HD36 - Virginia's Correctional System: Population Forecasting and Capacity

  • Published: 1985
  • Author: Joint Legislative Audit and Review Commission
  • Enabling Authority: Appropriation Act - Item (Regular Session, 1984)

Executive Summary:
Recently, Virginia's correctional system has undergone a period of steady growth. New facilities have been constructed to accommodate an increase in inmate population. Two medium-security facilities were added in FY 1983, another has opened in FY 1985, and the final facility that is currently planned will open in April 1986. The correctional system planners must now begin to examine the sufficiency of the system's capacity to house the inmate population expected in the future.

This report, the second in a series on corrections in Virginia, addresses two fundamental components of any viable plan: the capacity of the system and the inmate population forecast. The body of the report concentrates on an evaluation of the Department of Corrections' method for determining the capacity of the correctional system, and the department's method for forecasting inmate population. Before discussing these evaluations, this summary will provide an overview comparing inmate forecasts with capacity.

Alternative Scenarios for the Correctional System

Previous DOC forecasts indicated that the inmate population would reach as high as 15,000 by 1990. Recent analysis, however, indicates that the inmate population growth evidenced in the seventies has slowed. After evaluating technical aspects of forecasting and capacity calculations, JLARC staff corrected technical errors and refined DOC's methods to develop alternative scenarios for the correctional system in 1990.

After the interim briefing on this report (November 1984), the JLARC and DOC staffs worked cooperatively. During that period, new data became available, and DOC implemented most of the recommendations in the report. While a few possible refinements are still being considered for later inclusion, DOC now has a technically adequate methodology.

The revised DOC forecast methodology projects 11,225 inmates by 1990. This represents a growth of more than a thousand inmates over five years. As with any forecast, of course, this forecast assumes that conditions that existed in the correctional system in the past will continue to hold.

Several factors related to this leveling off of the expected growth have been cited in other states and are occurring in Virginia. The crime-prone population is growing more slowly since the "baby boom" children have passed out of this age group. Virginia's economy is expected to be strong, keeping unemployment and the tendency to commit crime low. Changes in statutes, such as increased serving time before parole eligibility for recidivists, have apparently been offset by other actions such as changes in the good conduct program and the Parole Board administrative changes that reduce serving time.

The forecast can be combined with some alternative configurations of the system's capacity to yield several possible scenarios for the next five years. Three possible scenarios are discussed below, and each is illustrated graphically in an adjacent figure. These scenarios are illustrative of the possible future direction for the system; they do not exhaust the range of alternatives.

SCENARIO 1

The Department of Corrections currently employs a measure of capacity called "operational capacity." Operational capacity is defined as the level of occupancy at which the facilities can be safely operated. The measure generally includes one inmate per cell, some number of special purpose (medical, isolation, segregation, etc.) beds, and multiple beds in dorms. While some inconsistencies exist in the way the capacity is calculated in different facilities (see the next section on the evaluation of capacity for detail), the measure does reflect DOC's desired operating level.

The operational capacity is plotted in Figure 1 (shown on page II of the report). The graph shows an increase in the capacity for FY 1986 due to the opening of the new facility at Augusta. All four of the recently constructed medium security institutions (MSIs) are included, at 500 beds each. The operational capacity also reflects DOC's current plan to close only the "A" building of the Penitentiary during this time frame.

The comparison of operational capacity and the revised DOC inmate population forecast indicates that a bed space shortage will occur by 1990. The capacity at that time will be 9,791 beds, while the population is forecast to be 11,225, leaving a shortage of 1,434 beds. That year has the largest shortfall in this scenario; it occurs because the population is gradually increasing while capacity remains level after FY 1986.

SCENARIO 2

DOC's definition of operational capacity does not include double-bunking at the MSls, and contains an inconsistent amount of space for each inmate in dormitory areas, especially in field units. JLARC therefore developed another capacity term, "reserve capacity." Reserve capacity involves the use of a level of double-bunking that has been achieved in the MSls in the past and the use of a maximum amount of space per inmate in dormitory areas.

