The instructor of my freshman physics laboratory hammered on one point: the abstract is the most important part of a lab report. The same is true in any discipline, and applies equally to a journal article, a proposal, or a report to a granting agency. Your abstract will be read by ten or twenty times as many people as will any other words in the article. So, if you want to make a positive impression, or just convey information, here's where to really pay attention to writing!
The purpose is not to tell the reader you did something: it's to tell her what you did in the simplest, most informative way possible. Too many abstracts begin: "This article investigates the determinants of drug use." With just a few more words you can tell the reader what you are actually doing: "I use a model of rational addiction to show how institutions and tastes affect drug consumption." Similarly for the concluding sentence. Instead of: "The results show that choice among drugs depends on the institutional framework and taste parameters," how about: "The results show that the substitution of designer drugs for marijuana increases with media coverage of drug issues and with the propensity to schizophrenia."
Should you hold back your punch line to increase suspense? It's tempting, but an economics article is not a Sherlock Holmes story. Readers are too busy to appreciate the excitement of your little mystery tale: they want quick information. If it suits them, they will read on for the details, appreciating that you've already made it easier for them by explaining where you're headed.
Should you display your copious bibliographic knowledge in the abstract? Not necessary; it's easy to flip to your bibliography. If your article is specifically an advance on one or two published works, it's fine to cite those.
The introduction and conclusion are the next most likely parts of your document to be read. Think of these sections as extensions of the abstract, carefully coordinated to entice the reader further and further into the details of your work. They can do this by amplifying the core intuition in a slightly more leisurely manner-but not if your abstract was just a verbatim repeat of parts of these sections. (And not if the conclusion merely states that all depends on lambda and the coefficient of EGGSBRAK, the definitions of which are hopelessly buried in sections 4.1 and 6.2.3(a)-but that's another essay.)
Abstracts for submission to meetings serve a somewhat different purpose, because the reader doesn't expect the results to be all worked out. Here it's more important to focus on the significance of the idea and how you will improve on existing knowledge. So a few citations are warranted, enough to identify the literature to which you are contributing and to explain how your approach differs from others. If you have a great new data set, say how it is compiled. If you can, explain what conclusions may emerge and what implications they would have.
I often write or revise abstracts last. The ideal time for me is after working through the entire article or proposal. Not immediately after, when I'm tired and may be tempted to dash off something quick; but just long enough for it all to percolate and brew up a clear vision of what I've accomplished. If such vision appears, I just do my best-or possibly decide the article really lacks coherence and needs yet another thorough overhaul! If the vision does appear, I try to capture a good snapshot for the harried potential reader, hoping at least one of us will benefit.
* This article was originally published as: Small, Kenneth A., "How to Write an Abstract," American Economic Association Committee on the Status of Women in the Economics Profession Newsletter (Feb. 1997), p. 22.