<p>When universities report class sizes, they report them based on the percentage of total classes that are a certain size. Yet, since large classes have many more students than small classes, having 4% of students in classes over 100 can really mean that 20% of the average student's classes will have more than 100 people.
To illustrate: imagine a very small school, with say, 200 students, each taking only one class. Say that there are 11 classes, 1 with 100 students, and 10 with 10 students each. Though only 9% of classes would have 100+ students (and 91% would have 10 students), 50% of the students would be in a 100 student class, while an equivalent 50% would be in a 10 student class.
Using the principle above and info from the common data set, I calculated the "real" class size ranges at three elite private research universities (Yale, Princeton, Stanford), 3 elite LACs (Amherst, Williams, Swarthmore), 2 elite tech schools (Caltech, and MIT), and 3 elite public research universities (Berkeley, Michigan, and Virginia).
In making my calculations, I used the common data set listings of number of classes in certain size ranges. I assumed that classes with 2-9 students would average 5, classes with 10-19 students average 15, 20-29 produces an average of 25, 30-39 makes an average of 35, 40-49 makes an average of 45, and 50-99 makes an average of 75. For classes with more than 100 students, I assumed an average of 150 at small schools (below 10,000 undergrads) and an average of 200 at large schools (above 10,000), since these large classes are generally lower-division introductory classes that are larger at larger schools (ie, all psych majors take intro to psych at both Berkeley and Princeton, but there are a lot more psych majors at Berkeley, so the class will be bigger). I calculated the approximate number of students in each class size range at a university, added these together to produce total student-classes. I divided by the number of undergrads to make sure these results were reasonable (when they did not seem to be, I have made a note). I then divided number of student-classes in each size range by total student-classes to produce the following percentages:</p>
<p>Yale
2-9: 7.25%
10-19: 32.03%
20-29: 12.37%
30-39: 5.86%
40-49: 5.65%
50-99: 16.27%
100+: 20.56%
Total: 99.99%</p>
<p>Stanford (note: produces a large number of classes/student, so may be off)
2-9: 8.10%
10-19: 22.35%
20-29: 8.34%
30-39: 7.07%
40-49: 8.26%
50-99: 20.19%
100+: 25.70%
Total: 100.01%</p>
<p>Princeton
2-9: 5.58%
10-19: 29.80%
20-29: 8.70%
30-39: 7.78%
40-49: 5.00%
50-99: 17.76%
100+: 25.37%
Total: 99.99%</p>
<p>MIT (note: produces a surprisingly large number of classes/student, so may be off)
2-9: 6.69%
10-19: 13.32%
20-29: 13.20%
30-39: 7.45%
40-49: 8.10%
50-99: 27.31%
100+: 23.94%
Total: 100.01%</p>
<p>Caltech (uses 2005 data, the latest CDS I found) (further note: produces a surprisingly large number of classes/student, even more than MIT so may be significantly off)
2-9: 7.07%
10-19: 18.94%
20-29: 15.57%
30-39: 16.50%
40-49: 5.30%
50-99: 21.46%
100+: 15.15%
Total: 99.99%</p>
<p>Williams
2-9: 11.65%
10-19: 30.46%
20-29: 18.93%
30-39: 13.65%
40-49: 7.23%
50-99: 9.47%
100+: 8.61%
Total: 100.00%</p>
<p>Amherst
2-9: 5.69%
10-19: 35.39%
20-29: 21.68%
30-39: 12.94%
40-49: 8.32%
50-99: 13.86%
100+: 2.13%
Total: 100.01%</p>
<p>Swarthmore (produces a number of classes per student that seems low, may be off)
2-9: 11.21%
10-19: 38.95%
20-29: 25.72%
30-39: 11.21%
40-49: 1.52%
50-99: 8.85%
100+: 2.53%
Total: 99.99%</p>
<p>UC Berkeley
2-9: 5.27%
10-19: 12.57%
20-29: 11.65%
30-39: 6.95%
40-49: 5.19%
50-99: 17.91%
100+: 40.44%
Total: 99.98%</p>
<p>University of Michigan
2-9: 1.78%
10-19: 12.48%
20-29: 16.15%
30-39: 8.37%
40-49: 4.70%
50-99: 20.03%
100+: 36.49%
Total: 100%</p>
<p>University of Virginia
2-9: 2.27%
10-19: 13.13%
20-29: 14.49%
30-39: 9.25%
40-49: 8.85%
50-99: 18.69%
100+: 33.33%
Total: 100.01%</p>
<p>Note: I'm well aware that I don't have accuracy to two decimal places (numbers probably vary by up to 10% of the value provided, ie 100+ at UVA is between 30 and 36%, most likely). I calculated this way to ensure that my numbers added up right, and I don't feel like rounding, so I left as is.
Further note: Obviously, these numbers do not mean one university is "better" than another because of class size numbers. As vicissitudes pointed out in another thread, large class sizes have many advantages.</p>