Two major changes to the UCR county-level files were implemented with the 1994 data and are continued through the 2012 data. A new imputation algorithm to adjust for incomplete reporting by individual law enforcement jurisdictions has been adopted. Also, a new Coverage Indicator has been created to provide users with a diagnostic measure of aggregated data quality in a particular county. These developments are described in greater detail below. The changes were instituted in response to comments from a number of users and after almost a year of discussions by UCR file users, the FBI, the Bureau of Justice Statistics, and the Inter-university Consortium for Political and Social Research. These changes result in a break in series from previous UCR county-level files. Data from earlier year files should not be compared to data from 1994 and subsequent years. Changes in procedures used to adjust for incomplete reporting at the ORI or jurisdiction level may be expected to have an impact on aggregates for counties in which some ORIs have not reported for all 12 months. However, the new adjustment procedures should result in county-level data that are less sensitive to changes between years in the extent of reporting by ORIs within a county. Consequently, data from 1994 forward should be more accurate estimates for longitudinal analysis. The crimes reported data prior to 2009 included the variables INDEX and MODINDX. For 2009 forward these variables, INDEX (The sum of variables MURDERS through MVTHEFT) and MODINDX (The sum of variables MURDERS through ARSON), were replaced with VIOL (The sum of variables MURDERS through AGALST) and PROPERTY (The sum of variables BURGLRY through MVTHEFT). While variable ARSON is listed separately for the crime reported data, it is included in property crimes for the arrest data.

IMPUTATION PROCEDURES USED FOR 1977-1993 UCR COUNTY-LEVEL FILES: The data for any ORI reporting 12 months were used for county aggregation as submitted. Data for an ORI reporting six to 11 months were increased by a weight of [12/months reported]. Data for any ORI reporting less than 6 months were deleted from the county total, and the population served by that ORI was deleted from the county population total to help control for differential data quality across counties. The aim of this procedure was to produce comparable data cross-sectionally across all counties. However, if there were major changes in the ORIs that reported in a county across years, artifactual changes in the longitudinal data for a county could be introduced because of potential variation in the type of ORIs used to compute imputed county totals and rates each year. IMPUTATION PROCEDURES USED FOR 1994 UCR COUNTY-LEVEL FILES AND ONWARD: The data for any ORI reporting 12 months were used for county aggregation as submitted. Data for an ORI reporting 3 to 11 months were increased by a weight of [12/months reported]. For ORIs reporting 0 to 2 months in the arrest files or 1 to 2 months in the crimes reported files, data for these ORIs were set to zero and then data were estimated using rates calculated from ORIs reporting 12 months of data located in the ORI's geographic stratum based on UCR Population Groups within their state. UCR Population Groups are defined as follows: Population Group -----------------

1. Cities 250,000 and over
2. Cities 100,000 - 249,999
3. Cities 50,000 - 99,999
4. Cities 25,000 - 49,999
5. Cities 10,000 - 24,999
6. Cities 2,500 - 9,999
7. Cities under 2,500
8. Non-MSA counties and non-MSA State Police
9. MSA counties and MSA State Police

There is an important distinction in estimation procedures for the arrest versus the crimes reported files. For the arrest files (Parts 1-3 and 5-7), data were estimated for agencies reporting 0-2 months based on the procedures mentioned above. However, due to the structure of the original data received from the FBI, ICPSR was only able to estimate data for agencies reporting 1-2 months in the crimes reported files (Parts 4 and 8). No estimations are provided for agencies reporting 0 months. To assist users, the county population figures and number of agencies reporting for both the arrest and crimes reported data files were added to the crimes reported files. These variables provide users with information that can be used to determine the proportion of the population or agencies not included in the crime reported files due to the absence of records for agencies reporting 0 months. Users should keep in mind these differences when doing their analyses.

