Coding

Do you need the best software engineers for coding? Don't look at Stanford and Berkeley.

It should be safe to assume that the average Stanford University computer science graduate would ace a coding proficiency test like the one given to entry-level software engineers at companies like Square. The majority of them do. However, they are not the best of the best on average. Stanford computer science graduates do not even make the top ten list of high scorers on the General Coding Assessment. A coding proficiency test designed by CodeSignal and administered to software applicants at the majority of major tech companies. Neither do those traditional universities such as the University of California, Berkeley, which is tied with Stanford for the second-best engineering program in the United States News and World Report college rankings, trailing only MIT.

Companies that have used or are currently using CodeSignal assessments in hiring include Meta, Robinhood, Square, Uber, Instacart, Zoom, and Asana. Each year, CodeSignal publishes these ranking reports to persuade these companies and the rest of the industry that recruiting primarily from universities with prestigious software engineering reputations is an inefficient use of resources. Stanford produces graduates with generally high scores, but the report aims to show that many other schools train students who are just as competitive, if not more so.

At the University of Central Missouri, which graduates more computer scientists than Stanford and Harvard combined, you could find a whole bunch of amazing software engineers. Companies spend millions of dollars chasing Ivy League graduates, and they don't always recruit from other schools. It's insane in this incredibly competitive market. Over the last few years, the racial, ethnic, and socioeconomic makeup of the tech industry has remained relatively stable. For tech companies that want to change that, recruiting from schools other than the stereotypically elite institutions could be one of the simplest ways to do so.

Unlike traditional university rankings, which consider graduation rates, job placement rates, peer reputation, and funding, the CodeSignal report considers only one data point: how people perform on the company's standardized assessment of the General Coding Framework. The industry’s widespread use of CodeSignal assessment has resulted in a statistically significant data set that employers and job applicants can rely on. Students and entry-level engineers all over the world study for this exam, which is used to screen applicants by all types of tech companies. More than 160,000 engineers have taken CodeSignal's assessment, and the company estimates that more than half of graduating computer science students take it. Most computer science programs in college teach algorithmic problem-solving skills, and the test is designed to assess those skills rather than knowledge of a specific language such as Java or Python.

CodeSignal scores range from 600 to 849, with scores above 800 indicating excellent problem-solving skills equivalent to the 84th percentile, according to the company. The university ranking list is based on the number of test-takers from each school who scored above 800 out of the total pool of people from that school. In 2022, an impressive 43 percent of UVA Charlottesville test-takers scored above 800, while Swarthmore's 22 percent is just above Stanford and on par with the California Institute of Technology. Swarthmore, a small liberal arts college, will graduate just over 50 students this year, outperforming not only Stanford and Berkeley but also the Georgia Institute of Technology and other large engineering institutions. Swarthmore CS graduates excel in more than just the CodeSignal test; at the North American championship for the International Collegiate Programming Competition last year, a Swarthmore team finished fourth, making it the only liberal arts college in the United States to qualify for the world championship.

The college's computer science department chair, Andrew Danner, speculated that the school's emphasis on algorithmic problem-solving rather than teaching specific languages could explain its success. Our introductory course is taught in Python, but the goal here is to teach you enough Python so that you can solve some computational problems with it. We do that again in our intermediate courses where we switch the language and teach them C and C++ so that they see a variety of languages throughout their career. There are also a lot of schools where you come in and start learning Java, and you spend your entire time learning Java, and you know that language extremely well and may not know how to adapt to other languages.

Computer science graduate is the most rapidly growing undergraduate degree program at almost every school that offers it. Even though Swarthmore does not actively recruit students interested in CS, it is now one of the top three degree programs. Students at the school have certain advantages over those at larger schools with well-known degree programs due to the school's small size. While a student at an elite research university might take a 300-person CS class with teaching assistants, Swarthmore's largest class is around 60 students, and everyone learns from the professor.