Maximum use of the reserve capacity would involve the addition of over 1,300 beds to operational capacity. However, this level of use is not considered feasible for planning purposes. The maximum reserve capacity allows a short-term solution for unforeseen circumstances and probably could never be fully utilized.

A moderate use of reserve capacity, however, could be used for planning purposes, especially when the inmate population and the capacity are reasonably close. JLARC used these assumptions to develop a second scenario, employing the same population projection as in the first. Figure 2 (shown on page III of the report) shows the bed space comparison if part of the reserve capacity were included. The capacity shown in FY 1990 could be realized by double-bunking the four MSls for a total capacity of 615 inmates each, and by using a guideline of 55 square feet maximum per inmate in the field units, and 70 square feet maximum per inmate in major institutions. These changes would add 514 beds to the operational capacity in 1990. The use of a portion of the reserve capacity reduces the bed-space shortage to approximately 900 beds in 1990.

SCENARIO 3

Scenario 3 offers a look at the correctional system with another assumption altered. The "A" building at the Penitentiary is currently scheduled to close by June 1986. The loss of 316 beds is reflected in the operational capacity. The Appropriations Act indicates that "B and C buildings" should be scheduled for closure by FY 1990. The third scenario uses the partial reserve capacity, but removes part of the remaining 552 Penitentiary beds beginning in FY 1989 and completes the closure in FY 1990.

The result of making this change is shown in Figure 3 (shown on page IV of the report). With the partial use of reserve capacity, the bed spaces would fall by 552, yielding a net deficit of 1,472 bed spaces.

These three scenarios illustrate three possible outcomes for the correctional system in FY 1990. While all of the scenarios indicate a bed-space shortage by 1990, the magnitude of the shortage is different. There are several alternatives for meeting the shortages, some involving the expansion of capacity and some the reduction of inmates. Continuing the use of the Penitentiary beyond 1990, reestablishing an institution at Deep Meadow, expanding one of the current facilities, or building a new facility are options for increasing capacity. Increased use of community corrections, sentencing guidelines, or a cap on the number of inmates are alternatives used in other states to limit prison population. An evaluation of the full range of alternatives should be conducted by DOC.

Evaluation of the Correctional System's Capacity

DOC'S operational capacity has generally been determined by departmental judgments about what capacity should be for each facility, rather than by empirical evidence. The judgments are based on the department's considerations for maintaining safety in the new facilities. The judgments do not reflect the maximum capacity which might be achievable during periods of heavy demand for extra inmate housing.

The department's approach to operational capacity has important consequences. First, the lack of a consistently applied method has meant that there is significant variation in the amount of space which is provided per inmate in dormitories, even in similar housing units. For example, the average amount of space per inmate in large permanent field units ranges from 47.9 square feet in Caroline to 69.7 square feet in Tazewell.

Second, the fact that operational capacity reflects DOC judgments about desirable, safe operating levels has meant that these figures in many cases do not reflect the levels at which many facilities have been operated. In 10 of the 18 major institutions, for example, the actual average daily population exceeded operational capacity for at least one month during FY 1984. The most dramatic examples were Buckingham and Brunswick, where the actual average daily population levels for June of 1984 were over 690, yet the operational capacities were 500. The reason for this major difference is that DOC does not include the double-bunking of single occupancy cells in its operational capacity, even though double-bunking is being done. At Buckingham and Brunswick, approximately 40 percent of the general population cells are currently double-bunked.

To supplement DOC's operational capacity, JLARC developed an alternative measure of capacity called reserve capacity. Reserve capacity reflects how far capacity can be reasonably increased to accommodate short-term population increases or forecast errors. JLARC's first step was to identify some guidelines which could be used to identify potential reserve capacity. In the new medium-security institutions, a level of double-bunking which had already been achieved was included as part of potential reserve capacity. For identifying a potential reserve in dormitories within major institutions, 60 square feet per inmate was set as a maximum amount of space, and 50 square feet per inmate was used for field units. The guideline for field units, for example, indicated a potential reserve of 237 beds over operational capacity.

