Richard Dawkins himself is by now overexposed, but in his quest to convert the world to the atheist meme he, along with most of the rest of the world, has neglected his own theory of the extended phenotype. Like the metaphor of the selfish gene, this is an insight with many very important consequences. So far we've just scratched the tip of this iceberg.
Yoram Barzel is one of the most important economists of the last half century, but few people know this. Along with the much more famous Friedrich Hayek, Barzel is probably the economist who has influenced me the most. He has had a number of important insights related to transaction costs which I've confirmed repeatedly in my own observations and readings in economic history. Probably his most important insights relate to the importance and nature of value measurement. (“Measurement Cost and the Organization of Markets,” Journal of Law and Economics, April 1982 Reprinted in The Economic Foundations of Property Rights: Selected Readings, SPejovich, ed., Edward Elgar Publishing Company, 1997. And in Transaction costs and Property rights, Claude Menard, Editor, Edward Elgar publishing Company, 2004.) Another is his analysis of waiting in line (queuing) versus market prices as two methods of rationing. Waiting in line is the main means of rationing in a socialist economy but is surprisingly widespread even in our own. (“A Theory of Rationing by Waiting,” Journal of Law and Economics, April 1974 Reprinted in Readings in Microeconomics, Breit, Hochman, and Sauracker, 3rd ed., 1986).
Advanced cryptography over the last twenty-five years has come with a large number of important protocols. Public key encryption got all the hype, but it was just a warm-up. Alas, the perceived esoteric nature of the field and magical function of its protocols has apparently intimidated regular programmers away from implementing most of them. Appearances to the contrary they are based on solid math, not magic. I've blogged on some of these, but there is far more out there waiting to be picked up by mainstream computer security. Of all the advanced cryptography protocols secure time-stamping and multiparty secure computation are probably the most important.
The ideas of voluntary oblivious compliance and admonition systems from the capability security community. These are very useful in elucidating interelationships between wet protocols (e.g. manual procedures and law) and security. (By contrast I think capability security itself, and indeed the entire area of access control, is over-hyped: there are important roles for such things, for example in internal corporation access control, secure operating systems, and services where trust by all parties in corporate systems administrators is satisfactory, but they are nowhere close to being the be-all and end-all of distributed computing as often seems to be implied. Rather it's advanced cryptography where the most important advances in distributed security lie).
The whole topic of franchise jurisdiction, crucial to understanding medieval and Renaissance English law, has nevertheless long been almost completely neglected. It is also a very different paradigm of the power of courts that has been forgotten by the modern legal and political communities in favor of the Roman model of governmental sovereignty.
Almost any topic that requires knowledge from more than one academic specialty has been overly neglected: so much so that most such topics don't even exist but should.
The specific ideas mentioned are just the tip of the iceberg of underappreciated topics. Please let me know what ideas you think have been the most underappreciated. There are, I'm sure, many, many ideas out there that I should know about but don't.