The introduction of an academic paper is arguably more of an art than a science. That said, there is a commonly used formula that I recommend to pretty much everyone (and use myself in most of my papers). I do not claim that this is the right formula for every paper. But it works well in my experience, and I believe the closer you try to stick to this formula, the more effective your introduction will be. Don’t try to “cheat” by making paragraphs very long.

Paragraph 1: Big-picture motivation. Make sure to include a clear statement of what is missing from our current state of knowledge.

Paragraph 2: What the paper does. Could be two paragraphs if the approach is really complicated, but ideally should be one paragraph. Try to give the reader a clear picture of methods, but avoid excessive detail about data, etc.

Paragraph 3: What the paper finds. Don’t just write down a list of findings, make sure you place them in context and state clearly what they imply/why they’re important. A list of things that your main findings are robust to can also go into this paragraph (in the concise form of “Our findings are robust to [list of variations].”)

Paragraphs 4-X: Extensions/heterogeneity analyses/mechanisms (“extensions+”). Don’t mention these in the first “what-the-paper-does” paragraph. Let the reader absorb the main findings first, then build on them. With extensions+, two approaches are possible. If the extensions+ are closely related, you can describe all of them in one paragraph and summarize the findings/takeaways in the subsequent paragraph. If the extensions are not closely related, you can devote a paragraph to each extension, stating what you do followed by what you find. The optimal number of paragraphs depends on how many extensions you have, but it’s hard to see >4 paragraphs devoted to this being optimal.

Paragraphs X+1/X+2: Literature contributions. See tips on how to write these above.Try to limit yourself to two paragraphs on literature contributions. Avoid a separate “literature review” section.

Last paragraph: The rest of the paper is structured as follows. Yes, whether to have this paragraph is somewhat controversial. But I like having a road map, and papers have different structures/different numbers of sections. So I suggest having this paragraph, at least as a PhD student (and all my papers have had it).

Now for some bonus content, I’d like to share some common mistakes that I see made in introductions. Will your paper get rejected if you make one of these? Maybe not. But if you avoid these, your introduction will undoubtedly be stronger.

Taking too long to get to what the paper does. My “formula” is to state what I do in paragraph 2 (and to have the first paragraph take up no more than half a page, ideally a third or less). This is hard to do (and I have papers that violate this formula), but it usually makes for a much more effective intro. See “my suggested formula for introductions” below.

Having a lot of citations in the first few paragraphs. Not only does this make the paper more difficult to read right away, but it also makes the reader feel like a lot of work has been done on these topics already and makes it harder to demonstrate a contribution. If a motivational citation is needed or helpful, one is sufficient (preceded by an “e.g.,” if it is one of many papers). Leave most of the intro section citations for the contribution part of the intro.

Not making it clear what the missing knowledge is up front. Has the question you’re asking been difficult to answer because it requires hard-to-get data (which you managed to get)? Because credible exogenous variation has been difficult to come by (but you have found such a source)? Because no one has figured out computational feasibility (but you have!)? Even if there’s no particular reason why no one has answered this question so far, having a clear statement of the knowledge gap will help prepare the reader for what you do (and will hopefully make them excited to read about it!)

Having a strawman research question. “Does air pollution affect human health?” is a very easy research question for reviewers to strike down as making an insufficient research contribution because almost everyone is already convinced the answer is “yes”. To avoid this situation, either rephrase your research question to be a quantitative one (e.g., “How much does air pollution affect human health?”) or briefly discuss why the answer is not obvious (e.g., what are some offsetting factors/mechanisms that could make the answer go one way or another?).

Not having clearly stated contributions in the contribution paragraph(s). Many students (and some professors!) write their contributions as “Doe et al. (2012) do X, but we do Y”, without making it clear why Y is important. Doing something different will not be automatically considered a contribution by the readers.

Spending too much time discussing what other papers do. For example, you might see something like this in the contribution section: “Doe et al. (2012) build a general equilibrium model in which there is a non-tradable service sector and a tradable goods sector. They find that, without a border adjustment, a domestic carbon tax has almost no effect on global emissions because of leakage.” and only then a discussion of the current paper’s contribution. In many cases, such summaries are better off being framed in terms of the paper’s own contributions. For example, replacing the sentences above with the following will make for much more effective writing: “Unlike Doe et al. (2012), our model includes a tradable service sector. We demonstrate that this modification overturns the finding that a domestic carbon tax without a border adjustment results in almost 100% leakage.”

