On Diversity as Strength

Reminder: LTC Heffington’s letter  and  Superintendent’s responses:

The Academy leadership says, in response to LTC Heffington’s concerns about a degraded Academy, of the cadets:

They are the most diverse group in the history of West Point and we are stronger for it.

This is something of a non-sequitur to the issues of conduct and performance in Heffington’s letter, and the unsubstantiated assertion (“we are stronger for it”) is out of place for an Academy is producing officers who, we are often reminded, will be in combat, leading other soldiers, and who are responsible for winning wars / crushing our enemies. But if the Supe brings it up, we’ll address it. The first priority ought to be combat effectiveness. If diversity contributes directly to combat effectiveness, then USMA is right on. But this should not be assumed a priori.

And if it is not, then the situation certainly fits with LTC Heffington’s assessment that:

…a culture of extreme permissiveness has invaded the Military Academy, and there seems to be no end to it. Moreover, this is not unintentional; it is a deliberate action that is being taken by the Academy’s senior leadership, though they refuse to acknowledge or explain it.

But what does it mean by “Stronger”? What exactly does Diversity mean in this context? In most bureaucratic environments, diversity refers to race/sex population percentages that fit with some model of “how things should be.” But surely an Army general wouldn’t pursue arbitrary class compositions of race and gender, for no discernible reason other than to have an arbitrary class composition of race and gender, and especially if that impacted performance, would he? Ideally, the performance of the whole group should be high, regardless of race or sex.

After all, the response letter says:

This commitment to excellence must permeate everything we do. I will not compromise my decision to advocate winning in accordance with our values of duty, honor and country. It is what America expects of our Army and of its leaders. The crucible of ground combat is unforgiving, and we owe it to our cadets to give them every opportunity to learn how to succeed.

Yet USMA has been trying to up its Diversity recruiting, saying

The West Point Center for Science, Technology, Engineering Math (STEM) Education, West Point Center for Leadership and Diversity in STEM, Urban Leadership Initiative, Project Outreach, Diversity Leadership Conference, SAT/ACT Score Improvement program… are some of our most effective instruments in achieving a lasting impact in our communities. Their work engages community influencers and students from diverse backgrounds to provide a unique West Point experience that generates awareness, cultivates interest, and most importantly, builds relationship.

By working together, we will expand our community outreach programs increasing West Point’s reach into under-served communities, enabling us to gain access to high quality candidates with diverse backgrounds. The long-term benefits extend well beyond West Point recruiting, creating valuable relationships to help demystify the military and inspire service to the Nation.

It is not exactly clear what this all means in the context of “commitment to excellence must permeate everything we do,” since, by every observable measure, diversity at West Point is measured in race and sex, and has definitely been changing. But perhaps the Sup really is tapping into hitherto un-realized “high quality candidates.” Let us see.

Class Composition

Below is a table of class composition by race since 2000.

Following paragraph updated April ’22 with duplicate records removed.

We see that the Corps has indeed become more diverse. In fact there has been a tremendous restructuring of the Corps’s demographics.

The change highlights include:

• White attendance has dropped from 79% to 64%. Assuming consistent self-identification from cadets, this is a 19% drop.
• The increase of Hispanic cadets as percentage of the attended class (6%->9%, or +50%; again assuming consistent self-identification by cadets).
• The increase of black students of 8% in 2000 to a low of 6% in 2007 to 14% in 2020, or a net 75% increase
• The increase in Asian attendance steadily from 5% to 9%, or an 80% increase

Turning for a moment to sex:

The composition of classes by sex has remained largely constant with women at around 15-16% for most of the years. We see this go up in 2018-2020 by 18%+ vs prior years. Perhaps it is the start of a trend.

Demographic Goals

I have heard, but have not seen a document (source suggestions are welcome), that the Academy tries to mirror the demographic composition of the Army. This composition is shown in a 2017 document from the Army G1:

So the Academy is tracking fairly well to that somewhat arbitrary (after all, we have a volunteer army) goal of race/sex diversity, and is correct that the most recent classes are the most diverse (as measured by demographic population spread) in history.

However, as West Point says, (emphasis added)

There are no appointments, vacancies, or nominations designed exclusively for minority groups. However, cultural and socio-economic backgrounds are given appropriate consideration while evaluating all applicants. West Point makes a concerted effort to inform minorities of admissions opportunities, reaching many that might not otherwise apply for admission. Today, minorities represent about 15-20 percent of the Corps of Cadets.

This indicates that the official line is that USMA gives minorities no special advantages in admission – that there are no reserved slots, and that the Academy tries to inform candidates of opportunities. This is found to be untrue here and by other assessments.

For example, we see in the 2017 Board of Visitors presentation (p 59) that there are very definite goals for racial class composition: 

It’s clear that there are goals for “diversity” purposes. If the goal were input-based, such as number of majority-minority schools canvassed, West Point’s claim that it makes a special effort to “inform” minorities would carry more weight. But its goal is output-based, measuring minority admissions.

We conclude that West Point is being dishonest, or at very least disingenuous, in claiming “no special vacancies” exclusively for minority groups at the same time its admissions success is measured in goals by racial group admittees. 

The Cadet Honor Code notwithstanding, perhaps that’s forgivable if there are benefits to diversity-advantaged admissions. This brings us to the second part of the Academy official’s claim about diversity: “… we are stronger for it.”  How can we verify or disprove that claim?

We turn again to the Academy’s own measures of performance for evaluation, including admissions rates, CQPA, and graduation rates.

Admissions:

First, in the pursuit of Army-defined “diversity,” let’s be frank:  there are different standards in recruiting for the different races and sexes.

We use here SAT scores as a proxy for overall strength of the candidate. While this is of course not the whole picture, it is a highly correlated indicator of performance as we saw in the “On Admissions Excellence” post. We look at classes ’10-’20 arbitrarily for a multi-year average of scores and attendee percentages. A year-by-year breakdown of admissions percentages is here and SAT profile distributions by race are here. We of course remind everyone that these are aggregate group profiles, and not reflective of any individual.

From a review of the chart – although not a detailed statistical review, which is beyond the scope of this post – we note that women are at a ~3-4% disadvantage for admissions probability compared to men, based on similar SAT scores.

We note that Asians are admitted at lower percentage rates than Whites of comparable test scores; Asians need at least ~150pts of additional SAT to reach comparable White admissions rates.

With a test standard deviation (SD) of ~200 points, we see that Blacks are admitted at about .33 SD (or 66 pts) lower scores than Whites and 1 SD lower than Asians. And they are admitted at close to 20% rates at those lower scores, compared to White and Asian admission rates of 11%. Hispanics are admitted at slightly higher scores 20-30 pts) but at much higher percentages than Whites (18% vs 11% for men, or ~75% more for the same scores).

This difference in standards–lower scores for higher admissions rates–shows that the Academy intentionally seeks diversity in race and alters its admissions standards to achieve them. It would be very improbable for those results to be achieved accidentally and consistently, or even through race-neutral recruiting, though if the Academy is using those, we welcome explanation of the techniques. The question is whether West Point has evidence to show that this makes the Corps (and Army) stronger. Our data indicate otherwise.

Grades

In CQPA, there are observable differences in group performance, particularly race. To illuminate this we can turn to the data. The below table shows average CQPAs by race for 2010-17. We do not include ’18, ’19, or ’20 because the classes had not graduated, and therefore CQPAs were not complete; but 2010-2017 is still a large sample size. We also do not filter by status code, to capture all cadet statuses including turnbacks and similar.

CQPA Standard Deviation is around .5-.6 CQPA points. The CQPA trend – recalling again that this is a weighted measure of Academic, Physical, and Military program scoring – reflects the SAT predictions. It is the Academy’s metric for student performance and quality while at USMA, and is used to determine class rank, Army branch options, and post. So the measurement of achievement is important to USMA.

We see that Blacks and Native Americans are low performers, Asians and Whites are toward the top, and Hispanic cadets are about halfway somewhere between. B/N CQPAs are .5 points, or about a full standard deviation+, lower than W/A performance. CQPA performance at USMA impacts cadets’ odds of graduating, their options coming out of school, and, if the metric is any good, predicts their performance in the Army. So to see large discrepancies is concerning.

CQPA differences by Sex are not so pronounced. Over the same time period:

Indicating that females have generally the same performance expectations as men at USMA.