For releases of UCR county-level files before 1994, data from jurisdictions reporting less than 6 months of data were not included in county totals in an effort to ensure cross-sectional data comparability and quality. With the new procedures to adjust for incomplete reporting, data will be provided for each active ORI that reports less than 12 months of data, whether through weighting of partial year data or substitution of a value based on population group and state. Instead of exercising an a priori judgment that 6 months of data is the minimum threshold for acceptable data quality, a new Coverage Indicator variable has been created that will allow users to set their own threshold for acceptable data quality and to include or exclude data based on the standards they set themselves. The Coverage Indicator variable represents the proportion of county data that is reported for a given year. The indicator ranges from 0 to 100. A value of 0 indicates that no data for the county were reported and all data have been imputed. A value of 100 indicates that all ORIs in the county reported for all 12 months in the year. Coverage Indicator is calculated as follows: CI_x = 100 * ( 1 - SUM_i { [ORIPOP_i/COUNTYPOP] * [ (12 - MONTHSREPORTED_i)/12 ] } ) where CI = COVERAGE INDICATOR x = county i = ORI within county Some ORIs do not have a population associated with their jurisdiction. These ORIs report for jurisdictions such as national parks, colleges and universities, toll bridges and tunnels, and most state police departments. As the coverage indicator is based on months of reporting and the population of each agency, this variable will not show estimation that did occur for statewide ORIs and for ORIs as listed above that do not have a population but reported 3 to 11 months of data. Conversely, the coverage indicator will indicate that estimation has occurred for ORIs with a population that reported 3 to 11 months of data even if the ORIs actually reported no crimes or arrests. Similarly, the coverage indicator will indicate that estimation had occurred for ORIs with a population that reported 0 to 2 months of arrest data and 1 to 2 months of crimes reported data, even though no rate was calculated to estimate data because of the lack of ORIs in the agency's geographic stratum reporting 12 months of data. Finally, since data for ORIs that reported 0 to 2 months of arrest and 1 to 2 months of crime are set to zero, users should be aware that no estimation of data was possible for ORIs in these categories without a population. In the crimes reported files, no estimation was possible for agencies reporting 0 months because population figures for these agencies were not provided in the original files from the FBI.

"Crime in the United States 2012" is a publication prepared by the Federal Bureau of Investigation and provides estimations of national reported crime activity and arrest statistics from law enforcement agencies in the UCR Program. Users of this data collection prepared by ICPSR may not be able to match the statistics presented in "Crime in the United States 2012" due to several factors:

1. The UCR staff continue to update agency records when additions or corrections are received by the UCR Program. The FBI statistics presented in "Crime in the United States 2012" are based on the data that the FBI has received prior to their established publication deadlines. The data used by ICPSR to prepare the data in ICPSR 35019 may contain additions or corrections to the data submitted to the FBI after their publication deadline for producing "Crime in the United States 2012."
2. The new imputation algorithm that ICPSR implemented in 1994 approximates the estimation procedures that the FBI uses to adjust for incomplete or unreported data from individual law enforcement jurisdictions. Limited or no crime or arrest data were available for some states in 2012. For some tables in "Crime in the United States 2012" (for example, Section II, Table 2: Total Crime Counts in the United States, and Section IV, Table 29: Estimated Arrests), the FBI provided estimated crime counts and arrest totals for some of these states (see Appendix I -- Methodology for the FBI's estimation methods). The ICPSR algorithm estimates data for agencies reporting less than three months of data by calculating rates from agencies reporting 12 months of data located in the agency's geographic stratum within that state. Since data for these states were very limited or not present in the original data, estimates for the entire state could not be produced by ICPSR.
3. Some tables in "Crime in the United States 2012" were prepared using data only from agencies submitting complete reports for the 12 months for 2012 prior to the publication deadline (for example, Section IV, Table 69: Arrests by State). For this ICPSR data collection, all law enforcement agency records present in the original FBI data were used in the aggregation to the county level. County records with a coverage indicator value of 100 contain only 12-month reporting agencies (see last paragraph of the Coverage Indicator section in the Introduction for possible exceptions). Agencies reporting 12 months of data within a county with a coverage indicator value less than 100 cannot be separately identified in this dataset aggregated at the county level.

