Eric R. Jorgensen and Dr. Jay Goodliffe, Political Science
As we recently observed the one-year anniversary of the embassy bombings in Africa, thoughts turned again to our vulnerability to terrorist attacks. One year ago President Clinton vowed to “wage war on terrorism.” It was at this point that many began to question the effectiveness of military deterrence on terrorist activity. This research examined the question of effectiveness. My hypothesis has been that military strength has no measurable effect in the deterrence of terrorist activities. I proposed instead that terrorists strike powerful, industrialized, democratic countries because their attacks appear more devastating when conducted against these states.
To investigate my hypothesis, I used a statistical technique known as linear regression as a predictive model. Linear regression uses an algebraic line equation (in the form y = mx + b) to explain and predict individual observations. The first step in this process was to collect sufficient data. I collected data concerning number and severity of significant terrorist incidents as well as demographic data on individual nations (military expenditures, GNP, military strength, population, relative freedom, etc.). Size, wealth, military strength and several other factors all seem to contribute to a nation’s overall prowess, both in theory and in statistical analysis. Thus, though several models were attempted, the variables that best fit my final model of terrorist activity were military expenditures as a percentage of GNP, nuclear capability and GNP itself. These are the independent variables. Because I created two models for terrorist activity, I use two different dependent variables with the independent variables, one in each model. The dependent variables are number of terrorist attacks and severity of terrorist attacks.
With the variables and methodology explained, we can now examine the models. The first equation models the number of terrorist attacks.
This shows that using 46 (n=46) cases or observations, this model explains 74% (R2=.740) of the variance in terrorist attacks. This is a significant amount of variance to be explained by only three variables. The t and SE statistics, located beneath each coefficient, tell us if each variable is statistically significant, or if they are statistically different than zero. For the number of observations we are using, the standard degree of significance would require the absolute value of t to be over 2.0. Now looking at the equation itself, we see that all the coefficients are statistically significant. The first number, the constant, represents how many terrorist attacks a country would have if all the variables were set to zero. Thus, if a state did not have a nuclear capability, spent no money on the military and had zero GNP, it would likely experience an average of 2.32 terrorist attacks. Of course such a nation does not exist, so we will examine the other variables. Nuke is a dummy variable, or a variable that take a value of one or zero (one if the nation has nuclear capability and zero if it does not). Thus, having nuclear capability raises the probable number of terrorist attacks by almost nine per year, or from just over two per year to just over eleven. The coefficient for military expenditures as a percentage of GNP (MilExp) is negative, meaning that as spending increases, the number of terrorist activities will fall, which would seem to refute my hypothesis. However, it is a matter of scale. The coefficient is about – 0.5, which means that all else held constant, to drop one terrorist attack per year, the military spending would have to take 2 percent more of GNP [-0.5×2(%)=-1(terrorist attack)]. For the US, this would mean an increase in military spending of about $143 billion or over 50 percent. Thus, this appears to uphold my hypothesis that military spending does little to affect terrorist activity. GNP appears to have a similar problem though statistically significant, because of the rational range of GNP it does not seem to affect terrorist activity much. Thus, the major influence on the number of terrorist attacks seems to be nuclear capability. It seems ironic that the symbol of ultimate deterrence should be also a symbol of the power that attracts terrorism.
Some scholars have indicated that while the number of terrorist activities is not affected by deterrence, the severity of the attacks is (1). My second model addresses this idea:
Once again, nearly this model explains 75 percent of the variance and all the variables are statistically significant. To understand this equation, however, we must understand the scale. Severity is an aggregate of the individual scores. Individual scores range from 0 to 3. For example, murder ranks 2.94. Thus, once again GNP becomes insignificant. Military expenditures as a percentage of GNP remain negative, however it would take a 50 percent rise in military spending to produce a significant change. As before the possession of nuclear weapon is the most influential factor on severity.
Thus, these two equations support my hypothesis that terrorist activities are not significantly affected by military deterrence as measured by nuclear potential, military spending and GNP. Further research should also be done in finding variables to account for the rest of the variance in the equations. Though my measures of free press proved insignificant in these models, more refined measures may help to better explain terrorist activities. More sophisticated measures of deterrence may also produce more complete data. Regional differences have also proven interesting and should be followed up. In short, though I feel confident in my research, there is much more work to do in producing a statistical model of terrorist activity.
References
- Mohr, Philip B. and Henry W. Prunckun, Jr. 1997. Military deterrence of international terrorism: An evaluation of Operation El Dorado Canyon. Studies in International Conflict and Terrorism 20, no.3:267-280.