In Part 1 of this post, I again discussed the former trader, author, statistician and somewhat cranky philosopher, Nassim Nicholas Taleb, and the ideas he presented in his book Fooled By Randomness, and then developed further in The Black Swan. Taleb is an admirer of hedge fund manager and fellow author Victor Niederhoffer, who Taleb says derives his knowledge of the world from past data that is “stripped of preconceptions, commentaries and stories.”
A crucial part of Niederhoffer’s dogma, according to Taleb, is “that any ‘testable’ statement should be tested, as our minds make plenty of empirical mistakes when relying on vague impressions … How many effects we take for granted might not be there?”
Here is a testable statement, “one that can be broken down into quantitative components and subjected to statistical examination,” says Taleb:
Automobile accidents happen closer to home.
Perhaps you’ve heard this one before. You can test this statement by averaging the distance between the accident and the driver’s home. Say that 20 percent of accidents happen within a 12-mile radius of home; how to interpret this? It would be a mistake, Taleb insists, to assume that you are more likely to have an accident driving near home than in remote places. This is what he calls, somewhat disparagingly, as “naïve empiricism.”
The fact is, accidents may happen closer to home because people spend most of their time driving close to home — especially if people spend 20 percent of their time driving within a 12-mile radius of home anyway.
Maybe we can make up a statement: “Enrollment never drops 20 percent between one year and the next.” Just because it’s never done that before in no way proves that it can’t do that. In 2016, CNN Money pointed out that U.S. college enrollment is falling, after a peak in 2010 of just over 21 million students. By the fall of 2014, there were 812,069 fewer students, according to the most recent data by the government’s National Center for Education Statistics. To paraphrase Taleb, we may not know much about enrollment management from historical information.
CNN Money offers many possibilities for the drop. One of the schools that have seen a precipitous drop is for-profit schools; naturally, that could be due to the bad publicity from the controversy of the “worthless degree.” It could be the financial expense, especially for low-income students. It could even be the fact that an upturn in the economy, which leads students to forego college in favor of getting a job.
It could be all of those. It could be none of those. It needs to be tested.
As we’ve pointed out before, Capture Higher Ed relies on a more sophisticated look at the historical data than a few “common sense” points. To not rely on intuition is the definition of “empirical.”
In building our predictive models, we reject theoretical thinking that seeks to construct a narrative for why a student is likely or unlikely to enroll at a given school. Instead, we rely on a strictly empirical approach for determining what features make the model and empirically test our models to see if they work on a holdout set of data.
By Sean Hill, Senior Content Writer, Capture Higher Ed