As a starting point, the incarceration rate for the U.S. is vastly higher than for the usual comparison group of high income countries--the number of inmates per 100,000 population is about six times higher in the U.S. than the average for OECD countries.
The U.S. has historically tended to be above-average in its incarceration rates, but this extreme outlier status is relatively recent. Back in the 1970s, for example, the U.S. incarceration rate was in the range of 250 per 100,000 people. By about 2005, the incarceration rate had more than tripled.
Of course, a tripling of the incarceration rate means roughly a tripling of the costs of running prisons, too--with a total cost now at about $80 billion per year.
There seems to me solid evidence that the rise in U.S. incarceration rates in the 1980s and 1990s are part of what helped to bring down the crime rate during that time. But the social costs of imprisonment are much larger than the bills paid by government: they include the reduction in future income earned after being released from prison, the costs of loss of freedom to those who are imprisoned, and costs that absence imposes on families and friends. With the monetary and nonmonetary costs in mind, it seems to me that the U.S. has been resorting to imprisonment far too easily for those who have committed non-violent offenses. If the goal is to continue the reduction in crime rates, finding a way on the margin to spend less on prisons and more on police is probably a productive tradeoff.
For some earlier posts on U.S. rates of imprisonment, see "Reducing the Federal Prison Population," (November 8, 2013), "Too Much Imprisonment" (November 30, 2011) and "U.S. Imprisonment in International Context: What Alternatives?" (May 31, 2012).