The Business of College Majors is Business

I looked at the distribution of college majors recently, with this question in mind: does the ordinary experience of most undergraduates reflect the left-leaning slant of American college faculty? Do students, on average, encounter faculty who are more conservative than the typical faculty member?

The question arises from the suspicion that students can be isolated in conservative ideological bubbles, hiding in plain sight within a larger institutional context that skews left.  Scratching around that question, I found an enormous amount of longitudinal data on college majors.

You can find analytical support for a lot of important questions in these data. A good source can be found here, and some brilliant data visualizations from Prof. Benjamin Schmidt at Northeastern University. In my simple graph, I want to make one point: most U.S. undergraduates choose a major with a vocational emphasis, and these kinds of majors constitute the entirety of the increase of undergraduates since 1971.

I start with the list of 33 majors that appear on the National Center for Education Statistics report summarizing bachelor’s degrees since 1971. The first thing to notice is the popularity of business degrees, which has held around 22% since the 1980’s. Also notice the category “Health professions and related programs”, which has grown from 3% to 9%.

I lump these categories into a group “occupations”. I also included smaller categories such as communication and journalism, law enforcement, and visual and performing arts. Not included are science and engineering, which seems strange at first because we tend to think their job potential exceeds everything on this list. We could include them without diminishing the effect visible in the graphs. The other groups are: liberal arts, science, social science, and other.

Here are three time series plots.  First, the annual breakdowns of the absolute number of graduates (from four year colleges) by group.

a

This series shows the percentages of total graph.  The rise we see in absolute numbers disappears, but occupational majors still capture more than half of all degrees

b

This graph compares majors to the U.S. population.  The graph shows a more extreme effect in recent years of the growth of occupational majors.

c

 

We could fiddle with these groupings on Ben’s interactive graphing tools, but here’s my takeaway:

1. Occupational majors dominate choices of majors in American colleges, and the dominance is increasing.
2. All other groups are relatively flat by comparison.
3. The effect isn’t quite as visible when you just look at the percentages of all majors. A better way to think of it: there are more people attending four-year colleges, and the increase is primarily an artifact of occupational majors.

R source files and NCES spreadsheet found here.

How to Lie with Statistics: Ice Bucket Challenge Edition

Look at this graph from a revisionist Vox piece on ALS donations:

If you’re teaching or taking elementary statistics, at any level, in any school, you can use this graph to demonstrate the “gee-whiz” deception technique which Darrel Huff described back in 1954 in How to Lie with Statistics.  Notice that the area of these circles bear no relationship to the numbers.

That’s because the proportions seem to be determined by the diameters of the circles.  Compare the orange circle to the pink one in the left column.  The pink circle represents a figure about two times the orange circle, but the area (and the visual cue) suggests it’s only 1/4 as large.  In other words:  1/3 becomes 1/9, 1/4 becomes 1/16, and so on.

It’s the oldest “How to Lie with Statistics” trick in the book: deceptive scaling.  It’s right there on page 66 of the Darrell Huff 1954 classic.

Reasonable people can disagree about which diseases should be prioritized for donations.  For my part, I disagree strongly with this article’s direction.  I have a relative who has been diagnosed with ALS, so my own disagreement is energized, but I still think that “death count”, or even “quality adjusted life”, are terrible proxies for “potential impact.”

Whatever you think of the article’s claims, this graph sensationalizes them.  I don’t know if the author and Vox are intentionally trying deceive readers, but deception is the result.

Update 8/28 10:19:  The article was posted on 8/20 and other observers have made the same complaint about the graphic (e.g., http://themendozaline.tumblr.com/post/95757674381/this-bubble-chart-is-killing-me).

The author pasted the graphic from the CDC, I’m guessing.  No excuse.

Updates on Washington 4th and Oregon’s Measure 90

Shortly after this month’s Washington primary, the Oregonian’s Jeff Mapes looked at Washington’s Fourth District and some implications for Oregon’s Measure 90, this year’s yet-another-top-two-primary initiative.

Thanks to Washington’s top-two primary system, the fourth congressional district will see two Republicans compete in the general election (Clint Didier and Dan Newhouse).  A frequent argument for top-two primary systems is the presumed moderating effect of including independent voters in the primary.  This result could also encourage candidates in elections like WA-04 to cross over and compete for the other party’s supporters.

