Roberts takes on the commonly made assertion that in contemporary warfare (since the early 1900s)that 80-90% of war victims/casualties have been civilians. This would suggest a civilian-military death ratio of 9:1. It is easy for such a startling statistic to take on a life of its own - and it has, according to Roberts, having been cited without empirical support in venues as influential as the British Medical Journal.
He traces the origins of the claim back to a set of statements in a couple of influential reports in the early 1990s that were broad inferential estimates based on some questionable assumptions - not verified counts of any kind - and that used a very broad definition of what constitutes a war casualty - including suffering from famine or displacement.
Roberts also traces some of the skepticism of these claims by the International Committee of the Red Cross (ICRC), contradictory data from the World Health Organization (WHO), and the intellectual melee that ensued from the Human Security Report, which has argued that in today's wars "actual death tolls are relatively small—and have been decreasing" (p.2) and said the notion that 90% of those killed by fighting in today’s wars are civilians is a "myth". Her goes on to present evidence of civilian and military casualties from particular wars that suggest a much lower ratio.
He concludes with a spirited call for better empirical data from actual wars and a caution that propagating the 9:1 ratio proposition - though intended to focus attention on protecting civilians - may have had at least three deleterious effects:
Firstly, it has not merely reflected, but also perpetuated, a misleadingly homogenised view of contemporary wars, when in reality each of them (and even each party to them) is unique in its character and in its consequences for civilians. Secondly, it has obscured significant achievements in civilian protection resulting from actions by states, international organisations and non-governmental organisations (NGOs). And thirdly, it has diverted attention from substantial issues to disputes about numbers and methodologies (pp. 128-129).
Adam Roberts (2010). Lives and Statistics: Are 90% of War Victims Civilians? Survival, 52 (3), 115-136 : 10.1080/00396338.2010.494880