DataandMethods

2

Does Staffing Shortages Intensify the Harmful Effects of Overcrowding in U.S.

Correctional Facilities

Data and Methods

This paper examines the impact of staffing deficits and the negative impact of

overcrowding in American prisons. Through the national secondary data, it is possible to create

an evaluation of the situation in organizations over a broad range of institutions. The discussion

relies on publicly available datasets that are generated by the Bureau of Justice Statistics (BJS)

and the U.S. Census Bureau, which can be detailed in measuring the facility capacity, population

pressures, staffing patterns, and institutional harms. The combination of these sources provides a

complete understanding of the influence of structural constraints on safety and operations in jails.

Data Sources

The data analysis in this paper will rely on the Census of State and Federal Correctional

Facilities (CSFCF). The 2005 CSFCF is a very comprehensive source of information on the

facility features, capacity, staffing and program availability (Stephan, 2008)!Since the CSFCF

contains data on security level, inmate characteristics, and organizational characteristics, it can

be viewed as a central data to analyzing the organizational attributes of overcrowding and

staffing.

The analysis uses the BJS Prisoners in 2020 statistical tables to contextualize recent

trends in population and confirm interventions regarding capacity and custody counts by

showing significant decreases in the American prison population under the impact of the

C0VID-19 pandemic (Carson, 2021b). These data sets offer credible state-level indicators of

3

capacity, custody levels, and changes in patterns of imprisonment, helpful to supplementary

models used to test powerfulness.

The Mortality in State and Federal Prisons, 2001-2019 statistical tables provide detailed

information about the deaths by suicide, homicide, illness, intoxication, and accidents, which is

used to measure institutional harm (Carson, 2021a). The use of mortality patterns assists in the

capture of the extent of institutional strain that is severe and provides an outcome measure that is

attached to overcrowding as well as staffing.

Lastly, to complement the use of cross-referencing on the facility-level and enhancing the

quality of institutional identifiers, the paper uses the report by the U.S. Census Bureau on the

Coverage of Prisons and Detention Facilities in the 2020 Census. This dataset evaluates the

coverage of facilities and the accuracy of the classification, demonstrating that census

enumeration covered virtually all facilities identified by DOJ, but uncovered significant gaps in

small or specialized facilities (Garcia et al., 2024). The data aid in the proper connection of

facility attributes across sources. All the datasets which were used are for public use and do not

have any identifiable personal information.

Key Variables

Dependent Variables

In accordance with the BJS reporting standards, all harm indicators are then translated into

rates per 1,000 inmates to accommodate population variations (Carson, 2021b). There are a few

indicators of institutional harm:

Inmate-on-inmate assaults (CSFCF).

Inmate-on-staff assaults (CSFCF).

4

Critical negative events, such as suicide, murder, or drunkenness (Mortality in State and

Federal Prisons).

Operational disruptions, such as lockdowns or major rule violations (CSFCF).

Independent Variables

1. Overcrowding

Overcrowding will be measured as the percent of rated capacity filled:

Overcrowding = \frac{Average Daily Population} {Facility Capacity} x 100

Facilities with a greater than 100 percent capacity will be deemed overcrowded.

Sensitivity analyses will also be done in a continuous measure. This action is in line with

national records on documentation that indicate that facilities in certain states are running beyond

their capacity (Stephan, 2008).

2. Staffing Shortfalls

The increased ratios and vacancy rates show more serious staffing gaps. The shortfall in

staffing is determined by:

Vacancy rate of correctional officers (CSFCF).

The prison-to-staff ratio which has been complemented with state figures of Prisoners in

2020 (Carson, 2021).

3. Interaction Term

The main hypothesis is that the negative impact of overcrowding is enhanced by low staffing

levels. The following interaction term will be built:

Overcrowding x Staffing Shortfall

S

A significant positive coefficient would support the research hypothesis.

Control Variables

Control variables include:

- Security level

- Facility size and age

- Gender composition

- Geographic region

- Availability of treatment and rehabilitative programming (Stephan, 2008)

- Proportion of inmates convicted of violent offenses

Analytic Strategy

The effects of overcrowding will be assessed using a multivariate regression model to

determine whether staffing shortages increase the impact of overcrowding. Ordinary least

squares (OLS) models describe continuous outcomes, whereas count-based incident measures

are referred to negative binomial regression, which takes into account that BJS data show that

incidents in prisons and mortality are frequently over-dispersed (Carson, 2021).