The second step in the process was to estimate a portion of the potential reserve capacity which might be used for capacity planning purposes. To identify this partial reserve, additional capacity factors were considered in interviews and observations at a number of facilities. Factors considered included: the availability of floor space for additional beds; potential security problems; work, educational, or recreational opportunities for inmates; and the capacity of support facilities, such as freshwater and waste-water capacity or food services. The fieldwork indicated that for some facilities, additional resources might be necessary to achieve additional capacity.

Table 1 (shown on page V of the report) shows operational, potential reserve, and partial reserve capacities for DOC facilities in FY 1985. The partial reserve reflects a 25 percent level of double-bunking at the medium security institutions. JLARC recommends that DOC review the capacity issue and contribute its perspective to defining reserve capacity for the system..

Recommendations 1-2. DOC should report a figure for the reserve capacity of its facilities to supplement operational capacity. Reserve capacity should reflect how far capacity can be reasonably increased to accommodate short-term population increases or forecast errors. Reserve capacity should be based on double-bunking of the cells in facilities like Buckingham and Brunswick. Reserve capacity should also be based upon precisely defined guidelines for DOC facility ward areas, such as square footage per inmate.

Evaluation of DOC's Forecast Methodology

The current methodology used by DOC to forecast inmate population was originally developed within the Florida Department of Corrections. The methodology was adapted for use in Virginia by consultants and DOC staff. The adaptation of the Florida model, dubbed SLAM II (Simulated Losses and Admissions Model II), began in spring of 1983, and the first forecast was produced by December.

The December forecast predicted 2,500 fewer inmates by 1990 than the previous forecast. Given the lower projection and the spate of recent corrections construction projects that were nearing completion, the General Assembly asked DOC for a short-term forecast and a description of the methodology employed to make the forecast (HJR 152). Also, the General Assembly requested in the Appropriations Act that JLARC examine the forecast and its impact on staffing and prison design.

The current forecast methodology consists of three components: admissions, releases, and manual adjustments. The admissions component forecasts the number of admissions that DOC can expect annually. The release component computes a probability of stay for each inmate currently confined or expected to be confined in the system, and the probabilities are then summed to yield a forecast of the number of inmates incarcerated in each month in the future. Manual adjustments attempt to correct the forecast for legislative and administrative changes that have occurred since the data base for the other components was established.

Admissions Component. The admissions component is based upon a sound methodology; however, there are several technical problems with the way the method is carried out. First, admissions are inconsistently measured during the forecast period. An attempt has been made to estimate the number of felons backed up in local jails, but the data cannot be verified. Furthermore, an estimate of parole violators is included as an adjustment to the data. Neither the need for the adjustment nor the method of making the adjustment appears to be justified.

Recommendations 3-6. Admissions should be measured from existing, verifiable data sources. Another method for estimating the number of felons backed up in local jails should be developed separately from the current admissions component.

Currently, two separate equations are used to predict admissions, one for whites and one for non-whites. Recent increases in the number of females incarcerated in the State system, and technical problems with the inclusion of females in the equations using predictor factors based upon the male population, indicate the need for a separate equation for females.

Recommendation 7. Females should not be included in the admissions equations which are forecast by male admissions. A separate equation should be developed for female admissions.

The admissions equations used by DOC include the crime-prone population and unemployment as predictors of admissions for the first two years of the forecast. After that period, unemployment is dropped from the equations and admissions are forecast by crime-prone population alone. Leaving out unemployment removes the effect of a factor that provides a linkage between economic conditions and incarcerations.

The omission of unemployment from the equations further exacerbates technical problems in the admissions component. The problems can be related to the omission of other relevant variables, such as crime rate and commitment rate. Using these factors should be explored by DOC as a long-term solution to the technical problems. In the short term, incorporating the number of admissions in the previous year will partially account for the omissions and improve the statistical properties of the model.

