Local-Area Crime Survey, [United States], 2015, 2016 (ICPSR 38920)

Version Date: Apr 29, 2024 View help for published

Principal Investigator(s): View help for Principal Investigator(s)
United States. Bureau of Justice Statistics; Westat

https://doi.org/10.3886/ICPSR38920.v1

Version V1

Slide tabs to view more

LACS, 2015 and 2016

The Bureau of Justice Statistics (BJS) entered into a cooperative agreement with Westat to develop and evaluate a lower-cost, subnational companion survey of victimization as one piece of the subnational estimates program. The Local-Area Crime Survey (LACS) was fielded in 2015 and 2016 and is intended for use by states, municipalities, or other jurisdictions and entities to assess levels and trends in public safety. The LACS is modeled in part after the National Crime Victimization Survey (NCVS), conducted for BJS by the U.S. Census Bureau. One of the two major statistical programs on crime produced by the U.S. Department of Justice, the NCVS is the nation's primary source of information about criminal victimization, whether reported or not reported to police. The core NCVS methodology includes a mix of in-person and telephone interviews with household members age 12 and older selected from an area probability sample to produce reliable national-level estimates. As another part of the subnational estimates program, BJS worked with the Census Bureau to enhance and reallocate the NCVS sample to support subnational estimates for the 22 most populous states and potentially substate areas within those states. For the most part, this direct estimation component of the program will not support estimates at the local level. See the NCVS Subnational Estimates webpage on the BJS website for more information.

The goals of this research were to (1) develop and test a relatively inexpensive survey design (2) that could be administered by local jurisdictions or their vendors (3) to produce cross-jurisdiction estimates and estimates of change over time within jurisdictions that may be compared with similar estimates using NCVS data. In addition to questions about victimization experiences, the LACS included questions about perceptions of community safety and police efficacy. The rationale for including these items was that they were relevant to all households, not just victims. The hope was that these items would increase survey response rates as non-victims would have important questions to answer. The LACS served as a platform for assessing the value of these questions for the planned NCVS instrument redesign. For more information, see the NCVS Instrument Redesign webpage on the BJS website.

United States. Bureau of Justice Statistics, and Westat. Local-Area Crime Survey, [United States], 2015, 2016. Inter-university Consortium for Political and Social Research [distributor], 2024-04-29. https://doi.org/10.3886/ICPSR38920.v1

Export Citation:

  • RIS (generic format for RefWorks, EndNote, etc.)
  • EndNote
United States Department of Justice. Office of Justice Programs. Bureau of Justice Statistics

Core based statistical area (CBSA)

Access to these BJS-sponsored data is restricted. Users interested in obtaining these data must complete a NACJD Restricted Data Use Agreement available from the ResearchDataGov website, specify the reasons for the request, and obtain IRB approval or notice of exemption for their research.

Restricted Data Use Agreements available on the NACJD website are provided for reference only. Please visit the ResearchDataGov website to download the appropriate Restricted Data Use Agreement and submit your request. Once approved, data may be accessed from a requester secure site via ICPSR's secure download procedures.

Inter-university Consortium for Political and Social Research
Hide

2015-09 -- 2015-12, 2016-09 -- 2016-12
2015-09 -- 2015-12, 2016-09 -- 2016-12
Hide

The goals of this research were to (1) develop and test a relatively inexpensive survey design (2) that could be administered by local jurisdictions or their vendors (3) to produce cross-jurisdiction estimates and estimates of change over time within jurisdictions that may be compared with similar estimates using NCVS data.

The sample universe for the field test included residential addresses in the 40 largest CBSAs (as of 2015) in the United States. All adults in residential households were eligible for the study; those living in institutional settings, including group quarters, were excluded from the study. The sampling frame for both years used an address-based sample (ABS), with a U.S. Postal Service (USPS) list of addresses. In year 1 of the study, three of the CBSAs were oversampled to generate sub-CBSA estimates. These substrata were defined based on police jurisdiction boundaries in Chicago, Los Angeles, and Philadelphia. In year 2 of the study, oversamples were included for Chicago and Philadelphia, and the year 2 sample included a 25% overlap with the year 1 sampled addresses (the remaining addresses were selected from the most recent, available address frame).

