Some thoughts on taking notes

Five years out of grad school, I realized the importance of being able to learn quickly when you’re working in job that requires a PhD. (Emphasis on quickly as opposed to learn.)

I was going through the tenure process. I no longer had professors telling me what to do (other than journal referees, all of whom hated me with a passion), my service and advising duties were steadily increasing, and I had a young child at home. It was quite a change. When you’re in grad school, you spend all your time learning. Even when I had reached the stage where I was “doing research”, it was mostly just a different type of learning, or trying/pretending to learn.

By the time I reached the tenure stage - and my experience was by no means unique in this - running the university came to occupy a big part of my job description. Service, advising grad students, advising undergrads, participating in all sorts of conversations and decisions that had nothing to do with research or teaching. When I could have really used a couple hours of work in the evening, I was choosing to spend my non-office time with my kid. It was tough to find the hours and energy just to get the urgent stuff done.

My field is macroeconomics and I work with time series data. Empirical macroeconomics is a technical field with new methods to learn all the time. None of the important tools I use in my research today were around when I was taking classes as a grad student. Since then, we experienced the Great Recession, energy economics exploded as a field, Bayesian inference went from unusual to the status quo, there was a “credibility revolution”, and the new researchers were working with new econometric tools, new programming languages, and big data. It’s not easy to keep up with all that. And to be honest, it doesn’t make sense to even try. Nonetheless, it was a challenge for someone in my position to keep up with enough of it to remain relevant.

One of my disadvantages is that I’m a slow learner. In grad school, I was able to substitute hours for intelligence. That didn’t work given my time constraints. I had to learn how to pick up new ideas quickly. Central to this was developing a good notetaking system. That wasn’t required when I was a student, when I only took notes during the lecture. All I had to do was combine my lecture notes with the readings and I did well enough to pass. Learning on my own meant I had to figure out what to learn and how to learn it. Once I had an understanding of the important concepts, I had to make connections with other ideas.

In those days I took my notes on paper. My system worked by capturing ideas as they came in.

What to learn. I found survey papers and job market papers to be a good indicator of the tools others were using and the questions they were trying to answer. When I was reading a paper and thought something looked important, I wrote it down, along with any thoughts I had. Figuring out what to learn isn’t difficult as long as you have a good notes system and you use it regularly.

How to learn it. This was more challenging. I would look for textbooks, literature reviews, original articles, and software provided by others. It’s not perfect today, but there was no Github, no culture of sharing back then. The little bit of software that was shared back then was for the most part poorly written and specialized to a particular problem. I did my best to collect these resources when I came across them.

Make connections with other ideas. This was something I didn’t do at first, probably because I was using paper, and the biggest problem with paper is that it’s not easy to make connections between notes.

These days I only take notes digitally. Writing on paper is the best way to learn, but you lose a lot in the process. You can’t search your notes, you can’t group similar items, you can’t query by tag, it’s hard to edit/extend paper notes, and most importantly, you can’t create hyperlinks. I want to link an empirical paper with notes on an econometrics paper that introduced the methods they used, or to a program I wrote to understand the econometric method.

In my experience, the details of the system don’t matter very much as long as it has search capabilities and the ability to add hyperlinks. Software like Evernote and Notion work fine, Emacs org-mode works fine if that’s your thing, wikis like PmWiki, Dokuwiki, and TiddlyWiki all work fine. SQL and Mongo databases work fine if you’re comfortable with databases. Fossil provides version control, a database, and wiki all in one. Go with the system you think is the most fun.

I do my best to keep my notes system simple. Plain text and its cousins markdown and org-mode are more than sufficient - anything other than the most basic formatting is a waste of time. I keep all my notes in one place. I use a system that will scale to thousands of pages (all of the above software will do that easily). My goal is a single database of notes rather than a bunch of notes scattered across computers, devices, directories, and apps. Taking notes is only the first step in the process. I also have to be able to access and query them without much overhead. That only works if everything’s in one place.

I version control my notes. That gives me a complete history of everything and useful metadata for free like automatic dating of notes. Even though I’m an economist, I don’t add anything but simple equations to my notes. I’ve found it’s better to write out the equations on paper and keep them with the original reference, or to scan them with my phone app into OneNote and then link to them. Since I don’t need to search my equations (I don’t think I’ve ever done that even once) and equations don’t contain hyperlinks, there’s really no point typing them into a text editor. I’d resist taking notes if it meant typing in lengthy equations.