In our recent post “Call for Sunlight…”, we outlined which groups went into which Army branches to put context around USMA’s stated need for preferences in admissions. Finding that certain minority groups were less represented in non-combat arms than in combat arms, we also speculated that one driver of the low number of minorities in combat arms was cadet branching preferences. In this post, we find validation of that assertion. We do this by analyzing the West Point classes of 2013-2017 branching data by CQPA and race.
Briefly, we find:
- There is a clear hierarchy of branch rankings by CQPA, indicating branch popularity.
- However, both combat arms and non-combat arms branches are available to wide ranges of cadet achievement.
- Due to the wide ranges of cadet achievement present in each branch, the numbers of slots available for each branch, and the “branch CQPA” ranking, we conclude that there are no structural barriers preventing cadets from going into either combat arms or non-combat-arms if they prefer to.
- From this we conclude that clustering of certain groups in combat or non-combat arms is a function of the preference of those groups.
For this analysis we used the classes of 2013-2017 to have the most recently graduated 5-class cohort available in our dataset. We explore which branches tend to have the highest performing cadets, the spread of talent within those branches, and the racial composition of the branches.
We use the five classes of 2013-2017 because the USMA-provided dataset had class data from 2000-2020, but graduated classes only through 2017. We filter out known duplicate records in our analysis and rely on the branch coding field in the tables. Note that there are likely some errors in the data that we have not identified which may prevent us from arriving at the precise numbers shared by USMA in other settings. However, the size of the dataset used (5 classes) ought to allow us to arrive at directionally correct findings for our purposes.
Branch codes include basic branch codes and do not (at least in examples we checked) indicate a Branch Detail branch. So if a cadet were to branch Military Intelligence and be detailed Infantry, the code would show MI.
Further, we exclude records that did not have Branch codes, indicating that they did not graduate, or were turnbacks, or had other issues. These were a significant number of each class, roughly corresponding to the number of separations we observed. But we did not analyze records with missing branch codes in detail and excluded them here from conclusions. However, this likely creates some selection effect when we look at CQPAs and test scores.
Last, we do not analyze branches by test scores such as the SAT. We used CQPA because it is the holistic measure of performance used by West Point. While there is correlation between test scores and CQPA, we use only CQPA here as the measure of performance at USMA and to determine eligibility for a branch selection. The question of which branches are “smartest” closely but not exactly tracks the CQPA rankings, due to the inclusion of the military and physical grade averages.
USMA’s files included branch codes presumably for the cadets who made it through the branch selection process. This was a while ago for us, but we recap from the best of our recollection. If a reader has an updated or more accurate process, please let us know and we’ll update this.
Branching happens in the latter part of the cadets’ senior year. Each branch is allotted a certain number of slots to fill. These slots are tied to available locations. So, for example, there may be 10 Infantry slots for Fort Bragg, NC with the 82nd Airborne. Cadets are allowed to put in their ranked branch and location preferences in an online tool. The tool takes the ranked preferences and matches cadets with branch/location openings by taking the the class rank of the cadet; searching the cadet’s preferences in order for an open slot; and assigning the cadet to that branch. That slot is then removed from the available selection pool for the next cadet on the ranked list. Then on branch night, the cadets find out what branches they’ve been assigned to.
So we see that there are several elements of this to consider: First, the cadet’s class rank, determined by CQPA. Second, the availability of slots for a given branch, and for a branch at a given location, dictate what opportunities a cadet at a given class position has.
Next, we keep in mind that the selection process is mostly a black box: cadets don’t see what the choices of their peers are while they are selecting branches and only receive the outcome of the matching algorithm. Of course they can coordinate outside of the selection tool (“Hey Bill, let’s both select Bragg and Infantry”), but this becomes difficult for large groups given the number of cadets and the number of branch/location options available. However, it is known that some branches are more popular from year to year influenced by current sentiment in the Corps, the Army, the news cycle, popular instructors, movies, or what-have-you.
CQPA by Branch
The Cumulative Quality Point Average (CQPA) is comprised of weighted Academic, Physical, and Military grade point averages. It determines the cadets’ class ranks and therefore their selection options.
Below, we compile the CQPA averages by branch and rank order them in descending order.
The branches are probably familiar to the readers, but we’ll refresh the codes used: MS: Medical Service; AV: Aviation; MI: Military intelligence; CY: Cyber Corps; FI: Finance; EN: Engineers; QM: Quartermaster; IN: Infantry; AR: Armor; AG: Adjutant General; AD: Air Defense; SC: Signal Corps; OD: Ordnance; TC: Transportation Corps; MP: Military Police; CM: Chemical Corps; FA: Field Artillery.
We see a clear hierarchy of branches. We see that MS and AV are on average the highest scoring branches and Field Artillery, the least. What’s interesting is that the first 5-10 branches have roughly equivalent max CQPAs, though the minimums keep dropping significantly. Various non-combat-arms seem to cluster in the 2.87-2.95 CQPA range. We see the scoring bands tighten up the further down we go.
