Rules for Presentations
The following rules are mandatory.
1. Plan your presentation to leave sufficient time for questions and discussion. This means that presentations should have no more than 15-20 slides. More than 20 are allowed only if there are a lot of figures. 25 is the absolute maximum.
2. Use no more than 5 slides for the introduction (including literature review), and preferably less. It is critical to get to your actual work as soon as possible.
3. No more than 1 slide, and no more than 3 minutes, on literature. Of course, you should be familiar with the related literature in order to answer questions about it.
4. Only put material on slides that you are actually going to talk about. In particular:
a. Only display equations that you will go through in detail. No general definitions, first- order conditions, step-by-step derivations etc. unless you explain them.
b. No walls of numbers (such as results from 17 regressions only one of which is the main result). Only numbers that you will actually talk about should appear on the slides.
5. When describing regression results, focus on the economics, not the statistics. In particular:
a. Describe every coefficient by a sentence like "if x goes up by .... then y...". The sentence should remind the audience of the units of measurement (percent, dollars, etc.).
b. Provide a sense of economic significance. In what sense is the coefficient large or small? For example, clarify the magnitude of a coefficient using summary statistics on x and y.
Detailed Suggestions
• General points about slides:
o Use a huge font and a simple slide layout. Remove any information or visual element that is not strictly necessary for understanding the slide.
o Minimize text: the fewer bullets and the fewer words per slide the better. No full sentences on slides. The slides should support your talk, not replace it.
o Use informative titles for your slides. Ideally, your main message should be decipherable by simply reading the titles of your slides in succession.
o Never use math symbols that have not been introduced.
• On the introduction:
o The introduction is important to tell the audience where things are going. At the same time, it is a trap where one can lose a lot of time. Work on it carefully.
o The introduction should contain (i) a brief statement of your research question, (ii) a brief description of the formal exercise you do (e.g. "part 1 has a simple model and part 2 is an IV regression" or "a calibrated dynamic model of xyz"), (iii) a brief summary of the main result (in economic terms; if your result is quantitative, then the main result should be a headline number), and (iv) a literature review that stresses what is new about your own work.
o Do not use quotes for motivation. The motivation can be either (i) a fact that has no obvious explanation or (ii) an open conceptual (e.g. policy) question. Do not provide quotes from dead economists, politicians etc.
• On the structure of your talk:
o Structure your talk into clearly identified segments, and make sure that your audience knows where you are. It can be useful to provide an outline slide that you can return to.
o Do not mix model setup and results. The setup/regression design should be explained first, and the results afterwards.
o No mystery novels. When presenting results, do not start with a sequence of steps that eventually culminate in a result. Announce the result (theorem or numerical) first, then explain it.
• On presenting empirical results/facts:
o When presenting a figure, first say what will be the point of the figure. Then state the variables on the axes, and the units of measurement (unless those are obvious from before, which is usually not the case). Then sequentially describe all the lines in the figure. Finally, say again what the point of the figure was.
o When presenting a table, follow the same basic pattern. First say why we need to look at the table. Then describe the layout of the table (e.g. "in rows are independent variables and in columns different regression specifications"). Also explain the units in which the variables are measured to the extent this is needed for interpreting the coefficients. Then walk though the important numbers. Finally state again the punch line.
o All regression coefficients should be interpreted in words, by providing a sentence of the type "if x goes up by ... then y goes up by ..., holding fixed z... ".
o Do not show irrelevant results. If OLS is not your preferred estimator, do not show OLS results, except perhaps as part of summary statistics if those are useful somewhere later.
• General presentation tips:
o Stand up, stand next to the screen, and look at your audience. Use your fingers to point. No laser pointers.
o Do not expect the audience to memorize math symbols. To address this:
? Economize on symbols.
? When talking, if possible refer to symbols by their economic meaning ("high risk aversion coefficient", not "high gamma" etc.).
o Remember that in 45 minutes you probably cannot present your entire paper. Make a conscious decision on which sections, derivations, results etc. can be omitted, instead of trying to fit too much and running out of time before getting to the main results. There is no penalty for finishing early.
o The easiest way to prepare a presentation is to imagine a smart economist who works on another topic than you: pick a name and imagine that person sitting in front of you. The goal of your presentation is to help this economist score high on a test that will ask him to write down your setup and explain the main results.
o Transitions are key. Tell the audience where you are going. Say when one subject is done and you are moving to the next subject. Say things like: "I have shown you the model. Now we are going to ..." Use back references. For example, say "I told you that investors tend to buy securities on their birthdays. The way I capture that in the model is ..."
o Can you make the same point with a graph instead of symbols? If so, do it.
o Mention the weaknesses in your paper. Trying to hide these weaknesses makes it more likely that your talk will end in disaster. Instead, show the audience that you are reasonable and on top of things, including the counterfactual implications of your model. Talk about them openly - your audience will respect you for it and it will help them better understand your results. Keep in mind that all models are wrong and it's the model's fault that it's wrong, not yours. The same is true for empirical results: it's the data, not you.
o You won't be able to avoid the situation that somebody asks you a question that highlights a bad feature of your model. This is not a problem- again, all models are wrong, not just yours! In answering questions of this type, use the principle "first the
bad news, then the good news". You first admit that your model is wrong. (Do this without hesitation, because your audience needs to understand that you are a reasonable person and are aware of where your model fails.) So, say something like: "yes, I agree with you that the assumption that agents die deterministically at the age of 60 is counterfactual." After you are on the same page with the audience, you can now say something positive about the model, like "But I still think that the assumption is useful for my purpose, because I am mostly interested in modeling the choice of education, which is likely to be less affected by what happens at the end of life." Or you say: "The main benefit from this assumption is that I can solve for agent's saving choices in closed form." You should end in a positive note to avoid the impression that your model is useless.