In this data collection, zeroes may represent both true zeroes or missing data, and it is possible to distinguish between the two. In the arrest files (Parts 1-3 and 5-7), the coverage indictor alone does not indicate whether or not zeroes are true or missing values. A county can have a coverage indicator of zero, but still contain arrest data. ICPSR can estimate arrest data for a county that did not report any arrests based on data for other counties of comparable population size in the same state that did report 12 months of data. In order to distinguish true zeroes from missing values, the user has to look at the coverage indicator for a county in conjunction with the arrest variables. If an arrest variable for a particular county has a value of zero and the coverage indicator is greater than zero, then the arrest variable contains a true zero. This true zero is not necessarily a reported zero as it may have been estimated. If an arrest variable for a particular county has a value of zero and the coverage indicator for that county is zero, then the user must check the values of ALL of the arrest variables. If some arrest variables contain zeroes and some contain other values, then the zeroes for this county are true. If ALL arrest variables and the coverage indicator for the county in question contain zeroes, then these indicate missing data. In the crimes reported file (Part 4), data can not be estimated for non-reporting agencies. Therefore, counties with no population for reporting agencies will also have a coverage indicator of zero and all crime variables will have a value of zero. These zeroes indicate missing data. If a county has one or more offense variables with a value of zero, but has a value of greater than zero for the population and coverage indicator variables, then the zeroes for these particular offense variables should be considered true zeroes for the purposes of analysis. SUMMARY OF HOW TO IDENTIFY MISSING DATA ARREST FILES: TRUE ZERO Coverage indicator > 0 or Coverage indicator = 0 AND at least one arrest count variable > 0 MISSING DATA Coverage indicator = 0 AND all arrest count variables = 0 ------------------------------------------------------ CRIMES REPORTED: TRUE ZERO Coverage indicator > 0 MISSING DATA Coverage indicator = 0 AND all other crime count variables will necessarily = 0

Cities designated by the Census Bureau as independent cities are reported separately and have their own "county" codes (see Appendix). Some jurisdictions, such as state parks and some state police, provide data only on a statewide basis. In these cases, data are allocated to counties proportionate to their share of the total state population. Parts 5 to 8 contain the amount of statewide data allocated to each county. To obtain the county total without the statewide allocation, the statewide values can be subtracted from the county data. State Police data for Vermont were not reported within a county and the State Police data for Alaska and Connecticut are not allocated to the counties. These three State Police records are identified by the county code 999. In the 1997 data and onward another 999 county code was added to these data files. The New York/New Jersey Port Authority reported data in 1997 through two agencies, one in New York and one in New Jersey. The FBI included the New York Port Authority agency as part of New York County. However, the FBI did not assign a UCR county code for the New Jersey Port Authority and did not provide a population figure for the area covered by this jurisdiction. Therefore, the New Jersey/New York Port Authority (NJNYPOA) is a separate record in the data files from 1997 onward with a county code of 999. New York City Metro Transportation Authority (NY03083) is a separate record from 2007 onward with a county code of 999. From 2008, tribal agencies began to report offense data. Data for those agencies in each state will be in one record identified by the county code of 777. ICPSR uses FIPS Publication 6-4 "Counties and Equivalent Entities of the United States, Its Possessions, and Associated Areas" from the National Institute of Standards and Technology (NIST), U.S. Department of Commerce, as a reference for assigning FIPS codes for the counties in this data collection. ICPSR consults this publication annually to note changes in FIPS codes that may affect this collection. The borough of Denali, Alaska (FIPS code 02068) was created from part of the Yukon-Koyukuk Census Area (FIPS code 02290) and unpopulated part of the Southeast Fairbanks Census Area (FIPS code 02240) effective December 7, 1990. The Denali Borough is NOT a record in the data. However, since no agency record from Denali was present in the original FBI data for either the arrest or crimes reported files, no data are missing from the ICPSR data. In May 1995, NIST announced that the independent city of South Boston, Virginia reverted to town status and became part of Halifax County as of June 30, 1995. The Virginia FIPS county code of 780 no longer exists and South Boston is included in FIPS county code 083. ICPSR did not adjust this data collection to reflect this change so that the data files would remain consistent with earlier years. Users who wish to use the current FIPS codes can combine the data from South Boston with Halifax County. In July 1999, NIST announced that Yellowstone National Park, which had been treated as a county equivalent, is legally part of Gallatin County and Park County, Montana. Yellowstone National Park was left as a county equivalent in this data collection to remain consistent with earlier years. In July 1999, NIST announced that as of November 13, 1993 Dade County, Florida officially changed its name to Miami-Dade County. The county was assigned a new FIPS code (086) to maintain the alphanumeric sequence of counties. In the 1997 data collection and onward, ICPSR also changed the Miami-Dade county code from 025 to 086 to match the change in the FIPS publication. However, the 1993-1996 UCR county-level data collections still used the old name and FIPS county code. In July 2001, NIST announced that the independent city (county-equivalent) of Clifton Forge, Virginia, had reverted to town status, effective midnight, July 1, 2001. Clifton Forge is now an incorporated place within Alleghany County, rather than a separate county-equivalent surrounded by Alleghany County. This action will reduce the number of Virginia independent cities to 39 and the number of United States counties and equivalent areas to 3,141. The action reduces the total number of independent cities in the United States to 42. The FIPS county code of 560 for Clifton Forge, Virginia, is deleted. The FIPS-55 class code will change from C7 to C1. The census place code of 0285 and FIPS-55 code of 17440 are unaffected. The FIPS county code of 005 for Alleghany County remains unchanged. ICPSR did not adjust this data collection to reflect this change so that the data files would remain consistent with earlier years. Users who wish to use the current FIPS codes can combine the data from Clifton Forge with Alleghany County. In January 2002, NIST announced that Broomfield County, Colorado, had been created from parts of Adams (001), Boulder (013), Jefferson (059), and Weld (123) counties effective November 15, 2001. The boundaries of Broomfield County reflect the boundaries of Broomfield city legally in effect on November 15, 2001. To maintain the alphanumeric sequence of counties, Broomfield County will have a code of 014 for FIPS 6-4, Counties and Equivalent Entities of the United States, Its Possessions and Associated Areas. ICPSR created county code 014 for Broomfield County to match the change in the FIPS publication.

