Results of research on the causes of political violence are questionab
le owing to poor reliability of data and the tendency to combine viole
nce perpetrated by several different actors in a general measure of ''
political violence.'' Dynamic models and a focus on the context of vio
lence allow examination of these issues. I compare an overall count of
political deaths from The New York Times Index with a comparable coun
t from a more complete database. The statistical inferences generated
using the two measures are virtually identical. More important, disagg
regating political deaths by the agency responsible for the deaths sho
ws that anti-state violence, pro-state vigilante counterviolence, and
state violence follow separate, distinctive dynamics. Although standar
d newspaper sources may provide reliable measures of political violenc
e, they do not guarantee a complete examination of its complex dynamic
s.