Home > News > Militaries: An industry in decline
310 views 5 min 0 Comment

Militaries: An industry in decline

- August 10, 2013


Apropos of nothing in particular that’s in the news (except maybe this), here is a graph of how two measures of military effort have evolved from 1945 to 2007, by region.  (I’m working on a project that has gotten me mired in available data on military spending and force sizes, and I just thought this was interesting.)

The black line is the average across countries of military spending as a percentage of GDP, using the Correlates of War (COW) estimate of total spending divided by World Bank GDP figures (which only start in 1960).  The red line is the average across countries of armed forces per 1,000 population, again using COW estimates.

You see really striking long-run declines in the West, Eastern Europe and the former Soviet Union, and Asia.  In these areas it almost looks as if demobilization from World War II has taken place gradually and over 60+ years.  In Latin America and North Africa/Middle East,  you see pretty striking declines since the end of the Cold War, and perhaps some decline in subSaharan Africa since around 2000.

Why the long-run declines?  Many factors, surely, but on the international side it’s plausible to credit the disappearance of intense conflict among the militarily strongest states, which completely dominated international politics before 1946.  US-Soviet conflict was pretty intense into the mid-1960s, but since then the major powers have been less and less concerned about being invaded by each other.  I’d credit the nuclear revolution above all else, although there’s a lot of debate on this question and even without nukes there are probably other things that have been pushing in the same direction.  Such as, perhaps, democracy …

On the domestic side of things, there is pretty good evidence that the spread of democracy has been a significant factor.  Not worth getting into the details here, but if you look at the data country by country you find that on average, when countries transition to democracy their military spending and army sizes go down, quite substantially.*  In fact they tend to go down when they transition from very autocratic to only somewhat autocratic (that is, to “anocracies”, or semi-democracies using the Polity data).  The effect of a democratic transition on arms levels in the state in which the transition occurs looks to be larger than the effect of transitions in neighbors on a state’s own military spending, although this is hard to be sure about statistically due to endogeneity issues.  I would guess that most of the democracy effect is a domestic matter — for instance, autocracies want bigger militaries to help put down domestic opposition or to pay off cronies, or democracies want smaller militaries to lower coup threats — but some of it might also be an international effect.  That is, if democracies want smaller militaries then this could reduce the demand for big armies in their neighbors.

The graph also shows some interesting variation across regions.  E. Europe/FSU and N. Africa/Middle East stand out for high levels of military spending during the Cold War, though both now appear to be converging towards the rest of the world (except maybe for army sizes in the Middle East).

Update:  Mark in comments asked what the data for the US looks like, so at risk of the Wrath of Gelman I’ve added these to the graph for the West.  We spend and hire considerably more than other countries, both in absolute terms (which is well known, I think), and relative to GDP and population (maybe less so).  Note also the upward movement following 9/11, especially in military burden.

Cleaner pdf version here:  milburbyregion


*This is based on models with country and year fixed effects, so it’s probably not just that there is a coincidental global trend up in democracy and down in arms spending.  Benjamin Goldsmith reported the same pattern concerning democracy and the military spending in his 2003 JCR article “Bearing the Defense Burden,” (gated), looking at data from 1869-1989 (though he didn’t include time fixed effects in his model).