Thursday, June 14, 2012

Why LIFO beats FIFO

When I write a blog post, it is usually about the last paper I read. I draw from a pile that goes by the LIFO principle, "last in, first out." That guarantees that my "stories" are super-fresh, but when there is nothing I go down the pile. I could use the FIFO principle, "first in, first out," which would give each paper a fair chance to be featured, but then I may end up with papers that are always older, instead of just occasionally (there are several dozen papers on the pile). Well, it turns out my principle is in fact better than FIFO (well not exactly, my case is different, but anyway).

Trine Tornøe Platz and Lars Peter Østerdal model a situation where people queue for service (say, boarding an airplane), where there is a bottleneck but it is open at all times. Agents then decide to queue depending on the way they are served. Suppose the cost of waiting is linear in time, people like to be served early, and everyone can be served. Then it is better to implement LIFO than FIFO. Indeed under FIFO, there is no reason for people to wait, they all come at the earliest possible time and have to wait the longest possible period. Under LIFO, there is less of an incentive to come early, indeed the first ones are not served first if anybody arrives right after. Waiting time is then reduced and everybody is better off, in expected terms.

No comments: