Victimization in Different Types of Areas in the United States: Subnational Findings from the National Crime Victimization Survey, 2010–2015

This report was produced by the University of Missouri-St. Louis for the Bureau of Justice Statistics under an Interagency Personnel Agreement. This report describes methods used to create “generic area” typologies based on geographic, social, economic, or demographic characteristics. These generic areas can represent places that are similar to each other based on the characteristics of interest. This report creates generic areas using three geographic indicators currently available in public-use National Crime Victimization Survey data: census region, metropolitan area status, and population size for the central cities of metropolitan areas.


Introduction
Crime in the United States has typically been examined using data from two national indicators-the Uniform Crime Reporting (UCR) program and the National Crime Victimization Survey (NCVS) (Planty, Langton, & Barnett-Ryan, 2014). 1 Data from the UCR program provide information about crimes that come to the attention of the police and that police departments voluntarily provide to the Federal Bureau of Investigation. A key advantage of UCR crime data is the ability to examine crime rates across law enforcement jurisdictions and geographic entities, such as cities, counties, states, and regions.
The NCVS gathers information from a nationally representative sample of persons age 12 or older to produce estimates of criminal victimization. Unlike the UCR program, the NCVS is able to collect information on crimes that occur regardless of whether they are reported to police. The NCVS also provides important details about crime incidents and victims that are not fully available in the current UCR program, such as victim characteristics, the victim-offender relationship, use of weapons, and whether the victim was injured during the incident. To protect the confidentiality of respondents, NCVS public-use data do not include detailed geographic information that would allow for comparisons of victimization rates across specific jurisdictions.
Enhancing the ability of the NCVS to provide information about the levels, nature, and consequences of victimization across different types of places has long been viewed as an important need by stakeholders (Langton, Planty, & Lynch, 2017). A key reason for this is that many areas do not have the resources to conduct their own state or local surveys, especially on an annual basis. Greater availability of subnational NCVS data would give state and local policymakers additional knowledge about how victimization rates and patterns, crime reporting rates, and use of victim services in their own areas compare with other places across the United States. In addition, information about the type of place in which individuals reside can be used to study whether sociodemographic predictors of victimization risk, such as sex or age, are correlated with victimization in similar ways across different areas.
The Bureau of Justice Statistics (BJS) is currently using four main approaches to develop NCVS subnational estimates of criminal victimization (Langton, Planty, & Lynch, 2017). 2 The information generated by these approaches will be useful for studying subnational patterns and informing federal funding allocations for crime prevention and related resources.
• The first approach includes an increase and reallocation of the NCVS sample. BJS used this approach to provide direct estimates of victimization based on aggregations of multiple years for the 22 largest states and the largest cities and metropolitan areas in the United States. 1 The UCR program has partnered with the Bureau of Justice Statistics on the National Crime Statistics Exchange and transitioned the UCR program to a National Incident-Based Reporting System (NIBRS)-only data collection, as of January 1, 2021. For more information, see https://bjs.ojp.gov/data-collection/national-incident-based-reportingsystem-nibrs. 2 For more information, visit the NCVS Subnational Estimates page on the BJS website at https://bjs.ojp.gov/subnational-estimates-program.
• The second approach uses statistical methods to develop model-based estimates for all 50 states and large core-based statistical areas (see Fay & Diallo, 2015). • The third approach involves reweighting NCVS data to produce representative estimates at lower levels of geography for areas where the existing sample is not representative, particularly for data years that precede the sample redesign. • The fourth approach, which is the focus of this report, creates "generic area" typologies based on geographic, social, economic, or demographic characteristics. These typologies are then used to generate subnational information about crime and victimization from the NCVS data. These generic areas can represent places that are similar to each other based on the characteristics of interest. This report creates generic areas using three geographic indicators currently available in public-use NCVS data: census region, metropolitan area status, and population size for the central cities of metropolitan areas.
To assess the feasibility of using NCVS data to generate reliable subnational estimates for these geographic generic areas, sample coverage, weighting, and variance estimation procedures in each of these areas were examined (Shook-Sa, Lee, & Berzofsky, 2015). These assessments indicated that NCVS sample coverage was generally adequate for producing reliable multiyear estimates of victimization within these areas, and that sample weights and direct variance estimation should be used for victimization estimates and confidence intervals. This report follows the recommendations from the feasibility analysis and presents subnational estimates of victimization for 20 different types of areas in the United States defined on the basis of the NCVS indicators of region, metropolitan area status, and population size.
Defining different types of areas in the NCVS public-use data  During this same period, a total of 18.2% of persons age 12 or older in the NCVS sample lived in the Northeast, 22.6% in the Midwest, 36.0% in the South, and 23.2% in the West (figure 2). The differences between the NCVS sample distribution and the U.S. population percentages are due to variations in the proportions of persons who are age 11 or younger across regions. The NCVS metropolitan-area indicator measures whether respondents live in areas with a large population nucleus (of at least 50,000 persons) together with adjacent communities and counties that have a high degree of economic and social integration (Office of Management and Budget, 2010) (for a map visualization, see https://www2.census.gov/geo/pdfs/mapsdata/maps/reference/ma_1999.pdf).
