Excellent Economist article on over-imprisonment in America (though it sort of mixes this issue up with the separate question of vaguely worded federal criminal statutes, which I think is a different problem):
Massachusetts is a liberal state, but its drug laws are anything but. It treats opium-derived painkillers such as Percocet like hard drugs, if illicitly sold. Possession of a tiny amount (14-28 grams, or ½-1 ounce) yields a minimum sentence of three years. For 200 grams, it is 15 years, more than the minimum for armed rape. And the weight of the other substances with which a dealer mixes his drugs is included in the total, so 10 grams of opiates mixed with 190 grams of flour gets you 15 years. [...]
Prosecutors may charge him with selling a smaller amount if he agrees to “reel some other poor slob in”, as Ms Dougan puts it. He is told to persuade another dealer to sell him just enough drugs to trigger a 15-year sentence, and perhaps to do the deal near a school, which adds another two years.
Severe drug laws have unintended consequences. Less than half of American cancer patients receive adequate painkillers, according to the American Pain Foundation, another pressure-group. One reason is that doctors are terrified of being accused of drug-trafficking if they over-prescribe.
The result is a terrifying wasting of financial resources and human potential. Sticking to tradeoffs inside the realm of crime prevention, it would clearly make more sense to increase the quantity and quality of police officers and parole/probation supervisors than to be handing out these endless jail sentences. Even for legitimately serious violent crimes, it’s more important to catch and prosecute people quickly and effectively than to lock people away forever and ever. In formalistic terms, a 50% chance of a 2-year sentence and a 2% percent chance of a 50-year sentence are the same thing. Real people, however, discount the future. And the sort of people who commit violent crimes—young men with poor impulse control—are especially likely to do so.
Mark Kleiman’s When Brute Force Fails makes these points (and more) and is by far the best thing I’ve read on the subject of crime and crime control in 21st century America.