Models will be estimated sequentially:

1. Baseline model with overcrowding and controls.

2. Staffing model adding staffing shortfall.

3. Full model including the interaction term.

6

Year fixed effects capture the effects of time variations in the prison populations, including

the variation related to the COVID-19 interruptions (Carson, 2021). These standard errors are

state clustered.

Missing Data and Quality Checks

Multiplex imputation is applied in cases where there is more than 5 percent missingness

and it is observed to be random. The distributional assumptions and multicollinearity are

evaluated using descriptive statistics and variance-inflation diagnostics. The Census Bureau data

consists of cross-validated facility identifiers, which is why the institutions are correctly aligned

(Garcia et al., 2024).

Ethical Considerations

Since every source of data is public-use and includes no identification information, the

project is not human subjects research and thus does not need the IRB.

7

References

Carson, E. A. (2021). Mortality in state and federal prisons, 2001-2019 — Statistical tables (NCJ

300953). Bureau of Justice Statistics. https://bjs.ojp.gov/content/pub/pdf/msfp0119st.pdf

Carson, E. A. (2021). Prisoners in 2020— Statistical tables (NCJ 302776). Bureau of Justice

Statistics. https://bjs.ojp.gov/content/pub/pdf/p20st.pdf

Garcia, M. M., Finlay, K., Speer, C. E., Willhide, E., Patti, K. N., & Loveless, T. A. (2024).

Coverage ofprisons and detention facilities in the 2020 Census. U.S. Census Bureau.

https ://www2 . census.gov/library/working-papers/2024/econ/coverage-of-prisons-and-

detention-facilities-2020-census.pdf

Stephan, J. J. (2008). Census of state and federal correctional facilities, 2005 (NCJ 222182).

Bureau of Justice Statistics. https://bjs.oip.gov/content/pub/pdf/csfcf05.pdf

2

Does Staffing Shortages Intensify the Harmful Effects of Overcrowding in U.S.

Correctional Facilities

Data and Methods

This paper examines the impact of staffing deficits and the negative impact of

overcrowding in American prisons. Through the national secondary data, it is possible to create

an evaluation of the situation in organizations over a broad range of institutions. The discussion

relies on publicly available datasets that are generated by the Bureau of Justice Statistics (BJS)

and the U.S. Census Bureau, which can be detailed in measuring the facility capacity, population

pressures, staffing patterns, and institutional harms. The combination of these sources provides a

complete understanding of the influence of structural constraints on safety and operations in jails.

Data Sources

The data analysis in this paper will rely on the Census of State and Federal Correctional

Facilities (CSFCF). The 2005 CSFCF is a very comprehensive source of information on the

facility features, capacity, staffing and program availability (Stephan, 2008)!Since the CSFCF

contains data on security level, inmate characteristics, and organizational characteristics, it can

be viewed as a central data to analyzing the organizational attributes of overcrowding and

staffing.

The analysis uses the BJS Prisoners in 2020 statistical tables to contextualize recent

trends in population and confirm interventions regarding capacity and custody counts by

showing significant decreases in the American prison population under the impact of the

C0VID-19 pandemic (Carson, 2021b). These data sets offer credible state-level indicators of

3

capacity, custody levels, and changes in patterns of imprisonment, helpful to supplementary

models used to test powerfulness.

The Mortality in State and Federal Prisons, 2001-2019 statistical tables provide detailed

information about the deaths by suicide, homicide, illness, intoxication, and accidents, which is

used to measure institutional harm (Carson, 2021a). The use of mortality patterns assists in the

capture of the extent of institutional strain that is severe and provides an outcome measure that is

attached to overcrowding as well as staffing.

