holds when the correlation is with attempted coups, a non-significant r = -.181). The
conclusion is that, at least as measured, any finding of increased coup activity cannot be
attributed to more complete news coverage of such events.
Valuable evidence cross-validating our coup measure comes from comparing it
with Dix (1994) who, also drawing on multiple sources, measured coups in these
countries for the years 1967-1991. His score of coups-by-country correlates with ours at
r = .919, significant at .01. This is strong independent support for the validity of our coup
measure.
4
A spiked diamond represents the occurrence of one or more coups for the designated
year and country.
5
According to the standardized residual plot, which follows roughly a straight line, the
data are neither oversmoothed or undersmoothed (see Jacoby, 1997, p.74).
6
Notably, the differenced series is characterized by a moving average process.
7
Although civil wars usually entail continuous action against the government, not all
such actions are serious attempts to overthrow the government. Much of the action can
be characterized as harassment, such as attacks on infrastructure, kidnappings, bomb
detonation, and civilian killings. Regardless of whether a civil war was underway or not,
these incidents were not considered attempted coups. To quantify attempted coups
during times of continuous fighting we considered a discrete incident (of an attempted
coup) to be a concerted attack on the government or military. If activities characteristic
of an attempted coup were constantly underway, an attempted coup was recorded,
assuming it was apparent that rebel forces had made a concrete effort to overthrow the
government. Such an effort counts as one attempted coup rather than recording each
Convention