According to report by the Human Security Centre of the University of British Columbia, most forms of political violence are in decline in the world since 1990. Among its findings are that "the number of genocides and violent conflicts dropped rapidly in the wake of the Cold War," and that "wars are not only far less frequent today, but are also far less deadly." Deadly to whom? The report says that "90% of those killed in today’s wars are civilians, and that women are disproportionately victimized by armed conflict." This makes it a bit unclear why these developments are characterized as an "improvement in global security." The Centre's press release (linked above) describes the results as "surprising" and as "confounding conventional wisdom." I would have to agree that it is surprising and confounding, but this may well be because it will provoke criticism of its findings and especially its methodology. I have not yet read the entire report (available in thirteen PDF files) but strongly suspect that many of the results are driven by how the researchers measured (Did India really have 156 "conflict years" between 1946 and 2003? We would have to know what the category is and whether it makes sense.). It doesn't help that a section of the promotional web page titled "data sources" contains only the text "Coming soon." Anyone who is interested enough to read the whole thing, please share your assessment in the comments.
Update: For an alternative representation of the frequency of genocide, try this table prepared by Gregory Stanton's Genocide Watch. For my part, it's hard to avoid the standard methodological objection that all ethnographers have: lots of cases means a poverty of empirical detail.