The NCVS metropolitan-area measure further denotes whether persons live in the core central city of a metropolitan area, in geographically adjacent counties surrounding the central city, or in places outside of a metropolitan area. The majority of the U.S. population resides in metropolitan areas. During 2010-15, a total of 84.5% of NCVS respondents lived in metropolitan areas (32.8% in the central cities and 51.7% in the adjacent areas), and 15.5% lived in places outside of metropolitan areas (figure 3). The NCVS measure of population size can be used to further divide the core central cities of metropolitan areas into three groups: small cities of 249,999 or fewer persons, mid-size cities with 250,000 to 999,999 persons, and large cities with 1 million or more persons. During the 2010-15 period, 15.4% of the population lived in small core cities of metropolitan areas, compared to 9.5% in mid-size core cities and 7.8% in large core cities.
The combination of these three NCVS public-use measures permits a 20-category classification for the type of area in which respondents live. For ease of readability in this report, the core central cities of metropolitan areas are labeled "urban" areas, areas adjacent to the core cities of metropolitan areas are labeled "suburban" areas, and places outside of metropolitan areas are labeled "rural" areas. However, land use and residential patterns may vary greatly within cities, within their adjacent metropolitan counties, and in places outside of metropolitan areas.
Western U.S. counties that are part of metropolitan areas tend to be much larger in geographic size than similar counties in the Northeast and Midwest; population density patterns also vary notably across categories of metropolitan areas. Though the majority of persons living outside of metropolitan areas are in places designated by the U.S. Census Bureau as rural, a notable proportion of these persons live in places that are considered urbanized or urban clusters. For example, during 2010-15, a total of 62% of NCVS respondents living outside of metropolitan areas were in places designated as rural areas, while 38% were living in urbanized areas and urban clusters. In addition, 23% of respondents living in areas adjacent to core cities of metropolitan areas were living in areas with rural land use patterns.
In 2010, there were 366 designated metropolitan areas in the United States (Wilson, Plane, Mackun, Fishetti, & Goworowska, 2012). To note the places that are categorized into each of the 20 categories, a few examples for each region are presented here. In the Northeast, large urban areas (i.e., metropolitan core cities with populations of more than 1 million persons) include New York City (NY) and Philadelphia (PA). Suburban places in the Northeast include those areas outside of the cities of New York and Philadelphia, but within the New York-Newark-Jersey City ( Population sizes vary in the 20-area types, and this affects the corresponding NCVS sample sizes in the areas (table 1). While the NCVS provides unbiased victimization estimates for the areas, standard errors will be larger in places with smaller samples. As a result, when rates for areas with smaller populations are compared to national rates, some of the differences may not be statistically significant. Victimization rates in different types of areas Violent victimization rates were higher than the national average in many of the small and mid-size urban areas during 2010-15 Violent victimization rates, which include rape and sexual assault, robbery, and aggravated or simple assault, were significantly higher in small and mid-size urban areas in the Northeast, Midwest, and West than the national average rate of 21.6 per 1,000 persons ages 12 or older (table 2; figure 4). In small urban areas, rates averaged 30.3 per 1,000 in the Midwest and 29.9 per 1,000 in the Northeast. In mid-size urban areas, violent victimization rates were 33.0 per 1,000 in the Midwest, 30.4 per 1,000 in the Northeast, and 31.0 per 1,000 in the West. However, rates in small and mid-size urban areas in the South were not significantly higher than the national average. A greater percentage of violent crime in mid-size (54.5%) and the large (52.3%) urban area in the Midwest was serious violent victimization, compared to the national average (33.6%). Rates of violent victimization were significantly lower than the national average in rural areas in the South (15.3 per 1,000) and Midwest (15.7 per 1,000), suburban areas in the South (17.1 per 1,000), and large cities in the Midwest (16.1 per 1,000).  The rate of serious violence, which includes rape and sexual assault, robbery, and aggravated assault, was highest in mid-size urban areas in the Midwest (18.0 per 1,000 persons age 12 or older), and this rate was significantly greater than for the nation as a whole (7.3 per 1,000) (figure 5). Serious violence rates also were significantly higher than the national average in midsize cities in the West (12.1 per 1,000) and small urban areas in the Northeast (11.0 per 1,000). In comparison, rates of serious violent victimization were significantly lower than the national average in rural areas in the South (4.7 per 1,000), suburban areas in the South (5.7 per 1,000), and rural areas of the Northeast (4.9 per 1,000). Household property crime rates were higher than the national average in many mid-size and large urban areas during 2010-15 Household property crime rates, which include burglary, household larceny, and motor vehicle theft, were significantly higher than the national average rate of 129.8 per 1,000 households in the West in all places except rural areas: 164.3 per 1,000 in suburban areas, 182.5 per 1,000 in small urban areas, 192.6 per 1,000 in mid-size urban areas, and 167.2 per 1,000 in large cities ( figure 6). In the South, household property crime rates were significantly greater than the national average in mid-size (156.4 per 1,000) and large (224.4 per 1,000) urban areas.
In the Midwest, households in mid-size cities also experienced higher-than-average rates of property crime (226.4 per 1,000). In comparison, rural (90.8 per 1,000), suburban (86.0 per 1,000), and large cities (85.4 per 1,000) in the Northeast had household property crime rates that were significantly lower than the national average. Other areas with lower-than-average household property crime rates included suburban areas in the South (111.3 per 1,000) and Midwest (105.2 per 1,000) and rural areas in the Midwest (101.4 per 1,000).

