Nailing Down Vulnerability

The main topic of my research is the vulnerability of urban areas for terrorism. But what exactly is vulnerability, and how can it be defined and operationalized? There are many discussions going on about these questions in the social sciences community. Albeit only very few of them are related to terrorism threats, the ideas and approaches presented there turn out to be very useful for me in finding my very own definition – in nailing down the quite abstract idea of vulnerability into a quantifiable measure.

One of the aforementioned scientific articles that deals with the operationalization of vulnerability is a paper by Piegorsch et al. (2007), published in Risk Analysis in 2007. The title is “Benchmark Analysis for Quantifying Urban Vulnerability to Terrorist Incidents” and hence in this case is even related to terrorism vulnerability. Some of the ideas presented in this article had been introduced in two other publications by Cutter et al. (2003), and Borden et al. (2007). Not all of the methods presented there will be transferable to my research and I disagree with some of the points they made, but I highly respect their efforts in vulnerability research.

Basically the main paper by Piegorsch et al. (2007) is an attempt to introduce a quantitative methodology to characterize vulnerability to terrorist attacks using a place-based vulnerability index and a database of historic terrorist incidents. They aimed at studying the relationship between vulnerability and terrorist incidents and employed a benchmark approach, which originates from epidemiological sciences and health risk assessment, to the vulnerability levels they calculated before (cf. Piegorsch et al. 2007, 1411).

To do so they selected and spatially defined their study areas, a number of 132 urban areas. The detailed selection process in described in Borden et al. (2007, 3). In addition to that they collected historic data about terrorist incidents from two widely renowned terrorism databases: the Terrorism Knowledge Database (TKB) of the Memorial Institute for the Prevention of Terrorism (MIPT), and the Global Terrorism Database (GTD) by the U.S. National Center for the Study of Terrorism and Responses to Terrorism (START). In total they gathered a number of 1,098 incidents in the aforementioned study areas. After doing so they created two binary indicators whether incidents (one or more) have happened in a certain area and whether there have there been casualties or not. Following that they calculated the probability for occurrence of a terrorist incident and related casualties in every location and related these probabilities to a place-based vulnerability index (PVI).

The most interesting part for me and my upcoming research was their description of how they calculated this index value. It consists of three components: a social vulnerability index (SoVI), a natural hazard vulnerability index (HazVI), and a vulnerability index of the built environment (BEVI). While all three make up a holistic image of the actual vulnerability situation in a give location, I focused for now on the calculation of the SoVI value. It is explained in great detail, together with a list of variables and factors and the reasons for their inclusion in the index, in Cutter et al. (2003, 245-254). The index focuses on the fact that vulnerability is caused by inequalities of the affected. Those inequalities can be divided into social inequalities, like age, gender, health, etc.; and place inequalities, like the level of urbanization, population growth rates, and economic vitality, amongst others. While the former define the susceptibility to harm, and the ability to respond, the latter define the characteristics of communities and the environment (cf. Cutter et al., 2003, 243).

The paper goes great lengths do describe both the disagreements in the vulnerability research society about specific variables, and the consensus about major factors. Those factors are as diverse as the lack of access to resources (e.g. information, knowledge, technology), to political power and representation, or social capital (including social networks and connections); beliefs and customs; building stock, and age; existence of frail and physically limited individuals; type and density of infrastructure and lifelines.

In the case of Cutter et al. (2003) the authors went from as much as 250 relevant variables, which had been collected from socio-economic datasets, to 42 independent ones after computation and normalization, and further reduced this data using factor analysis to 11 factors: personal wealth; age; occupation; single-sector economic dependance; race (african american, asian); ethnicity (hispanic, native american); housing stock & tenancy; density of built environment; infrastructure dependance.

Piegorsch et al. (2007) used this SoVI, together with the aforementioned HazVI and BEVI values to create a more general place-based vulnerability index, PVI. Analysis (which will not be explained here in more detail) then showed that the weighted PVI measure can significantly explain the relationship between the index value and actual terrorist incidence. Also, the benchmarking approach helped to identify places at risk, where the 
PVI value was greater than a certain benchmark value.

As I mentioned in my introduction, these three papers can prove to be very interesting and influential for my further studies. Yet, there are a number of shortcomings or critiques that I identified so far. First, this is the US-centric approach of the research, which is most obvious in the selection of the variables and factors used to calculate the vulnerability indices. While it makes sense to focus on society-specific factors, it also means that the analysis and selection of variables for other geographies has to be done in an equally careful way. Second, the spatial resolution of the analysis presented is very rough. While all three articles focus on the city level, I want to analyze vulnerability inequalities inside urban systems, and will hence focus on a more macro scale. Whether the necessary data to achieve this will be available has yet to be determined. Another major point of critique is that the simplified measure of terrorist incidents does not reflect the severity of incidents at all. Therefore the failed abduction attempt of a diplomat will be represented in the same way as devastating events like those of September 11th 2001. Also, some of the factors chosen don’t seem to be useful to represent the very specific vulnerability for terrorism. For example the idea of attractiveness for a terrorist attack is not represented in the model at all.

My next steps from here on will be to analyze the feasibility of a 1:1 reproduction of the aforementioned studies for Japan, preferably combined with an identification of relevant factors and variables for Japanese urban areas, that build upon the ones shown in the SoVI and PVI approaches. Also, I want to research ways to include the dimension of attractiveness, and to add the relevance of critical infrastructures into my framework.


Borden, K. A., Schmidtlein, M. C., Emrich, C. T., Piegorsch, W. W., Cutter, S. L. (2007). Vulnerability of U.S. Cities to Environmental Hazards. Journal of Homeland Security and Emergency Management, 4(2).

Cutter, S. L., Boruff, B. J. & Shirley, W. L. (2003). Social Vulnerability to Environmental Hazards. Social Science Quarterly, 84(2), 242-261.

Piegorsch, W. W., Cutter, S. L. & Hardisty, F. (2007). Benchmark Analysis for Quantifying Urban Vulnerability to Terrorist Incidents. Risk Analysis, 27(6), 1411-1425.

Presentation Slides

A slightly modified version has been presented at the Annual Tsukuba University SIS Summer Seminar in Yamanakako, Japan, on July 18th, 2011.

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.