Doing Economics

3
Giving Talks
After writing papers, presenting them—giving talks—is the second most
common form of communication in which economists engage. Even an
economist who has no intention of ever joining academia will almost surely
have to prepare at least one research presentation in her career—her job
talk, or a presentation of her research for an audience of people who are
considering hiring her as a colleague.
Like writing papers, how to give talks is something the average
economist is expected to learn on her own, learning both from others and by
doing, with little to no formal guidance except the odd blog post. This
chapter provides guidelines for giving talks, from your first presentation at
your department’s graduate-student brown-bag seminar to full-fledged
invited seminars, and from technical disciplinary research talks to outreach
talks given to people who know little or nothing about economics.
Presenting a paper involves the distillation of the results in that paper, so
much of the advice given in the previous chapter for writing papers (e.g.,
structure, content) naturally extends to giving talks. One difference between
presenting and writing—indeed, one advantage of giving talks over writing
papers—is that giving a talk is an occasion to engage in a dialogue with
would-be readers, and maybe even with potential editors and reviewers.
1
3.1 Invited Seminars
I use invited seminars as the benchmark against which I will compare other
types of talks, because an invited seminar is the ideal format for a talk in
that it allows you to present a paper fully, or nearly so. Moreover, it is a
format most economists will be familiar with from presenting at
departmental brown-bag seminars.

---

An invited seminar lasts anywhere from 60 to 90 minutes,
2
and the
presenter is expected not only to present a paper but also to answer the
audience’s questions about that paper. As such, it is the ideal setting to
communicate the contents of a paper from start to finish.
Thus, the first step involved in successfully giving invited seminars is to
know precisely what the norms are in the department that has invited you to
present your work. Though this obviously includes asking how much time
you will have to present, it also involves asking what the ground rules are
regarding questions. It used to be a distinguishing trait of economics
seminars that questions would be asked and answers given in a free-for-all
format,
3
but that is no longer the case as some seminar series have moved to
a format where no questions are allowed, say, for the first 15 minutes of
presentation to allow the speaker to set the stage for her work, or to a
format where only clarification questions are allowed throughout the talk,
with more substantial questions being relegated to a formal question-and-
answer session at the end.
Generally speaking, an updated version of the one-slide-per-minute rule
of thumb applies to giving talks. Though a 75-minute seminar will rarely
need as many as 75 slides, a 15-minute conference presentation should have
no more than 15 slides. Thus, the updated version of the rule is something
like “You should only have as many slides as you have minutes to present,
and the shorter the talk, the more likely this constraint is to be binding.”
Although the advice given to people making their first deck of slides is
invariably of the “less is more” variety when it comes to making each
individual slide, economists tend to be more comfortable with more text on
slides as well as with fewer images than, say, business executives. It is thus
perfectly acceptable for slides to have three to four bullet points, each
containing a full sentence. This serves a dual purpose: First, it means that
the speaker wastes no time memorizing what she wants to say. Second, it
also helps the audience understand what is going on in the paper, which in
turn helps the speaker by giving her more time to present instead of
answering clarification questions.
Like a paper, a talk—whether an invited seminar, a conference talk, or
any other kind of talk—is not a murder mystery. As such, a talk on a given
paper should be structured just like the paper it is meant to introduce to a

---

new audience. Specifically, an invited seminar presentation should
generally have the following structure:
1. Title. A single title slide with the title of the paper, the name of
the authors, and perhaps the date and location of the talk. If you
do choose to include the date and location of the talk, make sure it
is up to date.
2. Introduction. Do not make the mistake of assuming your audience
cares about what you are talking about just because you were
invited to give a seminar. It is your job to convince audience
members that your question is important, and that they care about
the answer more than they had hitherto realized. Use descriptive
statistics, broad trends, stylized facts, and so on to motivate your
work. Your introduction should consist of as many slides as
necessary to motivate the talk for the audience it is delivered to.
Generally, an invited seminar draws people from all kinds of
fields, so motivating the talk identically to how the paper
motivates what it does is fine. One way to draft the introduction
slides is to take the first sentence of every paragraph in the
introduction and turn it into a bullet point, perhaps shortening it in
the process. If a paper is written so that each paragraph introduces
a separate idea (or a separate piece of an argument), taking the
first sentence of each paragraph should yield a workable draft of
the introduction slides, with only a few gaps left to be filled.
Taking Keith Head’s introduction formula as an example, the
introduction slides should be separated into a broad hook meant
to make the audience care, a clear statement of the research
question the talk is meant to answer, a short (i.e., one- or at most
two-slide) discussion of antecedents (i.e., the five to ten most
relevant related contributions), a clear statement of the paper’s
value added (i.e., how it improves on those related contributions),
and a roadmap slide outlining the remainder of the talk. Much
like the roadmap paragraph at the end of the introduction,
opinions vary as to the usefulness of the outline slide. My view is
that it is better to include it and skip over it than not have it and
have some audience members confused about where they are in
the talk. It is also ideal to give a preview of your results, just in

