JEE EXTRACTS METHODOLOGY SAMPLE DESIGN The JEE Extracts data are assembled from data collected through the U. S. Census Bureau's annual surveys of government finances and public employment. The samples of local governments for those surveys are drawn from the Quinquennial Census of Governments. The samples consist of all large local general purpose governments above a certain population threshold (certainty units) plus a sample below the certainty level. The samples also include certain independent school districts and special districts, for which justice data are not collected. For detailed information about the sample for any given year, visit www.census.gov/govs/estimate. THE SURVEY PERIOD The federal government expenditure data are for the fiscal year, which ended on September 30 of the year indicated. For example, 2008 data are for the period from October 1, 2007, to September 30, 2008. The state expenditure data presented cover fiscal years ending June 30 for all states except four, whose fiscal years ended as follows: New York, March 31; Texas, August 31; and Alabama and Michigan, September 30. For local governments, the fiscal years reported are those that closed between July 1 and June 30. Most municipalities and counties end their fiscal years on December 31 or June 30. Thus, some local jurisdictions that ended their fiscal year on December 2007 are included in the 2008 spreadsheets. The fiscal years reported for Washington, D.C., ended on September 30 of the year indicated. Tables 14 and 15 indicate the month the fiscal year closed for the local governments. Some agencies operate on a different fiscal year basis from the rest of the parent government. In such instances, figures included are for the agency's fiscal year that ended within the parent government's regular fiscal year. The employment data are for October of the year indicated if prior to 1997; in 1997 the reference month changed to March. DATA COLLECTION The JEE Extracts data are from a special compilation of data and sources available from the Census Bureau's regular surveys of government finances and public employment. The recurrent survey of government finances and public employment provide data on expenditure and employment, by function of the federal, state, and local governments (counties, cities, townships, independent school districts, and special districts). The data collection procedures of these surveys for the expenditure and employment data are described below. Annual Government Finances Survey: Federal government financial data were obtained from actual data presented in The Budget of the United States Government for each fiscal year displayed. The methodology for federal financial data was changed in 1997. Previously ,the Census Bureau performed an extensive compilation of the federal budget for the purpose of bringing it in line with Census definitions. Beginning in 1997, the data were taken directly from the budget using the definitions of justice functions contained in the budget. The Census Bureau included justice expenditures of nonjustice agencies. Most of this expenditure is coded under other than justice functions in the federal budget. State finance statistics as well as those for large counties and cities were compiled by Census Bureau representatives from official reports and records, with the advice of state and local officers and employees. The data were compiled from state government audits, budgets, and other financial reports, either in printed or electronic format. The compilation generally involved recasting the state financial records into the classification categories used for reporting by the Census Bureau. The initial local government data collection phase used two methods to obtain data: mail canvass and central collection from state sources. In about 30 states, all or part of the data for local governments was obtained from cooperative arrangements between the Census Bureau and a state government agency. These usually involved a data collection effort carried out to meet the needs of both agencies--the state agency for purposes of audit, oversight, or information, and the Census Bureau for statistical purposes. Data for the balance of local governments in the annual surveys were obtained via mail questionnaires sent directly to county, municipal, township, special district, and school district governments. The mail canvass involved the use of detailed Census Bureau schedules with related reporting instructions. Census Bureau examiners reviewed the mail reports intensely and used extensive correspondence to supplement and verify incomplete and questionable information. As with mail canvass questionnaires, centrally collected financial data sometimes needed supplementation for such items as debt, assets, or particular functional expenditures or revenue items. Census Bureau staff obtained these supplementary data from special tabulations in other state offices, printed reports, secondary sources, or from mail requests directly to the county, municipal, or township governments. Through these efforts, current year expenditure information was obtained for the federal government, all state governments, and all large county and city governments. Small nonrespondent local governments had either the prior year's data or data for similar jurisdictions imputed for them. See the Nonrespondent Jurisdictions section below for a description of how this was done. Annual Public Employment Survey: Federal government civilian employment data were obtained from records maintained by the U.S. Office of Personnel Management. These records did not provide the information necessary to compute federal full-time equivalent employment. State government data were collected through a mail survey of all state departments, agencies, and institutions. In over half of the states employment data are collected centrally for the state government. In some cases, additional mail supplementation is necessary. As with the Finance Survey, these states change from year to year. A sample of local governments also was surveyed by mail questionnaire. State agencies and local governments that did not respond by the close of the request period received follow-up requests. Second request mail consisted of postcard reminders. Third request mail was a second mailing of the original request with a survey form. Mail returned to the Census Bureau because of address problems was readdressed and remailed. In addition, large governments that had not responded for several years were contacted by telephone to identify the appropriate office or individual to receive and complete forms. After extensive nonresponse follow-up, useable replies were received from more than 91% of the sample canvassed for the 2008 Survey of Public Employment, and similar response levels were achieved in other years. Nonrespondent local governments had either the prior year's data or data for similar jurisdictions imputed for them. See the Nonrespondent Jurisdictions section below for a description of how they were