Doing Economics

2.11 Writing Papers for General-Science Journals
Increasingly, economists interested in improving the visibility of their work
submit to and publish in general-science journals such as Nature, PNAS,
and Science. Writing for those journals, however, is dramatically different
than writing articles for economics journals. I thus briefly discuss how to do
so to close this chapter.
General-science journals care much more about your findings and their
implications than they do about methods. Methods are important, but they
are relegated to the end of the paper (or to supplemental materials), for
specialists who may be interested in seeing them.
Structure-wise, a paper in a general-science journal has four sections: (i)
introduction and motivations, (ii) results, (iii) discussion, and (iv) methods.
In the results section, you should only report results. The interpretation of
those results should be kept for the discussion section, where they can be
put in context. You can even afford to speculate, since unlike in an
economics journal, not everything you state needs to be backed up by a
number of robustness checks. But even in general-science journals, your
core results need to be robust, and speculation should only be relied on to
explain the implications of these results.
Understand: it is very challenging to write for general-science journals,
as you will need to appeal to researchers beyond economics. As a colleague
with experience publishing in both top general-economics and general-
science put it: “Publishing in PNAS requires just as much effort as
publishing in a general interest economics journal. Don’t wing it.”
Notes
1
. Knowing is half the battle, so knowing that many readers will read your paper inspectionally can
make you a more effective writer because it forces you to put more thought into writing your
introduction, your methods and results sections, and your conclusion. If you know that many of

---

your readers are unlikely to bother with reading, say, your background section, you should state
the most important facts of that section in your introduction. The greatest sin an academic writer
can commit is the sin of omission, which consists in leaving important information out of a paper.
The second greatest sin an academic writer can commit is one of commission, and it consists in
forcing the reader to rifle through the paper hunting for a specific bit of information. The
opportunity cost of a reader’s time is high, so the average reader is more likely to give up on
reading a paper than to hunt for information. This is especially true when a relatively junior writer
(e.g., a PhD student or an assistant professor) is writing to impress more senior readers (e.g., a
faculty advisor, a journal editor, or journal reviewers).
2
. This is not to say that there are no good papers looking at several research questions at once. But
at this point in time, what tends to be rewarded by the economics profession is answering a single,
relatively narrow research question well.
3
. One possibility is to make a theoretical argument without math, in words, and to leave the math to
an appendix. See, for instance, Sánchez de la Sierra (2020).
4
. Comparing means across treatment and control groups is the strict minimum when it comes to
testing for balance. A more restrictive approach consists in running a joint test (i.e., F-test) of
whether all means are simultaneously the same across groups. Another, more restrictive approach
consists in conducting tests of equality of distributions for pairwise comparison using a
Kolmogorov-Smirnov test or using Bera et al.’s (2013) smooth test for equality of distributions.
5
. In his classic On Writing Well, Zinsser (2006) writes: “[T]he whole purpose of tenses is to enable
a writer to deal with time in its various gradations, from the past to the hypothetical future.” My
editor at MIT Press also tells me that analysis of literature and film also uses the present tense
when describing a book or a film.
6
. This presumes that your research design is the best thing out there. In cases where your research
design is second- (or third-, or n
th
-) best, unless you significantly improve on external validity,
you will need to adjust your set of target journals downward.
7
. I talk explicitly of statistical endogeneity—what makes Cov(D, ϵ|x) ≠ 0—because many research
economists still confuse theoretical and statistical exogeneity. Theoretical exogeneity is when a
given variable is determined outside of a given theoretical framework (e.g., prices and income in
the typical utility-maximization problem). Statistical exogeneity is when Cov(D, ϵ|x) = 0 in the
regression framework we have been considering. Though the two share the same “exogeneity”
and “endogeneity” terminology, there is little overlap in their respective meanings. It is a poor
empirical economist who says his results are causally identified because his treatment variable is
theoretically exogenous.
8
. A colleague who has run numerous RCT notes that in his experience, adding controls tends to
have almost no effect on the standard errors, with the only exception being when the baseline
value of the outcome is added as a control variable in an ANCOVA (analysis of covariance) setup
(McKenzie 2012).
9
. This section’s placement in this chapter in no way indicates that you should think about
mechanisms ex post. It merely indicates that the section in which you empirically test for
mechanisms should come after your main results. Ideally, the presumed mechanisms whereby
your treatment variable affects your outcome of interest should be discussed as part of your
theoretical framework.
10
. Lab-in-the-field experiments are lab experiments that are conducted with “real” subjects (e.g.,
firm managers) in the field, outside of the experimental lab.
11
. A collider is a variable caused by two separate, possibly unrelated variables. Conditioning on a
collider or on a variable that lies on the causal path between the treatment and outcome variables
is problematic because it introduces bias (Morgan and Winship 2015).

---

12
. That being said, at the time of writing, the AEA journals have moved away from relying on
asterisks (i.e., “stars”) or other typographical symbols to denote statistical significance. It remains
to be seen, however, how many other journals will follow suit. Here, the best thing to do is to read
a journal’s formatting guidelines before submitting.
13
. One is tempted to add “in any discipline except economics,” as it is only in economics that it is
seemingly more important to please five to ten gatekeepers so as to get in the right journals than it
is to actually have an impact.
14
. Field journals are defined in relation to general journals. Whereas the latter might be open to
publishing articles on most if not all areas of economics, the former publish articles in a given
field (e.g., agricultural economics, economic history, monetary economics).
15
. Traditionally, the top five journals in economics have been, in alphabetical order, the American
Economic Review, Econometrica, the Journal of Political Economy, the Quarterly Journal of
Economics, and the Review of Economic Studies.
16
. The most famous of such urban legends is probably that surrounding Jack Kerouac’s On the
Road. From a story on National Public Radio (Shea 2007): “Legend has it that Kerouac wrote On
the Road in three weeks, typing it almost nonstop on a 120-foot roll of paper. The truth is that the
book actually had a much longer, bumpier journey from inspiration to publication, complete with
multiple rewrites, repeated rejections and a dog who—well, On the Road wasn’t homework, but
we all know what dogs do.”

---