---

case people have to leave early, or in case you cannot make it
through your entire talk.
3. Theoretical Framework. Presenting a theoretical framework in
support of an applied paper can be tricky. A common mistake is
to go into so much detail that one ends up in the weeds, and thus
wasting time with details that are, at best, left for the curious to
read for themselves. Another common mistake when presenting
theory—one that typically stems from the speaker’s insecurity,
and so one that is especially common among graduate students
and early-career researchers—is to go too quickly over the
theoretical framework in an effort to prevent audience members
from criticizing it. A middle-of-the-road approach works best.
First, this section should introduce the primitives, or basic
building blocks of the theoretical framework (i.e., the agents
considered, the commodities under study, the structure of the
market, the agent’s preferences or the technology they are using
to produce, the various constraints they face, and so on) as clearly
and concisely as possible. Second, the various assumptions made
about those primitives (e.g., whether a utility function is concave,
whether a production function exhibits constant returns to scale,
and so on) should also be stated clearly. Finally, the relevant
results should be stated roughly in the form “If assumptions 1, 2,
. . . are satisfied, the following result is obtained.” That’s it. This
is not the place to show how much tedious algebra you needed to
do to get at a specific result. If you feel the need to include a
specific proof, the place for it is your slide deck’s appendices,
which you can refer to in case someone asks for it.
4. Data and Descriptive Statistics. This section should present, in
bullet point format, all of the details necessary for an audience
member to understand the empirical work that follows and its
limitations. Where do the data come from, both geographically
(e.g., India) and institutionally (e.g., the US Census Bureau)?
What was the initial purpose of the data collection effort that led
to the data set? What is the basic unit of analysis? How were the
units selected for inclusion in the sample? What time period(s) do
the data cover? Were units followed over time, or is part or all of
the sample in a time period freshly sampled? For longitudinal

---

data sets, how much attrition was there between the beginning
and the end of the sample? Were the units that attrited replaced by
new ones and, if so, how were those replacement units selected?
What is the final sample size? Which observations, if any, were
dropped? Which variables are transformed using a common
transformation like a logarithm? After answering those questions,
it is customary to present a slide of descriptive statistics. Only the
variables retained for analysis should be included here—
outcomes, treatments, controls and, if applicable, instruments and
mediators—no more, no less. As with the outline slide at the end
of the introduction, the descriptive statistics slide tends to be
skipped (“Unless anyone insists on seeing these, I am going to
skip directly to my empirical framework”) unless the speaker
wishes to bring attention to any specific feature of the data. This
section also should feature slides for the relevant balance tests,
and the talk should clearly discuss the results of those tests—and
their consequences for the empirics. The one sin many commit
here is to present the tables in the paper “as is,” which is to say
much too small for anyone not having preternatural eyesight to
see. Unlike in a manuscript, it is perfectly fine here to break
tables into several slides for readability. Finally, this section is
also where limits to external validity should be discussed.
5. Empirical Framework. As with papers, the empirical framework
should be divided into two broad sections, namely estimation and
identification. The estimation framework should show the core
equations whose results are presented in the following section,
how those equations are estimated (i.e., least squares, maximum
likelihood, generalized method of moments) and which estimator
is used (e.g., linear probability model, probit, logit), clearly define
each variable used in those equations, clearly explain what the
subscripts (and superscripts) refer to, and so on, just as in the
paper. One thing to be aware of in this section is how variable
names are chosen, with the norm being either to use variable
names exactly as they are used in the theoretical framework or to
use fresh new variable names for the empirics. A common
mistake is to repurpose some variable names used in the
theoretical framework to refer to something completely different

---

in the empirical framework; for instance, to use x to refer to
quantity demanded in your theoretical framework, and then use x
to refer to your vector of control variables in your empirical
framework. After presenting the estimable equations, there should
be a thorough discussion of identification—that is, of the source
of (plausibly) exogenous variation used to estimate the causal
effect of interest, and why the results about to be presented have a
shot at being internally valid. For papers featuring a lab or field
experiment, this can be kept short. But for papers that rely on a
feature of observational data (e.g., the staggered rollout of a
program, a discontinuity in the data, some naturally occurring
source of variation), the onus is on the speaker to convince her
audience that that feature allows (getting as close as possible to)
estimating a causal effect as possible. Often, this section of the
talk is where much of the dialogue between the speaker and her
audience will take place.
6. Results and Discussion. This section is in principle the pièce de