terrorism: Global Terrorism Database yearly summaries
In Ecdat: Data Sets for Econometrics

Description Usage Format Details Author(s) Source References Examples

The Global Terrorism Database (GTD) "is a database of incidents of terrorism from 1970 onward". Through 2015, this database contains information on 141,966 incidents.

terrorism provides a few summary statistics along with an ordered factor methodology, which Pape et al. insisted is necessary, because an increase of over 70 percent in suicide terrorism between 2007 and 2013 is best explained by a methodology change in GTD that occurred on 2011-11-01; Pape's own Suicide Attack Database showed a 19 percent decrease over the same period.

1
2
3

  data(terrorism)
  data(incidents.byCountryYr)
  data(nkill.byCountryYr)

incidents.byCountryYr and nkill.byCountryYr are matrices giving the numbes of incidents and numbers of deaths by year and by location of the event for 206 countries (rows) and for all years between 1970 and 2016 (columns) except for 1993, for which the entries are all NA, because the raw data previously collected was lost (though the total for that year is available in the data.frame terrorism).

NOTES:

1. For nkill.byCountryYr and for terrorism[c('nkill', 'nkill.us')], NAs in GTD were treated as 0. Thus the actual number of deaths were likely higher, unless this was more than offset by incidents being classified as terrorism, when they should not have been.

2. incidents.byCountryYr and nkill.byCountryYr are NA for 1993, because the GTD data for that year were lost.

terrorism is a data.frame containing the following:

year

integer year, 1970:2016.

methodology

an ordered factor giving the methodology / organization responsible for the data collection for most of the given year. The Pinkerton Global Intelligence Service (PGIS) managed data collection from 1970-01-01 to 1997-12-31. The Center for Terrorism and Intelligence Studies (CETIS) managed the project from 1998-01-01 to 2008-03-31. The Institute for the Study of Violent Groups (ISVG) carried the project from 2008-04-01 to 2011-10-31. The National Consortium for the Study of Terrorism and Responses to Terrorism (START) has managed data collection since 2011-11-01. For this variable, partial years are ignored, so methodology = CEDIS for 1998:2007, ISVG for 2008:2011, and START for 2012:2014.

method

a character vector consisting of the first character of the levels of methodology:

c('p', 'c', 'i', 's')

incidents

integer number of incidents identified each year.

NOTE: sum(terrorism[["incidents"]]) = 146920 = 141966 in the GTD database plus 4954 for 1993, for which the incident-level data were lost.

incidents.us

integer number of incidents identified each year with country_txt = "United States".

suicide

integer number of incidents classified as "suicide" by GTD variable suicide = 1. For 2007, this is 359, the number reported by Pape et al. For 2013, it is 624, which is 5 more than the 619 mentioned by Pape et al. Without checking with the SMART project administrators, one might suspect that 5 more suicide incidents from 2013 were found after the data Pape et al. analyzed but before the data used for this analysis.

suicide.us

Number of suicide incidents by year with country_txt = "United States".

nkill

number of confirmed fatalities for incidents in the given year, including attackers = sum(nkill, na.rm=TRUE) in the GTD incident data.

NOTE: nkill in the GTD incident data includes both perpetrators and victims when both are available. It includes one when only one is available and is NA when neither is available. However, in most cases, we might expect that the more spectacular and lethal incidents would likely be more accurately reported. To the exent that this is true, it means that when numbers are missing, they are usually zero or small. This further suggests that the summary numbers recorded here probably represent a slight but not substantive undercount.

nkill.us

number of U.S. citizens who died as a result of incidents for that year = sum(nkill.us, na.rm=TRUE) in the GTD incident data.

NOTES:

1. This is subject to the same likely modest undercount discussed with nkill.)

2. These are U.S. citizens killed regardless of location. This explains at least part of the discrepancies between terrorism[, 'nkill.us'] and nkill.byCountryYr['United States', ].

nwound

number of people wounded. (This is subject to the same likely modest undercount discussed with nkill.)

nwound.us

Number of U.S. citizens wounded in terrorist incidents for that year = sum(nwound.us, na.rm=TRUE) in the GTD incident data. (This is subject to the same likely modest undercount discussed with nkill.)

pNA.nkill, pNA.nkill.us, pNA.nwound, pNA.nwound.us

proportion of observations by year with missing values. These numbers are higher for the early data than more recent numbers. This is particularly true for nkill.us and nwound.us, which exceed 90 percent for most of the period with methodology = 'PGIS', prior to 1998.

worldPopulation, USpopulation

Estimated de facto population in thousands living in the world and in the US as of 1 July of the year indicated, according to the Population Division of the Department of Economic and Social Affairs of the United Nations; see "Sources" below.

worldDeathRate, USdeathRate

Crude death rate (deaths per 1,000 population) worldwide and in the US, according to the World Bank; see "Sources" below. This World Bank data set includes USdeathRate for each year from 1900 to 2014.

The WorldDeathRate here were read manually from a plot on that web page, except for the the number for 2015, which was estimated as a reduction of 0.73 percent from 2014, which was the average rate of decline (ratio of two successive years) for 1990 to 2014. The same method was used to estimate the USdeathRate for 2015 as the same as for 2014.

