These are some opinionated comments on aspects of professional life—writing, giving talks, reviewing, judging colleagues, and so forth. They are based on my own ideas and experience and consequently, are undoubtedly customized to my interests in discussing economics experiments and interdisciplinary work across the psychology-economics boundary (and reflect battle scars from the latter). I realize it is a little pretentious to post these, but many students ask for advice and a little bit of it may save you from some errors. At the same time, please don’t take my opinions as special or definitive. Talk to fellow students, colleagues, and advisors as well. Borrow from those you admire, and avoid errors you perceive others making. Many of the same themes occur again and again in the passages below. Feedback welcome, of course, especially about other sites or written sources that people might find helpful.

Other sources: Caltech colleague Matt Jackson’s tips on seminars are at http://www.hss.caltech.edu/~jacksonm/Jackson.html. Leigh Thompson, a very successful organizational psychologist, has some professional advice at http://www.leighthompson.com/index.htm

Hal Varian, a well-known economist (who is famous for his introductory theory book and also writes columns for the NYTimes) has a great article on how to write papers in a collection, Readings in Games and Information, edited by Eric Rasmusen (see http://php.indiana.edu/~erasmuse/GI/rcontents.htm; it is a wonderful collection with many gems besides Hal’s great article, for those interested in game theory and applications).

Picking topics and writing papers

Picking topics:

First, write about what you are genuinely interested in. Much of research is conducted on your own, so intrinsic curiosity is very powerful. If you ask “Is this a hot topic?” you are missing the point. If your advisor sticks you with something you are bored by, everyone will be able to tell—especially you.

Second, choose something that is not too hard and not too easy. The overwhelming amateur mistake is to pick something too difficult and broad. Start with something so simple it is almost laughable. There is a profound asymmetry here: It is much easier to scale up something simple, and harder and frustrating to cut back. (Remember that in good restaurants people start by chopping vegetables for a year. The idea is to hone your skills and come to appreciate the raw ingredients and how basic skills matter. Pick something doable that builds your confidence.) If you are modelling complicated games, start with 2 players. If it’s dynamic, start with 2 periods. If these are ridiculously simple and easy to generalize—good for you—you’ll know that right away.

Pick a topic that is about something interesting in the world, rather than just the internal academic world. A small step will lead to bigger ones.

In choosing a thesis topic some other guidelines are useful. Pick a topic that stretches you and forces you to learn some new tools. At the same time, if you are planning a job market topic, show off what you do well and who you are. The transition from graduate student, to job market candidate, to tenured faculty member, is like a funnel: It is ok to dabble in lots of areas when you are in graduate school, learning and thinking, and after you have tenure. (That’s what tenure is for—to license you to think very broadly.)

In writing articles, audience design is key. Pick a prospective journal and a title. You don’t really understand your paper until you have a good title that is both apt and precise—it really describes what you have done--, provokes curiosity, and is zesty. If you pose a question be sure to clearly answer the posed question in the abstract of the paper, the introduction, and in the conclusion.

Remember that when you become a professor you’re a professional writer. Part of your job is to write well and clearly for your target audience. Feel free to use analogies, examples, and be playful and engaging. But if statistics are persuasive, use those too. Be graphical. (Economics is a seriously under-visualized field; use graphs which are easily interpreted to make your point.)

Seek feedback before you submit your article. Beg people you know to be constructively critical to tell you what they really think. Brace for some painful criticism and accept it. If you are really wounded, put the criticism aside until the sting wears off; then take it seriously and respond to it.

In the behavioral economics wars c 1980-95, we took the view that the easiest way to win an argument was to run another regression or experiment, or generalize the theorem. If a referee says “I think this result is sensitive to how much money subjects earn” try hard to find a source of money for a higher-stakes replication. Regression-running is even easier. If it is really difficult to generalize a theorem, explain why that’s so.

Giving talks

Science is a conversation; so conveying your ideas verbally and in writing is important. It is crucial for getting a job (your “job market talk”—practice it!) and getting your research to have impact (meaning that others understand it, cite it, describe it accurately to colleagues and students, and potentially build on what you have learned). Talking out loud is also a good way to cement your ideas and learn from feedback. (It is easier to get people to react to you when they are sitting and listening than when they are sitting reading your paper.)

In economics there are two basic talks—90 minute seminars, and short seminars at large conferences (typically 20 minutes).

Build either type of talk around a beginning and an end. Be sure to state very clearly why you are doing the research. What question do you hope to answer and why is it important? Revisit this statement in the conclusion— you can simply show your opening slide again— and present a clear answer. Keep in mind that in most audiences (and in most readerships), most of the people listening and reading are not specialists in the area. An ideal talk engages the full audience for at least half the talk, and also impresses the specialists in the other, more detailed, half (ideally, the specialists then “certify” whether your work is solid to the nonspecialists). If you have to use technical jargon—usually unavoidable—define it clearly and use mnemonic notation that is easy to remember, or refer back to what the notation means frequently. (Don’t worry about boring the specialists who know what you mean by “Lebesgue measure” or “epsilon-equilibrium” or “amygdala”; they won’t be bored by your repetition and nonspecialists will appreciate the reminders. Nobody ever gave a talk that specialists sniffed at because it was “too clear” unless it is obviously shallow.)

A key feature of your talk is “audience design”: Who is the audience? How much do they know? This will often vary at different universities and you should adjust accordingly. If there is a local angle—like a person there (even if deceased) who worked on your area, mention in passing how their work and yours fit together. Often there is one or more people in the audience you really want to reach or impress; don’t be shy, as a heuristic to guide your thoughts, to imagine yourself talking directly to that person.