Listing too many papers that have used similar techniques before. Yes, it helps with credibility, but it also takes away from novelty (at least subconsciously). In many cases, when discussing the method you use, you can just do an “(e.g., ….)” cite of one paper that used the technique and discuss other papers in the empirical strategy section. This is not to say you should avoid citing relevant papers. We must absolutely give credit where credit is due. But the citations can either be in the contribution paragraphs (if relevant to the contribution) or in a different section of the paper where you go into more details.

Listing citations in the contribution paragraphs without putting them in context. This can make the reader feel like you’re writing in a crowded literature and make them wonder what your contribution could possibly be. Only papers that are directly relevant to yours should be cited in the contribution paragraph and it should be clear why they are being cited. For example: “By focusing on adults, our study complements work studying the effects of pollution on infant mortality (Doe et al. 2012; Smith et al. 2013).” is much more effective than “Other studies that study the effects of pollution on mortality include Doe et al. (2012) and Smith et al. (2013).”

Explicitly criticizing previous literature. Of course, you want to appear to be making advances relative to previous literature. But the way to do that is to highlight your paper’s accomplishments rather than previous papers’ shortcomings. Criticizing previous papers’ shortcomings not only takes the focus away from your work (you will likely end up discussing other papers more than you should) but also makes you (and/or your coauthors) look mean-spirited. Using phrases like, “We build on the work of Doe et al. (2012) by …”,  “We extend the results of Doe et al. (2012) by…” or “Or findings complement those of Doe et al. (2012), who …” will highlight the contributions of your paper without explicitly beating up on Doe et al. (2012). Even better is to say, “We are the first to [insert your awesome contribution here].” This is not to say that you should never point out something that prior literature did not do. Just stick to neutral statements in such cases and avoid strongly critical phrases like “Prior literature has failed to […]” or “Prior estimates are of little value because they […].”

Stating results without conveying why they are important or what they imply. Even if you’ve motivated your research question well, if your description of the results boils down to “We estimate that the price elasticity of demand is -0.05 in the short run and -0.75 in the longer run,” your reader will be underwhelmed. Make sure you state explicitly what your results imply for our understanding or for policy (for example, by adding “These results imply that it takes consumers a nontrivial amount of time to adjust their behavior.” Try to relate the interpretation of results to your motivation and contribution.

Omitting important details from the introduction or having unimportant details in the introduction. The right amount of detail is a hard balance to strike. On one hand, there is no need to discuss your data in great detail, mention every single result, or do a deep dive into your identification strategy. On the other hand, you also want to make sure that the reader understands what you did and is as convinced by it as possible given introduction space constraints. For example, if you use an instrumental variable, you need to mention what that instrumental variable is in the introduction and briefly explain why it’s a valid one. If you anticipate certain obvious criticisms, you can also mention key robustness or placebo checks in the introduction, but leave less obvious ones for the body of the paper (A good generic phrase is “Our results are robust to a battery of robustness checks, including [list a few examples here]”).

Having too many footnotes. This is a bad idea for any part of the paper, but it’s especially bad for the introduction, which is supposed to provide your reader with a quick, accessible, easy-to-read, and exciting summary of your paper. Footnotes slow the reader down and complicate the text. An average of one footnote per page or less is a good rule-of-thumb in general and for the introduction I would consider if you can get rid of footnotes completely. Consider whether each footnote can be deleted, moved to the text, or moved to a different part of the paper. The most appropriate place to have a footnote in the introduction is the literature contribution paragraphs, where you may want to mention tangentially related literature in a footnote.

Discussing things that are not relevant to the research question at hand. I was personally guilty of this one in my job market paper, when I discussed the literature on the optimal generosity of unemployment insurance over the business cycle. My paper was about how hurricanes affect local economies, and the only connection to unemployment insurance was that I showed increased expenditure on such programs as a result of hurricanes. Make sure everything in your introduction is directly related to what you do.

Using passive/third-person voice. It’s ok to say “This paper” instead of “I” or “We” occasionally, but maintaining such third-person language throughout the paper will diminish its impact. Passive voice (where “regressions are estimated” and “results are discussed”) has an even larger negative impact. While there are disciplines where third-person language is expected, in economics first-person language is generally encouraged.