Lastly, to complement the use of cross-referencing on the facility-level and enhancing the

quality of institutional identifiers, the paper uses the report by the U.S. Census Bureau on the

Coverage of Prisons and Detention Facilities in the 2020 Census. This dataset evaluates the

coverage of facilities and the accuracy of the classification, demonstrating that census

enumeration covered virtually all facilities identified by DOJ, but uncovered significant gaps in

small or specialized facilities (Garcia et al., 2024). The data aid in the proper connection of

facility attributes across sources. All the datasets which were used are for public use and do not

have any identifiable personal information.

Key Variables

Dependent Variables

In accordance with the BJS reporting standards, all harm indicators are then translated into

rates per 1,000 inmates to accommodate population variations (Carson, 2021b). There are a few

indicators of institutional harm:

Inmate-on-inmate assaults (CSFCF).

Inmate-on-staff assaults (CSFCF).

4

Critical negative events, such as suicide, murder, or drunkenness (Mortality in State and

Federal Prisons).

Operational disruptions, such as lockdowns or major rule violations (CSFCF).

Independent Variables

1. Overcrowding

Overcrowding will be measured as the percent of rated capacity filled:

Overcrowding = \frac{Average Daily Population} {Facility Capacity} x 100

Facilities with a greater than 100 percent capacity will be deemed overcrowded.

Sensitivity analyses will also be done in a continuous measure. This action is in line with

national records on documentation that indicate that facilities in certain states are running beyond

their capacity (Stephan, 2008).

2. Staffing Shortfalls

The increased ratios and vacancy rates show more serious staffing gaps. The shortfall in

staffing is determined by:

Vacancy rate of correctional officers (CSFCF).

The prison-to-staff ratio which has been complemented with state figures of Prisoners in

2020 (Carson, 2021).

3. Interaction Term

The main hypothesis is that the negative impact of overcrowding is enhanced by low staffing

levels. The following interaction term will be built:

Overcrowding x Staffing Shortfall

S

A significant positive coefficient would support the research hypothesis.

Control Variables

Control variables include:

- Security level

- Facility size and age

- Gender composition

- Geographic region

- Availability of treatment and rehabilitative programming (Stephan, 2008)

- Proportion of inmates convicted of violent offenses

Analytic Strategy

The effects of overcrowding will be assessed using a multivariate regression model to

determine whether staffing shortages increase the impact of overcrowding. Ordinary least

squares (OLS) models describe continuous outcomes, whereas count-based incident measures

are referred to negative binomial regression, which takes into account that BJS data show that

incidents in prisons and mortality are frequently over-dispersed (Carson, 2021).

Models will be estimated sequentially:

1. Baseline model with overcrowding and controls.

2. Staffing model adding staffing shortfall.

3. Full model including the interaction term.

6

Year fixed effects capture the effects of time variations in the prison populations, including

the variation related to the COVID-19 interruptions (Carson, 2021). These standard errors are

state clustered.

Missing Data and Quality Checks

Multiplex imputation is applied in cases where there is more than 5 percent missingness

and it is observed to be random. The distributional assumptions and multicollinearity are

evaluated using descriptive statistics and variance-inflation diagnostics. The Census Bureau data

consists of cross-validated facility identifiers, which is why the institutions are correctly aligned

(Garcia et al., 2024).

Ethical Considerations

Since every source of data is public-use and includes no identification information, the

project is not human subjects research and thus does not need the IRB.

7

References

Carson, E. A. (2021). Mortality in state and federal prisons, 2001-2019 — Statistical tables (NCJ

300953). Bureau of Justice Statistics. https://bjs.ojp.gov/content/pub/pdf/msfp0119st.pdf

Carson, E. A. (2021). Prisoners in 2020— Statistical tables (NCJ 302776). Bureau of Justice

Statistics. https://bjs.ojp.gov/content/pub/pdf/p20st.pdf

Garcia, M. M., Finlay, K., Speer, C. E., Willhide, E., Patti, K. N., & Loveless, T. A. (2024).

Coverage ofprisons and detention facilities in the 2020 Census. U.S. Census Bureau.

https ://www2 . census.gov/library/working-papers/2024/econ/coverage-of-prisons-and-

detention-facilities-2020-census.pdf

Stephan, J. J. (2008). Census of state and federal correctional facilities, 2005 (NCJ 222182).

Bureau of Justice Statistics. https://bjs.oip.gov/content/pub/pdf/csfcf05.pdf