NOTE: USdeathRate is to two significant digits only, unlike WorldDeathRate, which has four significant digits.

worldDeaths, USdeaths

number of deaths by year in the world and US

worldDeaths = worldPopulation * worldDeathRate.

USdeaths were computed by summing across age groups in "Deaths_5x1.txt" for the United States, downloaded from http://www.mortality.org/cgi-bin/hmd/country.php?cntr=USA&level=1 from the Human Mortality Database; see sources below.

kill.pmp, kill.pmp.us

terrorism deaths per million population worldwide and in the US =

0.001 * nkill / worldPopulation

pkill, pkill.us

terrorism deaths as a proportion of total deaths worldwide and in the US

pkill = nkill / worldDeaths

pkill.us = nkill.us / USdeaths

As noted with the "description" above, Pape et al. noted that the GTD reported an increase in suicide terrorism of over 70 percent between 2007 and 2013, while their Suicide Attack Database showed a 19 percent decrease over the same period. Pape et al. insisted that the most likely explanation for this difference is the change in the organization responsible for managing that data collection from ISVG to START.

If the issue is restricted to how incidents are classified as "suicide terrorism", this concern does not affect the other variables in this summary.

However, if it also impacts what incidents are classified as "terrorism", it suggests larger problems.

Spencer Graves

National Consortium for the Study of Terrorism and Responses to Terrorism (START). (2017). Global Terrorism Database [Data file]. Retrieved from http://www.start.umd.edu/gtd [accessed 2018-04-08].

The world and US population figures came from "Total Population - Both Sexes", World Population Prospects 2015, published by the Population Division of the Department of Economic and Social Affairs of the United Nations, accessed 2016-09-05.

The World and US death rates came from the World Bank, accessed 2016-09-05.

Human Mortality Database. University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany).

Robert Pape, Keven Ruby, Vincent Bauer and Gentry Jenkins, "How to fix the flaws in the Global Terrorism Database and why it matters", The Washington Post, August 11, 2014 (accessed 2016-01-09).

data(terrorism)
##
## plot deaths per million population 
##
plot(kill.pmp~year, terrorism, 
     pch=method, type='b')
plot(kill.pmp.us~year, terrorism, 
     pch=method, type='b', 
     log='y', las=1)
     
# terrorism as parts per 10,000 
# of all deaths 

plot(pkill*1e4~year, terrorism, 
     pch=method, type='b', 
     las=1)
plot(pkill.us*1e4~year, terrorism, 
     pch=method, type='b', 
     log='y', las=1)
     
# plot number of incidents, number killed, 
# and proportion NA

plot(incidents~year, terrorism, type='b', 
      pch=method)

plot(nkill.us~year, terrorism, type='b', 
      pch=method)
plot(nkill.us~year, terrorism, type='b', 
      pch=method, log='y')

plot(pNA.nkill.us~year, terrorism, type='b', 
      pch=method)
abline(v=1997.5, lty='dotted', col='red')

##
## by country by year
##
data(incidents.byCountryYr)
data(nkill.byCountryYr)

yr <- as.integer(colnames(
  incidents.byCountryYr))
str(maxDeaths <- apply(nkill.byCountryYr, 
                       1, max) )
str(omax <- order(maxDeaths, decreasing=TRUE))
head(maxDeaths[omax], 8)
tolower(substring( 
  names(maxDeaths[omax[1:8]]), 1, 2))
pch. <- c('i', 'g', 'f', 'l', 
          's', 'c', 'u', 'p')
cols <- 1:4

matplot(yr, sqrt(t(
  nkill.byCountryYr[omax[1:8], ])),
  type='b', pch=pch., axes=FALSE, 
  ylab='(square root scale)   ', xlab='', 
  col=cols,
  main='number of terrorism deaths\nby country') 
axis(1)
(max.nk <- max(nkill.byCountryYr[omax[1:8], ]))
i.nk <- c(1, 100, 1000, 3000, 
          5000, 7000, 10000)
cbind(i.nk, sqrt(i.nk))
axis(2, sqrt(i.nk), i.nk, las=1)
ip <- paste(pch., names(maxDeaths[omax[1:8]]))
legend('topleft', ip, cex=.55, 
       col=cols, text.col=cols)

Loading required package: Ecfun

Attaching package: 'Ecfun'

The following object is masked from 'package:base':

    sign


Attaching package: 'Ecdat'

The following object is masked from 'package:datasets':

    Orange

 Named num [1:206] 6208 26 4266 0 846 ...
 - attr(*, "names")= chr [1:206] "Afghanistan" "Albania" "Algeria" "Andorra" ...
 int [1:206] 86 132 1 3 177 130 191 136 141 53 ...
         Iraq       Nigeria   Afghanistan       Algeria         Syria 
        13076          7773          6208          4266          3916 
    Nicaragua United States      Pakistan 
         3617          3003          2874 
[1] "ir" "ni" "af" "al" "sy" "ni" "un" "pa"
[1] 13076
      i.nk          
[1,]     1   1.00000
[2,]   100  10.00000
[3,]  1000  31.62278
[4,]  3000  54.77226
[5,]  5000  70.71068
[6,]  7000  83.66600
[7,] 10000 100.00000