In all types of areas, rates of violent and household property victimization remained stable or decreased significantly over time
Rates of violent victimization and household property crime in each area during 2010-15 were compared to the rates for 1998-03 and 2004-09 to assess changes over time across the 20 areas (table 3). None of the 20 areas exhibited significantly higher rates of either violent or property victimization during 2010-15 than in the previous periods.

Rates of violence involving a weapon and rates of firearm violence were highest in mid-size metropolitan cities in the Midwest
During 2010-15, rates of violent victimization involving a weapon were highest in mid-size urban areas in the Midwest (11.9 per 1,000 persons age 12 or older), and this rate was significantly higher than the rate of weapon violence for the nation (4.5 per 1,000) (table 4).
Other areas with significantly higher rates of weapon violence than the nation as a whole included small urban areas in the Northeast (8.8 per 1,000) and large urban areas in the South (8.6 per 1,000). The lowest rates of violence involving a weapon (3.0 per 1,000) were found in large urban areas in the Northeast. However, these rates were not significantly lower than the national average. The national rate of violence involving a firearm during 2010-15 was 1.5 per 1,000 persons age 12 or older. Mid-size urban areas in the Midwest (6.1 per 1,000) exhibited significantly higher firearm violence rates than the national average. Rural areas in the Midwest reported a significantly lower rate of firearm violence (0.7 per 1,000) than the national average.

Rates of violence resulting in injury to the victim were significantly greater than the national average in mid-size urban areas in the Midwest and large urban areas in the South during 2010-15
The national rate of violence resulting in injury to the victim during 2010-15 was 5.4 victimizations per 1,000 persons age 12 or older (table 5). Two areas had rates of injurious violence that were significantly higher than the national average: mid-size urban areas in the Midwest (10.0 per 1,000) and large urban areas in the South (7.9 per 1,000). Other areas also had higher rates of violence resulting in injury. However, due to their smaller populations and sample sizes, their rates were not significantly greater. Large urban areas in the Northeast (3.7 per 1,000) and suburban areas in the South (4.2 per 1,000) had significantly lower rates than the national average.