Recommendations 8-12. The admissions component should use crime-prone population, unemployment, and the previous years' admissions to forecast admissions throughout the forecast period. The effect of different expectations for unemployment should be routinely examined and reported. Other factors, particularly those related to the commission of the crime and the adjudication process, should be tested to see if they would improve the forecast.

Release Component. The release component of the forecasting model uses the amount of time served by inmates confined and released during FY 1982 to estimate the time to be served by inmates throughout the forecast period. The model assumes the inmates' sentences and serving times in this period will be the same in future years. Analysis of sentencing patterns over the past six years indicates that sentencing patterns have changed; therefore this change should be accounted for in the release component.

The complexity of the component may contribute to a significant problem in the future. The component is not flexible enough to easily make changes in the assumptions and look at the impact on the forecast. The error rate is difficult to estimate and the documentation is not sufficient to replicate or update the component. This creates a high risk of error in producing the forecast.

The complexity and structure of the component restrict the ability to estimate the impact of administrative and legislative changes on the inmate population. The need for five manual adjustments to the model attests to this problem. The lack of flexibility of the component makes it necessary to have two forecast approaches, one for the official forecast and one for legislative impact statements. The two are inconsistent in their assumptions and are not used to cross-check one another.

Recommendations 13-20. For the short-term, the release component should be updated by DOC to include the most recent years' data and to modify the assumptions for sentence distribution to more closely resemble the recent past. This should be done before significant decisions are considered regarding building new facilities.

By the end of the 1985, the release component should be replaced with a simpler, more flexible forecast model. Data and resources needed to update the component should be available to produce and publish a forecast update by November of each year. The forecast should indicate the impact of anticipated policy changes or the sensitivity of the forecast to changes in the model's assumptions.

Manual Adjustments. Manual adjustments have been added to the results of the admissions and release components to account for five administrative and statutory changes:

• recidivist serving time requirements,
• Parole Board administrative actions;
• firearm legislation;
• the Community Diversion Incentive Act;
• three serious offenses law.

The net effect of the adjustments is to reduce the forecast through 1990.

Recommendations 21-37. The manual adjustments should be incorporated into the other components of the model as soon as data is available. The actual impact of the changes should be estimated after they have been in place for a full year.

Model Performance and Maintenance.

The final forecasting issues in the report deal with the performance and maintenance of the model. If a model is to be useful for planning purposes, it must produce accurate results. For the model to continue to be reliable, it must be adequately maintained and updated.

The data which is available to date indicates that DOC's model has performed reasonably well. The department's revised admissions and population forecasts are shown in Table 2 (shown on page VIII of the report).

A simulation of the model's performance showed an average error of about three percent between July 1982 and May 1984. This is significantly better than the department's objective of keeping the error of the forecasting at less than ten percent.

However, there are several concerns about forecasting issues. First, DOC's performance objective of keeping error within ten percent allows for an error of up to 1,000 inmates. Second, a simulation of DOC's model indicates that it tends to consistently over-predict the population. This consistent over-prediction may weaken the credibility of the model. Third, there is a concern as to whether a high level of performance can be sustained with DOC's model. A long-term, high-level of performance is difficult to achieve with any model, as assumptions prove invalid and factors that are not accounted for begin to affect the system. This difficulty may be particularly acute with the DOC model due to the model's complexity. It is for this reason that DOC's efforts to maintain the model are also a performance concern.

The maintenance of a forecast model involves an ongoing effort to understand the theoretical basis of the model, the execution of the model, and the ways in which the model needs to be adjusted to account for changes. The key concern about the maintenance of DOC's model is whether or not sufficient priority will be placed on maintaining the model. A priority on maintenance was not placed on the initial DOC forecast model.

This report identifies several components which appear essential for the adequate maintenance of a forecast model. These components lead to several recommendations for the maintenance of DOC's model.

Recommendations 38-41. A single detailed source document on the DOC forecast model needs to be compiled. This document should contain all the material necessary to explain how to replicate and update the model. An ongoing commitment to forecasting should be made by DOC. Emphasis should be placed on anticipating factors which may change the forecast. To achieve this goal, it may be useful for DOC to involve participants from outside the department.