Appendix 4

Census of Jails, Annual Survey of Jails, and
Survey of Inmates in Local Jails

Methodology and survey sampling procedures

[ pdf version ]

Note: The following information was excerpted from U.S. Department of Justice, Bureau of Justice Statistics, Jails and Jail Inmates 1993-94, Bulletin NCJ-151651 (Washington, DC: U.S. Department of Justice, April 1995), pp. 14-16; Profile of Jail Inmates, 2002, Special Report NCJ 201932 (Washington, DC: U.S. Department of Justice, July 2004), pp. 11, 12; Prison and Jail Inmates at Midyear 2004, Bulletin NCJ 208801, pp. 11, 12; 2005, Bulletin NCJ 213133, p. 12; 2006, Bulletin NCJ 217675, p. 10 (Washington, DC: U.S. Department of Justice); Jail Inmates at Midyear 2007, Bulletin NCJ 221945 (Washington, DC: U.S. Department of Justice, June 2008), p. 6; and information provided by the U.S. Department of Justice, Bureau of Justice Statistics. Non-substantive editorial adaptations have been made.


Census of Jails

The Census of Jails, previously known as the National Jail Census, is taken every 5 to 6 years and is conducted for the U.S. Department of Justice, Bureau of Justice Statistics (BJS) by the U.S. Census Bureau. Data are presented for censuses conducted in 1983, 1988, 1993, 1999, and 2005. Questionnaires were mailed to all locally administered jails in the Nation. The number of jails included in the four previous censuses and the response rates are: 1983, 3,358 jails with 99% responding; 1988, 3,316 jails with 100% responding; 1993, 3,304 jails with 90% responding; and 1999, 3,084 jails with 99% responding.

Beginning in 2005, in order to reduce respondent burden and improve data quality, the census was split into two parts: the Census of Jail Inmates, 2005 (conducted June 30, 2005), and the Census of Jail Facilities, 2006 (conducted Mar. 31, 2006).

The 2005 census included all locally administered confinement facilities (under the authority of 2,853 local jurisdictions). These facilities are intended for adults but sometimes hold juveniles. They hold inmates beyond arraignment (usually more than 72 hours) and are staffed by municipal or county employees. The census also included 42 jails that were privately operated under contract to local governments, 65 multijurisdiction jails and 13 facilities maintained by the Federal Bureau of Prisons and functioning as jails.

Excluded from the census were physically separate temporary holding facilities, such as drunk tanks and police lockups, that do not hold persons after being formally charged in court (usually within 72 hours of arrest). Also excluded were State-operated facilities in Alaska, Connecticut, Delaware, Hawaii, Rhode Island, and Vermont, which have combined jail-prison systems. However, 14 independently operated jails in Alaska were included.

All jail jurisdictions responded to the 2005 census, resulting in a 100% response rate.

Estimation procedures

Because there was nonresponse and incomplete data on some census items, national totals had to be estimated. The following procedures were used to estimate totals when data were incomplete:

1. Each item was assessed for coverage and internal consistency. To estimate totals, extreme values were examined and verified by checking other census information and originally submitted forms. Detailed categories also were checked to determine if they summed to the reported totals.

2. To provide national and State total estimates of staff, item values were summed and then multiplied by a nonresponse adjustment factor (NAF). The NAF was a ratio of the total number of inmates in all jails to the number of inmates in jails that reported valid staff data.

3. All estimates were rounded to the nearest 100.

4. All rates, ratios, and percentage distributions were based on reported data only.

Annual Survey of Jails

In each of the years between the full censuses, a sample survey of jails is conducted to estimate baseline characteristics of the Nation's jails and inmates housed in the jails. Data from the Annual Survey of Jails (ASJ) are presented for 1984-87, 1989-92, 1994-98, 2000-2004, 2006, and 2007. The reference date for each of these surveys was June 30, except 1990, 2001, and 2007 when the reference date was June 29, and 1991 and 1996 when it was June 28. All surveys prior to the 1994 survey were based on all jails in jurisdictions with 100 or more jail inmates and a stratified random sample of jurisdictions with an average daily population of less than 100 inmates. For 1984, 1,164 jails in 893 jurisdictions were included; in 1985, 1,142 jails in 874 jurisdictions were included; in 1986, 1,137 jails in 868 jurisdictions were included; in 1987, 1,135 jails in 866 jurisdictions were included; in 1989, 1,128 jails in 809 jurisdictions were included; in 1990, 1,135 jails in 804 jurisdictions were included; in 1991, 1,124 jails in 799 jurisdictions were included; and in 1992, 1,113 jails in 795 jurisdictions were included.

A new sample of jail jurisdictions was selected for the 1994-98 surveys using information from the 1993 Census of Jails. A new sample was again selected and used for the 2000-2004 surveys based on information from the 1999 Census of Jails. A jurisdiction is a county (parish in Louisiana) or municipal government that administers one or more local jails. The sample included all 940 jail facilities in 878 jurisdictions. All 55 multijurisdiction jails (jails operated jointly by two or more jurisdictions) were included in the sample with certainty.

Jails in 356 other jurisdictions were automatically included in the sample if the jail held juveniles and had an average daily population of 250 or more inmates on June 30, 1999 or if they held only adults and had an average daily population of 500 or more.

The remaining jurisdictions were stratified into two groups: jurisdictions with jails holding at least one juvenile on June 30, 1999, and jurisdictions with jails holding adults only. Using stratified probability sampling, 467 jurisdictions were then selected from 10 strata based on the average daily population in the 1999 jail census.

Data were obtained by mailed and web-based survey questionnaires. After followup phone calls, the response rate for the 2003 and 2004 surveys was 100% for critical items such as rated capacity, average daily population, and number of inmates confined.

