 1988; Mickolus and Simmons, 1997, 2002), newspaper articles, and RAND terrorism experts. A full list of all 670 sources consulted in the construction of the homebase data is available from the authors upon request.

⁹ For information regarding the economic cost of terrorism in Israel see Berrebi and Klor (2005), Eckstein and Tsiddon (2004), and Fielding (2003a, 2003b).

¹⁰ The ratio of population between any two localities remained fairly constant over this period of time. Moreover, most of the attacks in our sample happened during the last decade (after the Russian immigration of the 1990s), so that total population was reasonably constant over the period of interest.

¹¹ Distance is computed as the shortest distance between two points and does not reflect potential geographical or physical barriers such as rivers or fences.

¹² 10 to 1 odds against attack mean the area can expect an attack 1 out of every 11 times.

¹³ In particular, these are the mean marginal effects, computed across all observations.

¹⁴ Only years at risk were considered. A locality is at risk for a terrorist attack from the date of its founding, or from the date of its first attack, whichever is earlier.

¹⁵ We also estimated the effect of distance to a military checkpoint, but found it to be a widely insignificant predictor of risk.

¹⁶ In principle, the authorities could exploit such a systematic basing strategy, and use it to discover terrorist groups. In practice, however, terrorists often situate their homebases in densely populated areas, where detection and removal is much more difficult even if the authorities are aware of a general location.

¹⁷ A well‐suited method for data that varies gradually, like risk, is the spline. The spline method fits a minimum‐curvature surface through the econometrically estimated risk points. More specifically, it fits a mathematical minimum curvature, two‐dimensional, thin‐plate Spline to a specified number of mapping points, while still passing through all estimated points.

One limitation of this approach was our inability to break down Palestinian areas into polygons that were as small as their Israeli counterparts. Owing to a lack of detailed demographic data, the Palestinian mapping polygons (the map contains four) frequently contained several major urban areas each. Therefore, taking our estimates at face value produced an unsatisfactory figure, with a hot spot in the center of each (large) Palestinian polygon, and continuously diminishing risk away from the center, even though the center of the polygon may not contain any inhabitants while the peripheries may contain a major settlement. As an alternative, for each Palestinian polygon, we imputed a risk level for each major urban area within it, under the assumption that risk was equally divided across urban areas within the same polygon. The map is based on these imputed values for the Palestinian cities.

¹⁸ The shortest time we allow between attacks is one‐third of a day. Therefore, an attack the next day should not be construed as immediate.

¹⁹ See Note 18.