An old paper by Akerlof modeled procrastination in a simple way.  Suppose that doing a task today involves costs that are more salient today than the costs that would be incurred from doing the task tomorrow.  (Hyperbolic discounting is the modern way of introducing this disparity.) And suppose you are naive in that you are bad at predicting your behavior in the future.  Then every day you will make an optimal plan for when to do the costly task and every day your optimal plan will be to do it tomorrow.  So it will never get done.

Unless there is a deadline.  The day of the deadline you will know that the task will not get done tomorrow if you don’t do it today, so that is the day when you will finally do it. Ryan Sager describes an experiment with a similar finding.  In fact the same effect holds with enjoyable tasks, not just pure costs.  (Jennie and I never visited many of the tourist sites of San Francisco until a month before we moved away from the Bay Area.)

Take an experiment conducted by two marketing professors, Suzanne Shu and Ayelet Gneezy, set to be published in a forthcoming issue of the Journal of Marketing Research (and written up in the May Atlantic by Virginia Postrel). Shu and Gneezy gave 64 undergraduates coupons for a slice of cake and a beverage at a local French pastry shop. Half got a certificate that expired in three weeks, half got one that expired in two months. While students were sure they were more likely to use the certificate with the more generous timeframe, the results were clear: The shorter timeframe made the students much more likely to redeem their certificates; 10 of the 32 students (31%) redeemed the three-week certificates, only two of the 32 students (6%) redeemed the two-month certificates.

You can imagine that the coupon is lost with some probability every day until it is used.  Thus, the longer deadline means less chance of getting to the moment of truth.  Here is a link to the original research.