For the 2006 and 2007 ASJ, a new sample of 874 jurisdictions and 936 jail facilities was used based on the 2005 Census of Jail Inmates. Local jail jurisdictions include counties (parishes in Louisiana) or municipal governments that administer one or more local jails. The sample included all jails operated jointly by two or more jurisdictions (multijurisdiction jails), with certainty (63). Other jail jurisdictions included with certainty (269) were those that (a) held juveniles at the time of the 2005 Census of Jail Inmates and that had an average daily population of 500 or more inmates during the 12 months ending June 30, 2005, or (b) held only adults and the average daily inmate population was 750 or more.

The remaining jurisdictions were stratified into two groups: jurisdictions with jails holding at least one juvenile on June 30, 2005, and jails holding only adults on that date. Using stratified random sampling, 542 jurisdictions were selected from 8 strata based on the two conditions enumerated above and 4 strata based on their 2005 average daily inmate population, which was derived from the 2005 Census of Jail Inmates.

Data were obtained from sampled jurisdictions by mailed and web-based survey questionnaires. After followup phone calls to respondents, the response rates for the 2006 and 2007 surveys were 100% for critical items such as the number of inmates confined, average daily population, and rated capacity.

Sampling error

Survey estimates have an associated sampling error because jurisdictions with smaller average daily populations were sampled for the survey. Estimates based on the sample survey may differ from the results of conducting a complete census. Different samples could yield somewhat different results. Standard error is a measure of the variation among the estimates from all possible samples, stating the precision with which an estimate from a particular sample approximates the average of all possible samples.

Readers interested in standard error estimates should consult electronic versions of the following reports: Prison and Jail Inmates at Midyear 2004, p. 14. Available:; Prison and Jail Inmates at Midyear 2006, p. 22. Available:; and Jail Inmates at Midyear 2007, pp. 10,11. Available:

Measures of jail population

Two measures of jail inmate population are used: the average daily population for the year ending on the survey/census reference date and the 1-day inmate count on the survey/census reference date of each year. The reference date for the annual surveys and the jail censuses is usually June 30 or the last business day in June. The average daily population balances out any extraordinary events that may render atypical the inmate count on the reference date. The 1-day count provides data on characteristics of inmates, such as race, Hispanic origin, and age, that may not be available on an annual basis.

In 1995 the Annual Survey of Jails obtained, for the first time, separate counts of the total number of offenders under jail jurisdiction, those held in jail facilities, and those supervised outside of jail facilities. Previous surveys and censuses included a small but unknown number of offenders under community supervision. To estimate the percent change from 1994 to 1995 in the jail population, the 1995 survey included a count of inmates held at midyear 1994.

In the 1996 survey the number of persons supervised outside a jail facility included for the first time persons under drug, alcohol, mental health, or other medical treatment. Comparison with 1995 estimates should exclude these persons.


In annual jail surveys beginning in 1994 and in jail censuses beginning in 1999, jail authorities were asked to report the number of inmates under age 18. Most, but not all, States defined a juvenile as a person under age 18 who is subject to juvenile court jurisdiction. Exceptions usually depend on offense severity or an offender's adjudication history.

Statutes and judicial practices sometimes allow youths to be held in adult jails. Often juveniles accused of acts that are crimes for adults may be held in jails or police lockups, given certain conditions: separation by sight and sound from the general population and detention for a limited time (typically less than 6 hours).

Survey of Inmates in Local Jails

The 2002 Survey of Inmates in Local Jails was conducted for BJS by the U.S. Census Bureau. Similar surveys of jail inmates were conducted in 1972, 1978, 1983, 1989, and 1996. Interviews for the 2002 survey were conducted from January through April 2002.

The sample for the 2002 survey was selected from a universe of 3,365 jails that were enumerated in the 1999 Census of Jails. The sample design was a stratified two-stage selection where jails were selected in the first stage and inmates to be interviewed were selected in the second stage. In the first stage, six separate strata were formed based on the size of the male, female, and juvenile populations. In two strata all jails were selected--those jails housing only females and those with more than 1,000 males or more than 50 females or both. In the remaining four strata a systematic sample of jails was selected. Each jail within a stratum had an equal probability of selection. Overall, 465 jails were selected. Interviews were conducted in 417; 39 jails refused or were excluded for administrative reasons, and 9 were closed or had no inmates to survey.

In the second sampling stage, interviewers from the U.S. Census Bureau visited each selected facility and systematically selected a sample of male and female inmates using predetermined procedures. A total of 6,982 inmates were interviewed and 768 refused to participate, resulting in a second stage nonresponse of 9.9%.

Interviews were about 1 hour long and used computer-assisted personal interviewing (CAPI). With CAPI, computers provide the interviewer questions, including followup questions tailored to preceding answers. Before the interview, inmates were told verbally and in writing that participation was voluntary and that all information provided would be held in confidence. Based on the completed interviews, estimates for the entire jail population were developed using weighting factors derived from the original probability of selection in the sample. These factors were adjusted for variable rates of nonresponse across strata and inmate characteristics. Further adjustments were made to conform the survey estimates to counts of jail inmates obtained from the 1999 Census of Jails and the 2001 Annual Survey of Jails.

Accuracy of the survey estimates

The accuracy of the estimates from the 2002 Survey of Inmates in Local Jails depends on two types of error: sampling and measurement. Sampling error is variation that occurs by chance because a sample rather than a complete enumeration of the population was conducted. Sampling error is measured by estimated standard error and varies by the size of the estimate and the size of the base population. Measurement error can be attributed to many sources, such as nonresponse, differences in the interpretation of questions among inmates, recall difficulties, and data processing errors. In any survey the full extent of the measurement error is never known.