I wrote an article a few days ago published here that strongly criticized political intolerance and suggested alternative peaceful coexistence. Right after I published the article, I came across a tweet from Dr. Sommers that almost made me take down my previous article in fear of sounding like a hypocrite. At face value, Dr. Sommers seems genuinely concerned about college enrollment rates for 18–to 24-year-old white men. However, the tweet was counterintuitive to the preconceived notion of the white man at the top of the patriarchal society. The tweet also didn’t bother pointing out white women were enrolled at much higher rates than any of the other demographics. So, my data science instincts immediately kicked in and started firing questions like what is the historical rate of college enrollment for white males? Why didn’t the author include the enrollment percentage for Black and Hispanic men? Fortunately, the author reveals the Department of Education as her source. I encourage you to take a closer look at the source for yourself because I did.
Before I lay out my arguments for and against the seemingly benign tweet, let me give you some background information on the author. The author of the tweet is Dr. Christina Sommers, a former philosophy professor at Clark University who is now employed by the conservative think tank American Enterprise Institute. She is the author of a book titled “The War Against Boys”. If you don’t have time to read the book, check out her article on The Atlantic that summarizes the main points. There are several data analysis tactics Dr. Sommers used to spread misinformation and I will break down the tactics below. You may follow along with the simple analysis I did here.
1. Presenting Too Many Variables To Misrepresent Data
As I mentioned earlier, it seemed odd that men and women statistics were shown on the same plot in addition to racial variability. Why does that matter? It matters because there are two variables on the graph: race and sex. This leads to unreliable results due to a confounding variable which is a fancy way of saying you didn’t account for an extra factor. It also seemed strange that the author picked a single year to do her comparison. So, how about we untangle the race/sex variables while we look at the historical differences among the races and sexes?
1a. Enrollment Differences among the Races
When removing the gender variable, both white men and women have a higher enrollment rate than black and Hispanic men/women. To my surprise, the differences among men have been widening since the 1970s, the post-Vietnam war generation. As we move past the economic stagnation of the 70s recession, the enrollment gap between white women (blue) and the rest widens significantly in the years 1978–2008. This clearly shows why people of color are at a disadvantage when it comes to college enrollment. After 2008, the gap in enrollment between white men or women vs people of color narrowed appreciably which is a cause for celebration. It should be obvious that the trend for all the races is going up which is another cause for celebration.
A few interesting questions I asked myself while staring at the graph were: how does the college completion rate look among the different races? What is the driving factor for the dropout rates? What is the return in college education among the races?
1b. Enrollment Differences among the Sexes
Lastly, let’s compare the historical enrollment among the sexes, and this may be the main concern of Dr. Christina Sommers. I encourage you to check out her channel on youtube: The Factual Feminist.
The data shows women’s college enrollment has been trending up at a higher rate than men since 1987. According to research, some of the stated reasons are an increase in high school performance by young women, policies, and changes to college entrance requirements favoring females and the rise of unaddressed problematic behaviors in males. I think the college enrollment differences among the sexes is a cause for concern because research on understanding the discrepancies seem to be lacking in depth.
2. Omitting Data to Tell a Story that Fits a Narrative
The graph she posted is a truncated graph with a shifted baseline of 30% which I recreated using the data from the Department of Education. The decision to use a truncated graph may seem innocent to untrained eyes, however, the purpose is to exaggerate group differences and skew the graph to fit her narrative. In addition, the author purposefully didn’t present the data in its entirety to fit her agenda. Do you see why she omitted black and Hispanic men from her plot on the tweet? The picture looks really bad with those demographics included. However, the intentional omission glaringly shows what she is most concerned with: white young men. If she cared enough about American college men in general, she would have shown the complete picture.
In lieu of representing the data accurately, look at the graph with a baseline of 0% and all the demographics included. The differences don’t seem as stark as the graph Dr. Sommers showed despite using the same exact data. Contrary to the point she is making, Black and Hispanic men are disproportionately enrolled at a lower percentage than any other demographics.
I really hope that I have demonstrated how Dr. Sommers intentionally misleads audiences by misrepresenting the data. In the information age, we need to be on the lookout for misinformation based on data misrepresentation and fight the spread of fake news on social media. Unfortunately, educated people like Dr. Sommers could sometimes be part of the problem instead of being the gatekeepers of unbiased science and research. However, you and I could arm ourselves against misleading data representations with simple analytical skills to stop these bad actors from spreading false narratives.