Prepare slides carefully. When you are starting out, look the slides over at least a day before and catch any typos. The usual mistake is to put too much material on the slides. One rule of thumb that professional speakers advocate is to put no more than 8 lines on a page, and 8 words per page. Don’t xerox pages from your paper unless they are slides or tables. If you have a complicated proof, sketch the main steps and go through them slowly.

Narrate what is on your slides if they are tables and graphs. Say what is in the columns and rows of a table, and remind listeners and readers what is on the x- and y- (and z-) axes. I once gave a talk where somebody in the audience was blind. This was a tremendous help for everyone, because I was forced to articulate precisely what was on the graphs! (As in, “The graph shows that over time, prices tend to converge toward 80, the Nash equilibrium prediction.”) Even those who could see, but were confused or tired or sitting in the back, got the point more clearly. A similar lesson applies when speaking to an audience who are not all fluent in your native language: Slow down your speech and put all the relevant ideas on your slides, so if they cannot follow your rapid speech they can at least read what you have said (most non-native speakers can read more rapidly than they can process your speech).

Buy a laser pointer and use it. (I am amazed these have not become standard in economics; in “real sciences” like neuroscience everyone uses them, so using them is not “unscientific”.) It guides the listener’s eye toward what you want them to see, especially in a busy graph or table.

Questions are your friend. They show the audience is engaged. Questions often remind you of an important feature you skipped over or did not explain clearly; they help you gauge what your audience has missed and what you can skim over more rapidly. Be prepared for some oddball or hostile questions. Don’t back down from an interesting skirmish, which often clarifies precisely what you have done that is interesting and new and—hopefully!—controversial. If you are really baffled or not in a fighting mode, politely say “That is a good question I’ll have to think about” or “let’s talk after”—and when you say that, cash the check by reaching out to the person afterward and trying to understand what they are saying.

While you want to be generous about allowing questions, feel free to take control of the seminar if you think the discussion is wandering and your time is being frittered away. (Knowing how to do this takes practice and some nerve. It’s true some people like to hear themselves talk or have to say something; whether they are doing so or really striking at the heart of a flaw in your work that you need to confront takes some practice. Also, if you are on the job market or giving some other nerve-wracking talk ask your sponsor or hosts if there is any sort of surprise question you are unlikely to anticipate and should be especially prepared for. A good thoughtful response to an anticipated question is impressive.)

The 90-minute talk: A good rule of thumb is that you can cover one reasonable-length slide every 3 minutes (more typically 5 minutes). Get a quick start—people often waste a couple of minutes on a nervous preamble, naming your coauthors and reciting your title which everyone can see. Avoid that and jump right in. Giving an experimental talk, I like to start with the design immediately to give the audience a subjects’-eye view of what the subjects were seeing and thinking when they were plunked down in an experiment. (Showing screenshots of experimental software is a good way to do this.) I call this the “James Bond start” after the style in recent James Bond movies. They often begin with a 10-minute scene of James doing some stunt and getting in trouble. After James is safe, then the opening credits roll.

Even in a 90-minute talk, you will typically be rushed at the end. Have a triage or bail-out plan: If you have 15 minutes and are already halfway done, what do you do? Feel free to skip ahead to the basic points you absolutely have to make. I often put key slides that are important toward the end so I can overcome the natural inclination to keep chugging along, in favor of jumping ahead to my favorites I really want to mention.

The 20-minute talk: Many of the comments above apply to the 20-minute talk, in spades. The 20-minute talk should *not* be the first 20 minutes of your 90 minute talk, a random sample of the longer talk, or a sped-up version (don’t give the 90-minute talk and speak 4-1/2 times as fast). Your goal in the 20-minute talk is like the advertising “trailer” or commercial for a movie—you want the audience to get the central point of your talk: What is your question and what did you learn? You want to whet their appetite for reading the paper, rather than trying to cram the whole thing in. If you have 6 experiments, discuss only 2 in detail and mention the others. Similar heuristics apply for empirics and theory. *Never* give details of a proof unless you think the audience is skeptical [in a job talk, for example, walk through a proof so they know that you know what you’re doing] or the proof has some really clever twist.

Refereeing

First, know the audience for the journal and its standards. A paper you should reject for Econometrica might be an obvious acceptance at a more specialized journal like, say, Games and Economic Behavior. Most journals have a target acceptance rate they are aiming for (say 10% at top econ journals). Many journals print statistics on turnaround times (also help for deciding where to submit) and acceptance rates once a year, so find these numbers and use them to guide the standard you apply. Keep in mind that referee reports are only weakly correlated, so if every referee applied a 10% standard, and journals want to see say 2 of 3 positive reports to move ahead, you are being too harsh by applying an internal 10% standard (aim for maybe 25%).

Keep in mind that journals are loosely divided into wide-audience journals and specialized journals. The wide audience ones include: American Economic Review, Quarterly Journal of Economics, Journal of Political Economy, Economic Journal, Review of Economic Studies (more oriented toward European theory), and Econometrica (more technical). If you aren’t familiar with the journal you are reviewing for, look at some past issues to get a feel for what they like. The wide-audience journals are looking for the best papers in the specialized journals that their wide audience should see. For example, if you are reviewing experiments for AER you should think: They will probably publish 10 experimental articles this year; is this one of the 10 that a broad audience should see?