January 11, 2001
War casualties may fall by 30-50% or more per decade for the next few decades, because there will be fewer young men relative to old men in the world.
It does seem that wars are more common between neighboring countries than distant ones, that more powerful nations are more likely to start wars, that wars rarely occur between nations where one is much more powerful than the other, and that wars are more likley within nations with greater ethnic clevages . Wars may also occur when a rising power is disatisfied with the status quo and the falling power has many allies . This does not seem like a lot to go on, however, when wondering whether wars will become more or less frequent in coming decades. Policy advice abhors a vacuum, however, so we still see forecasts based on the weak correlations that have been observed, such as predicting fewer wars because there are more democracies, or more wars because of the end of the cold war.
Recently however we have seen a strong apparently causal relation between demography and war. And this relation allows us to make stronger forecasts.
This next diagram is from their 1999 paper  and describes civil war combat deaths among 153 countries from 1989 to 1998.
While sometimes frustratingly sloopy in their presentation, Mesquida and Weiner have nevertheless collected an impressive variety of data sources showing similar results. A similar correlation appears in war deaths among 30 nations from 1981 to 1993, in a sample of 15 island nations, in combat deaths among 25 countries from 1960 to 1969, in 15 republics of the former Soviet Union from 1989 to 1993, and in 12 tribal societies. The correlation remains strong after controlling (individually) for average wealth, wealth inequality, democracy scores, and population density. A similar pattern appears in crime statistics.
It seems natural to interpret this correlation as demographics causing war. After all, demographics change slowly while wars seem to develop quickly, and few wars kill enough people to substantially change demographics. It also seems plausible to interpret this in evolutionary psychology terms, as an evolved vestige tendency among young men, when not restrained by old men, to seek mates and the resources that attract mates via all means possible.
Although Mesquida and Weiner do not offer a clear estimate, a linear approximation to the above graph suggests that median war casualties fall by roughly a factor of two to three when a nation's MAR falls by ten units. (Alternatively one might interpret it as a discrete threshold, where say median war casualties fall by a factor of a hundred when MAR falls below a threshold somewhere near 70.)
Here is a graph describing how world average MAR is forecasted to fall, and then histograms describing the world distribution of MAR from 1980 to 2020 in more detail.
(National MAR is rounded here to the nearest ten.) These histograms show that MAR is expected to fall relatively consistently across the entire distribution, though a small tail seems to remain at high MAR. A detailed table of MAR by year and country is also available.
More generally, if data comparing countries at the same time can inform us about how countries change over time, then falling MARs suggest a dramatic and hopeful prediction: median war casualties around the world should fall by roughly a factor of two or more per decade from 1990 to 2020! And they should have fallen by roughly the square root of that factor (about 30%) from 1980 to 1990. These predictions stand in sharp contrast to predictions of increasing war deaths due to the end of the stabilizing influence of the cold war .
If the entire distribution fell together as a unit, average war casualties would also fall by the same factor. The upper MAR tail should dominate average casualties, however, and that tail seems more resistant to decline. So we might scale back our prediction, perhaps to a 30% decline per decade. Nevertheless, the overall picture is very hopeful. Of course if deaths given MAR are really lognormally distributed, as they seem, then average war casualties could still be highly variable and hence hard to predict.
Deaths across the world over the last century due to wars, massacres and atrocities have been graphed by Matthew White . He says that world per capita deaths were 2.3% from 1900-1925, 4.2% from 1925-1950, 1.6% from 1950-1975, and 0.3% from 1975-2000. One might interpret this as showing that war deaths were relatively trendless until about 1980, when MAR started to fall substantially. (Better data on world MAR and war casualties over the last century might help us to better test this theory.)
On the other hand, William Eckhardt has estimated that the twentieth century was by far the most deadly in the last fifty centuries in terms of per capita war deaths world-wide . And average MAR has almost surely been very high through most of those centuries. Since the world was very poor in previous centuries, perhaps both wealth and youth make war more likely.
If war casualties really do fall by large factors over the next few decades because of falling ratios of young men to old men, the world economy should be more peaceful and prosperous, national boundaries should not change substantially, and forms of government will mostly change peacefully or not at all. Military expenditures might also be safely reduced (though that doesn't mean they will be). This may be disappointing new for those who hope for revolution, but be good news for those who hope for peace. If it does happen, however, those who run the new world order will likely credit their wise leadership.
 Woosang Kim, James Morrow, "When Do Power Shifts Lead to War?", American Journal of Political Science, 36(4):896-922, Nov. 1992.
 Christian G. Mesquida and Neil I. Weiner, "Human Collective Aggression: A Behavioral Ecological Perspective", Ethology and Sociobiology 17:247-262, 1996.
 Christian G. Mesquida and Neil I. Weiner, "Male Age Composition and Severity of Conflicts", Politics and the Life Sciences, 18(2):113-17, September 1999.
 U.S. Census Bureau, Online Demographic Aggregation, http://www.census.gov/ipc/www/idbagg.html, 2000.
 CIA, Global Trends 2015, December 2000.
 John J. Mearsheimer, "Why We Will Soon Miss The Cold War", Atlantic Monthly, July 1990.
 Matthew White, Wars, Massacres and Atrocities of the Twentieth Century, http://users.erols.com/mwhite28/war-1900.htm, 1999.
 William Eckhardt, War-related Deaths Since 3000 B.C., Bulletin of Peace Proposals 22(4):437-443, 1991.
Thanks to Chris Hibbert and Hal Finney for comments on this.