---
title: "When is a class a “commodity” and when does it depend on the instructor?"
slug: when-is-a-class-a-commodity-and-when-does-it-depend-on-the-instructor
url: https://listedarticles.com/articles/when-is-a-class-a-commodity-and-when-does-it-depend-on-the-instructor
canonical_url: https://statmodeling.stat.columbia.edu/2026/11/13/when-is-a-class-a-commodity-and-when-does-it-depend-on-the-instructor/
content_type: blog_post
language: en
published_at: 2026-11-13T14:20:52.000Z
updated_at: 2026-11-14T13:49:21.812Z
author: "Andrew Gelman"
authored_by: human
publisher: "Statistical Modeling, Causal Inference, and Social Science"
publisher_url: https://statmodeling.stat.columbia.edu/
topics: ["statistics", "education", "teaching"]
license: all-rights-reserved
word_count: 628
reading_minutes: 3
citation: "Andrew Gelman, Statistical Modeling, Causal Inference, and Social Science. \"When is a class a “commodity” and when does it depend on the instructor?.\" 13 Nov 2026. https://statmodeling.stat.columbia.edu/2026/11/13/when-is-a-class-a-commodity-and-when-does-it-depend-on-the-instructor/ (all-rights-reserved)"
# The full text follows. The web page shows an extract and sends readers
# to the source above; quote the citation and link the canonical URL.
---

# When is a class a “commodity” and when does it depend on the instructor?

> Andrew Gelman explains why he asks students to take one of his classes before joining a project: unlike calculus or linear algebra, statistics courses vary hugely from instructor to instructor in what actually gets taught, not just in teaching style.

# When is a class a “commodity” and when does it depend on the instructor?

Sometimes I get emails from students asking to work on a project with me, and I usually respond that they should take one of my classes first. If they try to work on a project without having taken a class first, I usually end up having to teach them a bunch of things. Sure, they could read my books, but I guess people don’t really like to read, also there’s something extra you get from hearing it live, doing the homework, getting feedback in class, etc.

This always puzzled me, that students would think they could jump in a project without taking any of my classes first . . . but then I realized that it’s usual to think of a math class as being pretty much the same no matter who teaches it. You take calculus or linear algebra or real analysis or whatever, and it can be taught well or poorly, there are good teachers and bad teachers, but they’re pretty much all teaching the same thing.

But statistics is different. Even if you limit yourself to a narrow topic–for example, regression analysis or survey sampling or Bayesian statistics (just to name three fairly generic-sounding classes that I’ve taught)–the course varies a huge amount from instructor to instructor. Not just teaching style (cos, yeah, my style is much different from that of every other statistics teacher I’ve seen. I’m not saying it’s better, mind you, just that it’s different. You can get a sense of it by reading Part 1 of [Active Statistics](https://avehtari.github.io/ActiveStatistics/)), but actually what gets taught. First off, most people don’t teach regression out of Regression and Other Stories, and other regression textbooks don’t look much like ours. Second, somehow what gets covered in class isn’t just what’s in the book. This is even true with Bayesian statistics. Lots of people teach out of Bayesian Data Analysis–and I hope that, soon, lots of people will be teaching from Bayesian Workflow–but there are still so many different directions to go.

And I’m not saying I’m unique here. In general, if you want to work with statistician X, it’s probably a good idea to first take a course from X, for just about any X.

In that way, a statistics class is less like a math class or a physics class or a chemistry class and more like a history class or a literature class. Even if the books are set ahead of time, different teachers teach it so differently that they’re just different courses entirely.

But I could see how many students, whose experience is taking straight math courses in high school and college, and maybe some very empty course such as AP statistics or the college equivalent, would look at the roster of statistics classes at a university and just see a collection of topics rather than what it really is, a collection of topics interacted with instructors. And our course descriptions usually don’t do such a good job of conveying that either. Hence this post.

**P.S.** I’m not saying that a course being a standardized commodity is a bad thing. In a way, that what my colleagues and I have been trying to do with [our textbooks](https://sites.stat.columbia.edu/gelman/books/), as well as with [Active Statistics](https://sites.stat.columbia.edu/gelman/active-statistics/): we’re trying to give other people the tools to teach as we do. But, the way things stand, it still matters a lot who teaches your class–maybe not such a bad thing, as the students should get some benefit from their big tuition payments. It just seems like a secret to many students that the instructor matters, not just in how well the topic is taught but also in the content of the course.