Will these candidates try to win Democratic support, along with the district’s 36% Latino population?  Maybe, but there are reasons for skepticism.

  • Neither candidate will want to lose large numbers of their combined 57% primary support, not to mention supporters of other Republican candidates.  Republican bona fides are important in the upcoming election, in the House, and future elections.
  • Didier can run as a celebrity without moderating his positions (he has two more Super Bowl rings than most candidates) or he hope for some kind of crossover appeal from the tea party agenda.

As the Measure 90 campaign is already facing challenges, supporters and opponents can use this race for some indication about how a top-two system changes general election strategy.  Given that most voters struggle to locate the ideological positioning of congressional candidates, supporters have something to prove.  On the other hand…in WA-04, to paraphrase Barry Goldwater, there are a lot of unattended ducks on the Democratic pond.

WA04: For 2014, Primary Election Reformers Get Their Wish

For the first time since Washington state’s top-two primary emerged in 2008, two Republicans will face off in the general election (Clint Didier and Dan Newhouse).   No Democrat will appear on the general election ballot.

The race will be an interesting one-off test of Washington’s top-two primary system.  Proponents hope the system will attract independent voters into the primary process, create competitive races in general elections in places where one party tends to dominate, and motivate candidates to compete for centrist voters who are ignored in a traditional closed primary.

Did the open primary format promote more moderate candidacies, as proponents hope it will?  Will one of the candidates (Newhouse, presumably) take advantage of the district’s large Latino population, as reformers would expect to see?  A single race reveals little, but here are a few considerations I would want to analyze.

1.  On the surface, we only see two paleo-conservatives, and if neither tries to moderate or change the conversation, the desired moderating effect will be hard to notice.

But WA-04 is a conservative district; Romney beat Obama 59.7 – 37.9.  Of the 226 districts where Romney beat Obama, WA-04 is 100th (close to the median).   Maybe the result does reflect some kind of centering.

But perhaps not, because (arguably) the result is going to reflect only the result you would see in an ordinary Republican primary.

2.  The combined Didier/Newhouse total is 57%-ish.   The result is (arguably) a result of the open seat effect, and competition between Didier and Newhouse.

If so, Democrats might say “We would rather influence a race between two Republicans who might compete for our voters, rather than hoist a sacrificial lamb in the general election.”  So we should notice if (and how) either candidate tries to attract Democrats.

3.  The turnout in this election was very low.  Where were the independents who supposedly sit out during closed primaries?

4.  Didier’s minor celebrity status could give him a way to appeal to non-Republicans (or at least trigger their memory when choosing), raising this question: does valence give a candidate a way to attract centrist voters without appearing to be less radicalized?  In other words, can Didier protect his conservative bona fides by leveraging his celebrity status?  Does celebrity status, or the use of any other alternative to the first dimension of ideology, dampen the primary’s theoretical moderating effect?

5.  Finally, if the race is positioned as a “tea party vs. mainstream GOP” battle, with Didier the clear Tea Party favorite, does Newhouse (perhaps slightly less of a tea party favorite than Didier) try a Thad Cochran strategy of some sort?  WA-04 is 36% Latino.  If Newhouse tries the Cochran strategy with Latino voters, primary reformers have something to cheer about.

The last possibility, or some variation of it, should quicken the pulse of politics geeks everywhere.  Will the Chinese proverb May you find what you are looking for look like a blessing, or a curse?

Getting Your Money’s Worth from College: Don’t Forget to Graduate!

Will I graduate is the best opening question when choosing a college. Aggregate graduation percentages are widely available, and the percentages reveal something about overall college quality. Are these percentages reflected in college rankings?

This question occurred to me last Thursday, when Portland’s Oregonian published the editorial Getting your money’s worth out of college in Oregon isn’t a simple equation. Unobjectionably, the editors argue that students should bake their employment prospects and costs into their college selection. The editorial includes a list of Oregon colleges and universities that around in national ranking among based on an index from an organization called Educate to Career (ETC). This index includes 1222 degree-granting four year institutions in the United States.