Separations

What about separation rates? We look at the number and reasons for separations. Ideally, we would want low separation / high graduation rates; after all, every cadet takes someone else’s slot–potentially someone who is more qualified–and we want those who are admitted to stay the course and become Army officers. High attrition may make for bragging rights (“It’s tough to get through”) but is bad from a resource management point of view.

And in a summary table of separation categories for classes 2010-17 :

The separation rates are not uniform by race. For instance, Black cadets separate at about a one-third rate, while Asian cadets separate at less than a one-fifth rate. The biggest differences in separation rates are seen in Academic-related separations.

So for every 10 Black cadets admitted, we would only need to admit 8 Asians or White cadets to get the same graduating yield (think of the recruiting cost savings!), and probably with better CQPAs.

Therefore, we know that admissions are influenced by race, and therefore we know that USMA is deliberately accepting higher failure rates for some groups over others for the sake of diversity (or, if we’re cynical, for sports).

What we don’t see is evidence for LTC Heffington’s assertion that separation rates should be even higher, that the class yields of 75-80% should be even lower because standards are too low. This is not something we can verify from the data, but we can say that USMA is deliberately bringing people likely to fail into USMA in order to satisfy de facto quotas.

Graduations

The next set of metrics is around graduations. It’s unclear why and where this varies exactly from separation reasons, but there is some slight variance in the data. The number of non-graduated statuses does not match the number of separation reasons. Perhaps this may be due to turnbacks or reasons for not graduating that may not be termed “separations.” So, separations rates plus Graduation Rates do not add up to a 100% due to turnbacks, remedial requirements, or other non-graduated statuses. The resulting differences are minor at <3% on average and do not impact research conclusions. Regardless, graduation rates by race are similar to that indicated by the separations data, again looking at class years 2010-17.

There is clear difference in graduation expectancy for different race groups but we see less of a difference by sex. This would be predicted, again, by the CQPA distributions and SAT scores.

It doesn’t have to be this way; as shown in the Admissions post, there are many qualified individuals who could be admitted with better outcomes. This doesn’t help the cadets who are admitted with high chances of failure; the academics and lifestyle are very rigorous, and it’s a disservice, if not outright immoral, to put people (especially 18-19 year old kids!) in positions where they are likely to fail.

Conclusion

In summary, we see that:

• USMA is deliberately pursuing class composition goals for minorities by, in addition to minority recruitment, employing race-based admissions strategies.
• Data indicate that the race-based practices include, where Admissions has complete discretion in selecting which candidates will be offered admission, extending preferences to minority candidates who, on average, have significantly lower SAT scores than non-minority candidates who have higher SAT scores.
• USMA’s diversity admissions goals are explicitly identified as such in documents like the Board of Visitors report.
• To achieve what are called its “class composition goals” for minorities, and despite Academy officials and others saying otherwise, USMA is admitting marginally qualified candidates with a known and higher probability of failure.
• The marginally qualified candidates are failing by the Academy’s own metrics, and more marginal performers are graduating into the Army.
• The marginally qualified candidates consume resources to try to get to graduation, and take up slots that other, available, better qualified individuals could have filled.
• The failures take up years of kids’ lives that could have been more productive had they gone somewhere that was a better fit.
• There are real and tangible costs to pursuing the diversity strategy, and no supporting quantifiable argument offered for pursuing the strategy.

I leave it to the reader (or, better, USMA leadership) to explain how this policy of encouraging academic mismatch and deliberately not accepting better qualified candidates “makes us stronger” or how it is a “commitment to excellence… [permeating] everything we do.”

Put another way: USMA (and by extension the officials in charge) is deliberately seeking and tolerating low performance and accepting high failure rates that hurt cadets to meet arbitrary student body composition goals by skin color. The claim of diversity leading to better outcomes notwithstanding, when the evidence is examined critically, it demonstrates the opposite: worse outcomes. It does, however, make for better optics, celebrated in slides like this:

The Academy likes to show off the results of its racial preferences but not how it is achieving them. Perhaps this pursuit of non-outcome-related optics is influencing LTC Heffington’s concerns about the Academy’s focus. The Academy ought to explain what it is doing in pursuing a class composition strategy solely based on a social engineering concept instead of on the effect it will have on our nation’s Army.

Thoughtful criticism and factual corrections are welcome.