How to read a paper

Most of the paper I’ll assign this term are in a standard format. Once you learn that format, you’ll be able to “read” a paper quickly (indeed, potentially more quickly than me because you’ll learn how to skip technical material and so so, whereas sometimes I’ll go through those details).

  1. Read the conclusion and then the introduction (doing it the other way around is OK, too). Sometimes they’ll be the same, wasting space and the reader’s energy (hey, in a short paper do you really need to repeat things?). The better written the paper, the easier it is to skim.
  2. Next — and I will take you through the process — scan the tables. This won’t always be helpful — the variable names may be computer-generated gibberish, which I find inexcusable in the days of ready editing. And it may not be clear what the tables are trying to convey — the title may be clear, but you won’t know which variables are central, and which are there as controls.
  3. Then … there will be a literature survey, something that will place the paper in context. Sometimes it will be terse, assuming you’re generally familiar with the field, and hence not very helpful. But try…
  4. Then comes stuff on methodology — what is the model, and perhaps details of what they’ll do in terms of statistical tests. If the paper is well-written, you can still read the first and last paragraphs, and (sometimes) the first and last paragraphs of sub-sections.
  5. You will find a description of data, typically with a table of sample means and so on. Always worth scanning.
  6. The penultimate section (or two ) will be the results. You really ought to skim. When there are two sections, one will be the main results, the other will consist of robustness checks.

The goal? — reading a paper in 5 minutes. Hard to achieve, but you may be able to approach that for your term paper, when you’re reading [scanning!] multiple papers on the same topic.

Addendum: you need to learn how to read tables, not a problem if you’ve had Econ 203 (econometrics), and comparatively easy when they use asterisks to denote statistical significance, harder when you only get a standard error. As noted above, I’ll take you through examples in class.

Leave a Reply

Your email address will not be published. Required fields are marked *