treated. DATA REVIEW AND ADJUSTMENTS Once the data are collected, intensive computer editing of the data at various stages of processing minimizes errors that may be introduced during processing in the annual finance and employment survey programs. For the mail portion of the surveys, figures reported by government officials are generally accepted as being substantially correct. In some cases, varying interpretations of the instructions or deficiencies in the responding governments' records may make it difficult for officials to render complete and accurate reports for their governments. These difficulties are handled by-- -- careful definition of terms and detailed instructions in difficult cases -- supplemental correspondence and telephone followup to officials -- intense examination of data collected through verification of internal consistency and comparison with previous reports and other sources of data. The data extracted from the regular Census Bureau programs for the JEE Extracts received additional examination. After compilation for the general finance survey, a special reviewer examined in detail the expenditure data for state governments and large local governments and investigated special data compilation problems and adjusted the data as needed. The local government expenditure and employment estimates were reviewed, potential problems were investigated, and data were revised, where necessary. Data for each of the states and large counties and cities displayed individually in the JEE Extracts tables and spreadsheets were scrutinized and compared to prior years' data. A separate computer edit was then performed. Where possible, both expenditure and employment data were adjusted to correct errors, reclassify activities, and narrow differences with the JEE Surveys. Specific procedures included referring to alternate sources of data, estimating missing data, and refining data through proration (for example, sheriff offices, where some employees perform police functions, others perform judicial functions (bailiffs), and others perform corrections (jail) functions.) NONRESPONDING JURISDICTIONS Prior to 1994, nonrespondent jurisdictions had the prior year's data imputed for them for purposes of estimating national, state and local, and local totals. For 1994 and subsequent years, data for similar jurisdictions were used to develop a per capita figure, and that was used to impute data for nonrespondent jurisdictions for the purposes of estimating national, state and local, and local totals. The imputed data are not displayed on any published tables nor on the spreadsheets on the BJS website. DATA LIMITATIONS The survey sample for the local government JEE Extracts justice expenditure and employment estimates was not designed specifically to produce data on these activities. Thus, the sampling variability, or "standard error," for the justice sectors is apt to be larger than for the major categories in the Census Bureau's regular surveys and for the same functions in the JEE Survey series. The "standard error" is a measurement of variation among the estimates from all possible samples, of which this is one, having the same size and selected using the same sampling design. Estimates derived from the different samples would vary from each other (and also from a complete census using the same data collection procedures). The standard error, therefore, measures the precision with which an estimate from one of these samples approximates the average result of all the possible samples. Coefficients of variation (CVs) expressed as percents, are available on the Census Bureau's Governments Division website (http://www.census.gov/govs/), along with instructions on how to use them. Interval estimates with a prescribed confidence level can be calculated for each statistic by using the sample estimate and the standard error as estimated from the sample. For example, a 90% confidence interval can be constructed by adding 1.6 times the estimated standard error to the estimate and subtracting 1.6 the estimated standard error from the estimate. If intervals were constructed in such a manner for all possible samples of the same design and size, about 90% of them would include the complete enumeration statistic. In reviewing the sample-based estimates in this report, note that because state government figures are not subject to sampling variation, the state-local aggregates shown for individual states are more reliable (on a relative standard error basis) than the local government estimates they include. Conversely, the sampling variability for smaller components, such as type of local government detail, is likely to be greater than that for the state and local total estimates. Because the national estimates of local government expenditure and employment are based on summations of individual state data, they are more reliable than the state-area data. The data also are subject to possible inaccuracies in classification, response, and processing. Every effort was made to keep such errors to a minimum through care in examining, editing, and tabulating the data submitted by government officials. Follow-up procedures were used extensively to clarify inadequate and inconsistent survey returns. Readers should be generally cautious in comparing governments, because differences among states and local governments in functional responsibilities, governmental structure, degree of urbanization, and population density can affect the comparability of expenditure and employment data. For example, some state governments directly administer certain activities that elsewhere are undertaken by local governments, with or without fiscal aid, and the same variation in the division of responsibilities exists for counties and cities. DATA DIFFERING FROM OTHER PUBLICATIONS The JEE Extracts data differ in some cases from data in the Census Bureau's annual finance and employment surveys because of the more extensive review procedures used for this special compilation, the refinements of data involved, and the definitional differences discussed in the Comparability Issues file (JEECOMP.TXT). The JEE Extracts trend data for 1985, 1988, and 1990 also differ from those published by BJS in its JEE Survey series. Those JEE Survey data should only be used in trend analyses with other JEE Survey data for 1971-79. The 1985, 1988, and 1990 JEE Extracts data in the spreadsheets here should be used in trend analyses only with other JEE Extracts data. Federal government figures presented in the JEE Extracts 2008 are directly from the Budget of the United States, using the definitions of justice functions contained in the Budget. Prior to 1997, the Census Bureau performed an extensive compilation of the federal budget for the purposes of making the data fit the definitions used in the Census Bureau's surveys of government finances and public employment. Therefore, "all governments" and "federal government" expenditure data prior to 1997 are not directly comparable to those reported for 1997 and subsequent years.