Methodology

How did we get these numbers?

Data for this tool came from the 2012 National Corrections Reporting Program (NCRP), which is the most recent year NCRP data are available. The 2012 NCRP includes data from 47 states. For the Prison Population Forecaster, we included states that (1) provided at least seven years of data (i.e., 2006–12), (2) had no or very few missing data, and (3) broke down their data by admission type (e.g., new crime versus revocation) and primary offense type (e.g., drug, property, violent, other).

After removing states that did not meet these criteria, we were able to include 15 states in the forecaster. Most of the other states were excluded as a result of missing or otherwise insufficient data. The 15 states included in the tool represent nearly 40 percent of the total US population of state prisoners; are of varying sizes; and have prison populations that are geographically, culturally, and demographically diverse.

The projections in this feature are based on trend analyses conducted by Urban research staff. We examined historical trends (giving more weight to recent trends) in each state’s prison stock population, admissions, and inmate lengths of stay. We then projected states’ baseline scenarios—that is, what their prison populations would be if the same trends were to continue through 2021. Finally, we altered (1) the number of admissions and (2) the lengths of stay during the same period of time to project example policy scenarios.

As a measure of quality control, we compared our projected numbers with actual population numbers provided by each state’s department of corrections (e.g., population counts in 2014), as well as projected future population numbers provided by state departments of corrections or other state forecasting agencies. We found that our projections were generally within a small margin of error (plus or minus 5 percent) of the state-provided numbers. In the few cases where our numbers had greater margins of error, we adjusted the forecasted baseline scenarios to bring them into closer alignment with the state-provided numbers. We tended to estimate more conservative population growth, or greater population declines, than what was projected by the states.

Although the projections are based on historical trends in each state, it is important to note that interpretations of these trends are subjective. We were not able to account for any of the major changes that have occurred in states since 2012, or those that will occur in the future. It is unlikely that our projections will mirror the actual changes that will occur in the states’ prison populations between 2012 and 2021. Thus, the projections should be used for illustrative purposes and for informing the conversation about how to most effectively reduce the size of state prison populations.

For further information on the methodology or the underlying algorithms employed, please contact Bryce Peterson.