University Admissions and Selection Bias

I'm at the age where I see many teenagers, both relatives and children of friends, heading off to university this fall. As all students and parents know, it is quite a challenge to find the best or most appropriate school. However this comes after having made applications to the preferred institutions, some or most of which are rejected. One of the criteria used is a measure of student performance.

The better the school, or often just the better the reputation, the highest the admission standards. When universities are ranked, it should be no surprise that the higher those standards, the higher the achievement of its graduates. This is entirely expected from a purely statistical perspective, reflecting the selection bias of the admission standards. In other words, unless the university is truly awful, good students ensure good graduates. It does not necessarily reflect the superior ability of the school to educate.

As a hiring manager in the technology industry this attribute of these schools has always posed a dilemma.  Some managers will unabashedly skew their preferences toward graduates from those schools, and especially if that school was also their alma mater. Yet this is unfair to other candidates since they are often of higher quality. There are a couple of reasons for this. One is that not all high-achievers are interested in or can afford to attend these schools, and may for financial reasons stick to where the cost is low. They may choose, for example, to live with their parents and attend the school in their home town.

The second reason has two parts: the quality of the learning environment and the ability of cream to always rise to the top. A consequence of selection bias is that it can hide a mediocre learning environment. Good, intelligent students will do well often in spite of the environment, in any school. This should not be surprising since that is likely what they've already done in high school: they've learned to productively learn at their own initiative. That initiative will continue to serve them well in university and beyond.

We should expect that a good school which admits an A student will, several years later, produce an A graduate. Anything less should be a travesty providing the student's life has not taken an abrupt turn in the interim. But a good school, no matter their standards, will also turn a B student into an A graduate, and even raise a C student into at least a B graduate. If you only admit A students, the grade point averages of the graduating class tells little on its own about the ability of the school to educate.

All the selection bias (high admission standards) of a school tells us is that the student has already proven himself or herself to be a high-achieving high school graduate. The standards themselves add nothing to this evaluation. Yet those standards do have one important value that is undeniable: they ensure that students are surrounded by other high-achieving students. Anyone who has gotten a science or engineering degree will know that some of the best learning comes from rubbing shoulders with intelligent peers. This alone can make up for many deficiencies in the formal teaching environment.

It is possible to ignore the high-admission standards of brand name institutions if they can meet just a few important performance metrics. Here is my list:

• Professors should be known to be good teachers and accomplished in their fields of study. A good proportion of them should have experience outside of academia, especially in the private sector. Many excellent professors can be found in lesser institutions since there are only so many positions in the select few top schools, and so good people are found in other places.
• The admission standards should be just high enough to ensure that a proportion of the students are high-achievers, to help create the needed learning environment. The sad truth is that even in schools with low standards, that proportion will increase after the freshman year since the poor students will not progress.
• Good facilities are needed to make research, studying and living both convenient and reasonably comfortable. Happily this is easier today than when I attended university since the internet adds tremendous ability to ensure that current research and communication with experts is much the same from any school.

There are of course other important attributes, although I tend to dislike those measured in most surveys. Use what you will, but just don't get blinded by the selection bias of high admission standards.