The year 1 target number of completed surveys for each of the 37 non-oversampled CBSAs (all CBSAs except Philadelphia, Los Angeles, and Chicago) was 2,100 completed surveys. The target sample sizes for the three oversampled CBSAs were 7,500 completed surveys for Los Angeles, 7,500 for Chicago, and 9,363 for Philadelphia. Based on a general assumption of a 50% response rate and 11% vacancy rate, the starting sample was 229,475.

The year 2 sample design was similar, although there were 38 non-oversampled CBSAs (Los Angeles was not oversampled in year 2). The year 2 target sample sizes for the non-oversampled and two oversampled CBSAs were the same as for year 1. The study design included randomly assigning a portion of the sampled addresses to be included in both years. The 25 percent overlap sample was selected from the year 1 sample while the new portion of the year 2 sample was selected from the most recent USPS address lists using the same procedures as used in year 1.

Sampled addresses were randomly assigned to equal-sized treatment groups by sorting the sample within CBSA (or CBSA stratum) and instrument (person-level survey or incident-level survey) by state, county, Census tract and Census block group, then numbering the addresses from 1 to 'x' repeatedly until all addresses were assigned a number (where 'x' equals the number of experimental treatments in that year of the field test). The research team developed two instrumentation approaches for collecting victimization data. One instrument version was based on collecting information about individual incidents (coined the "incident-level survey" or ILS). In this version, the household proxy respondent was asked to enumerate victimization incidents, then link those incidents to the relevant adult victim described in a roster. A second version was based on collecting victimization prevalence for each adult household member (the "person-level survey" or PLS).

The population of inference for the LACS was all residential adults, and a single household member acted as a proxy for the remaining adult household members. In year 2, this was done separately for the new and overlap portions of the sample. The starting sample in year 1 was 229,475, split between the PLS and the ILS instrument. In year 2, the starting sample of 217,250 was slightly smaller because one of the CBSAs oversampled in year 1 (Los Angeles) was not oversampled in year 2.

Cross-sectional

The target population is all persons 18 or older living in households in the 40 largest core based statistical areas (CBSAs). The universe of respondents for the LACS was all residential adults. A single household member acted as a proxy for the remaining, adult household members. Excluded are persons not living within these 40 CBSAs and persons under age 18, as well as those living in group quarters, who are crews of vessels, in institutions (e.g., prisons and nursing homes) or members of the armed forces living in military barracks.

crime incident, person, household

  • The incident-level survey (ILS) retains the core NCVS approach of using victimization probes followed by questions about reported incidents, although with less detail than the NCVS. The response structure is to ask about incidents and link them to the adults in the household who experienced them. This design can support incident-, person-, and household-level estimates of victimization. The ILS instrument does not capture all the incidents a victim experiences as the NCVS does, so data from the ILS and incident-level data from the core NCVS are not comparable.
  • The person-level survey (PLS) asks about the victimizations each adult has experienced, changing the focus from the incident to the person. The PLS approach begins with property crime at the household level and then asks about each adult's victimization experiences. The PLS crime questions include sufficient detail to assess whether the household or the individual adult experienced victimization in the 12-month reference period.
  • The Year 1 sample included almost 230,000 addresses with just under 94,000 households returning a survey. The overall response rate for the LACS is 47.1 percent using the American Association of Public Opinion Research (AAPOR) method RR3 (AAPOR 2016). Response rates varied greatly by CBSA (from 28 percent to 59 percent). Due to the Year 2 sample and data collection experiments, the Year 2 sample yielded a lower response rate than Year 1. The overall Year 2 response rate was 40.9 percent (RR3). For more information on the LACS response rates, see National Crime Victimization Survey Local-Area Crime Survey: Field Test Methodology Report (NCJ # 254519).

