Barbara Ericson’s 2015 AP CS demographics analysis: Still No African-Americans Taking the AP CS Exam in 9 States
November 9, 2015 at 7:28 am 8 comments
Normally, this is the time of the year when Barb writes her guest post about the AP CS exam-taker demographics. She did the analysis, and you can get the overview at this web page and the demographics details at this web page.
But before we got a chance to put together a blog post, Liana Heitin of EdWeek called her for an interview. They did a nice job summarizing the results (including interactive graphs) at the article linked below.
Some of the more interesting points (from Liana’s article):
No girls took the exam in Mississippi, Montana, or Wyoming. (Though Montana had no test-takers at all, male included, this year. Wyoming, which previously had no students take the test, had three boys take the exam in 2015).
Hawaii had the largest percentage of female test-takers, with 33 percent.
The overall female pass rate went up 3 percentage points, to 61 percent, from the year before.
Twenty-four girls took the test in Iowa, and 100 percent of them passed.”You don’t usually see 100 percent passing with numbers that big,” said Ericson. “Maybe five out of five pass. But 24 out of 24 is pretty cool.”
No African-American students took the exam in nine states: Idaho, Mississippi, Montana, New Hampshire, New Mexico, North Dakota, South Dakota, Utah, and Wyoming. That’s better than last year, though, when 13 states had no African-American test-takers.
Notably, Mississippi has the highest population of African-Americans—about half of the state’s high school graduates last year were black, according to the Western Interstate Commission for Higher Education. Yet of the five AP computer science test-takers, all were white or Asian and male.
Entry filed under: Uncategorized. Tags: APCS, BPC, computing for all, computing for everyone, NCWIT, public policy, women in computing.
Computing Ed Research - Guzdial's Take 
1.
mgozaydin | November 9, 2015 at 7:41 am
Why?
Very interesting statistics .
CS and engineering are good Money making professions .
It is sad the people coming from African American families are not attracted to Money making professions . WHY ???
2.
gasstationwithoutpumps | November 9, 2015 at 11:40 am
Rather bogus statistics—still emphasizing the results that aren’t statistically significant and glossing over the meaningful ones. For Mississippi, having 0 test takers in a group where you’d expect 2.5 is not significant underrepresentation—it is well within the variation you would expect by chance. The tiny number of test takers in Mississippi is a meaningful number, but the composition of that tiny group is not.
Having all 24 test takers in a group pass, when pass rates are 61%, is significant, even after multiple hypothesis correction for many states and many subgroups being examined.
I complained in previous years about the abuse of statistics (https://gasstationwithoutpumps.wordpress.com/2014/01/17/cs-commenters-need-to-learn-statistics/), but the message doesn’t seem to have gotten through.
I’d like to see an E-value on each of the factoids reported. How many such extreme under representation or over representations would you expect to see by chance among 50 states and the number of subgroups studied? I suspect that the biggest states will be the ones that show statistically significant racial disparity, not the states where there are only 5 test takers (where the race may indeed be a problem, but one that is not visible in these statistics, because no one is taking the test).
3.
Mark Guzdial | November 9, 2015 at 3:49 pm
I did hear you. As I said last year, we do try to consider the demographic statistics in the context of representation in the state. We didn’t compute the E-value in our 2012 SIGCSE paper because I’d never heard of an E-value until I read your article. I would be interested in computing the statistics you describe on the 2015 AP CS results, but I haven’t tried to figure out how yet.
As you say in your article, “zero” has power in terms of marketing, in terms of getting attention. Barbara starts getting queries about her analysis within days of the College Board’s release of the data. People use her analyses to inform policymakers about the status of computing education in their state. Barb puts in a lot of effort to put together her report each year — the College Board doesn’t make it easy. We could wait to release the results until after we do the tests for statistical significance, or we could release the basic tabulations as soon as we can and let people make the arguments that they will. Many of the results are useful even without tests for statistical significance. For example, at least one researcher in Pennsylvania has been using Barb’s analyses to figure out trends over time.
Even in what she has released so far, not everything that she wants to provide is there yet. It’s harder to compute the Hispanic numbers because of the categories that the College Board uses, so we still don’t have those results. Several states have asked us for the Hispanic numbers. She’s making available what she has as she completes it.
I did hear your message. It’s one of many messages that we receive. We are trying to provide useful information as soon as we can. If I can figure out how to compute the E-values too, and find the time to do it (the final revisions on my book on computing education research is due next Monday), I will.
4.
gasstationwithoutpumps | November 9, 2015 at 7:03 pm
Ah, lack of time is something I can relate to!
For figuring out E-values for under-representation calculations, you might want to look at the homework assignment my students are currently working on:
https://users.soe.ucsc.edu/~karplus/bme205/f15/Palindrome.html
(For the questions Barbara is addressing, it might be better to use the binomial distribution explicitly, rather than the normal approximation I have them use.)
5.
gasstationwithoutpumps | November 9, 2015 at 11:47 am
mgozaydin, your statement “It is sad the people coming from African American families are not attracted to Money making professions . WHY ???” makes the assumption that students have free choice about whether to study AP CS in high school or not. In fact, students living in impoverished school districts (as many African-American and Hispanic students do) often do not have the opportunity to take CS in high school, nor do they have the financial resources to study it on their own. CS is primarily offered to rich white and Asian students, because of the geographic distribution of wealth in school districts, and the perception that AP CS is a luxury course offering.
6.
gflint | November 9, 2015 at 4:45 pm
“No African-American students took the exam in nine states: Idaho, Mississippi, Montana, “. No one in Montana took the APCS test so the statement is correct. I am not aware of any school offering APCS in Montana. The one school that did no longer offers it. Looking at APCS seems to be the only measurement available for looking at the growth or decline of CS. Too bad. Montana does not do APCS because we offer dual-credit CS courses. Our dual-credit courses are growing rapidly wherever we can find teachers qualified to teach them..
7. Top Ed-Tech Trends of 2015: 'The Employability Narrative' | Co-Opt-Ed | December 8, 2015 at 8:43 am
[…] athletes from the least diverse of the big four professional sports). Female students and students of color remain underrepresented in computer science majors and classes, and many simply don’t see […]
8.
AP Computer Science Demographics Report for 2015 completed #CSEdWeek | Computing Education Blog | December 11, 2015 at 7:11 am
[…] see link here. We don’t yet have the statistical tests that Kevin Karplus asked for (see post here), but Barbara did list the percentage of Hispanic exam takers with their proportion of the […]