The main takeaway here is that every branch has a significant CQPA range for entry. Those branches at Quartermaster and below have basically the graduating minimum CQPA as eligible CQPAs. Those branches include both combat (IN, FA, AR, AD) and non-combat arms (AG, MP, OD, etc) . So CQPA is not a barrier for a cadet wanting to go into either category.
Branch CQPA Density
We turn to another question – how popular is each of the branches? We gauge this by assuming that, as a group, all cadets would select the same branches is they had opportunity to do so. As a result, the most-qualified cadets (highest CQPA) would tend to go to the popular branches first, raising the CQPA of that branch. The least popular branches would have the lowest CQPAs.
To visualize this we graph the branches in descending order by average CQPA, with a bubble size representing the number of cadets going into the particular branch.
As we showed earlier, there is a distribution of ability in each branch, so not everyone in FA is dumb, but more “lower-skilled” cadets tend to go FA. Assuming rough homogeneity of branch preferences among the cadets, this means that the branches on the left are most popular because the cadets with the highest CQPAs and most options pick them, and branches on the right are least popular. Medical Service may be the exception as it has a relatively small number of slots, and cadets must have a course of study that supports the branch choice and subsequent schooling, so they have to be working for it from very early in their four years.
One key observation here is the large size of the bubbles for Infantry and Field Artillery. There are lots of cadets in each, and they are at very different spots in the CQPA curve. The mid-range non-combat arms (AG, AD, SC, OD, TC, MP) together roughly equal the number of cadets in the Infantry bubble. Again, this means anyone who wants to go into a combat arm or non-combat arm can indeed do so. They may not get their first posting preference, but they can certainly choose whether to be at the tip of the spear or not.
Branch Racial Composition
We now turn to who goes into these branches. Tables are below showing the percentages of:
- Table 1: Branch breakdowns by race (top left table, read across the rows)
- Table 2: Race breakdowns by branch (top right table, read down the columns)
- Table 3: Combat Arms (CA)/ Non-combat arms (NCA) breakdowns by race (bottom left table, read across rows). Here we note that there are Combat Arms (e.g. Infantry), Combat Support branches (Chem, Intelligence, Military Police), and Combat Service Support (e.g. Transportation, Finance) branches. We are mostly interested here in the cadets’ demonstrated preferences between combat-arms and non-combat arms.
- Table 4: Race breakdown by CA / NCA (bottom right, read down columns)
A chart of Table 1 makes visual the branch breakouts:
And when we consolidate branches into the categories of “Combat Arms” or “Non-Combat Arms”, we find that racial groups have different branch-type rates:
We see that White and Hispanic cadets tend to go overwhelmingly (69% and 71% respectively) into Combat Arms. “N” cadets, being a little less than 1% of the overall population, tend to do so as well but we consider it a small sample size. The Asian and Black cadets have a distinctly larger preference to go into NCA branches, with CA participation at 53% and 62% respectively. So White and Hispanic cadets tend to go into CA branches at about 1.14x the rate of blacks and 1.32x that of Asians.
Why these groups prefer to go into different broad categories of branches is far beyond the scope of this blog. But we note the persistent difference in context with the opportunity to select among the CA / NCA branches. We conclude that the difference is indeed a result of group preference, even after accounting for the different performance levels exhibited by each group.
In an earlier post we noted that the Army considers the lack of minority (and specifically Black) representation in the higher officer ranks a problem. Such worry is demonstrated in pieces like this USA Today article with quotes like:
Not enough, some argue. Consider the new commanders of the Army’s operational brigades, including front-line units such as infantry, artillery and armor: There are 96 such brigades of about 4,000 soldiers each led by a colonel. Two of the incoming commanders of those units are Black.
and the New York Times.
African Americans are highly represented in the military, but almost invisible at the top.
Leaving aside topics such as why other minority groups (for example Asians) are not the subject of such worry, we posited that the voluntary preferences of groups to join combat arms will impact eventual leadership representation.
We found this to be the case. If cadets of all groups joined combat arms at equal rates, then, all else being equal, we would expect leadership representation to mirror the demographics of the graduating class. But equal branch preferences are not what happens.
For example, to build the tables above, we used raw numbers of branched cadets:
The overall ratio of White to Black cadets is (4274 / 537) = 7.96. So one might think that senior leadership composition should reflect that. But after we take preferences into account for combat arms, and assume that all General Officers come from combat arms (not true, but a simplifying assumption), we find:
(4274 White * .71 CA preference ) / (537 Black * .62 CA preference) = 9.11.
At a minimum then we would expect a 9x difference rather than a 8x. And this of course excludes any discussion of validity of CQPA on careers, retention, etc; it is only to indicate the effect of the preferences on composition goals.
The branch composition breakdown is a useful tool to check whether admissions preferences and lowered standards are achieving ends that support the stated goals. We find that groups prefer to branch into combat arms at different rates. Because the branch choice affects opportunities for promotion, this is a confounding factor to West Point’s admissions goals and practices.
As always, thoughtful criticism and factual corrections are welcome.