    Hide

    2024-04-29

    2024-04-29 ICPSR data undergo a confidentiality review and are altered when necessary to limit the risk of disclosure. ICPSR also routinely creates ready-to-go data files along with setups in the major statistical software formats as well as standard codebooks to accompany the data. In addition to these procedures, ICPSR performed the following processing steps for this data collection:

    • Performed consistency checks.
    • Checked for undocumented or out-of-range codes.

    Hide

    Since the main purpose of the LACS was to develop a methodology to produce local (e.g., CBSA or stratum within CBSA) estimates, the weighting procedures were constructed primarily to support CBSA-level estimates. This approach required 40 separate weightings, one for each CBSA, without regard to any data collected from the other CBSAs. Separate weights were developed for each condition version of the instrument. The rationale was that different instrument structures could substantially affect estimates, especially of victimizations.

    In Year 1, the weighting procedure implemented for each of the 40 CBSAs, separately for the ILS and PLS, involved the following steps:

  • The base weight for the sampled address was computed as the ratio of the number of addresses in the CBSA stratum to the number of sampled addresses.
  • The base weights were adjusted for household unit nonresponse within weighting classes within CBSA and stratum.
  • Missing values for data items used in household raking were replaced with imputed values.
  • Responding household weights were raked to match CBSA-level household control totals from the U.S. Census Bureau's American Community Survey (ACS).
  • A record was created for each adult, and missing values for data items to be used in adult raking were replaced with imputed values.
  • The household-level weight attached to each adult was raked to match CBSA-level adult population control totals from the ACS. This raking yielded a final adult-level weight.
  • In Year 2, the base weight calculation for the sampled address was more complex due to the overlap portion of the sample. Otherwise, the weighting steps were the same as for Year 1.

    All estimates have errors due to sampling, rather than observing, the full population of households and adult persons in the CBSA. To estimate the sampling errors (the standard errors of the estimates), a replication method of variance estimation was implemented. Replicate variance estimation allows for easily computable design-consistent variance estimators for a wide variety of descriptive and analytic statistics. A grouped jackknife variance estimator was used that involved creating a set of replicate weights. Rust and Rao (1996) call this the stratified jackknife (JK2) method. These replicate weights are used to produce standard errors of CBSA-level estimates.

    A separate set of weights was created for each instrument (ILS and PLS) so that each set sums to the number of households and adults for each CBSA. For more information on the LACS weighting and methodology, see National Crime Victimization Survey Local-Area Crime Survey: Field Test Methodology Report (NCJ # 254519).

    Final weights and replicate weights on the Year 1 data files include:

  • HHFINWT = The final raked and trimmed household weight for each Year 1 household-level file.
  • PFINWT0 = The final raked and trimmed person weight for each Year 1 person-level file.
  • HHREPFINWT1 - HHREPFINWT150 = 150 raked and trimmed household replicate weights for each Year 1 household-level file.
  • PFINWT1 - PFINWT150 = 150 raked and trimmed person replicate weights for each Year 1 person-level file.
  • Final weights and replicate weights on the Year 2 data files include:

  • HHFINALWTYR2 = The final raked and trimmed household weight for each Year 2 household-level file.
  • PFINWTYR2 = The final raked and trimmed person weight for each Year 2 person-level file.
  • HHREPFINWTYR2_1 - HHREPFINWTYR2_150 = 150 raked and trimmed household replicate weights for each household-level file.
  • PREPFINWTYR2_1 - PREPFINWTYR2_150 = 150 raked and trimmed person replicate weights for each person-level file.
  • With these replicate weights, the jackknife variance estimator can be computed using any of a number of standard packages such as SAS, SUDAAN, R, STATA, and WesVar.

    Hide

    Notes

    • The public-use data files in this collection are available for access by the general public. Access does not require affiliation with an ICPSR member institution.

    • One or more files in this data collection have special restrictions. Restricted data files are not available for direct download from the website; click on the Restricted Data button to learn more.