The ordering in the Oregonian’s list is surprising, given the relative stature of various Oregon colleges:

85th    Western Oregon University
154th    Oregon Institute of Technology
210    Linfield College (Not included in the Oregonian article).
233th    George Fox University
360th    Pacific University
431st    University of Oregon
475th    University of Portland
613th    Eastern Oregon University
627th    Portland State University
686th    Southern Oregon University
929th    Willamette University
1076th    Lewis and Clark College
1219th    Reed College

According to ETC, their, “…college ranking system empirically determines the economic value added, by each college ranked within our system. We define economic value added as being the improvement in earnings and employability of graduates; measured against the total cost of the education.”

ETC describes the “characteristics of colleges by their rank” as follows (emphasis mine):

Colleges in the top 1/3 of the ETC Index

  • A relatively high percentage of graduates will be employed in their field of study.
  • A majority of students will graduate in 4 or 5 years.
  • Earnings of graduates are relatively high.
  • Loan default rates are very low.

Colleges in the lower 1/3 of the ETC Index

  • A relatively high percentage of recent graduates are delinquent or in default on student loans.
  • Most of the students will graduate in 6 years, or later, if at all.
  • Recent graduates have high student loan balances
  • Graduates are not employed in occupations that utilize their field of study

No one should over-interpret college listicle fodder, and usually, rankings can be entertaining (especially this one from 2009). Users of ETC’s website can assess the quality of their products for themselves. (Disclosure: I have taught at Lewis & Clark College the last two years, which ETC ranked 1076 of 1222). But do the comprehensive index-based lists reflect differences in graduation rates? I compared the ETC index to data compiled from the National Center for Education Statistics for 2010 by the The Chronicle of Higher Education.

Some of the colleges on the Oregonian’s list have relatively low rates of graduation, and I looked at ETC’s rankings in relation to four-year and six-year graduation rates.

ETC Rank Institution 4 year Graduation % 6 year Graduation %
85 Western Oregon University * 17.4 41.5
154 Oregon Institute of Technology * 18.9 41.5
210 Linfield College 57.6 63.2
233 George Fox University 59.2 67.0
360 Pacific University 55.8 64.6
431 University of Oregon * 43.9 67.8
475 University of Portland 67.0 76.6
613 Eastern Oregon University * 20.8 32.1
627 Portland State University * 14.2 36.0
686 Southern Oregon University * 13.4 30.9
929 Willamette University 70.8 78.2
1076 Lewis and Clark College 68.2 76.0
1219 Reed College 59.8 78.7
  • Public university. ETC did not include Oregon State University in its index.

v1

Overall, graduation rates don’t count for much in these ratings, and taken literally, the ETC claims about graduation rates are incorrect. A significant number of colleges in the top 1/3 of the ETC Index (113, or 27.9%) graduate less than half enrollees after six years, and 204, or 50.1%, of the colleges in the lower 1/3 graduate half or more. Presumably, ETC is trying to describe a tendency: they really mean that, generally, higher index values correlate to higher graduate rates.

Nevertheless, the relationship between ETC indexes and both sets of graduation rates is very weak. The relationship for six-year graduation for public colleges is somewhat stronger, but non-existent for private colleges.

v2

  • Lowess curves show 99% confidence intervals.
  • R2 (SE) = -.010 (.029) R2 (SE) Private = .008 (.037) R2 (SE) Public = .319 (.043)v3
  • Lowess curves show 99% confidence intervals.
  • R2 (SE) = -.092 (.029) R2 (SE): Private = .013 (.037) R2 (SE) Public = .450 (.041)

The more serious concern is the significant number of colleges that rank very high on the ETC index with graduation rates below fifty percent. The 113 institutions noted above comprise 9.2% of all the colleges in ETC’s rankings.

Is it possible that any college with less than 50% graduation should be considered one of the upper third of a comprehensive college rating, let alone 113 of them? The top 10% of ETC’s rankings still include 29 colleges (or 24%) with 6 year graduation rates below 50%.

Graduation rates aren’t a perfect indicator of quality. The institution may be admitting students whose needs are not reflected in graduation statistics, such as part time students or those looking to take particular classes or extension programs. Or they may just be unable to compete for good students because their reputation has not caught up with their quality.   In the case of the ETC index, graduation rates may not reflect its goals in the usual way.

But a low graduation rate is serious, and the ETC Index does not weight it very heavily. Choosing a school with a low graduation rate is a leap of faith, and the Oregonian should have said something about it.