In the UCR county-level arrest files, the population and data for jurisdictions located in multiple counties are provided only in the county containing the largest population component of the jurisdiction. Counties containing smaller population components of multiple-county jurisdictions will contain no population or arrest data for these jurisdictions. Data in counties affected by one or more multiple-county jurisdictions are indicated by a multi-county jurisdiction flag variable. In the county-level crimes reported files, the population and crime data for jurisdictions located in multiple counties are provided by the UCR proportioned to each county (maximum of three) in which the jurisdiction is located. Drunkenness (FBI offense code 23) is not considered a crime in some states. States that do not consider drunkenness a crime, but which report data through the NIBRS system, such as North Dakota, may show arrest data for the offense in this data collection. This is because drunkenness may be listed as an incident in the NIBRS reporting system, even though there was technically no arrest in these states. Agencies in states where drunkenness is not a crime may use their own discretion in reporting drunkenness as an incident. Users should exercise caution when analyzing this variable because of these differences in reporting. The original data from the FBI contain one record for New York City. Data from New York City are allocated into New York City's five counties on the basis of the proportion of the population in each county. For example, the population for Queens county is divided by the total population of New York City and the resulting proportion is multiplied with data from each of New York City's arrest and offense categories to apportion data to Queens county. The total population for New York City and the population for each of the five counties from the 2010 U.S. Census were used to calculate these proportions. The population of Goodlettsville City, Tennessee is proportioned to both Davidson County and Sumner County. Population from the 2010 U.S. Census was used to calculate the population breakdown of Goodlettsville, Tennessee. Seven dummy records were created for the arrest data files and 58 dummy records were created for the crimes reported data files for counties not represented in the original FBI file. These county records provide only the state and county FIPS codes with the rest of the variables following county FIPS in the data file filled with zeros. No arrest data were provided for Alabama and Florida. Limited arrest data were provided for Illinois and South Dakota. Limited crimes reported data were available for Alaska, Mississippi, and South Dakota. Each arrest data file (Part 1, 2, or 3) has 56 variables. The crimes reported data file (Part 4) has 21 variables. Each allocated statewide data for arrests file (Part 5, 6 or 7) has 57 variables and the allocated statewide data for crimes reported file (Part 8) has 22 variables. All data files contain 3,178 cases.