First in a series of analyses of private-prisons in America ● Jul. 16, 2014 ● View other tutorials and examples on my blog.
By Carl V. Lewis ● Jul. 16, 2014
While much of the national debate over the mass privatization of state and federal prison systems in recent years has focused on cost per inmate – whether private prisons save public funds or are costing taxpayers more, as most economists believe – few mainstream narratives have surfaced concerning other, less quantifiable effects of prison privitization. For example, a helpful start might be to see if any correlation exists between incarceration rates in areas with private prisons and areas with public prisons. As it turns out, five companies alone have a monopoly on the private prison industry, each raking in millions of dollars of annual revenue in the business of incarceration.
In the choropleth, color-blind friendly map above, state colors indicate state incarceration rates, which were calculated by the number of incarcerated individuals divided by the total state population estimate for 2012. The darker red a state is colored, the higher its incarceration rate. Hover over any state to see more detailed information. The colored points overlaying the states indicate the locations of all 137 state-contracted private-prisons in the U.S., with different colors representing which of the five private-prison megacompanies owns the jail. Zoom in to get a closer view of the desnsity of locations.
The map visualizes two different datasets – 2012 incarceration rates by state and the locations and owners of all 137 private prisons nationwide. A preliminary analysis of these two variables taken together would, in fact, lend credence to the claim that private prisons incarcerate more people on average than their public counterparts do, presumably because they have a greater financial incentive to keep prisoner occupancy rates high becausee Zoom in to the map to the regions with the highest density of locations. What do you notice about those states?
Of course, that doesn't necessarily mean other confounding variables couldn't be at least partially responsible for the correlation, which although for this explanatory variable is fairly high at .72, certainly doesn't prove causation. There could very well be deeper systemic issues at work in the justice system or social attitudes in states with high prison populations, which tend to be concentrated in the southeastern part of the country and especially Texas. Or, it could always be a chicken-and-egg scenario; high crime rates and prison populations in those states may make it more efficient for cash-stapped state-budgets to outstource imprisonment.
Applying multiple linear regression using Python, as Gavin Hackeling has recently taught me how to do, I should be able to increase the correlation rate closer to the point of cautious causation. In my next post, I'll go over step-by-step this process, so that other data journalists can bolster the accuracy and meaning of their visualizations based upon objective statistical modeling.