The perils (and necessity) of forecasting how many people will be in jail
There’s an old saw often attributed to Yogi Berra: “It’s tough to make predictions, especially about the future.” Yet forecasting is something we do every day of our lives as we anticipate what consequences our actions will have. We also depend on other people, businesses and organizations to make reasonable predictions. And when the meteorologist predicts a blizzard that results in us opting out of studying for that exam, we get upset when not an inch of snow falls.
The challenge of forecasting is front and center within the criminal justice system, including forecasting future aggregate crime rates, criminal behavior and — in my own research with James Alan Fox — population demographics and homicide trends. It is also especially important among those concerned with housing people in jails and prisons. Correctional administrators must make educated predictions about how many people will be in jail and for how long in order to have the necessary bed space so as to not exceed capacity, to ensure adequate staffing and to provide the necessary services.
How many people we forecast the jails will need to hold over the next decades is essential in deciding what and how big we build. As New York City readers of this essay are well aware, this is a pressing matter as the city moves to shut down Rikers by 2027, replacing it with a four-borough jail system designed to handle a total of 3,300 people. Whatever the perils of forecasting, it is a necessity to have some idea of the range in the numbers of people the city’s jails may hold over the next decades so that, as governments invest billions in building new corrections infrastructure, there is sufficient space for decent housing, adequate staffing and necessary programming.
Many readers may not understand or fully appreciate all that goes into forecasting jail populations. Let’s take a look.
The first set of inputs into a jail population comprises three sets of interrelated things: first, how much crime leads to arrests which lead to jail; second, the volume of persons entering jail either through jail sentences and/or pretrial detention rather than some sort of diversion and/or pretrial programming; and thirds, how long they stay once incarcerated. We must bear in mind that the total number of those sent to jail is not only about pretrial incarceration, meaning persons who cannot pay bail or are remanded while awaiting court proceedings; it’s also a function of the number of persons who are sentenced for offenses where the sentence is by law less than a year. Jail populations comprise what criminologists call “stock” and “flow,” with the former being the number of persons in jail at any given time and the latter being the number of persons who are admitted to jail at any given time. Put more simply, population is a function of who goes in and how long they stay.
Whatever the perils of forecasting, it is a necessity to have some idea of the range in the numbers of people the city’s jails may hold over the next decades so that, as governments invest billions in building new corrections infrastructure, there is sufficient space for decent housing, adequate staffing and necessary programming.
A second set of inputs would be how laws and policies influence who gets jailed and for how long. Oftentimes, policies are made that may affect the length of stay two months or two years down the road in ways that a jail administrator may not be able to anticipate. But this is not an easy calculation to make, and it is not linear. New York State’s bail laws, for instance, drove significant reductions in pretrial incarceration — and backlash to those laws resulted in rollbacks that may have altered numbers further.
A third input would be a world-altering event, like the COVID-19 pandemic, that could significantly alter how court systems and jails operate. As we all know, the initial lockdowns and shutdown months after the onset of the pandemic in the U.S. slowed the criminal justice process such that jail populations declined significantly — only to rise thereafter, as my research in Miami-Dade showed. In New York City, the pandemic clogged court proceedings and prompted officials to release many incarcerated individuals. While no one can guess when the next pandemic might arrive, administrators now must be aware that it could happen at any time, requiring them to plan for a worst-case scenario to figure out what will be needed right away with respect to temporary holding space, adequate staffing and any type of medical needs.
Fourth, we must understand that it’s not as though we’re plugging these inputs into a well-established mathematical equation that yields one and only one answer. We have a range of forecasting methods and processes. As William Sabol and Miranda Baumann noted, there are many ways of conducting these projections and none of them are perfect. For example, while we can pretty well project the estimated size of the population that may be in their age-prone crime years, say mid- to late adolescence, that still does not always consider how social changes may affect offending patterns, both in type and level over time, as Robert Sampson and L. Ash Smith have observed.
If policymakers project lower jail populations, they might force consideration of reforms that might not otherwise be on the table. This is a high-risk, high-reward proposition.
Forecasts do not only tell policymakers and the public how many beds are required, but how many people they need to run facilities and how much funding to provide the necessary services. More people in jail may necessitate a budget increase, which means adjusting budget projections accordingly. With more population, there is a need for more services — especially educational opportunities, because in their absence, there may be an increased risk of misconduct or violence within facilities as well as a higher risk of recidivism.
Some might view forecasts as a lever to force policy changes. If policymakers project lower jail populations, and therefore build facilities that will only accommodate those smaller targets, they might force consideration of reforms that might not otherwise be on the table. This is a high-risk, high-reward proposition. Forecasters should play it straight, aiming for as much accuracy as possible. Given all the uncertainty already baked into the process, that’s difficult enough.
Some years ago, my colleagues and I prepared a report to help the Broward County, Florida, Sheriff’s Office manage future jail populations. In it, we provided a 10-year jail population forecast, conducted a cost-benefit analysis for jail alternatives compared to incarceration and assessed the level of predictive accuracy and validation of a risk assessment tool used to inform pretrial release decision-making.
In our forecast, using several different models, we anticipated a jail population of 4,745 jail inmates by 2020. That was higher than what turned out to be a jail population of between 3,700 and 3,900 in 2019, depending on when the count was obtained, to just over 4,000 in the first quarter of 2020 prior to the onset of the COVID-19 pandemic.
Was being on the high side a good thing or a bad thing? That is for policymakers to decide, but it is worth noting that our estimates were not too far off. Had we underestimated by 500-1000, then there would have been significant overcrowding.
There is no getting around the need to forecast. Much like hurricanes, we are good at predicting the next few hours and the next day. But as the days go out, the cone of uncertainty gets larger. As was the case with Hurricanes Ian and Charley in Florida, last-second changes in wind direction and speed combined with warmer waters can alter the shape of history, a coastline and its people. But predictions for hurricane landfalls that are 48 hours away have been cut in half, and as our methods continue to improve, forecasts in criminal justice generally, and among jail populations, will improve. Notwithstanding the wisdom of Yogi Berra.