Phil Arena has been playing around with alternative measures of military power. He begins with the straightforward observation that one current and popular measure of military power, the CINC scores in the Correlates of War project, list the United States as having fallen behind the People’s Republic of China in its military capability. As Phil writes, this is not a conclusion that most, if any, observers of world politics would endorse–and that even if it is true in the broadest sense (that in some total war between the United States and the PRC, the United States might not be able to conquer China) that it is not particularly useful.
Phil lays out the broader points of his critique–namely, that CINC overweights raw materials and does not adjust for quality of militaries–in his post. Using a new measure based on COW data, computed as he describes at his link, he proposes an alternate measure–one I think that the Duck’s readership should at least be aware of. This measure may get closer to the notion of who has the most usable military power at any given point in time. I’ve redone his graphics slightly but the data is all his.
In the first chart, we see the post-World War II relationship among Roosevelt’s “four policemen.” This chart, I think, accords pretty well with our understanding of the period: the United States begins and ends the period as the world’s most powerful military force, but during the Cold War its military potential (in conventional terms) was on par with the Soviets. (I know this point could be debated, and has been, ad nauseam, but it is not prima facie invalid.)
The value of the alternate measure becomes a little clearer when we consider the period 1816 to today.
 Here, we see the long peace of the 19th century reflected in the gradual build-down of European militaries (although note Germany’s relative rise over the late-nineteenth century). We also see, as we would expect, that actual forces-in-being peak in the two world wars (and that the United States, in both cases, emerges as the world’s leading military power–although it dismantles its military quickly after 1919). More important, we note that the leading powers of the 21st century — the United States, China, and Russia — are quite literally not on the map in the 19th (and if Japan were ever to begin behaving as realism says it should, it would join the new quartet). Moreover, eras of multipolarity, bipolarity, and unipolarity are fairly easily identifiable in this chart.
The question of how to measure international capabilities is a tough one, but I tend to think that decomposing military strength and military potential is a useful start. (In the short run, we care a lot about strength; in the long run, we care a lot more about potential.) Since all such measures–even GDP!–are ultimately somewhat arbitrary, it is at least useful to have a debate about what we should include in each. Duck readers, what would you include in your measures of international power?
As rough quantitative indicators go, this looks like an improvement on CINC, “comprehensive national power” and others. One major problem that it shares with them, though, is that it seems to seriously underplay the significance of nuclear weapons.
I agree this is an improvement on CINC, for which we ought to thank Phil. One problem, though, is that it is based almost entirely on military expenditure (milex) — for which the underlying data are pretty awful. There is a pretty extensive literature showing just how bad the milex data are — i.e., incomparable across countries, and even across time within a country. At the risk of immodesty, let me point to my own contribution, published last year in JPR:Â
https://jpr.sagepub.com/content/48/4/547.abstract
There are systematic reasons why milex is worse than other measures,
like GDP for instance — countries have stronger incentives to lie about
their milex (and worse, they might exaggerate at some points in time
and then hide milex at other times). Even if states are trying to be honest, there are some big difficulties in getting to comparability, such as how different states report the health costs of troops (which can be a big component of milex), whether to use PPP or market currency exchange rates, and other issues that are NOT well addressed by standard sources like SIPRI and IISS.
There’s also the nuclear weapons issue that Dani Nedal points to. Still, with a little tweaking, Phil could do us all a favor by giving the profession a better “capabilities” measure than CINC …
Took a quick look at Professor Arena’s post, and I agree with Professor Colgan’s praise. Still, I would add to the issues raised above by pointing out that spending per unit soldier doesn’t mean much. I would imagine there are pretty significant disparities in terms of military capabilities per unit of currency spent. Moreover, other questions arise. Is that spending on equipment or on training? How effective is training in developing both an esprit de corps as well as combat effectiveness? How effective are combined arms (the US is pretty good here, the Chinese not so much)? How does the measure account for the outsourcing of significant portions of logistics by the US military (which would change the military spending per soldier)? Indeed, how does it account for logistics capabilities at all? Since I suspect Professor Arena and I look at IR and see two very different things, it probably doesn’t come as a surprise that I think materiality is only part of the story of military capabilities, and without measuring the other more ideational/social aspects any measure will be woefully incomplete…
Thanks everyone for the feedback.  You all raised excellent points. Sorry I didn’t reply sooner.
I acknowledge that the measure is hopelessly incomplete. Â That said, I’m less worried about it being incomplete than I am distorting (the two are not synonymous). Â Depending on what you want to do with the measurement, the possibility that military expenditures for any given country could well be off by 20% or more might or might not be an issue. Â That possibility is very much worth being aware of, and I thank Jeff for the link to his article, which I’d be sure to reference if I decide to follow up on this in the future. Â But I think the measure as currently constructed remains useful for giving us a very big picture view of the relative standing of the major powers. Â I’d be reluctant to read too much into any analysis where the results would differ markedly if one of the key variables had been mismeasured by 20%. Â That goes for this indicator as well as any other.
With respect to nuclear weapons, I’m not sure I agree with Dani that it’s “a major problem.” Â What I’d say is that it prevents the measure from being used to answer certain questions. Â Does the fact that information about nuclear arsenals is missing lead us to incorrectly view the US as being dominant over China at the dawn of the 21st century? Â If you start from the assumption that threats to use nuclear weapons in order to ensure national survival are credible, whereas threats to use nuclear weapons in pursuit of more limited objectives are not (which many would argue is the case, though I acknowledge there’s plenty of room for debate about the issue of credibility of nuclear threats) then this only means that my measure is more suitable for analyzing the relative ability to coerce states over issues of more limited salience than it is for discussing the relative ability of states to handle threats to their survival. Â I’m okay with that. Â In fact, that’s exactly what I was after.
As for the points Jarrod raises, I confess that there’s no adjustment for logistics or esprit de corps or other admittedly relevant issues. Â Again, the degree to which that would bother me depends on the task to which the data are being put. Â I can imagine situations where the answer to the question being asked would critically depend on the omission or inclusion of such information. Â I can also imagine many where it would not. Â Anyone looking to use these data should be aware of that distinction, and think carefully about it. Â But where there is cause to believe that the omission of such concerns would not distort the analysis, I think it can safely be ignored.
 “the degree to which that would bother me depends on the task to which the data are being put”
True. But then I suppose I would ask what is the purpose of the measure as you have developed it? To indicate military power? To what end? If it is to show who would win a fight, then you need those other things.
BTW, really liked the podcast you did with Dan. Bummer about the technical difficulties.
The primary thing I would be interested in using it for is to sort out, in broad strokes, the relative coercive ability of the major powers. Â You’re absolutely right that if you wanted precise point predictions about the likely outcome of a military contest, it would be beneficial to incorporate other information. Â However, people are already using CINC to construct proxies for the likely outcome of a conflict, and I’m not sure they’re better off with that than they would be with my measure. Â IF the only way to dramatically improve the measure would be to incorporate information that is not presently available across time and space, and IF ths measure were to provide a somewhat less inaccurate view than those based on CINC, I’d call it worthwhile. Â I confess I have yet to look into the matter deeply enough to determine whether either of those things are true though….
Glad you enjoyed the podcast. Â It was a lot of fun to record. Â We’re likely to do another one sometime. Â Hopefully with fewer technical difficulties. :)