Rates of violence committed by strangers varied across types of areas more than rates of violence committed by non-strangers
During 2010-15, rates of violence committed by strangers were significantly lower than the national average (8.1 per 1,000 persons ages 12 or older) in rural areas both in the Midwest (4.4 per 1,000) and the South (3.1 per 1,000) (table 6; figure 7). Stranger violence rates were also significantly lower than the national average in suburban areas in the Midwest (5.8 per 1,000) and the South (6.3 per 1,000).
Rates of stranger violence during 2010-15 were significantly higher than the national average in numerous urban areas. In the Northeast, the stranger violence rate was highest in mid-size urban areas (14.8 per 1,000), though this rate was not significantly higher than the national average due to the sample size in these areas. In small urban areas in the Northeast, where a greater proportion of persons live, the stranger violence rate was significantly greater (13.6 per 1,000) than the national average.
The stranger violence rate was highest in mid-size urban areas in the Midwest (13.5 per 1,000 per 1,000 persons age 12 or older), and this rate was significantly greater than the national rate.
In the South, the highest rate of stranger violence was found in large urban areas (13.9 per 1,000). In the West, urban areas of all sizes had significantly greater rates of stranger violence than the national average: 12.1 per 1,000 in smaller urban areas, 15.9 per 1,000 in mid-size urban areas, and 14.0 in cities with the largest populations.

With one exception, rates of intimate partner violence did not vary significantly across areas during 2010-15
Compared to the patterns for stranger violence, there was less variation in rates of intimate partner violence across the 20 types of areas. During 2010-15, the national rate of violent victimization committed by intimate partners was 2.9 per 1,000 per persons age 12 or older. Nineteen of the 20 areas exhibited statistically similar rates of intimate partner violence (table 6; figure 8). The only area where rates of intimate partner violence differed significantly from the national average was in large urban areas in the Northeast, where intimate partner violence rates were significantly lower (1.0 per 1,000 persons age 12 or older). With two exceptions, rates of violence committed by well-known or casual acquaintances also showed little variability across types of places (table 6; figure 9). Large urban areas in the Northeast exhibited significantly lower rates (3.0 per 1,000 persons age 12 or older) of this type of violence during 2010-15 than the national average (7.3 per 1,000), and rates of violence by well-known or casual acquaintances were significantly higher in rural areas in the West (12.1 per 1,000). Rates of violence committed by relatives other than intimate partners were lower than those committed by intimate partners, well-known or casual acquaintances, or strangers. However, none of the areas exhibited rates of violence by other relatives that were significantly different from the national average (1.5 per 1,000) (table 6; figure 10).

The percentage of violent crime reported to police did not vary significantly across areas during 2010-15
During 2010-15, a total of 47.1% of all violent crimes in the United States were reported to police (table 7). Levels of violent crime reporting across the 20 different types of areas ranged from a low of 40.2% in mid-size urban areas in the Northeast to a high of 54.6% in Midwest rural areas. However, none of the area-reporting rates were significantly different from the national average.
For the nation as a whole, a lower percentage of household property crime victimization (36.2%) than violent victimization (47.1%) was reported to police. The percentage of property crime reported to police ranged from a low of 28.3% in large urban areas in the Northeast to a high of 39.4% in suburban areas in the South. Areas that reported household property crimes at a significantly lower percentage than the national percentage included mid-size (29.3%) and large (28.3%) metropolitan cities in the Northeast and urban areas of all population sizes in the West: 32.2% in small urban areas, 32.7% in mid-size urban areas, and 28.9% in large urban areas. In suburban areas in the West, the percentage was 33.9%. Rural areas in the Midwest (42.1%) reported household property crimes at a higher level than the national average during this period, as did suburban areas in the South (39.4%).

Conclusion
Using information available in public-use NCVS data files, this report examined victimization across different types of areas in the United States to help further BJS's efforts to develop subnational estimates of criminal victimization. By using survey information on region of the country, metropolitan area status, and population size of the core cities in metropolitan areas, respondents were classified as residing in 1 of 20 different types of areas. Use of this classification revealed that rates of violent and household property victimization varied significantly from the national average in some types of places and that rates of some forms of violence, such as non-stranger violence, varied little across types of places.
Additional analyses would be necessary to determine what factors might account for the variation in the rates shown in this report, and some of these factors can be examined using the public-use NCVS data. For example, in the Midwest and Northeast regions of the country, higher rates of serious violence were found in the small and mid-size cities of metropolitan areas, while in the South and West, higher rates were found in the mid-size and large cities.
Understanding why city size is differentially associated with serious violence across the regions requires further examination of factors such as population composition and other city-level differences. For instance, age and household income are known to be correlates of violent victimization, and higher-rate areas may have proportionately larger populations of young persons and poorer persons than areas with lower rates. The variation in these rates, however, may also be associated with characteristics of the areas themselves (such as housing and population density, economic and investment activity, and other conditions). These types of analyses would require the use of city-identifying information available in the NCVS geographic micro-data restricted use files rather than public-use NCVS data (see, e.g., Xie & Planty, 2014), and externally available data that can be linked to the NCVS city victimization rates. 3